Scale-Up Study of a Multiphase Photocatalytic Reactor—Degradation

Article pubs.acs.org/est

Scale-Up Study of a Multiphase Photocatalytic ReactorDegradation of Cyanide in Water over TiO2 Mahsa Motegh, J. Ruud van Ommen,* Peter W. Appel, and Michiel T. Kreutzer Department of Chemical Engineering, Delft University of Technology, Julianalaan 136, 2628 BL Delft, The Netherlands S Supporting Information *

ABSTRACT: This paper provides an integrated view on various aspects of reactor design for photocatalytic reactions and presents a scale-up study of photocatalytic reactors. This study focuses on degrading organic pollutants in the effluent of an integrated gasification coal combustion plant over TiO2, with the target of degrading cyanide to below its allowable emission threshold set by European legislation. Here, we show the interplay of different efficiencies that affect the overall apparent photonic efficiency and the reactor volume required to achieve a certain objective in conversion. The chosen reactor configuration is rectangular slurrybubble-columns-in-series to ensure a good mass transfer rate per photoreactor while approaching plug-flow behavior as a sum, and a high reactor surface-area-to-volume ratio for a good capture of incident photons. We consider a simple 1D photonic description of a photoreactor, in the direction of incident solar light, and implement a bidirectional scattering model for photocatalytic particles and bubbles to calculate the local rate of photon absorption and the photon absorption efficiency in the photoreactor. We show that, implementing the principles of process intensification, the large scale degradation of cyanide to below European emission limits is achievable.

■

INTRODUCTION Heterogeneous photocatalysis is a promising technology in the field of environmental remediation especially for purifying water from trace persistent organic pollutants.1−6 Even though a great body of literature exists on the fundamental aspects of photocatalysis such as kinetics, the developed strategies for the scale-up of photocatalytic reactors are scarce.7−16 Consequently, there are not many successful implementations of photocatalytic processes at large scales.7 Many researchers have focused on a specific issue such as determining the effect of initial reactant concentration, pH, temperature, catalyst loading, or light intensity on the kinetics of degradation of a certain compound in a given setup.17−23 A successful implementation of photocatalytic processes at large scales, however, calls for a global interdisciplinary view over the interplay between all the important parameters that affect the capture and utilization of photons in a photoreactor. Further, the engineering aspects of photocatalysis in terms of contacting patterns and mass transfer rates must be optimized to develop useful design procedures for large-scale operations. In this paper, we show how design of photoreactors using chemical engineering principles can result in a more cost-effective design by decreasing the total reactor volume required for achieving a predefined level of conversion. As a representative example for the degradation of organic pollutants, we selected the degradation of cyanide in water. Cyanide is a very toxic pollutant that is present in the effluents of coal gasification,24 mining of precious metals and electroplating processes.25 It has a stringent threshold for discharge to surface water of 0.1 ppm, set by the European Commission.26,27 Current physical, chemical, and biological methods to remove the dissolved cyanide are ineffective.28,29 Photocatalytic © 2013 American Chemical Society

degradation, however, has proved successful in fully mineralizing cyanide.25 The objective of this paper is to present a scale-up strategy for photocatalytic processes. To achieve this, we combine a detailed analysis of the photonic efficiency in photoreactors with knowledge of multiphase catalytic reactor design. The fate of all photons is followed via a bidirectional scattering model30 that leads to practical photoreactor models that are neither too simplistic nor too complex to be useful in photoreactor design, optimization and scale-up. Moreover, we show how an educated choice of reactor configuration and contacting patterns can result in a more cost-effective photoreactor. Our scale-up strategy is shown by initially extracting data from literature on batch photoreactors at lab and bench scale, translating such data into intensified engineering of continuous reactors at bench scale. Then, we extend our study to the design of a large-scale photocatalytic reactor to degrade cyanide in the effluent of a power plant under solar radiation, the treatment cost of which is reported in the Supporting Information (SI). It is demonstrated where the main photon losses are taking place in the chain of photocatalytic events and how we can improve them in order to achieve the minimum reactor volume necessary to comply with the environmental legislations on cyanide. Received: Revised: Accepted: Published: 1574

July 30, 2013 December 16, 2013 December 20, 2013 December 20, 2013 dx.doi.org/10.1021/es403378e | Environ. Sci. Technol. 2014, 48, 1574−1581

Environmental Science & Technology

■

Article

Optical Properties of Photocatalyst Particles and Bubbles. To set up photon balances based on the bidirectional scattering model, we need to know the extinction coefficient βs and the scattering albedo ωs of photocatalyst particles. The extinction coefficient denotes the loss of intensity per unit length of slurry and the scattering albedo is the ratio of scattering to extinction of light ωs = σs/βs. These optical properties can be acquired experimentally in dilute suspensions, see, for example, refs 36 and 37. Moreover, we need to know the probabilities of backward and forward scattering, which can be obtained from the average cosine of scattered rays.37 The bubbles are assumed not to absorb photons in the UV range. Thus, the scattering albedo of bubbles is unity, meaning that the extinction of light by bubbles occurs only via scattering. The scattering coefficient of bubbles depends on the bubble size, shape, gas fraction, and the probability of backward scattering (See ref 30 for detailed explanation). Photoreactor Model. In degrading low concentrations of contaminants, wherein oxygen deprivation is not an issue, tubular or thin-film photoreactors are reported to be ideal.14 However, for the photocatalytic degradation of a high level of organic pollutants in water, oxygen must uniformly and constantly be provided to the photoreactor. This is vital for the progression of the photocatalytic reaction, and guarantees the degradation of pollutants to values set by environmental legislation. Slurry-bubble columns (i.e., unified photoreaction and aeration unit) operating in a homogeneous regime offer great advantages in terms of constant and uniform supply of oxygen throughout the photoreactor and a high oxygen and reactant mass transfer rate. Importantly, a suitable photocatalytic reactor must have as large as a surface to volume ratio as possible to ensure sufficient capture of photons. In solar applications, when light-concentrating lenses or mirrors are not applied, a high surface to volume ratio in an aerated photoreactor means a flat geometry, that is, rectangular slurry-bubble column. The liquid and gas phase in such photoreactors are axially dispersed. Stiegel et al. introduced the following dimensionless correlation based on Reynolds and Peclet numbers for calculating the axial dispersion coefficient in rectangular bubble columns,38

