The study on the effect of hydrodynamic flow on the photocatalytic performance of photocatalytic reactors

The scalability of photocatalytic reactor is the major challenge due to the inability of the light penetration when the laboratory scale reactors are ascended. It is well known that the characteristic length of photocatalytic reactor is one of the influential parameters determining the efficiency of light penetration and also affects the hydrodynamic fluid flow behaviour inside the reactor. This study visualizes hydrodynamic flow behaviour of three differently configured photocatalytic reactors through numerical simulation of the fluid mass transport inside the reactors. The three photocatalytic reactors are (1) concentric cylindrical glass tube micro-reactor (CGTR), (2) UV-LED strip photocatalytic reactor (STR) and (3) classical annular reactor (CAR) from our previous work. The simulations of flow behaviour confirmed that CGTR exhibited plug flow regime, STR exhibited arbitrary flow (in-between continuously stirred tank reactor and plug flow reactor) and CAR behaved like continuously stirred tank reactor. Also, interpretations of numerical modelling were validated through plotting experimental residence time distribution curve. Furthermore, the comparison of performance parameters revealed that the highest modified space time yield (STYmodified) 225 s−1and highest modified photocatalytic space time yield (PSTYmodified) 9.375 s−1/kW/m3 was obtained for plug flow reactor when compared to continuously stirred tank reactor (CAR) and arbitrary flow reactor (STR). The study confirms that decrease in characteristic length of photocatalytic reactor onsets plug flow regime, which has highest photocatalytic performance efficiency.


Introduction
Conventional wastewater treatment consists of a combination of physical, chemical and biological processes (Saleh et al. 2020). Physical, chemical and biological wastewater treatment systems are well established in real-time industrial applications. The examples of these physical, chemical and biological treatment systems include screen chamber, primary clarifier, coagulation and flocculation, chlorination reactors, Activated sludge system, extended aeration, sequencing batch reactor (SBR) and moving bed bioreactor (MBBR) (Singh et al. 2022). These processes remove settleable solids, suspended solids, organic matter and bacteriological contaminants from wastewater. On the other hand, heterogeneous photocatalysis proves effective for the oxidation of organic recalcitrants which are difficult to degrade in the conventional processes. Even though there are a lot of research publications which have focused on excellent photocatalyst material development and characterization, the main challenge faced is the scalability of photocatalytic reactors. So far, there are various laboratory-scale photocatalytic reactors exhibiting excellent performance. The reason for this is due to the intrusion of light photons in the process, which has major influence over the performance. Therefore, the need of the hour is to design scalable photocatalytic reactors.
It is imperative to construct effective photoreactors to attain maximum efficiency in the performance of photocatalytic processes. The photocatalytic performance is evaluated with respect to energy efficiency (photocatalytic space time yield) and apparent reaction rate (space time yield) (Claes et al. 2019). The main challenge in the inefficacious performance of photocatalytic reactor is due to timeframe difference between light-catalyst interactions (~ 1 µs) and the substrate diffusion inside the porous catalyst (~ 1 s) (Turchi and Ollis 1990). The decrease in characteristic length of photocatalytic reactor increases the possibility of substrate integration with photocatalyst. The optimization of characteristic length might overcome this challenge without compromising on the volume throughput. Numerous applications of microreactors promise to bring technological advancements in terms of faster reaction rate which are not feasible with conventional reactors.
Microreactors were initially used for screening in microanalytical chemistry (Yi et al. 2014), biological investigation of cells and proteins (Tu et al. 2010) and studies on reaction kinetics and mechanisms (Sandel et al. 2012). Microreactors are the best choice due to unique characteristics for quick reactions (Yoshida et al. 2008), highly exothermic reactions (Pelleter and Renaud 2009) and explosive reactions (Zhang et al. 2004;Wootton et al. 2002).
Recently, microreactors have attracted a lot of research publications due to their high surface-to-volume ratio. The main concern in microreactor engineering is the throughput when scaled up to large industrial applications (Claes et al. 2019). Photocatalytic microreactors have high reaction speed; however, they are not very energy-efficient in terms of light utilization. Numerous thousands of microreactors must operate simultaneously in order to reach an industrial throughput. However, it is difficult to efficiently distribute light photons in a multiple scaled-up reactors. Therefore, the employment of a very thin substrate layer and suspended catalyst systems along with high throughput might address the problem of photon transfer, mass transfer and scalability. Due to the simple design features of annular reactors, they are the most common slurry reactors and have received extensive focus in computational fluid dynamics (CFD) modelling (Casado et al. 2016;Venier et al. 2021). Previous studies have discussed modelling in relation to irradiance, chemical reactions, hydrodynamics and mass transfer. There are few works which demonstrated that plug flow regime exhibited by microreactors/micro-channelled reactors show better performance in various applications including photocatalysis (Santos and Kawaji 2010;Wen et al. 2022). But there are no publications on the correlation of hydrodynamic flow regime and performance of photocatalytic reactor to the best of our knowledge.
This research work focuses on unique approach to correlate between characteristic length and photocatalytic performance (Wen et al. 2022;Wang et al. 2021). The flow behaviour is majorly influenced by the reactor geometry.
The prior computational simulation of hydrodynamic flow regime of photocatalytic reactors can facilitate the designers of photocatalytic reactors the ease in designing the reactor geometry as there are no standards in designing of photocatalytic reactors so far. The numerical modelling of three reactors with different geometries and correlation of photocatalytic performance with the hydrodynamic flow behaviour inside the reactors were evaluated. The three types of reactors, namely, cylindrical concentric glass tube micro-reactor (CGTR), strip photocatalytic reactor (STR) and classical annular reactor (CAR), which has different hydrodynamic flow pattern, were chosen in this study from our previous work (Sundar and Sellapa 2022a).

