PII:

Chemical Engineering Science, Vol. 53, No. 16, pp. 3007—3021, 1998 ( 1998 Elsevier Science Ltd. All rights reserved Printed in Great Britain S0009–2509(98)00119–5 0009—2509/98/$—See front matter

A laminar falling film slurry photocatalytic reactor. Part II—experimental validation of the model Gianluca Li Puma and Po Lock Yue* Department of Chemical Engineering, The Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong (Received 3 December 1997; received in revised form 15 April 1998; Accepted 20 April 1998) Abstract—The dimensionless mathematical model (LSSE—LSPP model) developed in Part I of this paper for a laminar falling film slurry (LFFS) photocatalytic reactor provides a method for scale-up and design of this type of reactor. Part II of the paper reports the experimental validation of the LSSE—LSPP model and a study of the sensitivity of the model with variations in the model parameters. The LSSE—LSPP model was used to calculate the rate of photocatalytic oxidation of salicylic acid in a pilot-scale LFFS photocatalytic reactor over a wide range of experimental conditions. The model results were found to fit the experimental data well, under conditions with varying substrate concentration, intensity of the incident radiation, catalyst loading, flow rates, recycle ratio and reactor geometry. The sensitivity analysis of the model parameters indicated that the reaction kinetics parameters are the major source of uncertainty in the modelling of photocatalytic reactors. The model revealed that, in the range of radiation wavelengths of 310—380 nm, light scattering in water suspensions of TiO (Degussa 2 P25) is negligible at high optical thicknesses but cannot be neglected at low optical thicknesses. ( 1998 Elsevier Science Ltd. All rights reserved. Keywords: Photocatalytic reactor; pilot plant experiments; laminar falling film. INTRODUCTION

The design of large-scale wastewater treatment plants employing photocatalytic reactors is best assisted with pilot plant experiments and a reliable reactor model. In Part I of this paper, a simple dimensionless mathematical model (LSSE—LSPP model) for LFFSIW photocatalytic reactors was developed from basic concepts with the aim of making the model readily usable for design and scale-up. Model simulation results for different reaction kinetics over a range of values of the model parameters were presented and discussed. In this part of the paper, the LSSE—LSPP model is used to predict the rate of photocatalytic oxidation of salicylic acid in a pilot-scale LFFSIW photocatalytic reactor using TiO as the photocatalyst. The experi2 mental conditions investigated included the effects of substrate concentration, intensity of the incident radiation, catalyst loading, flow rates, recycle ratio and reactor geometry on reactor performance. Salicylic acid was selected as a model organic substrate for the photocatalytic experiments as it provided a suitable substrate that is easy to analyse and is

*Corresponding author.

safe to handle in large volumes required for the pilot plant experiments. Salicylic acid has a relatively complex molecular structure that makes it similar to many toxic aromatic compounds. Matthews (1990) has shown that the rates of photocatalytic oxidation of salicylic acid in TiO aqueous suspensions are compa2 rable with those of chlorophenols and other aromatic substrates. The sensitivity of the model to variations in the model parameters was studied to provide information on the relative importance of these parameters. EXPERIMENTAL EQUIPMENT AND METHOD

A schematic representation of the pilot plant LFFSIW photocatalytic reactor is shown in Fig. 1. The reactor column was made from a single piece of a 316 stainless-steel tube with a smooth surface. The length of the column was 1600 mm and the diameter 108 mm. The column was housed in a structure to enable the exact vertical alignment of the reactor. The lamp was inserted inside a Pyrex UV-transparent sheath with a sealed end and was mounted exactly in the middle of the reactor column. An inert atmosphere around the lamp was created by continuous sparging of nitrogen through the bottom end of the sheath. This prevented the possible formation of small traces of ozone and ensured the temperature

3007

3008

G. L. Puma and P. L. Yue

Fig. 1. Schematic representation of the pilot plant LFFSIW photocatalytic reactor rig.

of the lamp to be kept at optimal conditions for irradiation. The reactor was operated under three different arrangements by using lamps of different lengths (Table 1). The ultraviolet lamps were of the fluorescent blacklight type. The emission spectrum of a blacklight lamp is a continuous broad band from 300 to 410 nm with a peak at approximately 355 nm as shown in Fig. 2. However, as the lamp was inserted inside a Pyrex UV-transparent sheath, only radiation of wavelengths higher than 310 nm could be transmitted through the sheath. The intensity of the incident radiation at the lamp wall was varied by powering the lamp with different ballast systems that were connected in series/parallel combinations to provide a number of different power input arrangements. UV energy fluxes were measured using a Cole and Parmer VLX-3W radiometer fitted with a 365 nm sensor. The relative spectral response of the 365 nm sensor (bandwidth from 355 to 375 nm) is shown in Fig. 2. To increase the flexibility of the pilot plant, the photocatalytic system was designed for operation in either the batch or continuous mode. When working in the continuous mode, the system could be operated in one pass through the reactor or in a recycle mode. Slurry suspensions of TiO in water were prepared in 2 a 480 l feed tank. The TiO suspension from the feed 2 tank was pumped to the liquid distribution system of the reactor and collected in a 6 l receiver tank at the bottom outlet of the reactor column. The solutions