MODELING APPROACH Photocatalytic Reaction Rate. Cyanide is photo-oxidized over TiO2 under UV irradiation to form cyanate, which could be further oxidized on the surface of photocatalyst to produce nitrate and carbonate.22 Alternatively, cyanate may be hydrolyzed to produce ammonium and carbonate.22 The rate of a photoreaction depends not only on the concentration of reactants, but also on the volumetric rate of photon absorption, denoted as ea. In this paper, we implement a type of photokinetic rate expression based on quasi-steady-state approximation originally derived5 for strongly adsorbing compounds,31 r = k1f (C)k 2ε( −1 +

1 + 2ϕea/k 2ε )

(1)

where the group k1 f(C) represents the chance that an electron/ hole pair undergoes a photoreaction at the surface of photocatalyst, the parameter ϕ represents the wavelengthdependent probability that a photon creates an electron/hole pair, ε is the photocatalyst volume fraction, and k2 is a coefficient that depends on material properties and increases with decreasing primary particle size (dp). Equation 1 serves as a good minimal representation of several important features of photocatalytic reactions, and for the reaction that we consider here, cyanide removal, kinetic data were successfully reported in this form, even though the adsorption is not very strong. Reaction rate’s dependence on ea reduces to ϕea at small values of irradiation intensity, and to (2k2εϕea)1/2 at high values of irradiation intensity. In a photoreactor usually both regimes coexist with square-root dependence on ea in the vicinity of irradiation source, and linear dependence far from the source. The transition from square-root to linear dependence on ea occurs at ea ∼ k2ε/ϕ. At values higher than this transition rate the recombination of holes and electrons becomes prominent in competition with surface reaction. Therefore, photons are used less efficiently for the photocatalytic reaction at higher intensities. Bidirectional Scattering Model. To calculate the rate of photon absorption, which influences the reaction rate, we use the bidirectional scattering model, developed by Motegh et al.,30 based on the work of Brucato et al.32 In this model, we assume that both scattering and absorption events take place within the photoreactor. The angular distribution of scattering is then captured in two main directions: backward (opposite to incident) and forward (in the direction of incident light). In bubbly slurry photocatalytic reactors, as photons pass through the reactor in the forward direction, they are either scattered backward by photocatalyst particles and bubbles, absorbed by photocatalyst particles, scatter forward by bubbles and photocatalyst particles or pass through the reactor unchanged. The backward scattered photons also go through the same events as they pass through the reactor. Setting up photon balances based on mentioned events, the rate of photon absorption by photocatalyst is calculated as the net gradient of intensity in the forward and backward direction. For a unidirectional incident light, the volumetric rate of photon absorption, ea, and the relevant optical properties of bubbles and photocatalytic particles are reported by Motegh et al.30 To calculate the volumetric rate of photon absorption, one may also consider the extension of bidirectional scattering, the sixflux model with a three-dimensional approach to radiation modeling.33−35

PecL = a1ReLb1ReGc1

(2)

with a1 = 8.2 × 10−4, b1 = 1.44, and c1 = −1.12. The equations required to calculate the gas fraction, axial dispersion coefficient, mass transfer coefficient and Peclet numbers in slurry-bubble columns are reported in Table S1 in the SI. The high liquid axial dispersion in rectangular bubble columns causes the liquid to be almost perfectly mixed. For well-mixed reactors, the reactant concentration in the reactor and at the outlet (Cex) is equal. When total degradation of a pollutant (Cex ∼ 0) is intended, such photoreactors will suffer from a very low surface reaction efficiency, that is, low k1 f(CCN−). Therefore, the overall photonic efficiency is very low is such reactors. This has consequences on the total reactor surface area that is required to achieve high conversions. To minimize the reactor volume, reactor frontal area, and the residence time required to achieve a desired level of conversion, it is more attractive to operate in the poorly mixed plug flow regime. We can approach the behavior of a plug flow reactor by putting a number of well-mixed slurry-bubble columns in series. Residence Time. The total residence time τres for the first order cyanide degradation rate to achieve a certain conversion 1575

dx.doi.org/10.1021/es403378e | Environ. Sci. Technol. 2014, 48, 1574−1581

Environmental Science & Technology

Article

Table 1. Photoreaction Rate Expression

The rate constants α1 are attained at pH of 12. bThe rate constants α1 and reaction order were attained by fitting eq 1 to kinetic data reported in ref 40. a

