UNIVERSITY OF CINCINNATI

___June 8____ , 2001____

I,______Lev Davydov______, hereby submit this as part of the requirements for the degree of:

______Doctor of Philosophy______in: ______Chemical Engineering______It is entitled: _Photocatalytic Degradation of Organic Contaminants: _Novel Catalysts and Processes______

Approved by: ______Prof. Panagiotis Smirniotis_ Prof. William Krantz______Prof. Allan Pinhas______Prof. Sotiris Pratsinis______

Photocatalytic Degradation of Organic Contaminants: Novel Catalysts and Processes

A dissertation submitted to the

Division of Research and Advanced Studies of the University of Cincinnati

in partial fulfillment of the requirements for the degree of

DOCTOR OF PHILOSOPHY

In the Department of Chemical Engineering of the College of Engineering

2001 by Lev Davydov

ME/Diploma, Mendeleev Institute, Moscow, Russia, 1993

Committee: Professor P.G. Smirniotis Professor W. Krantz Professor A. Pinhas Professor S.E. Pratsinis ABSTRACT

Photocatalysis has recently emerged as an advanced oxidation process. The

present dissertation aims at the practical increase of the energy efficiency of

photocatalysis using fundamental tools.

A comprehensive kinetic model was developed to describe primary photocatalytic

processes taking place on the surface of semiconductors. The steady-state assumption for all the intermediate species in the system allowed to find the lumped kinetic parameters and elucidate the relative extent of electron-hole recombination reactions. It is important

to utilize single-stage oxidation reactions to test this kinetic model and obtain kinetic parameters. This simple method also allowed to experimentally determine the rates of radical generation as well as the rate of electron-hole recombination for a number of commercial titania photocatalysts.

When interfaced with the continuous flow reactor design, the model allowed to predict the optimal radiation profiles in photoreactors, which would significantly increase the reactor output. Such profiles are represented by a combination of exponential functions, and they prescribe more radiation at the beginning of the reactor in comparison with that at the end. Furthermore, it was found that when a continuous-flow reactor with recycle, the use of the optimal profile can produce a major enhancement of the output in comparison with the uniformly illuminated photoreactor.

The expansion of the working range of photocatalysts to enable them to utilize visible light was also undertaken. To achieve this, a combination of doping and sensitization properties was needed. It was found that titania-loaded transition metal MCM-41 materials can allow for such combination. The heterojunction with the transition metal substituted MCM-41 works as a “sensitizer” for the titania deposit, while the extraframework transtion metal inclusion can diffuse inside the titania loading and work as a dopant. The latter allows to effectively utilize visible light to perform photocatalytic reactions. Similar effect of sensitization was found for Cd-MCM-41,

which enhances the photoactivity of titania in ultraviolet light.

Conclusively, it has been shown that the use of optimal reactor design as well as

novel catalysts containing active supports can significantly increase the reaction rates and

exclude the dependence of the process on the artificial sources of energy.

ACKNOWLEDGEMENTS

The author of this dissertation wishes to acknowledge a number of sponsors for their support of this research:

TAPPI Foundation

Procter&Gamble

NATO Science for Peace Program

US Department of Army

UC Distinguished Dissertation Fellowship

UC University Research Council

The author also wishes to thank his advisor, Prof. Panagiotis Smirniotis of UC as well as Prof. Sotiris Pratsinis of ETH Zurich, Dozent Roumen Tsekov of the University of Sofia, Dr. Alexandre Vorontsov of Boreskov Institute of Catalysis, and Dr.

Padmanabhareddy Ettireddy of UC.

TABLE OF CONTENTS

Table of contents 1

List of figures and tables 3

Introduction 9

Section 1 Photocatalytic reactor design

Chapter 1.1. Novel differential reactor for the measurement of overall quantum

yields 19

Chapter 1.2. The intrinsic catalytic activity in photoreactors 47

Chapter 1.3. Optimal radiation field in continuous heterogeneous photoreactors

77

Chapter 1.4. Sonophotocatalytic reactor for VOC destruction 107

Section 2 Kinetic modeling of photocatalytic processes

Chapter 2.1. Quantification of primary photocatalytic processes using single-stage

oxidation reactions 133

Chapter 2.2. Photocatalytic destruction of diethylsulfide in gas phase 165

Chapter 2.3. Radical generation during sonophotocatalytic destruction of VOCs

on zeolite-supported titania 189

1 Section 3 Novel photocatalyst design and testing

Chapter 3.1. Transition metal substituted MCM-41 as photocatalysts of aqueous

VOC oxidation in visible light: Synthesis and characterization 211

Chapter 3.2. Transition metal substituted MCM-41 as photocatalysts of aqueous

VOC oxidation in visible light: Photocatalytic activity 237

Chapter 3.3: Sensitization of TiO2 with a transition metal substituted MCM-41

support for enhanced activity in UV light 255

Conclusions and Future Work 263

Appendix 1: Liquid Phase Photocatalysis – Experimental Details 267

Appendix 2: Gas Phase Photocatalysis – Experimental Details 277

Appendix 3: Radiation Model – Computational Details 283

Appendix 4: Radiation Field Optimization - Computational Details 293

2 LIST OF TABLES AND FIGURES

Figure I.1. Simplified schematic of semiconductor photocatalysis

Table 1.1.1. Properties of titania powders employed in the present study. Figure 1.1.1. Conventional annular photocatalytic reactor configuration employed in the present study. The UV lamp is placed in the center of the annular region. Figure 1.1.2. Experimental reactor configuration with variable reaction zone thickness and length used in the present study Table 1.1.2. Setup parameters and experimental conditions for four photocatalytic reactors employed in the present study. Figure 1.1.3. Experimentally measured axial radiation profiles inside the photocatalytic reactor: 200 W light source, catalyst concentration 0.25 g/l. Figure 1.1.4. Experimentally measured axial incident radiation profiles for partially covered 200 W light source. Table 1.1.3. Axial uniformity factor for the photocatalytic reactors employed in the present study (Aldrich anatase, 0.25 g/l, 200 W lamp) Figure 1.1.5. Time on stream behavior of the photocatalysts utilized in the present study in the conventional photocatalytic reactor (catalyst concentration 0.25 g/l, 2,4,6- trichlorophenol 2 mM, t=30±3 °C, pH=3.75). Figure 1.1.6. Dependence of average overall quantum yield on the thickness of the reaction zone in the photocatalytic reactors with variable reaction zone length for 0.25 g/l Aldrich anatase and 2 mM of 2,4,6-trichlorophenol, t=30±3 °C, pH=3.75 Figure 1.1.7. Dependence of average overall quantum yield on the length of the reaction zone for the thinnest annulus (Aldrich anatase – 0.25 g/l, 2,4,6-trichlorophenol – 2 mM; t=30±3 °C, pH=3.75) Table 1.1.4. Comparison of overall quantum yields for TiO2-assisted photodegradation of 2,4,6-trichlorophenol in conventional (apparent) and differential (intrinsic) photoreactors (catalyst concentration 0.25 g/l); the reaction zone sizes are specified in parentheses.

Table 1.2.1. Photocatalytic powders of titania employed in the present study. Table 1.2.2. Geometric parameters and experimental conditions for four different photocatalytic setups employed in the present study. Figure 1.2.1. Experimental reactor configuration with variable reaction zone thickness and length Figure 1.2.2. Experimentally measured radiation profiles inside the photocatalytic reactor: 450 W light source; incident radiation profile for 100 W light source Table 1.2.3. Parameters of equations 7 and 8 used in the present study. Figure 1.2.3. Lower curve: Dependence of average overall quantum yield on the thickness of the reaction zone in the photocatalytic reactor with variable reaction zone (Figure 1.2.1), 6 cm length for 0.5 g/l Degussa P25 and 2 mM of phenol, t=30±3 °C, pH=3.75, P=100W; Upper curve: dependence of first order kinetic constant on the reaction zone length for the same conditions, obtained in reaction zone thickness 0.15 cm.

3 Table 1.2.4. Comparison between photocatalytic effectiveness factors (eq. 7) and intrinsic kinetic constants for several reaction zones (initial concentration of phenol 2 mM) Table 1.2.5. Comparison between theory and experiment for three different photocatalysts and three different light sources. a. 100 W lamp, 0.5 g/l of photocatalyst b. 200 W lamp, 0.25 g/l of photocatalyst c. 450 W lamp, 0.1 g/l of photocatalyst Figure 1.2.4. Dependence of photocatalytic effectiveness factor on design criterion: rhombs: Degussa P25, squares: Aldrich anatase, asterisks: Aldrich anatase 325

Figure 1.3.1. Schematic of the class of photocatalytic reactors analyzed in the present study and variables used Figure 1.3.2. Comparison of numerical solution of equation 2.13 at k1=0.01 Einstein/(L 1 - Cos(npq) min) and T=0.5 min for f *(q) = (Case a is for n=0.5, Case b – n=1, Case 1 - Sin(np) / np -4 -3 c – n=10, Case d – n=100): A - C0=10 M, B - C0=10 M Figure 1.3.3. Case I - comparison of the optimal radiation profiles in the reactor at -3 4 -1 -1 8 -1 -1 C0=10 M, k1=0.01 Einstein/(L min): A - k6=10 M min , B - k6=10 M min Figure 1.3.4. Output of a uniformly illuminated photoreactor calculated by the model of -4 4 -1 -1 Case I as a I as a function of process parameters at C0=10 M, k6=10 M min Figure 1.3.5. Case II - comparison of the optimal radiation profiles in the reactor at -3 4 -1 -1 8 -1 -1 C0=10 M, k1=0.01 Einstein/(L min): k6=10 M min , k6=10 M min Figure 1.3.6. Output of a uniformly illuminated photoreactor calculated by the model of -4 4 -1 -1 Case II as a function of process parameters at C0=10 M, k6=10 M min Figure 1.3.7. Comparison of the numerical solution of the full model with approximate analytical solutions for a uniformly illuminated photoreactor at k1=0.01 Einstein/(L min), T=1 min (a – numerical solution, b – Case I approximation, c – Case II approximation) Figure 1.3.8. The degree of enhancement due to optimal radiation profile in the reactor -4 4 -1 -1 (vopt/v1) for Case I: C0=10 , k6=10 M min Figure 1.3.9. The degree of enhancement due to optimal radiation profile in the reactor -4 4 -1 -1 (vopt/v1) for Case II: C0=10 M, k6=10 M min Figure 1.3.10. Optimal parametric curve for the degree of enhancement (vopt/v1) for Case -4 3 -1 -1 4 -1 -1 8 -1 -1 II: C0=10 M (a - k6=10 M min ; b - k6=10 M min ; c - k6=10 M min , d - 10 -1 -1 k6=10 M min )

Table 1.4.1. Properties of the photocatalysts employed in the present study Figure 1.4.1. Schematic of the sonophoto-catalytic reactor employed in the present study: 7 UV lamps and one ultrasonic probe Figure 1.4.2. Photocatalytic degradation of salicylic acid over different concentration of TiO2 Hombikat: Initial salicylic acid concentration: 1 mM, Reaction time: 3 hours, Oxygen flow rate: 0.5 L/min, 28 W UV-lamps, 100 W ultrasound (US). Figure 1.4.3. Relative enhancement of the rate of the photocatalytic reaction of salicylic acid over TiO2 Hombikat by US: Initial concentration: 1 mM, Reaction time: 3 hours, Oxygen flow rate: 0.5 L/min.

4 Figure 1.4.4. Photocatalytic degradation of salicylic acid over different titanias (0.25 g/L): a – Aldrich, b – Degussa P25, c – Ishihara; Initial concentration: 1 mM, Reaction time: 3 hours, Oxygen flow rate: 0.5 L/min, 28 W UV-lamps, 100 W US. Figure 1.4.5. Photocatalytic degradation of salicylic acid over different titanias: Catalyst weight: 0.1 g/L, Initial concentration: 1 mM, Reaction time: 3 hours, Oxygen flow rate: 0.5 L/min, 28 W UV-lamps, 100 W US. Table 1.4.2. Zero order reaction rates (mol/Lmin) of the photodegradation of salicylic acid (maximal error 9 %) Figure 1.4.6. TGA analysis of spent Hombikat TiO2 at 3 hours of reaction: Initial concentration: 1 mM, Reaction time: 3 hours, Oxygen flow rate: 0.5 L/min, 28 W UV- lamps, 100 W US. 2+ Figure 1.4.7. Photooxidation of Fe over Hombikat TiO2: Catalyst weight – 0.25 g/L, 2+ [Fe ]0=5 mM, Initial concentration: 1 mM, Reaction time: 3 hours, Oxygen flow rate: 0.5 L/min, 28 W UV-lamps, 100 W US. Figure 1.4.8. Photooxidation of formic acid over Hombikat TiO2: Catalyst weight – 0.25 g/L, [HCOOH]0=10 mM, Initial concentration: 1 mM, Reaction time: 3 hours, Oxygen flow rate: 0.5 L/min, 28 W UV-lamps, 100 W US.

Table 2.1.1. Photocatalysts employed in the present study Figure 2.1.1. Evolution of the concentration of reactive oxygen species in slurry: 2+ [Fe ]0=0.03 M, catalyst – Degussa P25 (a), Aldrich anatase (b), lamp power – 450 W, temperature 30±3oC Figure 2.1.2. Average zero-order rate of photodegradation of formic acid versus square - root of the absorbed photon flux: [HCOO ]o=30 mM, lamp power – 450 W, temperature 30±3oC. - Figure 2.1.3. Time course of the concentration of formic acid: [HCOO ]o=1.5 mM, lamp power – 450 W, temperature 30±3oC, catalyst concentration – 0.25 g/L Table 2.1.2. Comparison of the parameters of equation 3.5. Figure 2.1.4. Langmuir parameter of equation 3.5 versus catalyst concentration: [HCOO- o ]o=1.5 mM, lamp power – 450 W, temperature 30±3 C Figure 2.1.5. Reactive oxygen species generation rates (equation 3.8) as a function of photocatalyst concentration: lamp power – 450 W, temperature 30±3oC. Figure 2.1.6. Electron-hole recombination rates (equation 3.9) versus photocatalyst concentration: lamp power – 450 W, temperature 30±3oC. Table 2.1.3. Quantum yields of radical generation and electron-hole recombination. Table 2.1.4. Quantum yields of formic acid degradation.

Figure 2.2.1. Reaction scheme of photocatalytic degradation of diethyl sulfide based on the products detected. Figure 2.2.2. Exit concentrations of diethyl sulfide and products of its destruction over TiO2 Hombikat UV 100 as a function of operation time. Humidity <1%, temperature of catalyst 25°C. Lamp power 8 W. Figure 2.2.3. Exit concentrations of diethyl sulfide and products of its destruction over TiO2 Hombikat UV 100 as a function of operation time. Humidity 59%, temperature of catalyst 25°C. Lamp power 8 W.

5 Figure 2.2.4. Exit concentrations of diethyl sulfide and products of its destruction over TiO2 Hombikat UV 100 as a function of operation time. Humidity <1%, temperature of catalyst 40°C. Hg lamp power 450 W. Figure 2.2.5. Exit concentrations of diethyl sulfide and products of its destruction over TiO2 Hombikat UV 100 as a function of operation time. Humidity 10 %, temperature of catalyst 400C. Hg lamp power 450 W. Table 2.2.1. Cumulative consumption of diethyl sulfide and production of gaseous products during the period of photocatalytic reaction of 25 to 480 min expressed in moles for all compounds Figure 2.2.6. Exit concentrations of diethyl sulfide and products of its destruction over homemade TiO2 as a function of operation time. Humidity 19%, temperature of catalyst 250C. Lamp power 8 W. Figure 2.2.7. Exit concentrations of diethyl sulfide and products of its destruction over 2 TiO2 Degussa P25 50 m /g as a function of operation time. Humidity 19%, temperature of catalyst 25°C. Lamp power 8 W. Figure 2.2.8. Exit concentrations of diethyl sulfide and products of its destruction over 2 TiO2 Degussa P25 75 m/g as a function of operation time. Humidity 19%, temperature of catalyst 25°C. Lamp power 8 W. Figure 2.2.9. Exit concentrations of diethyl sulfide and products of its destruction over TiO2 Hombikat UV 100 as a function of operation time. Humidity 59%, 75% of feed stream passed through 30% aqueous solution of H2O2, temperature of catalyst 25°C. Lamp power 8 W.

Table 2.3.1. BET surface areas of the catalysts before and after the reaction Figure 2.3.1. X-ray diffraction patterns for the catalysts before and after reaction: (a) 25%TiO2/USY; (b) 25%TiO2/MCM-41 Figure 2.3.2. Intensity of the main XRD peak of MCM-41 (2q=2°) versus ultrasonic power input to volume ratio during ultrasonication Figure 2.3.3. Degradation of salicylic acid by UV light and combination of UV light and ultrasound on: (a) 25%TiO2/USY; (b) 25%TiO2/MCM-41 Figure 2.3.4. Degradation of salicylic acid on 25%TiO2/b and neat titania (Hombikat) Table 2.3.2: Initial reaction rates (mol/gTiO2 min) of salicylic acid under UV-light (28 W) and combination of UV (28 W) and ultrasound (100 W) Figure 2.3.5. Isotherms of adsorption of toluene from toluene/water solution by the zeolite supports employed in the present study: initial concentration of toluene – 0.5 mM, concentration of zeolite – 2 g/L, temperature – 20ºC Figure 2.3.6. Time course of the photo- and sonophotodegradation of formic acid on neat titania and MCM-41 supported titania: initial concentration of HCOOH – 10 mM, concentration of HK – 0.25 g/L, concentration of 25%TiO2/MCM-41 – 1 g/L, ultrasonic power input – 100 W, UV power input – 28 W Figure 2.3.7. Time course of the photo- and sonophotodegradation of formic acid on b- zeolite and USY supported titania: initial concentration of HCOOH – 10 mM, catalyst concentration – 1 g/L, ultrasonic power input – 100 W, UV power input – 28 W Figure 2.3.8. Time course of the photo- and sonophotooxidation of Fricke solution on neat titania and MCM-41 supported titania: initial concentration of HCOOH – 10 mM,

6 concentration of HK – 0.25 g/L, concentration of 25%TiO2/MCM-41 – 1 g/L, ultrasonic power input – 100 W, UV power input – 28 W Figure 2.3.9. Time course of the photo- and sonophotooxidation of Fricke solution on b- zeolite and USY supported titania: initial concentration of HCOOH – 10 mM, catalyst concentration – 1 g/L, ultrasonic power input – 100 W, UV power input – 28 W

Figure 3.1.1. XRD diffractograms of Me-Ti-MCM-41 (Si/Me-80, Si/Ti=40) supports Figure 3.1.2. FTIR spectra of a) MCM-41; b) Cr-Ti-MCM-41; c)25 wt% TiO2/Cr-Ti- MCM-41 Table 3.1.1. BET areas, peak pore sizes, and metal dispersions of the catalysts used in the present study Table 3.1.2. Comparison of reaction rates of degradation of different VOCs in water by the visible- irradiated composite catalysts of the present study. The performance of Degussa P25 in UV is shown as a benchmark Figure 3.1.3. UV-Vis diffuse reflectance spectra of the transition metal substituted MCM-41 supports (Si/Me=80, Si/Ti=40) Figure 3.1.4. UV-Vis diffuse reflectance spectra of the titania loaded transition metal substituted photocatalysts (titania loading – 25 wt%) Figure 3.1.5. UV-Vis diffuse reflectance spectra of the chromium substituted MCM-41 (Si/Cr=80) that have undergone the extraction experiment Figure 3.1.6. TPR profiles of catalysts: a) (Cr(NO3)3 on MCM-41; b) CrO3 on MCM-41; c) Cr-Ti-MCM-41; d) Cr-Si mixed oxide; e) Cr-MCM-41 Figure 3.1.7. TPR profiles of 25 wt % TiO2 loaded catalysts: a) Cr-MCM-41; b) Cr-Ti- MCM-41; c) (Cr(NO3)3 on MCM-41; d) CrO3 on MCM-41; e) Cr-Si mixed oxide Figure 3.1.8. Deconvoluted TPR profile of 25%TiO2/Cr-Ti-MCM-41 Figure 3.1.9. TPR profiles of catalysts after photocatalysis: a) 25%TiO2/Cr-Ti-MCM-41; b) 25%TiO2/(Cr(NO3)3 on MCM-41; c) CrO3 on MCM-41 Figure 3.1.10. Deconvoluted XPS spectra for O 1s peak: a - Cr-Ti-MCM-41; b - 25%TiO2/Cr-Ti- MCM-41 Figure 3.1.11. XPS spectra for Si 2p peak: a) Cr-Ti-MCM-41; b) 25%TiO2/Cr-Ti- MCM-41 Figure 3.1.12. Deconvoluted XPS spectra for Cr 2p peak: a) Cr-Ti-MCM-41; b) 25%TiO2/Cr-Ti- MCM-41

Figure 3.2.1. Scheme of the photocatalytic reactor employed in the present study Figure 3.2.2. Emission spectrum of the mercury lamp and transmission spectrum of the light filter Figure 3.2.3. UV-Vis diffuse reflectance spectra of the titania loaded transition metal substituted photocatalysts (titania loading – 25 wt%) Table 3.2.1. BET areas, peak pore sizes, and metal dispersions of the catalysts used in the present study Table 3.2.2. Surface atomic ratios of the select catalysts as determined by XPS Figure 3.2.4. Time course of the degradation of formic acid on the V-, Cr-, and Fe- substituted catalysts (Table 3.2.1), Degussa P25 in UV is shown for comparison (pH=4, T=25±3°C, catalyst concentration – 1 g/L)

7 Figure 3.2.5. Time course of the degradation of 2,4,6-trichlorophenol on the Cr- substituted catalyst of Table 3.2.1, Degussa P25 in UV is shown for comparison (pH=6, T=25±3°C, catalyst concentration – 1 g/L) Figure 3.2.6. Time course of the degradation of 4-chlorophenol on the Cr-substituted catalyst of Table 3.2.1, Degussa P25 in UV is shown for comparison (pH=6, T=25±3°C, catalyst concentration – 1 g/L) Figure 3.2.7. Time course of the degradation of phenol on the Cr-substituted catalyst of Table 3.2.1, Degussa P25 in UV is shown for comparison (pH=6, T=25±3°C, catalyst concentration – 1 g/L) Figure 3.2.8a. The projected structure of transition metal (example: chromium) substituted MCM-41 photocatalysts Figure 3.2.8b. The projected structure of titania loaded transition metal (example: chromium) substituted MCM-41 photocatalysts Figure 3.2.9. Proposed mechanism of the photooxidation on TiO2/Cr-(Ti)-MCM-41 with the molecular excitation of CrO3

Table 3.3.1. Physical and physicochemical properties of the catalysts utilized in the study Figure 3.3.1. XRD Patterns of the catalysts employed in the present study Figure 3.3.2. XPS spectra for Si 2p (left) and Cd 3d (right) for the catalysts and supports utilized in the study: a- Cd-MCM-41, b – 25%TiO2/Cd-MCM-41 Figure 3.3.3. DR UV-Vis spectra of the catalysts employed in the study Figure 3.3.4. Photocatalytic activity of the catalysts employed in the present study

Figure A.2.1. Schematic representation of the experimental set-up. 1- air cylinder, 2 – pressure regulator, 3 – mass flow controllers, 4 – liquid infusion pump, 5 – saturator with water, 6 - 6-way crossover valve, 7 – photochemical safety cabinet, 8 – photocatalytic reactor, 9 – GC HP 6890 equipped with gas sampling valve. Figure A.2.2. (A) Cross sections of the photocatalytic reactor with the bottom for vibrofluidization and (B) the bottom for ultrasonication. 1 – two UV lamps, 2 – pyrex window, 3 – four gas outlets at 90o with each other, 4 – thermocouple, 5 – four gas inlets at 90o with each other, 6 – Teflon membrane with catalyst over it, 7 – speaker, 8 – stainless steel plate, 9 – Teflon membrane, 10 – tip of the ultrasound transducer.

8 INTRODUCTION

Widespread air and water pollution has plagued the planet for many years. As a

response to the looming threat, the humankind has been expediting its efforts in pollution

abatement. Several approaches are used: to utilize environmentally benign processes, to

provide in-situ destruction of pollutants during the process, and to decontaminate the air or water stream emanating from the high throughput production facility. The latter two approaches can be addressed using photocatalysis.

Photocatalysis is a relatively new technique of decontamination of aqueous and

air streams. Photocatalytic water splitting on TiO2 was first discovered in 1972 by

Fujishima and Honda [1]. Since then, it has drawn considerable academic interest as a

very attractive, non-selective room-temperature process for the degradation of organic

pollutants [2, 3]. However, it has found a very limited degree commercialization [4]. This trend is primarily related to the low reaction rates exhibited by commonly known photocatalysts.

Semiconductors can provide light-induced charges for redox processes, which is primarily due to their electronic configuration [2]. In particular they are characterized by

a filled valence band and empty conduction band [5]. The elementary mechanism of photocatalytic transformation includes a number of steps, which have been exhaustively described in the literature [2, 6]. All photocatalysts must possess semiconducting properties in order to be able to perform photoinduced reactions. A simplified diagram of the photocatalytic mechanism is presented in Figure 1. It first starts with the absorption of light by the semiconductor particle. This may cause the excitation of the semiconductor,

9

Figure I.1. Simplified schematic of semiconductor photocatalysis

which promotes an electron from the valence band to the conduction band of the

semiconductor:

+ - TiO2 + hn ® h + e

In order to perform such transition the absorbed light must be of energy higher than the

bandgap energy of the semiconductor (for example, for TiO2 it must be above 3.2 eV).

Some of the electrons and holes can recombine to emit a quantum of energy or to release

heat, depending on the type of semiconductor used:

h+ + e- ® heat

The unrecombined electron-hole pairs can diffuse to the surface of the semiconductor

particle. Such electron-hole pairs can undergo the interface charge transfer and react with

the surface species. In particular, hole can react with the adsorbed organic compound or

surface hydroxyl group: h+ + RH ® R· + H+

10 h+ + >OH- ® > ·OH

And electron can reduce the surface titanium atom or attack the adsorbed oxygen or

another electron scavenger. In the first case, the titanium atom is re-oxidized by the next

available hole (which is essentially the same as recombination):

Ti+4 + e- ® Ti+3

Ti+3 + h+ ® Ti+4

In the case of an electron reacting with oxygen, a radical anion is formed, which can

further react to form hydrogen peroxide or other active oxygen radicals:

- · - e + O2 ® O2

· - + HO2 + e +H ® H2O2

The radicals formed by holes and electrons can attack the organic (or oxidizable and

reducible inorganic) species present in the solution or gas phase. It is widely accepted

that the role of hydroxyl radicals is to abstract the hydrogen atom from the a-carbon [6]:

· · OH + RH ® R + H2O

Then the resulting radical can be attacked by any of the reactive oxygen radicals forming

hydroxy-substituted species:

·OH + R· ® ROH

or tetraoxide species [7]:

· · R + O2 + HO2 ® ROOOOH

This latest reaction leads to the release of CO2. This way all organic compounds can be

oxidized to CO2 and water.

TiO2 is by far the most studied and known photocatalyst [8]. It combines the chemical stability and relatively high activity for water splitting as well as ease of

11 preparation and low price. The above factors, however, do not outweigh the need to use ultraviolet radiation. Other transition metal oxides, such as WO3 [9], Fe2O3 [10], Cu2O

[11], Bi2O3 [12], In2O3 [13], SnO2 [14], ZnO [15], and ZrO2 [16] as well as a number of mixed oxides were also tested for photocatalytic water splitting and/or decontamination of aqueous and gas streams. None of them, however, exhibit structural stability to photocorrosion. Some of them (WO3, Fe2O3) show good visible light response without photocorrosion, but certain modifications of the reaction medium are required.

Transition metal sulfides [17] exhibit excellent visible light absorption, and some of them (such as CdS) have a favorable position of the energy bands to be able to photosplit water. All of the chalcogenide semiconductors, however, are subject to severe photocorrosion and self-oxidation [18], since sulfur (or Se and Te) in them can be easily oxidized. Therefore, these materials have not found an application as visible light photocatalysts. A number of attempts [19, 20] have been made to expand the working range of TiO2 to visible part of the solar spectrum. Two approaches can be followed, namely, doping and sensitization. The first approach involves the incorporation of foreign transition metal ions into the structure of titania in order to create an impurity energy level between its valence band and conduction band. This way it is possible to promote an electron from valence band to conduction band via two-step excitation absorbing two quanta of lower energy. This arrangement was fully tested by Lawless et al [19], and it was found that the impurity levels created by dopant ions act as recombination centers for the photogenerated holes and electrons as well as traps for the electrons. In the latter process, the second step of the excitation does not take place since the electron gets trapped on the impurity level [2]. Therefore, charge separation in thereby photoexcited

12 material is insufficient to perform photocatalytic reactions (in particular, transfer electron

to oxygen), let alone photosplit water.

The second way of expanding the photoresponse of a semiconductor is sensitization. A number of attempts were made to use easily excitable dyes as sensitizers

[21]. The mechanism consists of the photoexcitation of the dye in visible light with the subsequent charge injection into the host semiconductor [22]. This way it is possible to

achieve the full separation of charges in the semiconductor in two steps. However,

although the use of dyes allows the utilization of visible light, they are organic

compounds and can be subject to photodegradation [23]. This limits their use for

photocatalysis, and dye-sensitization has found its primary use in photovoltaic cells [24].

A number of variables has been introduced to characterize the performance of

photocatalysts/photoreactors. The most common are overall quantum yield and reaction

rate constant. The first variable is equal to the number of moles converted by one mole of

photons absorbed [25]. As observed by a number of researchers, this quantity is

dependent on the absorbed light intensity [26]. This limits its applicability to the

photocatalyst comparison. On the other hand, quantum efficiency (moles reacted per

mole of incident photons) is a good measure of photoreactor efficacy to destroy a

particular compound. The latter variable is also dependent on the light intensity with

square-root dependence for high intensities and linear dependence for lower intensities

[26], which again limits its applicability. Yet another variable widely used for catalysts

assessment is first-order reaction rate constant [27]. The meaning of such rate constant

does not differ from traditional reaction engineering, and it is even more dependent on

13 the light intensity [28]. Therefore, a variable, which would be less dependent or even

independent of light intensity or reactor setup and configuration, is highly desirable.

The reaction engineering of photocatalytic systems is a well-developed field [29], and it is primarily concerned with the methods of TiO2 distribution in photoreactors [30] as well as with the irradiation methods to achieve the maximum quantum efficiency [31].

The most “popular” laboratory configuration is an annular aqueous slurry photocatalytic reactor [32]. Such system possesses several disadvantages for commercial application, such as high optical thickness, the necessity of slurry filtration, and catalyst precipitation

[33]. There are also research issues associated with such reactors. It is difficult to assess the role of illuminated areas and dark areas in the reaction [34]. The latter arise due to severe non-uniformity of the radiation field and end effects in the reactor [27]. As widely accepted, higher radiant intensity leads to reaction rate proportional to square root of intensity, while lower light intensity obeys linear dependence [35]. Therefore, the combination of both types of relationships in a non-uniformly irradiated photo reactors will result in misleading kinetic information.

Low quantum yields are major shortcomings of the photocatalytic degradation

[4]. A number of approaches have been followed to increase the energy utilization. One of them is using periodic illumination in hope to enhance the performance. It consists of alternating of light and dark regions in a photoreactor [36]. Such configuration allows for as much as five-fold enhancement of the quantum efficiency. Two mechanisms have been proposed to explain this effect. First, it is proposed that the intermediate oxygen species are stable enough to continue the oxidation even in the dark [37], which allows saving light energy. The second hypothesis is that the dark recovery time provides the

14 replenishment of the photocatalytic surface by the reactant [38], thus increasing the

reaction rate. However, the periodic radiation profile is empirical and may not be optimal,

especially for the first hypothesis. That is why there is a need to deduce the optimal radiation profile from theoretical considerations.

Ultrasound has been widely known to induce radical reactions [39]. This useful

property has found its applications in sonolysis of water [40], sonolytic degradation of

aqueous organic pollutants [41], and sonochemical synthesis of chemicals [42]. The

underlying phenomena include cavitation, microstreaming, and localized supercritical

· · conditions. These phenomena lead to sonolytic splitting of water (H2O ® H + OH) as

well as pyrolysis of a vaporized molecule. Photocatalysis, on the other hand, can offer the

generation of surface radicals. Therefore, the combination of both processes may lead to

the enhanced hydroxyl radical production, as their generation will take place both in the

bulk solution and on the surface.

Conclusively, several important problems in photocatalysis have to be addressed

in order for it to become a commercially viable alternative to the conventional methods of

pollution abatement. First, new accessible methods of catalyst testing and comparison

have to emerge. Second, new reaction systems have to be developed and the utilization of

light has to be optimized. Third, new photocatalysts utilizing visible light and thus

requiring no artificial sources of light even indoors are necessary. These issues are

tackled in the following sections of the dissertation.

References

1 Fujishima, A. and Honda, K., Nature, 238, 37 (1972)

15

2 Hoffmann, M.R., Martin, S.T., Choi, W., and Bahnemann, D.W., Chem. Reviews, 95, 69 (1995)

3 Peral, J., Domenech, X., Ollis, D.F., J.Chem.Tech.Biotech, 70, 117 (1997)

4 Surender, G.D., Fotou, G.P., and Pratsinis, S.E.,Trends Chem.Eng., 4, 145 (1998)

5 Kamat, P.V., Chem. Rev., 93, 267 (1993)

6 Turchi, C.S. and Ollis. D.F., J. Cat., 122, 178 (1990)

7 Gerischer, H. and Heller, A., J.Phys.Chem., 95, 5261 (1991)

8 Alfano, O.M., Bahnemann, D., Cassano, A.E., Dillert, R., Goslich, R., Cat.Today, 58, 199 (2000)

9 Ashokkumar, M. and Maruthamuthu, P., Int. J. Hydrogen Energ. 16, 591 (1991)

10 Pal, B. and Sharon, M., J.Chem.Technol.Biotechnol., 73, 269 (1998)

11 de Jongh, P.E., Vanmaekelbergh, D., Kelly, J.J., Chem.Comm., 1069 (1999)

12 Maruthamuthu, P., Gurunathan, K., Subramanian, E., Sastri, M.V.C., Int.J.Hydr.Energy, 19, 889 (1994)

13 Poznyak, S.K., Golubev, A.N., Kulak, A.I., Surf.Sci., 454, 396 (2000)

14 Tennakone, K., Bandara, J., App.Catal.A: Gen., 208, 335 (2001)

15 Sivakumar, T., Shanthi, K., Samuel, T.N., Bioproc.Eng., 23, 579 (2000)

16 Navio, J.A., Hidalgo, M.C., Colon, G., Botta, S.G., Litter, M.I., Langmuir, 7, 202 (2001)

17 Khairutdinov, R.F., Colloid J., 59, 535 (1997)

18 Fermin, D.J., Ponomarev, E.A., Peter, L.M., J.Electroanal.Chem., 473, 192 (1999)

19 Serpone, N., Lawless, D., Disdier, J., and Herrmann , J.M., Langmuir, 10, 643 (1994)

20 Navio, J.A., Testa, J.J., Djedjeian, P., Padron, J.R., Rodriguez, D., Litter, M.I., App.Catal.A: Gen., 178, 191 (1999)

21 Cho, Y.M., Choi, W.Y., Lee, C.H., Hyeon, T., Lee, H.I., Environ.Sci.Tech., 35, 966 (2001)

16

22 Persson, P., Bergstrom, R., Lunell, S., J. Phys. Chem B, 104, 10348 (2000)

23 Martyanov, I.N., Savinov, E.N., Catal.Today, 39, 197 (1997)

24 Ren, Y., Zhang, Z., Gao, E., Fang, S., Cai, S., J.Appl.Electrochem, 31, 445 (2001)

25 Calvert, J. G. and Pitts, J. N., Photochemistry. John Wiley & Sons, Inc., 1967.

26 Riegel, G. and Bolton, J. R., J. Phys. Chem., 1995, 99, pp. 4215-4224

27 Fotou, G. P., and Pratsinis, S. E., Chem. Eng. Comm., 1996, 151, 251-269

28 Okamoto, K., Yamammoto, Y., Tanaka, H. and Itaya, A., Bull. Chem. Soc. Jpn., 1985, 58, 2023-2028

29 Cassano, A.E., Alfano, O.M., Catal.Today, 58, 167 (2000)

30 Vorontsov, A.V., Barannik, G.B., Snegurenko, O.I., Savinov, E.N., Parmon, V.N., Kinet.Catal., 38, 84 (1997)

31 Martin, C.A., Camera-Roda, G., Santarelli, F., Catal.Today, 48, 307 (1999)

32 Romero, R.L., Alfano, O.M., Cassano, A.E., Ind. Eng. Chem. Res., 36, 3094 (1997)

33 Ray, A.K., Catal.Today, 44, 357 (1998)

34 Heit, G. and Braun, A.M., Wat. Sci. Tech., 35, 25 (1997)

35 Sun, L. and Bolton, J.R., J. Phys. Chem., 100, 4127 (1996)

36 Buechler, K.J., Noble, R.D., Koval, C.A., and Jacoby, W.A., Ind. Eng. Chem. Res., 38, 892 (1999)

37 Sczechowski J.G., Koval C.A., Noble R.D., J. Photoch. Photobio. A, 74, 273 (1993)

38 Upadhya S, Ollis D.F., J. Phys. Chem. B, 101, 2625 (1997)

39 Suslick, K.S., Science, 247, 1439 (1990)

40 A. Kotronarou, G. Mills, and M. R. Hoffmann, J. Phys. Chem., 95 (1991) 3630

41 C. Petrier, M.-F. Lamy, A. Francony, A. Benahceene, B. David, V. Renaudin, and N. Gondreson, J. Phys. Chem., 98, 10514 (1994)

42 Sweet, J.D., Casadonte, D.J., Ultrason.Sonochem., 8, 97, (2001)

17

SECTION 1. PHOTOCATALYTIC REACTOR DESIGN

Chapter 1.1. Novel differential reactor for the measurement of overall quantum yields

INTRODUCTION

The photocatalytic degradation of organic contaminants in aqueous solutions commonly involves irradiation of titania particles in near UV region at 300-400 nm wavelength gap [1,2]. Several variables have been introduced to describe photocatalytic activity. Primary quantum yield represents the number of radicals produced in the primary process on the surface of a photocatalyst per single photon absorbed [3,4]. It can be obtained by indirect methods, such as the use of “spin-traps” [5]. Overall quantum yield or quantum efficiency represents the number of molecules degraded per photon sent into the system or absorbed by the system [3,4] and is the most common value for the characterization of photocatalysts. Apparent reaction rate constants are also widely used for evaluation of photocatalysts for degradation of a particular organic molecule. Since most phototocatalytic reactions are pseudo-first order for dilute organics, it is quite attractive to compare different photocatalysts on the basis of reaction rate constants.

They, however, differ significantly with radiative power of the lamp and concentration of the catalyst employed in the same reactor [6] which limits their use as a characteristic value. Thus, in numerous previous kinetic investigations the effect of the radiation field and geometry of each reactor was incorporated in the reaction rate constant or overall quantum yield. This effect is strongly related to the radiation source in the reactor and to the radiation extinction due to the presence of a suspension. [6] reported parabolic axial

19 profiles for the light distribution in a conventional annular photocatalytic reactor with

strong radial attenuation, which were predicted theoretically by [7]. These profiles are

non-uniform partly due to the small size of the lamp in comparison with that of the

reactor. The assessment of the photocatalytic activity of semiconductors based on results

obtained from different reactors can thus be misleading since different locations in

photoreactors are subject to different values of light intensity. Apparently, a valid

comparison of various photocatalysts would require the uncoupling of the intrinsic

properties of the catalyst from the reactor configuration and radiation field.

A number of attempts have been made to quantify the intrinsic reaction rate.

Palmisano et al [8] defined quantum yield as a ratio of “intrinsic reaction rate” to rate of

photon absorption. In this work, the intrinsic reaction rate corresponded to the

photochemical reaction rate obtained from laboratory kinetic data. Karakitsou and

Verykios [9] proposed an expression for the intrinsic rate of hydrogen production in the

photocatalytic cleavage of water. This intrinsic rate was represented by a ratio of the

hydrogen production rate to the optical transparency (ratio of transmitted radiation to

incident radiation) of the slurry. Serpone [10] introduced “relative photonic efficiency” in an attempt to better describe the photocatalytic behavior of different semiconductors.

This value for a standardized system (catalyst, reactor, and actinometer) was considered to be unity. The quantum yield of a particular photocatalytic system then becomes a product of its relative photonic efficiency with the apparent quantum yield. Davydov et al

[11] introduced a photocatalytic effectiveness factor, which allows to isolate intrinsic reaction rate constant from apparent kinetic data obtained in commonly used photocatalytic reactors.

20 Overall, the aforementioned problems lead to serious discrepancies in kinetic

constants and quantum efficiencies reported by different research groups for the same

catalysts and the same organic reactants. Moreover, quantum yields for the same

photocatalytic system but different reactor sizes can be drastically different as shown

below. For this reason, we have adopted the following approach to obtaining overall

quantum yields. The performance of TiO2 powders in the photocatalytic degradation of

2,4,6-trichlorophenol as a probe molecule in an annular photocatalytic reactor is used in order to obtain kinetic data. The variation of the reaction zone in the present study is achieved by decreasing its thickness and height to the point where the average overall quantum yields approach their maximum values. These values can be considered intrinsic since the non-uniformity of the radiation field will be minimized. Thus, the most favorable system will be isolated for kinetic studies and determining the intrinsic value of the overall quantum yield.

EXPERIMENTAL

The photocatalytic degradation of 2,4,6-trichlorophenol (reagent grade, Fisher) over various TiO2 suspensions was carried out in an annular slurry quartz reactor [12].

The physical properties of the catalysts are presented in Table 1.1.1. The suspension

(0.25 g/l) was prepared with the titania particles dispersed ultrasonically in the aqueous solutions containing 2 mM 2,4,6-trichlorophenol. This reactant was chosen because it dissolves well in water and cannot be stripped by the oxygen flow in the reactor. The pH of each solution was adjusted to 3.75 by 0.5 M H2SO4. Before each experiment, the UV lamp was warmed up for 5 min. Samples were withdrawn from the reactor and filtered

21 with 0.2 mm membrane filters (Gelman Sciences) to remove the titania particles. The

suspension temperature was maintained at 30+3 oC. The concentration of the unreacted

2,4,6-trichlorophenol was measured utilizing a Gas Chromatograph (HP-6890 series,

equipped with an FID) by direct injection. 4-Chlorophenol (reagent grade, Fisher) was

used as an internal reference peak. It was chosen because of its retention time in the

column (between water and 2,4,6-trichlorophenol). The results of a blank run (water

only) were subtracted from each chromatogram. A capillary column (Chrompack 7584)

was used for separation, and helium was used as a carrier gas. The reaction mixture was

tested for the presence of intermediates at different time on stream by direct injection into

a Gas Chromatograph (HP 5890 Series II) equipped with a Mass Spectrometer (HP-5972

series) and no detectable amounts of intermediate organic compounds were found.

Table 1.1.1. Properties of titania powders employed in the present study. BET Surface Area, Primary Particle Anatase Content, Catalyst m2/g Size, nm wt. % Degussa P25 75 21 80-85 Aldrich Anatase 10 156 99 SG* 147 N/A ~15 * Prepared by sol-gel method in accordance with [13]

The photocatalytic reaction aiming at obtaining a priori non-intrinsic data took place in the annular region formed between the outside wall of the immersion vessel and the inside wall of the reaction well (Ace Glass, Inc.), hereinafter referred to as a conventional photocatalytic reactor (Figure 1.1.1). The inside and the outside radii of the reactor were 2.7 cm and 3.3 cm, respectively. A double-walled quartz immersion well was placed in the middle of the reaction vessel. The purpose of the quartz immersion well

22 was to allow the circulation of water for cooling the light source and the solution.

Moreover, the water prevented the heat of the lamp from entering the reaction zone. The

UV lamp radiation was filtered by a Pyrex filter (7740, Ace Glass, Inc.) placed between the lamp and the immersion well. The thickness of the Pyrex filter was 2.38 mm, the inside diameter was 26 mm. The immersion type UV-radiation source used was 200 W

(medium pressure, Ace Glass, Inc.) mercury vapor quartz lamp powered through a transformer (7830-60-C1, Ace Glass, Inc.). The radius of the lamp was 1 cm and its emissive length was 13 cm. Circulation of the suspension within the reactor assured com-

Figure 1.1.1. Conventional annular photocatalytic reactor configuration employed in the present study. The UV lamp is placed in the center of the annular region.

23 plete mixing. Oxygen (Wright Brothers, 99.5 %) was sparged into the solution at 500 cm3/min through a fritted glass tube. This amount of oxygen was well in excess of what is necessary to oxidize the reactant since the increase in oxygen flow rate above that value did not lead to increased activity of the photocatalysts. With this flow rate no catalyst stripping or deposition on the walls of the reactor was observed. The entire pho-

Figure 1.1.2. Experimental reactor configuration with variable reaction zone thickness and length used in the present study

24 toreactor system was kept inside a UV-safe cabinet.

As mentioned above, the lamp used in our study had a cooling jacket with circulating water, which absorbed the IR part of the spectrum (heat). Also, the thick

Pyrex tube surrounding the lamp acted as a filter by not allowing radiation with l < 320

nm to pass through. It was demonstrated [14] that the emission spectrum of mercury UV-

lamps, transmission spectrum of Pyrex, and absorption spectrum of titania overlap yielding a narrow band of radiation (about 350-360 nm) utilized in photocatalytic

reactors. Since our reactor was made of the same materials as the one in the above study,

we can treat the UV radiation passing through the suspension as almost monochromatic.

The photocatalysts were also evaluated in a variable reaction zone catalytic setup

(Figure 1.1.2) with the same lamp and cooling jacket but different sizes of the annulus. It

is worth noting that the length/width ratio of the reactor is much larger than shown. Table

1.1.2 presents the parameters of this setup (made of Plexiglas) as well as those for the

conventional photoreactor. With this variable reaction zone configuration we were able to

decrease light transfer limitations. It thus allowed to minimize the non-uniformity of the radiation field in both axial and radial directions. A conventional magnetic stirrer vigorously agitated the suspension in the reactor. A laboratory peristaltic pump

(Masterflex 7017-21) was used to run the suspension through the reactor and an intermittent vessel used for sample withdrawal. A magnetic stirrer also agitated this vessel in order to avoid sedimentation of titania. No deposition of titania was observed during the entire experiment (15 min). Flexible plastic tubing connected the pump, vessel, and reactor.

25 Table 1.1.2. Setup parameters and experimental conditions for four photocatalytic reactors employed in the present study. Setup# Reaction Total Oxygen Liquid flow, Ultra- zone reactor flow, ml/min ml/min sonic bath, thickness volume, L min RO-RI, mm 1 (Fig. 1.1.2) 1.5 0.325 200 150 5 2 (Fig. 1.1.2) 3.0 0.415 300 150 10 3 (Fig. 1.1.2) 4.5 0.500 400 150 15 Conventional 6.0 0.650 500 N/A 20 (Fig. 1.1.1)

Cabrera et al [15] showed that the use of homogeneous actinometry (potassium ferroxalate) for the determination of photon fluxes is misleading since this method does not take into account the above effects in a real photocatalytic system. This method is

also non-selective with respect to the wavelength of the incident light in the range used

for photocatalysis. Therefore, it provides information about the whole spectrum of the

light source the major part of which may have energy too low for the excitation of the

photocatalyst. For this reason, a radiometer (International Light, Inc. Model IL 1700)

connected with a flat detector (International Light, Inc., Model SED033 #3435) was

employed to measure the photon flux in the reactor. A UV-filter was used with the

detector in order to measure only the light of interest and it exhibited a strong maximum

at the wavelength of 355 nm with isotropic angular response. It was set parallel to the

axis of the reactor against the point at which the measurement was taking place. Values

of incident radiation at RI (on the jacket) as well as the outgoing radiation at RO (on the outer wall) were measured at various locations along the reactor length by the detector.

The measurement was repeated for a different batch of Degussa P25 and the variance was found to be insignificant.

26 RESULTS AND DISCUSSION

The axial radiation distributions measured by radiometry in the conventional photocatalytic reactor (Figure 1.1.1) at the jacket as well as at the outer wall of the reactor filled with two different titania slurries (Aldrich anatase and Degussa P25) is shown in

Figure 1.1.3. The profiles are represented by bell-shaped curves with the maximum located at the center of the light source. The local radiant flux at the reactor ends can be up to two orders of magnitude smaller than that in the middle (end effects). It is also worth noting that the radiation decreases about one order of magnitude having passed through the 6 mm layer of 0.25 g/l suspension of titania in the radial direction (Figure

1.1.3). Moreover, the radiation reduction by attenuation in the ambient suspension can be significantly larger for higher concentrations of the photocatalyst (e.g. Fotou and

Pratsinis, reported more than 2 orders of magnitude decrease [6]). This indicates that different parts of the conventional photocatalytic reactor receive significantly different numbers of photons. On the purely qualitative basis one can isolate “high intensity” and

“low intensity” regions (Figure.3) in the conventional photocatalytic reactor (Figure

1.1.1). This leads to different reaction rate laws [16] in these two regions, which are averaged in the determination of quantum yields. On the contrary, if we manage to keep the radiation at any point within the reaction zone at the “high intensity” limit, we will be able to obtain the intrinsic value of the quantum yield, although incurring losses in the radial direction.

Apparently, the problem related to the end effects and radial attenuation must be circumvented in order to construct a reactor, which will yield intrinsic photocatalytic properties. A number of measurements were performed to compare the axial radiation

27 profiles at the inner wall of the reactor for various openings (Figure 1.1.4). The masking

of light was achieved by covering the pyrex filter (Figure 1.1.2) with aluminum foil. As

one can observe, the specific radiation G impinging on the inner wall of the reactor outside the specified reaction zone opening, drops significantly by decreasing the

distance in the z-direction. This part of the radiation corresponds to the beams travelling with an angle different than 90o with respect to the axis of the lamp. One would desire to have a sharp decrease of the radiation at the boundary of the opening, which is indeed the case for the smaller reaction zone openings (Figure 1.1.4). Furthermore, for the smallest reaction zone thickness (0.15 cm) involved in the present study, the photocatalytic effect of these beams is minimized for geometrical reasons. In order to quantify the role of the

0.1

2 0.01

1E-3

Incident Radiant Flux, W/cm Aldrich Anatase Degussa P25 1E-4

0 5 10 15 20 25 Reactor Height, cm

Figure 1.1.3. Experimentally measured axial radiation profiles inside the photocatalytic reactor: 200 W light source, catalyst concentration 0.25 g/l.

28 sphericity of radiation within the reaction zone, one can introduce a uniformity factor as a

ratio of the longitudinal average photon flux to the maximum attainable photon flux on

the inner wall of the reactor. The values of this factor are presented in Table 1.1.3. One

can observe that it indeed approaches unity for smaller openings. This indicates that the

radiation within the reaction zone becomes axially uniform with the decrease of the

opening. Thus, the sphericity effects related to the nature of light emission are also

minimized. It should be mentioned that the flat detector utilized in the present paper

(Model SED033 #3435) is characterized by isotropic angular response, thus providing

0.04

0.03 2

0.02

0.01

Radiant Flux, W/cm 4 cm open 2 cm open 0.00 1 cm open

8 10 12 14 16 Reactor Height, cm

Figure 1.1.4. Experimentally measured axial incident radiation profiles for partially covered 200 W light source.