Numerical methods
The simulation of the reactors was performed considering a three-dimensional, steady-state and laminar flow. Ansys Fluent solves conservation equations for mass and momentum. (1) for conservation of mass, or continuity equation, can be exhibited as follows:

Mass conservation Equation
where ρ is the fluid density v is the velocity vector, and S m is addition of mass.

Momentum conservation
The Navier-Stokes equations are utilised for the momentum balance at steady-state conditions (Janczarek and Kowalska 2021). Equation (2) for conservation of momentum can be written as follows: where P is the fluid pressure, τ is the viscous stress tensor, and ρg and F is the gravitational body force and external body forces.
The segregated steady-state solver was used to solve the governing equations. First-order upwind discretization scheme was employed except for pressure for which the second-order method was selected. The SIMPLE algorithm was chosen for the pressure-velocity coupling. Convergence of the numerical solution was ensured by monitoring the scaled residuals to a criterion of at least 10 −6 for the continuity, momentum variables with 1000 iterations. The photocatalytic reactors were simulated in a steady-state mode with a 1-s time step.

Residence time distribution
To confirm the hydrodynamic behaviour and hydrodynamic flow pattern of the three reactors, an RTD approach was performed. One of the most important attributes that need to be considered while examining the hydrodynamic behaviour of the reactor is residence time distribution (RTD) using methylene blue tracer (Casado et al. 2016). RTD is the primary method for determining the time for which the several reactants stayed in the reactor, the mixing of elements inside the reactor, flow of reactants and scaling up of the reactor. The three models used commonly are plug flow, completely stirred flow and arbitrary flow (Fig. 1). The exit age distribution (E(t)) was calculated using Eq. (3): The mean residence time (T) and variance (σ 2 ) was calculated according to Eqs. (4) and (5): where C(t) is the concentration of tracer at the outlet of the reactor. Starting from the system working under stationary circulation of water, 0.05 mol/m 3 methylene blue (SRL chemicals) was injected into the reservoir with a vigorous stirring to assuming perfect mixture in the vessel at time t = 0 (Casado et al. 2016). The absorbance at 660 nm in the reactor outlet was monitored in continuous mode with a time interval of 1 s.

Modified performance parameters
A cubic element of fluid is considered for the calculation of modified space time yield (STY modified ) and modified photocatalytic space time yield (PSTY modified ). The characteristic length of reactor is assumed to be the equivalent to the dimensions of cube as shown in Fig. 2.
Assumptions are made that the fluid elements are travelling at the constant velocity. The modified space time yield (STY modified ) and modified photocatalytic space time yield (PSTY modified ) parameters are correlated with the characteristic length of the photocatalytic reactors below: where t is the residence time, v is the fluid velocity, and l is the characteristic length.

Computational fluid dynamics of reactors
The CFD models of CGTR, STR and CAR have been developed using the commercial software Ansys Student software 2022 (Ansys Inc.). The reactor models of respective dimensions were defined (Sundar and Sellapa 2022b) and modelled using the AnsysSpaceclaim tool (Fig. 3).
The inlet and outlet tubes were modelled for the three reactor models as they have a great impact on the flow field. The quartz tubes in CAR were simulated as an empty cylinder. The meshed reactors are displayed in Fig. 4.
The element size was defined as 2 mm, and high-quality smoothening was utilized for building mesh. The reactor volume was discretized in approximately 940,000 structured and unstructured volume cells using Ansys Meshing tool. This number of cells has been found to be high enough to give mesh-independent results, corresponding to a mean cell volume of 0.01 mm 3 . A denser mesh was defined at the inlet and outlet as there may be greater velocity gradient at both inlet and outlet areas.
The fluent analysis was used to predict the hydrodynamic behaviour of the three photocatalytic reactors. Figure 5 shows the streamlines of velocity magnitude calculated for the CGTR, STR and CAR assuming a steady-state model. In case of CGTR, the streamlines of flow are disturbed at the inlet due to the sudden change in flow direction from vertical inlet to circular pathway, and further through the flow, the streamlines become smooth and even. It is observed that in CGTR, there is minimal amount of transverse mixing. The Reynolds number (R e ) of CGTR is calculated to be 300, which makes it clear that the flow is laminar. The low R e is due to the less characteristic length of the micro-depth reactor of 2 mm (0.002 m). The maximum velocity (0.15 m/s) was taken for the calculation of R e for all reactors. The theoretical narrow laminar profile (in RTD curve) was not reached due to the axial dispersion.
It was observed that at CGTR inlet (Fig. 6), there were small eddies causing some extent of mixing, and further, the flow velocity became laminar throughout the reactor. In case of STR, initially, the flow is even, and later, the flow becomes transient. There is dead zone found at the junction between glass tube and inlet/outlet, where the velocity is lowered due to the sudden increase in the cross-section of flow. The characteristic length dimension of STR is calculated to be 0.0175 m. Reynolds number of STR is calculated to be 2625, which makes it clear that the flow is a transition between laminar and turbulent flow. In the case of CAR, it was observed that velocity streamlines are non-uniform through the annular region. There are dead zones found at the walls of reactor and at the walls of quartz tube where velocity magnitude is very low. The R e was calculated to be 9300, which is very high implying that CAR follows turbulent flow. The characteristic length of CAR was calculated to be 0.