XCN in a highly dispersed rectangular bubble column is calculated as τres =

XCN (1 − XCN)kapp,[CN−]εsl

in the so-called homogeneous regime with small, nearly monodisperse bubbles. This makes that liquid-gas mass transfer is not limiting, and can be ignored. The photocatalytic reaction takes place on the photocatalyst surface among the adsorbed species. Equilibrium conditions are achieved between the adsorbed and bulk concentrations of the organic compounds. A steady-state approximation is applied for unstable reaction intermediates such as radicals and excited electrons and holes. The consumption of oxygen at pilot scale is low, due to the low reactant concentrations and slow reactions. However, under realistic conditions typically more than one contaminant is present. For example, in our large-scale study, the wastewater contains both a high formate concentration and a low cyanide concentration. This means that the oxygen consumption is high, and the system must be aerated continuously. The contraction of bubbles is negligible. The adsorption of oxygen on the catalyst surface sites does not compete with the organic compounds.5 The catalyst loading is ∼1 g/L, hence, εs ≪ 1, and its effect on liquid viscosity or density is negligible. The kinetics of cyanide degradation is of first order in reactant concentration, while the degradation of formate, which is also present in the feed stream is of zero order (See Table 1). Due to the small photocatalyst particle size ∼1 μm, the catalyst sedimentation and maldistribution, as well as the liquid-solid mass transfer can be neglected. The liquid and gas phase in each slurry tank is axially dispersed. The kinetic rate constants α1 and α2 for the degradation of cyanide over Aldrich nanoparticles are reported in Table 1, taken from ref 25. The α1 coefficient is related to the surface reaction at the photocatalyst surface, and α2 constant is related to the electron-hole trapping properties of photocatalyst particles. The dark adsorption of cyanide is reported to be negligible, resulting in a first order reaction rate in cyanide concentration. For the degradation of formates over TiO2 nanoparticles in slurry there was no reported kinetic rate expression in the form of eq 1. Therefore, we fitted such a kinetic rate expression to the data on the degradation of formic

(3)

where εsl is the slurry holdup, and kapp,[CN−] is the apparent depth-averaged kinetic rate constant of cyanide defined as, kapp,[CN−] =

α1,[CN−]SgCcat L

∫0

L

⎛ ⎛ ⎞1/2 ⎞ ⎜ −1 + ⎜1 + α2ea ⎟ ⎟dx ⎜ ⎜ CcatSg ⎟⎠ ⎟ ⎝ ⎝ ⎠ (4)

Sg is the specific surface area of catalyst, Ccat the catalyst loading, L the photoreactor depth, α1,[CN−] and α2 are kinetic rate constants presented in Table 1. In case of multicomponent photocatalytic degradation, kapp,[CN−] shall be multiplied by (1 + ∑iKads,iCi)−1. In a series of rectangular bubble columns with identical residence time τres,i, the conversion for a first order reaction is then calculated as: XCN = 1 −

1 n

(1 + εslkapp,[CN−]τres, i)

, τres, i =

V ϕv n

(5)

where, n is the number of tanks in series, ϕv is the volumetric flow rate of liquid, and V is total reactor volume. The residence time to achieve conversion XCN in an ideal plug flow reactor based on the chosen kinetic rate expression in this work is calculated as τres,plug = =

ϕv

∫ L 0

εslkapp,[CN−] −ϕv

εslkapp,[CN−]L

XCN

1 dX 1−X

log(1 − XCN)

(6)

Assumptions for the Photoreactor Model. Because of the small gas load required, the slurry-bubble column is operating 1576

dx.doi.org/10.1021/es403378e | Environ. Sci. Technol. 2014, 48, 1574−1581

Environmental Science & Technology

Article

acid over TiO2 Degussa P25 reported by Chen et al.40 At high concentrations of formic acid the degradation of formate is independent of formate concentration (i.e., zero order reaction rate). We calculated the rate of photon absorption in the labscale photoreactor of Chen et al.40 based on the bidirectional scattering model and fitted the α1 coefficient. The value of α2 for Degussa P25 particles was taken from literature.39 The formic acid adsorption on TiO2 Degussa P25 is reported to be 1.21 × 103 L mg−1.41 When formates are present in addition to cyanide, the degradation rate of cyanide must be corrected for the adsorption by formates on the photocatalyst. In that case, the k1 f(C) in eq 1 for the degradation of cyanide will be changed from k1CCN− to α1,[CN−]α2 CCN − k1f (CCN−) = 2 1 + K ads, F[HCOO−] (7)

Figure 1. Residence time (red) and total required surface area of photoreactor (blue) as a function of number of rectangular slurrybubble columns in series for the continuous treatment of 30 ppm of cyanide in 2 L h−1 of water over Anatase at photocatalyst loading of 0.5 g L−1 under lamp illumination, I0 = 1.91 × 10−8 Einstein cm−2 s−1.