29 Table 1.1.3. Axial uniformity factor for the photocatalytic reactors employed in the present study (Aldrich anatase, 0.25 g/l, 200 W lamp) m Reaction zone length, cm GI/GI 1 0.96 2 0.88 4 0.81 8 0.65 24 0.43

objective measurement of the radiation flux density impinging at any point. One should

1

Degussa P25 C/Co Sol-Gel Aldrich Anatase

0.1 0 20 40 60 time, min

Figure 1.1.5. Time on stream behavior of the photocatalysts utilized in the present study in the conventional photocatalytic reactor (catalyst concentration 0.25 g/l, 2,4,6- trichlorophenol 2 mM, t=30±3 °C, pH=3.75).

30 also consider that the determination of quantum yields presented in this study, was based on the entire photon flux sent into the annular region of the photoreactor for each reaction zone opening.

Typical representative time on stream behavior of the catalysts employed in this study in the conventional photocatalytic reactor (Figure 1.1.1) is presented in Figure

1.1.5. They follow first order kinetics, which was also observed by other researchers [17] for chlorinated phenols. In order to avoid the change of the state of the catalyst (such as in the case of the sol-gel made catalyst) and subsequent deviation from linearity several experimental measures were undertaken. Firstly, the lamp was warmed up for five minutes before the reaction suspension was introduced into the reactor. Secondly, short times on stream (15 min) were selected for the determination of the kinetic constants in order to reduce the influence of catalyst aggregation. The use of kinetic data obtained at low conversion was also suggested by Braun et al. [18] for the quantum yield determination. Thirdly, straight lines were extrapolated to zero time regardless of the initial value of concentration.

As mentioned above, researchers have extensively used overall quantum yields

(f) to characterize the performance of photocatalysts [4]. The radiation field is, however, incorporated into this value since the volumetric average activity is divided by the volumetric average photon absorption rate. Thus, it becomes a strong function of the reactor setup because of the averaging procedure. A number of experiments with the variable reaction zone catalytic setup (Figure 1.1.2) were conducted in an attempt to relate the overall quantum yield with the reactor configuration. Several different reaction zone widths (0.15, 0.3, 0.45, and 0.6 cm) and reaction zone heights (2, 4, 6, and 24 cm)

31 were selected. As mentioned above, the thickness of the reaction annulus was varied by

utilizing Plexiglas tubes of different diameters. The length of the reaction zone was

changed by masking the UV-lamp by applying aluminum foil to the pyrex filter. It should

be reminded that for the determination of quantum yields, the conversion should be relatively low [18], so fifteen minutes on stream were selected. Average overall quantum yields were calculated in this work for each system as follows:

x A × CA ×V f = O (1) (GI × AI - GO × AO )× t where xA denotes reactant conversion, t is reaction run time, AI, AO are the surface areas of the jacket and the outer wall, respectively, V is the volume of the reactor, and GI, GO are the average values of the radiation in the z-direction in the particular reactor on the jacket and on the outer wall, respectively. Other researchers [](Cassano et al, 1995) adopted a similar definition of the overall quantum yield. It should be noted that the calculated quantum yield is an average value for 15 min since expression (1) does not incorporate the ratio of the rates of reaction and photon absorption explicitly.

Alternatively, it can be necessary to use Beer’s law to obtain GO from GI in (1) when the direct measurement of the former is difficult. This was the case for our variable reaction zone photocatalytic setup due to the partial light absorption (15-30 %) by the outer tube made of Plexiglas. One can estimate the average value of b (extinction

coefficient in Beer’s law) by knowing the dimensions of the laboratory reactor and by

measuring the incident photon flux at RI and the photon flux at RO under reaction

conditions (Figure 1.1.4). We utilized this methodology and obtained the extinction

coefficients for our photocatalytic slurries in the conventional photoreactor and used

32 those values in calculations of quantum yields. By rearranging Beer’s law for annular

geometry one can obtain the following equation for the average value of the extinction

coefficient of the slurry:

d(Gr) = -bG r (2) dr

In the above relation G is z-average radiant flux at radius r of the reaction annulus. By determining the incident radiation at the inner and outer walls of the reactor (GI and GO) and integrating equation 2 for the entire volume of the reaction zone one can obtain the average value of the extinction coefficient. Thus determined value of b encompasses the

net loss of radiant power in the radial direction of the reactor, namely, absorption,

scattering, and losses due to electron-hole recombination. Although this is not the true

extinction coefficient commonly obtained in spectrophotometric measurements, it is the

most realistic value for the determination of quantum yields since we are mainly

interested in the light attenuated in the reaction zone. Moreover, Bohren and Huffman

[19] suggested that scattered radiation is not irretrievably lost because it can be absorbed

at a different location in the reactor or scattered again. The shapes of the radiation

profiles of Figure 1.1.3 indicate that there is an almost constant radial decrease of photon

flux at any axial coordinate of the reactor subject to radiation. One would expect

significant flattening of the profiles at RO in comparison with those at RI. Furthermore, the value of the extinction coefficient determined only from the data at the midpoint of the reactor (z=12 cm) alone should in this case be expected to have the largest value. Our calculations showed, however, that, the variance of the values of extinction coefficient b

obtained from individual points of Figure 1.1.3 for Aldrich anatase was around 0.18 cm-1.

This indicates that the majority of scattered radiation is utilized effectively inside the

33 photocatalytic reactor and the overall net influence of scattering effects in such system is minimized.

When the radiation enters the inner wall of the reactor, part of it will be back- scattered. The back-scattered radiation will travel both in z- and r-direction and will enter again the reaction zone due to the cylindrical shape of the reactor configuration. This radiation can be considered part of the original radiation, which is simply utilized at a different location in the reactor. Hence, one can safely assume that the back-scattered radiation does not alter the net result. Furthermore, Romero et al [20] stated that for annular photoreactors with an opaque lamp the effect of the back-scattered radiation is negligible. We performed an experiment in order to determine whether UV light from another lamp can pass through an irradiating mercury lamp (in other words to determine its opacity to UV). Two mercury lamps (200 W and 100 W) were placed parallel to each other at a distance of 2 cm. The radiation detector was masked to leave only narrow vertical slit (0.25 cm) of its surface open and placed on the jacket of the first lamp. The first measurement involved only the 200 W lamp irradiating. The second measurement was undertaken with both lamps emitting light. The enhancement of the radiation flux emitted by the first lamp in the second experiment was found to be in the vicinity of 5% of that when only one lamp was turned on. This clearly shows that the passage of significant portions of back-scattered radiation through the lamp is very unlikely and it will reflect from the lamp and re-enter the reaction zone thus minimizing the net losses of radiation due to back-scattering.

Romero et al [20] presented extensive modeling of the radiation field inside annular photocatalytic reactors. It was based on the estimation of local volumetric rates of

34 energy absorption (LVREA) as functions of the location in the reactor by solving the

radiative transfer equation (RTE). This powerful technique indeed allows to predict the

radiation field on the basis of the specific power of the lamp. The approach adopted in the

present study is purely experimental and is based on the measurement of the incident

radiation values on the inner and outer walls of our reactors with the assumption of radial

exponential decay within the reaction zone. Moreover, the detector utilized in the present

study (described in the experimental section of the paper) can capture light impinging at

any angle, fully simulating the photocatalytic particle in the reaction annulus. In this

manner, our experimental measurements constitute an accurate means to assess LVREA.

Nevertheless, the presence of back-scattering at the inner wall may lead to overestimated attenuated photon fluxes, which in turn result in consistently lowered quantum yields.

However, the present approach does not rely on rigorous modeling to quantify the radiation field inside the photoreactor, but it is an experimental attempt to measure intrinsic overall quantum yields.

Figure 1.1.6 shows a family of dependencies of the quantum yield on the reaction

zone thickness for several reaction zone lengths. It should be noted that light extinction

(Beer’s law) is first order with respect to incident radiation and chemical reaction is first

order with respect to reactant conversion. The quantum yield (as defined by eq. 1) is

formed by the ratio of the moles of reactant converted to the number of photons absorbed.

As a result, the average absorbed photon flux and reaction rate increase simultaneously as

we reduce the reaction zone thickness. This leads to the linearity of the above

dependencies at the wider lamp openings (24 cm and 6 cm). As will be shown below, the

35 0.14 2 cm open 0.12 4 cm open 6 cm open 0.1 24 cm open 0.08 0.06 , mol/Einstein

f f 0.04 0.02 0 1 2 3 4 5 6 DR, mm

Figure 1.1.6. Dependence of average overall quantum yield on the thickness of the reaction zone in the photocatalytic reactors with variable reaction zone length for 0.25 g/l Aldrich anatase and 2 mM of 2,4,6-trichlorophenol, t=30±3 °C, pH=3.75 information from the quantum yield calculation (Figure 1.1.6) a strong function of the particular setup. This discrepancy arises from the highly non-uniform radiation distribution along the axis of the reactor. The value of the lowest point of the graph

(corresponding to the conventional reactor with the open lamp) agrees with the ones obtained by others [6] for the photocatalytic degradation of phenol. One can observe, for example, that for the open lamp reactor (24 cm) the quantum yield increases twofold when reducing the reaction zone thickness from 0.6 cm to 0.15 cm. Only slight increase is observed with the decrease in the lamp opening to 13 cm (corresponding to the

36 emissive length of the lamp). The increase in quantum yield becomes more significant for

4 cm and 2 cm opening while the twofold increase in the range 0.6 cm – 0.15 cm reaction zone thickness still holds. The upper curve of Figure 1.1.6 starts with a steep slope, but levels off for the reaction zones of 0.3 cm and 0.15 cm. The shape of the curve indicates that for openings 2 cm and below and reaction zone thicknesses 0.15 cm and below the photocatalysis occurs at maximum efficiency. Thus, the kinetic data obtained from this reaction zone can be considered intrinsic.

0.16 0.14 0.12 0.1 0.08 0.06 , mol/Einstein f f 0.04 0.02 0 0 5 10 15 20 D L, cm

Figure 1.1.7. Dependence of average overall quantum yield on the length of the reaction zone for the thinnest annulus (Aldrich anatase – 0.25 g/l, 2,4,6-trichlorophenol – 2 mM; t=30±3 °C, pH=3.75)

37 The values of f (Figure 1.1.6) are lower than unity. Similar values for quantum yields were reported by other researchers [21] for phenolic compounds. The fact that the quantum yields are smaller than unity indicates that not every photon is effectively transformed to an active radical. Since hydroxyl radicals (not photons themselves) are primarily responsible for the photocatalytic degradation [22], one can

conclude that more than one hydroxyl radical is necessary to degrade one molecule of

2,4,6-trichlorophenol. It is also seen from Figure 1.1.6 that different locations in the

annular photocatalytic reactor will have different values of local reaction rates. It is also

worth noting that such significant variation in quantum yields depending on the reaction

zone size is not due to the effect of periodic irradiation followed by dark recovery.

Indeed, even for the smallest reaction zone the space time was in the vicinity of 2

seconds. The effects of periodic irradiation take place only when the time of exposure is

less than 150 ms followed by dark recovery time in the range of seconds [23]. In order to

substantiate this point additional experiments were performed using a 0.25 g/l slurry of

Degussa P25 at 20% of original flowrate through the system (reaction zone 0.15 cm thick

and 2 cm long). The quantum yield was found to decrease by 11 % only.

It can be seen from Figure 1.1.6 that the system with the highest quantum yield is

close to representing intrinsic catalytic activity since the curve is approaching a plateau

value; the reaction volume then should be treated as a differential reaction volume. In our

case it corresponds to the reaction zone of 0.15 cm thickness and 2 cm length. Since the

quantum yield is such a strong function of the length and width of the reaction zone, one

may expect it to acquire higher values at (RO-RI)® 0. Nevertheless, the decrease of the optical thickness by lowering the photocatalyst concentration (lower extinction

38 coefficient) in the thinnest reaction zone does not lead to any further significant increase of the quantum yield, as will be shown later in the present study. This is because the entire reaction volume obeys the same rate law with respect to light intensity (contrary to the conventional photoreactor). Thus, for our system (Aldrich Anatase – 2,4,6- trichlorophenol) the average overall quantum yield becomes independent of the reactor system in the vicinity of 0.135 moles/Einstein. It can be also justified by the fact that the reaction in this case occurred only in a small volume with little radiation transfer resistance and negligible dissipation. A similar trend (plateau value of the overall quantum yield) was obtained by Davydov et al [11] for the photodegradation of phenol in a 0.5 g/l slurry of Degussa P25.

Figure 1.1.7 presents the quantum yields obtained in the thinnest reaction zone

(0.15 cm) as a function of the reaction zone length. It should be noted that the profile of the incident radiation varies significantly with the opening of the lamp (Figure 1.1.4),

resulting in different values of the quantum yield. As seen from Table 1.1.3 (column 2)

the uniformity of the radiation in the reaction zone converges to unity by decreasing the

reaction zone opening. For example, the value of the above factor is 0.96 for the opening

of 1 cm, close to complete axial uniformity of radiation in the variable length reaction

zone. Considering the leftmost points of Figure 1.1.7 one can observe that the quantum

yields for 3 cm and 1 cm are slightly larger than that for 2 cm. Thus, the values of the

quantum yield f converge longitudinally to a plateau value. This signifies that the

intrinsic value of the average overall quantum yield for the photodegradation 2,4,6- trichlorophenol has been achieved.

39 A number of experiments were performed in the aforementioned differential

volume with the other two catalysts employed in the study. They resulted in the following

values of 15-min average overall quantum yield: 0.1376 for Degussa P25 and 0.1320 for sol-gel made. These values are consistent with the trend observed previously by Fotou

and Pratsinis (1996) for the photocatalytic degradation of phenol by titania slurries in the

conventional reactor. A comparison of the above values with those obtained from the

conventional photocatalytic reactor (Figure 1.1.1) is presented in Table 1.1.4.

One can observe that the quantum yields obtained from a differential volume are

7-10 times higher than those obtained by conventional means. This is a result of almost

complete uniformity of the radiation field in the former reactor. As mentioned above,

although some radial losses of radiation are incurred in this differential volume, the

Table 1.1.4. Comparison of overall quantum yields for TiO2-assisted photodegradation of 2,4,6-trichlorophenol in conventional (apparent) and differential (intrinsic) photoreactors (catalyst concentration 0.25 g/l); the reaction zone sizes are specified in parentheses. b f (0.6x24 cm) f (0.15x2 cm) Catalyst , 1/cm app int moles/Einstein moles/Einstein Degussa P25 3.50 0.0189 0.1376 Aldrich Anatase 2.25 0.0131 0.1305 SG 1.125 0.0128 0.1320

radiation remains at the “high intensity” limit at the outer wall of the reactor and the

reaction rate law with respect to the absorbed photon flux does not change. On the

quantitative basis, this can be substantiated by experiments with lower concentrations of

the photocatalyst, which will lead to lower radial attenuation of light through the reaction

zone (lower b). Such experiments with 0.1 g/l of Degussa P25 (the most attenuating

catalyst, according to Table 1.1.4) revealed an 8.5 % increase of the overall quantum

40 yield of 2,4,6-trichlorophenol photodegradation in comparison with the one reported in

Table 1.1.4.

The difference between the quantum yields for Aldrich anatase and Degussa P25 in the conventional reactor is more significant than in the differential reactor. Apparently, this is an outcome of increased multiple scattering in the conventional reactor due to the significantly smaller primary particle size of Degussa P25 (Table 1.1.1) which leads to absorption of scattered radiation rather than incident in the conventional reactor

(increased b). Degussa P25 also possesses the higher BET surface area in comparison with Aldrich anatase, but at the same time the lower anatase content (Table 1.1.1). Such close values of the intrinsic quantum yields for the above photocatalysts indicate the significance of the crystallinity of the semiconductor powders for their photocatalytic activity. Although the high BET surface area sol-gel made catalyst exhibited relatively low activity for the degradation of 2,4,6-trichlorophenol (Figure 1.1.4) due to low anatase content (15 %), its light extinction properties are also small. This leads to the intrinsic quantum yield (determined from the differential volume) in the same vicinity as those for

Aldrich anatase and Degussa P25. Overall, the minimization of light transfer limitations in photoreactors “rectifies” the kinetic parameters of the reactor geometry and reduces the influence of the radiation field on these parameters. In the ideal case of perfectly uniform radiation in both radial and axial directions one should expect higher quantum yields for

Aldrich anatase (99 % anatase) in comparison with that of Degussa P25 (~75 % anatase) and SG (~15 %anatase).

41 The above discussion strongly suggests that a differential reaction volume does exist for photocatalytic reactors. The use of such differential reactors for kinetic studies results in useful information on the photocatalytic properties of different semiconductors.

CONCLUSIONS

The traditional reaction-engineering concept of a differential reaction volume was extended to photocatalytic systems. It was demonstrated that proceeding toward such a differential volume can minimize radiant transfer resistances and subsequently reduce the non-uniformity of the radiation field, thus yielding intrinsic properties. In order to test this idea, a variable reaction zone photoreactor was constructed and employed for the

UV-assisted photocatalytic degradation of 2,4,6-trichlorophenol in titania slurries. The local values of radiant flux density were also measured to experimentally quantify the uniformity of the radiation field. It was shown that the average overall quantum yield achieves a plateau value with a decrease in the reaction zone size. The methodology of obtaining kinetic data from a differential volume of uniform radiation field, which simultaneously offers minimal extinction of light, can be used for the determination of intrinsic catalytic properties.

ACKNOWLEDGEMENTS

The authors wish to thank TAPPI foundation for support. The authors also wish to thank Hewlett Packard Company for the donation of A HP-6890 Series gas chromatograph under the HP U.S. Instructional Equipment grants. The award (HP

42 001770318) was allocated to Prof. P. G. Smirniotis for the project entitled “Enrichment

of the Chemical Engineering Curriculum”.

NOMENCLATURE

A = surface area, cm2

L = reactor length, cm

G = monochromatic specific incident radiation rate, W/cm2

G = z-average radiant flux, W/cm2 r = radial coordinate, cm t = time on stream, min

V = volume of the slurry reactor surrounding the lamp, cm3

z = axial coordinate, cm

Greek Letters

b= extinction coefficient, cm-1

fappt = apparent overall quantum yield (conventional reactor), moles/Einstein

fint = intrinsic overall quantum yield (plateau value), moles/Einstein

Subscripts

I= pertaining to the inner wall of the reaction annulus

O = pertaining to the outer wall of the reaction annulus

m = maximal value in the range

43 REFERENCES

1 Okamoto, K.; Yamammoto, Y.; Tanaka, H.; Tanaka, M.; and Itaya, A., Bull. Chem.

Soc. Jpn., 1985, 58, 2015-2022.

2 Ollis, D. E., Environ. Sci. Technol., 1985, 19, 480-488.

3 Calvert, J. G. and Pitts, J. N., Photochemistry. John Wiley & Sons, 1967.

4 De Bernardez, E. R.; Claria, M. A.; and Cassano, A. E. In "Chemical Reaction and

Reactor Engineering", (J. J. Carberry and A. Varma, Eds.), Marcel Dekker, New York,

1987, Vol. 26, p. 839-921.

5 Sun, L. and Bolton, J. R., J. Phys. Chem., 1996, 100, pp. 4127-4134.

6 Fotou, G. P., and Pratsinis, S. E, Chem. Eng. Comm., 1996, 151, 251-269.

7 Tsekov, R. and Smirniotis, P. G., Chem. Eng. Sci., 1997, 52, pp. 1667-1671.

8 Palmisano, L.; Augugliaro, V.; Campostrini, R.; and Schiavello, M., J. Catal., 1993,

143, pp. 149-154.

9 Karakitsou, K. and Verykios, X. E., J. Catalysis, 1995, 152, p. 360-367.

10 Serpone, N., J. Photochemistry Photobiology A: Chem., 1997, 104, pp. 1-12.

11 Davydov, L.; Smirniotis, P.G.; and Pratsinis, S.E., Chem. Eng. Sci., 1998, submitted

12 Fotou, G. P.; Vemury, S.; and Pratsinis, S. E., Chem. Eng. Sci, 1994, 49, p. 4939-

4948.

13 Coil, R. F., M.S. Thesis, University if Cincinnati, 1997

14 Dibble, L. A. and Raupp, G. B., Catalysis Letters, 1990, 4, 345-354.

15 Cabrera, M. I.; Alfano, O. M.; and Cassano, A. E., Ind. Eng. Chem. Res. 1994, 33,

3031-3042.

44

16 Hoffmann, M.R., Martin, S.T., Choi, W., and Bahnemann, D.W., Chem. Reviews, 95,

69 (1995)

17 Jardim, W.F.; Moraes, S.G.; and Takiyama, M.M.K., Wat. Res., 1997, 31, pp. 1728

1732

18 Braun, A.; Maurette, M. T.; and Oliveros, E., Technologie Photochemique, 1st edition; Presses Polytechniques Romandes: Lausanne, Switzerland, 1986

19 Bohren, C. F. and Huffman, D. R., Absorption and Scattering of Light by Small

Particles. John Wiley & Sons, 1983.

20 Romero, R.L.; Alfano, O.M.; and Cassano, A.E., Ind. Eng. Chem. Res., 1997, 36, pp.

3094-3109

21 Matthews, R. W., J.Catal., 1988, 111, 264-272.

22 Turchi, C.S. and Ollis. D.F., J. Catal., 122, 178 (1990)

23 Sczechowski, J. G.; Koval, C. A.; and Noble, R. D., Che. Eng. Sci., 1995, 50, pp.

3163-3173

45

Chapter 1.2. The intrinsic catalytic activity in photoreactors

INTRODUCTION

Several variables have been introduced to describe photocatalytic activity. Primary

quantum yield represents the number of radicals produced in the primary process (most

frequently hydroxyl radicals) on the surface of a photocatalyst per single photon absorbed

[1]. It is probably the most accurate way to characterize photocatalytic activity. This

variable is, however, difficult to obtain in simple photocatalytic experiments since only

indirect methods are currently available to quantify primary processes. One of them is the

use of a probe molecule capable of admitting only one hydroxyl radical (such as methanol)

after which the oxidation ceases completely [2]. Overall quantum yield or quantum

efficiency represents the number of molecules degraded per photon sent into the system or

absorbed by the system [1]. It is probably the most common value to be used by different

researchers for characterization of photocatalysts, and it was shown [3] to vary

significantly with the reaction zone length and thickness. Apparent reaction rate constants

are also widely used for evaluation of photocatalysts for the degradation of particular

compounds since most phototocatalytic reactions are pseudo-first order for dilute organics.

They, however, differ significantly as a function of the radiative power of the lamp and

concentration of the catalyst employed in the same reactor [4], which restrains their use as

a universal characteristic.

Several researchers attempted to relate photocatalytic reaction rate with radiation

intensity. Okamoto et al attempted [5] to incorporate the formation rate of active hydroxyl radicals into the reaction rate equation by:

dC n - = F · I f (c,P ) (1) dt OH O2

47 where F·OH is the quantum yield for the formation of hydroxyl radicals, I is the rate of

radiation absorption, PO2 is the oxygen partial pressure, and n is a constant.

A number of attempts have also been made to define a common means of

comparison of reaction rates for different photocatalysts. For instance, an expression was

proposed [6] for the intrinsic rate of hydrogen production in the photocatalytic cleavage of

water. It represented the ratio of the hydrogen production rate to the optical transparency

(ratio of the transmitted radiation to the incident radiation) of the slurry. Relative photonic

efficiency [7] was introduced as yet another means to compare the photocatalytic behavior

of different semiconductors. Its value for a standardized system (catalyst, reactor, and

actinometer) was proposed to be unity. The quantum yield of each particular photocatalytic

system then becomes a product of its relative photonic efficiency with the apparent

quantum yield. Rigorous kinetic approach coupled with photon balance was also used [8]

to model the photocatalytic degradation of trichloroethylene in water in a reactor with a

parabolic reflector. Local volumetric rate of energy absorption was used to characterize

the radiation field inside the photoreactor, and major species’ balances accounted for the

photochemical transformation.

The present study attempts the determination of intrinsic reaction rate constants for

photocatalysts from their apparent activity and the radiation field in the photoreactor. The

performance of various TiO2 powders in the photocatalytic degradation of phenolic

compounds (shown to obey first order rate law [3,4]) in an annular reactor is used to test

our proposal. The present method is based on measurements of the incident photon fluxes at

different reactor locations and uncoupling the non-uniformity of the radiation field from apparent activity. The reaction rate constants obtained from our model are then compared

48 to the experimental ones obtained for select catalysts in a differential photocatalytic

volume [3].

EXPERIMENTAL

Commercial titania powders were evaluated for the photocatalytic degradation of

phenol. The titania powders utilized in the present study were Aldrich anatase, Aldrich

anatase 325, and Degussa P25; their characteristics are shown in Table 1.2.1.

Table 1.2.1. Photocatalytic powders of titania employed in the present study. BET Surface Primary Particle Anatase Content, Area, m2/g Size, nm wt% Degussa P25 75 a 21b 70-80 b Aldrich Anatase 10 a 156c 99 b Anatase 325 mesh 9.75a < 44 mm b 99 b a. Measured by nitrogen adsorption Micromeritics Gemini 2320 surface analyzer at 77 K b. According to manufacturers’ specification c. Equivalent diameter

The choice of the catalysts was based on their different photocatalytic properties (Degussa

P25 the highest, Aldrich anatase 325 the lowest). The photocatalytic degradation of phenol

(reagent grade, Fisher) over various TiO2 suspensions was carried out in an annular slurry quartz reactor [4]. The suspensions (0.1, 0.25, and 0.5 g/l) were prepared with the titania

particles dispersed ultrasonically in the aqueous solutions containing phenol. The pH of

each solution was adjusted to 3.75 by 0.5 M H2SO4. Before each experiment, the UV lamp was warmed up for 5 min. Samples were withdrawn from the reactor and filtered with 0.2 mm membrane filters (Gelman Sciences) to remove the titania particles. The suspension temperature was maintained at 30+3 oC. The concentration of the unreacted compound was

measured on a GC (HP-6890 series, FID) by direct injection using 4-chlorophenol (reagent

49 grade, Fisher) as a reference peak. A capillary column (Chrompack 7584) was used for

separation, and helium was used as a carrier gas. Other experimental conditions can be

found in Table 1.2.2. The reaction mixture was tested for the presence of intermediates at

different time on stream by direct injection into GC (HP 5890 Series II) equipped with a

MS (HP-5972) and no detectable amounts of intermediate compounds were found.

Table 1.2.2. Geometric parameters and experimental conditions for four different photocatalytic setups employed in the present study. Setup# Annulus Reactor Oxygen flow Slurry thickness volume, rate, flow rate, DR, mm liters ml/min ml/min 1 1.5 0.200 200 150 2 3.0 0.325 300 150 3 4.5 0.425 400 150 Conventional 6.0 0.650 500 n/a (Ace Glass)

A conventional liquid-phase photocatalytic reactor (Ace Glass, Inc., Cat. No.

7868-10) was used to obtain clearly apparent kinetic data as described elsewhere [4,5]. A

double-walled quartz immersion well was placed in the middle of the reaction vessel. Its

purpose was to allow the circulation of water for cooling the light source and the solution

and also filtering off the IR part of the lamp emission spectrum. The UV lamp radiation

was filtered by a Pyrex filter (7740, Ace Glass, Inc.) placed between the lamp and the

immersion well. The thickness of the Pyrex filter was 2.38 mm, the inside diameter was 26

mm. The immersion type UV-radiation sources were 100 W, 200 W, and 450 W (medium pressure) mercury vapor quartz lamps. The radius of the lamps was 1 cm and their emissive length was 13 cm (200 and 450 W) and 7 cm (100 W). Circulation of the suspension within the reactor assured complete mixing. Oxygen (Wright Brothers, 99.5 %) was sparged into the solution at 500 cm3/min through a fritted glass tube. With this flow

50 rate no deposition on the walls of the reactor was observed. Further increase of the oxygen

flow rate did not lead to discernible activity increase, since the amount of oxygen supply to

the suspension exceeded by far the stoichiometrically needed. The entire photoreactor system was kept inside a UV-safe cabinet.

The titania powders were also evaluated in a variable reaction zone photoreactor (Figure

1.2.1) with the same lamp and cooling jacket but different thickness of the reaction zone. It is worth noting that the length/width ratio of the reactor is much larger than shown. Table

1.2.2 presents the parameters of this setup. The thickness of the reaction zone was varied by using different outer tubes (Plexiglas), while the height was adjusted by putting aluminum foil on the pyrex filter and black electric tape on the outer wall of the cooling jacket. The latter was done to minimize the effect of sphericity of light emission and thus be able to use the realistic reaction zone height in further calculations. With this variable reaction zone configuration we were able to minimize light transfer limitations and also approach the uniformity of radiation both along z- and r-directions. A conventional magnetic stirrer vigorously agitated the suspension in the reactor. A laboratory peristaltic pump (Masterflex 7017-21) was used to run the suspension through the reactor. An intermittent vessel was used for degassing of the suspension and sample withdrawal. This vessel was also agitated by a magnetic stirrer in order to avoid sedimentation of titania and provide better dispersion of oxygen bubbles. Flexible plastic tubing connected the pump, vessel, and reactor. The kinetic constants resulting from the above experiments were corrected for the illuminated volume of the reactor.

51 Figure 1.2.1. Experimental reactor configuration with variable reaction zone thickness and length

A radiometer (International Light, Inc. Model IL 1700) was used to determine local

incident photon flux densities in the reactor connected with a detector (International Light,

Inc., Model SED033 #3435). A UV-filter was employed in order to measure only the light

of interest and its response exhibited a strong maximum at the wavelength of 355 nm. The

diameter of the detector was 1 cm, and the inner diameter of the reaction zone was 5.4 cm.

The incident radiation at RI as well as the outgoing radiation at RO was measured at various locations along the reactor length by the detector at the outer wall of the quartz

52 cooling jacket. The attenuated photon flux can thus be obtained by measuring the incident

radiation flux and the outgoing radiation flux at the outer radius of the reactor (RO) when the suspension is present.

THEORY

Numerous works have elaborated on the kinetics of photocatalytic transformations

[9]. In the present study we will adopt the following simplified kinetics when considering photocatalytic processes. Quanta of light are absorbed by the photocatalyst particle producing one electron-hole pair per photon. The latter engages into reactions with water and oxygen to produce active oxygen species (we will limit them to OH-radicals for simplicity). The rate of generation of these radicals can be expressed similarly to equation

1 [10] for the heterogeneous photodegradation of methylene blue with ZnO as a catalyst. In their studies the rate of generation of surface oxygen species was equal to the product of the quantum yield F·OH, measured in molecules per photon with m×G, which is the local

absorbed photon flux density in the reactor:

d[·OH ] = F · mG dt OH (2)

It was shown experimentally [12] that hydroxyl radicals attain a steady state

concentration almost instantaneously in photocatalytic systems. At such steady state the

consumption and recombination balance the generation of radicals. The recombination rate

is accounted for by the experimentally obtained primary quantum yield (in the form of [2]),

while the local rate of radical consumption is represented by the following equation for the

bimolecular interaction with the reactant:

53 d[ A] - = k [· OH ][A] dt S (3)

where [A] denotes the organic substrate concentration, t – time, kS – bimolecular reaction

rate constant, and [·OH] – hydroxyl radical concentration. It should be noted that no

dependence of the reaction rate on the oxygen concentration (or partial pressure) is

assumed since an excessive amount of O2 is supplied to the reactor. For the general case of

polychromatic radiation, an integral form of equations 2 and 3 over the entire range of

frequencies is required [12]. Furthermore, considering an “instantaneous” rate of the

pollutant degradation at time t, we can equate the above two relations for the entire volume

of the reactor:

l2 · F · (l)m(l)G(l)dVdl = k [ OH ][A]dV òò OH ò S l1V V (4)

where m – absorption coefficient, l – radiation wavelength, and V – reaction volume.

One can observe that kS represents the bimolecular reaction rate constant of the organic and hydroxyl radical in the adsorbed state. However, only the product of kS with

[·OH] can be determined experimentally and is usually referred to as apparent reaction rate constant (kapp). Furthermore, the steady-state concentration of active radicals differs significantly with absorbed radiant power [5] as well as catalyst concentration and reactor geometry. This leads to different values of the apparent reaction rate constant reported for the same photocatalyst and VOC using different photocatalytic reactors. Obtaining the value of kS from the above equation is a difficult task since it requires in situ measurements

of the concentration field of active radicals. Alternatively, one can introduce a uniformly

54 illuminated “ideal” photoreactor (loaded with the same concentration of the same catalyst)

wherein no light transfer resistances occur. This will make a uniform radiation profile (for

flat and cylindrical geometry, provided the thickness of the annulus is small) and yield

intrinsic photocatalytic properties. In such ideal system the concentration of active radicals

will also be uniform. Therefore, the above equation will take a different form for the ideal

reactor:

l2 0 0 0 · 0 F · (l)m (l)G (l)dl = k [ OH ] [A] ò OH S l1 (5)

Relying on the traditional chemical reaction engineering approach, we propose to introduce an effectiveness factor for photocatalytic systems, which will take into account the radiation field non-uniformity inside the reactor. Using such effectiveness factor we assume that the intrinsic reaction rate can be related with the apparent one by equation (6), obtained by dividing equations (4) and (5):

l 2

F · (l)m(l)G(l)dV dl ò ò OH k [ ·OH ][A]dV l V ò S h = 1 = V l · 0 (6) 2 kS [ OH ] [A]V 0 0 0 V F · (l)m (l)G (l)dl ò OH

l1

The right hand side of the above relation represents the ratio of reaction rates in a real system and reference-ideal system, which is directly related to the radiation field

(uniform in the second case). The left-hand side is the ratio of the net rate of radical generation in the real photoreactor to that in the ideal one. The hydroxyl radical concentration in the ideal case [·OH]0 represents that corresponding to the maximal attainable rate of radiation absorption throughout the differential reaction zone for the

55 particular lamp. Thus, h becomes a properly formulated function of the reactor geometry, size, and radiation field. Since the latter is a function of the catalyst type and concentration, one has to estimate h for each reactor geometry and catalyst in order to determine its activity at uniform irradiation.

1.E-01

1.E-02 2

1.E-03 Radiant Flux, W/cm 1.E-04 Incident 450W Outgoing (AA) Outgoing (P25) Incident 100W

1.E-05 0 5 10 15 20 25 Reactor Height, cm Figure 1.2.2. Experimentally measured radiation profiles inside the photocatalytic reactor: 450 W light source; incident radiation profile for 100 W light source

Based on equation (6), h is expected to be smaller than unity. Radiation scattering

and absorption by the suspension, and also the effect of the reactor geometry and size are

present in a real system. These factors do not allow the maximal UV-light utilization.

· Although the above analysis does not allow determining kS or [ OH] explicitly, it allows to estimate the ratio of the average concentration of hydroxyl radicals in the real photocatalytic system to that in the ideal case of the uniform light conditions throughout the

56 reactor. Observing the photon fluxes allows one to conclude from Figure 1.2.2 that

different locations in the conventional annular photocatalytic reactor [3,4] will have

different values of local reaction rate. As mentioned, detailed information about the

emission spectrum of the light source is necessary to find the true rate of generation of

active species in the reactor. It was shown [14] that only a very narrow band of radiation

(about 350-360 nm) is utilized in photocatalytic reactors with titania as a catalyst when a mercury lamp is used in a pyrex reactor. Since our reactor was made of the same materials as the one in the above study, we can treat the UV radiation passing through the suspension as almost monochromatic at l=355 nm. Hence, the integral over the range of wavelengths in equation 6 can be eliminated in order to simplify our model.

The above theory is expected to be valid for any reactor geometry and size. For the case of an annular reaction zone additional simplifications can be applied. The primary quantum yield is related to the effectiveness of the catalyst to generate active radicals when quanta of radiation reach the catalyst surface. Moreover, the quantum yield for ·OH generation was shown [2] to exhibit insignificant dependence on the photocatalyst concentration. Therefore, it can be considered a sole function of the catalyst, which remains the same in the real and reference-ideal systems within a reasonable range of catalyst concentrations, regardless of the location in axial and radial direction.

Furthermore, the concentration of a VOC in the slurry is very low when pollutant degradation is considered and the absorption of light by the dissolved organic can be neglected. The photocatalyst is present in our suspension in low to medium concentrations

(0.1-0.5 g/l) which renders unreasonable to expect nonlinear optical behavior. Thus, it is equitable to assume that the absorption coefficient is independent of the reactor system and

m will be the same for the ideal and real cases.

57 Local radiant flux densities (incident radiation) G do differ with different catalysts, their concentrations, and reactor geometries. We therefore assume that in the ideal case with no radiation transfer resistances G is equal to the maximal incident radiant flux in the reactor. Furthermore, the rate of radiation absorption is the same throughout the reaction zone. The length of the latter in the real system is determined by extending the lateral lines of descent in Figure 1.2.2 to the x-axis. For the real reactor both the geometry and radiation transfer resistances must be taken into account. Due to the annular geometry utilized in the present study we expect the beams of radiation reflected at the inner wall to compensate each other (as suggested in [14]), and the back-scattered radiation is not lost. A comparison of the shapes of the axial radiation profiles of Figure 1.2.2 indicates a constant radial decrease of photon flux at any axial coordinate. This means that the presence of the photocatalyst does not qualitatively change the symmetric axial radiation profiles in the reactor. Indeed, if the extent of angular scattering were high, one would observe a significantly flatter profile at RO in comparison with that at RI. This indicates that the majority of angularly scattered radiation is utilized effectively inside the conventional photocatalytic reactor. Therefore, the major input of scattering takes place in the forward and backward direction.

With the incorporation of the above assumptions equation 6 for our system reduces to:

òGdV h = V (7) ò G 0dV V

58 The local value of incident radiation for the ideal case (G0) should be independent of z and equal to that at the midpoint of the reactor at the radius RI. Such radiation intensity in the ideal case is assumed not to decay with distance due to the absorption and/or scattering.

One can measure experimentally the radiant flux on the cooling jacket using a radiation detector, thus obtaining a realistic profile of radiation entering the reaction zone

(Figure 1.2.2, upper curve and lowest curve). According to the above discussion, angular scattering can be neglected and two-flux radiation model can be applied [15] for such one- dimensional propagation of radiation. It assumes that the scattering remains isotropic throughout the vessel and the absorption and scattering of radiation are accounted for by two independent coefficients. Moreover, this model lumps together all forward-scattered beams into one flux and backward-scattered into another. We will therefore adopt the

equation derived in [16] for the two-flux radiation transfer in a suspension of microorganisms with a modification for annular geometry. This relation is given by:

RI (1+ a)exp[-d(r -1)]- (1- a)exp[d(r -1)] G(r,z) = 2G I (z) 2 2 r(RO - RI ) + RI (1+ a) exp(d) - (1- a) exp(-d)

(8a)

where GI denotes the incident radiation (or radiant flux at the inner radius of the inner wall

RI), z – axial coordinate. The values of parameters a, d, and r are determined from the

following relations:

r - R r = I RO - RI a = (1- w)1/ 2

d = bC cata(RO - RI ) (8b)

59 where RO is the radius of the outer wall, r - radial coordinate, w =s/b– scattering albedo,

b – extinction coefficient of the slurry, and Ccat – photocatalyst concentration. Substituting

equation 8a into equation 7 will give us the value of h for the annular photocatalytic

reactor. In order to use equation 8a one needs to estimate the values of b and w from

independent spectrophotometric experiments without and with an integrating sphere,

respectively. In the present study the extinction coefficient was determined for the range of

concentrations 0.025-0.2 g/l at reaction conditions (in the region of linear response of the spectrophotometer). The scattering albedos for the catalysts used were obtained from

literature [2, 7]. The parameters of equations 8a and 8b are summarized in Table 1.2.3.

The value of h defined by equation 7 can be now calculated from simple experimental measurements. Therefore, the intrinsic reaction rate constant and the turn over

frequency (TOF) of each catalyst (measured in molecules site-1s-1) can be determined.

These two parameters are independent of the individual reactor setup and present universal

characteristics of photocatalysts. The estimation of the TOF for our catalysts was based on

initial reaction rate of phenol degradation, which was then divided by the number of sites.

The estimation of the number of sites was based on the BET surface area of the catalyst, its

anatase content, and surface site concentration of 5×1014 sites/cm2 [17].

RESULTS AND DISCUSSION

The axial radiation distribution measured for a 100 W and 450 W medium pressure

mercury lamp in the conventional photocatalytic reactor are shown in Figure 1.2.2. Similar

60

Table 1.2.3. Parameters of equations 7 and 8 used in the present study. Parameter Catalysts AA b P25 b AA325b b, L/(g cm) a 12.85 16.0 13.98 c d c w 0.478 0.586 0.478

Lamps (Watts) 450 200 100 L, cm 19.5 e 16.5 e 12e m 2 GI , W/cm 0.051 0.048 0.00465 a. Measured using UV-Vis spectrophotometer Shimadzu U160 at 355 nm b. AA = Aldrich anatase P25 = Degussa P25 AA325 = Aldrich anatase 325 mesh c. From reference [2] d. From reference [7] e. Measured as the opening between the two intersects with the x-axis of the parabolic incident radiation profile

profiles have been obtained for a 200 W lamp [3]. The profiles are represented by bell-

shaped curves with the maximum located at the center of the light source. One can observe

that the radiant flux at the reactor ends can be two orders of magnitude smaller than that in

the middle. It is also worth noting that the radiation decreases about one order of magnitude

in the radial direction when 0.25 g/l titania suspension is used (Figure 1.2.2, second from

the top curve). Moreover, the radiation reduction by attenuation in the ambient suspension

can be even larger for higher concentrations of the photocatalyst [4], adding to the radiation

field non-uniformity.

Overall quantum yields (f) are extensively used to compare the performance of

photocatalysts [18]. The radiation field non-uniformity is, however, incorporated into this

value since the volumetric average activity is divided by the volumetric average photon

absorption rate, thus making it a strong function of the reactor setup [3]. A number of experiments with the variable reaction zone photoreactor (Figure 1.2.1) were conducted in

61 an attempt to relate the overall quantum yield of phenol degradation with reactor

configuration. Several different reaction zone widths (1.5, 3, 4.5, and 6 mm) and reaction

zone heights (2.25, 4, 6, 8, and 12 cm) were selected. It should be noted that fifteen minutes

on stream were chosen for the determinations of quantum yields, so as to keep conversions

relatively low [19]. Average overall quantum yields were calculated in the present study

for each system from the relation:

x C V A AO f = (9) m ò GdV × t V

where xA denotes reactant conversion, t is reaction run time, AI and AO represent the

corresponding light transmitting surface areas, m is the absorption coefficient of the slurry

and G is the local radiation intensity in the particular reactor.

Figure 1.2.3 shows the average overall quantum yield as a function of the reaction

zone thickness for the reaction zone length of 6 cm (lower curve). It clearly shows that the

information from the quantum yield calculation can be very subjective and is a strong

function of the particular setup. One can observe, that by increasing the reaction zone

thickness the quantum yield decreases, thus providing lower activity per photon absorbed.

This difference arises from the highly non-uniform radiation distribution along the radius of the reactor due to its high attenuation (up to 2 orders of magnitude). In order to find a reactor configuration, which will ensure photocatalytic activity approaching intrinsic we chose to use 0.5 g/l of Degussa P25 powder with a 100 W mercury lamp. This selection is based on the fact that low incident photon flux and high attenuation of light are the least favorable conditions for obtaining a uniform radiation profile. Indeed, Degussa P25 is the

62 DL, cm 1 6 11 0.13 0.19

0.12 0.17 0.15 0.11 0.13

0.1 1/min ,

mol/Einstein 0.11 k , 0.09

phe 0.09

f f vs R 0.08 k vs L 0.07 0.07 0.05 0.1 0.2 0.3 0.4 0.5 0.6 0.7 DR, cm

Figure 1.2.3. Lower curve: Dependence of average overall quantum yield on the thickness of the reaction zone in the photocatalytic reactor with variable reaction zone (Figure 1.2.1), 6 cm length for 0.5 g/l Degussa P25 and 2 mM of phenol, t=30±3 °C, pH=3.75, P=100W; Upper curve: dependence of first order kinetic constant on the reaction zone length for the same conditions, obtained in reaction zone thickness 0.15 cm.

most light attenuating powder among the catalysts tested (Table 1.2.3) and at this

reasonably high concentration it gives around two orders of magnitude radial decrease in

the conventional photocatalytic reactor [4]. The choice of the weakest lamp (100 W) is

also justified by the fact that it will create the highest non-uniformity of the radiation field.

Therefore, if such least favorable system approaches intrinsic reaction rate constant for

some reactor configuration, other systems (photocatalyst, lamp) will definitely provide

same. One can observe that the value of the average overall quantum yield arrives at a

plateau around 0.1 moles/Einstein for the reaction zone thickness of 0.3 and 0.15 cm. This

63 behavior indicates that the intrinsic activity is achieved since the further reduction of the reaction zone does not lead to higher activity.

The thinnest reaction zone can thus be considered a differential reaction volume. It was tested for several reaction zone lengths (12, 8, 6, 4, and 2.25 cm). One can observe that the reaction zone reduction beyond 6 cm does not increase the reaction rate constant

(Figure 1.2.3, the upper curve), which signifies that the kinetic data obtained from the smaller three reaction zones can be considered intrinsic. The plateau value of the reaction rate constants in a thin reaction zone (0.15 cm) occurs as we decrease its length. This is because having a fraction of emissive length of the lamp covered (6, 4, and 2.25 cm opening) the radiation output of the 100 W lamp (or any lamp used in this study) becomes almost completely uniform. For this reason, in the case of smaller lamp openings the reaction rate constant becomes directly proportional to the reaction zone length, which in turn leads k (equal to kapp corrected for the irradiated volume fraction) to a plateau. We will therefore use the largest reaction zone (6 cm) of the above three in order to minimize the experimental error.

Taking all the above into account, we shall therefore treat the first order reaction rate constant obtained from the highest points on Figure 1.2.3 as fully representing intrinsic activity of the catalyst employed. This can be also justified by the fact that the reaction in this case occurred only in a small volume with insignificant radiant transfer resistance. We will therefore use the kinetic constants obtained from these data to validate equation 6 experimentally. We consider high concentration of a highly extinctive photocatalyst irradiated by a low-intensity UV-lamp the least favorable system. More specifically, we chose 0.5 g/l of Degussa P25 with 100 W lamp. According to Figure 1.2.3, a plateau value of the photoactivity has been achieved in the above reaction zone with such system. The plateau value of the photoactivity is therefore expected to be valid for other catalysts under

64 comparable conditions because the radial light distribution is more uniform for less

extinctive reaction media (smaller b). Similarly, higher incident light intensity (achieved

by using a higher wattage lamp) also adds to more uniformity of the radiation field in the

radial and axial directions. Since we isolated the differential reacion zone for the least

favorable system, this same reaction zone will hold for all the systems utilized in the

present study, which will allow us to validate our theory. One can observe from equation

6, that the numerator of the right hand side is a commonly measured apparent reaction rate.

The denominator of the above equation represents the maximal attainable apparent reaction rate, which is considered the ideal case of uniform reaction field. Once h is determined

from the first part of equation 7, it becomes possible to calculate the intrinsic initial

reaction rate of the reactant degradation from apparent kinetic data.

Table 1.2.4 presents the effectiveness factors determined by equation 7 and several

intrinsic reaction rate constants for several reaction volumes approaching a differential

one. The constants were determined by equation 6 as follows. The initial apparent reaction

rate (numerator of the right-hand side of equation 6) was measured experimentally.

Utilizing the effectiveness factor h as described above, the initial intrinsic reaction rate was obtained. In Table 1.2.4 such rate is represented by the first order reaction rate

· 0 constant, which is a product of kS and [OH] in equation 6. The constants were then

divided by the catalyst concentration and corrected for the illuminated volume fraction

since not the whole reactor length was subject to radiation. It can be observed that by

decreasing the reaction volume (length and thickness of the reaction zone) the effectiveness

factor h increases monotonically. Subsequently, the intrinsic reaction rate constant

increases. This is because the system chosen for comparison is least favorable for the

determination of catalytic properties, as described above. Indeed, such low wattage of the

65 lamp in conjunction with the high concentration of highly extinctive titania powder lead to the values of the effectiveness factor ~0.55 and below.

Table 1.2.4. Comparison between photocatalytic effectiveness factors (eq. 7) and intrinsic kinetic constants for several reaction zones (initial concentration of phenol 2 mM) Reaction Zone Parameter 0.5 g/l P25 100 W 0.1 g/l P25 450 W Lamp Lamp h 0.5587 0.8474 0.15x6 cm kint, 1/min 0.58 1.62 h 0.4060 0.7531 0.3x6 cm kint, 1/min 0.49 1.43

0.45x6 cm h 0.3077 0.6771 kint, 1/min 0.45 1.29 Conventional h 0.1665 0.4266 Reactor kint, 1/min 0.38 1.35 Note: In the reaction zone column the first dimension corresponds to the thickness, the second one to the length; the reaction zone parameters for the conventional reactor are 0.6x12 cm for 100 W lamp and 0.6x19.5 cm for 450 W lamp.

Furthermore, more favorable systems (higher wattage lamp, differential reactor, low catalyst concentration) should yield h approaching unity and kint converging to a single value. One can also observe from Figure 1.2.2 that the peak irradiance of a 100 W lamp is more than order of magnitude smaller than that of a 450 W lamp. This leads to low penetration length for the radiation emitted by the 100 W lamp into an optically dense 0.5 g/l slurry of Degussa P25. As a result, the illuminated volume for which the reaction rate constants are corrected is overestimated, and the latter attain lower values as the reaction zone is decreased radially. The presence of such defunct volume in the reaction zone explains the discrepancy between the reaction rate constants observed from Table 1.2.4 for the 100 W lamp. On the contrary, the last column of Table 1.2.4 corresponds to the most favorable (in terms of light transfer) operating conditions. One can clearly observe that the

66 intrinsic kinetic constants converge to a much greater extent than those obtained utilizing

100 W for rather concentrated slurry. This illustrates that in order to achieve the intrinsic

activity the whole reaction zone should be irradiated uniformly. For this reason, higher

power lamps and lower concentration slurries are preferable since they allow to correctly

estimate the illuminated volume (which is difficult for 100 W lamp and 0.5 g/l of P25) in

our case.

A comparison of kint calculated by equation 6 from apparent data for three catalysts

and three different light sources are presented in Tables 1.2.5a, 5b, and 5c. The values of h

calculated by equation 7 for the conventional reactor are about 2.5-7 times smaller than unity suggesting that one cannot rely on the apparent reaction rate constants or overall quantum yields as measures of photocatalytic activity. The photocatalytic effectiveness factors are generally smaller for Degussa P25 than those of Aldrich anatase and Aldrich anatase 325. This is caused by different light transfer properties mainly due to the different primary particle size. One can also observe that the effectiveness factors for both anatase powders in all three systems have very close values, slightly smaller for AA325. This is because the optical properties of their slurries (Table 1.2.3) are not significantly different.