could be recycled from the receiver tank back to the liquid distributor system and/or discharged from the system through an overflow. The contents of the feed and the receiver tanks were thoroughly mixed with mixer units, oxygen spargers and heating/cooling facilities. All wetted parts of the pilot plant were made of 316 stainless-steel, Teflon, Viton or borosilicate glass to eliminate leaching of organics into the solution. Salicylic acid (Riedel-de Hae`n*99.8%) was used as a model organic substrate for the photocatalytic reaction experiments. The photocatalyst was untreated Degussa P25 titanium dioxide. The samples collected during the steady-state experiments were promptly analysed by a Hewlett-Packard 1090 Series II Liquid Chromatograph and by a Shimadzu TOC5000 Total Organic Carbon Analyzer. Both analytical instruments were equipped with autosampling units. The degradation of salicylic acid and the formation of intermediate products were followed by HPLC analysis. The degree of mineralisation of salicylic acid to carbon dioxide was followed by total organic carbon (TOC) analysis. CONVERSIONS AND MATERIAL BALANCE

The oxidation of salicylic acid yields intermediate products before mineralisation. The reaction may be represented by a series-reaction as shown in the following stoichiometric relation: C H O #7O N I12n N 7CO #3H O. 7 6 3 2 2 2

(1)

Laminar falling film slurry photocatalytic reactor—II

3009

Table 1. Specification of ultraviolet lamps used and the geometrical parameters for each reactor geometry

Lamp

Nominal power (W)

Length (*) (mm)

Bulb diameter (mm)

Length/ diameter

a (*)

b (**)

36 18 15

1200 560 360

28 28 28

42.9 20 12.9

1.33 2.86 4.44

22.22 10.37 6.67

Philips TLD36W/08 blacklight Philips TLD18W/08 blacklight Avis FL15 BLB black light blue

Note: (*) Effective bulb length of lamp in which radiation is emitted. (**) Dimensions of reactor column: H"1.6 m, R"0.054 m.

Fig. 2. Spectral power distribution of a PHILIPS TLD36W/08 blacklight lamp. The inset shows the relative spectral response of the 365 nm sensor. The response band-width is from 355—375 nm.

Quantitatively, the ‘apparent’ concentration of all the intermediate products at the reactor outlet may be expressed in terms of equivalent amounts of salicylic acid as shown in the following equations: 138.12]*TOC I"*C! "*C!1.643]*TOC 7]12.011

(b) of salicylic acid to carbon dioxide, s S!-*#:-*#!#*$NCO2 *TOC *TOC s "1.643 " (6) S!-*#:-*#!#*$NCO2 C0 TOC0 (c) of salicylic acid to intermediates, s

S!-*#:-*#!#*$NI s "s !s . (7) S!-*#:-*#!#*$NI S!-*#:-*#!#*$ S!-*#:-*#!#*$NCO2

(2) EFFECT OF ADSORPTION

*C"C0!C

(ppm)

(3)

065 *TOC"TOC0!TOC (ppmC) (4) 065 where eqs (3) and (4) represent, respectively, the removal of salicylic acid and of the total organic carbon as measured by the HPLC and TOC analyses. The results of the oxidation experiments conducted in the pilot plant may therefore be expressed in terms of conversions s: (a) of salicylic acid, s , S!-*#:-*#!#*$ *C s " S!-*#:-*#!#*$ C0

(5)

As salicylic acid was found to be readily adsorbed on titanium dioxide, the concentration of salicylic acid C as measured by HPLC analysis could not be %2 used for calculating its actual removal rate. The effect of adsorption should be taken into account. Figure 3 shows the results of the experiments of ‘dark’ adsorption of salicylic acid on TiO (Degussa P25) for 2 a catalyst concentration of 2.75 g l~1. The Freundlich model was used for treating the ‘dark’ adsorption data. The overall concentration of salicylic acid C 07%3!-was calculated from: C "C #c k C1@k2, 07%3!-%2 T*O2 1 %2

(8)

3010

G. L. Puma and P. L. Yue

Fig. 3. Dark adsorption of salicylic acid on water suspension of TiO (Degussa P25) at a temperature of 2 22—24°C. TiO loading 2.75 g l~1. The solid line is a least-squares fit of the Freundlich equation. 2

where C was the concentration of salicylic acid in %2 equilibrium with the solid phase as measured by HPLC, c was the concentration of the catalyst, T*O2 and k and k were the Freundlich coefficients. The 1 2 calculation of the overall concentration of salicylic acid in the sample collected after reaction was not a simple procedure since it was not possible to establish the effect of the reaction intermediates on the adsorption of salicylic acid on titanium dioxide. It was assumed that, for these samples, the adsorption of salicylic acid was not influenced by the presence of the reaction intermediates.