where Kads,F and [HCOO−] are the adsorption coefficient and concentration of formate. Comparing eq 1 and the kinetic rate expressions in Table 1 gives αα k1 = 1 2 (8) 2

coefficient of bubbles are calculated assuming an average bubble diameter of 4 mm and the probability of backward scattering of 5%.30,43,44 The photocatalyst is Anatase from Aldrich, with scattering albedo of 0.75, specific extinction coefficient of 3.8 × 104 cm2 g−1 at 365 nm, and specific surface area of 7.1 m2 g−1.36 Regarding the residence time, a single slurry-bubble column performs poorer than the batch bench-scale photoreactor proposed by Marugan et al.42 Since in a slurry-bubble column, due to a high axial dispersion, the concentration in the reactor is the exit concentration, the photoreaction rate is lower as compared to the batch reactor wherein the reactant concentration drops gradually in time. However, as the number of slurry-bubble columns in series increases, the photoreactor performance becomes closer and closer to that of ideal plugflow photoreactor. As shown in Figure 1, 8 tanks-in-series reasonably approach the ideal plug flow behavior, and results in a smaller residence time as compared to the batch operation with aeration and reaction units separated. The required surface area to capture enough photons to continuously degrade 30 ppm of cyanide in 2 L h−1 of water also decreases as the number of tanks-in-series increases. However, it remains higher in continuous operation as compared to batch. The lower overall photonic efficiency caused by the lower average surface reaction efficiency in slurry-bubble columns in series results in a higher required photoreactor surface area (See Table 2). Even though the required surface area to capture photons in the batch reactor is smaller, the total volume is larger compared to the continuous operation. The total reactor volume in the batch operation for treating 2 L h−1 of water is 4 L,42 while for 8 tanks-in-series this volume reduces to 2.5 L. The photoreactor in the batch operation has a volume of 1.25 L, but it has a big storage/aeration tank of 2.75 L that does not do any photoreaction. As the volume of this storage tank increases (i.e., to treat a larger volume of water), the residence time increases accordingly, while in the continuous operation the residence time remains the same. It is also difficult to make a one-to-one comparison between the batch photoreactor with a separate aeration unit and the continuous operation in slurrybubble columns-in-series, since the former has a cylindrical and the latter has a flat geometry. Solar vs Lamp Illumination at Bench Scale. In this section we discuss the photon losses at different steps of the chain of photocatalytic events45 under solar or artificial irradiation. In Table 2 detailed efficiencies for cyanide

and, k 2ε =

■

2ϕSg Ccat α2

· (9)

RESULTS AND DISCUSSION Continuous Operation of Bench-Scale Photoreactor. Marugan et al. presented an interesting scale-up study of a photoreactor to degrade cyanide in water from lab scale to bench-scale operation.42 The bench-scale photoreactor was operated in batch and reduced cyanide (conversion ∼95%) in 4 L of water within 2 h. It consisted of an aeration vessel (2.75 L) and an annular photoreactor (1.25 L) with a lamp on the cylinder axis. This bench-scale photoreactor can be further optimized in two aspects. First, the aeration and photoreaction units can be combined to have a single unit bubbly slurry photoreactor. We have shown previously that, for practical gas fractions, bubbles do not affect the overall photonic efficiency of photoreactors significantly.30 Second, the batch operation can be converted to a continuous operation with the capacity of treating 2 L h−1 of water from cyanide. As discussed earlier, we can approach plug-flow behavior by putting a number of slurrybubble columns (i.e., well-mixed tanks) in series. In Figure 1, we show the effect of the number of tanks-inseries11 on the total residence time and the total surface area required to degrade 30 ppm of cyanide. Whereas in catalytic reactors, the reactor costs is estimated by reactor volume, in photocatalytic reactors, the cost of photoreactor depends more on the photoreactor surface normal to irradiation, since this surface area must be made of UV-transparent materials such as quartz or borosilicate glass. We compare the results of tanks-inseries to the reported data for batch operation by Marugan et al.42 and to the outcome from an ideal plug-flow photoreactor. Similar to the bench-scale photoreactor, the rectangular slurrybubble columns in series are operating at optical thickness of τ ∼ 19, with photocatalyst loading of 0.5 g L−1 and photoreactor depth of 1 cm. The superficial air velocity is set at 2 cm/s, resulting in a ∼10% gas fraction inside each column, calculated based on the equations presented in SI Table S1. The scattering 1577

dx.doi.org/10.1021/es403378e | Environ. Sci. Technol. 2014, 48, 1574−1581

Environmental Science & Technology

Article

reaction rate and a smaller residence time. However, the electron-hole trapping efficiency is lower for lamp illumination as compared to solar radiation. The higher the intensity, the faster the electron-hole pairs are generated, ultimately faster than they can migrate to the photocatalyst surface and react, resulting in a more prominent electron-hole recombination. The surface reaction efficiency is lower for the continuous photoreactor. Due to the high axial dispersion of liquid in each slurry bubble column, they are quasi well-mixed. Thus, the concentration of cyanide in each photoreactor-in-series is the exit concentration. This results in a lower surface reaction efficiency per tank, and consequently, a lower average surface reaction efficiency as compared to the batch operation where the concentration of cyanide drops steadily in time. The overall photonic efficiency is lower in the continuous operation based on tanks-in-series as compared to the batch operation, if both are illuminated by lamp. This means that for reaching the same level of conversion in the continuous mode, a larger photoreactor surface area is required to capture the photons as compared to batch operation. However, since the continuous configuration does not have a separate aeration vessel, the total volume is smaller. Figure 2a presents the electron-hole trapping efficiency of Anatase from Aldrich photocatalyst particles as a function of dimensionless photon flux Υ = 2βtI0ϕ(k2ε)−1.30 As the dimensionless photon flux increases the electron-hole recombination becomes more and more prominent and the dependence of reaction rate on ea shifts from linear to square-root. This makes that the electron-hole trapping efficiency is higher under solar illumination than under lamp illumination. Figure 2b compares the surface reaction efficiency (k1 f(C) in eq 1) in batch vs continuous operation. The continuous operation is demonstrated for 8 rectangular slurry-bubble columns in series as well as the ideal plug flow regime. The separation of reaction rate dependence on concentration and on ea in eq 1 results in surface reaction efficiency to be independent of the irradiation source. As seen in Figure 2b, the total residence time to degrade 30 ppm of cyanide is the shortest in an ideal plug-flow photoreactor, owing to a maximized surface reaction efficiency. An inefficient use of total reactor volume in the batch photoreactor (i.e., a separate aeration unit) leads to a longer residence time than that of plug-flow. The slurry-bubble columns in series, however, approaches the plug-flow behavior with an improved residence time as compared to batch.