At the same time, the activity of AA325 is slightly lower than that of AA. A comparison of the BET surface areas of the two leads to a conclusion that the only source of the increased activity of AA is its much narrower particle size distribution. Indeed, according to the manufacturer’s specification this value is 0-44 mm for AA325. The determination of

67

Table 1.2.5. Comparison between theory and experiment for three different photocatalysts and three different light sources. a. 100 W lamp, 0.5 g/l of photocatalyst Parameter Catalyst AA325 AA P25 h (0.15x6 cm) 0.3864 0.3968 0.5587 kint, L/(min gcat) 0.36 0.46 0.58 h (0.6x12 cm) 0.1440 0.1542 0.1665 kint, L/(min gcat) 0.36 0.37 0.38

TOF, 1/s 0.15 0.15 0.027

b. 200 W lamp, 0.25 g/l of photocatalyst

Parameter Catalyst AA325 AA P25 h (0.15x6 cm) 0.7253 0.7367 0.7117 kint, L/(min gcat) 0.64 0.77 0.98 h (0.6x16.5 cm) 0.2629 0.2748 0.2577 kint, L/(min gcat) 0.76 0.74 1.17

TOF, 1/s 0.3150 0.2982 0.0837

c -450 W lamp, 0.1 g/l of photocatalyst

Parameter Catalyst AA325 AA P25 h (0.15x6 cm) 0.8570 0.8633 0.8474 kint, L/(min gcat) 0.77 1.12 1.62 h (0.6x19.5 cm) 0.4360 0.4458 0.4266 ’ kint , L/(min gcat) 0.83 0.83 1.35 TOF, 1/s 0.3458 0.3382 0.0961

68 surface acidity of all the photocatalysts (according to [20]) yielded negligible difference

between AA and AA325. As mentioned above, the photocatalytic activity is also strongly related to the BET surface areas of the titania powders. The difference in this value for both anatase powders and Degussa P25 is more than sevenfold. It contributes significantly to the increased activity of Degussa P25.

Comparing the value of effectiveness factors h and intrinsic reaction rates for the

three systems employed in the study (Tables 1.2.5a, 5b, 5c) elucidates the following

phenomena. As mentioned above, the increase of the lamp wattage leads to a more uniform

radiation profile (compare upper and lower curves of Figure 1.2.2). Also, the decrease of

the catalyst concentration results in smaller values of attenuation per unit volume (smaller

b). Evidently, the above information indicates that the third operation system employed in

the study (450 W lamp and 0.1 g/l of catalyst) should be the most favorable for the

determination of intrinsic properties. In fact, such system gives the values of h close to

unity for the differential volume (Table 1.2.5c).

The values of kint (Tables 1.2.5a, 5b, 5c) correspond to the smallest and largest

reaction volumes. One can observe that for 100 W lamp and 0.5 g/l of catalyst (Table

1.2.5a) the values of the intrinsic reaction rate constants are in very good agreement for

each individual catalyst in the small and large reactors. Some discrepancy arises for the

system with the high concentration of Degussa P25 photocatalyst. As discussed above, this

is the most unfavorable system characterized by the highest non-uniformity of the radiation

field, and the effect of a smaller reaction zone played its role. The second operating system

(Table 1.2.5b) shows that the reaction rate constants for individual catalysts converge to

comparable values. The third system (highest incident radiation and lowest catalyst

concentration used in this study) presented in Table 1.2.5c is expected to be the most

69 favorable because of the highest values of the effectiveness factor (~0.9). Indeed, we observe that the intrinsic reaction rate constants determined form the smallest and largest reaction volumes converge, especially for the anatase powders (AA and AA325).

The anatase content and BET surface area also play a significant role in photoactivity. Therefore, an effective measure of the activity of catalytic sites is an absolute necessity, and turnover frequencies (TOF) can serve as such measure; their values for each catalyst are also presented in Tables 1.2.5a, 5b, and 5c. The turnover frequencies determined are higher than previously reported [21] which were in the range 10-6 to 10-2

(molecules/site s) for photocatalytic reactions over TiO2 or ZnO catalysts, although their data were based on apparent reaction rates. For the determination of the turn over frequencies only the anatase phase of the catalysts was included in the calculations, since it was found that only the anatase phase is mainly responsible for the photocatalytic behavior

[22,23]. It could be inferred from Tables 1.2.3 and 5a that AA and AA325 have sites more active in comparison with Degussa P25. In contrast, the value of the apparent reaction rate constants for AA and AA325 are lower than that of Degussa P25 [4,5]. This is because their BET area (which is related to the active sites density) is about 7.5 times lower than that of P25. Hence, the relatively low TOF of Degussa P25 in comparison with that of

Aldrich anatase powders can be attributed to the more efficient generation of radicals from the UV light of the latter catalysts. However, higher BET surface area, synergism of rutile and in its structure [4], and also higher extinction coefficient of Degussa P25 under identical incident light conditions serve as a counterbalance to the lower activity of its catalytic sites. Therefore, the difference in the apparent activity of the catalysts employed in the present study is not as drastic as in the activity of the sites. One can also observe that the turnover frequencies increase with the increase of the lamp power and decrease of the

70 catalyst concentration. This occurs because more photons are available per average photocatalytic particle in this system.

The parametric dependence of the effectiveness factor h for all the photocatalytic systems and reactors employed in the presence study is shown in Figure 1.2.4. The variable parameter used can be obtained from equation (8a) for a particular case of zero scattering

1 h h

0.1 0 1 2 3 4 5 (Gm/Gav)×(bDR) Figure 1.2.4. Dependence of photocatalytic effectiveness factor on design criterion: rhombs: Degussa P25, squares: Aldrich anatase, asterisks: Aldrich anatase 325

(a=0 in equations 8a and 8b). Then equation 7 becomes easily integrable and one can obtain the following expression:

71 G 1- Exp[-d] G æ d d2 ö h = = ç1- + +...÷ (10) ç ÷ G m d G m è 2 6 ø

The right-hand side of this equation represents a Taylor series expansion of h about the point d=0. Thus, as a first approximation h becomes proportional to the product of average incident radiation (G) with the parameter d over the maximal incident radiation in the

reactor (Gm). It should be noted that the criterion of Figure 1.2.4 encompasses the two

following important factors. The ratio of the average incident radiant flux to the maximum

(along the length) incident radiant flux expresses how uniform the axial radiation profile is.

The parameter d (calculated by equation 8b) incorporates the factors contributing to the

radial non-uniformity of light distribution (extinction coefficient, reaction zone thickness).

Thus, the criterion utilized in Figure 1.2.4 fully represents the radiation field within

photoreactors. One can observe from Figure 1.2.4 that the photocatalytic effectiveness

factor behaves the same way as the traditional effectiveness factor. Starting at the value of

unity (which can be readily seen by extrapolation to d=0) it slowly decreases, then drops

to almost zero value. It can be also observed that the trend holds for all three catalysts in a

fairly wide range of values of the criterion, corresponding to different reaction zone lengths

and thicknesses.

The above discussion demonstrates that utilizing intrinsic kinetic constants as well

as turnover frequencies result in useful information on the photocatalytic properties of

different semiconductors. The comparison of intrinsic reaction rate constants for the

catalysts employed in the present study shows that the activity of the two kinds of Aldrich

anatase constitutes about 60-70 % of that of Degussa P25 in a wide range of reactor sizes

and lamps utilized. On the contrary, the average activity of individual sites (TOF) of

Degussa P25 is considerably smaller (by a factor of 4-5), which suggests that the majority

72 of its catalytic sites are not utilized. Quantum efficiencies, on the other hand, may allow

one to compare different reactor geometries and configurations when only one catalyst is

employed in the study under identical conditions. Furthermore, the introduction of the

photocatalytic effectiveness factor can provide qualitative information for reactor design.

CONCLUSIONS

A methodology for determining intrinsic activity in aqueous heterogeneous

photocatalytic reactors was introduced. The most common commercial titania powders

(Degussa P25, Aldrich Anatase, Aldrich anatase 325) with high BET surface area were

evaluated for the UV-induced heterogeneous degradation of phenol to test the proposed methodology. It was demonstrated that kinetic data obtained from commonly used photocatalytic reactors (such as liquid phase reactor) do not correspond to the intrinsic activity of catalysts due to strong non-uniformity of the internal radiation field. In contrast to the vast majority of the previous studies, which report scattered data for different reactors, the present work offers an objective means to characterize photocatalysts. The data obtained from a variable reaction zone were found to approach intrinsic in a differential volume (annular reaction zone of 0.15 cm thickness and 6 cm length). A correction factor (h) representing the ratio of the radical generation rate in the laboratory

to that in the reference-ideal photocatalytic reactors was introduced. This allowed to obtain the intrinsic activity (kint) of the photocatalyst from apparent kinetic measurements

(kapp). The proposed methodology was tested for an annular reactor for the above reaction

(first order) and showed good agreement with the data. The intrinsic reaction rate constants

kint were found to be independent of the reactor size for cylindrical geometry for three light

sources employed (100, 200, and 450 W) and variable catalyst concentration. The

73 calculated turnover frequencies (TOF) allowed for a realistic comparison of the intrinsic

activity per unit surface area of the individual sites. The sites of Aldrich anatase powders

were found to be more active (twofold) than those of Degussa P25.

ACKNOWLEDGEMENT

The authors wish to acknowledge Technical Association of Pulp and Paper Industry

(TAPPI) for supporting this work through grant No. PE-380-97.

REFERENCES

1. Calvert, J. G. and Pitts, J. N., Photochemistry. John Wiley & Sons, Inc., 1967.

2. Riegel, G. and Bolton, J. R., J. Phys. Chem., 1995, 99, pp. 4215-4224.

3. Davydov, L., Smirniotis, P.G., Pratsinis, S.E., Ind. Eng. Chem. Res., 1998 (submitted)

4. Fotou, G. P., and Pratsinis, S. E., Chem. Eng. Comm., 1996, 151, 251-269.

5. Okamoto, K., Yamammoto, Y., Tanaka, H. and Itaya, A., Bull. Chem. Soc. Jpn., 1985, 58, 2023-

2028.

6. Karakitsou, K. and Verykios, X. E., J. Catalysis, 1995, 152, p. 360-367.

7. Serpone, N., J. Photochemistry Photobiology A: Chem., 1997, 104, pp. 1-12.

8. Alfano, O.M., Cabrera, M.I., Cassano, A.E, J. Catal., 1997, 172, 370-379.

9. Hoffmann, M.R., Martin, S.T., Choi, W., and Bahnemann, D.W., Chem. Rev., 1995, 95, 69-96

10. Hacker, D. S., and Butt, J. B., Chem. Eng. Sci. 1975, 30, 1149-1158.

11. Schwartz, P.F., Turro N.J., and Bossmann, S.H., J. Phys. Chem., 1997, 101, 7127-

7134

12. Claria, M. A., Irazoqui, H. A., and Cassano, A. E., A.I.Ch.E. J. 1988, 34, 366-382.

13. Dibble, L. A. and Raupp, G. B., Catal. Let., 1990, 4, 345-354.

74 14. Romero R.L., Alfano O.M., and Cassano A.E., Ind. Eng. Chem. Res., 1997, 36, 3094-

3109

15. Siegel, R. and Howell, J. R., Thermal Radiation Heat Transfer, Hemisphere

Publishing, 1992

16. Cornet, J.F., Dussap, C.G., and Dubertret, G., Biotech. Bioeng., 1992, 40, 817-825

17. Pruden, A. L., and Ollis, D. F., J. Catal., 1983, 82, 404-417.

18. De Bernardez, E. R., Claria, M. A., and Cassano, A. E. in "Chemical Reaction and

Reactor Engineering", Marcel Dekker, New York, 1987, 26, p. 839-921.

19. Braun, A., Maurette, M. T., and Oliveros, E., Technologie Photochemique, Presses

Polytechniques Romandes: Lausanne, Switzerland, 1986

20. Tanabe, K., Misono, M., Ono, Y., Hattori, H., Studies Surf. Sci. Cat., 1989, 51, 1-25

21. Childs, L. P., Ollis, D. F., J. Catal., 1980, 67, 35-48.

22. Augugliaro, V., Palmisano, L., Sclafani, A., Minero, A. and Pelizzetti, E., Toxicol. Environ.

Chem., 1988, 16, 89-109.

23. Okamoto, K., Yamammoto, Y., Tanaka, H., Tanaka, M. and Itaya, A., Bull. Chem. Soc. Jpn.,

1985, 58, 2015-2022.

75

Chapter 1.3. Optimal Radiation Field in Continuous Heterogenous Photoreactors

INTRODUCTION

Photocatalysis has recently emerged as a novel technique of pollution abatement

[1]. However, some intricacies of photocatalytic processes are yet to be explored.

Sczechowski et al [2] discovered the phenomenon of periodic illumination in

photocatalysis. This phenomenon allows increasing the quantum efficiency of typical

photocatalytic reactors by a factor of about five by alternating irradiation for

microseconds with dark recovery time in the order of hundreds of milliseconds. The

mechanism of such enhancement is still unclear. The above authors attribute it to better

utilization of the hydroxyl radicals generated during the irradiation period. In other

words, when the light has been turned off, the hydroxyls can still oxidize the organic molecule for some time. This is unrealistic when the characteristic times of radical processes are taken into account. For example, the time of charge-carrier recombination on the surface of titania is in the order of 10-100 ns [1], which is orders of magnitude lower than the optimal dark recovery time found by Sczechowski et al [2]. A different approach was used by Upadhya and Ollis [3], who theoretically studied the photooxidation of formate ion on titania in a periodically illuminated system. They attributed the photoefficiency enhancement by periodic illumination to the replenishment of the reactants on the surface of the solid catalyst during the dark recovery time.

Buechler et al [4] studied the effect of periodic illumination on gas phase photodegradation of trichloroethylene over TiO2 and found the doubling of the quantum yield. It should be noted, however, that the above researchers used the compounds

77 containing chlorine (such as trichloroethylene) as their probe molecules. As a result, this

may have led them to multiple secondary reactions involving chlorine radical, which is

more stable than hydroxyl and also an oxidizer. Thus, an experiment involving primary

oxidation only would be desirable for the true estimation of the periodic illumination

effect.

The kinetic models governing photocatalytic processes have been extensively

studied [1,5]. They represent systems of non-linear first-order differential equations, which hints at the inherent possibility of optimization. From the theoretical point of view, one may be able to discern whether the effect of the illumination profile is related with the intrinsic behavior of these equations or it is related to either external processes (such as adsorption of reactants) or secondary reactions (such as the effect of chlorine) taking place. Also, if the effect does exist, an optimum spatial distribution of radiation in photoreactors would be very helpful in the reactor design. The present study attempts to answer these questions by analyzing the system of equations governing primary photocatalytic processes. The approximate solutions of this system will be developed, and the reactor output function will be optimized with respect to the radiation distribution function.

THEORY

The following elementary steps of photocatalytic transformation will be considered [1,5]. Although this approach can be applied to any photocatalytic system, we will exemplify TiO2-mediated photocatalysis since it this process is the most studied.

78 · A photon of radiation with energy higher than the bandgap of the semiconductor

(l<385 nm for titania) impinges on a particle producing an electron-hole pair:

hn + - TiO2 ® h + e R1=k1G (1.1)

The rate of the above reaction is proportional to the photon absorption rate (assuming that

the excitation is 100% efficient) as the concentration of the semiconductor in the slurry is

constant throughout the reaction vessel under adequate mixing.

· Some of the resulting electrons and holes recombine releasing heat

+ - + - h + e ® TiO2 + heat R2=k2[h ][e ] (1.2)

The rate of non-radiative recombination is dependent on both the number of holes and

number of electrons in the bulk of the semiconductor particle.

· Valence-band holes form primary radicals

h+ + OH- ® ·OH

+ · + + h + H2O ® OH + H R3=k3[h ] (1.3)

The concentration of ions and water in the aqueous solution is abundant. Many gas phase

photocatalytic studies also provide water as an incoming reactant. For this reason we

assume the rate of surface reaction of holes be proportional to the hole concentration on

the surface of the semiconductor.

· Conduction-band electrons are scavenged by oxygen forming multiple peroxide

species by the reactions

- · - e + O2 ® O2

+ · - · H + O2 ® HO2

+ · - - - 2 H + O2 +e ® H2O2 R4=k4[e ] (1.4)

79 It should be noted that the limiting step of the above reaction scheme is the surface

reaction between the electron and surface oxygen assuming that all peroxide species are

equally active. It was shown [1,6] that the limiting stage in radical formation is the

electron transfer between the solid surface of titania and adsorbed molecular oxygen.

Furthermore, the rate of this consecutive reaction (1.4) must be equal to the electron

transfer rate (the first reaction).

· · Ñ · Some of the reactive oxygen species OH, HO2, H2O2 (denoted by O ) recombine

with the surface of titania:

Ñ Ñ TiO2 + O ® TiO2 + ½ O 2 R5=k5[O ] (1.5)

The recombination of surface radicals can be represented by electron transfer between the

surface radical and semiconductor surface (similarly to recombination at the wall in

radical chemistry). The two groups of radicals (produced by reaction 1.3 and reaction 1.4)

can be lumped into one group when considering their recombination with the surface of

titania. This is because their recombination takes place via electron transfer, and the rate

of recombination must be equal for both groups in order to maintain the balance of

charges in the particle.

· The remaining radicals engage into chemical reaction with the organic (or inorganic)

reductant (A) present in the system:

Ñ Ñ A + O ® mH2O + nCO2 R6=k6[O ][A] (1.6)

The rate of this reaction is proportional to both the concentration of surface radicals and

the concentration of the reactant in the solution.

For further analysis we will denote the concentration of the reactive oxygen

Ñ Ñ species produced by reaction 1.3 as [O1 ], those produced by reaction 1.4 as [O2 ], and

80 the concentration of the organic reactant undergoing degradation as C. We will also

denote the temporal distribution of radiation in the reactor as f(t), considering that the

number of available photons per reaction run is bounded. This is due to the finite power

and length of the lamp and the uniformity of the light absorption in the reactor. The

representation of the above distribution function is shown in Figure 1.3.1, which depicts

the general heterogeneous continuous photoreactor configuration. This configuration

corresponds to plug-flow regime, and the realistic examples for an aqueous system could be a thin-film reactor, thin-layer annular slurry reactor, falling film reactor, and spherical

Figure 1.3.1. Schematic of the class of photocatalytic reactors analyzed in the present study and variables used

81 reactor. For gas phase photocatalytic reactions one can envision moving bed, fluidized

bed as well as thin film reactors as those for which the axial distribution of light is of cru-

cial importance. As can be seen from Figure 1.3.1, any lamp of length comparable with

the reactor length will create non-uniformity of the radiation profile. As mentioned

above, the total number of photons emanating from the lamp per unit time is a bound

value. Thus, instead of making radiation a function of the axial distance one can envision

it as a function of time t that the flow front has spent in the reactor. Furthermore, the

overall residence time in the reactor T will be dependent on the flowrate (T=V/Q). Now

we can construct the following system of equations that governs our photocatalytic

transformation based on the reactions shown above.

+ d[h ] + - + = k1f(t) - k 2[h ][e ]- k3[h ] (2.1) dt

- d[e ] + - - = k1f (t) - k2[h ][e ] - k 4[e ] (2.2) dt

Ñ d[O1 ] + Ñ Ñ = k3[h ] - k5[O ] - k6[O ]C (2.3) dt 1 1

Ñ d[O2 ] - Ñ Ñ = k 4[e ] - k5[O ] - k6[O ]C (2.4) dt 2 2

dC Ñ Ñ = -k6 ([O ] + [O ])C (2.5) dt 1 2

These equations are subject to the following initial conditions: [h+](0)=0, [e-](0)=0,

Ñ Ñ [O1 ](0)=0, [O2 ](0)=0 and C(0)=C0 and t is time variable ranging from 0 to T. The light distribution can be represented as an external force term f*(q) with q being dimensionless time q=t/T, since it is independent of the concentration of any species present in the

82 reactor. In this case the normalizing factor for the radiation distribution (equal to the axial average of the light intensity in the reactor) will be incorporated in k1. It should be also noted that since the total number of photons available in the system is limited and function f*(q) is positively defined and normalized (where k1 is the normalizing factor), i.e.

1 ò f *(q)dq =1 (2.6) 0

The above system of differential equations of first order is nonlinear and is not solvable analytically. Certain simplifications based on the physics of the process can be made to this system of equations in order to be able to solve it analytically. First, a steady state of the light-induced intermediates (electrons and holes) can be assumed with

(d[h+]/dt»d[e-]/dt»0), which is frequently done in this type of studies [1,5]. Second,

+ - analyzing equations 2.1 and 2.2 yields k3[h ]»k4[e ] and we can lump the balance of charges into one equation. This leads to the conclusion that light generates two equally valued opposite charges on the photocatalytic particle [1], and their concentrations in the reactor upon excitation should also be equal. This cannot, however, guarantee the equality of the steady state concentrations of electrons and holes [7]. Furthermore, since the rates of generation of these species are equal, the rates of their disappearance should also be equal in order to preserve the balance of charges in a photocatalytic particle.

Hence, the concentrations of reactive oxygen species produced by route 1.3 and 1.4 can

Ñ Ñ Ñ be lumped into one (2[O ]» [O1 ]+[O2 ]). Therefore, our system of equations will acquire the form:

83 k 2k3 + 2 + 0 = k1f(q) - [h ] - k3[h ] (2.7) k4

d[OÑ ] + Ñ Ñ = k3[h ] - k5[O ] - k 6[O ]C (2.8) dt

d ln C Ñ = -2k6[O ] (2.9) dt

The multiplier 2 in equation 2.9 results from the addition of the two rates of degradation

of the reactant (equations 2.3 and 2.4). Solving (2.7) for [h+] yields:

æ ö + k4 ç k1k2 * ÷ [h ] = ç 1 + 4 f (q) -1÷ (2.10) 2k2 ç ÷ è k3k4 ø

Introducing this solution into eq.(2.8) one gets

Ñ æ ö d[O ] k3k4 ç k1k2 * ÷ Ñ Ñ = ç 1 + 4 f (q) -1÷ - k5[O ] - k6[O ]C (2.11) dt 2k2 ç ÷ è k3k4 ø

Now substituting here [OÑ] from eq.(2.9) and integrating once over time yields

q dlnC C k k4k 6 æ k k ö + k ln + k (C - C ) = - 3 ç 1+ 4 1 2 f *(x) - 1÷dx (2.12) t 5 6 0 ò ç ÷ d C0 k 2 0è k3k4 ø or in dimensionless form

C d ln q 1 C k C C k k4T æ k k ö 0 + 5 ln + C ( -1) = - 3 ç 1 + 4 1 2 f *(x) - 1÷dx (2.13) q 0 ò ç ÷ k6T d k6 C0 C0 k2 0è k3k4 ø

RESULTS AND DISCUSSION

We first would like to ascertain whether the effect of periodic illumination takes place in a system with no mass-transfer limitations as the one described by the above

84

Figure 1.3.2. Comparison of numerical solution of equation 2.13 at k1=0.01 Einstein/(L 1 - Cos(npq) min) and T=0.5 min for f *(q) = (Case a is for n=0.5, Case b – n=1, 1 - Sin(np) / np -4 -3 Case c – n=10, Case d – n=100): A - C0=10 M, B - C0=10 M

85 system of equations. The behavior of the numerical solution of equation 2.13 for C/C0

(the dimensionless reactant concentration) is presented in Figures 1.3.2a and 2b for

1 - Cos(npq) f *(q) = (periodic function simulating alternate light) for several choices 1 - Sin(np)/(np)

of the parameter n (0.5, 1, 10, 100). In this case the number n corresponds to the number

of flashes of light that the any particle of the fluid sees when passed through the reactor.

One can observe that the output concentration does change with the radiation distribution,

achieving the minimal value at n=10. The profiles for n=10 and n=100 overlap in Figure

1.3.2a and merge at the end of the reactor in Figure 1.3.2b, thus producing the same

reactor output. Comparing Figures 1.3.2a and 2b also shows that the effect of the

4 radiation distribution is more pronounced for “slower” reactions (k6=10 ). This is because

the lower value of the bimolecular reaction rate constant increases the importance of the

non-linear terms of equation 2.13. If mass transfer were the limiting factor of the process,

one should observe the opposite picture, namely, more effect of periodic illumination for

faster reactions. This is because the dark time will allow for the re-adsorption of the

reactant on the catalyst surface.

It is impossible to get an analytical solution of equation 2.13. However, to obtain

an approximate analytical solution and optimize the output concentration with respect to

the axial radiation distribution in the reactor, one can consider two particular cases.

Case I. Since the photocatalytic process (or any waste treatment process) aims at

achieving small relative output concentration v=Cout/C0, it follows that Cout/C0<<1.

Therefore, C/C0 of the third term in the right hand side of equation 2.13 can be neglected.

This approach (as will be seen later) is valid after the flow has spent some time in the

86 reactor, sufficient to significantly reduce the concentration of the reactant. After the

above simplification it becomes possible to solve equation (2.13) and simplify the

solution by integration by parts:

æ éqæ k k ö ùö ç ç 1 2 ÷ x - - q · ÷ êò ç 1 + 4 f (x) -1÷d Exp[ k5T ] ú ç k k k T ê k k ú÷ C ç k6C0 3 4 6 0è 3 4 ø ÷ = Exp (1 - Exp[-k 5Tq]) - ê ú C ç k k k q æ ö ÷ 0 5 2 5 ê ç k1k 2 * ÷ ú ç ê· Exp[k 5Tx] 1+ 4 f (x) - 1 dx ú÷ ç ò ç k k ÷ ÷ è ë 0 è 3 4 ø ûø (3.1)

Then the output relative concentration becomes:

æ 1 ö ç k 6C0 k3k4k6T é k1k2 * ù ÷ v = Expç (1- Exp[-k5T]) - ò ê 1+ 4 f (t) -1ú(1- Exp[k5T(q -1)])dq÷ ç k k k ê k k ú ÷ è 5 2 5 0 ë 3 4 û ø (3.2)

Once we have obtained a closed-form solution for the relative output concentration of the photoreactor, we can optimize its the radiation distribution to achieve minimal v. Since v is a monotonous function of f*(q), the minimum of the reactor output will correspond to the maximum of the integral

1æ k k ö ç 1 2 * ÷ - x - x ò ç 1 + 4 f (x) -1÷(1 Exp[k5T( 1)])d (3.3) 0è k3k 4 ø

with the necessary condition (2.6). The integral 3.3 will then be chosen as the

performance index [8] of our reactor. This variation problem could be solved by the

Lagrange method [9]. It consists of forming an “energy integral” (general form):

q2 I = ò g[q, y,f *(q)]dq (3.4) q1

87 q2 subject to the constraint ò f *(q) = const . Then, the function minimizing thus expressed q1

“energy” can be found by solving Euler-Lagrange equation:

¶ (g(q, y, f *(q))- ef *(q))= 0 (3.5) ¶f *

where e is a parameter.

To minimize our function C/C0(q) of equation (3.5) let us introduce a functional formed from equations 3.3 and 3.5:

* k1k 2 * k1k 2 * F(f (q)) = (1 - Exp[k5T(q - 1)]) 1+ 4 f (q) - 2a f (q) (3.6) k3k 4 k3k 4

where a is a constant. Then we are looking to its maximum in variation of f*(q) defined

as

æ ö ç ÷ * k1k 2 ç1 - Exp[k5T(q - 1)] ÷ * dF(f (q)) = 2 ç - a÷df (q) (3.7) k3k 4 ç k1k 2 * ÷ ç 1 + 4 f (q) ÷ è k3k 4 ø

The above relation can be equated to zero as the necessary condition for an extremum

and then solved for f*(q). Hence, the function f*(q) maximizing the functional F(f*(q)) is

k k æ (1 - Exp[k T(q -1)])2 ö f *(q) = 3 4 ç 5 - 1÷ (3.8) ç 2 ÷ 4k1k2 è a ø

The constant a can be obtained from the following equation using the condition 2.6:

2k T + 1 - (2 - Exp[-k T])2 a2 = 5 5 (3.9) k1k 2 2k5T(1+ 4 ) k3k 4

Thus, the optimal distribution of radiation acquires the form:

88 2 k k4 æ 2k T(1 - Exp[k T(q -1)]) k k ö f *(q) = 3 ç 5 5 (1+ 4 1 2 ) - 1÷ (3.10) 4k k ç 2 k k ÷ 1 2 è 2k5T +1 - (2 - Exp[-k5T]) 3 4 ø

One can isolate several groups of constants in the above expression. The first one, k3k4/k2, is a measure of the rate of radical formation relative to the recombination of electron-hole pairs. This quantity can be determined experimentally [10] and is in the

-4 order of 10 M/min. The second constant, k5, is not easy to be determined experimentally, although the ratio of k6/k5 can be determined at very low concentrations of the reactant (usually in the order of 103-104 M-1). We will thus fix the above ratio of the bimolecular reaction rate constant k6 (which is known for most solutes) to the radical recombination constant, and then vary k6. The typical values of these constants (k3k4/k2

-4 4 -1 =10 M/min and k6/k5=10 M ) will be used in further calculations.

The parametric dependence of the optimal radiation profile is visualized in

Figures 1.3.3a and 3b. One can observe that the behavior of the optimal distribution function for “slow” and “fast” reactions, respectively, is different. The reactions having lower bimolecular rate constants are better suited for a monotonically decreasing radiation profile (Figure 1.3.3a). On the contrary, the “fast” reactions favor a uniform radiation profile (f*(q) slightly exceeding 1) with a sharp decrease at the values of x approaching unity. The value of the normalizing constant in these calculations (k1=0.01

Einstein/(L min)) was chosen to correspond to a total radiant flux emitted in the near-UV range by a 450 W mercury lamp (Hanovia) in a standard 0.65 L photoreactor (Ace

Glass). It should also be noted that the optimal distribution profile lowers with the increase of the residence time in the reactor. As expected, the profile becomes more

89

Figure 1.3.3. Case I - comparison of the optimal radiation profiles in the reactor at -3 4 -1 -1 8 -1 -1 C0=10 M, k1=0.01 Einstein/(L min): A - k6=10 M min , B - k6=10 M min

90 uniform, thus flattening the reaction rate. Overall, both types of profiles (for “slow”

reaction and “fast” reaction) prescribe more radiation to the reaction zone with higher

concentration of the reactant, which in our case is the beginning of the reactor.

Introducing now equation 3.10 in the expression for v (equation 3.2) the latter

acquires the form

æ ö ç k k k4 k k4k6T ÷ 6 (C - 3 )(1 - Exp[-k T]) - 3 · ç k 0 k k 5 k k ÷ ç 5 2 5 2 5 ÷ vopt = Exp (3.11) ç æ k k ö 2k T + 1 - (2 - Exp[-k T]) 2 ÷ ç· ç + 1 2 ÷ 5 5 ÷ ç ç1 4 ÷ ÷ è è k3k4 ø 2k5T ø

As a benchmark of performance we will use the uniformly illuminated photoreactor (f*=1). This will allow us to assess the efficiency enhancement due to the use of optimal radiation distribution. The output concentration of the reactant in the uniformly illuminated photoreactor can be obtained:

æ k k k T æ öæ - - öö ç k6 3 4 6 ç k1k 2 ÷ 1 Exp[ k5T] ÷ v1 = Exp C0 (1- Exp[-k5T]) - 1+ 4 -1 ç1- ÷ ç k k k ç k k ÷ç k T ÷÷ è 5 2 5 è 3 4 øè 5 øø (3.12)

A typical plot of the uniformly illuminated photoreactor output (v1) as a function

of experimental parameters, such as residence time T and k1 is presented in Figure 1.3.4.

These parameters correspond to reactor length (or fluid velocity through the reactor) and

the power of the lamp. The ranges chosen were from instantaneous pass through the

reactor to typical 1 min per pass and from negligible radiation intensity to k1=0.01

Einstein/(L min) corresponding to a typical 450W lamp. It can be readily seen from the

Figure 1.3.that little or no conversion is observed at shorter residence times (up to 0.2

min). This is because the first term of equation 3.12 dominates at these conditions, and

91 the expression under the exponential becomes a small positive number. Furthermore,

when the residence time is increased, the conversion in the uniformly illuminated

photoreactor increases approaching 99%. This is the region where the approximation of

Figure 1.3.4. Output of a uniformly illuminated photoreactor calculated by the model of -4 4 -1 -1 Case I as a I as a function of process parameters at C0=10 M, k6=10 M min

Case I is the most accurate. The increase of the radiant intensity leads to rapid decrease of the output concentration.

Case II. Another case to consider would be that of low conversion (high v). This takes place in conventional photoreactors with circulation of the reaction medium, where conversions per pass are very small. Thus, we will use the same methodology for Case II, wherein C/Co is very close to unity. In the vicinity of unity (low conversion) lnv can be approximated as follows:

2 æ ö æ C ö C ç C ÷ ç ÷ ln = ç -1÷ - 0.5ç - 1÷ + ... C0 è C0 ø è C0 ø

92 We will take the first term of the Taylor series and substitute it into equation 2.13 yielding:

C d q 1 C æ k öæ C ö k k T æ k k ö 0 + ç 5 + C ÷ç -1÷ = - 3 4 ç 1 + 4 1 2 f *(x) -1÷dx (3.13) q ç 0 ÷ç ÷ ò ç ÷ k6T d è k 6 øè C0 ø k 2 0è k3k 4 ø

This equation can be solved as follows:

æqæ k k ö k ö ç ç 1 + 4 1 2 f (x) - 1÷dx - Exp[-k T( 5 + C )q] ·÷ çò ç ÷ 6 0 ÷ C k k T è k3k4 ø k6 = 1 - 3 4 ç0 ÷ (3.14) q C0 k5 ç k æ k k ö ÷ k2 ( + C0) ç· Exp[k T( 5 + C )x]ç 1 + 4 1 2 f *(x) -1÷dx ÷ k6 ç ò 6 k 0 ç k k ÷ ÷ è 0 6 è 3 4 ø ø

Using the methodology developed in Case I, one can also optimize the output of the reactor expressed by equation 3.14. Substituting x=1 into the above equation will produce the sought performance index v. Then the “energy” functional will take the form:

* k5 k1k 2 * * F(f (q)) = (1 - Exp[k6T( + C0)(q - 1)]) 1+ 4 f (q) - af (q) (3.15) k6 k3k 4

æ ö ç k1k2 k5 ÷ 2 (1 - exp[k6T( + C0)(q -1)]) * ç k3k4 k6 ÷ * dF(f (x)) = ç - a÷df (q) (3.16) ç * k1k 2 ÷ ç 1 + f (q) ÷ è k3k 4 ø

From which the optimal f*(q) can be determined:

2 4k k æ k ö k k f *(x) = 1 2 ç1 - exp[k T( 5 + C )(q - 1)]÷ - 3 4 (3.17) 2 ç 6 0 ÷ a k3k 4 è k6 ø 4k1k 2 where the parameter a2 can be determined using the constraint (2.6) by the following equation:

93 2 æ 4k k ö æ k k ö ç 1 2 ÷ ç 5 5 2 ÷ ç ÷ ç2k6T( + C0) + 1 - (2 - exp[-k6T( + C0)]) ÷ è k 3k 4 ø è k6 k6 ø a2 = (3.18) k5 æ 4k1k2 ö 2k 6T( + C0)ç1+ ÷ k6 è k3k4 ø

Then, the final expression for the optimal radiation profile becomes:

æ k5 k5 2 k1k2 ö ç 2k6T( + C0)(1 - Exp[k6T( + C0)(q - 1)]) (1 + 4 ) ÷ * k3k4 ç k6 k6 k3k4 ÷ f (q) = ç - 1÷ 4k1k2 k5 k5 2 ç 2k6T( + C0) + 1 - (2 - Exp[-k 6T( + C0)]) ÷ è k6 k6 ø

(3.19)

The parametric dependence of the optimal radiation profile is visualized in Figures 1.3.5a and 5b. This function described by equation 3.19 behaves similarly to that of Case I, and the similarities can be seen in the expressions themselves. In Case I the main factor of the

* equation for f (q) was k5T, and for Case II it is (k5+k6C0)T. Therefore, the main input in

the difference of these two functions can be attributed to the influence of the initial

concentration of the reactant in the system. Considering that the typical value for the

-4 -3 k5/k6 ratio is 10 M and that of C0 is 10 M, the latter variable can have a significant effect. Comparing the behavior of the optimal distribution function for “slow” and “fast” reactions shows that the profiles for the “fast” reactions are almost identical for both cases (Figure 1.3.3b and 5b). They slightly exceed unity for the major part of the reaction zone, while abruptly falling at the end of the reactor. On the contrary, the profiles for

“slow” reactions are more linear-like for the second case (Figures 1.3.5a and 3a, respectively). As expected, the profile becomes more uniform, thus flattening the reaction

94

Figure 1.3.5. Case II - comparison of the optimal radiation profiles in the reactor at -3 4 -1 -1 8 -1 -1 C0=10 M, k1=0.01 Einstein/(L min): k6=10 M min , k6=10 M min

95 rate. As in case I, both types of profiles (for “slow” reaction and “fast” reaction) prefer more radiation at the beginning of the reactor. Such optimal radiation profile can be approached in practice as a number of lamps of decreasing power in series.

Now, introducing the optimal radiation distribution function (equation 3.19) into the approximate low-conversion case solution (equation 3.14) one can obtain the profile of concentrations of the reactant:

æ ö ç k5 k5 2 ÷ 2k6 ( + C0 )T +1- (2 - Exp[-k6 ( + C0 )T]) ç æ k1k2 ö k6 k 6 ÷ ç ç1+ 4 ÷ -1+ ÷ ç è k3k4 ø k5 ÷ 2k6 ( + C0 )T k3k 4T ç k6 ÷ v = 1- ç ÷ opt k k ( 5 + C ) ç k5 ÷ 2 0 1- Exp[-k6 ( + C0 )T] k6 ç k ÷ ç + 6 ÷ ç k5 ÷ ç k6 ( + C0 )T ÷ è k6 ø

(3.20)

Furthermore, for the purpose of comparison we will calculate the output concentration at f*=1:

æ k5 ö ç 1- Exp[-k6 ( + C0 )T] ÷ k k T æ k k öç k ÷ = - 3 4 ç + 1 2 - ÷ - 6 (3.21) v1 1 ç 1 4 1÷ç1 ÷ k5 è k3k4 ø k5 k2 ( + C0 ) ç k6 ( + C0 )T ÷ k6 è k6 ø

Figure 1.3.6 depicts the concentration dependence of the output of the uniformly illuminated photoreactor (v1) at low conversion approximation as a function of the experimental parameters (residence time and lamp power). It can be seen that for such system for k1>5×10-4 Einstein/L min and T>0.5 min almost complete conversion is

96

Figure 1.3.6. Output of a uniformly illuminated photoreactor calculated by the model of -4 4 -1 -1 Case II as a function of process parameters at C0=10 M, k6=10 M min observed, which contradicts the main assumption of Case II. However, a curve corresponding to the minimal values of the parameters is also notable.

It would also be helpful to compare the approximate analytical solutions for the above extreme cases (equations 3.1 and 3.14) with the numerical solution of equation

2.13 in order to better determine the applicability ranges of each approximation. This comparison for f=1 is made in Figures 1.3.7a, 7b, and 7c. The first Figure 1.3.considers the case of a relatively “slow” reaction and lower initial concentration of the reactant.

The rapid decay of relative concentration expressed by the Case II model is notable. It is also clearly seen that the Case I model predicts fairly well the concentration profiles in the uniformly illuminated photoreactor. This is because the significance of the first term under the exponential of equation 3.1 is much lower at lower C0. On the contrary, in the

97 -4 A: C0=10 M

4 -1 -1 k6=10 M min

-3 B: C0=10 M

2 -1 -1 k6=10 M min

-3 C: C0=10 M

10 -1 -1 k6=10 M min

Figure 1.3.7. Comparison of the numerical solution of the full model with approximate analytical solutions for a uniformly illuminated photoreactor at k1=0.01 Einstein/(L min), T=1 min (a – numerical solution, b – Case I approximation, c – Case II approximation)

98 equation 3.14 the influence of the initial concentration of the reactant is in the

denominator, thus leading to a rapid decay. Figure 1.3.7b depicts the behavior of a yet

slower reaction for a larger initial concentration of the reactant (which is a typical value

in photocatalytic studies). This case yields very good agreement of both models with the

numerical solution. The third comparison (Figure 1.3.7c) was made at a typical value of

-3 10 -1 -1 the initial concentration (10 M) and very rapid reaction (k6=10 M min ). In this

situation the models exhibited the following range of applicability: Case I model

predicted the behavior at high conversions, and Case II model predicted those at low to

intermediate conversions. Thus, for such system the low conversion approximation offers

significant agreement with the full mathematical picture.

We now wish to analyze the enhancement achieved by the use of the optimal

radiation profiles in photocatalytic reactors. This will be done by calculating the ratios of

the outputs of the optimally illuminated photoreactor to that of the uniformly illuminated

one (vopt/v1), shown in Figures 1.3.8 and 9. In both cases the low power of the lamp and also the low residence time in the reactor produce no enhancement. This is consistent

with the experimental data obtained by Sczechowski et al [2], wherein little effect of

periodic illumination took place at low intensities of light. Figure 1.3.8 depicts the

enhancement for a high-conversion case in a typical parametric space for a “slow”

reaction and low initial concentration of the reactant. One can observe that the

enhancement grows monotonically with the increase of the residence time and lamp

power. This behavior should be expected, since the optimum profile in this case (Figure

1.3.3a) was close to linear-monotonic with more radiation at the beginning of the reactor.

99

Figure 1.3.8. The degree of enhancement due to optimal radiation profile in the reactor -4 4 -1 -1 (vopt/v1) for Case I: C0=10 , k6=10 M min

Figure 1.3.9. The degree of enhancement due to optimal radiation profile in the reactor -4 4 -1 -1 (vopt/v1) for Case II: C0=10 M, k6=10 M min

100 As a result, this profile matches better the concentration profile and thus leads to more

conversion. The behavior of the same system at low conversion (Figure 1.3.9) is more

peculiar. It starts from “no-enhancement” and proceeds to attain the maximal

enhancement, then returning to the “no-enhancement” situation. One can also observe that there exists an optimal curve in the parametric space corresponding to the maximal enhancement (or the ratio vopt/v1 equal to zero) of the photocatalytic reaction due to the

use of the optimal illumination strategy. The bimolecular reaction rate constant (k6) can

also vary in a wide range [11]. Figures 1.3.8 and 9 correspond to a “slow” reaction

4 -1 -1 8 (k6=10 M min ), and the enhancement is possible. For a relatively fast reaction (k6>10

-1 -1 M min ) the values of vopt/v1 were equal to unity.

Figure 1.3.10. Optimal parametric curve for the degree of enhancement (vopt/v1) for -4 3 -1 -1 4 -1 -1 8 -1 -1 Case II: C0=10 M (a - k6=10 M min ; b - k6=10 M min ; c - k6=10 M min , d - 10 -1 -1 k6=10 M min )

One can also calculate the optimal relation between the experimental variables

(lamp power and residence time) in the function of equation 3.19 corresponding to the low-conversion case. This can be done by equating the ratio vopt/v1 to zero (since this ratio cannot be negative) and solving for k1. This is given in Figure 1.3.10 for several

101 values of the bimolecular reaction rate constant. One can observe that for slower

reactions the optimal operation curve approaches the boundary of our domain {T, k1}. On

the other hand, for faster reactions the optimal curves converge. Furthermore, no

8 -1 -1 enhancement is observed for k6>10 M min since the decay of concentration occurs so rapidly that there is no difference between the use of optimal radiation profile and the uniform illumination profile.

The parametric behavior of the approximations described above hints the possibility of the effect of periodic illumination on the output of the reactor. Some optimal distributions of light consist of a plateau close to unity with the consequent sharp decrease (Figures 1.3.3b and 5b). Certain combination of parameters gives monotonic decrease from about 1.5 to zero (Figures 1.3.3a and 5a). Since the conversion per pass in photoreactors can be relatively low [12], and such radiation profile is optimal for one pass, one can envision the process occurring in such circulating reactor as a sequence of one-pass reactors. In this sequence each reactor should have the optimal radiation profile.

Then, the enhancement of conversion (based on the numbers of Figures 1.3.8 and 9) will

n be (vopt/v1) -fold if the optimal illuminating strategy is used. In such strategy one should also consider the change in C0 incurred as a result of reaction, although it can be neglected for lower conversion values.

CONCLUSIONS

A non-linear non-homogeneous kinetic model taking into account balances of reactants and active intermediates in photocatalysis has been developed. Two particular cases were isolated that allowed to develop approximate analytical solutions. These

102 solutions depended non-linearly of the axial radiation distribution inside the reactor. The

two particular cases were optimized for the light distribution using optimal control

methods. These optimal radiation profiles prescribed more radiation at the beginning of

the reactor for both cases and for both slow and fast reactions. The utilization of such

optimal light distribution can lead to a significant enhancement of performance. Such

optimal strategy for multi-pass photoreactors can give much greater results in terms of

reactor output.

ACKNOWLEDGEMENT

The authors wish to thank NATO Science for Peace Programme (Grant No. SfP-

974209) for partial support of this work. This research is also based upon the work supported by the U.S. Army Research Office under grant number 40414/CH/YIP. The authors (L.D. and P.G.S.) are grateful to the above award.

NOMENCLATURE

C – concentration of the reactant, M

C0 – initial concentration of the reactant, M

f(t) – axial distribution of radiant energy in photoreactors, Einstein/(L min)

f*(q) – normalized axial distribution of radiant energy in photoreactors g(q, y, f*[q]) – function of equation 3.4

G – radiant flux density, W/cm2

I – energy integral of equation 3.4 k1 – normalizing coefficient for f(x), Einstein/(L min)

103 -1 -1 k2 – constant of electron-hole recombination, M min

-1 k3 – constant of hole-driven radical generation, min

-1 k4 – constant of electron-driven radical generation, min

-1 k5 – constant of radical recombination with the solid, min

-1 -1 k6 – constant of surface reaction of the organic reactant, M min

n – number of light flashes in the reactor

t – time the fluid particle has spent in the reactor, min

T- space time of the reactor, min

v – output relative reactant concentration

v1 - output relative reactant concentration of a uniformly illuminated photoreactor

vopt - output relative reactant concentration of an optimally illuminated photoreactor

y – dependent variable of equation 3.5.

Greek Letters

a – parameter introduced in equations 3.6 and 3.16

e – parameter of equation 3.5

x – dummy variable of equation 2.12

F – functional of equation 3.6 and 3.16

q – dimensionless time

REFERENCES

1 Hoffmann, M.R., Martin, S.T., Choi, W., and Bahnemann, D.W., Chem. Reviews, 95,

69 (1995)

104

2 Sczechowski J.G., Koval C.A., and Noble R.D., J. Photoch. Photobio. A, 74, 273

(1993)

3 Upadhya S, Ollis D.F., J. Phys. Chem. B, 101, 2625 (1997)

4 Buechler, K.J., Noble, R.D., Koval, C.A., and Jacoby, W.A., Ind. Eng. Chem. Res., 38,

892 (1999)

5 Turchi, C.S. and Ollis. D.F., J. Cat., 122, 178 (1990)

6 Gerischer, H. and Heller, A., J.Phys.Chem., 95, 5261 (1991)

7 Minero, C., Catal. Today, 54, 205 (1999)

8 Bryson, A.E. and Ho, Y.C., Applied Optimal Control, Hemisphere Publ., 1975

9 Weinstock, R. Calculus of Variations, Dover Publ., Inc. (NY), 1974

10 Davydov, L. and Smirniotis, P.G., J. Catal., 191, 105 (2000)

11 Anbar, M. and Neta, P.,Int. J. Appl. Rad. Iso., 18, 493 (1967)

12 Davydov, L., Smirniotis, P.G., and Pratsinis, S.E., Ind. Eng. Chem. Res., 38, 1376

(1999)

105

Chapter 1.4. Sonophotocatalytic reactor for VOC destruction

INTRODUCTION

Many researchers have undertaken extensive studies of photocatalytic techniques

assisted by titanium dioxide (TiO2) for the removal of organic and inorganic

contaminants from aqueous streams. These techniques have been proven to be effective

for oxidative destruction of the most recalcitrant organic compounds like azo dyes [1],

TNT [2] and paraquat (1,1’–dimethyl-4,4’-bipyridinium dichloride) [3] and for the

reduction of several heavy metals [4]. Among those techniques, heterogeneous

photocatalysis over TiO2 or other semiconductors [5, 6, 7] and ultrasound [8, 9, 10, 11,

12, 13] have recently been attracting attention as advanced oxidation process for treating water. Photoexcitation of semiconductor TiO2, promotes valence band electrons to the

conduction band, thus leaving an electron deficiency or hole in the valence band.

· - Dioxygen provides a sink for electrons forming superoxides O2 thus leading to the

· protonated form hydroperoxide HO2 . Holes can react with water molecules (or

hydroxide anions) to form ·OH radicals, or can also be filled by an adsorbed organic

donor. Furthermore, the introduction of ultrasound as a source of agitation and activation

[14] has been reported in various applications.

The phenomena of cavitation i.e. the nucleation, growth and collapse of bubbles

in a liquid are due to the chemical effects of ultrasound [15, 16, 17]. The collapse of the

bubbles induces localized supercritical conditions: high temperature, high pressure, and

electrical discharge effects. The consequences of these extreme conditions are the

cleavage of dioxygen molecules and water molecules forming H· and ·OH radicals. From

107 · · · · - · the reactions of these entities (O , H , OH) with each other and with H2O and O2 , HO2

radicals and H2O2 are formed. Therefore, the combination of photocatalytic and ultrasonic irradiation can enhance the degradation of organic pollutants in water by the species, notably ·OH radicals, as was illustrated in a paper concerning monochlorophenols [18]. But distinct phenomena can also involve, e.g. direct electron transfer from the organic compounds to TiO2 in photocatalysis, and pyrolysis of the organic compounds in ultrasonic water treatment. Furthermore, one process is basically heterogeneous since it involves a solid and a liquid that is TiO2 powders in aqueous organic solution, whereas in the other process local heterogeneity is created by cavitation with ultrasound. We were therefore, very interested in investigating the effect of the combination of UV-irradiated aqueous titania and ultrasound on degradation of water polluting organic compounds.

Several aromatic compounds, which are classified as priority pollutants by the US

EPA and other related agencies, have been photocatalytically oxidized to CO2 and H2O

[19]. Among these, salicylic acid is also one of the most dangerous organic pollutants, because it leads to a range of undesirable outcomes in humans and animals, including fetal death, growth retardation, and congenital defects. Salicylic acid also exhibits strong fluorescence and hence can be readily measured at low concentrations. Moreover, the

UV-assisted photodegradation of this compound in TiO2 slurries has already been reported [20, 21]. However, no systematic study of its sonolysis and coupled sonolysis/photocatalysis has been conducted. Due to these reasons we decided to use salicylic acid as the probe molecule contaminant. In this study an attempt has been made

108 to show the photocatalytic activity of different titania particles in the degradation of

salicylic acid using combination of UV-light and ultrasound.

EXPERIMENTAL

Catalysts

The objective of the experimental work performed in this study was to evaluate

the photocatalytic degradation of salicylic acid with commercial titanias under

ultrasonication. Therefore, we used different commerial titanias to compare the results on

the destruction of salicylic acid. The titania powders utilized in the present study were

Aldrich anatase (AA), Hombikat UV-100 (HK), Ishihara ST-21 (ST), and Degussa P25

(P25). These catalysts are commercial titanias with different characteristics, which are

presented in Table 1.4. 1.

