E. coli strain DSM 30083 was used for all the bacterial inactivation studies. This laboratory strain, widely studied and with a sequenced genome, is a non-pathogenic and a primary model organism for laboratory research. The cells of E. coli were grown in sterile conditions in 100 ml of Luria Bertani (Miller’s LB Broth) medium at 37 °C. The bacterial growth was monitored by optical density (at 600 nm) in a spectrophotometer (UVmini-1240, Shimadzu). The culture in the stationary phase was diluted in a buffer solution (buffered sodium chloride/peptone/water, pH 7, Oxoid) to obtain a concentrated bacterial suspension of 10⁶ Npp/100 ml. Viable bacteria counting was carried out by the microplates MUG/E. coli method in accordance with the requirements of the standard NF EN 9308-3 (1999). Over the treatment time, samples of 18 ml were analyzed. A preliminary separation of the catalyst from the liquid phase was carried out (spontaneous decantation in the case of suspended catalysts). Therefore, only the viable bacteria concentration in the liquid phase was determined irrespective of the type of catalyst used.

Two different TiO₂ Aeroxides and a supported TiO₂ catalyst were investigated. The two Aeroxides were the Degussa P25 (20-30 nm primary particle size by transmission electron microscopy (TEM); 54 m² g⁻¹ specific surface area by BET; 78% anatase, 22% rutile by X-ray diffraction (XRD)) and the VP Aeroperl P25/20 (20-30 µm primary particle size by TEM; 46 m² g⁻¹ specific surface area by BET; 80% anatase, 20% rutile by XRD). The supported catalyst consisted of Millennium PC-500 immobilized on a flat cellulosic fiber support (specific surface area 98 m² g⁻¹, cellulosic fibers 38 g m⁻², TiO₂ 16.7 g m⁻², zeolite 2 g m⁻², and SiO₂ 13.3 g m⁻²).

The disinfection efficiencies of the catalysts were compared under identical conditions. In general, the capacity of a catalyst to degrade a pollutant (e.g. bacteria) depends on its ability to absorb light, which is in turn related to its physicochemical properties (e.g. specific surface area etc.). In this work, we ensured that the disinfection efficiency of the catalysts was determined when absorbing the same quantity of light. This was achieved by optimizing the loading of each catalyst in the solution to be treated via an optical method.

In a previous study, Minero et al. ^{[8, 14, 18]} showed that the optimal catalyst loading was the minimal quantity of catalyst required to ensure the total absorption of UV irradiation entering a reactor, the geometry of which defines the optical path length of the emitted light ^{[19, 20]}. Elsewhere, it has been shown ^[11] that in terms of photocatalytic water treatment efficiency, this optimal concentration should ensure the highest pollutant degradation rate in a given period.

As a first step, the ability of the catalysts to absorb light was ascertained using an optical arrangement (Fig. 1), which comprises an integrating sphere linked to a 1240 Shimadzu UV spectrophotometer (wavelength ranging from 250 to 1100 nm). The light was supplied by a source simulating solar radiation (1000 W m⁻², solar spectrum AM1.5).

Fig. 1. Optical set up for the catalyst characterization and loading optimization.

The supported catalyst was positioned at the inlet of the optical unit, allowing for the collection of both direct and scattered radiation ^{[21, 22]}. For the catalysts in powder form, various suspensions at different concentrations were placed in quartz containers at different thicknesses and then arranged in front of the opening of the integrating sphere. By measuring the light transmission through the different materials, a relationship involving the absorbance, the concentration of TiO₂, and cell thicknesses (pathlength) was determined.

The reactor was a cylindrical borosilicate glass tube (internal diameter 1.2 cm, length 80 cm) operating in a closed recirculating circuit with a stirred reservoir tank, with a total working volume of 1 L (Fig. 2).

Fig. 2. Experimental photoreactor.

The radiation source was a UV lamp (VL-330) with emission centered at 365 nm. After calibration with a 365 nm UV sensor (UVA 365 from Lutron Electronic Enterprise), the radiation flux density at the reactor axis was controlled between 5 and 35 Wm⁻², the range of values corresponding to solar UV irradiation. To illuminate the total surface of the reactor, an aluminum compound parabolic collector (CPC) was positioned just behind the reactor.

First, the ability of the suspended catalysts to degrade the bacteria was ascertained by photocatalytic tests carried out in the previously described photoreactor under different concentrations in a range between 0.5 and 3 g L⁻¹ and under a constant light flux density of 35 W m⁻². These photocatalytic tests were carried out to define experimentally the optimal concentration that ensures the optimal bacterial inactivation rate and to compare this new concentration to that found optically.