Table 2. Summary of the Detailed Efficiencies for the BenchScale Degradation of Cyanide, over Anatase from Aldrich at Photocatalyst Loading of 0.5 g L−1, under Lamp Radiation for Both Batch and Continuous Operation, and under Solar Illumination for the Continuous Operation Onlya efficiency I0 (Einstein cm−2 s−1), light intensity λave (nm), intensity-averaged wavelength ηabs = I−1 0 ∫ eadx, absorption efficiency ϕ, excitation efficiency ηEHT = ((∫ rdx/∫ eadx)/ (ϕk1 f(C))), electron− hole trapping efficiency ηSR = k1 f (C), Surface reaction efficiency (8 tanks in series in continuous regime) ηOv = I−1 0 ∫ rdx, overall efficiency

batch illuminated by lamp

continuous illuminated by lamp

continuous under solar illumination

1.91 × 10−8

1.91 × 10−8

8.96 × 10−9

365

365

356

0.805

0.805

0.805

1 0.0883

1 0.0883

1 0.124

0.178

0.157

0.157

0.0126

0.0112

0.0157

a The efficiencies are calculated using a rate expression of the form r = k1 f(C)k2ε[−1 + (1 + 2ϕea/k2ε)1/2], consistent with a quasi-steady state approximation of photocatalytic microkinetics.

degradation are given for a batch photoreactor under lamp illumination,42 and in continuous mode under both solar and lamp irradiation. The continuous mode is assumed to be carried out in eight rectangular slurry-bubble columns in series. The emission efficiency of lamp and solar sources are assumed equally to be 100%, since we are mainly interested in photon losses that are related to the photocatalyst properties or photoreactor design. The solar light intensity is assumed to be the standard 30 W m−2. The lamp intensity was reported to be 1.91 × 10−8 Einstein cm−2 s−1.42 The lamps emission peak was reported to be 365 nm42 at which the optical properties of TiO2 (Anatase from Aldrich) were acquired. However, such optical properties for solar application were attained at the solar intensity-averaged wavelength of 356 nm. The photon absorption efficiency depends on the photocatalyst properties and the optical thickness. For all reactor configurations considered the optical thickness is taken to be high (i.e., τ ∼19 with photocatalyst loading of 0.5 g L−1), resulting in a high photon absorption efficiency. The lamp illuminates at a higher intensity than sunlight in the UVA range, giving a higher

Figure 2. (a) Electron-hole trapping efficiency as a function of dimensionless photon flux for Anatase from Aldrich at photocatalyst loading of 0.5 g L−1. Symbols indicate the dimensionless photon flux for lamp with I0 = 1.91 × 10−8 Einstein cm−2 s−1 and for solar irradiation I0 = 30 W m−2. (b) The surface reaction efficiency for batch vs continuous operation as a function of residence time. 1578

dx.doi.org/10.1021/es403378e | Environ. Sci. Technol. 2014, 48, 1574−1581

Environmental Science & Technology

Article

Figure 3. Large-scale degradation of cyanide in the effluent of IGCC power plant under solar illumination over TiO2 Degussa P25 at photocatalyst loading of 0.4 g L−1 (a) The achievable conversion of cyanide as a function of total surface area of eight rectangular slurry-bubble column photoreactors in series for continuous removal of 99% of cyanide in 13 m3 hr−1 of water in optically thick photoreactors. (b) The conversion of cyanide and formate per tank in eight optically thick rectangular slurry-bubble column photoreactors in series to reach 99% cyanide removal.

resulting in a gas fraction of ∼10%, according to the calculations based on SI Table S1. Such a gas fraction provides sufficient oxygen to degrade cyanide and formates in water (eq S1 in the SI). The sunlight UVA intensity is assumed to be available for 6 h day−1. Even though the effluent flow rate is 13 m3 h−1, the photoreactor capacity must be 4 times as much, since the produced effluent in 24 h day−1 must be treated within 6 h of sunlight. Therefore, the flow rate of liquid to the photoreactor is ϕv = 52 m3 h−1. Photoreactor Dimensions. To degrade cyanide to below the discharge limit of 0.1 ppm, we need to aim at a 99% conversion. Similar to the bench scale, we use 8 rectangular slurry-bubble columns in series. Figure 3a shows the achieved conversion of cyanide as a function of photoreactor surface area. High surface areas are required, even though a more photocatalytically active photocatalyst was used. This is due to the low overall photonic efficiency, specifically the low surface reaction efficiency and the electron-hole trapping efficiency. Figure 3b shows the conversion of cyanide and formate per slurry-bubble column. The residence time for 99% degradation of cyanide in the effluent stream of Elcogas IGCC plant is calculated to be ∼5.4 h which is reasonable. Using Aldrich anatase42 the residence time would be ∼7.9 h. In the absence of formate and using Degussa P25, reducing cyanide by 99% takes just ∼2.4 h, showing the necessity to consider other relevant compounds. Discussion on Photonic Efficiencies. To make photocatalysis more attractive for industrial applications, a considerable increase of the photonic efficiency is required. The constituents of the overall photonic efficiency, excluding the (solar) irradiation efficiency, are the photon-absorption and reaction efficiencies. The high scattering albedo of common photocatalysts limits the improvement of the photon absorption efficiency. However, the reaction efficiency can be enhanced by, for example, using photocatalyst particles of smaller crystal size or doping by nobel metals,31 and by adjusting the point of zero charge. The former enhances the electron-hole trapping efficiency, and the latter improves the reactant adsorption at the surface of the photocatalyst (See SI). Scale-Up Strategy. The analysis of photon efficiencies in different steps of a photoreaction combined with the principles of process intensification leads to the following scale-up strategy:

Large Scale Operation under Solar Radiation. In this section, we present a scale-up study of photocatalytic reactors to degrade cyanide in the effluent of an integrated gasification combined cycle (IGCC) power plant. The cyanide contaminated water is coming from the syngas washing section and the blow down of slag water closed circuit. The goal is to degrade cyanide to below the maximum allowable discharge value of 0.1 ppm set by the European Commission.26,27 Assumptions of the Model. In order to model the degradation of cyanide in an effluent of a real stream we first need to focus on the feed composition. The inlet concentration of contaminants in the effluent of the ELCOGAS IGCC power plant prior to entering the photocatalytic reactor is reported to be for cyanide [CN−] = 10 mg L−1, and for formate [HCOO−] = 1700 mg L−1 at pH 9.5. The volumetric flow rate of water to be treated is 13 m3 h−1. The presence of other organic compounds can hinder the degradation of cyanide through competitive adsorption on photocatalyst surface. In case of IGCC effluent, formates are strongly adsorbed on TiO2,41 impeding the degradation of cyanide. Therefore, the surface reaction efficiency declines (see eq 7.), and a longer residence time and a larger photoreactor surface area are required to achieve the same level of conversion. Although the formate conversion is unavoidable, it is beneficial to the overall scheme of the effluent treatment. Since most common photocatalysts require UVA radiation to generate electron-hole pairs, they are limited to the UV range of solar spectrum. To calculate the degradation rate of contaminants, we assumed the solar intensity to be the standard 30 Wm−2 in the UVA range, considerably lower than the commonly used lamps. Therefore, photocatalysts for solar applications must have enhanced photon absorption and trapping efficiencies to allow for reasonable residence times. Using Aldrich Anatase42 would result in residence times longer than the duration of sunlight in a day. Instead, we use TiO2 Degussa P25 with a higher photocatalytic activity. Degussa P25 has an average scattering albedo of 0.8, and specific extinction coefficient of 5 × 104 cm2g−1 36 considering solar illumination. These optical properties are acquired at the intensity-averaged wavelength of solar UVA spectrum. We assumed that the photoreactor is optically thick, leading to a maximized photon absorption efficiency.30 The photocatalyst loading is set to 0.4 g L−1, and the photoreactor depth to 2 cm, giving an optical thickness of τ ∼40. The superficial gas velocity is set to 2 cm s−1 and bubble diameter to 4 mm,

• Intensifying the bench-scale reactors by combining aeration and photoreaction in one vessel decreases the 1579