Table 1.4.1. Properties of the photocatalysts employed in the present study Catalyst BET Anatase Primary Extinction Surface Content(2), Particle Coefficient(3)(3),, Area(1), m2/g % Size, nm L/(cm g) Aldrich 10 99 100-200(4) 15.3 DegussaP25 75 70-80 ~20(2) 23.5 Hombikat 313 99 <10(2) 7.4 Ishihara 64 95 ~20(2) 9.3 1. Measured by Micromeritics Gemini 2320 surface analyzer at 77 K 2. According to the manufacturer’s specification 3. Measured spectrophotometrically by Shimadzu 2501PC, equipped with an integrating sphere Shimadzu ISR 2200. 4. Determined by Transmission Electron Microscopy (Phillips CM-20)

BET surface area and pore size distribution:

109 BET surface area and pore size distribution of the titania powders was measured by nitrogen adsorption at 77 K using an Accelerated Surface Area and Porosity apparatus

(ASAP 2010, Micromeritics). Prior to analysis, 0.5 to 1 g of TiO2 powder was degassed at 473 K and 200 mm Hg for 2 hours. The adsorption isotherms of nitrogen were collected at 77 K using approximately 20 values of relative pressure ranging from 0.05 to

0.99. Pore size distribution of these titania powders were also measured.

X-ray diffaction (XRD)

X-ray diffraction (XRD) analysis was performed on a Siemens D500 diffractometer, using monochromated CuKa radiation and standard recording conditions.

XRD phases present in the samples were identified with the help of ASTM Powder Data

Files. The XRD patterns of titanias have been compared before and after the reaction.

The intensity of peaks of anatase and rutile were calibrated using MgO (Aldrich) as an internal standard.

UV-visible Spectroscopy

Optical properties of various titanias and concentration of salicylic acid were determined by using UV-vis spectrophotometer (Shimadzu UV-2501PC) equipped with an integrating sphere attachment Shimadzu ISR-1200. The powders were compressed into flat wafers and analyzed for diffuse reflectance. In other experiments, the powders were suspended in water at an appropriate pH (pH=6.5 corresponding to all photocatalytic experiments). Then the optical properties (extinctance and absorbance) of

110 the resulting slurry were measured for several concentrations of the powders. These measurements yielded the corresponding specific extinction and absorption coefficients.

Total Organic Carbon Analyzer (TOC)

The total organic carbon (TOC) of initial and irradiated samples was determined with a Shimadzu 5000 analyzer.

Thermal gravimetric studies

Thermogravimetric analyses (TGA) were conducted on a Perkin-Elmer TAS7

TGA apparatus to determine absorbed salicylic acid and its intermediates and to quantify the weight loss catalysts at elevated temperatures. Samples of Hombikat titania after degradation of salicylic acid were tested. The TGA analyses were performed using 10-12 mg of sample and a heating rate of 5°C min-1 up to a maximum temperature of 375°C and isothermal at 375°C for 5 hours and atmospheric pressure for all catalysts. A continuous stream of nitrogen was used to purge off-gases from the TGA electronics and sample region.

Photocatalytic studies

The photocatalytic activity of the selected titania powders was evaluated using salicylic acid in oxygenated aqueous suspensions. Salicylic acid (Fisher) and titania used in the study were of reagent grade. Photocatalytic experiments were performed in a custom-made ultrasonic reactor presented in Figure 1.4.1, which contained 7 pyrex immersion wells hosting the UV lamps. The reactor consisted of a working volume of 1 L

111 and was surrounded by a glass jacket to allow circulation of cooling water to maintain the reaction temperature at 30±2°C. Seven UV-irradiated black light fluorescent lamps (4 W each, manufactured by Wiko) with a peak intensity at 375 nm were used and were placed inside the pyrex immersion wells (cutoff wavelength 320 nm). However, only about 7 W out of 28 W was in the near-UV range. The agitating action of a stirrer and ultrasound assured adequate mixing of the suspension of titania particles inside the reactor. Pure oxygen gas (Wright Brothers, 99.5%) was sparged through the vessel utilizing a stainless steel bubbler at a rate of 500 cm3/min. The ultrasonic reactor was housed inside a UV- safety cabinet.

The experiments were performed with an aqueous solution of salicylic acid having initial concentration of 2 mM (276 mg/l). The catalyst concentration in the salicylic acid experiments was 0.25 g/l. Prior to the addition of slurry into the reactor the titania particles were dispersed in the solution by ultrasonication for 10 minutes in an ultrasonic bath (model Labline LC20H). The ultrasonic field generated from the ultrasonic bath provided satisfactory dispersion of the titania particles. Experiments with titania, UV-light, ultrasound, combination of UV-light + titania, and UV-light + ultrasound + titania were performed separately to determine which factor was having a major influence on the photocatalytic degradation of salicylic acid. These experiments were also used to determine the effect of each factor on the process of degradation.

Similarly to our previous work [22], the ultrasound which we have used in this experiment was produced by the UWR ultrasonic processor at an amplitude of ~ 50% corresponding to the power input of 100-110W/L at the frequency of 20 kHz. The UV- lamps were warmed up for 5 minutes before starting the experiments.

112 Samples (1-2 ml) were withdrawn from

the reactor by a syringe through the wall

of the cabinet. The contents of the

syringe were passed through 0.2 mm

membrane filters contained in a plastic

filter holder (Gelman Science) to remove

suspended titania particles and the

concentration of salicylic acid was

monitored by measuring the solution

absorbance at 296 nm [23] using a

spectrophotometer Shimadzu 2501PC. Figure 1.4.1. Schematic of the sonophoto- catalytic reactor employed in the present Quantitative absorption of light was study: 7 UV lamps and one ultrasonic probe calibrated with known concentrations of salicylic acid.

In order to identify the difference in the composition of the intermediates of sonophotocatalytic degradation a more advanced technique of analysis was needed.

However, the boiling point of salicylic acid is 375°C, which is above the limit imposed by most chromatographic columns. For this reason, exactly the same degradation experiments were repeated for phenol (reagent grade, Fisher). This was done in order to be able to discern phenol as well as the intermediate compounds of its degradation on the

GC (Shimadzu GC-17A) equipped with a capillary column (Supelco). The effluent of the

GC was sent to a mass-spectrometer (Shimadzu QP 5505). In the kinetic experiments the concentrations of phenol were tracked by measuring absorbance of solution at 270 nm

113 (Shimadzu 2501PC), and these same solutions were then introduced into GC/MS by

direct injection.

In order to obtain a better understanding about the radical generation by

ultrasound, irradiated titania, and the combination thereof, several additional reactions

were performed. The experiments with modified Fricke dosimeter [24] and formic acid followed a similar pattern as those with salicylic acid and phenol, with the different methods of analysis. The concentration of the unreacted formic acid was measured using a conductivity meter (VWR Scientific, golden cell, cell constant 10.25 cm-1). It was calibrated by standard solutions of formic acid with the concentrations ranging from

0.001 to 20 mM. A quadratic curve passing through zero and the experimental points was fitted with the accuracy of 0.999; this curve was used as the conductivity-concentration correlation. Modified Fricke dosimeter was used to determine the overall concentration of reactive oxygen species in irradiated aqueous suspensions of titania. It consisted of 5 mM of FeSO4 (Fisher), 10 mM of CuSO4 (Fisher), and 20 mM of H2SO4 (Fisher). During the

reaction of Fe2+ with reactive oxygen species Fe3+ ion is produced, and the latter was

quantified by spectrophotometry (Shimadzu PC 2501) at the wavelength of 304 nm.

Standard quartz cuvettes with the pathlength of 1 cm were used. The absorbance by Fe3+

was correlated with the concentration of Fe2+ reacted and further used for calculations.

RESULTS AND DISCUSSION

For the assessment of photocatalytic properties of the powders employed in the

present study their physical characteristics have been assessed (Table 1.4.1). These

catalysts are quite different in primary particle size (10-200 nm), anatase content (70-

114 100%), BET surface area (10-313 m2/g) as well as the optical properties of their

suspensions. This leads us to expect different reaction rates of the photodegradation of

salicylic acid, as well as different mechanisms and values of enhancement by ultrasound.

First, let us consider the blank experiments (Figure 1.4.2b, the upper two curves).

They were undertaken in the presence of 0.25 g/L of HK. One can observe that no

degradation takes place in the

100 absence of light and a ultrasound. This is expected, 90 with UV as the rate of thermal with UV + US 80 400mg heterogeneous catalysis in 100 titania slurries is negligible b

90 without UV + US [6]. However, the sonication

with US of the slurry in the dark gives 80 with UV 250mg % of salicylic acid with UV + US 100 a certain degree of c conversion. Sonolysis of 90 water was found [8] to with UV with UV + US · 80 100mg produce active radicals (H 0 30 60 90 120 150 180 and ·OH) via reaction 1, Time (minutes) which are capable of Figure 1.4.2. Photocatalytic degradation of salicylic acid over different concentration of TiO2 Hombikat: attacking the organic Initial salicylic acid concentration: 1 mM, Reaction time: 3 hours, Oxygen flow rate: 0.5 L/min, 28 W UV- compound in solution: lamps, 100 W ultrasound (US).

)))) · · H2O ® H + OH (1)

115 The presence of a solid phase enhances this phenomenon [12], as the microbubbles tend

to break up into smaller ones, thus increasing the total number of regions of high

temperature and pressure. This increases the number of hydroxyl radicals produced by

the system and leads to the oxidation of the reactant in the absence of light. Hydrogen

radicals from the above reaction can also interact with the oxygen present in the system

to form numerous peroxide radicals and hydrogen peroxide. It should also be noted that

the production of radicals is not the only route of sonolysis of organic molecules. This

process can also be due to the pyrolysis of vaporized molecules and shear stress [10].

Other researchers utilized substantially different systems to study the sonolysis of organic

molecules in water [25, 26], which limits the comparison of their studies with our results.

The general trend observed by the above research groups, however, was the first order

rate of disappearance of the reactant and high degree of conversion.

Now let us study the effect of the catalyst mass by comparing the behavior of HK

at 0.1, 0.25, and 0.4 g/L concentration (Figures 1.4.2a, 2b, and 2c, respectively). As

expected, the conversion after 3 hours is the greatest for the larger concentration of the

catalyst and the smallest for the smallest concentration of the catalyst. For some systems

(as described in [38]) the reaction rate decreases with the increase in the catalyst concentration after a certain threshold value. However, we observed that this threshold

value of catalyst concentration has not been achieved for the system HK/salicylic acid,

since the rate increases monotonically in the sequence 0.1, 0.25, and 0.4 g/L of titania

with and without the ultrasound. Moreover, comparing the enhancement of the reaction

rate (defined as the absolute difference between the two reaction rates divided by the rate

of photocatalysis alone) imparted by the presence of ultrasound reveals a different trend

116 (Figure 1.4.3). The enhancement is negligible for the low concentration of the catalyst

(see Figure 1.4.2a). This is due to the insufficient number of centers of bubble disruption

in the solution since the presence of solids significantly enhances this process [27]. As a

result, the ultrasound passes through the slurry without imparting energy into it and with

little formation of

active radicals. On the other

hand, relatively low UV=28W, US=100W 60 UV=20W, US=160W concentrations of solids usually

correspond to smaller aggregate 40 sizes [28], which are more

difficult to break in order to 20 expose more surface area to the

Relative Enhancement, % light. The enhancement of the 0 activity is the largest for the 0.0 0.1 0.2 0.3 0.4 Catalyst Concentration, g/L intermediate concentrations of

Figure 1.4.3. Relative enhancement of the rate of the titania (0.25 g/L) as shown in photocatalytic reaction of salicylic acid over TiO2 Hombikat by US: Initial concentration: 1 mM, Figure 1.4.2b. This is a combined Reaction time: 3 hours, Oxygen flow rate: 0.5 L/min. effect of the full attenuation of

light [29], aggregate breakage by ultrasound due to the action of its shear stress, and

generation of radicals by the ultrasound itself upon impingement onto solid particles. The

low value of enhancement by ultrasound for higher concentrated suspensions can be

explained by the fact that the working volume of the slurry becomes low [30]. Indeed,

from the high values of the extinction coefficients (Table 1.4. 1) one cannot expect the

117 light to penetrate far into such an optically dense medium. As a result, there is no enhancement due to aggregate breakage, as the zone of action of light is far away from the zone of action of ultrasound. Furthermore, although the attenuation of light is full, it is not in contact with the ultrasound, and the enhancement is only due to the production of hydroxyl radicals by the ultrasound itself. Therefore, one has to use higher ultrasound

power and less attenuating

100 powders to achieve adequate Aldrich (anatase) enhancement of photocata- 90 lytic activity in highly with UV 80 with UV + US a concentrated suspensions. As 100 Degussa P25 has been shown previously

90 [29] the “working zone” of

% of salicylic acid photocatalysis is significantly 80 b

100 decreased at higher Ishihara concentrations of catalysts in 90 the slurry. If one considers a

80 c suspension of 0.5 g/L of P25 0 30 60 90 120 150 180 (extinction coefficient 23.5 Time (minutes) L/g cm from Table 1.4.1) the Figure 1.4.4. Photocatalytic degradation of salicylic acid over different titanias (0.25 g/L): a – Aldrich, b – roughly predicted 99% Degussa P25, c – Ishihara; Initial concentration: 1 mM, Reaction time: 3 hours, Oxygen flow rate: 0.5 L/min, 28 attenuation of light (by W UV-lamps, 100 W US. Beer’s law) will take place at

118 0.4 cm from the lamp. Therefore, no combined effect of ultrasound and photocatalysis can be expected for such system.

The other three catalysts considered in the present study (ST, P25, and AA) were also tested for enhancement due to ultrasound. As in many photocatalytic studies, P25 exhibited the highest activity for the degradation of salicylic acid (Figure 1.4.4b).

Although ST exhibited activity comparable with that of P25 for the degradation of formic acid [38], this is not the case in the present study (Figure 1.4.4c) as we observed the activity of ST comparable with that of AA. Apparently, ST is more reactant specific than all other catalysts. It has been observed previously that surface complexation of TiO2 takes place in the process of photodegradation of salicylic acid [31], which may have a different effect on different catalysts. BET surface area and optical properties of the catalyst also play an important role in photocatalysic performance. Higher adsorbing titanias (such as HK) usually perform better than low adsorbing ones (such as AA) for a variety of reactions. It has been shown previously [38] that although HK has the highest surface area (313 m2/g), its activity is close to that of P25 (S.A.=75 m2/g) because HK cannot generate enough electron-hole pairs to engage into the reaction with all adsorbed species. This is due to the high light scattering and absorption (incorporated in the extinction coefficient of Table 1.4. 1).

In regards to the increase of the reaction rate by ultrasound, the catalyst with the largest particle size (AA) revealed negligible enhancement (Figure 1.4.4a). Such behavior should be expected, as the degree of aggregation of these particles is much smaller [28]

(smaller number of particles in the aggregate). This is because the rate of aggregation is proportional to the number concentration of particles squared [28]. As a result, the

119 breakage of aggregate should not play a major role in the activity. Furthermore, the total number of particles present in the system is much lower, as the primary particle size is larger. This decreases the efficiency of spreading microbubbles. Thus, the only effect of ultrasound in the case of AA is the production of radicals by the ultrasound itself

(reaction 1), which does not take place to a major extent, as will be seen further in the

study. P25 (Figure 1.4.4b) does

100 a show some enhancement due to

ultrasound as well as the

90

largest conversion among the

catalysts tested. The

80 ST enhancement by ultrasound is 100 b comparable with that of ST % of salicylic acid (Figure 1.4.4c). This should be 90

expected as their physical

UV + TiO 2 characteristics are very close UV + US + TiO 2 80 P25 (Table 1.4.1). The particle size 0 30 60 90 120 150 180 Time (minutes) is intermediate among the

Figure 1.4.5. Photocatalytic degradation of salicylic catalyst tested, and the relative acid over different titanias: Catalyst weight: 0.1 g/L, Initial concentration: 1 mM, Reaction time: 3 hours, enhancement is also Oxygen flow rate: 0.5 L/min, 28 W UV-lamps, 100 W US. intermediate. Further tests on

these catalysts at 0.1 g/L

(Figure 1.4.5a and 5b) support the conclusion made above about low concentrations of catalysts as the additional activity due to the action of ultrasound is very small. It should

120 be noted that all trends reported in the study have been confirmed by two independent

methods of analysis, namely, spectrophotometry and TOC measurements. Therefore, true

enhancement by ultrasound was obtained, even though in the region of low conversion

values. Furthermore, the agreement between the TOC and spectrophotometric

measurements indicate the absence of the intermediates in solution. As will be shown

further in the study by means of TGA, the intermediates are to be found only on the

surface of the catalyst.

Table 1.4.2. Zero order reaction rates (mol/Lmin) of the photodegradation of salicylic acid (maximal error 9 %) Catalyst 0.1 g/L 0.25 g/l UV UV+US UV UV+US Hombikat 0.033 0.033 0.055 0.083 Ishihara 0.053 0.054 0.045 0.066 Degussa 0.064 0.076 0.101 0.119 Aldrich 0.021 0.025 0.078 0.084

Table 1.4.2 summarizes the zero order reaction rates of all systems tested,

corresponding to the surface saturation of the catalyst. The rate of photodegradation of

salicylic acid was previously found to be of shifting order [32, 33], obeying zero-order

kinetics at relatively low conversions and first-order kinetics at higher conversions. Other

researchers observed zero order kinetics [31] or first order kinetics [21]. This is contrary to the results observed in our reactor, wherein the order of the reaction was zero even after 6 hours. It should be noted, however, that we worked in the low-conversion regime, thus not achieving the point of shifting order. The values of Table 1.4.2 were determined with the error below 9%. As mentioned above, P25 exhibited the highest activity for the degradation of salicylic acid and moderate enhancement by ultrasound, while HK showed

121 lower activity and the largest enhancement by ultrasound. None of the catalysts exhibited first-order behavior. This discrepancy cannot be attributed to the presence of mass transfer limitations in our system for two reasons. First, titania powders used are essentially non-porous [34] and open structures. Second, from the reaction engineering viewpoint, mass transfer limitations are known to shift the order of reaction from zero to one half rather than from first to zero.

12.0

UV + TiO2 11.5 UV + US + TiO2 11.0

10.5

10.0

9.5 Weight (mg) 9.0

8.5

8.0 0 100 200 300 400 500 600 700 Time (minutes)

Figure 1.4.6. TGA analysis of spent Hombikat TiO2 at 3 hours of reaction: Initial concentration: 1 mM, Reaction time: 3 hours, Oxygen flow rate: 0.5 L/min, 28 W UV- lamps, 100 W US.

The absence of the shifting order in the degradation of salicylic acid using photocatalysis and ultrasound can be also attributed to two effects associated with the interaction of ultrasound and solid matter. These are microstreaming and increased mass transport [35]. Microstreaming provides in-situ regeneration of catalyst surface as the cavitation near solid surface causes a jet of fluid directed onto the particle [35]. In this

122 manner, the partial blockage of the active sites of the photocatalyst (leading to the reduction of adsorption capacity) can be circumvented. Ultrasound can also enhance the mass transfer on the liquid-solid interface [36]. This supplies more reactant to the surface and effectively enhances the adsorption of reactant bringing it closer to saturation, which may be otherwise impaired due to the larger size of the molecule of salicylic acid. On the contrary, as will be seen further in the study, when formic acid or Fe2+ ions are used, the adsorption is facilitated, and the enhancement of mass transport may not play a role.

Yet another way of enhancement due to ultrasound in photocatalysis is the direct action of the cavitational effects on the organic reactant [10]. Since we observed a much lower degree of dark sonolysis of salicylic acid in comparison with that of the photocatalysis, such direct action on the reactant itself occurs to a much lesser extent than the phenomena described above. However, the partially hydroxylated intermediate compounds (such as dihydroxybenzoic acid and catechol [33]) are more labile and can possibly undergo the direct action of cavitation [35]. This is another reason why the enhancement of the photocatalytic reaction rate due to ultrasound was found to be the largest when the smallest primary particle size was used. Thermogravimetric analysis was performed on the adsorbed organics during the photodegradation of salicylic acid. The surface organic species on spent HK catalyst after 3 hours of reaction with and without ultrasound exhibited discernible difference (Figure 1.4.6). One can observe that the loss of mass constituted 7.8±0.1% for the used catalyst of sonophotodegradation (the upper curve). On the contrary, the used catalyst of photodegradation without ultrasound (the lower curve) lost 9.2±0.1% of mass at about 350°C. Therefore, one can conclude that the use of ultrasound during photocatalysis allows to lower the concentration of the organic

123 species on the surface of the catalyst. It should be noted that a high boiling point of

salicylic acid is beyond the temperature limit of common chromatographic columns,

which precluded us from using more discriminative GC/MS analysis of chloroform

extracts of the surface organic species.

For this reason the above sonophotocatalytic system was also tested for the

presence of intermediate compounds, which can be of great importance for specialized

treatment of recalcitrant wastewater or other streams. Phenol was chosen as a model

compound as its intermediates are very well known [37], and it is possible to detect them

by a GC/MS. Such experiment involving UV-assisted degradation under identical operating conditions revealed the following trends. First, the enhancement due to ultrasound (US/UV power input ratio of ~8) for the UV-irradiated slurry of 0.25 g/L HK

constituted about 100% for phenol and more than 50% for salicylic acid (Figure 1.4.3).

Second, it was possible to detect quinones in the reaction mixture by MS in the tail of the

phenol peak when using photocatalysis alone. However, none of these compounds were

present when the ultrasound accompanied the photodegradation. It is possible that the

intermediate compounds of phenol photodegradation are usually more labile than the

starting reactant, they can be degraded by the ultrasound alone. Indeed, quinones are

known to decay much more rapidly on the surface of the catalyst than phenol does [37],

but they may be stable far away from the liquid-solid interface. Therefore, the utilization

of ultrasound during photocatalysis allows to eliminate toxic intermediates of degradation

in the bulk solution.

Two methods were utilized in the present study aiming at the quantitative

determination of active radicals. These were Fricke dosimetry and HCOOH degradation.

124 The first one was based on the conversion of two-valent into three-valent iron ions. Two- valent iron ions can react with reactive oxygen species [24] of both hydroxyl and peroxyl nature forming two three-valent iron ions per electron-hole pair scavenged according to the following reaction scheme:

5 M -4 UV+TiO 2 UV+US+TiO 4 2 US+TiO 2

3 generated x 10 +3

2

1 Concentration of Fe 0 0 50 100 150 200 Time (minutes)

2+ Figure 1.4.7. Photooxidation of Fe over Hombikat TiO2: Catalyst weight – 2+ 0.25 g/L, [Fe ]0=5 mM, Initial concentration: 1 mM, Reaction time: 3 hours, Oxygen flow rate: 0.5 L/min, 28 W UV-lamps, 100 W US.

- Fe2+ + ·OH ® Fe3+ + OH (2.1)

2+ · + + Cu + HO2 ® Cu + H + O2 (2.2)

Fe3+ + Cu+ ® Fe2+ + Cu2+ (2.3)

2+ 3+ - 2Fe + H2O2 ® 2Fe + 2OH (2.4)

Therefore, this reaction will proceed even when the hydroxyl radicals are absent. Indeed, in the absence of UV light the transformation still takes place under ultrasound (Figure

1.4.7). The degree of dark reaction is much lower than that of the photocatalytic reaction

125 (upper curves). On the contrary, the degradation of formic acid is not possible without the

presence of hydroxyl radicals [38]:

100

UV+TiO 80 2 UV+US+TiO2

US+TiO2 60 US

40 % of formic acid 20

0 0 30 60 90 120 150 180 Time (minutes)

Figure 1.4.8. Photooxidation of formic acid over Hombikat TiO2: Catalyst weight – 0.25 g/L, [HCOOH] 0=10 mM, Initial concentration: 1 mM, Reaction time: 3 hours, Oxygen flow rate: 0.5 L/min, 28 W UV-lamps, 100 W US.

- · · - HCOO + OH + O2 ® H2O + CO2 + O2 (3)

Since the sonolytic degradation in these conditions does not take place for HCOOH

(Figure 1.4.8), one can conclude that no free hydroxyls are generated by ultrasound alone in our system. Furthermore, substantial rate enhancement (35 %) is observed when ultrasound is combined with photocatalysis. The effect of ultrasound in the degradation of salicylic acid is therefore due to the breakage of the larger molecule by the mechanisms described above or pyrolysis of a vaporized molecule [10], which cannot take place for iron ions and formic acid.

126 The comparison of the performance of our system for the single stage oxidation

· · reactions 2.1-2.4 (utilizing OH and H2O2) and 3 (utilizing only OH) elucidates the following phenomena. From general considerations, for such small entities as iron (II) and one-carbon molecule of formic acid the major direct effects of ultrasound (such as microstreaming, pyrolysis, and bubble implosion) cannot play a significant role due to the size of these species. Therefore, the dark sonolysis of these compounds can be attributed only to the attack by reactive oxygen species. As one can observe from Figure 1.4.7, the direct sonolysis of Fe2+ does in fact take place. The behavior of formic acid is different

(Figure 1.4.8). No conversion due to sonolysis was observed after 3 hours at 100 W/L and 200 W/L of ultrasonic input. Therefore, the generation of hydroxyls by ultrasound in our system is inhibited. One can recall from the literature that HCOOH can react only with hydroxyl radicals, and Fe2+ reacts with both hydroxyl radicals and hydrogen peroxide. Since no sonlolysis is observed for HCOOH and the additive effect of ultrasound and photocatalysis is observed for Fe2+, one can conclude that the hydroxyls generated in the bulk solution by ultrasound tend to recombine with the solid or with each other. The bimolecular reaction rate constant for the recombination of two hydroxyl radicals is two orders of magnitude larger than that of reaction 3 [39]. Furthermore, the generation of H2O2 by ultrasound in aqueous solutions has been previously reported [40].

Thus, the dark sonolysis of HCOOH is limited to the reaction with the unrecombined hydroxyl radicals in solution. On the contrary, for salicylic acid this factor is far less important since peroxides can attack the transient radicals of phenolic compounds [41], which can be formed during pyrolysis or bubble implosion. Therefore, we observe a small degree of conversion for salicylic acid.

127 As seen in Figure 1.4.7, the effect of ultrasound and photocatalysis on the Fricke reaction is additive and not synergistic. This is because Fe2+ can react with both ·OH and

· H2O2 (recombined OH). This is not the case for formic acid (Figure 1.4.8), which can react with ·OH only, and for which the effect is synergistic (35 % enhancement of the rate). Theron et al [26] proposed the deaggregation of the catalyst and the photocatalytic utilization of H2O2 produced by the ultrasound as two main reasons of the synergistic effect of ultrasound and photocatalysis. The additive effect of the latter two processes on the generation of Fe3+ supports the above conclusion since hydrogen peroxide is equally active for this reaction, and therefore the ultrasound simply serves as an additional source of radicals. If microstreaming, pyrolysis, or enhanced mass transfer were to play a major role in the rate enhancement, one would observe a synergistic effect for all reactions, including Fricke reaction. The synergistic effect of ultrasound on the photodegradation of formic acid also suggests that the utilization of H2O2 and defragmentation of the catalyst have the major impact on the performance of photocatalysis and ultrasound in the degradation of organic pollutants.

CONCLUSIONS

Sonophotocatalytic destruction of salicylic acid on four titania powders was studied. The relative extent of the possible reasons of the increased activity under ultrasonication was explored. The combination of the action of ultrasound and UV- assisted photocatalysis yielded synergistic effects for the catalysts with smaller particle size (Hombikat), while negligible enhancement was observed for large particle size photocatalyst (Aldrich anatase). Degussa P25 exhibited the highest overall activity for the

128 degradation of salicylic acid and moderate enhancement of activity by ultrasound.

Marginal enhancement was observed for smaller (0.1 g/L) concentration of the powders in the slurry. Single stage oxidation reactions were utilized for a mechanistic study, which allowed to isolate the main factors of the activity enhancement by ultrasound, namely, catalyst deaggregation and efficient utilization of recombined hydroxyl radicals.

Acknowledgements

This research is based upon work supported in part by the U.S. Army Research Office under grant number 40414/CH/YIP. The authors are also grateful to NATO (Grant No.

SfP-974209) for partial assistance.

REFERENCES

[1] M.S.T. Goncalves, A.M.F. Oliveira-Campose, E.M.M.S. Pinto, P.M.S. Plasencia, and M.J.R.P. Queiroz, Chemoshere, 39 (1999) 781

[2] D.C. Schmelling, and K.A. Gray, Wat. Res., 29 (1995) 2651

[3] E. Moctezuma, E.Leyva, E. Monreal, N. Villegas, and D. Infante, Chemsphere, 39 (1999) 511

[4] M.I. Litter, Appl. Catal. B, 23 (1999) 89

[5] P.V. Kamat, Chem. Rev., 93 (1993) 267

[6] M.R. Hoffmann, S.T. Martin, W. Choi, and D.W. Bahnemann, Chem. Rev., 95 (1995) 69

[7] A. Mills, and S. Le Hunte, J. Photochem. Photobio. A: Chmistry, 108 (1997) 1

[8] A. Kotronarou, G. Mills, and M. R. Hoffmann, J. Phys. Chem., 95 (1991) 3630

[9] N. Serpone, R. Terzian, P. Colarusso, C. Minero, and E. Pelizzetti, Res. Chem. Intermed., 18 (1992) 183

129

[10] C. Petrier, M.-F. Lamy, A. Francony, A. Benahceene, B. David, V. Renaudin, and N. Gondreson, J. Phys. Chem., 98 (1994) 10514

[11] N. Serpone, and P. Colarusso, Res. Chem. Intermed., 20 (1994) 635

[12] J.N. Jensen, Hazard. Ind. Wastes, 28 (1996) 265

[13] I. Hua, and M. R. Hoffmann, Environ. Sci. Technol., 31 (1997) 2237

[14] A.A. Atchley, and L.A. Crum, in Ultrasound-Its Chemical, Physical and Biological Effects, ed. K.S. Suslick, VCH, New York, 1988, ch.1, p. 1

[15] H.G. Flynn, in Physical Acoustics, ed. W.P. Mason, Academic Press, New York, 1964, vol. 1, pp. 58

[16] A.J. Walton, and G.T. Reynolds, Adv. Phys., 33 (1984) 595

[17] I. Hua, R.H. Hoechemer, and M. R. Hoffmann, J. Phys. Chem., 99 (1995) 2335

[18] N. Serpone, R. Terzian, H. Hidaka, and E. Pelizzetti, J. Phys. Chem., 99 (1995) 2634

[19] O. Legrini, E. Oliveros, and A.M. Braun, Chem. Rev., 93 (1993) 671

[20] V. Sukharev and R. Kershaw, J.Photochem.Photobiol. A: Chem, 98 (1996) 165

[21] G.L. Puma and P.L. Yue, Ind. Eng. Chem. Res., 38 (1999) 3246

[22] P.G. Smirniotis, L. Davydov, E.P. Reddy, and P. France, Proc.Ind. Appl. of Zeolites, Brugge, Belgium, Oct. 22-25, Technologisch Instituut, 2000, p. 233

[23] S. Tunesi, and M. Anderson, J. Phys. Chem., 95 (1991) 3399

[24] Fricke, H. and Hart, E.J. in Radiation Dosimetry Vol. II, Acad. Press, 1966, p.167

[25] L.K. Weavers, N. Malmstadt, and M.R. Hoffmann, Environ.Sci.Tech., 34 (2000) 1280

[26] P. Theron, P. Pichat, C. Guillard, C. Petrier, and T. Chopin, Phys.Chem.Chem.Phys., 1 (1999) 4663

[27] L. Thompson and P. Doraiswamy, Ind.Eng.Chem.Res., 38 (1999) 1215

[28] S. Friedlander in Smoke, Dust and Haze, Wiley and Sons Press, 1969

[29] L. Davydov, P.G. Smirniotis, and S.E. Pratsinis, Ind. Eng. Chem. Res., 38 (1999) 1376

130

[30] G. Heit, and A.M. Braun, Wat. Sci. Tech., 35 (1997) 25

[31] A.E. Regazzoni, P. Mandelbaum, M. Matsuyoshi, S. Schiller, S.A. Bilmes, and M.A. Blesa, Langmuir, 14 (1998) 868

[32] G.P. Fotou, and S.E. Pratsinis, Chem. Eng. Comm, 151 (1996) 251

[33] A. Mills, C.E. Holland, R.H. Davies, and D. Worsley, J.Photochem.Photobiol.A: Chem., 83 (1994) 257

[34] M.I. Cabrera, O.M. Alfano, and A.E. Cassano, J.Phys.Chem., 100 (1996) 20043

[35] A.J. Johnston and P. Hocking, in Hazardous waste management III, ACS Press (1993)

[36]T. J. Mason and J. L. Luche, in Chemistry under extreme or non-classical conditions, John Wiley, New York, 1997, p. 317

[37] K. Okamoto, Y. Yamammoto, H. Tanaka, M. Tanaka, and A. Itaya, Bull. Chem. Soc. Jpn., 58 (1985) 2015

[38] L. Davydov and P.G. Smirniotis, J. Catal., 191 (2000) 105

[39] M. Anbar and P. Neta, Int. J. Appl. Rad. Iso., 18 (1967) 493

[40] E. Gonze, L. Fourel, Y. Gonthier, P. Boldo, A. Bernis, Chem.Eng.J., 73 (1999) 93

[41] E. Pelizzetti and C. Minero, Coll.Surf.A: Physicochem.Eng.Asp., 151 (1999) 321

131

SECTION 2. KINETIC MODELING OF PHOTOCATALYTIC PROCESSES

Chapter 2.1. Quantification of the primary processes in aqueous heterogeneous

photocatalysis using single-stage oxidation reactions

INTRODUCTION

Reactive oxygen species [1] play a crucial role in heterogeneous photocatalysis

aimed at the degradation of organic compounds. They are formed on the surface of the

photocatalyst and react with either adsorbed or dissolved organic species. Turchi and

Ollis [2] showed that the major reaction route of photocatalytic oxidation involves the

hydroxyl radical attack. Several researchers proposed a number of methods to quantify

hydroxyl radical formation. a-Abstraction of hydrogen from methanol was proposed as

probe reaction [3], which leads to a presumably stable intermediate (formaldehyde)

according to the reaction:

· · · + OH, O2 + OH, O2 + OH, O2 CH3OH ® H2CO ® HCOOH ® CO2 (1) -H2O -H2O -H2O

This way it was possible to estimate reaction rates and quantum yields of hydroxyl radical production by titania in aqueous solutions. 3-Carboxyproxyl was also utilized as an ·OH trap [4], which forms a stable diamagnetic adduct. This allowed the estimation of the hydroxyl radicals concentrations in irradiated semiconductor slurries.

The above methods for quantification of primary radicals production can adequately characterize the “oxidative power” of a particular photocatalyst. However, the main disadvantage of both methods is that they rely on the formation of an intermediate compound, which can further degrade, especially under high intensity light conditions.

133 For this reason, a single stage oxidation reaction becomes highly desirable. There are

suggestions in the literature [5] that the photocatalytic degradation of formic acid

represents such reaction (it produces CO2 and H2O directly). Other researchers

investigated the photocatalytic and photoelectrocatalytic degradation of formic acid (the

last stage of reaction 1) in aqueous solution. For example, a ZnO photoelectrode was used for this reaction [6, 7]. It was shown that the electrons of the activated formate ion

(formyloxy radical) are released to oxygen when the latter is present in the system. This electron transfer happened in preference to the zinc oxide electrode as no significant increase in the anodic photocurrent took place. A study on the photodegradation of formic acid over platinized CdS semiconductor [8] revealed that the lower pH is favorable for this reaction. Immobilized TiO2 was also utilized to study the UV-assisted

degradation of formic acid [9, 10]. Benderskii et al [11] investigated the

photodegradation at high initial concentrations of formic acid.

Photochemical studies [12] and photoelectrocatalytic studies [6] suggest that the

last step of the above reaction (degradation of formic acid in the presence of oxygen) is a

single stage oxidation reaction. Moreover, oxygen prevents the formiloxy radicals from

dimerizing [12] making it an ideal candidate to study primary processes of

photocatalysis. As a supplement, chemical dosimetry will be used in the present study as

a means to quantify reactive oxygen species and primary photocatalytic processes

(electron hole recombination and reactive oxygen species formation) occurring in

irradiated semiconductors. The dependence of each reaction rate on the average

volumetric rate of radiant energy absorption as well as the BET surface area of the

catalysts will be elucidated.

134 THEORY

The present analysis will be based on the primary reactions presented below, which describe completely all the chemical features of the phenomenon under consideration. On the basis of the similarity of certain reactions (namely, radical formation and recombination) they were lumped into one reaction.

· A quantum of radiation with energy higher than the bandgap of the semiconductor

(l<385 nm) impinges on a titania particle producing an electron-hole pair:

hn

+ - TiO2 ® h + e Rex=m CcatG (2.1)

The rate of the above reaction is proportional to the photon absorption rate (assuming that the excitation is 100% efficient) as the concentration of the semiconductor in the slurry is constant throughout the reaction vessel.

· Some of the resulting electrons and holes recombine by the following reaction

+ - + - h + e ® TiO2 + heat Rrec=krec[h ][e ] (2.2)

The rate of recombination is dependent on both the number of holes and number of electrons in the bulk of the semiconductor particle.

· Valence-band holes form primary radicals

h+ + OH- ® ·OH

+ · + + h + H2O ® OH + H Rh+=kh+[h ] (2.3)

The concentration of ions and water in the aqueous solution is abundant. For this reason we assume the rate of surface reaction of holes be proportional to the hole concentration on the surface of the semiconductor.

135 · Conduction-band electrons are scavenged by oxygen forming multiple peroxide

species by the reactions

- · - e + O2 ® O2

+ · - · H + O2 ® HO2

+ · - - - 2 H + O2 +e ® H2O2 Re-=ke-[e ] (2.4)

It should be noted that the limiting step of the above reaction scheme is the surface

reaction assuming that all peroxide species are equally active. It was shown [1] that the

limiting stage in radical formation is the electron transfer between the solid surface of

titania and adsorbed molecular oxygen. Furthermore, the rate of this consecutive reaction

(2.4) must be equal to the electron transfer rate. Since an excessive amount of oxygen is

supplied to the reactor, the rate is proportional to the concentration of electrons.

· · Ñ · Some of the reactive oxygen species OH, HO2, H2O2 (denoted by O ) recombine

with the surface of titania:

Ñ Ñ TiO2 + O ® TiO2 + ½ O 2 Rct=kct[O ] (2.5)

The recombination of surface radicals is represented by electron transfer between the

surface radical and semiconductor surface. Furthermore, its rate is proportional only to

the concentration of surface radicals formed. The two groups of radicals (produced by

reaction 2.3 and reaction 2.4) can be lumped into one group when considering their

recombination with the surface of titania. This is because their recombination takes place

via electron transfer, and the rate of recombination must be equal for both groups in order

to maintain the electrical neutrality of the particle.

· The remaining radicals engage into chemical reaction with the organic present in the

system:

136 - Ñ - Ñ - HCOO + O ® H2O + CO2 + e RS=kS[O ][HCOO ] (2.6)

The rate of this reaction is proportional to both the concentration of surface radicals and

the concentration of the reactant in the solution. Others [1,2] proposed similar reactions for the description of the primary processes occurring in photocatalysis.

As a summary of the background knowledge about the oxidation of formic acid, the

following assumptions are adopted in the present study when developing the kinetic

model for the photodegradation of formic acid.

1. H2O, O2, and radicals produced are in adsorbed state as the lifetime of radicals in the

bulk solution is low [12].

2. Formic acid reacts with surface radicals, not with holes. The latter process may occur

at relatively high (1 M) concentrations of HCOOH [11] when formic acid can

compete with water for the adsorption sites of titania. This assumption will be

justified experimentally below.

3. The electron produced in reaction 2.6 is scavenged by oxygen [6, 12].

4. The rates of photolysis and dark heterogeneous degradation of formic acid are

negligible. They were found to be within the experimental error rendered by titration

in the degradation of formic acid in the absence of catalyst and at the concentration of

Degussa P25 of 0.25 g/l, respectively.

5. Steady state is achieved with respect to holes, electrons, and reactive oxygen species.

The following derivations are conducted in order to develop the reaction rate law

(based on known quantities) for the photodegradation of formic acid over irradiated

137 semiconductors (titania). From the balance of electrons and holes at steady state one can

deduce:

+ - + mC catG - k rec[h ][e ] - k h+ [h ] = 0 (3.1.1)

+ - - mC catG - k rec[h ][e ] - k e- [e ] = 0 (3.1.2)

+ - The above system yields kh+[h ]=ke-[e ]. Solving for the concentration of electrons and holes results in the following relations.

2 + k e- k e- k e- [h ] = - + 2 + mC catG 2k rec 4k rec k reck h+ (3.2.1)

k k2 k - = - h+ + h + + h+ m [e ] 2 CcatG 2krec 4krec krecke- (3.2.2)

The steady-state balance of reactive oxygen species can be derived from equations 2.3-

2.6 and it takes the form:

+ - Ñ Ñ - kh+[h ] + ke-[e ] – kct[O ] – kS[O ][HCOO ] = 0 (3.3)

Solving the above equation leads to the steady-state concentration of these species:

+ - Ñ k h+ [h ] + k e- [e ] [O ] = - k ct + k S [HCOO ] (3.4)

Hence, the local reaction rate of formic acid photodegradation by combining equations

2.6, 3.2.1, 3.2.2, 3.4 becomes:

138 k S [HCOO - ] é ù k ct k h+ k e- k rec R - = × × ê 1 + 4mC G -1ú HCOO k k cat k k 1 + S [HCOO - ] rec ëê h+ e - ûú k ct (3.5)

One can observe that the constant (kS/kct) of the first term in the numerator represents the ratio of photodegradation rate to the recombination of reactive oxygen radicals. The second multiplier represents the ratio of a product of radical generation rates to electron- hole recombination rate. These two complexes of kinetic constants (kS/kct and khke/krec) can serve the purpose of photocatalyst characterization in their combined form as one can expect high photodegradation rates from photocatalysts with high values of both complexes. We will attempt, however, to obtain individual reaction rates of radical generation and electron-hole recombination. For this reason two experimental conditions may be considered. The first condition would be the one of relatively high concentration of formic acid (greater than 5 mM, as will be proved experimentally), so that the unity in the denominator can be neglected. From such high concentration experiments it will be

(0) possible to determine the parameter kh+ke-/krec by plotting the reaction rate (R ) as a function of the incident photon flux.

In the second extreme case, low concentration of the reactant (below 0.5 mM, as will be shown later) make it possible to neglect the concentration term in the denominator, and the reaction becomes first order with respect to formic acid (R(1)). Thus, for the same concentration of catalyst, light source, and other experimental conditions

two reaction rates can be determined:

139 é ù (0) k h+k e - k rec R = × ê 1 + 4mC catG -1ú k rec ëê k h+k e- ûú (3.6)

k R (1) = S [HCOO - ]R (0) k ct (3.7)

By observing the steady state balance of radicals one can derive the total rate of radical

generation in a particular photoreactor. This will allow to compare the performance of

different photocatalysts on its basis. After these algebraic manipulations with the balance

of electrons and holes (equation 3.1) and active radicals (equation 3.3) one can arrive at

the following relations:

k 2 + S [HCOO - ] k R = R + R = ct R (0) gen h+ e- k 1 + S [HCOO - ] k ct (3.8)

k 2 + S [HCOO - ] k R = mC G - ct R (0) rec cat k 1 + S [HCOO - ] k ct (3.9)

Quantum yield of any photochemical process is usually defined as the ratio of the

rate of this process to the rate of energy absorption. If the absorption of radiation flux Ga

brings about the rate of transformation Rn, the expression of the quantum yield can be

written in the following fashion:

R F = n G a (3.10)

140 Overall, the above methodology allows assessing the rates and quantum yields of primary

photocatalytic processes, such as active radical generation and electron-hole

recombination from simple experimental measurements. The knowledge of these rates

will allow us to predict reaction rates and quantum yields in different photoreactors at

different catalyst concentrations and create a common means of comparing different

photocatalytic semiconductors.

EXPERIMENTAL

The experimental work performed in this study aims at testing the proposed methodology with selected photocatalysts. Two main categories of experiments were performed in the

present study: high initial concentration of formic acid and low initial concentration of

formic acid. Commercial titanias were evaluated for the photocatalytic degradation of

formic acid. The titania powders utilized in the present study were Aldrich anatase

(denoted as AA), Hombikat UV-100 (HK), Ishihara ST-21 (ST), and Degussa P25 (P25); their characteristics are shown in Table 2.1.1. As one can observe, these catalysts possess quite different optical and adsorption properties. The suspensions were prepared with the titania particles dispersed ultrasonically in aqueous solutions containing the appropriate amount of formic acid. Prior to each experiment, the UV lamp was warmed up for 5 min.

Samples were withdrawn from the reactor and filtered with 0.2 mm membrane filters

(Gelman Sciences) to remove the titania particles. The suspension temperature was maintained at 30+3oC. The concentration of the unreacted compound was measured by a

141 Table 2.1.1. Photocatalysts employed in the present study

Catalyst BET Surface Anatase Primary Extinction Scattering Area(1), m2/g Content(2), Particle Size, Coefficient b(3), Albedo w(3) % nm L/(cm g)

AA 10.0 99 100-200(4) 15.32 0.388 P25 75.0 70-80 ~20(2) 23.50 0.663 HK 313.2 99 <10(2) 7.42 0.467 ST 64.0 95 ~20(2) 9.32 0.562

1. Measured by Micromeritics Gemini 2320 surface analyzer at 77 K 2. According to the manufacturer’s specification 3. Measured spectrophotometrically by Shimadzu 2501PC, equipped with an integrating sphere Shimadzu ISR 2200. 4. Determined by Transmission Electron Microscopy (Siemens) conductivity meter (VWR Scientific, golden cell, cell constant 10.25 cm-1). It was calibrated by standard solutions of formic acid with the concentrations ranging from

0.001 to 20 mM. A quadratic curve passing through zero and the experimental points was fitted with the accuracy of 0.999; this curve was used as the conductivity-concentration correlation. The reaction mixtures were also tested for the presence of intermediates at different time on stream by direct injection into GC (HP 5890 Series II) equipped with a

MS (HP-5972). No detectable amounts of intermediate compounds were found.

A modified Fricke dosimeter [13] was used to determine the overall concentration of reactive oxygen species in irradiated aqueous suspensions of titania. It consisted of 5 mM of FeSO4 (Fisher), 10 mM of CuSO4 (Fisher), and 20 mM of H2SO4 (Fisher). During the reaction with reactive oxygen species Fe3+ ion is produced, and it was quantified by spectrophotometry (Shimadzu U160) at the wavelength of 304 nm. Standard quartz cuvettes with the pathlength of 1 cm were used. The absorbance by Fe3+ was correlated with the concentration of Fe2+ reacted and further used for calculations.

142 In the experiments of the photooxidation of nitrite ions, 30 mM concentration of

NaNO2 (Fisher) was subjected to photolyses by different powders of titania under the same operating conditions. The unreacted nitrite was determined by titration using 0.005 mM KMnO4 (Fisher) solution. The end point of titration was determined colorimetrically at the wavelength of 535 nm.

A conventional annular liquid-phase photocatalytic reactor (Ace Glass, Inc., Cat.

No. 7868-10) was utilized for the study as described elsewhere [14]. A glass stirring shaft

assured mixing and circulation within the vessel. A double-walled quartz immersion well was placed in the middle of the reaction vessel. Its purpose was to allow the circulation of water for cooling the light source and the solution. The UV part of lamp radiation was also filtered by a Pyrex filter (7740, Ace Glass, Inc.) placed between the lamp and the immersion well. The thickness of the Pyrex filter was 2.38 mm, the inside diameter was

26 mm and it allowed to eliminate the far and mid-UV bands (l<320 nm) of the emission spectrum of the lamp. The immersion type UV-radiation source was 450 W (medium pressure) mercury vapor quartz lamp (Ace Glass). Oxygen (Wright Brothers, 99.5 %) was sparged into the solution at 500 cm3/min through a fritted glass tube at the bottom of the vessel. With this flow rate no deposition on the walls of the reactor was observed.

Such oxygen flow rate did not lead to reactant stripping and entrainment into the gas phase.

The titania powders were also evaluated in a differential photoreactor [14] with the same lamp and cooling jacket but different thickness of the reaction zone. A conventional magnetic stirrer vigorously agitated the suspension in the reactor. A laboratory peristaltic pump (Masterflex 7017-21) was used to run the suspension through

143 the reactor. An intermittent vessel was used for degassing of the suspension and sample

withdrawal. This vessel was also agitated by a magnetic stirrer in order to avoid

sedimentation of titania and provide better dispersion of oxygen bubbles. Flexible plastic tubing connected the pump, vessel, and reactor. No reactant stripping was observed during the experiments.

A radiometer (International Light, Inc. Model IL 1700) was used to determine

local incident photon flux densities in the reactor. It was connected with a detector

(International Light, Inc., Model SED033 #3435). A UV-filter was employed in order to

measure only the light of interest and its response exhibited a strong maximum at the

wavelength of 355 nm. The incident radiation at RI as well as the outgoing radiation at RO was measured at various locations along the reactor length by the detector at the outer wall of the quartz cooling jacket.

Spectrophotometer Shimadzu 2501PC with an integrating sphere attachment ISR

2200 was used to determine the optical properties of titania powders and their suspensions. The diffuse-reflectance spectra of all four powders utilized in the study were found to be almost identical. This strongly suggests the absence of quantum confinement effects frequently occurring in semiconductor nanoparticles. To determine the absorption and extinction coefficients, aqueous slurries of the powders of several different concentrations (0.025-0.25 g/L) were pre-ultrasonicated for 1 min at pH=4. They were placed into 1 cm quartz cuvettes for absorbance measurements at 355 nm. The absorbance was then plotted versus catalyst concentration, and the slopes yielded the corresponding optical characteristics. The measurement with an integrating sphere attachment gave the absorption coefficient (m), and the measurement without the latter

144 gave extinction coefficient (b). Scattering albedo (w) was determined as a fraction of the

light scattered away, w=(b-m)/b.

RESULTS AND DISCUSSION

At the beginning of this section we wish to identify the dominating pathways for

the reaction scheme 2.1-2.6. In order to simplify the above reaction scheme one can use

the knowledge obtained from homogeneous photochemistry. First of all, it is known that

peroxide radicals and hydrogen peroxide do not react with formic acid in homogeneous

solution [12]. Our experiments with the initial concentration of formic acid of 30 mM and that of hydrogen peroxide of 30 mM proved the same fact since they yielded no discernible conversion after one hour. Hence, we can narrow down the number of reactive oxygen species to hydroxyl radicals. Furthermore, peroxide radicals (reaction

2.4) form hydrogen peroxide, which being irradiated by UV may produce two hydroxyl radicals according to the reaction:

hn · H2O2 ® 2 OH (4.1)

There are conflicting reports in the literature on the values of the quantum yield for this

reaction in homogeneous medium. Calvert and Pitts [15] showed, however, that the

absorption by hydrogen peroxide in the near UV range (320-400 nm) is negligible, and

the reaction 4.1 takes place primarily in the mid- and far UV radiation.

Independent experiments utilizing modified Fricke dosimeter [13] can be performed to quantify the concentration of reactive oxygen species in an irradiated slurry.