Second, to highlight the bacterial inactivation efficiency of each of the used catalysts, additional photocatalytic experiments were carried out using the photoreactor under different irradiation flux densities (between 5 and 35 W m⁻²) and under optimal catalyst loading. Using the suspended catalysts, the contaminated solution flowed through the reactor by means of a volumetric pump and was continuously mixed in the recirculation tank to ensure a constant optimal catalyst concentration throughout the treatment time. The supported catalyst was wrapped around a rod (diameter 5 mm) positioned at the axis of the reactor. The useful surface of the photocatalytic medium in the reactor was 0.02 m². The contaminated solution flowed under laminar conditions through the reactor. However, a major concern in this case was leaching, which leads to a reactivity modification due to the change in TiO₂ recovery. Because of this, a preliminary phase of 24 h of Milli-Q water recirculation was operated under laminar flow conditions. Temperature and pH were essentially constant over the treatment time with values of 25 °C and 6.9, respectively.

There is no universal kinetic expression to describe the photodisinfection process, and the literature proposes several different laws. These generally show that the photo decontamination rate is dependent on the contaminant concentration and the light flux density ^[23]. In our work, to facilitate the comparative analysis, the initial kinetics were described with a pseudo-first-order law with ‘constant’ irradiation-dependent rate constant, k(I) and a general expression is given by Eq. (1):

where C is the concentration of viable bacteria in the aqueous phase (Npp/100 ml) over irradiation time t (min), and k(I) is the initial disinfection rate constant (min⁻¹) expected to be a function of the irradiation flux density I (W m⁻²).

However, the validity of this model implies that the rate constant does not vary for a given irradiation flux density during the process. It was found experimentally (see below) that the curves obtained did not represent a single elemental first-order reaction step but rather two sequential steps. These were: a log-linear inactivation region depending on the light flux density and a final deceleration process, which is known as the tail. Considering that when using each of the catalysts, the majority of bacterial inactivation has occurred in the first sequence, only this log-linear inactivation region was taken into account.

To define experimentally the optimal loading for the suspended catalysts, the kinetics of the photodisinfection of E. coli were observed for different concentrations of TiO₂. The rate constants, k (min⁻¹), for the photodisinfection of E. coli using the suspended P25 catalyst are shown in Fig. 3 as an example.

Fig. 3. Kinetic constants for the inactivation of E. coli as a function of the concentration of P25 catalyst in suspension.

It is clear that an increase in the bacterial inactivation constant occurs with an increase in the concentration of the catalyst until a limiting concentration is reached, close to C_opt = 1.6 gL⁻¹. This value was then compared to the optical measurement (Fig. 4) indicating the optimal light absorption of the suspended P25 catalyst.

Fig. 4. Light transmission measurement as a function of the couple (concentration of TiO2 × light path length L = 1.2 cm) using P25 in suspension (1) and VP Aeroperl in suspension (2).

The optimal concentration should ensure a 99% absorption of radiation entering the tubular reactor, considering a transmittance of 0.01 and a reactor internal diameter = 1.2 cm. Using the P25 catalyst as an example, through photocatalytic experiments the optimal concentration was calculated as C_opt = 2/1.2 = 1.66 g L⁻¹. Comparing the two values of concentration obtained with the two different methods, it was concluded that the optimal values were consistent. This validates the possibility of using the optical method to define the optimal loadings of suspended catalyst without the need for further additional photocatalytic experiments.

To establish the correlation between the properties of a material and its efficiency to inactivate E. coli bacterial cells, it is essential to know the effect of the radiation on the material involved under its optimal working condition (optimal light absorption). The kinetics of photolysis and photocatalysis using the different materials are shown as a function of irradiation time in Figs. 5 and 6, respectively.

Fig. 5. (a) Photolysis kinetics of E. coli as a function of irradiation time under different irradiation flux densities; (a') The initial inactivation kinetics profiles. (1) 5 W m−2; (2) 10 W m−2; (3) 20 W m−2; (4) 35 W m−2.

Fig. 6. Photocatalysis of E. coli as a function of irradiation time using the suspended VP Aeroperl (a), the supported catalyst (b), and the suspended P25 (c) under different irradiation flux densities, where (a’), (b’), and (c’) are their initial inactivation kinetics profiles. (1) 0 W m−2; (2) 5 W m−2; (3) 10 W m−2; (4) 20 W m−2; (5) 35 W m−2.