dx.doi.org/10.1021/es403378e | Environ. Sci. Technol. 2014, 48, 1574−1581

Environmental Science & Technology

Article

(2) Herrmann, J. M. Heterogeneous photocatalysis: State of the art and present applications. Top. Catal. 2005, 34, 49−65. (3) Alfano, O. M.; Bahnemann, D.; Cassano, A. E.; Dillert, R.; Goslich, R. Photocatalysis in water environments using artificial and solar light. Catal. Today 2000, 58, 199−230. (4) Bahnemann, D. Photocatalytic water treatment: Solar energy applications. Sol. Energy 2004, 77, 445−459. (5) Turchi, C. S.; Ollis, D. F. Photocatalytic degradation of organic water contaminants: Mechanisms involving hydroxyl radical attack. J. Catal. 1990, 122, 178−192. (6) Jassby, D.; Farner Budarz, J.; Wiesner, M. Impact of aggregate size and structure on the photocatalytic properties of TiO2 and ZnO nanoparticles. Environ. Sci. Technol. 2012, 46, 6934−6941. (7) Van Gerven, T.; Mul, G.; Moulijn, J.; Stankiewicz, A. A review of intensification of photocatalytic processes. Chem. Eng. Process: Process Intensif. 2007, 46, 781−789. (8) Kamble, S. P.; Sawant, S. B.; Pangarkar, V. G. Novel solar-based photocatalytic reactor for degradation of refractory pollutants. AIChE J. 2004, 50, 1647−1650. (9) Subramanian, M.; Kannan, A. Photocatalytic degradation of phenol in a rotating annular reactor. Chem. Eng. Sci. 2010, 65, 2727− 2740. (10) Adesina, A. A. Industrial exploitation of photocatalysis: Progress, perspectives and prospects. Catal. Surv. Asia 2004, 8, 265−273. (11) Li Puma, G.; Toepfer, B.; Gora, A. Photocatalytic oxidation of multicomponent systems of herbicides: Scale-up of laboratory kinetics rate data to plant scale. Catal. Today 2007, 124, 124−132. (12) Li, D.; Xiong, K.; Li, W.; Yang, Z.; Liu, C.; Feng, X.; Lu, X. Comparative study in liquid-phase heterogeneous photocatalysis: Model for photoreactor scale-up. Ind. Eng. Chem. Res. 2010, 49, 8397−8405. (13) Mueses, M. A.; Machuca-Martinez, F.; Li Puma, G. Effective quantum yield and reaction rate model for evaluation of photocatalytic degradation of water contaminants in heterogeneous pilot-scale solar photoreactors. Chem. Eng. J. 2013, 215−216, 937−947. (14) Prieto-Rodriguez, L.; Miralles-Cuevas, S.; Oller, I.; Agüera, A.; Li Puma, G.; Malato, S. Treatment of emerging contaminants in wastewater treatment plants (WWTP) effluents by solar photocatalysis using low TiO2 concentrations. J. Hazard. Mater. 2012, 211−212, 131−137. (15) Braham, R. J.; Harris, A. T. Review of major design and scale-up considerations for solar photocatalytic reactors. Ind. Eng. Chem. Res. 2009, 48, 8890−8905. (16) Satuf, M. L.; Brandi, R. J.; Cassano, A. E.; Alfano, O. M. Scalingup of slurry reactors for the photocatalytic degradation of 4chlorophenol. Catal. Today 2007, 129, 110−117. (17) Yin, X.; Xin, F.; Zhang, F.; Wang, S.; Zhang, G. Kinetic study on photocatalytic degradation of 4BS Azo dye over TiO2 in slurry. Environ. Eng. Sci. 2006, 23, 1000−1008. (18) Jin, H. J.; Sang, D. K.; Tak, H. L.; Dong, H. L. Degradation of trichloroethylene by photocatalysis in an internally circulating slurry bubble column reactor. Chemosphere 2005, 60, 1162−1168. (19) Al-Rasheed, R.; Cardin, D. J. Photocatalytic degradation of humic acid in saline waters. Part 1. Artificial seawater: Influence of TiO2, temperature, pH, and air-flow. Chemosphere 2003, 51, 925−933. (20) Pujara, K.; Kamble, S. P.; Pangarkar, V. G. Photocatalytic degradation of phenol-4-sulfonic acid using an artificial UV/TiO2 system in a slurry bubble column reactor. Ind. Eng. Chem. Res. 2007, 46, 4257−4264. (21) Yoshikawa, N.; Kimura, T.; Kawase, Y. Oxidative degradation of nonionic surfactants with TiO2 photocatalyst in a bubble column reactor. Can. J. Chem. Eng. 2003, 81, 719−724. (22) Chiang, K.; Amal, R.; Tran, T. Photocatalytic oxidation of cyanide: Kinetic and mechanistic studies. J. Mol. Catal. A-Chem. 2003, 193, 285−297. (23) Muneer, M.; Theurich, J.; Bahnemann, D. Titanium dioxide mediated photocatalytic degradation of 1,2-diethyl phthalate. J. Photochem. Photobiol. A. 2001, 143, 213−219.

residence time, and enables the continuous operation of photoreactors. • Rectangular slurry-bubble-olumns-in-series with limited depth is an effective configuration for the degradation of high levels of organic pollutants in water. This configuration approaches the ideal plug-flow behavior that results in a maximized photoreaction efficiency. Thus, the surface area to capture photons, and the investment costs are reduced. Moreover, they allow for a high surface to volume ratio that facilitates an efficient capture of photons. • The overall photonic efficiency for photocatalytic degradation of pollutants must be enhanced. The main losses in the overall photonic efficiency are in the electron-hole trapping and surface-reaction steps. • The composition of the feed stream must be taken into account in the photoreactor design. The presence of other organic compounds, that undergo photocatalytic reactions, with competitive adsorption on photocatalyst surface can impede the photoreaction rate of the target compound. We showed that, following this strategy, the large-scale continuous degradation of cyanide to below legal emission levels is feasible. Even though, the photocatalytic reactor surface area for large-scale solar applications is large, such surface areas can be reduced significantly if photocatalysts are modified to have an increased surface reaction efficiency and electron-hole trapping efficiency. For example, the required surface area of the photoreactor can be reduced by 50% if Anatase photocatalyst with crystal size of 9 nm28 is used instead of Degussa P25 (See SI Figure S1). An important reason for the low photoreaction rate is that most common photocatalysts are limited to UVA radiation to generate electron-hole pairs. Coupling or doping photocatalysts with upconversion-activators can enhance their visible light and infrared response.46 This will increase the photoreaction rate and results in lower required surface areas and residence times in sunlight-driven water purification applications.

■

ASSOCIATED CONTENT

S Supporting Information *

Additional information as noted in the text. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS We are grateful to Mr. Francisco Garcia Pena, the Engineering R&D Director for providing the data on the flow rates and the level of contaminants in the waste stream of the Elcogas IGCC plant. Moreover, we thank Prof. Dr. J. Marugan from Universidad Rey Juan Carlos for very useful discussions on this work.