The following chemical transformations take place:

Fe2+ + ·OH ® Fe3+ + OH- (4.2.1)

145 2+ · + + Cu + HO2 ® Cu + H + O2 (4.2.2)

Fe3+ + Cu+ ® Fe2+ + Cu2+ (4.2.3)

2+ 3+ - 2Fe + H2O2 ® 2Fe + 2OH (4.2.4)

One can observe that two types of reactive oxygen species (hydroxyls and peroxide)

participate in the oxidation of Fe2+ producing three Fe3+ ions. The overall reaction is

independent of the oxygen concentration as it is liberated by reaction 4.2.2. Since one of

the three Fe3+ ions produced is scavenged by Cu+ ions (reaction 4.2.3), we have two

radicals producing two iron (III) ions. Therefore, the oxidation rate of iron ions represents

a realistic measure of the reactive oxygen species in photocatalytic solutions. Considering

the irreversible bimolecular reaction rate law for the reaction of ferrous ions with reactive

oxygen species:

2+ d[Fe ] 2+ Ñ - = k 2+ [Fe ][O ] dt Fe (4.3)

The above relation is analogous to equation 2.6, and rearranging it yields:

1 d ln[Fe 2+ ] [O Ñ ] = - k + dt Fe2 (4.4)

Replacing differentials in equation 4.4 with finite differences one can determine the

instantaneous concentration of radicals based on known concentrations of iron (II) ion.

The values of the bimolecular reaction rate constants for Fricke dosimeter are available in

the literature (we will use the value of 2.5×108 L/(mol min) reported in [16]). The input of

homogeneous photochemical oxidation (photo-Fenton process) was assessed for the

146

-12 2.5x10 AA 0.1 g/L AA 0.15 g/L -12 2.0x10 AA 0.25 g/L

-12 1.5x10

-12 1.0x10 Concentration, mol/L Ñ Ñ O -13 5.0x10

0.0 0 1 2 3 4 5 6 Time, min A

4.0x10-12 P25 0.1 g/L -12 3.5x10 P25 0.15 g/L P25 0.25 g/L 3.0x10-12

2.5x10-12

2.0x10-12

1.5x10-12

1.0x10-12 Concentration, mol/L Ñ Ñ O 5.0x10-13

0.0 0 1 2 3 4 5 6 Time, min B

Figure 2.1.1. Evolution of the concentration of reactive oxygen species in slurry: 2+ [Fe ]0=0.03 M, catalyst – Degussa P25 (a), Aldrich anatase (b), lamp power – 450 W, temperature 30±3oC

147 Fricke dosimetry reactions and it contributed to about 5-10 % of the activity. This is due

to the instability of the ferrous ion in solution under illumination. To minimize this effect

and to enable the dosimeter to work in less acidic media, cupric salts are added to the

solution. Transition metal ions can absorb UV radiation as well as titania. In order to test

the alteration of the radiation field in the reactor due to the absorption by the constituents

of the Fricke solution, its optical density was analyzed at the wavelength of 355 nm (peak

radiation in our reactor). This was done before the reaction and after 10 minutes of

reaction (Degussa P25, 0.25 g/L). The absorbances found were in the range from 0.078 to

0.101, respectively. This corresponds to the absorbance by 0.25 g/L slurry of Degussa

P25 in the vicinity of 2. Therefore, the maximum alteration of the photon flux by the

transition metal ions employed can reach about 5 % of the input of the absorption by

titania. Furthermore, the above species are present in the solution in ionic form and

therefore do not contribute to light scattering. This leads to the conclusion that the

radiation fields in the photodegradation of formic acid and oxidation of Fricke solution

can be considered identical. The time-dependence of the hydroxyl radical concentrations

(calculated by equation 4.4) for several concentrations of the photocatalysts is shown in

Figure 2.1.1a (P25) and 1b (AA). One can observe that the steady-state concentration of

these radicals is not achieved instantaneously. On the contrary, it passes through a

maximum for all concentrations of AA and P25; the maximum becomes less pronounced

for the lower concentrations of the catalyst. This effect is due to lower incipient surface

area in the reactor and consequently lower number of surface hydroxyls capable of

becoming hydroxyl radicals. One can also observe that the maximum is less pronounced

for AA. This is due to its lower BET surface area and lower number of active sites of the

148 latter catalyst. Similar trend (maximum of the concentration of primary radicals) was

qualitatively predicted by Riegel and Bolton [17] and attributed to the lower rate of

electron entrapment in comparison with that of hole entrapment. Thus, the trapped holes

form more radicals at the beginning of the reaction than electrons do, until the

recombination makes up for the excess of radical formation. Furthermore, these steady- state concentrations of primary radicals are in the order of 10-12 mol/L after several minutes of reaction. The total number of active sites (capable of becoming active radicals) of titania was estimated as 5×1018 m-2 by Pruden and Ollis [18]. This corresponds to the “concentration” of active sites equal to 1.55×10-4 mol/L when 0.25 g/L slurry of

P25 is irradiated. One can observe that this value is considerably higher than that of the steady state concentration of active radicals. Hence, a very small fraction of the catalytic sites of titania is actually active in the irradiated slurry. This fact is primarily due to inefficient light utilization and high electron-hole recombination rates (as will be shown later in this section).

In order to complete the above kinetic model, certain assumptions of the theoretical section must be clarified experimentally. There has been extensive discussion in the literature (summarized in [19]) regarding whether organic molecules react with surface radicals of the excited semiconductor or with its holes directly during photodegradation. From general considerations, the latter is justifiable when the organic molecule and water can compete for the active sites of the photocatalyst or replace hydroxyl groups on its surface [3]. This can be true for relatively high concentrations

(order of moles per liter) of organics in the aqueous system (such as in the one reported in

[11]). The vast majority of photocatalytic studies, however, deal with very low (order of

149 ppm) concentrations. Additional evidence of the participation of primary radicals in the

main oxidation reaction may be obtained by the photooxidation of inorganic ions, which

cannot occur via radical mechanism. Indeed, low photoreactivity of illuminated ZnO

electrode with respect to the oxidation of sulfite and bromide ions was observed [20].

Herrmann and Pichat [21] found no photocatalytic activity to oxidize chloride-ions by

titania. To assess further the extent of direct oxidation by valence band holes,

experiments were conducted utilizing oxygen as an electron scavenger and nitrite ion as a

labile reducing agent, which followed the mechanism followed the mechanism:

- + - + NO2 + h + H2O ® NO3 + H E=0.934 V (4.5)

The extent of nitrite photodegradation can elucidate to which degree the UV-produced holes participate in the reaction directly. Experiments with initial sodium nitrite concentrations of 30 mM in 0.25 g/L slurry of P25 showed very low rates of photooxidation. The latter were comparable with those of homogeneous oxidation by molecular oxygen in the dark with a total quantum yield in the order of 10-3 moles/Einstein, which is two orders of magnitude lower than that obtained with formic acid. This proves that the surface reactive oxygen species are the main oxidizing agents in aqueous photocatalysis.

To obtain the lumped parameters of equation 3.5 of average zero-order reaction rates can be plotted as a function of the volume-averaged square root of the absorbed photon flux in the reactor (Figure 2.1.2). The absorbed photon flux was varied by changing the photocatalyst concentration in the slurry. One can measure experimentally the radiant flux on the cooling jacket using a radiation detector, thus obtaining a realistic profile of radiation entering the reaction zone. However, one needs to utilize modeling in order to

150 determine the value of photon fluxes at different locations in the reactor (parameter G in

equation 3.5). Then this equation can be integrated for each specific case and its constants

can be determined. It was shown previously [14] that the shape of the axial radiation

profile emitted by our mercury lamps remains almost unchanged having passed through the reaction annulus. This leads to the conclusion that angular scattering can be neglected as the scattered beams compensate each other. The Schuster-Schwartzichild radiant transfer approximation was developed for this case [22], and it will be applied to our system to determine radiant fluxes at any location in the reactor. This two flux model assumes that the scattering remains isotropic throughout the vessel and the absorption and scattering of radiation are accounted for by two independent coefficients. Moreover, this model lumps together all forward-scattered beams into one flux and backward-scattered into another. We will therefore adopt the equation derived in [22] for the two-flux

radiation transfer in a suspension of microorganisms with a modification for annular

geometry. According to the model, the photon balance at each axial coordinate in the

reactor space is given by:

+ d(rF ) + 1 - + = -CcatmrF + Ccats(rF - rF ) (4.6.1) dr 2

- d(rF ) - 1 - + = CcatmrF + Ccats(rF - rF ) (4.6.2) dr 2

where F+ corresponds to the forward photon flux, and F- corresponds to the backward

photon flux. Equations 4.6.1 and 4.6.2 are subject to the following boundary conditions:

151

-4 2.0x10

-4 1.8x10 AA P25 -4 1.6x10 HK -4 1.4x10 ST

-4 1.2x10

-4 1.0x10

-5 8.0x10 , mol/(L min)

(0) -5

R 6.0x10

-5 4.0x10

-5 2.0x10

0.010 0.015 0.020 0.025 0.030

1/2 1/2 -1/2 Ga , Einstein (L min)

Figure 2.1.2. Average zero-order rate of photodegradation of formic acid versus square - root of the absorbed photon flux: [HCOO ]o=30 mM, lamp power – 450 W, temperature 30±3oC.

+ r=RI: F =GI(z) (4.6.3)

- r=RO: F =0 (4.6.4)

The solution of equations 4.6.1 and 4.6.2 for G=F++F- with the boundary conditions 4.6.3 and 4.6.4 yield the following relation:

R (1+ a)exp[-d(r -1)]- (1- a)exp[d(r -1)] G(r,z) = 2G (z) I I 2 2 r(RO - RI) + RI (1+ a) exp(d) - (1 - a) exp(-d)

(4.6.5)

where GI denotes the incident radiation (or radiant flux at the inner radius of the inner

wall RI, determined experimentally). The values of parameters a, d, z, and r are

determined from the following relations:

152 r - R r = I RO - RI z z = L a = (1- w)1/2

d = bCcata(RO - R I) (4.6.6) where RO is the radius of the outer wall, r - radial coordinate, w =s/b– scattering albedo, b – extinction coefficient of the slurry, and Ccat – photocatalyst concentration. In order to use equation 4.6.5 one needs to estimate the values of b and w from independent spectrophotometric experiments without and with an integrating sphere, respectively. The values of the optical properties of the suspensions used are summarized in Table 2.1.1.

One can observe form Figure 2.1.2 that for all four catalysts the linearity holds, which was also found by others [17]. Slight reproducible deviation from linearity can be seen for AA at lower (0.05-0.15 g/L) concentrations. As reported before [4], the square-root dependence of the reaction rate (or quantum yield) on the absorbed photon flux may not hold for the entire range of fluxes. More specifically, at lower absorbed intensities such dependence becomes linear, thus switching the kinetic regime with respect to the

“concentration” of absorbed photons. Furthermore, the passage of the lines through the origin at zero absorbed photon flux should not be expected since at lower values of radiant flux the kinetic regime is different (equation 3.5). Taking the slopes of these dependencies allow one to calculate the lumped kinetic parameter (kh+ke-/krec) for each photocatalyst. The values of this parameter determined with the degree of accuracy of

18% for AA and within 10% for the other three catalysts are presented in Table 2.1.2.

153 The numerator of this parameter represents the cumulative rate of formation of active

oxygen

1.0 ST HK 0.8 P25 AA

0.6 0

C/C 0.4

0.2

0.0 0 10 20 30 40 50

Time, min

- Figure 2.1.3. Time course of the concentration of formic acid: [HCOO ]o=1.5 mM, lamp power – 450 W, temperature 30±3oC, catalyst concentration – 0.25 g/L

species and its denominator corresponds to the rate of electron-hole recombination. The

values of this parameter are much lower than unity, which substantiates the widely

accepted notion of electron-hole recombination rate as the pivotal fact of photocatalyst efficiency [1]. One can also observe that this ratio increases in the following fashion:

AA

(such as rutile) enhances charge separation, which in turn leads to lower recombination rates [23]. This lumped parameter is not proportional to the BET surface area of the

154 photocatalyst. One would expect it to be proportional to BET area squared, since the rates

of radical formation are surface rates and recombination occurs primarily in the bulk of

the semiconductor. This behavior takes place because the entire surface of the high-BET area catalysts is not fully subjected to light in photoreactors. Indeed, the lifespan of a single particle of size 5 nm (HK-100) in slurry is extremely small, after which it forms an aggregate with another particle, thus reducing the incipient surface of the catalyst.

Table 2.1.2. Comparison of the parameters of equation 3.5. Catalyst kh+ke-/krec, m, L/(cm g) kS/kct, mol/(L min) L/(mol min)(1) AA 1.82×10-6 9.38 40815 P25 3.25×10-5 7.92 9756 HK 1.95×10-5 9.39 1443 ST 2.00×10-5 4.09 2477 1. Determined at the catalyst concentration of 0.25 g/L, lamp power=450 - o W, [HCOO ]0=1.5 mM, temperature = 30±3 C

As mentioned in the experimental section, experiments were performed with all

the photocatalysts at relatively low concentrations of formic acid (starting with ~1.5

mM). The rate equation proved to be of shifting order. Indeed, the first multiplier of

equation 3.5 is of Langmuir type; the reactant concentration-time dependence is illustrated in Figure 2.1.3. Similar trends were observed by Matthews [5], who also reported low agreement between the experimental values at low conversions and those calculated by Langmuirian model. This major discrepancy had the form of a “time-lag”

during the first few minutes. Moreover, since we observed steady state of hydroxyl

radicals only after 5 min on stream, we can integrate the time variable in equation 3.5

from some parameter td (“dead” time) instead of zero. Therefore, the integral Langmuir-

type model becomes:

155 [HCOO - ] [HCOO - ] t = k1 - + k 2 ln - + t d [HCOO ]0 [HCOO ]0 (4.7)

- (0) (0) where k1= -[HCOO ]0/R and k2= -1/(R kS/kct). Furthermore, the parameter kS/kct was extracted from equation 4.5 using curve-fitting software. The values of this parameter are

18000

16000 P25 14000 HK ST 12000

10000

, L/mol 8000 ct /k s

k 6000

4000

2000

0 0.0 0.1 0.2 0.3 0.4 0.5 0.6

C cat, g/L

Figure 2.1.4. Langmuir parameter of equation 3.5 versus catalyst concentration: - o [HCOO ]o=1.5 mM, lamp power – 450 W, temperature 30±3 C

shown in Figure 2.1.4 as a function of catalyst concentration. One can observe that this variable monotonically decreases with the increase in the photocatalyst loading for a wide range of concentrations (0.05-0.5 g/l). It should be noted that for AA the reactant concentration range corresponding to the first order rate law is much shorter than that for the other three catalysts (Figure 2.1.3), which did not allow us to elucidate a distinct trend of kS/kct for AA. Its value for all catalysts is much higher than unity, which shows that the

156 majority of the reactive oxygen species formed on the surface of the catalyst engage into the chemical reaction even at low concentrations of the organic. As the catalyst concentration increases, so does the extent of recombination due to lower concentration of the available organic per gram of catalyst. For this reason the monotonic decrease of the kS/kct should be expected.

The comparison of the parameter kS/kct for the catalysts employed in the study

(Table 2.1.2) shows that the extent of reaction with the ambient organic are lower for the higher BET surface area catalysts (HK

P25 [14].

Figures 2.1.5 and 6 show the rate of generation of active oxygen radicals

(equation 3.8) and electron-hole recombination (equation 3.9) as a function of photocatalyst concentration for the four powders involved in the present study. The rate of radical generation represents the oxidative power of a particular catalyst. One can observe that it is the lowest for Aldrich anatase, however, not proportional to the BET surface area of catalysts (Table 2.1.1). It also attains a plateau value at higher concentrations of photocatalyst as almost entire portion of the photon flux becomes utilized in the reactor. One should also expect the lowering of this value at even higher concentration of the photocatalyst since the smaller fraction of the reactor volume is

157 subject to light. Furthermore, for the highest BET surface area catalyst (HK) the

generation rate (equation 3.8) is approximately the same as for P25 and ST.

-4 3.5x10 ST HK 3.0x10 -4 P25

2.5x10 -4 AA

2.0x10 -4

1.5x10 -4 , mol/(L min)

gen -4

R 1.0x10

5.0x10 -5

0.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6

C cat, g/L

Figure 2.1.5. Reactive oxygen species generation rates (equation 3.8) as a function of photocatalyst concentration: lamp power – 450 W, temperature 30±3oC.

This again underlines the fact that the working surface area of photocatalysts is much

smaller than their total surface area, even for non-porous particles. The values of the

generation rate (order of 10-4 mol/(L min)) divided by the photon absorption (order of 10-

2 Einstein/(L min)) lead to the values of quantum yields in the order of 10-2 mol/Einstein

commonly encountered in heterogeneous photocatalysis. Therefore, in the absence of

secondary radical oxidation reactions this is the limiting value for the quantum yield.

Figure 2.1.6 shows the rate of electron-hole recombination as a function of the

photocatalyst concentration. It can be readily seen from this Figure 2.1.that the four

photocatalysts behave similarly. A monotonic increase in the recombination rate is

observed for all four catalysts with the increasing concentration; it arrives at a plateau

158

-3 2.5x10

-3 2.0x10

-3 1.5x10

, mol/(L min) -3 ST

rec 1.0x10

R HK P25 -4 5.0x10 AA

0.0 0.1 0.2 0.3 0.4 0.5 0.6

C cat, g/L

Figure 2.1.6. Electron-hole recombination rates (equation 3.9) versus photocatalyst concentration: lamp power – 450 W, temperature 30±3oC. value for concentrations of 0.25 g/l and higher. Furthermore, the average electron-hole recombination for all catalysts with different physical properties (Table 2.1.1) is in the order of 10-3. Such proximity of recombination rates is corroborated by the fact that the process of electron-hole recombination occurs in the bulk of the semiconductor.

Therefore, since the core of each semiconductor photocatalyst utilized in this study is represented by anatase, the recombination rates are expected to be close. The highest recombination rates are observed for AA in the whole range of concentrations. This is because the particle size (Table 2.1.1) of the latter catalyst is much larger than that of all others. This prompts much higher hole diffusion length and, as a consequence, higher probability of bulk recombination. By comparing Figures 2.1.5 and 6 one can observe that the electron-hole recombination rate is considerably higher than that of generation of

159 reactive oxygen species. This explains why quantum yields of photocatalytic degradation of organic compounds are lower than unity.

In order to compare the efficiency of each catalyst employed in the study for photochemical conversion of light one can also obtain quantum yields for primary processes such as that of radical generation and electron-hole recombination. These quantum yields are presented in Table 2.1.3 and they are independent of the catalyst concentration. One can observe that 6-7 fold increase in BET surface area (AA vs. P25 and ST, Table 2.1.1) lead to the similar increase in the quantum yield of generation.

However, the further increase of the surface area (up to 313.2 for HK) leads only to ~5 % enhancement. This apparently implies that the limit for our system (lamp, reactor, semiconductor species) has been achieved. Thus, the quantum yield of radical generation is an inherent characteristic of the oxidative powder of photocatalysts, and it is the highest number one could achieve in the absence recombination of radicals (reaction 2.5) and secondary oxidation reactions. Furthermore, for all photocatalysts the quantum yield of electron-hole recombination is substantially larger than that of radical generation, again underlining the limitations of photocatalytic degradation in terms of energy efficiency, as the energy from recombination is largely lost.

Table 2.1.3. Quantum yields of radical generation and electron-hole recombination.

Catalyst Quantum yield, mol/Einstein Generation Recombination AA 0.0285±0.007 0.972±0.007 P25 0.122±0.011 0.878±0.011 HK 0.185±0.029 0.816±0.029 ST 0.172±0.021 0.828±0.021

160 In our earlier work we found that the apparent quantum yield of phenol and 2,4,6- trichlorophenol photodegradation can vary quite substantially (up to six-fold) with the reactor setup [14]. To illustrate the applicability of the present method one can determine the quantum yields of the degradation of formic acid in two different photoreactors. The reactors tested corresponded to the conventional one (reaction zone size 24cm x0.6cm, total volume 0.65 L) and the differential one (reaction zone size 12cm x0.15cm, total volume 0.25 L). The quantum yields obtained for the four photocatalysts utilized in the present study at various concentrations (0.05-0.5 g/L) are presented in Table 2.1.4. The maximal difference in the quantum yields obtained from these two reactors is in the vicinity of 20 % (for HK). Such low difference is primarily due to the absence of secondary oxidation reactions taking place in the degradation of phenols. Heit and Braun

[24] studied photolysis in aqueous systems and isolated two reactor volumes. These volumes correspond to strongly irradiated primary reaction zone and weakly illuminated secondary reaction zone, which occupies the major part of the reactor volume. The latter zone is dominated by the oxidation reactions between the organic and dissolved oxygen, hydrogen peroxide, and relatively stable oxygen radicals (such as peroxyl). Furthermore,

Table 2.1.4. Quantum yields of formic acid degradation.

Catalyst Quantum yield, mol/Einstein Turnover -1 Conventional Differential Frequencies, s AA 0.017±0.002 0.018±0.004 0.035 P25 0.082±0.008 0. 090±0.008 0.022 HK 0.101±0.018 0.081±0.008 0.0037 ST 0.100±0.019 0.094±0.002 0.018

161 in the case when single stage complete oxidation of the reactant takes place, the

secondary reaction volume does not exist. This is the primary reason of the less

variability in the quantum yields of Table 2.1.4, and it justifies the validity of the

proposed methodology. Furthermore, this fact substantiates our assumption of “high”

photon flux in the reactor, which allowed us to neglect the unity under the square root in

equation 3.5. More specifically, one cannot expect the shift of the kinetic regime with

respect to photons absorbed in a thin annulus, as the radial attenuation of light is low. The

turnover frequencies for each catalyst have also been calculated based on the number of

sites calculated by Pruden and Ollis [18]. It was assumed that only the anatase phase of

the titania powders possesses active sites. The much higher turnover number for AA is

notable, although its apparent activity is the lowest. HK possesses the lowest turnover

frequency because the degree of utilization of its surface is very low. Conclusively, the

utilization of a single-stage oxidation reaction aids quantification of primary processes in heterogeneous photocatalysis and provides an elegant means of comparing different semiconducting powders.

CONCLUSIONS

The quantification of primary photocatalytic processes was performed for selected titania powders by utilizing a single-stage oxidation reaction, namely, the photodegradation of

HCOOH. It elucidated the effect of BET surface area and phase composition of the photocatalysts on their behavior. In particular, the concentration of reactive oxygen species was found to pass through a maximum and attain a steady state. Furthermore, the rate of generation of these species increases monotonically with catalyst concentration

162 and arrives at a plateau value in the concentration vicinity of 0.25 g/l. This generation rate follows the pattern: ST » HK » P25 >AA; its value for Aldrich Anatase is approximately three times lower that that for the other catalysts. On the contrary, the electron-hole recombination rate was the highest for the latter catalyst and changed in the following fashion: AA > P25 > ST » HK. The highest value for Aldrich Anatase is related to its large primary particle size, which leads to a higher diffusion length for the charges to reach the surface of the particle. Furthermore, the knowledge of primary rates of photocatalysis reveals the nature of low quantum yields commonly encountered and can provide an objective means to compare different photocatalysts.

ACKNOWLEDGEMENT

The authors wish to acknowledge the Technical Association of Pulp and Paper Industry

(TAPPI) for supporting this work through grant No. PE-380-97.

NOMENCLATURE

Ccat – photocatalyst concentration, g/l

F+ - forward photon flux density, Einstein/(cm2 min)

F- - backward photon flux density, Einstein/(cm2 min)

G – photon flux density, Einstein/(cm2 min)

Ga – absorbed photon flux density, Einstein/(L min)

2 GI –density of the photon flux entering reaction zone, Einstein/(cm min) k1 – constant of equation 4.7 k2 – constant of equation 4.7

163 kct – radical recombination (charge transfer) reaction (2.5) rate constant, 1/min

ke- - surface electron reaction rate constant, 1/min

2+ kFe2+ - bimolecular surface reaction rate constant of Fe ion oxidation, L/(mol min)

kh+ - surface hole reaction rate constant, 1/min

krec – electron-hole recombination rate constant, L/(mol min)

ks - bimolecular surface reaction (2.6) rate constant, L/(mol min)

L – reactor length, cm r – reactor radius, cm

R(0) – reaction rate of photodegradation (zero order), mol/(L min)

R(1) - reaction rate of photodegradation (first order), mol/(L min)

Rct – charge transfer reaction (5) rate, mol/(L min)

Re- - surface electron reaction rate, mol/(L min)

Rex – excitation rate, mol/(L min)

Rgen – radical generation rate, mol/(L min)

Rh+ - surface hole reaction rate, mol/(L min)

Rn - reaction rate of the nth process, mol/(L min)

Rrec – electron-hole recombination rate, mol/(L min)

Rs – bimolecular surface reaction (6) rate, mol/(L min)

RI – inner radius of the reaction zone, cm

RO – outer radius of the reaction zone, cm

t - time, min

td – dead time, min

z –axial coordinate, cm

164

Greek Symbols

a - parameter of equation 4.6.2

b – extinction coefficient of the photocatalyst slurry, L/(g cm)

d - parameter of equation 4.6.2 m – absorption coefficient of the photocatalyst slurry, L/(g cm) s – scattering coefficient of the photocatalyst slurry, L/(g cm) w – scattering albedo of the photocatalyst slurry

F – quantum yield, moles/Einstein

REFERENCES

1. Hoffmann, M.R., Martin, S.T., Choi, W., and Bahnemann, D.W., Chem. Reviews, 95,

69 (1995)

2. Turchi, C.S. and Ollis. D.F., J. Cat., 122, 178 (1990)

3. Sun, L. and Bolton, J. R., J. Phys. Chem., 100, 4127 (1996)

4. Schwarz P.F., Turro, N.J., Bossmann, S.H., Braun, A.M., Wahab, A.-M., and Dürr, H.,

J. Phys. Chem. B, 101, 7127 (1997)

5. Matthews, R.W., Wat. Res., 24, 653 (1990)

6. Morrison, S.R. and Freund, T., J. Chem. Phys., 47, 1543 (1967)

7. Micka, K. and Gerischer, H., J. Electroanal. Chem., 38, 397 (1972)

8. Matsumura, M., Hiramoto, M., Iehara, T., Tsubomura, H., J. Phys. Chem., 88, 248

(1984)

9. Aguado, M.A., Anderson, M.A., and Hill, C.G., J. Mol. Cat., 89, 165 (1994)

165

10. Ha, H.Y. and Anderson, M.A., J. Env. Eng., 217 (1996)

11. Benderskii, V.A., Zolotovskii, Ya.M., Kogan, Ya.L, Khidekel’, M.L., and Shub,

D.M., Doklady Akademii Nauk SSSR, 222, 606 (1975)

12. Draganiæ, I.G. and Draganiæ, Z.D. Radiation Chemistry of Water, Acad. Press (1971)

13. Fricke, H. and Hart, E.J. in Radiation Dosimetry Vol. II, Acad. Press, p.167 (1966)

14. Davydov, L., Smirniotis, P.G., and Pratsinis, S.E., Ind. Eng. Chem. Res., 38, 1376

(1999)

15. Calverts, J.G. and Pitts, J.N., Photochemistry. J. Wiley & Sons, Inc., 1967

16. Anbar, M. and Neta, P., Int. J. Appl. Rad. Iso., 18, 493 (1967)

17. Riegel, G. and Bolton, J.R., J. Phys. Chem., 99, 4215 (1995)

18 Pruden, A. L., and Ollis, D. F., J. Catal., 82, 404 (1983)

19. Pelizzetti, E. Sol. Ener. Mat. Sol. Cells, 38, 453 (1995)

20. Gomes, W.P., Freund, T., and Morrison, S.R., J. Electrochem. Soc., 115, 818 (1968)

21. Herrmann, J.-M. and Pichat, P., J. Chem .Soc. Farad. I, , 76, 1138 (1980)

22. Cornet, J.F., Dussap, C.G., and Dubertret, G., Biotech. Bioeng., 40, 817 (1992)

23. Kamat, P.V., Chem. Rev., 93, 267 (1993)

24. Heit, G. and Braun, A.M., Wat. Sci. Tech., 35, 25 (1997)

166 Chapter 2.2. Photocatalytic destruction of diethylsulfide in gas phase

INTRODUCTION

Reduced sulfur compounds, such as H2S, CH3SCH3, CS2, OCS, CH3SH,

CH3SSCH3 are significant contaminants of the Earth’s atmosphere. Their oxidation leads to the formation of tropospheric SO2, which eventually becomes H2SO4 [1], one of the main components in acid rains. Reduced organic sulfur compounds are mostly released in anaerobic biological activities. Sewage and industrial wastewater treatment plants are examples of places of their pronounced anthropogenic generation.

They have strong unpleasant odor and very low perceptibility thresholds, for example,

0.4 ppb for H2S, 0.07 ppb for CH3SH, and 3 ppb for CH3SCH3 [2]. These sulfur compounds are also highly toxic.

Reduced sulfur compounds have found their use as chemical warfare agents

(CWAs). The main component of well-known vesicant yperite, or mustard gas, is bis(2-chloroethyl)sulfide. Technical grade Russian yperite besides bis(2- chloroethyl)sulfide contains other vesicant constituents, such as bis(3- chloroisopropyl)sulfide, 2-chloroethyl-1-chloroisopropylsulfide [3]. Additional sulfur- containing impurities include 1,4-dithiane, bis(2-chloropropyl) sulfide, 2-chloroethyl-

2’-chlorobutyl sulfide, bis(2-chloroethyl) disulfide, bis(2-chloroethyl) trisulfide, 1,2- bis(2-chloroethylthio)ethane [4].

Currently applied methods of odor removal are biofiltration and bioscrubbing, activated carbon adsorption, wet chemical scrubbing, and thermal oxidation [2]. A new and highly promising universal method for the destruction of various airborne and dissolved organics is photocatalytic oxidation [5]. A wide set of classes of organic compounds was proved to be mineralized by means of photocatalysis. Suzuki [6]

167 demonstrated the photocatalytic destruction of airborne H2S and CH3SH in a static system over TiO2/honeycomb support. No catalyst deactivation was noted during 20 refill-destruction cycles for H2S. A detailed study on destruction of gaseous H2S in humid air in a flow reactor has been performed by Canela et al. [7]. The main product

2- of the reaction was surface SO4 species. TiO2 deactivation was observed only at concentrations of H2S as high as 600 ppm. Peral and Ollis [8] also did not detect TiO2 deactivation in photocatalytic oxidation of gaseous dimethyl sulfide in concentration

100 mg/m3. However, the conversion in this case was low (10%), and no surface sulfur was discovered on catalyst operated in the reaction for 5000 min. The authors suggested that sulfur was removed as SO2 and/or SO3 during the run.

Reduced sulfur compounds have poor solubility in water. Hence, their reactions in dissolved state are often studied in organic solvents, mostly acetonitrile, and mixtures of water with organic solvents. This approach has the advantage of detection of non-volatile intermediate products. Photooxidation of aryl methyl sulfides in CH3CN resulted mainly in formation of sulfoxides, sulfones, disulfides, and minor quantities of thiols. Sulfoxides and sulfones also were the products of photocatalytic oxidation of heterocyclic sulfur compounds (analogs of anthracenes)

[9]. However, introduction of groups that can be easily oxidized drastically changes the selectivity of the photocatalytic oxidation of sulfides. Aldehydes and acids were predominantly formed from hydroxyalkylaryl sulfide [10].

According to international treaties, production of CWAs is prohibited and existing stockpiles have to be destroyed. One of the methods used by US Army for the destruction of CWAs is incineration. This gives rise to public concern, and the search for other methods was launched. Chemical neutralization by nucleophilic substitution and oxidation is one of most suitable alternative methods for yperite disposal [11].

168 However, the reaction with the suggested nucleophile (monoethanolamine) leads to formation of large quantities of organic products that have to be disposed [3].

Furthermore, the oxidation by the developed agent, oxone, may result in equally toxic products, such as bis(2-chloroethyl)sulfoxide and divinyl sulfone [12].

Photocatalytic oxidation can be a substituent method for yperite annihilation.

First trials of photooxidation of yperite simulants, 2-chloroethyl ethyl sulfide and 2- chloroethyl methyl sulfide, in acetonitrile and water/acetonitrile media were undertaken by Fox et al. [13]. While sulfoxide and sulfone were main products in acetonitrile, a complex mixture of hydrolysis and oxidation products was formed in water/acetonitrile solvent.

Our work is centered on photocatalytic degradation of diethyl sulfide in the gas phase. This compound serves as a simulant of reduced sulfur air contaminants and yperite. We will describe detected gaseous and surface products of the reaction as well as routes of degradation. The extent and chemical nature of the catalyst deactivation is also studied. Several types of TiO2 and different conditions of the reaction were tested to find out the regime with minimal deactivation. Finally, recommendations for further development of photocatalytic method are presented.

EXPERIMENTAL

Photocatalytic experiments were performed using four different samples of titanium dioxide: Hombikat UV 100 (Sachtleben Chemie GmbH), Degussa P25 with

2 2 50 m/g, Degussa P25 with 75 m /g, and homemade TiO2. Diethyl sulfide (Fluka) with assay >98% and deionized water were used. Zero-grade air (Wright Bros.) was utilized as the main component of the reactor feed. Photocatalytic degradation of diethyl sulfide over a thin film of photocatalysts was accomplished in a flow reactor.

The set up used in this study has been described in detail in our earlier work [14]. The

169 reactor was a stainless steel cylindrical vessel illuminated through a Pyrex window placed on the top of the reactor. Before a typical run, 25 mg of photocatalyst was deposited on stainless steel plate of circular area of 9.1 cm2 using aqueous slurry method. Then, the catalyst was dried at 1100C to remove the excess of physisorbed water, and the plate was attached to the bottom of the reactor. Four nozzles at the bottom of the reactor provided a uniform feed flow pattern striking the catalyst. In order to prepare feed stream with certain diethyl sulfide and water vapor concentrations, zero grade air from a cylinder was passed through a water saturator, diluted, and then diethyl sulfide was added using a 7490 Cole-Parmer liquid infusion pump. In one set of experiments, the saturator was filled with 30% solution of H2O2 instead of water. All the gas lines were kept at 50-1000C with the help of heating tapes in order to avoid condensation and/or adsorption of reactants and products. The reactor feed flow rate used in all experiments was 20 cm3/min. The concentration of diethyl sulfide in the feed stream was 368 ppm. Two types of light sources were employed to illuminate the catalyst: two 4 W fluorescent lamps (Wiko, Japan) and

450 W mercury lamp (Hanovia). They provided light intensity of 1.1 and 10 mW/cm2, respectively, at the catalyst level in the reactor. It has been previously demonstrated

[14] that under these conditions there are no external mass transport limitations for acetone oxidation. Prior to starting the photocatalytic experiments, the catalyst in the reactor was kept in flow of reactants overnight in order to reach adsorption equilibrium.

Qualitative analysis of reactor feed and effluent streams was performed using on-line gas chromatograph HP 5890 equipped with a CP Poraplot U column and mass selective detector (HP 5972). The GC oven temperature program included a ramp from 35 to 190 0C at 6 0C/min and constant temperature at 190 0C for 30 min. For the

170 identification of the surface products adsorbed on the catalysts, extraction with isopropanol was performed. The resultant solution was analyzed on HP-5 column employing the temperature ramp 35 to 190 0C at 4 0C/min. Quantitative analysis of gas streams was done by an on-line gas chromatograph HP 6890 equipped with the CP

Poraplot U column, flame ionization (FID) and thermal conductivity detectors (TCD).

The chromatograph was calibrated for CO2, diethyl sulfide, ethanol, acetaldehyde, diethyl disulfide, ethylene, and water. The calibrations were performed by on-line injection of the mixtures of the respective gases with air into the GC.

Additional qualitative information about the product distribution was obtained by passing the effluent of the reactor through the solution of bromine (Fluka) in chloroform (Fisher) for 48 hours. This allowed to test the products of the reaction for the presence of C=C bonds by observing the discoloration of bromine as well as analyzing the solution on the GC/MS for the presence of brominated compounds. Gas chromatograph Shimadzu GC-17A equipped with a capillary column (Supelco) and mass-spectrometer Shimadzu QP-5505A were used for this analysis. The solution was directly injected into the splitless inlet of the GC.

RESULTS AND DISCUSSION

The blank experiments were initially conducted to prove the photocatalysis to be the only route of diethyl sulfide destruction on TiO2. Before switching the lamps of the reactor on, the composition of the reactor effluent was always measured and found to be identical with the composition of feed stream. Therefore, no dark reaction was detected over all catalysts studied at room temperature. The second type of blank experiment was carried out in order to check the possibility of photochemical transformation of diethyl sulfide under the light. No conversion of diethyl sulfide was

171 detected during this test at air humidity 25% (21°C). Hence, only heterogeneous photocatalytic processes cause all the transformation of diethyl sulfide over TiO2 under ultraviolet light in the system applied.

Switching on the lamps of the reactor initially led to high conversion of diethyl sulfide (up to 90%), which gradually decreased but did not fall to zero.

Photocatalytic destruction of diethyl sulfide resulted in the formation of the same gaseous products in different proportions over all catalysts tested in the current work.

Gaseous products comprised ethylene, carbon dioxide, acetaldehyde, ethanol, sulfur dioxide, S-ethyl ethanethioate, and diethyl disulfide. The extract in isopropanol from the surface of TiO2 Hombikat UV 100 operated during 23 hours at humidity 19% and temperature 26°C in diethyl sulfide destruction contained the following products: diethyl sulfoxide and smaller quantities of diethyl disulfide, diethyl sulfone, 2- ethylthioethanol, and diethyl trisulfide. Acid products and salts (such as sulfates),

Figure 2.2.1. Reaction scheme of photocatalytic degradation of diethyl sulfide based on the products detected.

172 which are produced at the final stage of the oxidation of sulfur in diethyl sulfide were not detected because of their extremely low volatility. Only diethyl disulfide was detected in both gaseous and surface products. Other products distributed between gas and surface according to their volatility.

On the basis of the set of detected products it is possible to make some conclusions about the routes of diethyl sulfide destruction (Figure 2.2.1). Such products as ethanol and ethylene clearly testify to the contribution of C-S bond cleavage (corresponding branches in Figure 2.2.1) which did not happen in the dark. The cleavage can be deemed to proceed through hydrolysis to form ethanol and rearrangement to form ethylene. However, the intermediate product of hydrolysis (ethylthiol) was not detected in the products. This may be related to the high lability of this intermediate at the reaction conditions. Very low yield of thiols during the photocatalytic oxidation of sulfides (less than 5%) was reported in [15], while the yield of disulfides was

· considerable (up to 21%). Dimerization of surface CH3CH2S species results in diethyl disulfide. Further C-S bond cleavage in this disulfide and interaction of

· · CH3CH2SS and CH3CH2S surface species can explain the formation of diethyl trisulfide. It should be noted that although diethyl disulfide was present in the initial diethyl sulfide feed, its concentration in the reactor effluent was well beyond its initial concentration, especially in experiments with high light intensity (as will be shown later in the study).

The second obvious route of diethyl sulfide photocatalytic destruction is the oxidation of sulfur atom. Due to their low vapor pressure, the products of this oxidation pathway (sulfoxide and sulfone) resided on the surface of catalyst, where they could be further oxidized leading to the formation of inorganic compounds.

173 The third route is the

40 a 350 oxidation of the - , 2 S 2

300 ) carbon, which resulted 5 H

30 2

CHO, ppm (C H ) S

3 2 5 2 250 CH CHO in the formation of low 3 OH, (C

C H 5 2 4 , ppm H 200 4 2

S and CH C H OH concentrations of H

2 2 5 20 2 )

5 (C H ) S H 2 5 2 2 2 150

and C CH3CH2-S-CO-CH3 in

100 10 the gas phase. Such

50 Concentration of C route was also present

Concentration of (C 0 0 0 500 1000 1500 2000 in the photocatalytic Time, min degradation of the Figure 2.2.2. Exit concentrations of diethyl sulfide and products of its destruction over TiO2 Hombikat UV 100 as oxygen analog of a function of operation time. Humidity <1%, temperature of catalyst 25°C. Lamp power 8 W. diethyl sulfide, diethyl

ether, and led to the 16 350

14 formation of , 2

300 (C H ) S S

2 5 2 2 ) CHO, ppm 5 3 CH CHO 12 3 H ethylacetate [16]. The 2 250 C2H4 C H OH 10 2 5 last route of diethyl OH, (C S and CH 5 2 200 ) (C H ) S , ppm

5 2 5 2 2 H 4

8 2 H H

2 2 150 sulfide photocatalytic 6 and C

100 4 oxidation is the

50 2

Concentration of C oxidation of the b- Concentration of (C 0 0 0 200 400 600 800 1000 1200 1400 carbon. Small Time, min quantities of Figure 2.2.3. Exit concentrations of diethyl sulfide and products of its destruction over TiO2 Hombikat UV 100 as CH3CH2SCH2CH2OH a function of operation time. Humidity 59%, temperature of catalyst 25°C. Lamp power 8 W. were detected on the surface of catalyst, which signifies the presence of this fourth route of degradation.

The proposed above four routes of the photocatalytic destruction of diethyl sulfide

174 represent only initial stages of transformation. During further deeper reaction, these routes can intersect each other. For example, after oxidation of b-carbon the oxidation of sulfur could take place. Finally, the photocatalytic oxidation results in inorganic

2- oxides, H2O, CO2, and SO2. Formation of surface SO4 ions is also hypothesized on the base of work [7] devoted to photocatalytic oxidation of H2S.

As it was noted above, catalyst deactivation takes place during the photocatalytic degradation of diethyl sulfide. From a practical point of view, it is important to carry out the reaction under conditions of minimal deactivation. The knowledge of the influence of operating conditions is also helpful in determining the mechanism of diethyl sulfide photocatalytic destruction. Since the deactivation of

TiO2 in this study was found to occur, it was meaningless to record the dependence of the initial rate versus conditions of the reaction, as this initial rate was true for only a short period of time. Instead, full time dependence of the rate was recorded for each set of conditions. This approach allows the integration of the kinetic curves in order to obtain the quantities of diethyl sulfide converted and products formed during a certain period of time. Since concentrations of products were followed in all cases at least from 25 to 480 min of illumination, this time interval was chosen for integration of concentrations in reactor effluent.

Among the photocatalysts tested, Hombikat UV 100 showed the highest activity in the diethyl sulfide degradation. For this reason the influence of operating conditions was studied for this type of titania. It should be emphasized that among all the products included in the scheme of Figure 2.2.1, only the following gaseous products of the branch “C-S cleavage” were measured reliably: ethanol, acetaldehyde, diethyl disulfide, and ethylene. The concentrations of other possible gaseous products were too low to be measured by either FID or TCD. As one can see from Table 2.2.1,

175 at least 50% of diethyl sulfide consumed was converted into these detected products.

Therefore the “C-S cleavage” branch can be considered to be the main route of diethyl sulfide destruction over TiO2.

Figure 2.2.2 demonstrates the dependence of concentrations of reactants and products in the reactor effluent versus time on stream during photocatalytic reaction over TiO2 Hombikat UV 100 at very low humidity level (less than 1 %) and low light intensity (1.1 mW/cm2). In this and the following figures in this paper, the upper bounds of the graphs correspond to the feed concentration of diethyl sulfide (368 ppm) to facilitate the visual estimate of conversion. Immediately after switching on the lamp of the reactor, the effluent concentration of diethyl sulfide became 393 ppm almost instantaneously, which is above the feed concentration. This is obviously due to thermal and/or photoinduced desorption of diethyl sulfide from the surface of the catalyst. Then, the concentration of (C2H5)2S fell below that in the feed stream, but continually approached it with time and almost reached it in 1500 min of reaction.

The effluent concentrations of diethyl disulfide, ethanol, and acetaldehyde attained maxima, whereas the concentration of C2H4 steadily decreased with time. The peaks in concentration versus time dependencies most probably reflect accumulation of the corresponding products on the surface of TiO2. Indeed, when the experiment is carried out at high humidity (19% and 59%) and low light intensity (1.1 mW/cm2), shown in

Figure 2.2.3, the peak of diethyl disulfide is less pronounced due to lower adsorption.

The peak of ethanol in this case could be related to the accumulation of this compound in water-dissolved state on the surface of TiO2. Total consumption of diethyl sulfide over TiO2 Hombikat UV 100 is about 1.6 times higher in air with considerable humidity levels such as 19 and 59% (Table 2.2.1). The production of ethanol and acetaldehyde is higher at high humidity, whereas the formation of

176 disulfide is about half as

350 much. The generation

150 , 2

300 S of CH did not 2 2 4 ) CHO, ppm 5 3 H 2 250 significantly change (C H ) S 2 5 2 100 OH, (C S and CH 5 2 200 , ppm ) 4 H 5 CH3CHO 2 with the humidity level. H H 2 2 C H 150 2 4 C H OH 2 5 and C The total quantity of (C H ) S 100 2 5 2 2 50 carbonaceous products 50 Concentration of C

Concentration of (C released at high 0 0 0 200 400 600 Time, min humidity is also lower.

Figure 2.2.4. Exit concentrations of diethyl sulfide and At high humidity, less products of its destruction over TiO2 Hombikat UV 100 as a function of operation time. Humidity <1%, temperature quantity of adsorbed of catalyst 40°C. Hg lamp power 450 W. diethyl sulfide is

80 expected. It is easy to 350 , 2 understand the low S

300 2 ) 5

60 H 2 CHO, ppm 3 250 formation of disulfide (C2H5)2S

CH CHO OH, (C 3 5 , ppm

200 4 H

S and CH C H 2 under this condition, H 2 2 4 40 2 )

5

H C2H5OH 2 150

(C2H5)2S2 and C because the bimolecular

100 20 dimerization of surface 50 Concentration of C · CH3CH2S species

Concentration of (C 0 0 0 200 400 600 800 Time, min requires their close

Figure 2.2.5. Exit concentrations of diethyl sulfide and disposition. Increased products of its destruction over TiO2 Hombikat UV 100 as a function of operation time. Humidity 10 %, temperature quantities of of catalyst 400C. Hg lamp power 450 W. acetaldehyde and ethanol testify to the increased catalytic activity in both hydrolysis and oxidation.

177

Table 2.2.1. Cumulative consumption of diethyl sulfide and production of gaseous products during the period of photocatalytic reaction of 25 to 480 min expressed in moles for all compounds.

Catalyst Hombikat UV 100 Degussa P25 TiO2 50 m2/g 75 m2/g

Humidity <1% 19% 59% <1% 10% 59%* 19% 19% 19% G** 1.1 1.1 1.1 11 11 1.1 1.1 1.1 1.1 -5 -5 -5 -5 -5 -5 -5 -5 -5 (C2H5)2S 1.89×10 3.01×10 2.99×10 4.97×10 4.31×10 3.31×10 1.18×10 1.66×10 1.80×10 -5 -5 -5 -5 -5 -6 -5 -5 CH3CHO 1.20×10 - 1.64×10 3.99×10 2.80×10 2.05×10 8.73×10 1.19×10 1.81×10 -6 -6 -5 -6 -6 -6 -6 -6 C2H4 2.95×10 - 3.22×10 1.28×10 9.48×10 3.29×10 1.30×10 1.77×10 2.53×10 -6 -6 -6 -6 -6 -7 -7 C2H5OH 1.03×10 - 1.49×10 1.89×10 1.07×10 1.89×10 5.10×10 7.04×10 - -6 -6 -6 -6 -6 -6 -6 -6 (C2H5)2S2 7.22×10 - 3.17×10 7.81×10 6.87×10 3.04×10 4.57×10 5.90×10 7.93×10 *** -5 -5 -5 -5 -5 -6 -5 -5 Xtot 1.52×10 - 1.37×10 3.51×10 2.61×10 1.59×10 9.84×10 1.31×10 1.82×10

* 75% of initial air feed flow was passed through a saturator filled with 30% aqueous H2O2 ** light intensity in mW/cm2 *** total conversion of (C2H5)2S

Two values of humidity, namely, <1% (Figure 2.2.4) and 10% (Figure 2.2.5),

were used at relatively high light intensity of 11 mW/cm2. The 10-fold increase in the

light intensity resulted only in 1.5 to 2.5-fold growth of diethyl sulfide consumption

(Table 2.2.1). The quantity of disulfide did not increase compared to the case of low

light intensity oxidation in dry conditions. Under high light intensity, the increase in

humidity decreased both diethyl sulfide consumption and production of carbonaceous

products as well as production of H2S. Probably water is not as necessary for high-

light-intensity oxidation as for low-light-intensity one. This may be associated with

increased production of water in photooxidation. The proportion of formed quantities

of ethanol to that of acetaldehyde is markedly lower for high light intensity conditions

compared to low light intensity ones. Thus, in the consecutive photocatalytic reaction

scheme (unbalanced):

178 CH3CH2SCH2CH3 +H2O ¾® CH3CH2OH + [CH3CH2SH] (1)

CH3CH2OH +O2 ¾® CH3CHO +H2O (2) the reaction (2) is relatively faster at high light intensity. This is the reason why the production of significant quantities of CO2 according to reaction (3) was not observed in this work.

O2 O2 CH3CHO ¾¾® CH3COOH ¾¾® CO2 + H2O (3)

As it was demonstrated in [16], ethanol and acetaldehyde are in competition for reaction sites on the surface of TiO2. It was also shown that ethanol adsorbs stronger than acetaldehyde. Hence, reaction (3) is suppressed until the concentration of ethanol is very low. Nevertheless, trace amounts of acetic acid were detected by mass-spectrometry. Homemade TiO2 is a high-crystallinity anatase powder with primary particles size 8-13 nm and specific surface area 120 m2/g, and it was tested for the same reaction for comparison purposes. One would expect faster deactivation of this titania compared to Hombikat UV 100 due to lower surface area available to host surface products. Indeed as Figure 2.2.6 and Table 2.2.1 show, the quantity of diethyl sulfide consumed over homemade TiO2 is lower than over Hombikat UV 100 under the same conditions. The estimates obtained from the kinetic curves demonstrate that higher quantities (in comparison with those over Hombikat UV-100) of gaseous products are formed. Most probably, the surface of this titania is more stable to deactivation due to its higher crystallinity compared to Hombikat UV 100.