Photolysis tests showed that UVA had a considerable direct action on bacterial cells but had a lower bacterial inactivation effect than photocatalysis. Referring to the literature ^{[24, 25]}, the damage caused by UVA light is mainly due to its absorption by cellular components called intracellular chromophores. The best known intracellular chromophore is probably L- tryptophan. In the presence of oxygen, UV irradiation contributes to the generation of reactive oxygen species (ROS), which induce oxidative stresses that damage the cell membranes and components.

The dark tests using each of the photocatalysts showed a decrease in bacterial concentration over the treatment duration. Because only viable bacteria in the liquid phase have been counted, this decrease could be due to an increasing bacterial adsorption on the surface of the catalysts over time. This result is interesting in so far as it highlights a non-negligible interaction between the target bacteria and the solid material used.

Plots of photocatalysis using each of the catalysts showed two different bacterial inactivation regimes: a log-linear inactivation region depending on the radiation flux density and a final deceleration process. The log-linear inactivation regimes showed that across the used light spectrum all catalysts were able to achieve more than 99% of bacterial inactivation with less than 120 min of UV irradiation exposure. However, none of them achieved a total bacterial disinfection over the treatment time used. This asymptotic limit of inactivation is known by the tail and could be related to the development of bacterial resistance during the disinfection treatment ^[26], or to an inhibition phenomenon produced by the competitive action of the organic products released to the medium as explained by Benabbou et al. ^{[27, 28]}.

The kinetic constants k (min⁻¹), derived from the kinetics of photodisinfection, are given in Fig. 7 as a function of the light flux density I (W m⁻²) applied. Irrespective of the material used, the kinetic constants were correlated in a linear fashion with the light flux density emitted at λ = 365 nm. It is also apparent that the rate of bacterial inactivation depends on the material used and is dependent of the structure of the material. In fact, as the amount of photons increase, more electron-hole pairs will be formed, leading eventually to the formation of more OH^. radicals, and, in turn, leading to a higher activity of bacterial photoinactivation.

Fig. 7. Kinetic constants of E. coli inactivation as a function of the light flux density with photolysis (1), P25 in suspension (2), supported catalyst (3), and VP Aeroperl in suspension (4).

Because of this linear correlation, it was possible to rearrange Eq. (1) to obtain Eq. (2):

where C is the concentration of viable bacteria in the aqueous phase (Npp/100 ml), k_t is the linear correlation constant (m²J⁻¹), and I is the incident light flux density (W m⁻²).

The values of k_t for each of the media used are summarized in Table 1. This parameter is a correlation kinetic constant connecting the microbial inactivation rate and the light flux density received over time. Based on the disinfection kinetic constants it emerges that: (1) both suspended catalysts showed almost the same disinfection efficiency (with a mean k_t = 33.5 x 10⁻⁴ m² J⁻¹); (2) the global effect of the photocatalysis degradation was 3.5 times higher than that of photolysis; (3) the disinfection efficiency shown by the supported photocatalyst was slightly less than that of the suspended Degussa P25 by a factor of 1.2 and less than that of the VP Aeroperl by a factor of 0.8. This result contrasts to those in the literature ^{[18, 29]} where catalysts in slurries show significantly higher degradation rate constants than do supported catalysts. In this work, the high disinfection efficiency of the supported catalyst could be related to the existence of other non-apparent intrinsic phenomenon such as its bacterial adsorption capacity, which will be investigated in future work.

Table 1 Optical properties and kinetic constants of the various media used: absorbency (A), slopes of the kinetic constant versus intensity (kt), and slopes of the kinetic constant versus cumulative energy (ke). Catalyst A Kinetic constants kt (10-4 m2 J-1) ke (10-3 L kJ-1) Degussa P25 0.99 35 60 VP Aeroperl 0.99 32 53 Supported catalyst (2D) 1 28 46 Photolysis — 9 15

Table 1 Optical properties and kinetic constants of the various media used: absorbency (A), slopes of the kinetic constant versus intensity (kt), and slopes of the kinetic constant versus cumulative energy (ke).