Control experiments The suspensions were illuminated with radiation of wavelength higher than 310 nm, under conditions in which TiO can sustain heterogeneous photo2 catalysis. Control experiments were performed in the absence of UV light and/or TiO to assess the extent 2 of side dark-reactions or direct photolysis of salicylic acid. In all of these experiments, negligible conversions of salicylic acid were observed within an experimental error of $2%. It was concluded that the oxidation of salicylic acid observed was the direct result of heterogeneous photocatalysis.

ing conditions and the predicted reactor conversions can be computed. The geometrical parameters a and b were calculated according to the dimensions of the reactor column and the effective length of the lamps from which radiation was emitted. Their values are shown in Table 1 for each of the lamps used. The parameters which describe the fluid dynamics of the liquid phase, l and o were assumed to be those of pure water at ambient temperature, i.e. l"1] 10~6 m2 s~1 and o"1000 kg m3. The value of the mass Napierian coefficient e , N!1 averaged in the range of wavelengths 310—380 nm for suspensions of TiO (Degussa P25) in deionised 2 water, was determined by the measurements of the intensity of the incident radiation with different catalyst loading. The measurements were performed in a special apparatus to simulate the TiO suspension 2 in the actual reactor, taking into account the effects of mixing and gas holdup. The TiO concentrations 2 used in these measurements were representative of those used in the reaction experiments. No extrapolation was made to higher concentrations of TiO . 2 The definition of the mass Napierian coefficient is as follows: !ln (I/I ) 0 (9) e " N!1 cl

Estimation of model parameters The parameters and constants that appear in the LSSE—LSPP model are summarised in Part I of this paper. Once these parameters have been estimated, the dimensionless parameters of the LSSE—LSPP model can be evaluated for given values of the operat-

where c is the concentration of the absorbing substance and l is the thickness of the solution traversed by the light, and I/I is the light attenuation. If c is 0 expressed in (kg m~3) and l in (m) e is therefore in N!1 (m2 kg~1). For the measurement of e , a PHILIPS TL6W/08 N!1 blacklight lamp and a sensor at 365 nm were mounted

REACTOR PERFORMANCE AND MODELLING

Laminar falling film slurry photocatalytic reactor—II

3011

Fig. 4. Apparatus for the measurement of the mass Napierian coefficient of suspensions of TiO . 2 Table 2. Measurements of the average mass Napierian coefficient, in the wavelength range from 355 to 375 nm of suspensions, of TiO (Degussa P25) in 2 deionised water Catalyst loading TiO Degussa P25 2 (g l~1) 0 0.005 0.01 0.02 0.03 0.04 0.05

Incident radiation 355—375 nm (mW cm~2) 0.528 0.421 0.339 0.216 0.135 0.085 0.054 Average Standard deviation

inside two quartz tubes and the two tubes were immersed in the apparatus as shown in Fig. 4. The results in Table 2 show that the value of the average spectral mass Napierian coefficient in the wavelength range from 355 to 375 nm does not vary with the catalyst concentration and is equal to 1189 m2 kg~1.

Mass Napierian coefficient (m2/kg~1) — 1191.9 1166.0 1176.1 1196.3 1201.6 1200.1 1188.7 0.144

The optical absorbance spectra for TiO (Degussa 2 P25) suspensions in a 1 mm path- length cell were measured by a Pharmacia Ultrospec III UV spectrophotometer to establish the wavelength of radiation at which this estimate could be averaged. The results in Fig. 5 show that, in the wavelength range from 355

3012

G. L. Puma and P. L. Yue

Fig. 5. Optical absorbance spectra for TiO (Degussa P25) suspensions in 1 mm path length cell. 2

to 375 nm, there was an almost linear dependence of the absorbance on the wavelength of the radiation. As the spectral response of the 365 nm sensor is symmetrical (cf. Fig. 2), the value of the average spectral mass Napierian coefficient e in Table 2 is taken to be the N!1 average at the wavelength of 365 nm. Since the TiO slurry is irradiated in the 2 wavelength range of 310—380 nm, the spectral mass Napierian coefficient, e , needs to be averaged over N!1 the entire range. To obtain this average, it is assumed that the quantum yield of the reaction in the radiation wavelength from 310 to 380 nm is constant. It was observed that, in the range of wavelengths from 330 to 380 nm, there is an almost linear dependence of the absorbance on the wavelength of the radiation (Fig. 5) and the spectral power distribution of the Philips TLD36W/08 blacklight lamp is symmetric with respect to the wavelength at 355 nm (cf. Fig. 2). If the small contribution to irradiation in the wavelength range of 310 to 330 nm (cf. Fig. 2) is neglected, the average spectral mass Napierian coefficient calculated at 355 nm should provide a good estimate over the entire range of wavelengths of radiation from the blacklight source. As the ratio of the absorbance at 355 and at 365 nm in Fig. 5 is constant with the catalyst concentration and is equal to 1.09, the e at 355 nm may be N!1 calculated from the measured e at 365 nm: N!1 Absorbance (355 nm) e N!1,355/. + "1.09. (10) e Absorbance (365 nm) N!1,365/. The average spectral mass Napierian coefficient at 355 nm is therefore as follows: e

N!1,355/.