The samples of Degussa P25 with even lower specific surface areas (50 and 75 m2/g compared to 120 m2/g for the homemade titania) were also tested for the gas-phase

179

photocatalytic degrada-

tion of diethylsulfide 30 ,

(C H ) S 2 2 5 2 (Figure 2.2.7 and Fig 8, S

300 25 2 ) CH3CHO 5 H C H 2 2 4 respectively). The 20 C 2H5OH OH, (C

(C2H5)2S2 5 200 , ppm 4

H quantities of products 2

15 H 2 CHO concentrations, ppm 3 formed and (CH CH ) S and C 3 2 2 10 100 S and CH 2

) consumed is smaller 5 5 H 2 Concentration of C (C compared to those 0 0 0 200 400 600 800 1000 Time, min described for all other

Figure 2.2.6. Exit concentrations of diethyl sulfide and titanias. The deactivation products of its destruction over homemade TiO2 as a function of operation time. Humidity 19%, temperature behavior of these two of catalyst 250C. Lamp power 8 W. samples of Degussa P25

is quite similar, and the 350 25 , 2 corresponding quantities

300 S 2 ) (C2H5)2S 20 5 H CHO, ppm 2 3 CH CHO 250 3 of the reagent and the

C2H4

15 OH, (C

C H OH 5 200 2 5 , ppm 4 products are directly H S and CH 2 2 (C H ) S H ) 2 2 5 2 2 5 H 2 150 10 proportional to specific and C

100 surface. This corroborates 5 50 Concentration of C the qualitative identity of

Concentration of (C 0 0 0 100 200 300 400 500 these samples: the Time, min difference is only Figure 2.2.7. Exit concentrations of diethyl sulfide and products of its destruction over TiO2 Degussa P25 quantitative. 50 m2/g as a function of operation time. Humidity 19%, temperature of catalyst 25°C. Lamp power 8 W. The addition of hydrogen peroxide into aqueous solutions is known to accelerate the photocatalytic

180 oxidation of dissolved

30 350 organics [17], primarily by , 300 (C H ) S 25 2 scavenging conduction 2 5 2 S 2 ) CH CHO 5

CHO, ppm 3 H 3 2 250 C H 2 4 20 band electrons according

C2H5OH OH, (C

200 5

(C H ) S , ppm 4

S and CH 2 5 2 2 H

2 to the scheme: 15 2 ) H

5 2 H

2 150 - · -

10 and C H2O2 + e ¾® OH + OH 100

5 The resulting hydroxyl 50 Concentration of C radical engages in the Concentration of (C 0 0 0 100 200 300 400 500 600 Time, min attack of the organic

Figure 2.2.8. Exit concentrations of diethyl sulfide and matter present in the products of its destruction over TiO2 Degussa P25 75 m2/g as a function of operation time. Humidity 19%, system. However, the temperature of catalyst 25°C. Lamp power 8 W. influence of hydrogen

20 peroxide on photocatalytic 350 , 2 S

2 oxidation in gas phase has )

300 5 H 15 2

CHO, ppm C H 3 2 4 250 received little attention. (C2H5)2S C2H5OH

CH CHO OH, (C 3 5

(C H ) S , ppm

2 5 2 2 H 4

200 2 The effect of HO vapors H 2 2 S and CH 2

2 10 )

5 H 2 150 and C added to the feed stream

100 5 was measured at humidity

50 Concentration of C 59% and low light

Concentration of (C 0 0 0 200 400 600 800 intensity (1.1 mW/cm2) Time, min over Hombikat UV 100 Figure 2.2.9. Exit concentrations of diethyl sulfide and products of its destruction over TiO2 Hombikat UV 100 (Figure 2.2.9). The as a function of operation time. Humidity 59%, 75% of feed stream passed through 30% aqueous solution of addition of hydrogen H2O2, temperature of catalyst 25°C. Lamp power 8 W. peroxide substantially changed the shape of the product curves (compare with Figure

2.2.3). The quantities of all compounds increased except disulfide. When comparing

181 the product distribution in the absence and in the presence of hydrogen peroxide

(Figure 2.2.3 and Figure 2.2.9, respectively), one can observe the following trends.

First, the steady-state concentration of diethyl sulfide is achieved much slower in the presence of H2O2. Previous studies [18] showed that titania deactivated by the photocatalytic oxidation of organic compounds in gas phase can be reactivated by saturation with hydrogen peroxide and subsequent irradiation in an air stream. This way the oxidation of “heavy” products blocking the active sites of the catalyst can be achieved. In our case, the peroxide may have performed in-situ partial re-activation of the catalyst. As a result, the attainment of the plateau concentration of diethyl sulfide is slower.

The second observable effect of the presence of hydrogen peroxide in the system is the shift in time and the reduction of the absolute values of the maxima of the product concentrations. This takes place for all degradation products (Figure 2.2.3 and Figure 2.2.6). Since the oxidation of diethyl sulfide can only be conceived as a consecutive reaction, the simultaneous lowering of the maxima and the shift of maxima to longer time on stream can be attributed to the decrease of both production rates and disappearance rates of the respective compounds. The production rate of the above products can be lower than in the absence of hydrogen peroxide since it can scavenge an electron, and the resulting OH-ion can scavenge a valence band hole

(unlike in the liquid phase, OH- cannot easily diffuse away from the surface). As a result of this reaction, two hydroxyl radicals are formed. Thus, the attack of the sulfur

+ - by h and O2 decelerates due to the lack of holes and electrons. This leads to the alteration of the kinetic curves observed. On the contrary, the rates of lateral C- oxidation are expected to increase due to increased production of hydroxyl radicals.

The effect of direct attack of hydrogen peroxide on the diethyl disulfide is infeasible

182 since it is a homolytic process. Overall, small quantities of gaseous hydrogen peroxide have beneficial effect on diethyl sulfide conversion and shift the target of secondary oxidation toward the oxidation of carbon atoms. The distribution of gaseous products at 480 min of reaction (Table 2.2.1) also supports the alteration of the reaction mechanism proposed above. The comparison of columns 4 (Hombikat, 59% humidity) and 7 (Hombikat, 59% humidity, equilibrated with 30% H2O2) show the simultaneous increase of the overall conversion and the concentrations of C-oxidation products (ethanol and acetaldehyde) in the latter case. On the contrary, the concentration of the primary S-oxidation product (diethyl disulfide) in gas phase is decreased. The above two phenomena are due to increased production of surface hydroxyl radicals and lack of free surface holes available for S-oxidation.

The last part of this section is devoted to discussion of the mechanism of diethyl sulfide photocatalytic degradation. As it was mentioned above, the photocatalytic degradation of diethyl sulfide does not occur either in the absence of

TiO2 or without ultraviolet light. Therefore, it is true photoassisted catalytic (or photocatalytic) process. The first stage of such reactions is the generation of free charge carriers, electrons and holes:

+ - TiO2 + hn ¾® h + e . (4)

Active consumption of gaseous diethyl sulfide starts only after switching on the ultraviolet illumination of the catalyst. This shows that dark adsorption of diethyl sulfide was low and adsorption is stimulated by irradiation. Low dark adsorption most likely results from strong chemisorption of water on the surface of TiO2. However, under the UV light, desorption of water was detected in earlier studies [19]. This may be explained by the following reaction:

183

HOH

IV - III O-Ti -O + e ¾® O-Ti -O + H2O (5)

The resultant TiIII sites can be restored to highly adsorptive TiIV sites by reaction with holes. Adsorbed diethyl sulfide, however, probably is not desorbed by this mechanism. Instead, it is oxidized by holes to form sulfide radical cation:

+ CH3CH2-S-CH2CH3 CH3CH2-S -CH2CH3 O-TiIV-O +h+ ¾® O-TiIV-O (6)

The formation of sulfur radical cation as intermediate in photosensitized oxidation of sulfides is commonly accepted and proved [20,21]. The redox potential of sulfides is about +1.8 Volt with respect to the normal hydrogen electrode and lower than redox potential of many dyes in excited state. It is also lower than the redox potential of holes in TiO2 (+3 Volt). One could speculate that singlet oxygen also might mediate the oxidation as it is in homogeneous photooxidation [22]. Studies by Davidson and

Pratt [23] and Fox and Abdel-Wahab [24] demonstrated that this is not the case for

TiO2-assisted photooxidation. Further transformations of the radical cation are not so clear. According to the scheme presented in Figure 2.2.1, hydrolysis must take place:

+ CH3CH2-S -CH2CH3 S-CH2CH3 IV IV + O-Ti -O + H2O ¾® O-Ti -O + CH3CH2OH + H , (7)

A competitive reaction is ethylene production:

+ CH3CH2-S -CH2CH3 S-CH2CH3 IV IV + O-Ti -O ¾® O-Ti -O + CH2=CH2 + H (8)

Surface species can also dimerize to form disulfide:

CH3CH2-S S-CH2CH3 IV IV IV IV O-Ti -O-Ti -O ¾® O-Ti -O-Ti -O + CH3CH2SSCH2CH3 .

(9)

184 Formation of trisulfide can be explained by further adsorption of disulfide, its dissociation, and dimerization with surface [CH3CH2S] species. Ethanol formed during hydrolysis reacts further by eqs. (2) and (3).

The second branch of oxidation results in diethyl sulfoxide and sulfone. Electrons from the conduction band of TiO2 react with oxygen forming anion radical:

- - O2 + e ¾® O2 (10)

It is reasonable to suppose the further mechanism to be similar to that in photosensitized oxidation [20]:

CH CH CH CH O-O O-O 3 2 2 3 CH CH CH3CH2 3 2 CH CH + - CH2CH3 2 3 S O-O S S O-TiIV-O-TiIV-O ¾® O-TiIV-O-TiIV-O ¾® O-TiIV-O-TiIV-O

(11)

O-O CH3CH2 CH2CH3 S IV IV O-Ti -O-Ti -O + (C2H5)2S ¾® 2(C2H5)2S=O

(12)

The reaction analogous to (12) but with diethyl sulfoxide instead of diethyl sulfide produces sulfone.

The third and fourth routes of oxidation to produce corresponding a-carbon and b-carbon oxidation products is easy to understand in terms of hydroxyl radical attack. The hydroxyl radical is considered as the major oxidant in aqueous photocatalytic oxidation [25]. However, in our case, production of hydroxyl radicals is relatively small and their impact is also small. This follows from consideration of product distribution. Disulfides are known to react about 50 times faster with ·OH than sulfides [26]. Nevertheless, diethyl di- and trisulfide are among major surface

185 products of diethyl sulfide photodestruction. The next reactions explain small quantities of S-ethyl ethylthioate in the gas phase and still smaller quantities of 2- hydroxyethylthioethane on the catalyst surface:

· · + OH,-H 2 O · + OH C2H5S-CH2-CH3 ¾¾¾¾¾¾® C2H5S-CH -CH3 ¾¾¾¾® C2H5S-CHOH-CH3

(13)

+ + C2H5S-CHOH-CH3 + 2h ¾® C2H5S-C(O)-CH3 + 2H

(14)

· · + OH,-H2O + OH C2H5S-CH2-CH3 ¾¾¾¾¾¾® C2H5S-CH2-CH2 ¾¾¾¾® C2H5S-CH2-CH2OH

(15)

This agrees with the fact that hydrogen abstraction from b-carbon is a minor channel compared to that from a-carbon [27].

CONCLUSIONS

This study has revealed deactivation of TiO2 catalyst during photocatalytic degradation of airborne diethyl sulfide in the concentration of hundreds of ppm. The distribution of products strongly suggests that the two main pathways of the degradation are the oxidation of sulfur and carbon atoms and hydrolysis of the C-S bond. The products of the complete oxidation (SO2 and CO2) were detected in very small quantities, most of the products correspond to C-S bond cleavage and partial oxidation. Further efforts should be devoted to make photocatalytic oxidation suitable for the complete mineralization of reduced sulfur compounds. The decrease in sulfide concentration can help to avoid deactivation, as it was for H2S photocatalytic oxidation [7]. Modification of the photocatalyst surface is another way to reduce the rate of deactivation. The co-feeding of hydrogen peroxide enhanced the conversion of diethyl sulfide.

186 ACKNOWLEDGEMENTS

This project has been funded in part by the National Research Council under the

Collaboration in Basic Science and Engineering Program (COBASE), which provided support for the first author (A.V.V.). The content of this publication does not necessarily reflect the views or policies of the NRC, nor does mention of trade names, commercial products, or organizations imply endorsement by the NRC. The authors are indebted to the Science for Peace Programme of NATO which initiated the collaboration between these two research groups through the award SFP-974209. This research is also based upon work supported in part by the U.S. Army Research Office

(Young Investigator Award to P.G.S.) under grant number 40414/CH/YIP.

REFERENCES

[1] C.Wilson, D.M.Hirst, Prog. Reaction kinetics, 21 (1996) 69.

[2] B.Mills, Filtration and Separation, 2 (1995) 147.

[3] I.N.Luganskii, V.V.Sheluchenko, I.N.Krotovich, V.V.Chebotaev, V.A.Kholodeva, O.A.Papkova, Zhurnal Ross.Khim.Ob-va im.D.I.Mendeleeva, 38 (1994) 34.

[4] B.A.Trofimov, N.K.Gusarova, I.A.Dorofeev, M.Ya.Khil’ko, E.B.Belousov, O.B.Bannikova, V.K.Novikov, I.N.Krotovich, V.V.Gormai, V.N.Fokin, N.F.Dolgov, A.V.Shantrokha, A.N.Denisenya, V.I.Kholstov, Russian Journal of Applied Chemistry, 67 (1994) 148.

[5] Hoffmann, M.R., Martin, S.T., Choi, W., and Bahnemann, D.W., Chem. Reviews, 95 (1995) 69

[6] K.Suzuki, in: D.F.Ollis, H.Al-Ekabi (Eds.), Photocatalytic Purification and Treatment of Water and Air, Elsevier, 1993, p.421.

[7] M.C.Canela, R.M.Alberici, W.F.Jardim, J.Photochem.Photobiol. A: Chem., 112 (1998) 73.

[8] J.Peral, D.F.Ollis, J.Molec.Catal. A:Chem., 115 (1997) 347.

187

[9] A.A.Abdel-Wahab, A.E.M.Gaber, J.Photochem.Photobiol. A: Chem., 114 (1998) 213.

[10] M.A.Fox, A.A.Abdel-Wahab, J.Catal., 126 (1990) 693.

[11] Y.-C.Yang, Chemistry & Industry, 9 (1995) 334.

[12] R.I.Tilley, D.R.Leslie, Aust.J.Chem., 48 (1995) 1781.

[13] M.A.Fox, Y.-S.Kim, A.A.Abdel-Wahab, Catalysis Lett., 5 (1990) 369.

[14] A.V.Vorontsov, E.N.Savinov, P.G.Smirniotis, Chem. Eng. Sci., 55 (2000) 5089.

[15] N.Somasundaram, C.Srinivasan, J.Photochem.Photobiol. A: Chem., 115 (1998) 169.

[16] A.V.Vorontsov, E.N.Savinov, G.B.Barannik , V.N.Troitsky and V.N.Parmon, Catalysis Today, 39 (1997) 207.

[17] A.Sclafani, L.Palmisano, E.Davi, New J.Chem., 14 (1990) 265.

[18] R.M.Alberici and W.F.Jardim, App. Catal. B: Env., 14 (1997) 55

[19] G.Munuera, V.Rives-Arnau, A.Saucedo, J.Chem.Soc., Faraday Trans.1, 75 (1979) 736.

[20] N.Soggiu, H.Cardy, J.L.H.Jiwan, I.Leray, J.P.Soumillion, S.Lacombe, J.Photochem.Photobiol. A: Chem., 124 (1999) 1.

[21] E.Baciocchi, C.Crescenzi, O.Lanzalunga, Tetrahedron, 53 (1997) 4469.

[22] S.M.Bonesi, M.Freccero, A.Albini, J.Phys.Org.Chem., 12 (1999) 703.

[23] R.S.Davidson, J.E.Pratt, Tetrahedron Lett., 52 (1983) 5903.

[24] M.A.Fox, A.A.Abdel-Wahab, Tetrahedron Lett., 31 (1990) 4533.

[25] C.S.Turchi, D.F.Ollis, J.Catal., 122 (1990) 178.

[26] R.A.Cox, D.Sheppard, Nature (London), 284 (1980) 330.

[27] A.J.Hynes, P.H.Wine, D.H.Semmes, J.Phys.Chem., 90 (1986) 4148.

188 Chapter 2.3. Radical generation during sonophotocatalytic destruction of VOCs on

zeolite-supported titania

INTRODUCTION

Semiconductor-loaded zeolites and mesoporous molecular sieves (such as MCM-

41) have recently drawn increased attention as potential photocatalysts due to their unique pore structure and adsorption properties. Titania has proved to be the most active photocatalytic semiconductor as it allows for the complete degradation of pollutants under ultraviolet irradiation. Xu and Langford [1, 2], and Cho et al [3] studied the photodegradation of several organic compounds on various TiO2 – loaded zeolites as well

as molecular sieves of MCM-41 type. The above researchers summarized the advantages of using zeolite or mesoporous support for titania in photocatalysis. They are as follows:

formation of ultrafine titania particles during sol-gel deposition; increased adsorption, especially for non-polar compounds; acidity which allows to enhance electron- abstraction; less UV light scattering as the main component of zeolites is silica.

The use of ultrasound for wastewater treatment has also been explored [4]. The action of ultrasound allows creating microbubbles in water at high temperature and pressure, leading to localized transient supercritical conditions. This leads to the production of active radicals (H· and ·OH), that take part in the degradation of ambient organic matter. The presence of solid particles in the aqueous system was found to enhance the production of microbubbles under ultrasound [5]. Therefore, the combination of these two ways of treatment may produce a synergistic effect [6, 7]. There are only a few reports on the sonophotochemical degradation of organic pollutants in water. For

189 example, Shirgaonkar and Pandit [8] studied the photodegradation of 2,4,6- trichlorophenol in irradiated TiO2 slurries with ultrasound and found enhancement of the quantum yield. Theron et al [9] studied the degradation of phenyltrifluoromethyl-ketone in water by simultaneous use of TiO2 photocatalysis and ultrasound. It was found that the presence of ultrasound significantly reduces the formation of toxic stable by-products during photocatalysis.

The combination of the effect of zeolites and TiO2 with ultrasound in the photocatalytic destruction of salicylic acid was studied in our previous work [15].

However, many effects of zeolite supports as well as the role of sonolysis have not been fully understood for this system. The present study attempts to combine the advantages of all three ways of pollution abatement: photocatalysis, sonolysis, and use of zeolite materials. More specifically, ultrasound resistant zeolitic and mesoporous supports have been isolated, loaded with titania, characterized, and then used for the degradation of salicylic acid, a common organic pollutant. Single stage oxidation reactions are utilized to assess the relative enhancement of radical generation by each particular system.

EXPERIMENTAL

Synthesis and Characterization

The USY zeolite of Si/Al=2.8 was obtained from PQ Corporation. Zeolite beta was synthesized using tetraethylammonium hydroxide as the template with the nominal

Si/Al ratio of 15. Siliceous and Al – substituted mesoporous molecular sieves were synthesized according to the recipes presented by Sayari et al [10] and Corma et al [11], respectively. In the case of MCM-41, the resulting catalyst was dispersed in ~100 ml of

190 isopropanol, and titanium isopropoxide (TIPO) was added to achieve 25 wt % loading.

The system was dried while stirring at ambient temperature. It was then placed in the oven to dry at 100°C for 1 hour. Water was added to hydrolyze the TIPO, and the system was stirred for 1 hour. The solids were then filtered off, washed, and dried for 1 hour at

100°C. They were then transferred into a boat-like crucible and calcined at 450°C for 3 hours with a temperature ramp of 2 °C/min. In the case of doped USY (Si/Al=2.6) and b, the deposition of titania took place in aqueous system using (NH4)2(TiO)(C2O4)2 as the source of titania as the estimated kinetic diameter of TIPO is larger than the pore opening of USY. The catalyst was dispersed in ~100 ml of water, the titanium source was added to achieve 25 wt % loading. The system was dried at ambient temperature to preserve the structure of USY. It was then placed in the furnace and calcined at 450°C for 10 hours with a temperature ramp of 5 deg/min.

All catalysts were characterized by Nicolet powder X-ray diffractometer equipped with a CuKa source to assess their crystallinity. MCM-41 powders were run from 2 to 7 degrees (2 q) to assess the crystallinity of the matrix and from 20 to 50 to assess the crystallinity of the TiO2 loading. USY-based samples were run from 20 to 50 degrees corresponding to both loading and support. Furthermore, the powders were characterized by UV-Vis Shimadzu 2501PC with an integrating sphere attachment ISR1200 for their diffuse reflectance in the range of wavelengths of 200 to 800 nm. BET and pore size distribution studies were also conducted (using Micromeritics apparatus) to characterize the synthesized MCM-41, b-zeolite (Si/Al=15) and USY (Si/Al=2.8) samples. Typical

XRD patterns of the powders synthesized and spent can be found in their corresponding figures.

191 A test to determine the hydrophobicity of each zeolitic support was performed in

order to assess their affinities toward specific groups of organic compounds (adapted from [12]). It involved competitive adsorption of toluene from aqueous solution. In a typical experiment 1 mM of toluene was dissolved in 100 mL of water, and the resulting solution was allowed to equilibrate with air for 5 h. After this period, approximately 0.2 g of the pre-ground support was introduced. In a timely fashion samples of the liquid were withdrawn, filtered, and analyzed spectrophotometrically at 204 nm for the remaining concentration of toluene in the solution.

Photocatalytic Experiments

The photocatalytic testing included the degradation of salicylic acid (reagent

grade, Fisher), which was performed in a batch multitubular pyrex reactor using seven 4

W UV-Visible mercury fluorescent lamps. Before the reaction the slurry (0.25 g/L of

solids, 1 mM of salicylic acid) was ultrasonicated for 10 min in the ultrasonic bath in

order to assure the breakage of catalyst aggregates. The suspended catalyst in aqueous

system was oxygenated at 0.5 L/min and irradiated during each reaction run. Certain runs

included ultrasonication of the reaction suspension, which was performed by the VWR-

600 ultrasonic processor with an extended probe. The concentration of salicylic acid in

the reactor was monitored by UV-Vis spectrophotometry at the wavelength of 296 nm

(Shimadzu 2501PC). The concentration of formic acid in the aqueous solution (initial concentration 10 mM, Fisher) was measured by an electric conductivity meter (VWR

Scientific, equipped with a golden cell, cell constant 10.25 cm-1). A number of experiments was performed with a cupric-ferrous aqueous dosimeter [13], which

192 consisted of 1 mM FeSO4, 5 mM CuSO4, and 0.8 M H2SO4 (all reagent grade, Fisher).

The concentration of the Fe3+ formed was monitored spectrophotometrically at the wavelength of 304 nm.

RESULTS AND DISCUSSION

The BET surface areas of fresh and used catalysts are depicted in Table 2.3.1. The surface area of TiO2 Hombikat (HK) and TiO2/MCM-41 did not change during the reaction. This is because of the weak adsorption of salicylic acid on HK due to its small primary particle size and on titania-loaded MCM-41 due to its silica nature. The decreasing surface area of 25%TiO2/USY catalyst after the reaction may indicate the partial deactivation of the catalyst with time. This was proved by the activity studies of this catalyst, which are presented below. However, the crystallinity of 25%TiO2/USY catalyst has not changed, as can be observed from Figure 2.3.1a. On the contrary, the surface area of 25%TiO2/MCM-41 remained almost unchanged throughout the reaction.

However, the rate of salicylic acid degradation over this catalyst decays significantly

(Figure 2.3.3). As shown by independent TGA experiments, the pores of the USY get blocked by the unreacted salicylic acid under reaction conditions, which does not allow the fresh reactant to access the active sites. Furthermore, the spent catalyst was filtered and dried at 150°C, the salicylic acid was removed, and the surface area remained same.

This was also proved by crystallinity of this catalyst that was analyzed by XRD. It should be also noted that USY lost about one third of its surface area upon loading TiO2, while

193 Table 2.3.1. BET surface areas of the catalysts before and after the reaction BET Surface area before BET Surface area after Catalyst reaction, m2/g reaction, m2/g

TiO2 Hombikat 313 312 USY zeolite 659 not applicable 25%TiO2/USY 415 379 MCM-41 941 not applicable 25%TiO2/MCM-41 669 668 -zeolite 525 not applicable 25%TiO2/-zeolite 231 258 b-zeolite lost one half. This leads to a conclusion that the titania precursor dispersed better inside USY than in b-zeolite. The increased hydrophilicity (see below) of USY also contributes to better dispersion of an inorganic agent. As a result of better dispersion of the titania precursor, the pores of USY are not blocked by titania, and this creates an opportunity for salicylic acid and the intermediates of its degradation to occlude.

The BET surface area of 25%TiO2/-zeolite after used for reaction is higher than that before the reaction, possibly because of wash away of titania from the titania doped

-zeolite. Since -zeolite is a relatively hydrophobic material, the inorganic titania precursor could not distribute well in the pores of this material, thus allowing for the presence of the islands of titania on the surface. The latter can also be justified by the fact that while performing the reaction in ultrasound, the titania was liberated from the catalyst, it was also explained by the XRD results in the coming paragraphs.

XRD patterns of 25%TiO2/USY and 25%TiO2/MCM-41 catalysts have been shown in Figure 2.3.1a and 1b, respectively. As shown in Figure 2.3.1a the crystallinity of 25%TiO2/USY catalyst before and after photocatalytic degradation of salicylic acid by combination of UV-light and ultrasound obeys a similar pattern. In the case of

25%TiO2/MCM-41 (see Figure 2.3.1b), the crystallinity of these materials was also

194 similar before and after the reaction. These XRD results indicate that the 25%TiO2/USY

and

a b

Spent 25%TiO2/USY zeolite

Spent 25%TiO2/MCM-41 Intensity (a.u.) Intiensity (a.u.) 25%TiO2/USY zeolite

25%TiO2/MCM-41

20 25 30 35 40 45 50 2 3 4 5 6 7 2q 2q

Figure 2.3.1. X-ray diffraction patterns for the catalysts before and after reaction: (a) 25%TiO2/USY; (b) 25%TiO2/MCM-41

25%TiO2/MCM-41 catalysts are stable under ultrasound since the structures of these catalysts did not change after 180 min of reaction. However, the XRD patterns of the spent 25%TiO2/ -zeolite and the fresh 25%TiO2/-zeolite showed the opposite trend.

More specifically, there was a loss of crystallinity of zeolite b under ultrasound.

Furthermore, -zeolite was not active for the degradation of salicylic acid, because it is the more hydrophobic material in comparison with USY (results shown below). Because of such hydrophobicity, the inorganic titania precursor used during the catalyst synthesis

may not have been admitted into the pores of this zeolite. On the other hand, an organic

precursor which would have more propensity to a hydrophobic material can have a

molecule insufficiently small to penetrate the pores of zeolite b.

A study was also undertaken to ascertain the stability of MCM-41 structure under

ultrasound. One can observe from Figure 2.3.2 that the ability of the mesoporous

195 structure to withstand ultrasonic waves monotonically decreases with the increase of the

power input per volume. This should be taken into account when designing

sonophotocatalytic reactors. The reactor utilized in the present study allowed for the

complete preservation of the structure of the TiO2-loaded MCM-41 photocatalysts.

The degradation of salicylic

3000 acid over 25%TiO2/USY

zeolite with UV light and

2000 combination of UV light and

ultrasound has been shown in

1000 Figure 2.3.3a. In the former case degradation of salicylic Intensity of XRD peak

acid was considerably lower 0 0.0 0.2 0.4 0.6 0.8 1.0 Power/Volume, W/L than that in the latter case. In

Figure 2.3.2. Intensity of the main XRD peak of the first experiment with MCM-41 (2q=2°) versus ultrasonic power input to titania-loaded USY (-o-) the volume ratio during ultrasonication: ultrasound – 100W, MCM-41 concentration – 1 g/L degradation was prominently

higher than the second experiment (-D-). This is because in the second experiment (-D-) the solution containing salicylic acid and the catalyst was kept under stirring for 1 hour before introducing into the ultrasonic reactor, whereas in the first experiment (-o-) the light and ultrasound were turned on immediately after the slurry was prepared. This was done in order to discern whether the decreasing concentration of salicylic acid was due to the adsorption or reaction on the catalyst surface. The rate of decreasing concentration of salicylic acid in the first 5 minutes is very high, further the rate decreasing continuously

196 with time. Moreover, the sharp decrease in the concentration of salicylic acid in the first 5

minutes was not due to the adsorption of salicylic acid on the surface of 25%TiO2/USY

zeolite, as we realized through the second experiment. Since the initial and terminal rates

of degradation of salicylic acid are within the experimental error, one can conclude that

the disappearance of the reactant was not due to the adsorption or photoadsoprtion on

catalyst surface.

100 a 100 b 98 98 UV 96 96 UV 94 94

92 92 UV + US UV + US 90 90 % of salicylic acid % of salicylic acid 88 88

86 86

84 84 0 30 60 90 120 150 180 0 30 60 90 120 150 180

Time (minutes) Time (minutes)

Figure 2.3.3. Degradation of salicylic acid by UV light and combination of UV light and ultrasound on: (a) 25%TiO2/USY; (b) 25%TiO2/MCM-41. Initial concentration of salicylic acid – 1mM, UV – 28W, Ultrasound – 100W

As shown in Figure 2.3.3b the degradation of salicylic acid over 25%TiO2/MCM-

41 with UV-light was less active than with combination of UV light and ultrasound. The rate of decreasing concentration of salicylic acid was very high in the first 5 minutes for both cases, but it is more predominant in the latter one. This clearly shows that the effect of ultrasound was the basis for faster decreasing concentration of salicylic acid. This is due to the propagation of a pressure wave in aqueous solution, which leads to the

197 formation of cavitation bubbles; a prerequisite for these bubbles is the presence of

dissolved gas and suspended solid [5]. The collapse of these bubbles spawn localized

extreme conditions such as very high temperatures and pressures, which in turn lead to

  the dissociation of H2O and the production of radical species such as OH, HOO etc in the presence of TiO2

containing catalysts.

Therefore, this leads to the 100 enhancement of the rate of b UV + US (25%TiO2/ ) 96 reaction under ultrasound UV (HK)

condition on supported TiO2 92 catalysts.

88 UV + US (HK) The photocatalytic % of salicylic acid

degradation of salicylic acid 84 0 30 60 90 120 150 180 over 25%TiO2/-zeolite with Time (minutes) combination of UV light and Figure 2.3.4. Degradation of salicylic acid on ultrasound is shown in Figure 25%TiO2/b and neat titania (Hombikat) Initial concentration of salicylic acid – 1mM, UV – 28W, 2.3.4. The concentration of Ultrasound – 100W salicylic acid remains the same after 3 hours reaction time. It may be explained that the structure of -zeolite supported TiO2 somewhat collapsed under the ultrasound, which was also proven by the

XRD patterns of 25%TiO2/-zeolite after the reaction as well as ICP studies. Therefore,

-zeolite is not a viable support for the photocatalytic degradation of salicylic acid under

198

Table 2.3.2: Initial reaction rates (mol/gTiO2 min) of salicylic acid under UV-light (28 W) and combination of UV (28 W) and ultrasound (100 W) Catalyst UV+ultrasound UV only

25%TiO2/-zeolite 0 0 TiO2 Hombikat 0.085 0.053 25%TiO2/MCM-41 0.58 0.27 25%TiO2/USY 0.84 0.056 ultrasound Moreover, zeolite b is known as more hydrophobic among the ones tested

[12], which will also be explored further in this study. As a result, the inorganic titania precursor may not have been distributed well inside the zeolite pores. Moreover, the hydrophobicity of zeolite b has a negative effect on the adsorption of salicylic acid, leading to negligible degradation under the present reaction conditions. The photocatalytic degradation of salicylic acid over TiO2 Hombikat with UV light and with the combination of UV light and ultrasound has also been presented in Figure 2.3.4. The concentration of salicylic acid decreased in both cases, this was more predominant in the case of combination of UV light and ultrasound. Theron, et al. [9] reported that the simultaneous use of photocatalytic and ultrasound treatment on any organic pollutant give an increasing rate constant value. In our study Hombikat TiO2 was more active for the degradation of organics by UV light and ultrasound combined. The rate of concentration decay of salicylic acid remains unchanged even after 3 hours of reaction time. Thus, Hombikat TiO2 is a stable and active catalyst for the degradation of salicylic acid by combination of UV light and ultrasound.

The initial reaction rate of salicylic acid with UV light and combination of UV light and ultrasound (adjusted for TiO2 content) over different catalysts are presented in

Table 2.3.2. As can be noted from this table, the initial rate of reaction of salicylic acid

199 over 25%TiO2/- zeolite is zero when used with UV light and combination of UV light and ultrasound. Although the pore sizes of the two zeolites studied in the present work are comparable [14], the more hydrophobic nature of -zeolite in comparison with USY

does not allow the efficient adsorption of the polar salicylic acid molecule on its surface.

The latter process always precedes the photocatalytic reaction. The initial reaction rate is

the highest for the MCM-41-based photocatalyst when only UV is employed and for the

USY based photocatalyst when UV in combination with ultrasound is employed. The most significant enhancement by ultrasound (more than 10-fold) is observed for

TiO2/USY. This can be attributed

to the facilitated adsorption of 1.0 Beta USY salicylic acid in the pores of 0.9 USY which otherwise does not

0.8 , C/Co 3 take place because of the CH 5 H

6 0.7 C comparable sizes of the

0.6 adsorbate molecule and the pores

0.5 of the adsorbent. The main 0 20 40 60 80 100 120 140 160 180

Time, min difference between zeolite b and Figure 2.3.5. Isotherms of adsorption of toluene from toluene/water solution by the zeolite supports USY structures is the presence of employed in the present study: initial concentration of toluene – 0.5 mM, concentration of zeolite – 2 supercages in the latter zeolite. g/L, temperature – 20ºC Apparently, their presence is

advantageous since the titania presursor can accumulate there and form TiO2 clusters

upon calcination, thus leaving the pore mouths free of titania. Such arrangement allows

for facilitated diffusion of the reactants to the titania cluster.

200 As mentioned in the experimental section of the paper, hydrophobicity tests were

performed in order to ascertain the affinity of different support to adsorb different organic

molecules as well as titania precursors. The method was adapted from previous work [15] and it included selective adsorption of toluene from its aqueous solution. The shapes of the adsorption isotherms for various supports are shown in Figure 2.3.5. One can observe that the isotherms are quite different. In particular, b-zeolite (Si/Al=15) adsorbs a

relatively large amount of toluene during the first 0.5 min followed by slow adsorption

thereafter, thus reaching the adsorption equilibrium very rapidly. On the contrary, the

adsorption by USY is monotonic. This attests to higher hydrophobicity hypothesis

proposed by Xu and Langford [2] to explain the low activity of TiO2 supported on zeolite

b. Moreover, [12] found negligible hydrophobicity of zeolite USY. There are several

consequences of the higher hydrophobicity of zeolite b. The inorganic titania precursor

may have difficulty adsorbing on a more hydrophobic material. At the same time an

organic precursor would not penetrate the pores of this zeolite. Moreover, a hydrophilic

organic solute (such as salicylic acid) will have less accessibility to the loaded TiO2. As a

result, the lowest reaction rate was observed on TiO2/b for the photodegradation of

salicylic acid. On the contrary, when USY was used as a support (having similar pore sizes) significant degradation was observed.

A number of auxiliary reactions was used to explore the relative extent of different radicals’ generation by zeolite-supported TiO2 and mass transfer limitations induced by the microporous supports. They involved small molecules and ions, thus excluding the mass transfer limitation effects. The first one allows to compare the action of hydroxyl radicals [7]:

201 · · HCOOH + OH ® COOH + H2O (1.1)

· · - + O2 + COOH ® O2 + CO2 + H (1.2)

The second reaction is widely used in radiation chemistry to indirectly determine

radiation doses (radiation dosimetry). It is based on the measurement of the radicals

produced in water by radiation. One of them can track the presence of hydroxyl radicals

and hydrogen peroxide produced in solution. It obeys the following scheme [13]:

- Fe2+ + ·OH ® Fe3+ + OH (2.1)

2+ · + + Cu + HO2 ® Cu + H + O2 (2.2)

Fe3+ + Cu+ ® Fe2+ + Cu2+ (2.3)

2+ 3+ - 2Fe + H2O2 ® 2Fe + 2OH (2.4)

100 100

95 95

US+TiO (Hombikat) 2 90 90 UV+25% TiO 2/MCM-41

UV+US+25% TiO 2/MCM-41

85 UV+TiO (Hombikat) 2 85 UV+US+TiO (Hombikat) 2 80 80 UV+25% TiO 2/USY UV+US+25% TiO /USY 2 % concentration of formic acid 75 75 UV+25% TiO /b-

% concentration of formic acid 2 zeolite UV+US+25% TiO /b- 2 zeolite 70 70 0 30 60 90 120 150 180 0 30 60 90 120 150 180 Time (minutes) Time (minutes)

Figure 2.3.6. Time course of the photo- Figure 2.3.7. Time course of the photo- and and sonophotodegradation of formic acid sonophotodegradation of formic acid on b- on neat titania and MCM-41 supported zeolite and USY supported titania: initial titania: initial concentration of HCOOH – concentration of HCOOH – 10 mM, catalyst 10 mM, concentration of HK – 0.25 g/L, concentration – 1 g/L, ultrasonic power input concentration of 25%TiO2/MCM-41 – 1 – 100 W, UV power input – 28 W g/L, ultrasonic power input – 100 W, UV power input – 28 W

202 The first reaction has been shown to the action of hydroxyl radicals [7] since the

superoxide, peroxyl radicals and hydrogen peroxide (resulting from the surface reactions

of conduction band electrons) do not act upon formic acid [16]. Therefore, one can

estimate the generation of hydroxyl radicals by supported or unsupported titania by

conducting the photo- and/or sonoassisted degradation of formic acid. Figures 2.3.6 and 7 compare the performance of various photocatalysts for the photodegradation of HCOOH.

As with salicylic acid (Figures 2.3.3-4) the rates of degradation are enhanced by ultrasound for all catalysts employed. Some features are different, however, which are due to distinct features of the formic acid molecule. As mentioned above, since it reacts with hydroxyls only, the stabilization of other reactive oxygen species (as suggested by

Xu and Langford [1,2]), which may be provided by a zeolitic support, is not applicable to this process. The decrease of the stability of intermediates suggested by the above authors is also out of the question, since many researchers could not detect any intermediates of the HCOOH photodegradation [17]. Furthermore, the internal mass transfer limitations that affected the performance of the photocatalysts in the case of salicylic acid are not applicable to formic acid since its kinetic diameter is in the order of 2 Å, which is much smaller than the pore diameters of the zeolite supports employed. As a result of these factors, we observe smaller degree of enhancement by ultrasound for the photodegradation of formic acid (Figures 2.3.6 and 7). Moreover, for the supported catalysts the degree of non-linearity of the kinetic curves is much smaller than that in the case of salicylic acid (Figures 2.3.2 and 3); the behavior exhibited by these catalysts is now close to linearity. This is due to the lesser effect of mass transfer limitations. The photodegradation of formic acid on titania-loaded MCM-41 is almost linear in the

203 absence of ultrasound (Figure 2.3.6) and becomes close to half-order when the ultrasound

is used. This can be attributed to the effect of microstreaming observed in ultrasonic

treatment of liquids [18]. As mentioned above, the molecule of HCOOH is small. At the same time the pores of neat MCM-41 are in the order of 4 nm, and the ultrasound may provide the opportunity for a “microstream” of solution to penetrate the pores of MCM-

41 at the beginning of the reaction (up to 30 min), thus giving rise to the enhancement observed. This may not be possible in the absence of ultrasound (linear behavior) since the mass transport is primarily due to adsorption forces. At the later stages, however, this effect diminishes because of the gradual decomposition of the MCM-41 structure. The latter phenomenon was detected by XRD analysis (not shown) and it took place only in acidic medium created by formic acid. It is probably due to the osmotic pressure caused by the enhanced ionic strength of the formic acid solution in comparison with that of the salicylic acid solution.

One more significant difference of the photo and sonophotoassisted degradation of formic acid is the behavior of titania-loaded b-zeolite. It was found inactive for the photodegradation of salicylic acid (Figure 2.3.2), but it outperforms all other supported catalysts for the degradation of formic acid. This may be due to the presence of cavities inside zeolite Y [14]. They may cause a different distribution of titania precursor on the support with the majority of it trapped in the cavities. As a result, the accessibility of

TiO2 by a much smaller molecule is better when it is supported on zeolite b, which leads to increased activity. Lastly, it is worth noting that the use of ultrasound alone with the presence of titania in the dark has produced no discernible effect on the concentration of formic acid. Apparently, this may be the result of the absence of hydroxyl radicals in the

204 sonicated solution to perform reaction 1 or the enhanced recombination of these radicals

with the solid and with each other. In order to test the formation of hydrogen peroxide as

a result of the recombination of hydroxyl radicals the same experiment was performed in

the presence of ammonium oxalate oxotitanate. Titanium ions in solution react with

hydrogen peroxide to form yellow-colored peroxocomplexes [19]:

2+ + TiO + H2O2 ® 2 H + TiO3 (3)

Since no yellowing of the solution was observed even after 3 hours of sonication, which would indicate the presence of hydrogen peroxide. One can conclude that the self- recombination of hydroxyl radicals does not take place in the sonicated slurry. Therefore, the absence of activity of ultrasound to destroy formic acid (contrary to salicylic acid) is

6 6 -4 UV+US+TiO (Hombikat) -4 UV+25% TiO /b- 2 2 zeolite UV+TiO (Hombikat) UV+US+25% TiO /b- 5 2 5 2 zeolite UV+25% TiO /MCM-41 2 UV+25% TiO 2 /USY UV+US+25% TiO /MCM-41 UV+US+25% TiO /USY 4 2 4 2 US+TiO 2 (Hombikat) generated X 10 +3 generated X 10

3 +3 3

2 2

1 1 Concentration of Fe 0 0 Concentration of Fe

0 30 60 90 120 150 180 0 30 60 90 120 150 180 Time (minutes) Time (minutes)

Figure 2.3.8. Time course of the photo- and Figure 2.3.9. Time course of the photo- and sonophotooxidation of Fricke solution on sonophotooxidation of Fricke solution on b- neat titania and MCM-41 supported titania: zeolite and USY supported titania: initial initial concentration of HCOOH – 10 mM, concentration of HCOOH – 10 mM, catalyst concentration of HK – 0.25 g/L, concentration – 1 g/L, ultrasonic power concentration of 25%TiO2/MCM-41 – 1 g/L, input – 100 W, UV power input – 28 W ultrasonic power input – 100 W, UV power input – 28 W

205 due to the recombination or radicals with the solid, which is abundant in the

photocatalytic slurry. It should be noted that the mobility of the hydroxyl radicals

produced by photogenerated holes is very limited due to the fact that titania is

· intrinsically hydroxylated [20]. On the contrary, the radicals of peroxide nature (HO2 ,

· O2) are only adsorbed on the surface of titania, and the presence of supports may

increase their lifetime and thus increase the likelihood of their reactions with the aqueous

organic pollutant.

The other auxiliary reaction (2.1-2.5) will be used to test the above hypothesis. It

is believed to account for hydrogen peroxide and hydroxyl radicals only [13], since the

radicals produced by the conduction band electrons are engaged into a redox cycle with

the copper ions as well as the further transformation to H2O2. As one can observe from

Figures 2.3.8-9, ultrasound has a deleterious effect on the photooxidation of iron ions by zeolite-supported titania. This may be due to the acidic environment and presence of transition metal ions, which allow to partially ion-exchange the aluminum ions in the

framework of the zeolites. This leads to the ease of framework destruction under the

action of ultrasonic waves, as shown by the XRD patterns of the spent catalysts. As to the

MCM-41 supported TiO2 the presence of ultrasound had no effect on the rate of

photooxidation of iron ions, contrary to neat titania (Figure 2.3.8). It should be noted that

ion exchange into siliceous MCM-41 matrix is very unlikely at these conditions.

Therefore, the lack of enhancement by ultrasound is due to the lack of adsorption of iron ions onto mainly hydrophobic siliceous MCM-41 support.

Conclusively, the use of ultrasound allows to enhance the rates of most reactions of the photodegradation of organic compounds on supported and unsupported titania. The

206 role of support involves preferential adsorption of organic molecules from solution. No

enhanced hydroxyl radical generation due to the presence of a zeolitic support was

observed. The production of reactive oxygen species whose lifetime in the bulk solution

is considerable was enhanced by the presence of the supports.

CONCLUSIONS

Titania-loaded zeolites and mesoporous molecular sieves have drawn increased attention of researchers as novel materials with enhanced photocatalytic properties. The

present work investigates Hombikat titania, TiO2-loaded USY and  zeolites, and TiO2-

loaded MCM-41 molecular sieve as catalysts for the UV-assisted aqueous

sonophotochemical destruction of salicylic acid. It was found that the presence of an

ultrasonic field enhances the rate of photodegradation for the catalysts employed.

Although the ultrasound creates pressure waves in the liquid that could possibly destroy

the framework, the zeolite and MCM-41 matrices were found to be stable under the

present reaction conditions. The BET surface areas of the catalysts utilized were

considerably lower than those of the pure supports, indicating the partial blockage of the

pores by titania. Furthermore, the loaded zeolite b was inactive for the destruction of

salicylic acid, and loaded MCM-41 exhibited the highest specific activity per gram of titania. Unlike Hombikat titania, all composite catalysts deactivated with time. This took place even under the ultrasonic field, although the initial rates of degradation were much higher for the composite catalysts. The deactivation was due to partial blockage of the pores with the unreacted salicylic acid and intermediate as well as the intermediates of its degradation.

207 Two types of single-stage oxidation reactions were employed to investigate the

possibility of enhanced radical generation by the composite catalysts. The effect of

supports on the hydroxyl radicals produced by the titania (tested by HCOOH) was found

to be minimal. No reaction was also observed in the dark, primarily due to the fast recombination of hydroxyl radicals with the solids present in the slurry. Moreover, the

· cumulative generation of OH and H2O2 species (as tested by Fricke dosimeter solution) was indeed affected by the presence of a support. The enhancement due to ultrasound

observed for the supported titania was found to be more significant than that for neat

titania.

ACKNOWLEDGEMENT

The authors wish to thank U.S. Department of Army (Grant No. DAAD 19-00-1-0399) and NATO (Grant No. SfP-974209) for the financial support of this work. The authors are also grateful to Mr. Paul France for fruitful discussions.

REFERENCE

1 Xu, Y. and Langford, C.H., J. Phys. Chem., 1995, 99 11501-11507

2 Xu, Y. and Langford, C.H., J. Phys. Chem., 1997,101, 3115-3121

3 Cho, C.H., Kwak, J.H., Ryoo, R., Ahn, W.S., Jung, K.Y., and Park, S.B., in Progress in Zeolite and Mesoporous Materials (eds H. Chon, S.K. Ihm, and Y.S. Uh) Elsevier Science B.V., 1997, 1617-1623

4 Savage, P.E., Chem. Rev., 1999, 99, 603-621

5 Suslick, K.S., Science, 1990, 247 1439-1445

6 Davydov, L., Reddy, E.P., and Smirniotis, P.G., App.Catal.B: Env., 2001, in press

208

7 Davydov, L. and Smirniotis, P.G., J.Catal., 2000, 191, 105-114

8 Shirgaonkar, I.Z. and Pandit, A.B., Ultrasonics Sonochemisty, 1998, 2, 53-61

9 Theron, P., Pichat, P., Guillard, C., Petrier, C., and Chopin T., Phys.Chem.Chem.Phys., 1999, 1, 4643-4668

10 Sayari, A., Liu, P., Kruk, M., and Jaroniec, M., Chem. Mat., 1997, 9, 2499-2506

11 Corma, A., Fornes, V., Navarro, M.T., and Perez-Pariente, J., J. Catal., 1994, 148 569-574

12 Stelzer J., Paulus M., Hunger M., Weitkamp J., Micro.Meso.Mater., 1998, 22, 1-8

13 Fricke, H. and Hart, E.J. in Radiation Dosimetry Vol. II, Acad. Press, 1966, 167-205

14 Szostak R., Molecular Sieves Principles of Synthesis and Identification, Van Nostrand Reinhold Catalysis Series, New York, 1989

15 Smirniotis, P.G., Davydov, L., Reddy, E.P., and France, P., Proc.Ind. Appl. of Zeolites, Brugge, Belgium, Oct. 22-25, Technologisch Instituut, 2000, 233-241

16 Draganiæ, I.G. and Draganiæ, Z.D. Radiation Chemistry of Water, Acad. Press, 1971

17 Matthews, R.W., Wat. Res., 1990, 24, 653

18 A.J. Johnston and P. Hocking, in Hazardous waste management III, ACS Press, 1993

19 Clark, R.J.H. Chemistry of Titanium and Vanadium, Elsevier, 1968

20 P.V. Kamat, Chem.Rev., 1993, 93, 267-296

209

SECTION 3. NOVEL PHOTOCATALYST DESIGN AND TESTING

Chapter 3.1. Transition metal substituted MCM-41 as photocatalysts of aqueous

VOC oxidation in visible light: Synthesis and characterization

INTRODUCTION

Photocatalysis has recently emerged as an alternative method of decontamination of air and water. Titanium dioxide has received by far the most attention of academic and industrial researchers because it combines high activity, chemical stability, and low price.

However, the widespread commercial utilization of this compound as a photocatalyst has serious drawbacks. The major problem is its large bandgap requiring ultraviolet light for its excitation. Therefore, no more than 5 % of the solar spectrum can be utilized for the process. If used indoors, it requires artificial sources of UV light, which can be harmful to living species and energy consuming. Thus, there is an urgent need to develop new generation of photocatalysts capable of utilizing ideally the full solar spectrum outdoors or requiring no special light sources indoors. This presents a challenge of modifying existing photocatalysts to enable them to absorb and convert visible light to chemical energy.

There is a number of semiconductors which can serve as candidates for this application. Cadmium sulfide has been studied for the photodegradation of organics in visible light, but its stability is very low due to photocorrosion and release of toxic cadmium ions into the reaction medium [1]. Many oxides of transition metals absorb part or all spectrum of visible light. Colloidal iron (III) oxide was found to be active for the photooxidation of salicylic acid and phenol in visible light [2, 3], and leaching of iron

ions was observed. Modified tungsten (VI) oxide was found active for the

211 photodegradation of oxalic acid in visible light [4], but the rates reported are very low.

Modified titanium dioxide was also explored for the degradation in visible light. Being doped with transition metals, TiO2 exhibits better absorption response in the visible part of the spectrum. It was found [5] that dopants considerably reduce the activity of titania since they act as charge carrier recombination centers. Titania having iron (III) oxide deposited on its surface has also been prepared, and it was found that up to a certain content it increases the activity of TiO2 in UV. Copper oxide (II) encapsulated in titania was reported [6] to be active for the photooxidation of ethanol in visible light. However, the catalyst was pre-irradiated in UV, thus making unclear the true source of activity.

The purpose of the present study is to synthesize and characterize novel mesoporous materials for the photocatalytic utilization of visible light. This will be achieved by using transition metal modified molecular sieves in conjunction with titania.

Then a number of experimental methods will be used to characterize these materials and explain the activity trends on the basis of such characterizations.

EXPERIMENTAL

Synthesis

Transition metal substituted MCM-41 supports with Si/Me=80 and Si/Ti=40 were synthesized as previously reported [7, 8] using Ludox HS-40 (DuPont) as the source of silica. The precursors used for the incorporation of transition metal oxides in the framework of MCM-41 were vanadia: VO(C3H7O)3 (Alfa); chromia: CrCl3 (Fisher); iron

(III) oxide: Fe2(SO4)3 (Fisher); titania: titanium isopropoxide (Aldrich). All samples were prepared in the presence of hexadecyltrimethylammonium bromide (Fluka) as a template.

212 The following is the typical preparation procedure: 35 grams of Ludox was added to

14.55 ml of water under stirring, and 18.2 ml of 40 % tetramethylammonium hydroxide

(Fluka) added. Independently, 18.25 g of the template was dissolved in 33 ml of water,

and subsequently 7 ml of 28 % NH4OH was introduced. Finally, the above two solutions

containing Ludox and template were mixed together. The corresponding amounts of each

transition metal oxide precursor dissolved in either ethanol or water (depending on the

nature of the precursor) were added drop-wise from a pipette to the resulting mixture.