■

REFERENCES

(1) Matthews, R. W. Photo-oxidation of organic material in aqueous suspensions of titanium dioxide. Water Res. 1986, 20, 569−578. 1580

dx.doi.org/10.1021/es403378e | Environ. Sci. Technol. 2014, 48, 1574−1581

Environmental Science & Technology

Article

(24) Durran, A.; Monteagudo, J. M.; San Martin, I.; Garcia-Pena, F.; Coca, P. Photocatalytic degradation of pollutants from Elcogas IGCC power station effluents. J. Hazard. Mater. 2007, 144, 132−139. (25) Marugan, J.; van Grieken, R.; Cassano, A. E.; Alfano, O. M. Intrinsic kinetic modeling with explicit radiation absorption effects of the photocatalytic oxidation of cyanide with TiO2 and silica-supported TiO2 suspensions. Appl. Catal., B 2008, 85, 48−60. (26) European Commission. Integrated Pollution Prevention and Control- Reference Document on Best Available Techniques in Common Waste Water and Waste Gas Treatment/Management Systems in the Chemical Sector. 2003; http://eippcb.jrc.ec.europa.eu/ reference/BREF/cww_bref_0203.pdf. (27) The European Parliament and the Council of the European Union. Directive 2010/75/EU of the European Parliament and of the Council of 24 November 2010 on industrial emissions (integrated pollution prevention and control) (Recast). OJEU 2010, 334, 17−119. (28) Karunakaran, C.; Gomathisankar, P.; Manikandan, G. Solar photocatalytic detoxification of cyanide by different forms of TiO 2. K. J. Chem. Eng. 2011, 28, 1214−1220. (29) Bahnemann, D.; Bockelmann, D.; Goslich, R. Mechanistic studies of water detoxification in illuminated TiO2 suspensions. Sol. Energy Mater. 1991, 24, 564−583. (30) Motegh, M.; van Ommen, J. R.; Appel, P. W.; Mudde, R. F.; Kreutzer, M. T. Bubbles scatter light, yet that does not hurt the performance of bubbly slurry photocatalytic reactors. Chem. Eng. Sci. 2013, 100, 506−514. (31) Alfano, O. M.; Cabrera, M. I.; Cassano, A. E. Photocatalytic reactions involving hydroxyl radical attack: I. Reaction kinetics formulation with explicit photon absorption effects. J. Catal. 1997, 172, 370−379. (32) Brucato, A.; Rizzuti, L. Simplified modeling of radiant fields in heterogeneous photoreactors. 2. Limiting “two-flux” model for the case of reflectance greater than zero. Ind. Eng. Chem. Res. 1997, 36, 4748− 4755. (33) Brucato, A.; Cassano, A. E.; Grisafi, F.; Montante, G.; Rizzuti, L.; Vella, G. Estimating radiant fields in flat heterogeneous photoreactors by the six-flux model. AIChE J. 2006, 52, 3882−3890. (34) Li Puma, G.; Khor, J. N.; Brucato, A. Modeling of an annular photocatalytic reactor for water purification: Oxidation of pesticides. Environ. Sci. Technol. 2004, 38, 3737−3745. (35) Li Puma, G.; Brucato, A. Dimensionless analysis of slurry photocatalytic reactors using two-flux and six-flux radiation absorptionscattering models. Catal. Today 2007, 122, 78−90. (36) Cabrera, M. I.; Alfano, O. M.; Cassano, A. E. Absorption and scattering coefficients of titanium dioxide participate suspensions in water. J. Phys. Chem. 1996, 100, 20043−20050. (37) Satuf, M. L.; Brandi, R. J.; Cassano, A. E.; Alfano, O. M. Experimental method to evaluate the optical properties of aqueous titanium dioxide suspensions. Ind. Eng. Chem. Res. 2005, 44, 6643− 6649. (38) Stiegel, G. J.; Shah, Y. T. Axial dispersion in a rectangular bubble column. Can. J. Chem. Eng. 1977, 55, 3−8 cited By (since 1996) 7. (39) Marugan, J.; van Grieken, R.; Pablos, C.; Satuf, M. L.; Cassano, A. E.; Alfano, O. M. Rigorous kinetic modelling with explicit radiation absorption effects of the photocatalytic inactivation of bacteria in water using suspended titanium dioxide. Appl. Catal., B 2011, 102, 404−416. (40) Chen, Q.; Chang, J.; Li, L.; Yuan, J. Y. A new kinetic model of photocatalytic degradation of formic acid in UV/TiO2 suspension system with in-situ monitoring. React. Kinet. Catal. L. 2008, 93, 157− 164. (41) Tôrres, A. R.; Azevedo, E. B.; Resende, N. S.; Dezotti, M. A comparison between bulk and supported TiO2 photocatalysts in the degradation of formic acid. Braz. J. Chem. Eng. 2007, 24, 185−192. (42) Marugan, J.; van Grieken, R.; Cassano, A. E.; Alfano, O. M. Scaling-up of slurry reactors for the photocatalytic oxidation of cyanide with TiO2 and silica-supported TiO2 suspensions. Catal. Today 2009, 144, 87−93. (43) Davis, G. E. Scattering of light by an air bubble in water. J. Opt. Soc. Am. 1955, 45, 572−581.

(44) He, H.; Li, W.; Zhang, X.; Xia, M.; Yang, K. Light scattering by a spheroidal bubble with geometrical optics approximation. J. Quant. Spectrosc. Radiat. Transfer 2012, 113, 1467−1475. (45) Motegh, M.; Cen, J.; Appel, P. W.; van Ommen, J. R.; Kreutzer, M. T. Photocatalytic-reactor efficiencies and simplified expressions to assess their relevance in kinetic experiments. Chem. Eng. J. 2012, 207− 208, 607−615. (46) Cates, E. L.; Chinnapongse, S. L.; Kim, J.; Kim, J. Engineering light: Advances in wavelength conversion materials for energy and environmental technologies. Environ. Sci. Technol. 2012, 46, 12316− 12328.

1581

dx.doi.org/10.1021/es403378e | Environ. Sci. Technol. 2014, 48, 1574−1581