"1.09e "1295.7 m2 kg~1. N!1,365/. (11)

The intensity of the incident radiation at the lamp wall I was estimated by eq. (12), where I is the w 365/. incident radiation at the lamp wall measured with the radiometer fitted with a 365 nm sensor; P is the j relative spectral response of the 365 nm sensor (bandwidth from 355 to 375 nm), ¼ is the radiant power of j the lamp at wavelength j, (cf. Fig. 2), 310 nm is the wavelength cut-off of Pyrex tubing and 380 nm is the highest wavelength that can photoactivate titanium dioxide photocatalyst: :380/. ¼ dj 310/. j I "I . w 365/. :375/. ¼ P dj 355/. j j

(12)

Using the procedure described in Part I of this paper, the coefficients m and n in the rate equation for the photocatalytic oxidation of salicylic acid in slurry suspensions of TiO were obtained from 2 linear regression of ln c vs ln I , (Fig. 6) and ln c vs j w j ln C0 (Fig. 7). These coefficients were estimated for j the disappearance of salicylic acid and its conversion to carbon dioxide. The remaining kinetic rate constant k was estimated by fitting the LSSE—LSPP T model to the results of Figs 6 and 7. By averaging the results of the third and fifth columns in Table 3, it can be seen that, within a close range of variation, k is T a constant. A summary of the kinetic parameters for the photocatalytic oxidation of salicylic acid in slurry suspensions of TiO in deionised water, under the illumina2 tion of light at wavelengths from 310 to 380 nm, is shown in Table 4. It can be seen that the dependence of the conversions on the intensity of the incident radiation follows an approximately half-order power law, which is consistent with reports in the literature for other organic substrates (Al-Sayyed et al., 1991; Blake et al., 1991; Kormann, et al., 1991). The index

Laminar falling film slurry photocatalytic reactor—II

3013

Fig. 6. Estimation of kinetic parameter m for photocatalytic oxidation of salicylic acid. C0 " S!-*#:-*#!#*$ 10.3 ppm; TiO "2.75 g l~1; QO"0.4 l min~1; g"0.95. 2

Fig. 7. Estimation of kinetic parameter n for photocatalytic oxidation of salicylic acid. I " w 7.66 mW cm~2; TiO "2.75 g l~1; QO"0.4 l min~1; g"0.95. 2

n in Table 4 is a negative number due to inhibition effects associated with the photocatalytic oxidation of salicylic acid. The inhibition effects are caused by the formation and accumulation of intermediate products during the early stages of the oxidation (Cunningham and Al-Sayyed, 1990). Such inhibition effects could be accounted for by introducing an inhibition term in a Langmuir type of surface kinetics model but would make the reactor model complicated. By representing the reaction kinetics with a power law, the inhibition

effect is reflected by the negative value of the index n in the model. It is important to recognise that the kinetic parameters in Table 4 should be applied within the range of concentrations and intensities at which they have been evaluated. Comparison of model results with experimental results Within the ranges of substrate concentrations and light intensities investigated, the LSSE—LSPP model satisfactorily fitted the data on the conversions of

3014

G. L. Puma and P. L. Yue Table 3. Estimation of model parameters, k , and k , Model parameters: T,S!-*#:-*#!#*$ T,S!-*#:-*#!#*$NCO2 a"1.33, b"22.22, H"1.6 m, N "1454, e "1295.7 m2 kg~1, q"1.71, m "0.50, Re N!1 S!-*#:-*#!#*$ n "!0.44, m "0.53, n "!0.16. Experimental conditions: N 2 S!-*#:-*#!#*$ S!-*#:-*#!#*$ CO S!-*#:-*#!#*$ TiO "2.75 g l~1, QO"0.4 l min~1, g"0.95 2 I "7.66 mW cm~2 w CO S!-*#:-*#!#*$ (ppm) 10.30 11.07 20.24 34.43 57.88 99.08 CO "10.3 ppm S!-*#:-*#!#*$ I w (nW cm~2) 5.767 7.664 7.664 8.422 11.673 12.637 Average Standard deviation

s S!-*#:-*#!#*$ (%)

k T,S!-*#:-*#!#*$ (]108)*

22.18 22.50 9.35 5.37 2.58 1.05

5.537 5.638 5.356 6.339 6.193 5.252

s S!-*#:-*#!#*$ (%) 18.83 22.18 22.50 21.98 24.72 26.67

k T,S!-*#:-*#!#*$ (]108)* 5.210 5.537 5.638 5.225 5.161 5.545 5.553 0.375

s

s

S!-*#:-*#!#*$NCO2 (%)

k T,S!-*#:-*#!#*$NCO2 (]108)*

12.03 11.21 5.39 3.04 1.73 1.00

7.259 6.704 6.624 6.706 6.772 7.502

S!-*#:-*#!#*$NCO2 (%) 9.76 12.03 11.21 12.57 13.61 14.47

k

T,S!-*#:-*#!#*$NCO2 (]108)* 6.674 7.259 6.704 7.260 6.701 6.898 6.922 0.308

*[kg(1~n) s~1 W~mm3m`3n~3].

Table 4. Kinetic parameters for the oxidation of salicylic acid k ]108* T Salicylic acid 5.55 Salicylic acidNCO 6.92 2 Range of validity of kinetic parameters IFF~UVA I.!9 w m/0 (mW cm~2) (mW cm~2) 5.77—12.64 2.22—4.85

m

n

0.50 0.53

!0.44 !0.16 C0 S!-*#:-*#!#*$ (ppm) 10.3—99.1

*[kg(1~n) s~1 W~mm3m`3n~3].