The final mixtures were stirred together for 30 minutes, then transferred into polypropylene bottle and treated under autogenous pressure without stirring at 90 - 100C for 3 days. The resulting solids were filtered, washed, dried, and calcined at 550C for 10 hours under airflow. The temperature profile was 2 °C/min up, 15 °C/min down.

The resulting catalyst (typicaly 1.5 g) was dispersed in ~100 ml of isopropanol, and titanium isopropoxide was added to achieve 25 % loading. The system was dried while stirring at ambient temperature. It was then placed in the oven to dry at 100C for 1 hour. They were then transferred into a boat-like crucible and calcined at 450C for 3 hours with a temperature ramp of 2 C/min. Certain samples of Cr-MCM and V-MCM underwent reduction in hydrogen at 380C. The temperature was chosen to avoid the formation of extra-framework crystalline chromia [9]. In this procedure, hydrogen

(Wright Brothers, 99.9 %) was run through a tubular reactor with a bed of catalyst with a heating rate of 2 degrees per minute and stay of 3 h at the final temperature.

213 Characterization

All catalysts were characterized using Nicolet powder X-ray diffractometer equipped with a CuKa source to assess their crystallinity. MCM-41 powders were run from 2 to 7 degrees (2q) to assess the crystallinity of the matrix and from 20 to 50 to assess the crystallinity of the TiO2 loading (XRD). Furthermore, the powders were characterized by UV-Vis spectrophotometer (Shimadzu 2501PC) with an integrating sphere attachment ISR1200 for their diffuse reflectance in the range of wavelengths of

200 to 800 nm. BaSO4 was used as the standard in these measurements. BET and pore size distribution studies were also conducted (using Micromeritics ASAP-2010 apparatus) to characterize the synthesized photocatalysts. The exact compositions of the photocatalysts were determined the ICP spectroscopy of their aqueous suspensions. The valence states of transition metal ions inside the solids were deduced from the diffuse reflectance spectra of the powders.

Temperature programmed reduction (TPR) experiments were carried out in a gas flow system equipped with a quartz micro-reactor, using custom-made set-up attached with TCD detector. Approximately 100 mg of sample was pretreated in 23 ml/min flowing of He at 350°C for 1 h. After pretreatment, the catalyst samples were tested in 6

3 vol% H2/He, 25 cm /min and increasing the temperature from 100 to 800 °C at 5°C/min and kept the temperature at 800°C for 2 h. The amount of hydrogen consumed by the catalyst sample in a given temperature range (in mmoles/g) was calculated by integration of corresponding TCD signal intensities taking in account calibrated values obtained in separate experiments.

214 X-ray photoelectron spectroscopy (XPS) was used to analyze the atomic surface

concentration on Cr-Ti-MCM-41 and 25%TiO2/Cr-Ti-MCM-41. The XPS analyses were

conducted on a Perkin-Elmer Model 5300 X-ray photoelectron spectrometer with MgKa

radiation at 300 W. Typically, 89.45 and 35.75 eV pass energies were used for survey

and high-resolution spectra, respectively. The effects of the sample charging were

eliminated by correcting the observed spectra for a C 1s binding energy values of 284.5

eV. The powdered catalysts were mounted onto the sample holder and were degassed

overnight at room temperature and pressures on the order of 10-7 Torr. The binding

energies and atomic concentrations of the catalysts were calculated via the XPS results

using the total integrated peak areas of the Cr 2p, Ti 2p, Si 2p, and O 1s regions.

4500 RESULTS AND DISCUSSION

The X-ray diffraction 3600 (XRD) analysis emloyed to 2700 characterize the crystallinity of

Cr-Ti-MCM-41 1800 the catalysts showed a number of

Intensity (a.u.) V-Ti-MCM-41 trends. First, the diffractograms 900 Fe-Ti-MCM-41 recorded from 2 to 7 degrees 0 2 3 4 5 6 7 (Figure 3.1.1) exhibited the same 2q location of peaks as siliceous Figure 3.1.1. XRD diffractograms of Me-Ti-MCM- MCM-41 [7]. However, the 41 (Si/Me-80, Si/Ti=40) supports, Me=V, Cr, Fe

intensities of these peaks are lower than those found for siliceous MCM-41. This can be attributed to the presence of

215 foreign ions in the gel during synthesis, which can hinder the structure directing action of

the template by changing the ionic strength of the medium. A similar trend (the lowering

of peak intensities) was observed by Kevan et al [14]. The second part of the XRD

analysis was performed in the range of 20 to 50 degrees in order to assess the crystallinity

of TiO2 loaded onto the transition metal substituted MCM-41 support. It showed that our samples exhibit low crystallinity of titania. Xu and Langford [10] observed a different trend when loading titania onto siliceous MCM-41. However, they utilized a different support (unmodified MCM-41) and different loading method (sol-gel precipitation with acid peptization). There are steric hindrances associated with the formation of titania clusters inside zeolites and mesoporous molecular sieves [10] due to the low pore size of these materials. This does not allow for the efficient crystal growth under the calcination conditions employed, and the overall crystallinity of TiO2 in our samples remains low. It

is also worth noting that no peaks corresponding to the transition metal oxides (those of

V, Cr, and Fe) were observed on our diffractograms. This indicates that our oxides were

either atomically dispersed in the framework of MCM-41 or attained an amorphous form

outside the framework. Transition metal ions are expected to be in intimate contact with

the loaded titania, provided that uniform distribution of titania on the pore walls of the

molecular sieve has been achieved. With this structure it will be possible to combine the

effect of TiO2 and the atomic dispersion of the transition metal inside MCM-41 achieved as a result of the work of a structure-directing agent. This may not possible for mixed oxides. It is expected that this special arrangement will yield increased activity for the photodegradation of aqueous organic pollutants in visible light.

216 The Fourier Transform Infrared (FTIR) spectra (Figure 3.1.2) of the three types of

the powders (MCM-41, Cr-Ti-MCM-41, and TiO2 loaded Cr-Ti-MCM-41) were recorded in the range of wavenumbers of 400-4000 cm-1 order to assess the trace amounts of the partially decomposed surfactant template remaining inside the MCM-41 material. We observed similar trends for all three materials with different intensities of the transmis-

sion peaks. The presence of a

peak at about 1620 cm-1 for all c MCM-41 materials clearly

shows the presence of b quaternary ammonium salts

(upper curve). The

Trasmittance enhancement of a peak at 1850 a cm-1 (leftover ammonia) for the

substituted MCM-41 material

1400 1600 1800 2000 (middle curve) leads to a Wavenumber (cm-1) conclusion that the ammonia Figure 3.1.2. FTIR spectra of a) MCM-41; b) Cr-Ti- leaving the template MCM-41; c)25 wt% TiO2/Cr-Ti-MCM-41

(quaternary ammonia salt) is

likely to form a salt with the amphoteric or acidic transition metal oxide (for example,

(NH4)2CrO4). The second calcination (upper curve) allows to remove thus chemisorbed ammonia. Furthermore, a new peak at 1550 cm-1 is significantly enhanced when titania is loaded onto Cr-Ti-MCM-41, while the one at 1620 cm-1 almost disappears. One can

217 attribute the first peak to Cr=O bond stretching, as we will see from the XPS results that

chromium species travel to the surface of the composite catalyst.

The BET surface areas (SA), peak pore sizes (PPS), and oxygen chemisorption

results (% of metal dispersion, MD) are summarized in Table 3.1.1. One can observe that

the presence of transition metal ions in the gel during synthesis lowers the SA of the

resulting MCM-41 material (for example, SA=941 m2/g for siliceous MCM-41 and

SA=703 m2/g

Table 3.1.1. BET areas, peak pore sizes, and metal dispersions of the catalysts used in the present study BET SA Pore volume Peak pore % of metal Catalyst (m2/g) (cm3/g) size (nm) dispersion* MCM-41 941 0.9 4.2 Not defined

25%TiO2/MCM-41 667 0.6 3.4 0.1 Cr-Ti-MCM-41 702 1.1 6.7 51.0

25%TiO2/Cr-Ti-MCM-41 571 0.7 5.1 23.4 Cr(III)-imp. MCM-41 1217 1.0 N/D 51.8

25% TiO2/Cr(III)MCM-41 932 0.7 2.7 34.2

CrSiOx 553 Not defined** Not defined 74.5

25%TiO2/CrSiOx 425 Not defined Not defined 195.3*** Fe-Ti-MCM-41 719 0.8 5.7 28.6

25%TiO2/Fe-Ti-MCM-41 536 0.7 5.1 37.8 V-Ti-MCM-41 742 1.3 7.1 17.7

25%TiO2/V-Ti-MCM-41 628 0.6 3.9 14.4 Note: * Metal surface area, determined by oxygen chemisorption, stoichiometric factor metal to oxygen equals to 1. ** Not defined due to the absence of well defined porosity. *** The true stoichiometric factor is equal to 3 for chromium, therefore the metal dispersion can attain values higher than 100%.

218 for Cr-Ti-MCM-41). Nonetheless, the BET surface area of the catalysts employed in the

present study is much higher than titanias conventionally used for UV-assisted

heterogeneous photocatalysis [11]. The PPS also changes with the introduction of the

transition metal. Since the same surfactant template was utilized for the synthesis of both

siliceous and transition metal substituted MCM-41 materials, one should expect to obtain

PPS in a close range for both types of catalysts. The above discrepancy is therefore due to the partial breakage of the tubular walls of the MCM-41 structure resulting in the formation of bigger pores (as well as lower SA).

It should be noted that some loss of SA is observed when titania is deposited on the MCM-41 support, as the pore diameters are expected to decrease. Indeed, the PPS of

MCM-41 also reduces with the loading of titania. This is a logical conclusion of the loss of surface area due to partial blockage of the pores. Comparing the Cr-substituted MCM-

41 samples elucidates the following phenomena. Smaller percentages of titania deposition (~10 %) led to almost negligible (~1%) loss of the SA. On the contrary, higher coverages (25 %) lead to substantial losses in the SA (~15%). This clearly indicates that the distribution of titania in the two samples is different. Low loadings lead to layer-type distribution of titania [10], which simply covers the pore walls. Higher loadings fill up some of the pores leading to their partial blockage, such that the removal of isopropyl alcohol upon calcination does not open the pore mouth.

The results of oxygen chemisorption (% of metal dispersion, MD) are also presented in Table 3.1.1. Naturally, no MD has been observed for the siliceous MCM-41 sample as well as that containing 25% of TiO2. However, all transition metal substituted

MCM-41 materials exhibited strong oxygen chemisorption. The greatest MD was

219 observed for chromium-incorporated samples, followed by iron and vanadium

incorporated ones. One can also observe that the MD of chromium in Cr-Ti-MCM-41 is

comparable with that of Cr(III) impregnated MCM-41 and greater than that in the mixed

chromia-silica. The former is due to the higher SA of the siliceous MCM-41 (which leads

to significant dispersion capacity), and the latter can be attributed to the presence of

individual clusters in the mixed oxide. The oxygen chemisorption results also show a pronounced effect of the TiO2 loading on the MD. For chromium incorporated and impregnated MCM-41 samples about 50% and 35%, respectively, of the MD is lost upon loading 25% of titania; the loss of MD is lower for V-Ti-MCM-41. This is due to partial blockage of the transition metal sites by the TiO2 loading, making them inaccessible to oxygen. It should be noted that Cr-MCM-41 as well as TiO2/Cr-MCM-41 exhibited exactly the same trends in spite of the lack of titanium atoms in the framework.

On the contrary, Fe-Ti-MCM-41 exhibits higher MD of iron on titania than inside

MCM-41, as the oxygen chemisorption is higher for the TiO2 loaded sample. The dispersion of chromium in the mixed chromia-silica significantly increases upon the loading of titania. This is because the clusters of chromia (VI) in the mixed amorphous oxide can disperse in the titania precursor during loading, which is further enhanced by calcination. As a result, we obtain values of MD over 100%, which is an overestimation due to the use of the stoichiometric ratio of oxygen to metal equal to one.

For the photocatalytic testing of the catalysts specified, several modifications of the conventional reactor setup were necessary. First, one needs to assure that only visible light reaches the catalyst suspension. The transmission spectrum of the filter exhibits a sharp decrease at 400 nm, thus effectively eliminating the ultraviolet part of the spectrum.

220 Considering that all of our catalysts absorb light significantly below 550 nm, the two

strong bands of the mercury lamp available for the reaction are 405 and 435 nm. In order

to have a quantitative means of comparison, the photodegradation of the respective

reactants on UV-irradiated Degussa P25 powder and visible-irradiated Degussa P25 was

undertaken as well. The results are shown in Table 3.1.2 in terms of the reaction rates.

The catalysts without loading of titania (neat V-Ti-MCM-41, Cr-Ti-MCM-41, and Fe-Ti-

MCM-41) were tested for the photodegradation of formic acid and yielded no discernible activity. All reaction solutions were tested for the leaching of transition metal ions. It is remarkable to note that the leaching was negligible for all TiO2 loaded catalysts, and it was considerable for the neat chromium substituted MCM-41 materials (see below).

Table 3.1.2. Comparison of reaction rates of degradation of different VOCs in water by the visible- irradiated composite catalysts of the present study. The performance of Degussa P25 in UV is shown as a benchmark Reaction Rate, mM/min Catalyst

HCOOH C6H6O C6H5OCl C6H3OCl3

P25 - Visible 0 0 0 3.3×10-4

-3 25% TiO2/Fe-MCM-41 3.3×10 0 0 0

-2 -4 -3 -3 25% TiO2/Cr-MCM-41 3.5×10 5.25×10 1.6×10 2.6×10 -4 25% TiO2/V-MCM-41 3×10 0 0 0 Degussa P25 - UV 3×10-2 3×10-3 2×10-3 4.45×10-3

As can be seen from Table 3.1.2, titania-loaded chromium-substituted MCM-41

materials are very active for the photodegradation of organic compounds in visible light.

For the decomposition of formic acid, its activity is comparable to the commercial titania

(Degussa P25) utilized in UV radiation. The iron and vanadium substituted analogs

221 exhibit only marginal activity for this reaction. Furthermore, the activity of 25% TiO2/Cr-

MCM-41 for the photodegradation of phenolic compounds in visible light was also

explored. It is comparable with Degussa P25 in UV for the chlorinated phenols and it is

considerably less for the neat phenol.

Three characterization

1.2 techniques, namely, diffuse UV Visible

1.0 reflectance UV-Visible V-Ti-MCM-41 Cr-Ti-MCM-41 spectroscopy (UV-Vis), 0.8 Fe-Ti-MCM-41 temperature programmed 0.6

reduction (TPR), and X-ray 0.4 Absorbance, a.u. photoelectron spectroscopy 0.2 (XPS) were utilized in an 0.0 200 300 400 500 600 700 800 900 attempt to elucidate the reasons Wavelength, nm of the remarkable activity of Figure 3.1.3. UV-Vis diffuse reflectance spectra of the transition metal substituted MCM-41 supports TiO /Cr-Ti-MCM-41 for the (Si/Me=80, Si/Ti=40) 2

photooxidation in visible light.

The UV-Vis spectra of the catalysts in the range 190-900 nm are shown in Figures 3.1.3-

5. The original transition metal substituted MCM-41 materials also exhibit absorption in visible light as well as in the UV range (Figure 3.1.3). The spectra of the corresponding neat oxides were also taken and allowed us to identify the species present in our materials. As mentioned above, the photocatalytic activity of these samples was negligible. This is because no actual semiconductor oxide was present in the powder. On the contrary, highly dispersed transition metal ions were in this case subjected to light,

222 the local excitation of which cannot produce a significant effect [12]. One can observe

significant absorption in visible light by all materials (Figure 3.1.4). This is consistent

with the absorption by

mixed oxides of vanadium, UV Visible chromium, and iron with

1 titania [13]. Overall, the

enhanced absorption of the TiO2 /V-Ti-MCM-41 TiO /Cr-Ti-MCM-41 2 composite catalyst is a TiO2 /Fe-Ti-MCM-41

Absorbance, a.u. necessary condition for the

photoactivity in visible

200 300 400 500 600 700 800 900 light, but not sufficient for Wavelength, nm its ability to perform Figure 3.1.4. UV-Vis diffuse reflectance spectra of the titania loaded transition metal substituted photocatalysts photocatalytic reactions in (titania loading – 25 wt%) visible light, as was

observed in the activity studies above.

One can observe that the photocatalysts employed in the present study (Figure 3.1.4)

drastically differ from neat titanium dioxide (not shown). Their absorption edges are in

the vicinity of 500-600 nm, thus shifting the bandgap position to ~2.0 eV. The latter can

be determined by the extension of the descending branches of the absorption spectra to

the axis of wavelengths. Then, from the point of the intersection one can obtain the

frequency and multiply by Planck’s constant to arrive at the bandgap energy. The low

bandgap of the materials utilized in the present study justifies why these photocatalysts

can be potentially good candidates for performing photochemical reactions in visible

223 light. The upper branch of the curves (200-350 nm) practically coincides with that of titania. Then, the absorption by titania itself sharply decreases, and the absorption of light

in the range of 350 to 450 nm is exhibited by the heterojunction of titania with the

corresponding transition metal oxide. The absorption after 450 nm is primarily due to the

transition metal oxide itself. It should be noted that the flat line of the absorption spectra

observed after about 450-500 nm does not mean the absence of absorption. Vanadium- substituted catalysts (Figure 3.1.4) allow to expand the absorption spectrum even further toward infrared part, especially in the reduced form. It should be noted that we have chosen to analyze 25 wt % loaded composites as to assure that the presence of a large quantity of the second oxide does not overshadow the true optical properties. One can observe that the loading of titania significantly alters the absorption spectrum of the composite. The neat V-Ti-MCM-41 shows the presence of V+5 species as the onset of its absorption spectrum (~625 nm) coincides with that of pure V2O5 (Figure 3.1.3). The presence of 25 wt % of TiO2, however, shifts the absorption edge from about 625 to about 450 nm, and the presence of vanadium is observable in the tail of the spectrum.

This clearly signals the unavailability of V+5 species inside the framework of MCM-41, signifying that all of these species are fully covered. The reduced TiO2/Cr-Ti-MCM-41 samples exhibited significant absorption in the visible range. Two shoulders are observable on this spectrum: in the range of350-400 nm and 550+ nm. This also coincides with the behavior of the reduced pure vanadium oxide prepared independently.

Iron-substituted MCM-41 samples behave similarly to vanadium-substituted ones

(Figure 3.1.4). The loading of titania simply shifts the absorption edge by about 100 nm towards UV. The spectra of both samples are rather monotonical, which does not allow

224 us to make qualitative conclusions about the nature and placement of iron species inside

MCM-41. One can symbolically divide the curve into three parts: 200-400 nm, 400-500

nm, and 500-600 nm. The first part will correspond to the absorption by pure titania, the

second – by iron doped titania, and the third – by iron oxide itself.

When considering Cr-Ti-MCM-41 based samples (Figures 3.1.3 and 3.1.4), the situation is somewhat similar to the prior work. It was found previously [14] that there are three species inside Cr-substituted MCM-41: framework Cr+3, extra-framework Cr+3 and extra-framework Cr+6. Neat Cr-Ti-MCM-41 exhibits absorption maximum at ~390 nm which corresponds to Cr+6 species. When loaded with titania, this maximum

+6 (corresponding to Cr ) is overshadowed by the absorption by TiO2, leaving only an absorption shoulder. Finally, the reduced sample of titania-loaded Cr-Ti-MCM-41 does contain a small remainder of the above maximum. This can be attributed to incomplete reduction of Cr+6 species. It was found [15] that some of the Cr+6 species inside Cr-Ti-

MCM-41 are non-reducible. Furthermore, the deposition of titania further deprives the reduction process as some of the Cr+6 becomes unavailable. Conversely, the presence of an absorption minimum in the vicinity of 550 nm (green) clearly indicates the presence of

Cr+3 species in abundance.

A comparison of the UV-Vis spectra of various neat oxide standards with chromium containing materials was made. There are absorption peaks at ~275 nm and ~

370 nm and shoulders at ~470 nm and ~600 nm on the spectra of Cr-Ti-MCM-41 (Figure

3.1.3). The same materials, but loaded with 25% TiO2 (Figure 3.1.4) exhibit higher absorption in the UV range due to the presence of titania. All the materials still retain high absorption in visible light (up to 600 nm) and have a distinct shoulder at ~370 nm. If

225 one compares the absorption spectrum of Cr-Ti-MCM-41 with three individual chromium oxides, several trends can be observed. The positions of peaks corresponding to

chromium species in the

1.2 MCM-41 based materials UV Visible

1.0 are shifted towards lower

Cr(NO3)3/MCM-41 Cr-Ti-MCM-41 Subst 0.8 wavelengths in comparison Cr-Ti-MCM-41 Leached with bulk chromium oxide 0.6

(Cr2O3, CrO2, and CrO3). 0.4 Absorbance, a.u. This is possibly due to local 0.2 molecular excitation effects 0.0 200 300 400 500 600 700 800 900 by the respective oxide Wavelength, nm [12], as the dispersion of Figure 3.1.5. UV-Vis diffuse reflectance spectra of the chromium substituted MCM-41 (Si/Cr=80) that have chromia inside MCM-41 is undergone the extraction experiment

high. Such comparison also allows to attribute the peaks of Cr-Ti-MCM-41 at 275 and 370 nm to Cr+6 and shoulder at

~475 nm to Cr+3 similarly to the previous study [15]. No enhanced absorption in the visible range due to Cr+4 is observed, contrary to the prior work [14], which found evidence of Cr+4 in Cr-MCM-41. In order to determine the ratio of Cr+6 to Cr+3 in our calcined Cr-Ti-MCM-41 samples a leaching experiment was performed, and the absorption spectrum of the resulting powder was compared with that of the original Cr-

Ti-MCM-41 (Figure 3.1.5). One can observe the disappearance of the absorption peaks at

275 nm and 370 nm (corresponding to Cr+6), while the shoulder at 470 nm and broad peak at 650 nm (corresponding to Cr+3) were retained by the material. The amount of

226 Cr+6 leached in two independent experiments constituted about 80% of the chromium

introduced during synthesis. Therefore, the majority of Cr+3 oxidized to Cr+6 during the

calcination of Cr-Ti-MCM-41. This again testifies for the high dispersion of chromium inside MCM-41 since it is impossible to obtain Cr+6 by calcination of common chromium

(III) salts [16]. Furthermore, no leaching of Cr+3 could be detected spectroscopically.

Similar leaching tests were performed on the most active catalyst, TiO2/Cr-Ti-MCM-41,

and no significant leaching was detected. Conclusively, the UV-Vis study allowed to

reveal the coexistence of Cr+3 and Cr+6 inside MCM-41, which contributes to its peculiar

photocatalytic behavior.

Temperature programmed

150 reduction (TPR) was used 440 800 in the present study to 120 investigate different

90 oxidation states of the e

60 transition metal substituted d Intensity (a.u.) c MCM-41 materials and 30 b relate these oxidation states a 0 with the photocatalytic 200 400 600 800 Isothermal (2 h) performance. Figure 3.1.6 Temperature (oC) compares the TPR profiles

Figure 3.1.6. TPR profiles of catalysts: a) (Cr(NO3)3 on for a number of Cr- MCM-41; b) CrO3 on MCM-41; c) Cr-Ti-MCM-41; d) Cr- Si mixed oxide; e) Cr-MCM-41 containing materials. One

227 can observe that the

210 reduction behavior is 443 638 800 180 almost identical. Two

150 major peaks can be found: e 120 570 440°C (corresponds to d

90 Cr+6®Cr+3, according to c

Intensity (a.u.) 60 [17]) and ~800°C. The first 500 b 30 peak validates our earlier a 0 observations by UV-Vis 200 400 600 800 Isothermal (2 h) about the presence of Cr+6 Temperature (oC) in the MCM-41 based

Figure 3.1.7. TPR profiles of 25 wt % TiO2 loaded catalysts: a) Cr-MCM-41; b) Cr-Ti-MCM-41; c) materials made with Cr (Cr(NO3)3 on MCM-41; d) CrO3 on MCM-41; e) Cr-Si mixed oxide (III) precursor. The second

peak was observed even in

siliceous MCM-41, so it 519 60 can be attributed to the 752 hydroxyl groups leaving 40 445 the surface of amorphous

305 silica. When TiO2 is loaded Intensity (a.u.) 20 onto our Cr-containing

materials, the TPR profiles

0 show marked difference 200 400 600 800 Isothermal (2 h) Temperature (oC) (Figure 3.1.7). We still

Figure 3.1.8. Deconvoluted TPR profile of 25%TiO2/Cr- Ti-MCM-41 228 observe the two major peaks at 440 and 800 degrees, but a shoulder appears before the

first peak and another peak appears between them. The shoulder in the range of 250-

350°C corresponds to the dehydroxylation of TiO2 surface and also the reduction of

titanium from +4 to +3 [18]. The position of the new peak is apparently due to the

transition Cr+3®Cr+2 [18]. It should be noted that the position of this new peak strongly

depends on the preparation method. For the titania loaded chromium impregnated MCM-

41 materials (the upper 3 curves of Figure 3.1.7), the peak is located at about 640°C. For

TiO2/Cr-MCM-41 and TiO2/Cr-Ti-MCM-41 this peak is much less pronounced and

shifted by about 70°C towards the lower temperatures.

This behavior can be seen in more detail in Figure 3.1.8 representing the

deconvoluted TPR diagram for the most active catalyst (TiO2/Cr-Ti-MCM-41). One can

observe from these profiles that the position of the central peak corresponding to

Cr+3®Cr+2 transition is clearly at 520°C, which is considerably higher than the reduction

temperature recorded by other researchers [18]. Apparently, highly dispersed Cr+3 (Cr+6

reduces at 440°C) inside MCM-41 facilitates the above transition, contrary to the

behavior of the materials simply impregnated with chromium precursors (TiO2/Cr(III)-

imp MCM-41, TiO2/Cr(VI)-imp MCM-41) or amorphous silica-chromia. Since the

composition of all the materials in Figure 3.1.7 is the same, the lower temperature of the

reduction of titania is due to a high degree of its interaction with the framework

chromium. Furthermore, when incorporated into MCM-41 during synthesis, chromium is

expected to attain the tetrahedral coordination [14], while the impregnated MCM-41 has

chromium in its native coordination. This may also contribute to the peculiar interaction

229 of chromium and titanium oxides in our active catalysts for the photodegradation of

aqueous organic pollutants in visible light.

The TPR diagrams of

spent chromium based 415 618 150 800 photocatalysts are presented

542 590 210 in Figure 3.1.9. It should be 100 537 c reminded that 25% TiO2/Cr-

b Ti-MCM-41 (as well as 25% Intensity (a.u.) 50

TiO2/Cr-MCM-41) exhibited a 0 significant activity for the 200 400 600 800 Isothermal (2 h) Temperature (oC) photodegradation of organic

Figure 3.1.9. TPR profiles of catalysts after compounds in visible light. photocatalysis: a) 25%TiO /Cr-Ti-MCM-41; b) 2 On the contrary, the other 25%TiO2/(Cr(NO3)3 on MCM-41; c) CrO3 on MCM-41

two catalysts (curves b and c

of Figure 3.1.9) showed negligible activity. As a result, the qualitative TPR behavior of

the inactive catalysts remained almost unchanged after the reaction. On the contrary, the

peaks corresponding to the interaction of chromium and titania (Cr+3®Cr+2) in the active

catalyst have disappeared, and the TPR profile became similar to that of TiO2/MCM-41.

The disappearance of the interaction described is apparently the reason for the

deactivation observed for this photocatalyst after 1 hour of photocatalytic reaction with

HCOOH. This was not the case when degrading phenolic compounds. No deactivation

was observed, and the color of the catalyst remained yellow as opposed to turning green

when photodegrading formic acid in visible light.

230 The samples of

25%TiO2/Cr-Ti-MCM-41

O1s from SiO2 and Cr-Ti-MCM-41 were

investigated by ESCA. The

XPS bands of O 1s, Si 2p and

Cr 2p core levels are shown a Intensity (a.u.)

O1s from TiO2 in Figures 3.1.10, 3.1.11, and

O1s from CrOx 3.1.12, respectively. The

b binding energy values of O

526 528 530 532 534 536 1s, Si 2p, Ti 2p and Cr 2p Binding energy (eV) photoelectron peaks and

Figure 3.1.10. Deconvoluted XPS spectra for O 1s Cr/Si and Cr/Ti surface peak: a - Cr-Ti-MCM-41; b - 25%TiO2/Cr-Ti- MCM-41 atomic concentration ratios as determined by XPS of the above catalysts are summarized in Table 3. All these figures clearly indicate that the XPS photoelectron peaks depend on the surface concentration of chromia in the Cr-Ti-MCM-41 and 25%TiO2/Cr-Ti-MCM-41. The O 1s profile, as shown in Figure 3.1.10, is due to the overlapping contribution of oxygen from silica and chromia in the case of Cr-Ti-MCM-41 and silica, chromia and titania in the case of 25%TiO2/Cr-

Ti-MCM-41, respectively. As shown in Figure 3.1.10a from the deconvoluted XPS spectra of O 1s corresponding to 25%TiO2/Cr-Ti-MCM-41, one can clearly detect that there are three types of O 1s peaks which binding energy values at 529.6 eV, 530.0 eV and 532.7 eV (Table 3) are belong to the oxygen atoms that are more bound to Cr (CrOx),

Ti (TiO2) [19] and Si (SiO2) [20], respectively. The binding energy values of three

231 different types of O 1s peaks can be judged from the difference in the electronegativity of

the elements [21]. It should be noted that there was only one oxygen photoelectron peak

at 532.7 eV belongs to SiO2 was observed in the case of Cr-Ti-MCM-41. The peak intensity of O 1s XPS band corresponding to SiO2 reduced drastically when titania was

loaded on Cr-Ti-MCM-41. This clearly indicates that the titania loading on Cr-Ti-MCM-

41 changing its structural behavior and also possible migration of oxygen atoms is taking place with respect to different structural modifications.

Figure 3.1.11 shows the

binding energy of Si 2p Si 2p photoelectron peak at 103.5

eV, which agrees well with the

values reported in the literature

[20]. The intensity of Si 2p is

Intenisty (a.u.) more predominant in the case a of Cr-Ti-MCM-41 than the b titania loaded Cr-Ti-MCM-41. 97.5 100.0 102.5 105.0 107.5 As seen from the XPS figures Binding energy (eV) presented, the ratio of the Cr/Si Figure 3.1.11. XPS spectra for Si 2p peak: a) Cr-Ti- MCM-41; b) 25%TiO2/Cr-Ti- MCM-41 for Cr-Ti-MCM-41 is much

less than that of the ratio of Cr/Si for 25%TiO2/Cr-Ti-MCM-41. As a matter of fact, it was due to the surface coverage of titania and chromia segregation on to the surface of

Cr-Ti-MCM-41.

232 As shown in Figure

3000 3.1.12 the surface Cr(VI) 2p Cr(VI) 2p 3/2 1/2 a Cr(III) 2p Cr(III) 2p 3/2 1/2 concentration of Cr 2p is 2500 more predominant in the case 2000 b of 25%TiO2/Cr-Ti-MCM-41 1500 than in case of Cr-Ti-MCM-

1000 41, but the Cr(VI) peak at Intensity (a.u.) 579.3 eV is more dominating 500 than Cr (III) in the latter case. 0 575 580 585 590 XPS analysis shows two Binding energy (eV) different Cr species or

Figure 3.1.12. Deconvoluted XPS spectra for Cr 2p chemical states on the peak: a) Cr-Ti-MCM-41; b) 25%TiO2/Cr-Ti- MCM-41 catalysts surfaces (Figure

3.1.12). The simple interpretation would be to assign the lowest binding energy peak at

576.7 eV corresponding to Cr (III) and the highest binding energy peak at 579.3 eV corresponding to Cr (VI). The co-existence of Cr2O3 and CrO3 has often been observed on different supports [22]. An increase of the Cr(III)/Cr(VI) ratio is observed when the surface concentration of Cr increases (Figure 3.1.12). The enhancement of the surface mobility and surface rearrangement of chromia was due to the loading of 25% TiO2 on

Cr-Ti-MCM-41. The possible explanation for these data is breaking of Si-O-Cr bonds in

Cr-Ti-MCM-41 pores with the formation of reduced chromia species in the form of Ti-O-

Cr on the surface of the catalysts. One can expect that possible oxygen migration leads to the formation of new oxide islands of lower coordination. This type of structural

233 modifications could occur and lead to variations in the dispersion. It could also reflect the

heterogeneous character of the surface chromia, which leading to the existence of two

separated electronic levels. Further confirmation was obtained by characterization of Cr-

Ti-MCM-41 and 25%TiO2/Cr-Ti-MCM-41 catalysts with UV-Vis and photocatalytic degradation of formic acid. The photocatalytic degradation of organic compounds becomes completely dependent on the surface concentration of chromia on the catalyst.

That is the reason why the photodegradation is the highest in the case of 25%TiO2/Cr-Ti-

MCM-41.

CONCLUSIONS

Several transition metal based titania loaded MCM-41 (Si/Me=80, titania content

– 25 wt%) materials were synthesized as potential photocatalysts for the degradation of

organics in visible light. The conventional MCM-41 synthesis was modified by the

addition of transition metal precursor and followed by the sol-gel deposition of titania.

The photocatalytic activity of these catalysts was tested for the degradation of formic acid

as well as phenolic compounds, and the cromium substituted catalyst was found to be

exceptionally active in visible light. An extensive characterization study was undertaken

into these materials. The UV-Vis spectrophotometric study revealed the coexistence of Cr

(III) and Cr (VI) oxides inside the substituted MCM-41 at the ratio of 0.25. TPR showed the specific interaction of Cr and Ti in these materials, which is believed to contribute to their photocatalytic activity in visible light. XPS revealed increased surface concentrations of Cr ions upon the loading of TiO2, which at the same time allows to

minimize the leaching of chromium ions into the aqueous solution.

234

ACKNOWLEDGEMENT

The authors are grateful to the Young Investigator Award of the United States

Department of Army (Grant No. DAAD 19-00-1-0399) and NATO Science for Peace

Program (Grant No. SfP-974209) for their support of this work.

REFERENCES

1. Khairutdinov, M., Colloidal J., 59, 535 (1997)

2. Pal, B. and Sharon, M., J.Chem.Technol.Biotechnol., 73, 269 (1998)

3. Chatterjee, S., Sarkar, S., and Bhattachatyya, J. Photochem. Photobiol. A: Chem., 81, 199 (1994)

4. Ashokkumar, M. and Maruthamuthu, P., Int. J. Hydrogen Energ. 16, 591 (1991)

5. Serpone, N., Lawless, D., Disdier, J., and Herrmann , J.M., Langmuir, 10, 643 (1994)

6. Sakata, Y., Yamamoto, T., Okazaki, T., Imamura, H., Tsuchiya, T., Chem. Lett., 1253 (1998)

7. Sayari, A., Liu, P., Kruk, M., and Jaroniec, M., Chem. Mat., 9, 2499 (1997)

8 Corma, A., Fornes, V., Navarro, M.T., and Perez-Pariente, J., J. Catal., 148, 569(1994)

9. Willi, R., Maciejewski, M., Gobel, U., Koppel, R.A., and Baiker, A., J. Cat., 166, 356 (1997)

10 Xu, Y., and Langford, C.H., J. Phys. Chem., 101, 3115 (1997)

11. Davydov, L. and Smirniotis, P.G., J.Cat., 191, 105 (2000)

12. Zang, L., Lange, C., Abraham, I., Storck, S., Maier, W., and Kisch, H., J.Phys.Chem. B, 102, 10765 (1998)

13 Serpone, N., Lawless, D., Disdier, J., and Herrmann , J.M., Langmuir, 10, 643 (1994)

14. Zhu, Z.D., Chang, Z.X., and Kevan, L., J. Phys. Chem. B, 103, 2680 (1999)

235

15. Ulagappan, N., Rao, C.N.R., Chem. Comm, 1047 (1996)

16. Maciejewski, M., Kohler, K., Schneider, H., and Baiker, A., J.Solid State Chem., 119, 13 (1995)

17. Uhm, J.H., Shin, M.Y., Zhidong, Z., Chung, J.S., Appl.Cat.B:Env., 22, 293 (1999)

18. Zhu, Z., Hartmann, M., Maes, E.M., Czernuszewicz, R.S., and Kevan, L., J.Phys.Chem.B, 104, 4690 (2000)

19. Reddy, B.M., Chaodhay, B., Reddy, E.P., Fernandez, A., J.Mol.Catal.A:Chem., 162, 431 (2000)

20. Reddy, B.M., Ganesh, I., Reddy, E.P., J.Phys.Chem.B, 101, 1769 (1997)

21. Imamura, I., Ishida, S., Taramoto, H., Saito,, Y., J.Chem.Soc. Faraday Trans., 89, 27 (1993)

22. Pradier, C.M., Rodrigues, F., Marcus, P., Landau, M.V., Kaliya, M.L., Gutman, A., Herskowitz, M., Appl.Catal.B:Env., 27, 73 (2000)

236 Chapter 3.2. Transition metal substituted MCM-41 as photocatalysts of aqueous VOC oxidation in visible light: Photocatalytic activity

INTRODUCTION

Photooxidation by irradiated semiconductors is a relatively new technique of pollution abatement. The vast majority of current studies employ TiO2 as the semiconductor due to its stability and relatively low price. However, it requires UV-light to be excited and become capable of photooxidation. Ultraviolet radiation is expensive as it requires additional sources of energy, and it is also harmful to living species. Therefore, there is an urgent need to develop new photocatalysts capable of working in visible light. There is a number of semiconductors which can serve as candidates for this application. Cadmium sulfide has been studied for the photodegradation of organics in visible light, but its stability is very low due to photocorrosion and release of toxic cadmium ions into the reaction medium [1]. Many oxides of transition metals absorb part or all spectrum of visible light. Colloidal iron (III) oxide was found to be active for the photooxidation of salicylic acid and phenol in visible light [2, 3], and leaching of iron ions was observed.

Modified tungsten (VI) oxide was found active for the photodegradation of oxalic acid in visible light [4], but the rates reported are very low. Modified titanium dioxide was also explored for the degradation in visible light. Being doped with transition metals, TiO2 exhibits better absorption response in the visible part of the spectrum. It was found [5] that dopants considerably reduce the activity of titania since they act as charge carrier recombination centers. Titania having iron (III) oxide deposited on its surface has also been prepared, and it was found that up to a certain content it increases the activity of TiO2

237 in UV. Copper oxide (II) encapsulated in titania was reported [6] to be active for the

photooxidation of ethanol in visible light. However, the catalyst was pre-irradiated in UV, thus making unclear the true source of activity.

Overall, there has been no catalyst prepared for work in the visible light, which would combine chemical stability and high activity. Xu and Langford [7] reported an activity enhancement of titania when it was loaded onto zeolitic or mesoporous silica support. Transition metal substituted zeolites and MCM-41 have been reported [8].

Substituting silicon with another atom during synthesis of MCM-41 allows to atomically disperse the transition metal in the framework of the zeolite [9] under the action of the template. This property is used in the present study of titania-loaded transition metal substituted MCM-41 molecular sieves as potential photocatalysts for the oxidation of organics in visible light [10]. Successful photooxidation in visible light can allow for more effective utilization of solar energy.

EXPERIMENTAL

Synthesis and Characterization

Transition metal substituted MCM-41 supports with Si/Me=80 and Si/Ti=40 were synthesized as previously reported in Chapter 3.1. The precursor of silica in MCM-41 was

Lodox HS-40 (Aldrich). The precursors used for the incorporation of transition metal oxides in the framework of MCM-41 were vanadia: VO(C3H7O)3 (Alfa); chromia: CrCl3

(Fisher); iron (III) oxide: Fe2(SO4)3 (Fisher); titania: titanium isopropoxide (Aldrich). The final gel mixture was treated under autogenous pressure without stirring at 90 - 100C for 3

238 days. The resulting solids were filtered, washed, dried, and calcined at 550C for 10 hours under air flow. The temperature profile was 2 °C/min up, 15 °C/min down.

The resulting catalyst (typicaly 1.5 g) was dispersed in ~100 ml of isopropanol,

and titanium isopropoxide was added to achieve 25 % loading. The system was slowly

dried while stirring at ambient temperature. It was then placed in the oven to dry at 100C

for 1 hour. The powder was calcined at 450C for 3 hours with a temperature ramp of 2

C/min. Certain samples of Cr-Ti-MCM-41 and V-Ti-MCM-41 underwent reduction in hydrogen at 380C for 3 hours.

All catalysts were characterized using a number of methods as described in

Chapter 3.1. Nicolet powder X-ray diffractometer equipped with a CuKa source was used to assess the crystallinity of the powders. Furthermore, the powders were characterized by

UV-Vis spectrophotometer (Shimadzu 2501PC) with an integrating sphere attachment. BET and pore size distribution studies were also conducted (using Micromeritics ASAP-2010 apparatus) to characterize the synthesized photocatalysts.

Temperature programmed reduction (TPR) experiments were carried out in a gas flow system equipped with a quartz micro-reactor, using custom-made set-up attached with

TCD detector. X-ray photoelectron spectroscopy was used to analyze the atomic surface concentration on Cr-Ti-MCM-41 and 25%TiO2/Cr-Ti-MCM-41. The XPS analyses were conducted on a Perkin-Elmer Model 5300 X-ray photoelectron spectrometer with MgKa

radiation at 300 W.

239

Figure 3.2.1. Scheme of the photocatalytic Figure 3.2.2. Emission spectrum of the reactor employed in the present study mercury lamp and transmission spectrum of the light filter

Photocatalytic Experiments

The photocatalytic testing included the degradation of organic compounds, which

was performed in a batch round flat-plate reactor (Figure 3.2.1) using 200 W medium

pressure mercury lamp (Hanovia) as the light source. A 0.25” thick plexiglas filter (US

Plastics, Cat. No. 44673) was utilized for the purpose of excluding ultraviolet radiation

when conducting visible-light experiments. The cooling jackets around the reactor and

around the lamp allowed to effectively preclude the IR part of the spectrum from

penetrating into the reaction solution and cooled the lamp. The emission spectrum of the

lamp (obtained from the manufacturer) as well as the transmission spectrum of the filter

(obtained spectrophotometrically) are presented in Figure 3.2.2. Several reactants were

tested for the photocatalytic degradation: 2,4,6-trichlorophenol (Fisher), 4-chlorophenol

(Aldrich), formic acid (Fisher). Although other researchers [5] used oxalic acid as their

240 probe molecule for photodegradation, the choice of our reactants was justified by the possibility of the leaching of transition metal ions from the catalysts into the solution. Thus,

to minimize such leaching chemicals that cannot act as chelates were chosen. Prior to the

reaction 0.5 L of the slurry (1 g/L of solids, 1 mM of the corresponding phenolic

compound, pH~6, or 10 mM of formic acid, pH~4) was ultrasonicated for 10 min in an

ultrasonic bath in order to assure the breakage of catalyst aggregates. The suspended

catalyst in aqueous system was oxygenated (Wright Bros, 99.9%) at 0.5 L/min to assure the

complete saturation. The temperature in the reactor was kept constant at 25±3°C. The

concentration of phenol in the reactor was monitored by UV-Vis spectrophotometry

(Shimadzu PC2501) at the wavelength of 270 nm, the concentration of 4-chlorophenol – at the wavelength of 274 nm. The concentration of 2,4,6-trichlorophenol was measured chromatographically (GC HP 6891 equipped with an FID and a capillary column) as described elsewhere [11]. The concentration of formic acid was tracked using a conductivity meter (VWR Scientific, cell constant 10 cm-1). The latter device was pre- calibrated for the measurements by standard solutions of formic acid. The presence of transition metal ions in the solution was also monitored by absorption spectrophotometry:

Cr+6 at 374 nm, Cr3+ at 460 nm, Fe3+ at 304 nm, and V+5 at 350 nm. Standard solutions of

(NH4)2CrO4 (Fisher), CrCl3 (Fisher), FeCl3 (Aldrich), and NH4VO3 (Aldrich), respectively, were used for calibration purposes.

241 RESULTS AND DISCUSSION

The exhaustive

characterization of the materials UV Visible utilized in the present study has

1 already been performed and

discussed (Chapter 3.1). Several TiO2/V-Ti-MCM-41 TiO /Cr-Ti-MCM-41 2 issues will be reiterated in order to TiO2/Fe-Ti-MCM-41

Absorbance, a.u. further explore the photocatalytic

activity of these materials. First, it

200 300 400 500 600 700 800 900 was observed that titania loaded Wavelength, nm transition metal substituted MCM- Figure 3.2.3. UV-Vis diffuse reflectance spectra of the titania loaded transition metal substituted photocatalysts 41 molecular sieves can absorb UV (titania loading – 25 wt%) as well as some portion of visible light (Figure 3.2.3). The lack of absorption in the visible by 25%TiO2/V-Ti-MCM-41 is notable.

Furthermore, the spectrum of the above catalyst is smooth, thus lacking shoulders observed for

25%TiO2/Cr-Ti-MCM-41 and 25%TiO2/Fe-Ti-MCM-41. Such shoulders usually hint at the existence of dopant levels inside the oxide matrix. As will be seen below, this property of

25%TiO2/V-Ti-MCM-41 affects its photocatalytic performance in visible light.

The metal dispersions of the materials utilized in the present study as well as several test materials as determined chemisorption are shown in Table 3.2.1. By analyzing the values of the transition metal dispersions one can note the significant difference of the vanadium-substituted catalyst. Its metal surface area is much lower than that of the rest of the catalysts and it is almost

242 Table 3.2.1. BET areas, peak pore sizes, unaffected by the loading of titania. This behavior and metal dispersions of the catalysts used in the present study may contribute to the low photocatalytic activity Catalyst % of metal of the above catalyst, as will be shown below. dispersion* MCM-41 Not defined The other important finding from

25%TiO2/MCM-41 0.1 characterization is surface elemental Cr-Ti-MCM-41 51.0 concentrations ratios as determined by XPS 25%TiO2/Cr-Ti-MCM-41 23.4 Cr(III)-imp. MCM-41 51.8 (Table 3.2.2). As seen from the table, chromium

25% TiO2/Cr(III)MCM-41 34.2 present in minute amounts on the surface of neat CrSiO 74.5 x Cr-Ti-MCM-41. The loading of TiO2, however, 25%TiO2/CrSiOx 195.3 enhances the surface Cr/Si ratio about 200-fold. Fe-Ti-MCM-41 28.6

25%TiO2/Fe-Ti-MCM-41 37.8 This contributes significantly to the unique V-Ti-MCM-41 17.7 photocatalytic activity of 25%TiO2/Cr-Ti-MCM- 25%TiO2/V-Ti-MCM-41 14.4 41 explored in the present study. Note: * Metal surface area, determined by oxygen chemisorption, stoichiometric factor metal to oxygen equals to 1.

Table 3.2.2. Surface atomic ratios of the select catalysts as determined by XPS Surface Atomic Catalyst Ratios Cr/Ti Cr/Si

25%TiO2/Cr-Ti-MCM-41 0.136 0.258 Cr-Ti-MCM-41 0* 0.0014 Note: * corresponding species not found on the surface of the catalyst

243 For the photocatalytic

testing of the catalysts

1.0 specified, several modifica-

0.9 tions of the conventional

0.8 reactor setup were necessary.

First, one needs to assure that 0.7

HCOOH C/Co Degussa P25 Vis only visible light reaches the 0.6 TiO /V-Ti-MCM Vis 2 catalyst suspension. The TiO2/Fe-Ti-MCM Vis TiO /Cr-Ti-MCM Vis 0.5 2 transmission spectrum of the Degussa P25 UV 0 20 40 60 80 100 120 140 160 180 filter exhibits a sharp Time, min decrease at 400 nm, thus Figure 3.2.4. Time course of the degradation of formic acid on the V-, Cr-, and Fe-substituted catalysts (Table effectively eliminating the 3.2.1), Degussa P25 in UV is shown for comparison (pH=4, T=25±3°C, catalyst concentration – 1 g/L) ultraviolet part of the

spectrum (Figure 3.2.2).

Considering that all of our catalysts absorb light significantly below 550 nm, the two strong

bands of the mercury lamp available for the reaction are 405 and 435 nm. In order to have

a quantitative means of comparison, the photodegradation of the respective reactants on

UV-irradiated Degussa P25 powder and visible-irradiated Degussa P25 was undertaken as well. The results are shown in all figures depicting the course of degradation (Figures

3.2.4-3.2.7). Moreover, catalysts without loading of titania (neat V-Ti-MCM-41, Cr-Ti-

MCM-41, and Fe-Ti-MCM-41) were tested for the photodegradation of formic acid and yielded no discernible activity.

244 All reaction solutions

were tested for the leaching of 1.0 transition metal ions. It is 0.8 remarkable to note that the

0.6 leaching was negligible for all C/Co 3

OCl TiO loaded catalysts, and it 3 0.4 2 H 6 C was considerable for the neat 0.2 Degussa P25 - Visible TiO /Cr-Ti-MCM-41 - Visible 2 chromium substituted MCM-41 Degussa P25 - UV 0.0 0 20 40 60 80 100 120 140 160 180 materials. On the other hand, Time, min reduced titania-loaded

Figure 3.2.5. Time course of the degradation of 2,4,6- vanadium-substituted catalyst trichlorophenol on the Cr-substituted catalyst of Table 3.2.1, Degussa P25 in UV is shown for leached into the reaction comparison (pH=6, T=25±3°C, catalyst concentration – 1 g/L) medium to a major extent.

The time course of the photodegradation of formic acid on visible-light irradiated catalysts is shown in Figure 3.2.4. Non-reduced (as calcined) chromium substituted MCM-

41 with loaded titania did exhibit activity comparable with that of Degussa P25 in UV

(Figure 3.2.4). For the TiO2/Cr-Ti-MCM-41 the concentration of formic acid decreases steadily for about 60 min, then levels off at approximately 25 % of conversion. This behavior is unexpected, since from a physics point of view the altervalent cations

(especially with the oxidation state of +5) would serve best as dopants [12], but we will see later that the primary doping species in the above catalysts is Cr+6. Iron substituted

MCM-41 sieves loaded with titania also exhibit some photoactivity (Figure 3.2.4). The concentration of formic acid monotonically decreases and levels off at about 6 % of

245 conversion, which was proved in three independent experiments. The maximum

degradation rate is considerably lower than that of neat titania (P25) in UV. Such behavior

can be explained on the basis of the structure of neat Fe-Ti-MCM-41. It was found [8] that

the majority of iron oxide is located outside the framework of MCM-41. This makes

possible the contact of iron oxide (and not iron ions in the framework) with titania. It is

known, however, that the presence of neat iron oxide (III) is deleterious for the activity of

the titania based mixed oxide photocatalysts in visible light [13]. The activity of TiO2/V-

Ti-MCM-41 was also explored under identical operation conditions (Figure 3.2.4).

However, no discernible conversion of formic acid under visible light was detected.