One of the long term purposes of this study is to scale up the photocatalytic reactor and to use real solar irradiation for wastewater disinfection. Given that solar irradiation flux is variable over time, it is necessary to rearrange Eq. (2) to compare the efficiency of each of the used catalysts when the received light flux density is variable. Plots as a function of irradiation time were not adequate to compare the media disinfection efficiency because variations in incident UV irradiation intensity are ignored. The disinfection kinetics involved a function of the cumulative light energy per unit volume of treated solution over a given irradiation time, Q_UV (kJ L⁻¹) following Eq. (4). The expression obtained is a slightly modified form of that presented by Goslich et al. ^{[29, 30]} and can be derived from Eq. (3):

where Q_UV = (S/V_T)∫0t I dt, k_e = k_t V_T/S, S is the irradiated surface (m²), V_T is the volume of the solution to be treated (L), and k_e is the volumetric energy constant (L kJ⁻¹).

The results of the bacterial inactivation rates as a function of the quantity of cumulative energy received per unit volume of the illuminated reactor are shown in Fig. 8. For each of the used catalysts, the curve depicts the experimental points of all tested irradiation flux densities. Derived from Eq. (4), the values of the constants k_e were shown in Table 1. As expected, the disinfection kinetics displayed almost the same profile regardless of the emitted irradiation flux density. This means that the photoinactivation was related exclusively to the cumulative energy received. This result is obviously in agreement with previous conclusions and confirms the possibility of using this comparative method to reveal the catalytic performance under variable-intensity solar irradiation.

Fig. 8. Initial kinetics E. coli inactivation as a function of the quantity of the energy received over time with photolysis (1), P25 in suspension (2), supported catalyst (3), and VP Aeroperl in suspension (4).

Our ultimate goal was to establish the performance of the catalysts in terms of how efficiently they use light. As demonstrated here, the used catalysts integrated into a photoreactor gave performance that was proportional to the flux across the whole irradiation spectrum, which corresponded to the light flux density of sunlight hitting the Earth’s surface. To compare the inactivation efficiency of the catalysts used irrespective of their characteristics (shape and size) and the irradiation conditions, we assessed their apparent quantum yields (η_m) given by Eq. (6). This yield, based on the number of photoinactivated bacterial cells divided by the number of absorbed photons (given by Eq. (5)), reflects the ability of these solids to degrade bacteria when the same quantity of UV light is absorbed ^{[29, 31, 32]}.

Here, N_ph is the number of absorbed photons at λ = 365 nm over the treatment duration, N_t is the cumulative number of the inactivated E. coli cells (Npp) at a given period of time, E_λ is the average energy of a photon in the ultra violet range (J), S is the total irradiated surface (m²), I is the light flux density (W m⁻²) over irradiation time Δt (s), and η₀ is the optical yield of the process.

The quantity of bacterial cells broken down was deduced from the kinetics of disinfection. The number of photons absorbed by each material tested was ascertained by the number of photons received multiplied by the optical yield of the process (η₀). The optical yield of the process (η₀), using each of the catalysts, was obtained in earlier work ^[33]. It indicates the actual quantity of photons reaching the surface of the medium, by considering the transfer of radiation through the various components comprising the experimental equipment. Three factors affect the overall optical yield (η₀): reflectivity of the CPC receptor (R = 0.95), transmission from the reactor wall (T_r = 0.95), and absorbency of the media (A) used. As described previously, the concentrations of both P25 and Aeroperl catalysts in suspension were chosen to ensure almost total absorption of the incident radiation (A_TiO2 = 0.99). The light absorption of the 2D cellulosic material was A = 1 ^[24].

For each of the materials used, a mean quantum yield value was assessed from the average yield obtained over the reaction duration. Both the optical yield of the process and the quantum yield of each of the materials used are shown in Table 2. Despite the spread of the experimental results recorded under given light flux densities, the quantum yields of the materials displayed almost constant values over time. This means that the quantum yield depends in a linear way on the quantity of photons absorbed irrespective of the light flux. The quantum yields were significantly lower than those found for the degradation of various chemicals ^[1]. This was due to a high number of photons or hydroxyl radicals required for the inactivation of one single cell. Some results were reported by Sun et al. ^[2] and by Cho et al. ^[3] dealing with the correlation between the inactivation of E. coli and the quantity of absorbed hydroxyl radicals. Finally, the inactivation performance of the supported catalyst was comparable to those of the suspended catalysts in terms of the efficiency with which light radiation is used to inactivate bacterial cells.

Table 2 Apparent quantum yield of the material (ηm) and optical yield (η0) of the process. Catalyst η0 ηm (10-14 degraded cells/ absorbed photons) Degussa P25 0.81 2.58 VP Aeroperl 0.81 2.03 Supported catalyst (2D) 0.91 1.70

Table 2 Apparent quantum yield of the material (ηm) and optical yield (η0) of the process.