salicylic acid as shown in Figs 8 and 9. In addition, the model results show that the mass balance on salicylic acid was successfully closed. The effectiveness of a falling film photocatalytic reactor operating in a laminar regime varies with catalyst loading. The effect of increasing the overall volumetric rate of energy absorption and the number of catalytic sites with increasing catalyst concentration is counterbalanced by the darkening of the flow field closest to the wall of the column. There should be a catalyst concentration at which the photocatalytic reaction attains an optimum. Table 5 shows the application of the LSSE—LSPP model to the results obtained at different catalyst loading in the range 0.1— 5 g l~1. Under these conditions the optical thickness q varied from 0.06 to 3.09. The results showed a good fit of the model for conversions of salicylic acid, s . However, the model S!-*#:-*#!#*$ underestimated the results of s when S!-*#:-*#!#*$NCO2

the catalyst concentration was higher than 1 g l~1. The optimal catalyst loading predicted by the model was found to be approximately to 4.25 g l~1, i.e. 54% higher than the value derived from the experiments. The results of s in the range of catalyst load4!-*#:-*#!#*$ ing from 2.75 to 5 g l~1 could not be measured because, for this range of catalyst concentrations, the liquid-phase concentration of salicylic acid in equilibrium with the adsorbate was below the HPLC detection limit. However, the overall conversion of salicylic acid is always greater than the conversion of salicylic acid to carbon dioxide, the difference being the salicylic acid that was converted to some intermediate products. This difference becomes less as the catalyst concentration is increased. The maximum conversion of salicylic acid predicted by the model is 38% which is consistent with the experimental observations (cf. Table 5). Hence for catalyst concentrations higher than 2 g l~1 s (in the third column of S!-*#:-*#!#*$

Laminar falling film slurry photocatalytic reactor—II

3015

Fig. 8. LSSE—LSPP model results of the effect of the intensity of incident radiation on photocatalytic oxidation of salicylic acid. CO "10.3 ppm; TiO "2.75 g l~1; QO"0.4 l min~1; g"0.95. Model S!-*#:-*#!#*$ 2 parameters: a"1.33; b"22.22; H"1.6 m; N "1454; e "1295.7 m2 kg~1; q"1.71. Kinetic paraRe N!1 meters from Table 4.

Fig. 9. LSSE—LSPP model results of the effect of inlet substrate concentration on the photocatalytic oxidation of salicylic acid. I "7.66 mW cm~2; TiO "2.75 g l~1; Qo"0.4 l min~1; g"0.95. Model w 2 parameters: a"1.33; b"22.22; H"1.6 m; N "1454; e "1295.7 m2 kg~1; q"1.71. Kinetic paraRe N!1 meters from Table 4.

Table 5) are shown as greater than the corresponding values of s shown in the fifth column of S!-*#:-*#!#*$NCO2 Table 5. The discrepancies are acceptable considering that the model adopts an approximate approach in accounting for radiation absorption and the representation of the reaction kinetics is oversimplified. Further experiments were performed to assess the validity of the model for changes in the feed flow rate,

the recycle ratio and consequently the Reynolds number. Table 6 shows the results for a feed flow rate varying from 0.48 to 0.2 l min~1 and for a recycle ratio in the range 0.68—0.97. Within these limits, the estimated thickness of the liquid film was in the range of 0.281—0.479 mm and the Reynolds number N was Re between 291 and 1434. Table 6 shows that although the results of s agreed well with the model, S!-*#:-*#!#*$

3016

G. L. Puma and P. L. Yue Table 5. Comparison between the LSSE—LSPP model results and the experimental results. Effect of the catalyst loading on the photocatalytic oxidation of salicylic acid. Model parameters: a"1.33, b"22.22, H"1.6 m, N "1415, e "1295.7 m2 kg~1. Experimental conditions: CO "9.95 ppm, I "7.66 mW cm~2, Re N!1 S!-*#:-*#!#*$ w QO"0.2 l min~1, g"0.97. Kinetic parameters from Table 4 TiO 2 (g l~1)

q

s (%) S!-*#:-*#!#*$ (Experimental)

s (%) S!-*#:-*#!#*$ (Model)

s S!-*#:-*#!#*$NCO2 (Experimental) (%)

0.1 0.5 1.0 1.5 2 2.75 3.5 4.25 5 8 10

0.06 0.31 0.62 0.93 1.24 1.70 2.16 2.62 3.09 4.94 6.17

14.63 25.62 29.41 34.10 36.49 *37.96 *37.40 *32.89 *31.95 Not available(*) Not available (*)

14.19 25.84 31.45 34.36 36.06 37.42 37.97 38.06 37.87 35.92 34.31

5.54 13.16 23.54 29.81 34.98 37.96 37.40 32.89 31.95 Not available (*) Not available (*)

s

S!-*#:-*#!#*$NCO2 (Model) (%) 6.57 13.29 16.96 18.99 20.23 21.25 21.68 21.78 21.62 20.34 19.26

(*) Experiment not performed.