Several other materials

were prepared and tested for 1.0 the photodegradation of formic 0.9 acid for comparison purposes. 0.8 The first group included

0.7 OCl, C/Co 5 TiO /V-Ti-MCM-41 and

H 2 6 C 0.6 Degussa P25 - Visible TiO2/Cr-Ti-MCM-41, but now

TiO2/Cr-Ti-MCM - Visible 0.5 Degussa P25 - UV both photocatalysts were 0 20 40 60 80 100 120 140 160 180 Time, Min reduced in hydrogen. No

Figure 3.2.6. Time course of the degradation of 4- activity was observed, which chlorophenol on the Cr-substituted catalyst of Table 3.2.1, Degussa P25 in UV is shown for comparison was also the case for the same (pH=6, T=25±3°C, catalyst concentration – 1 g/L) non-reduced vanadium-

246 substituted catalyst. The second group of materials was synthesized to compare directly with the most active catalyst, namely, the non-reduced TiO2/Cr-Ti-MCM-41. In particular, siliceous MCM-41 was impregnated with CrO3, Cr(NO3)3, and a mixed chromium-silicon

oxide was also synthesized

to achieve the ratio of Si/Cr

1.0 equal to 80. None of these

0.9 test materials, however,

0.8 exhibited any meaningful

0.7

O, C/Co activity for the 6 H 6 C 0.6 photodegradation of formic

P25-Visible acid in visible light. 0.5 TiO 2/Cr-Ti-MCM-41-Vis P25-UV Therefore, the considerable 0.4 0 50 100 150 photoactivity of non-reduced Time, min

TiO2/Cr-Ti-MCM-41 is not Figure 3.2.7. Time course of the degradation of phenol on the Cr-substituted catalyst of Table 3.2.1, Degussa only due to the presence of P25 in UV is shown for comparison (pH=6, T=25±3°C, catalyst concentration – 1 g/L) chromium ions and titania,

but also due to the special environment provided by the MCM-41 matrix.

The performance of the most active catalyst (TiO2/Cr-Ti-MCM-41) was also tested

for the photodegradation of phenolic compounds in visible light. As seen from Figure

3.2.5, the above specimen exhibits photoactivity to decompose 2,4,6-trichlorophenol. This

is also the case for 4-chlorophenol and phenol (Figures 3.2.6-3.2.7). The activity for the

degradation of these probe molecules in visible light is not as high as that of Degussa P25

in UV. This can be due to the presence of the benzene ring in these phenolic compounds. As

247 the result, the smaller-bandgap photocatalyst with a lower oxidation potential may not be as active in breaking the benzene ring as neat TiO2. Similarly to Degussa P25 in UV, the activity of 25%TiO2/Cr-Ti-MCM-41 in visible light acquires lower values with the smaller number of chlorine atoms in the reactant molecule. This behavior is primarily due to the lesser generation of secondary active chlorine radicals [14]. No deactivation is observed for the above three reactions in visible light, contrary to

the photodegradation of formic

acid. Apparently, the chlorine

atoms present in the molecule

help to re-activate the catalyst.

In order to explain the

promising activity of 25%

TiO /Cr-Ti-MCM-41 for the 2 Figure 3.2.8a. The projected structure of transition metal (example: chromium) substituted MCM-41 visible light assisted photocatalysts photodegradation of organic

contaminants it is helpful to

generalize the surface

characterization results

described above. The

negligible activity of

vanadium-substituted catalysts

Figure 3.2.8b. The projected structure of titania loaded transition metal (example: chromium) is likely to be due to the low substituted MCM-41 photocatalysts degree of interaction of the

248 loaded titania with the V-Ti-MCM-41 matrix. Considering the absorption spectrum of the above catalyst (Figure 3.2.3), one can observe that it is monotonic, thus lacking the absorption shoulders usually found in mixed oxides. This hints at the possibility of low interaction between the support and titania. The marginal activity of 25% TiO2/Fe-Ti-

MCM-41 is probably due to the lack of metal dispersion (Table 3.2.1), and as a result, the lack of the interaction with the loaded titania. Although the absorption of the latter catalyst contain the shoulders characteristic of mixed oxides (Figure 3.2.3), it is very likely that iron oxide did not incorporate into the framework of MCM-41 [15]. Therefore, the

interaction between Fe-Ti-MCM-41 can exist, but the area of the heterojunction may not be

sufficient to produce a significant effect.

When considering chromium-substituted catalysts, all the factors necessary for the photoactivity in visible light are present. There are absorption shoulders in the visible range (Figure 3.2.3), high metal dispersion (Table 3.2.1), specific transition Cr+3 ® Cr+2

(Figure 3.1.8), and surface enrichment of chromium species upon the loading of titania

(Table 3.2.2). Furthermore, from the XPS studies one can observe (Table 3.2.2) that some surface chromium is present in neat Cr-Ti-MCM-41 specimens. This can be attributed to extraframework chromium species. Therefore, two types of interactions are possible in

TiO2/Cr-Ti-MCM-41 or TiO2/Cr-MCM-41, namely Cr-doped mesoporous SiO2 (MCM-

41) with TiO2 and extraframework CrOx with TiO2. The projected structure of the above composite catalyst is presented in Figures 3.2.8a and 3.2.8b. It should be noted that the presence of tetrahedrally coordinated six-valent chromium of the catalyst is essential since its fully reduced form does not exhibit any photoactivity, as described above. The role of

249 three-valent chromium, however, is still unclear since the deactivated catalyst was rich in

Cr (III).

The latter phenomenon

leads to a conclusion that

Cr(VI) is the species

primarily responsible for the

photoactivity in visible light.

Figure 3.2.9 explains the

proposed mechanism of the

charge generation happening Figure 3.2.8. Proposed mechanism of the photo- at the heterojunction of the oxidation on TiO2/Cr-(Ti)-MCM-41 with the molecular excitation of CrO 3 loaded titania with chromium-

incorporated MCM-41. Chromium(VI) doped glasses [16] as well as mesoporous silicas

[17] are known for their tetrahedral coordination of chromium. Such coordination allows

for a special transition under visible light: Cr+6=O-2 ® Cr+5-O-1. Such transition when

happens in contact with amorphous TiO2 can produce an effect similar to that found in

platinum (IV) chloride modified amorphous titania [18]. In particular, the Cr+5 species can

-1 possibly donate an electron into the surrounding TiO2 and O can scavenge an electron

from the surrounding TiO2. In this case, the charge separation will occur, which will lead

to a hole and an electron in TiO2. If this process happens at or near the catalyst surface, the

charges can interact with the surface hydroxyl groups or adsorbed oxygen to produce active

oxygen radicals. As observed before, the photodegradation of certain test compounds leads

to the deactivation of 25%TiO2/Cr-Ti-MCM-41. Since the charge transfer to adsorbed

250 oxygen is the limiting stage in photocatalysis [19], one can expect some accumulation of electrons on the surface of the catalyst. These electrons can possibly reduce Cr(VI) to

Cr(III), which will lead to the depletion of the six-valent chromium and subsequent deactivation of the catalyst. In summary, the above mechanism is an attempt to clarify the mode of interaction between the amorphous titania and Cr-Ti-MCM-41 support. More information on the exact positions and coordinations of atoms is needed for this mechanism to be fully supported experimentally.

CONCLUSIONS

Several transition metal based titania loaded MCM-41 (Si/Me=80, titania content –

25 wt%) materials were tested for the degradation of organics in visible light. The chromium substituted MCM-41 was found to serve as the best support for titania to achieve the highest degradation rates of formic acid, 2,4,6-trichlorophenol, and 4-chlorophenol.

The change in the state of the catalyst during the reaction was observed for Cr substituted molecular sieves, which led to its deactivation. The UV-Vis spectrophoto-metric study revealed the coexistence of Cr (III) and Cr (VI) oxides inside the substituted MCM-41 at the ratio of 0.25. TPR showed the specific interaction of Cr and Ti in these materials, which is believed to contribute to their photocatalytic activity in visible light. XPS revealed increased surface concentrations of Cr ions upon the loading of TiO2, which at the same time allows to minimize the leaching of chromium ions. The mechanism of the photodegradation in visible light on titania loaded transition metal substituted MCM-41 is proposed.

251 ACKNOWLEDGEMENTS

The authors are grateful to the Young Investigator Award of the United States

Department of Army (Grant No. 40414/CH/YIP) and NATO Science for Peace Program

(Grant No. SfP-974209) for their support of this work. The authors thank Mr. Paul France for fruitful discussions.

REFERENCES

1. Khairutdinov, M., Colloidal J., 59, 535 (1997)

2. Pal, B. and Sharon, M., J.Chem.Technol.Biotechnol., 73, 269 (1998)

3. Chatterjee, S., Sarkar, S., and Bhattachatyya, J. Photochem. Photobiol. A: Chem., 81, 199 (1994)

4. Ashokkumar, M. and Maruthamuthu, P., Int. J. Hydrogen Energ. 16, 591 (1991)

5. Serpone, N., Lawless, D., Disdier, J., and Herrmann , J.M., Langmuir, 10, 643 (1994)

6. Sakata, Y., Yamamoto, T., Okazaki, T., Imamura, H., Tsuchiya, T., Chem. Lett., 1253 (1998)

7. Xu, Y., and Langford, C.H., J. Phys. Chem., 101, 3115 (1997)

8. Rey, F., Sankar, G., Maschtmeyer, T., Thomas, J.M., Bell, R.G., and Greaves, G.N., Topics in Cat., 3, 121 (1996)

9. Corma, A., Fornes, V., Navarro, M.T., and Perez-Pariente, J., J. Catal., 148, 569(1994)

10. Davydov, L., Smirniotis, P.G., and France, P., Photocatalytic Degradation of Organic Compounds, Patent Application, Procter & Gamble Case 8197

11. Davydov, L., Smirniotis, P.G., and Pratsinis, S.E., Ind.Eng.Chem.Res., 38, 1376 (1999)

12. Karakitsou, K.E. and Verykios, X.E., J.Phys.Chem., 97, 1184 (1993)

13. Litter, M.I. and Navio, J.A., J. Photochem. Photobiol. A: Chem., 84, 183 (1994)

252

14. Hoffmann, M.R., Hua, I., and Hochemer, R., Ultrason. Sonochem., 3, S163 (1996)

15. Carvalho, W.A., Wallu, M., and Schuchardt, U., J.Mol.Catal.A:Chem., 144, 91 (1999)

16. Koepke, C., Wisniewski, K., Grinberg, M., Majchrowski, A., and Han, T.P.J., J.Phys:Condens.Matter, 13, 2701 (2001)

17. Yamashita, H., Yoshizawa, K., Ariyuki, M., Higashimoto, S., Che, M., and Anpo, M., Chem.Comm., 435 (2001)

18. Zang, L., Lange, C., Abraham, I., Storck, S., Maier, W., and Kisch, H., J.Phys.Chem. B, 102, 10765 (1998)

19. Hoffmann, M.R., Martin, S.T., Choi, W., and Bahnemann, D.W., Chem.Rev., 95, 69 (1995)

253

Chapter 3.3: Sensitization of titania with Cd incorporated MCM-41 support for the enhanced activity in UV light

INTRODUCTION

In the recent past the discovery of silica-based mesoporous molecular sieves

M41S, including the hexagonal MCM-41 [1, 2] offered new opportunities for creating

highly dispersed and more accessible catalytic sites by incorporating transition-metal ions

into their silica based frameworks [3, 4, 5, 6, 7]. Many of the transition metal

incorporated within the framework of mesoporous materials show unique reactivities not

only for various catalytic reactions, but also for photocatalytic reactions under UV

irradiation [8, 9, 10]. An attempt to incorporate cadmium ions into aluminosilicate Al-

MCM-41 during synthesis and by ion exchange was also undertaken [7]. In the present

study, we have synthesized and characterized cadmium containing mesoporous silicate

(Cd-MCM-41) for the first time. Photocatalytic degradation of formic acid has also been

studied over titania loaded Cd-MCM-41. The important advantage of this titania loaded

Cd-MCM-41 is an active catalyst for photocatalytic degradation of formic acid under

mild conditions rather than titania loaded siliceous MCM-41 reported earlier [8, 10].

SYNTHESIS AND CHARACTERIZATION

Cd incorporated MCM-41 with Si/Cd = 80 was prepared by the following

procedure, 9.13 g of cetylammonium bromide [Alfa, 99 % CH3(CH2)15N(CH3)3] was

combined with 20 ml of water and 3.5 ml of ammonium hydroxide (Fisher, 28% NH4OH)

to form solution A; 17.5 ml of LUDOX HS-40 (Aldrich, colloidal silica 40 wt.% suspension in water) was combined with 10 ml water and 9.1 ml of tetraethylammonium

255 Table 3.3.1. Physical and physicochemical properties of the catalysts utilized in the study Surface XRD XPS Pore Area, Catalyst size, d a a Cd 3d 100 0 5/2 Cd/Si m2/g nm nm nm eV

25%TiO2/MCM-41 3.61 4.17 N/a N/a 941.0 4.2 Cd-MCM-41 3.61 4.17 405.2 0.02 703.5 5.2

25%TiO2/Cd-MCM-41 3.67 4.24 405.3 0.07 605.6 3.8 a calculated using a0 = 2d100/ 3 b zero-order reaction rate of UV-assisted photodegradation, conditions: catalyst - 1 g/L, HCOOH – 10 mM; UV light power - 28 W, oxygen flowrate 0.5 L/min, T - 300 K. hydroxide to form the solution B. A pre-determined amount of solution A and 0.5 g of cadmium acetate [Fisher, 99.87%, Cd(OCOCH3)2 ×2H2O] dissolved in 20 ml of water were introduced into solution B under vigorous stirring. After stirring for 30 minutes at

350 K, the resulting gel was transferred to the oven and kept for 72 h at 373 K under autogenous pressure. The aged gel was washed with deionised water and air-dried. The calcinations of the Cd-MCM-41 to remove the occluded template were carried out in air at ca. 823 K for 10 h with the temperature ramp of 2 K/min. The calcined Cd-MCM-41 sample was colorless, indicating the absence of colored CdO species (brown) outside the framework. This result was independently verified by XPS and UV-VIS spectroscopy.

25% titania loaded Cd-MCM-41 was synthesized as reported in our earlier work

[9]. The MCM-41-based supports (typically 1.5g) were dispersed in ~100 ml or isopropanol, and titanium isoproxide was added to achieve 25% loading. The system was dried while stirring at ambient temperature. It was then placed in the oven to dry at 373 K for 8 h. The samples were then calcined in air at 773 K for 3 h with a temperature ramp of 2 K/min.

256 An X-ray powder diffraction patterns (Nicolet, with CuKa radiation; l = 0.154

nm) of calcined Cd-MCM-41 and 25%TiO2/Cd-MCM-41 (Figure 3.3.1) are comparable to

the diffraction patterns of

pure silica MCM-41 [1, 2].

There is a decrease in d100 by

ca. 0.8 nm on Cd-MCM-41

and 0.2 nm on 25%TiO2/Cd-

MCM-41 (Table 3.3.1). The

Intensity (a.u.) unit-cell parameter increases 25% TiO2/Cd-MCM-41 when titania is loaded on Cd- Cd-MCM-41 MCM-41, demonstrating that 2 3 4 5 6 7 2q cadmium is incorporated in

Figure 3.3.1. XRD Patterns of the catalysts employed in the framework. The latter the present study. phenomenon can be also be justified by the fact that no peaks corresponding to CdO were observed on the diffractogram.

X-ray photoelectron spectroscopy (XPS) of Cd-MCM-41 and 25%TiO2/Cd-

MCM-41 show that cadmium is well dispersed in MCM-41 framework. The binding energy values of Cd 3d5/2 photoelectron peak (Figure 3.3.2 right) and surface atomic concentration ratio between Cd and Si for Cd-MCM-41 and 25% TiO2/Cd-MCM-41 are depicted in Table 3.3.1. The binding energy value of Cd 3d5/2 photoelectron peak at 405.2 eV suggests that cadmium is dispersed in its metallic state in both cases, which agrees

257 well with the value reported in the literature [11]. As shown in Table 3.3.1, the surface

atomic concentration ratio of Cd/Si is increased 3.5 times when titania loaded on Cd-

MCM-41. At the same time, the surface concentration of silicon atoms significantly

decreases with the loading of titania (Figure 3.3.2 left). The possible explanation for these

data is diffusion of Cd ions towards the surface of MCM-41 pores and interacting with

the loaded titania.

Si 2p Cd 3d5/2 Cd 3d 3/2

a

Intensity (a.u.) a b b

100 102 104 106 400 405 410 415

Binding energy (eV)

Figure 3.3.2. XPS spectra for Si 2p (left) and Cd 3d (right) for the catalysts and supports utilized in the study: a- Cd-MCM-41, b – 25%TiO2/Cd-MCM-41

To the best of our knowledge, these diffusion effects have never been reported in the literature. The XPS results perfectly agree with pore size distribution and BET surface area values represented in Table 3.3.1. The average pore diameter and BET surface areas decreased when titania was loaded on Cd-MCM-41. One can conclude that the loading of titania transfers cadmium to the surface of the catalyst and partially blocks the pores of

Cd-MCM-41. Furthermore, the color of the composite catalyst remained white (as also shown in Figure 3.3.3), which is contrary to the color of bulk CdO. This is due to high

258 dispersion and possibly to a different coordination number of cadmium. Inside the

framework of MCM-41 cadmium has to attain a tetrahedral coordination, whereas CdO is cubically coordinated. Such arrangement may enhance the diffusion of ions and contribute to the sensitization of the composite catalyst.

The diffuse

1.5 reflectance UV-Vis spectra Cd-Ti-MCM-41

TiO2 /Cd-Ti-MCM-41 of the catalysts were

TiO2 /MCM-41 examined in the range 200- 1.0

850 nm in order to assess

the light absorption 0.5

Absorbacen, a.u. characteristics of the

catalysts. One can observe

0.0 200 300 400 500 600 700 800 900 from Figure 3.3.3 that the Wavelength, nm absorption of UV light is Figure 3.3.3. DR UV-Vis spectra of the catalysts employed in the study very strong on

25%TiO2/Cd-MCM-41

when compared to the TiO2/MCM-41 and Cd-MCM-41 within the 280 to 400 nm wavelength region. A shoulder at about 300 nm specific for Cd-MCM-41 only is also observed. Its position is considerably different from the one observed on the spectrum of bulk CdO (not shown). This hints at a possibility that the photocatalytic activity of the

Cd-incorporated catalyst may be higher than that of the siliceous MCM-41.

259 PHOTOCATALYTIC ACTIVITY

As mentioned above, different physicochemical charac-teristics confirmed that in

Cd-MCM-41 and 25%TiO2/Cd-MCM-41 catalysts, the Cd ion is highly dispersed in the silica-based framework structure. The catalytic properties of Cd-MCM-41, 25%TiO2/Cd-

MCM-41 and 25%TiO2/MCM-41 were inspected by carrying out the photocatalytic degradation of formic acid (as an example of organic pollutants) under UV light in an aqueous medium.

The concentration

of formic acid was 1.00 analyzed using a 0.95 conductivity meter (VWR 0.90 Scientific, cell constant 10 0.85

cm-1). The initial rate of the 0.80 HCOOH, C/Co 25%TiO /Cd-Ti-MCM-41 reaction over Cd-MCM-41 0.75 2 25%TiO /MCM-41 2 Cd-Ti-MCM-41 0.70 and 25%TiO2/Cd-MCM-41

0 20 40 60 80 100 120 140 160 180 was the same until 15 Time, min minutes of reaction. Then Figure 3.3.4. Photocatalytic activity of the catalysts employed in the present study Cd-MCM-41 deactivated,

and the photodegradation stopped. Surprisingly, the rate of the reaction over 25%TiO2/Cd-MCM-41 was found to be 1.5 times as fast as that of 25%TiO2/MCM-41 even after 3 h (Figure 3.3.4). This finding is unique since other researchers [10] postulated that the presence of transition metals in the framework of MCM-41 is deleterious for the activity of titania-loaded

260 MCM-41 catalysts in UV light. The difference of our systems arises from the fact that the transition metal ions (such as cadmium) can diffuse to the surface, thus creating a new energy band. Since the valence band of CdO is lower than that of TiO2 [12] (+3.2 for

CdO and +2.8 for TiO2 V vs NHE), such arrangement may enhance the oxidation potential of photogenerated holes. Conclusively, our catalytic studies indicate

25%TiO2/Cd-MCM-41 to be a potentially useful catalyst for degradation of aqueous organic pollutants under UV light.

CONCLUSIONS

Cd incorporated MCM-41 material has been successfully synthesized, characterized, and catalytically tested for the first time. It was found that titania loaded

Cd-MCM-41 is more active than titania-loaded siliceous MCM-41 for the photocatalytic degradation of formic acid under UV light. The surface characterization of these materials revealed that the enhanced activity is due to the diffusion of cadmium ions to the surface of the catalyst during treatment.

ACKNOWLEDGEMENT

This work was supported by the United States Department of Army (DOA) through Young Investigator Program (Grant No. DAAD 19-00-1-0399).

REFERENCES

1 C.T. Kresge, M.E. Leonowicz, W.J. Roth, J.C. Vartuli and J.S. Beck, Nature, 1992, 359, 710.

261

2 J.S. Beck, J.C. Vartuli, W.J. roth, M.E. Leonowicz, C.T. Kresge, K.D. Schmitt, C.T.-W. Chu, D.H. Olson, E.W. Sheppard, S.B. McCullen, J.B. Higgins and J.L. Schlenker, J.Am. Chem. Soc., 1992, 114, 10834.

3 R.S. Mulukutla, K. Asakura, S. Namba and Y. Iwasawa, J. Chem. Soc., Chem. Commun., 1998, 1425.

4 P.T. Tanev, M. Chibwe, and T.J. Pinnavaia, Nature, 1994, 368, 321.

5 D. Zhao, and D. Goldfarb, J.Chem. Soc., Chem. Commun, 1995, 973.

6 A. Corma, A. Martinez and V. Martinez-Soria, J. Catal, 1997, 169, 480.

7 W.Y. Lin. Q.G. Cai, W.Q. Pang, and Y. Yue, J. Chem. Soc., Chem. Commun., 1998 2473.

8 L. Davydov, E.P. Reddy and P. Smirniotis, J. Catal., 2001, submitted.

9 P. Smirniotis, L. Davydov, E.P. Reddy, and P. France, Industrial Applications of Zeolites, 2000, 233-242.

10 Y.M. Xu and C.H. Langford, J. Phys. Chem. B, 1997, 101, 3115

11 D. Briggs, M.P. Seah, (Eds.), Practical Surface Analysis, 2nd Edition, Vol. 1 Auger and X-ray Photoelectron Spectroscopy, Wiley, New York, 1990.

12 P.V. Kamat, Chem.Rev., 1993, 93, 26

262 CONCLUSIONS AND FUTURE WORK

The major results presented in this dissertation highlight the importance of the

three research directions to make photocatalysis a feasible alternative to current pollution

abatement techniques. It has proven successful to apply conventional reactor design tools

to photocatalytic systems. The concepts of a differential reactor and effectiveness factors

have been shown to be applicable to light distribution as opposed to concentration or

temperature distribution commonly encountered. Furthermore, the utilization of reaction

engineering knowledge allows to increase the reaction rates of photocatalytic processes

by careful reactor design. The enhancement of reaction rates is essential for

photocatalysis to become a technique to be reckoned with. One more reactor design approach to follow is to increase the light-catalyst contact time to achieve deeper conversion per pass. A differential reactor is something that is always needed for the adequate and reproducible representation of the reaction rates. While commonly applied for the temperature and concentration distributions, this concept has a different meaning for photocatalytic systems. This is because the light transfer (and not heat or mass transfer) is the major issue in slurry photocatalytic reactors. Proceeding toward such differential volume (by reducing the reaction zone thickness and length) can minimize the light attenuation, thus reducing the non-uniformity of the radiation field and yielding intrinsic photocatalytic properties. The utilization of effectiveness factors in photocatalytic systems proposed in the present work can expand the knowledge of photoreactors and enable engineers to assess the performance of each particular system on the common basis, namely, intrinsic kinetic constants and turnover frequencies. When a turnover frequency of a particular photocatalyst is obtained from the intrinsic and not

263 apparent reaction rate, it becomes comparable for different reactor sizes and geometries.

This advantage can only be brought about by utilizing photocatalytic effectiveness factors. Apart from the adequate description of the photocatalytic process, the reaction engineering knowledge allows for the optimization of photoreactors. More specifically, using a rigorous kinetic model (which is non-linear and non-homogeneous) accounting for the balance of species in a continuous photoreactor gives the possibility to optimize the axial radiative distribution in order to achieve the maximal conversion of the reactant.

It was found that more radiation at the beginning of the reactor produces a more significant reactor output. The use of such optimal irradiation strategy can yield significant enhancement of the reactor output in comparison with the uniformly irradiated photoreactor. This is especially true for the multipass low-conversion photoreactors with recycle. An experimental study would be ideal to prove the theoretical optimal distribution of light using, for example, a number of lamps of decreasing power in series.

Another important feature of photoreactors is catalyst aggregation. The use of ultrasound as an auxiliary process in photocatalysis can reduce the effective aggregate size and reveal more surface area of the catalyst to be subjected to radiation. A synergistic effect is observed for a sonophotocatalytic system. Such effect is dependent on the catalyst concentration and passes through a maximum. Furthermore, the presence of ultrasound in a photoreactor can enhance the generation of reactive oxygen species. More experiments would be needed to elucidate the relative extent of the generation of different reactive oxygen species using the probe reactions developed in the present study.

The importance of kinetic modeling in photocatalysis is difficult to overestimate.

It allows uncoupling the different phenomena at hand, and the application of kinetic

264 modeling to photocatalysis yielded some important information. First, the measure of

active oxygen radical generation by photocatalysts has been assessed. It provides a

qualitative measure of the oxidative power of photocatalysts. Second, the individual

surface reaction and electron-hole recombination rates have been determined. The ratio of these rates is about 0.01 at best, hinting at the inherent limitations of the technique.

Moreover, the larger particle size photocatalyst experience a much larger degree of

electron-hole recombination. Third, kinetic modeling can suggest the pathways for the

rate enhancement. In particular, the use of ultrasound, zeolitic supports, and

nanostructured materials can provide such enhancement. More research is needed into the

experimental and theoretical understanding of the fundamentals, which will lead to more

efficient photocatalytic processes. The kinetic role of oxygen also has not been

sufficiently explored, thus leaving unclear their ultimate fate in the photocatalytic

process.

Yet another important issue is catalyst design. The optimally designed catalyst

would combine high activity, chemical stability, and energy efficiency. The supported

sensitized catalyst described in the present dissertation can effectively utilize visible light

with the reaction rate comparable to best UV-driven photocatalysts. This opens the way

to the synthesis of a whole new class of transition metal substituted mesoporous materials

that can impart such property to the loaded titania. The transition metal inclusion is

presumably tetrahedrally coordinated, something that is not achievable in bulk oxides. It

is of great importance, however, to use powerful X-ray absorption techniques to elucidate

the true reasons of exceptional activity of one group of transition metals over the other

group.

265

Appendix 1: Liquid Phase Photocatalysis - Experimental Details

1. Light Sources

· 450 W mercury vapor medium pressure lamp (Ace Glass#7825-34) manufactured

by Conrad-Hanovia. Lamp emissive length – 13 cm, diameter – 2 cm. Radiative

power within 300-400 nm wavelength range – 28 W (info provided by Ace Glass,

Inc.) Uniform radiant intensity within its emissive length – 13 cm opening used in

differential reactor setup. Aluminum foil was used to mask the ends of the lamp

and cut the non-uniform regions of the radiation profile. 450 W power source

(Ace Glass#7830-58) used with the lamp. Warm-up time – 5 min. Time to reach

steady-state intensity of radiation – 1-2 min depending on its initial temperature.

The experimental time started after the lamp broke down (the power source

started droning). Repetitive tickling of the power source was used to break down

the hot lamp.

· 200 W mercury vapor medium pressure lamp (Ace Glass#7825-32) manufactured

by Conrad-Hanovia. Lamp emissive length – 12.5 cm, diameter – 2 cm. Radiative

power within 300-400 nm wavelength range – 8 W (determined by radiometry

when 450 W power source was used). Worked at ~20% overvoltage with the 450

W power source. Warm-up time – 5 min. Time to reach steady state radiant

intensity – 2 min (determined by radiometry). Repetitive tickling of the power

source was used to break down the hot lamp. Very non-uniform radiation profile –

2 cm opening was used in the differential reactor setup. The above method was

used to eliminate the non-uniformity of the radiation profile. No increase in local

267 radiant intensity was observed when 2 cm opening was used. The experimental

time started after the lamp broke down (the power source started droning).

· 100 W mercury vapor medium pressure lamp (Ace Glass#7825-30) manufactured

by Conrad-Hanovia. Lamp emissive length – 8 cm, diameter – 2 cm. Radiative

power within 300-400 nm wavelength range – 1.53W (info provided by Ace

Glass). This lamp was not used for a differential reactor setup. 100 W uncased

Ace Glass power source was put together by Mr. Lev Davydov. This power

source could not be applied for 200 W lamp: the arc was not sustainable and

broke after ~ 1 min. Warm-up time for 100 W lamp – 5 min. Time to reach

steady-state radiant intensity – 5 min (output voltage measured by a conventional

tester). The cold lamp would not break down. Depending on the room

temperature, the lamp sometimes had to be pre-warmed in the oven at ~100 C.

The hot lamp required repetitive tickling of the power source to be broken down.

The experimental time started after the output voltage reached 90-95 V

(corresponding to the steady-state radiative output).

2. Photocatalytic reactors

· Conventional photocatalytic reactor (FR) was manufactured by Ace Glass, Inc.

consisted of a reaction vessel (ID 66 mm) and lamp immersion well (OD 54 mm).

The latter was inserted into the reaction vessel and sealed by means of 60/40-glass

seating. Several reactor outlets were used (see the schematic in the main section).

One outlet was occupied by a funnel, which would allow the excessive oxygen to

go out of the reactor at the same time not allowing anything to go in. The air tube

268 connecting the glass frit of the aerator with the oxygen tank employed the second

outlet. The third outlet was used for in-situ sampling. A plastic tube was inserted

into it. This tube was routed toward the wall of the photocatalytic safety cabinet

thus going through it and protruding to the outer side of the cabinet by about 7-10

cm. The impeller operated in the outer sleeve of the reactor and provided vigorous

circulation of the reacting fluid and good oxygen hold-up in the system. It was

connected to a conventional electric motor with variable RPM characteristics.

· Variable reaction zone photocatalytic setup was constructed “in-house”. Plexiglas

tubes (United Plastics Corp.) of different inner diameters (2.25”, 2.5”, 2.75”) were

cut to 14” length. These tubes constituted the outer wall of the reactor. A plexiglas

plate was cut to the size of 3.5x3.5” and served as a bottom of the reactor. It was

attached to the tube using Weld-On #16 acrylic glue. The glass frit (aerator) was

glued into the hole at 1” above the bottom of the reactor in order to leave

sufficient space for a magnetic stirring bar to operate. It was protruding through

the outer wall by 1”. In order to provide circulation of the reacting suspension,

holes 1” from the top and bottom of the reactor were drilled and copper unions

0.5x0.375” were twisted in. They were later connected to the pump by Tygon

plastic tubing. This photocatalytic setup utilized the same immersion well and

lamps as the conventional photocatalytic setup. The top of each reactor was fitted

to the 60/40 seating of the jacket by filling the void with liquid epoxy resin,

which, after hardening, yielded a satisfactory seal. These reactors operated with

the laboratory peristaltic pump, which provided circulation of the suspension. For

the purpose of degassing of the entering suspension and also for sampling, an

269 intermittent vessel was placed between the exit of the reactor and the entrance of

the pump.

3. Photocatalytic runs

· Conventional reactor

The runs had the following sequence:

- pouring the appropriate amount of water (0.65 L) into a 1L beaker

- adding the weighed amount of phenol to achieve 2 mM

- magnetic stirring

- adjusting pH with 0.5 M H2SO4 to achieve ~3.75 (approximately 1 droplet per 0.1L

of solution), monitoring pH with a portable Fisher pH-meter

- adding the appropriate amount of the catalyst, while continuing to mix

- mixing for 1 hour

- ultrasonication for 10 min in an ultrasonic cleaner (Labline LC-20H), while

suspending the beaker on a ring made of plastic foam, so its bottom does not touch

the bottom of the bath

- transfer the liquid into a pre-started reactor, shake while carrying

- start the oxygen flow

- pour the contents of the beaker into the reactor through the lateral hole using a funnel

- adjust the impeller speed and oxygen flowrate

- turn on the lamp and start the stopwatch

- withdraw samples through the sampling tube emanating from one of the lateral holes.

The reactor was ordered from Ace Glass (super-mix photochenical reactor), but has

270 since then been discontinued. The lateral impeller (all glass) was supplied by Fisher

along with the standard teflon bushing.

Variable reaction zone photocatalytic setup

- all steps remain the same through the ultrasonication

- start the peristaltic pump, magnetic mixers, and oxygen

- pour the liquid gradually into the intermittent vessel

- put the sampling tube into the intermittent vessel

- turn on the lamp

The corresponding outside plastic tubes made by US Plastics, Inc., glued by the acrylic glue (same manufacturer), the oxygenator (frit glass) was supplied by Fisher and put in using epoxy resin (Beck’s Hardware). The brass fittings for inlet and outlet were ordered from Cincinnati Valve Co. (1/4” pipe thread on one side, 1/8’ regular thread on the other).

4. Reagents

A large number of reagents were tested as candidates for the probe molecules to

perform photocatalytic degradation. Since an open system was used as a photoreactor,

such probe molecule has to combine excellent solubility in water, a polar group, and

low partial vapor pressure. As a result of this condition, the original tests performed

on benzene, toluene, xylene, and monochlorobenzene resulted in the enhanced

reactant stripping from the aqueous solution. On the contrary, if a very small

concentration of the above organic compounds was used, the extraction with an

271 organic solvent (chloroform) was required. The procedure typically involved the

addition of 2 mL of chloroform to 2 mL of solution. Then vigorous shaking and

overnigh equilibration was applied. It was followed by drying of the solvent in order

to increase the concentration of the analyte. During the drying the volume of the

sample was reduced to 0.5 mL. Thus prepared chloroform solutions were injected into

a gas chromatograph and the results showed the presence of the impurities of the

solvent. Therefore, this technique is plagued with significant flaws and was

abandoned. Only the solutes that combine the above three criteria were used, such as

phenolic compounds and salicylic acid.

5. Analyses

The analysis of phenol, 4-chlorophenol, and 2,4,6-trichlorophenol were performed using HP GC 6890 equipped with a TCD and FID. The column was capillary (HP-…).

The samples were collected into Ependorf plastic single-use capped tubes (1 mL). For each injection, a hole was punctured in the cap, and 1 uL of liquid was taken by a 10 uL syringe (Fisher) 10 times pre-washed with the liquid. The syringe was then inserted into the injection port of the GC (Splitless mode). The robust motion pushed the liquid out, after which the syringe was quickly withdrawn from the port and the start button pressed on the GC. Several approaches were utilized in analysis. First, internal reference

(monochlorophenol) was added to each sample and the ratio of the peak areas of 2,4,6-

trichlorophenol and 4-chlorophenol on FID served as the “concentration” of the former.

The method utilized for this was 3clph1.d (the printout attached). A typical

chromatogram of this analysis is also attached. The peak at 2.5 min corresponds to

phenol, and the one at 8 min is 4-chlorophenol. The second approach was to use water

272 detected by TCD as the internal standard. In this case the duration of each run could be

shortened. The method was levphe1.d, it was utilized to detect phenol (the duration was 5

min instead of 12, otherwise the same as 3clph1.d). A typical chromatogram is also

attached at the end of this section. The phenol peak corresponds to 2.5 min on FID, and

water - 2 min on TCD. The ratio of these peaks was calibrated against the phenol

concentration exhibiting a straight-line dependence.

The analyses of aqueous reagents (toluene, phenol, 4-chlorophenol, and salicylic acid) pertaining to chapters 1.4, 2.3, and 3.2 were made using the UV-Vis

spectrophotometer Shimadzu 2501PC. The wavelengths used were 204, 270, 274, and

296 nm, respectively. Deionized water was used as a standard in all experiments. Quartz

cuvettes with the pathlength of 1cm and capacity of 1 mL were used. Triple washing with

deionized water with subsequent drying by compressed air was necessary for the

complete removal of the species present in the cell. The sequence of analysis was the

following: starting from the higher concentrated solution (0 min) toward the lower

concentrated solution. Initially two measurements were taken for each sample, but due to

the excellent reproducibility of analysis this was found unnecessary. For each reagent a

calibration line was created and saved as a “standard” file, such as cro4.std. Then within

the “standard” file the “unknown” mode was evoked. This way the machine calculates

the concentration of the compound automatically.

The analyses of Chapter 2.1 involved conductivity measurements of the solutions

of formic acid as well as the spectrophotometric measurements of Fe(III) ions. The

concentration formic acid was measured by a conductivity meter (VWR Scientific Model

2052, golden cell, experimental cell constant 10.25 cm-1). It was calibrated by standard

273 solutions of formic acid with the concentrations ranging from 0.001 to 20 mM. A

quadratic curve passing through zero and the experimental points was fitted with the

accuracy of 0.999; this curve was used as the conductivity-concentration correlation. The

analytical procedure involved several steps. First, about 2-3 mL of the solution were

placed into the 1 cm test tubes. Then the conductivity probe was immersed into the tube.

To achieve the equilibrium, three upward and downward movements of the probe as well

as 30 second equilibration time was necessary. Then the reading was taken from the

conductivity meter. The meter was used in the temperature correction mode, which

allowed to adjust the conductivity for the fluctuations in the solution temperature. The

reaction mixtures were also tested for the presence of intermediates at different time on

stream by direct injection into GC (HP 5890 Series II) equipped with a MS (HP-5972).

No detectable amounts of intermediate compounds were found.

A modified Fricke dosimeter was used to determine the overall concentration of

reactive oxygen species in irradiated aqueous suspensions of titania, as described in

Chapter 2.1. It consisted of 5 mM of FeSO4 (Fisher), 10 mM of CuSO4 (Fisher), and 20

3+ mM of H2SO4 (Fisher). During the reaction with reactive oxygen species Fe ion is

produced, and it was quantified by spectrophotometry (Shimadzu U160) at the

wavelength of 304 nm. Standard quartz cuvettes with the pathlength of 1 cm were used.

The absorbance by Fe3+ was correlated with the concentration of Fe2+ reacted and further

used for calculations. Although considerably higher absorbance of Fe (III) ions is in the

mid-UV range (200+), the above range could not be used since cupric ions also highly

absorb at these wavelengths.

274 In the experiments of the photooxidation of nitrite ions, 30 mM concentration of

NaNO2 (Fisher) was subjected to photolyses by different powders of titania under the same operating conditions. The unreacted nitrite was determined by titration using 0.005 mM KMnO4 (Fisher) solution. The end point of titration was determined colorimetrically

at the wavelength of 535 nm. In a typical procedure, exactly 2 mL of the solution were

taken and placed into 1 cm tubes. A 0.5 cm magnetic stirring bar was placed into the tube

as well. The permanganate solution was added dropwise from a burette, and once the

solution was beginning to colorize, the tube was inserted into the colorimeter (Spectronic

20). When the absorbance of 0.5 was achieved, the titration stopped, and the reading was

taken.

275

APPENDIX 2: Gas phase photocatalysis – experimental details

4 To fume hood 3 2 3

6 5 1

9

8

7 Figure A.2.1. Schematic representation of the experimental set-up. 1- air cylinder, 2 – pressure regulator, 3 – mass flow controllers, 4 – liquid infusion pump, 5 – saturator with water, 6- 6-way crossover valve, 7 – photochemical safety cabinet, 8 – photocatalytic reactor, 9 – GC HP 6890 equipped with gas sampling valve.

Photocatalytic experiments were performed using four different samples of titanium dioxide: Hombikat UV 100 (Sachtleben Chemie GmbH), Degussa P25 with 50

2 2 m /g, Degussa P25 with 75 m /g, and homemade TiO2 (Chernogolovka, Russian

Federation). The specific surface area and pore volume distribution were measured with

ASAP 2010 instrument (Micromeritics) using degassing at 423 K for 4 h. Before placing

the catalyst into the reactor it was heated at 383 K for 10 min to remove physisorbed

water. Thin catalyst film deposited on a stainless steel round plate located at the bottom

of the reactor was the formulation used in the study. In order to prepare film of titanium

277 dioxide a conventional aqueous slurry method was used. Granules of TiO2 were prepared by pressing powder at pressure 32,000 bar for 30 second, grinding the tablet, and sieving with standard sieves.

Diethyl sulfide (Fluka) with assay >98% and deionized water were used. Zero- grade air (Wright Bros.) was utilized to prepare reactor feed. Photocatalytic degradation of diethyl sulfide over a thin film of photocatalysts was accomplished in a flow reactor.

The set-up used in the present study is schematically represented in Fig. 1. All tubing used was made of stainless steel 316. Zero grade air (99.9%) from a cylinder was passed through two mass-flow controllers (MKS), one of which was connected to a saturator filled with water. The second line was used to provide additional air and thus alter the concentration of water fed into the reactor. A 74900 Cole-Parmer liquid infusion syringe

pump was employed to inject liquid reactant into the air stream at the appropriate rate.

The resultant gas mixture entered into the photocatalytic reactor placed into a

photochemical safety cabinet for protection of personnel from UV light. Fig. 2A shows a

cross section of the photocatalytic reactor used. The main cylindrical stainless steel body

of the reactor has length 7.6 cm and internal diameter 3.8 cm. Two 4-Watt fluorescent

lamps (Wiko, Japan) with a maximum in emission spectra at 350-360 nm (information

provided by the manufacturer) illuminated the reactor from the top through a pyrex

window sealed with Teflon covered o-rings. The reaction mixture enters the bottom of

the reactor through four nozzles forming 900 angle with each other. The inlet gas streams

strike the catalyst which is either lying / deposited on the bottom of the reactor or is

vibrofluidized. The temperature of the catalyst bed was measured by means of a thin

278 thermocouple touching the bed. The reaction mixture exits the reactor through four nozzles situated at the top of the reactor. To prevent the entrance of catalyst granules

1

2 3

A 4

5 6

7

B 8 9

10 Figure A.2.2. (A) Cross sections of the photocatalytic reactor with the bottom for vibrofluidization and (B) the bottom for ultrasonication. 1 – two UV lamps, 2 – pyrex window, 3 – four gas outlets at 90o with each other, 4 – thermocouple, 5 – four gas inlets at 90o with each other, 6 – Teflon membrane with catalyst over it, 7 – speaker, 8 – stainless steel plate, 9 – Teflon membrane, 10 – tip of the ultrasound transducer.

279 into the nozzles, a controlled amount of fiber glass was placed inside them. The flow rate

of 20 cm3/min was used in all experiments unless specified otherwise. Test experiments

were performed to ensure that the inflow and outflow of gases through all the nozzles are

distributed uniformly.

The reactor was a custom-made stainless steel cylindrical vessel (Peppers

Machine Shop, Cincinnati, OH) illuminated through a Pyrex window placed on the top of

the reactor. Before a typical run, 25 mg of photocatalyst was deposited on stainless steel

plate of circular area 9.1 cm2 using aqueous slurry method. Then, the catalyst was dried at

1100C to remove the excess of physisorbed water, and the plate was attached to the

bottom of the reactor. Four nozzles at the bottom of the reactor provided a uniform feed flow pattern striking the catalyst. In order to prepare feed stream with certain diethyl

sulfide and water vapor concentrations, zero grade air from a cylinder was passed through

a water saturator, diluted, and then diethyl sulfide was added using a 7490 Cole-Parmer

liquid infusion pump. In one set of experiments, the saturator was filled with 30%

0 solution of H2O2 instead of water. All the gas lines were kept at ~65 C with the help of

heating tapes in order to avoid condensation and/or adsorption of reactants and products.

The reactor feed flow rate used in all experiments was 20 cm3/min. The concentration of

diethyl sulfide in the feed stream was 368 ppm. Two types of light sources were

employed to illuminate the catalyst: two 4 W fluorescent lamps (Wiko, Japan) and 450 W

mercury lamp (Hanovia). They provided light intensity at the catalyst’s level in the

reactor at 1.1 and 10 mW/cm2, respectively. Prior to starting the photocatalytic

experiments, the catalyst in the reactor was kept in flow of reactants overnight in order to

reach adsorption equilibrium.

280 Qualitative analysis of reactor feed and effluent streams was performed using on-

line gas chromatograph HP-5890 equipped with a CP Poraplot U column and mass selective detector (HP 5972). The GC oven temperature program included a ramp from

35 to 190 C at 6 C/min and constant temperature at 190 C for 30 min. For the identification of the surface products adsorbed on the catalysts, extraction with isopropanol was performed. The resultant solution was analyzed on HP-5 column employing the temperature ramp 35 to 190 0C at 4 0C/min. Quantitative analysis of gas

streams was done by an on-line gas chromatograph HP-6890 equipped with the CP

Poraplot U column in series with HP-5 column (Part No. 19091J-413), flame ionization

(FID) and thermal conductivity detectors (TCD). The chromatograph was calibrated for

CO2, diethyl sulfide, ethanol, acetaldehyde, diethyl disulfide, ethylene, and water by

varying the flowrate of the corresponding liquid compound through the syringe. The

calibrations were performed by on-line injection of the mixtures of the respective gases

with air into the GC.

Additional qualitative information about the product distribution was obtained by

passing the effluent of the reactor through the solution of bromine (Fluka) in chloroform

(Fisher) for 48 hours. This allowed to test the products of the reaction for the presence of

C=C bonds by observing the discoloration of bromine as well as analyzing the solution

on the GC/MS for the presence of brominated compounds. Gas chromatograph Shimadzu

GC-17A equipped with a capillary column (Supelco) and mass-spectrometer Shimadzu

QP-5505A were used for this analysis. The solution was directly injected into the splitless inlet of the GC.

281 A pre-calibrated IL 1700 research radiometer/photometer (International Light,

USA) connected with a flat detector (International Light, Inc., Model SED033 #3435) was used to measure light intensity. The angular response of this detector was 180 degrees, allowing to measure the true radiant flux density. Two light intensities at the bottom of the reactor were used in experiments: 0.85 and 1.0 mW/cm2.

282 APPENDIX 3: Radiation Model - Computation Details

The kinetic approach of Chapters 1.2 and 2.1 involved a two-flux radiation model.

Shuster-Schwartzschild approximation for light transfer (two-flux model) is based on the following assumptions:

· The light propagation is isotropic

· The absorption of light takes place as well as scattering, and these two phenomena

are accounted for by two independent coefficients (Es and Ea)

· Light can be grouped into two fluxes: backward and forward

· Forward flux at the entrance of the reaction zone is known

· Backward flux at the exit of the reaction zone is equal to zero

Under the above assumptions the balance of photons in any differential volume of the absorbing and scattering medium will acquire the following form described below. For

the simplicity of the derivations we will take G = Gforw + Gback and G = Gforw - Gback,

2 where G denotes radiant flux density in W/cm and Cx denotes the concentration of solids

Planar geometry:

dG ^^ += dZ E a C x G E s C x G ^ d G ^ = dZ E a C x G ^ (0 GG () 0) =- 2 ^ (1 GG () 1) =+ 0

One can deduce from the first equation that

283 ^ 1 dG G = ()+ dZ C x E a E s

1 d 2G - ()+ dZ E a C x C x E a E s

d 2G =- a ( + )C 0 dZ E a EE s x G

( ) ( ) -+-+ dd ZZZZ G C x aa EEE s C x aa EEE s +=+= C e C e C e C 2121 e

-dd ZZ ^ - dd d -- dddd G = C e C 21 e = a -=- ZZZZ )()( ()+ ()+ C e C e C e C 2121 e C x E a E s C x E a E s

a( ) =--+ 2 C C C C2121 -- dddd a =--+ 0 C e C e ()C e C 2121 e

( ) ( aa ) =++- 211 C C 21 -dd ()()aa =-++ 011 C e C 21 e

d (1+ a) -= e CC 12 -d e ()1-a

284 d æ ()1+a 2 ö ç()1 a -- e ÷ = 2 C1ç -d ÷ è e ()1 -a ø

-d æ ()12 -a ö = ç e ÷ C1 ç - dd 22 ÷ è ()()ee 11 +-- aa ø

-d æ ()12 + a ö -= ç e ÷ C 2 ç - dd 22 ÷ è ()()ee 11 +-- aa ø

( 1 dd () 1-- ) ( ) (11 +-- aa ) ZZ = 2 ee G - dd 22 e ()()e 11 +-- aa

In cylindrical geometry one must account for the dissipation of light due to the expansion of the reaction zone: d Gr )( ^^ += dr E a C x Gr E s C x Gr ^ d Gr )( = dr E a C x Gr

G R G21 )()( =- 2RR III

G R G R0201 )()( =+ 0

^ 1 d Gr)( G r = ()+ dr C x E a E s

^ If we denote Gr G1 Gr == G2

285

' G1 G 2 = ()+ C x E a E s

Then we can follow the “planar” derivations:

-dd += rr G C e C 211 e

-dd a -= rr G (C e C 212 e )

-- dddd a RRRR IIII =--+ 2 C e C e (C e C 2121 e ) R I -- dddd a RRRR 0000 =-++ 0 C e C e ()C e C 2121 e

-dd ( ) RR II ( aa ) =++- 211 C e C21 e RI -dd ()()RR 00 aa =-++ 011 C e C 21 e

d R0 (1+a) -= e CC 12 -dR0 e ()1- a

dR0 2 d (1+ a) -d R ()1 a -- e RII = 2 C e C11 -dR0 e RI e ()1-a

-d (12 -a) R0 = RI e C1 )( 2 dd 00 --- RRRR II )( 2 e ()()e 11 +-- aa

286 d (12 +a) R0 -= RI e C 2 )( 2 dd 00 --- RRRR II )( 2 e ()()e 11 +-- aa

dd --- ( ) R r (11 +-- aa ) R00 r)()( = 2 RI e RI e G1 )( 2 dd 00 --- RRRR II )( 2 e ()()e 11 +-- aa

dd --- ( ) R r (11 +-- aa ) R00 r)()( = 2 R I ee G )( 2 dd 00 --- RRRR II )( 2 r e ()()e 11 +-- aa

A sample Mathematica 3.0 source code (scattP25-2.nb) used for the calculations of radiant fluxes follows below. The first three cells (titled 450W Lamp, 200W lamp, and

100W Lamp) calculate the radiation profiles and absorbed radiant fluxes in the conventional photoreactor and its effectiveness factor. They are utilized in Chapter 1.2.

The last two cells calculate and compare the effectiveness factors for the conventional and differential photocatalytic reactors

287

288

289

290

291

APPENDIX 4: Radiation Field Optimization - Computation Details

The kinetic approach of Chapter 1.3 involved a non-linear non-homogeneous system of ordinary differential equations based on the rates defined in Chapter 2.1. A sample Mathematica 3.0 source code (comparison-aux.nb) follows below. It is used for the calculations of optimal axial radiant distributions according to the “high” and “low” conversion cases of chapter 1.3 (cell 1). Cell 2 compares the output of the uniformly irradiated photoreactor for the above cases. Cell 3 introduces the theoretically optimal radiation distribution and calculates the reactor output. Cell 4 compares the ratios of the outputs of the uniformly irradiated photoreactor to that of the optimally irradiated photoreactor for “high” and “low” conversion cases. Cells 5 and 6 are auxiliary, and they test the parametric sensitivity of the model. Cell 7 calculates the optimal parametric space for the low conversion case.

293

294

295

296

297