Table 6. Comparison between the LSSE—LSPP model results and experimental results. Experiments performed under different inlet flow rates and recycle ratios. Model parameters: a"1.33, b"22.22, H"1.6 m, e "1295.7 m2 kg~1. N!1 Experimental conditions: CO "9.9 ppm, I "7.66 mW cm~2, TiO "2.0 g l~1. Kinetic parameters from Table 4 S!-*#:-*#!#*$ w 2 QO (l min~1)

g

N Re

q

s S!-*#:-*#!#*$ (Experimental) (%)

s S!-*#:-*#!#*$ (Model) (%)

0.48 0.48 0.4 0.3 0.2

0.68 0.86 0.94 0.96 0.97

291 684 1258 1434 1415

0.73 0.97 1.19 1.24 1.24

12.54 16.22 22.76 28.52 37.38

14.14 16.72 21.68 27.55 36.25

the results of s diverged from the model S!-*#:-*#!#*$NCO2 at conversions higher than 13%. As the maximum value of s from Table S!-*#:-*#!#*$NCO2 3 from which the rate constant k T,4!-*#:-*#!#*$NCO2 was obtained is also approximately 13%, it is possible that at higher s accumulation of S!-*#:-*#!#*$NCO2 intermediate products might cause a change in the kinetics which would be reflected in a change in . Conversely, due to the high adsorpk T,4!-*#:-*#!#*$NCO2 tion coefficients of salicylic acid on TiO (Degussa 2 P25) (cf. Fig. 3), intermediate products were believed to have insignificant influence on the total amount of salicylic acid adsorbed and therefore on the kinetics of initial degradation of salicylic acid. Consequently the kinetic constants for salicylic acid conversion in Table 4 would have a higher range of validity and the model could predict s with a good accuracy at 4!-*#:-*#!#*$ values as high as 36% (cf. Tables 5 and 6). The design and scale-up of photocatalytic reactors would be best assisted by a mathematical model which has been experimentally validated on reactors of different irradiation arrangements. The geometry of a LFFSIW photocatalytic reactor is relatively simple and is described by the parameters a and b in the LSSE—LSPP model. The magnitude of these two

s s S!-*#:-*#!#*$NCO2 S!-*#:-*#!#*$NCO2 (Experimental) (%) (Model) (%) 6.47 8.66 13.01 20.14 34.98

6.73 8.19 11.04 14.58 20.33

parameters was varied using blacklight lamps of different lengths (Table 1) and the experimental results were compared with the LSSE-LSPP model prediction. In these experiments a varied from 1.33 to 4.44 and b varied from 6.67 to 22.22 as shown in the last two columns of Table 1. Figure 10 shows the results of the experiments conducted at given substrate concentrations and Fig. 11 shows the results at given intensities of the incident radiation. The results of Figs 10 and 11 and of other experiments are also plotted in Figs 12 and 13 in terms of the conversions s S!-*#:-*#!#*$ and s as a function of the Damko¨hler S!-*#:-*#!#*$NCO2 number N . The LSSE-LSPP model calculations Da were found to agree satisfactorily with the experimental data within the experimental errors. This agreement is acceptable despite the assumption that the lamp was modelled as a line source of radiation. All the results presented in this section are consistent with the experimental validation of the LSSE—LSPP model. Sensitivity analysis An analysis of the sensitivity of the model to variations in the parameters m, n, k , and e , with reT N!1 spect to the values in Table 4 and eq. (11), was

Laminar falling film slurry photocatalytic reactor—II

Fig. 10. LSSE—LSPP model results of the effect of lamp length and intensity of the incident radiation on the photocatalytic oxidation of salicylic acid. The solid lines are model results for s , the S!-*#:-*#!#*$ broken lines are for s . CO "10.3 ppm; CO "10.3 ppm; S!-*#:-*#!#*$NCO2 S!-*#:-*#!#*$, &03L/1.2. S!-*#:-*#!#*$, &03L/0.56. CO "10.8 ppm; TiO "2.75 g l~1; QO"0.4 l min~1; g"0.95. Model parameters: S!-*#:-*#!#*$, &03L/0.36. 2 H"1.6 m; N "1454; e "1295.7 m2 kg~1; q"1.71. Kinetic parameters from Table 4. Geometrical Re N!1 parameters from Table 1.

Fig. 11. LSSE—LSPP model results of the effect of the lamp length and the inlet substrate concentration on the photocatalytic oxidation of salicylic acid. The solid lines are model results for s , the S!-*#:-*#!#*$ broken lines are for s . I "7.66 mW cm~2; I "9.0 mW cm~2; S!-*#:-*#!#*$NCO2 w, &03 L/1.2. w, &03 L/0.56. I "11.2 mW cm~2; TiO "2.75 g l~1; QO"0.4 l min~1; g"0.95. Model parameters: w,&03 L/0.36. 2 H"1.6 m; N "1454; e "1295.7 m2 kg~1; q"1.71. Kinetic parameters from Table 4. Geometrical Re N!1 parameters from Table 1.

3017

3018

G. L. Puma and P. L. Yue

Fig. 12. Conversion of salicylic acid as a function of the Damko¨hler number for different lamp lengths. The solid lines are LSSE—LSPP model results.

Fig. 13. Conversion of salicylic acid to carbon dioxide as a function of the Damko¨hler number for different lamp lengths. The solid lines are LSSE—LSPP model results.

performed to assess their relative effect on model predictions. The results in Table 7 which are based on the results of Figs 8 and 9, show that the sensitivity of the model to parameters variations of #10% decreases in the following order: m'n'k 'e . (13) T N!1 As a consequence, m is the parameter that should be estimated with a high degree of accuracy and e is N!1 the parameter that can tolerate a lower degree of precision. A 10% increase in m was found to give rise

to 32—65% variation in model predictions, depending on the experimental conditions. In sharp contrast to this, the model was relatively insensitive to variations in e . The maximum variation on model predictions N!1 was 2.6% in the range of !10%(e (#100%. N!1 A maximum negative variation on model predictions of 13.8% was found with a 50% decrease in e . N!1 The results in Table 7 show that the stiffness of the model increases as the intensity of the incident radiation is increased, and decreases with inlet substrate concentration. In addition it is observed that larger

19.84 22.23 23.07 26.14 26.93

22.21 9.66 4.74 2.32 1.11

I * w (mW cm~2)

5.77 7.66 8.42 11.67 12.64

CO s S!-*#:-*#!#*$ (ppm) 10.30 20.24 34.43 57.88 99.08 #7.77 #8.99 #9.42 #9.57 #9.52

#7.98 #7.75 #7.67 #7.37 #7.30

10% increase k T

#32.18 #44.71 #63.72 #65.56 #65.37

#48.51 #48.04 #47.82 #46.89 #46.63

10% increase m

!15.08 !14.54 !5.82 !5.43 !9.13

!15.42 !15.08 !14.96 !14.50 !14.39

10% increase nt

*CO "10.3 ppm. sI "7.66 MW cm~2. tn"!0.40. S!-*#:-*#!#*$ w

s S!-*#:-*#!#*$ (Model) (%) q"1.713

#0.88 #1.02 #1.05 #1.08 #1.09

#0.90 #0.87 #0.86 #0.83 #0.82

10% increase e N!1 q"1.88

!11.69 !13.18 !13.74 !13.82 !13.64

!11.98 !11.69 !11.58 !11.20 !11.10

50% decrease e N!1 q"0.86

!1.23 !1.41 !1.48 !1.48 !1.43

!1.27 !1.23 !1.22 !1.17 !1.16

10% decrease e N!1 q"1.54

#2.06 #2.40 #2.53 #2.59 #2.62

#2.13 #2.06 #2.03 #1.95 #1.92

50% increase e N!1 q"2.57

#0.58 #0.72 #0.76 #0.79 #0.81

#0.60 #0.58 #0.56 #0.53 #0.52

100% increase e N!1 q"3.43

Table 7. Sensitivity analysis. Model results are taken from Figs 8 and 9. Model parameters: a"1.33, b"22.22, H"1.6 m, e "1295.7 m2 kg~1. N!1 Experimental conditions: TiO "2.75 g l~1, QO"0.4 l min~1, g"0.95. Kinetic parameters from Table 4 2

Laminar falling film slurry photocatalytic reactor—II 3019

3020

G. L. Puma and P. L. Yue Table 8. Sensitivity analysis. Model results are taken from Table 5. Model parameters: a"1.33, b"22.22, H"1.6 m, N "1415, e "1295.7 m2 kg~1. Experimental conditions: CO "9.95 ppm, I "7.66 mW cm~2, Re N!1 S!-*#:-*#!#*$ w QO"0.2 l min~1, g"0.97. Kinetic parameters from Table 4 % variation on model predictions and optical thickness 50% decrease e

N!1

10% decrease e

N!1

50% increase e N!1

100% increase e N!1

TiO 2 (g l~1)

q

s S!-*#:-*#!#*$ (model) (%)

% variation

q

% variation

q

% variation

q

% variation

q

0.1 0.5 1.0 1.5 2

0.06 0.31 0.62 0.93 1.24

14.19 25.84 31.45 34.36 36.06

!25.57 !21.22 !17.81 !15.11 !12.78

0.03 0.15 0.31 0.46 0.62

!4.26 !3.31 !2.59 !2.04 !1.58

0.06 0.28 0.56 0.83 1.11

#17.94 #12.88 #9.29 #6.64 #4.47

0.09 0.46 0.93 1.39 1.85

#31.93 #21.72 #14.70 #9.64 #5.58

0.12 0.62 1.24 1.85 2.47

deviations of the model to changes in e occur at low N!1 values of the optical thickness (i.e., q"0.86). This last effect can be seen more clearly in Table 8 where the sensitivity of the model to changes in e was assessed N!1 in relation to the experiments conducted at different catalyst concentrations and therefore different values of the optical thickness. The results are consistent with the observation that the LSSE—LSPP model becomes stiffer in response to changes in e as the N!1 optical thickness is increased. It should also be noted that with systems of very high optical thickness and short physical thickness, the geometry of the radiation model should be less important. DISCUSSION AND CONCLUSIONS

This study has provided an experimental validation of the LSSE—LSPP model described in Part I of this paper for a wide range of experimental conditions. A method for the evaluation of the model parameters from the results of the operation of a LFFSIW photocatalytic reactor has been shown. The LSSE—LSPP model fitted the rate of photocatalytic oxidation of salicylic acid successfully within the range of validity of the estimated model parameters. Outside the range of validity, it was discovered that the conversion of salicylic acid could still be represented with the same model parameters, however, this was not suitable for the representation of the conversion of salicylic acid to carbon dioxide. This effect was attributed to the accumulation of intermediate products that might have caused a change in the reaction kinetics. The results of the sensitivity of the model to variations in the model parameters suggested that the major sources of uncertainty in the LSSE—LSPP model are the kinetic parameters, particularly m, which should be determined with a high degree of accuracy. This leads to the conclusion that the effects of factors such as velocity gradients on the radial particle distribution along the reactor are minor. Thus, treating the slurry system as homogeneous will not introduce any significant error. Another conclusion that can be drawn from the results of the sensitivity analysis is that light scattering

effects with suspensions of TiO (Degussa P25) in 2 water illuminated with UV in the range 310—380 nm are negligible at high values of optical thickness but cannot be neglected at low values of optical thickness. As a consequence, it is not possible to avoid the mathematical modelling of the radiation field for the estimation of the intrinsic kinetic parameters of TiO (Degussa P25) photocatalysed reactions. Diffi2 culties arise from the fact that the observed reaction rates are dependent on the local volumetric rate of energy absorption, the evaluation of which is intrinsically linked with the mathematical modelling of the radiation field. Attempts to perform kinetic experiments using reactors of low optical thickness so that there is little decay in intensity of the incident radiation within the traversed thickness of the reactor would incur considerable errors due to light scattering. In addition to the above, LFFSIW photocatalytic reactors which use TiO (Degussa P25) slurries are 2 not significantly affected by the supposed redistribution of the local rate of energy absorption due to light scattering. Under optimised conditions, LFFSIW photocatalytic reactors would be operated at high values of the optical thickness, at which light scattering effects are negligible. This suggests that, complex modelling of the radiation field is not required and a simplified approach that neglects light scattering or, as an alternative, uses a parameter such as the mass Napierian coefficient, is appropriate for the modelling of LFFSIW photocatalytic reactors. In conclusion, the major obstacle to the practical use of reliable mathematical models of photocatalytic reactors rests in the evaluation of accurate kinetic parameters. This is likely to be the major source of error in photoreactor modelling. Highly complex mathematical models, which may provide a more rigorous representation of the radiation field in a heterogeneous photoreactor, would still be limited by the errors introduced in the evaluation of the kinetic parameters. The LSSE—LSPP model presented in this paper provides a simple model with readily measurable parameters suitable for use in preliminary

Laminar falling film slurry photocatalytic reactor—II

design, provided that the representation of the reaction kinetics is sufficiently accurate. NOTATION

c C H I I.!9 m/0 k T k 1 k 2 l ¸ m n N Da N Re P Q r r l R TOC » l ¼ z

absorbing substance concentration, kg m~3 substrate concentration, kg m~3 length of the reactor column, m intensity of the incident radiation, W m~2 maximum value of intensity of the incident radiation at the free surface of the falling film, W m~2 kinetic constant, kg(1~n) s~1 m3m`3n~3 W~m coefficient of Freundlich adsorption model, m3 kg~1 coefficient of Freundlich adsorption model, dimensionless thickness of the solution traversed by light, m lamp length, m order of the reaction with respect to LVREA, dimensionless order of the reaction with respect to substrate concentration, dimensionless Damko¨hler number ["k (kI.!9 )m CO n~1H/ T m/0 j l ], dimensionless z,.!9 Reynolds number ("4dl /l), dimenz,!7%3!'% sionless relative spectral response of the sensor, dimensionless volumetric flow rate, m3 s~1 radial coordinate, m lamp radius, m distance between the lamp axis and the free surface of the film, m total organic carbon, kg m~3 volume, m3 fluid velocity, m s~1 radiant power, W axial coordinate, m

Greek letters a geometrical parameter ("H/¸), dimensionless b geometrical parameter ("¸/R), dimensionless c conversion per pass, dimensionless d film thickness, m *C substrate removal, kg m~3

*TOC e N!1 g k k N!1 l m o q s

3021

TOC removal, kg m~3 mass Napierian coefficient, m2 kg~1 or l g~1 cm~1 recycle ratio, dimensionless absorption coefficient, m~1 mass Napierian extinctance ("c]e ), m~1 N!1 kinematic viscosity, m2 s~1 film radial coordinate ("r!R), m fluid density, kg m~3 optical thickness ("ce d), dimensionless N!1 total conversion for the system, dimensionless

Subscripts average average value eq equilibrium j substrate I intermediate products out system outlet z direction along the axial coordinate w lamp wall j wavelength, nm m direction along the radial coordinate Superscripts max maximum value O system inlet REFERENCES

Al-Sayyed, G., D’Oliveira, J. C. and Pichat, P. (1991) Semiconductor-sensitized photodegradation of 4chlorophenol in water. J. Photochem. Photobiol. A: Chem. 58, 99—114. Blake, D. M., Webb, J., Turchi, C. and Magrini, K. (1991) Kinetic and mechanistic overview of TiO 2 photocatalyzed oxidation reactions in aqueous solution. Solar energy material 24, 584—593. Cunningham, J. and Al-Sayyed, G. (1990) Factors influencing efficiencies of TiO -sensitized photo2 degradation. Part 1—Substituted Benzene acids: Discrepancies with dark-adsorption parameters. J. Chem. Soc. Faraday ¹rans. 86, 3935—3941. Kormann, C., Bahnemann, D. W. and Hoffmann, M. R. (1991) Photolysis of chloroform and other organic molecules in aqueous TiO suspensions. En2 viron. Sci. ¹echnol. 25, 494—500. Matthews, R. W. (1990) Purification of water with near-U.V. illuminated suspensions of titanium dioxide. ¼at. Res. 24, 653—660.