Residence time distribution
The study of hydrodynamics of reactors is very important for the reactions involving fluids. The residence time distribution indicates the type of chemical reactor, and Reynolds number indicates the type of flow.
To confirm the hydrodynamic behaviour and flow pattern of the three reactors, residence time distribution (RTD) approach was performed. One of the most important attributes that need to be considered while examining the hydrodynamic behaviour of the reactor is residence time distribution (RTD) using methylene blue tracer. Figure 7 shows the concentration curve of RTD data, where there is a short delay during which little or no tracer was detected at the outlet till 4 s. At the 5th second, there was Furthermore, Fig. 8 shows the exit age distribution E(t) for three types of reactor from experimental analysis. The RTD of CGTR denoted that all the tracer fluid elements systematically passed the reactor without transverse mixing. The RTD of CAR indicated the profile of ideal continuously stirred tank reactor which is characterised by perfect back mixing. At t = 0 s, the tracer concentration was identified at the outlet. This showed that CAR is perfectly mixed flow, and each portion of the fluid elements has the same chance to be discharged at the outlet, regardless of how long it has already been inside the continuously stirred tank reactor. There is a slow decrease in the tracer concentration at the outlet till 4 s.
The RTD of STR showed that there was sudden rise in tracer concentration at outlet at 1 s; later, there was a decline in the concentration of tracer till 4 s. There was some degree of plug flow due to the initial sharp peak, and there was also some degree of mixed flow due to the slow decrease in concentration. This showed that STR was

Comparison of reactors
In our previous work (Sundar and Sellapa 2022a), it was shown that photocatalysis with simultaneous adsorption and hydrodynamic caviation (PCSA + HC) exhibited the maximum colour removal of 100% in CGTR. Similarly, under optimised conditions, experiments ((1) photocatalysis with initial adsorption -PCIA, (2) photocatalysis with initial  (4) photocatalysis under simultaneous adsorption and hydrodynamic cavitation -PCSA + HC) including control experiments (photolysis and hydrodynamic cavitation -HC) were performed for STR and CAR. The % colour removal achieved in the three reactors is illustrated in Fig. 9. The experiments optimised conditions were initial pH 11 and graphene dosage 0.3 g/L, AO dye concentration 10 ppm and irradiation time 35 min (Sundar and Sellapa 2022b).
The experiments were carried out in triplicate to obtain the standard deviation error of < 5%. The STR exhibited the second highest removal (85.8%) under PCSA + HC. In CAR, 57% of colour was removed under PCSA + HC. These results implied that CGTR performed better than STR and CAR. Under PCSA + HC, there is synergetic effect of photocatalysis, adsorption and hydrodynamic cavitation for increased degradation rate (Asgari et al. 2021).
To get a better understanding of the performance and energy efficiency among micro-reactor (MR) (Visan et al. 2014), packed bed reactor (PBR) [1], CAR, STR and CGTR, the modified space time yield (STY modified ) and modified photocatalytic space time yield (PSTY modified ) is plotted in Fig. 10 and represented in Table 1.
Space time yield and photocatalytic space time yield is a benchmark proposed by Leblebici et al. (2015), which is dependent on the hydrodynamic behaviour (either continuously stirred tank reactor or plug flow reactor) and reaction rate constant. Space time yield is dependent on reaction rate constant and hydrodynamic flow behaviour whereas photocatalytic space time yield is dependent on space time yield and illumination efficiency. The modified parameters STY modified and PSTY modified are calculated using the hydrodynamic flow behaviour influential parameter, which is called characteristic length of photocatalytic reactor. The reactor which performs best in pollutant removal (productivity) is present on the top side of graph whereas a reactor which performs best in energy utilisation is present on the right side of graph.
It was observed from Fig. 10 that CAR and PBR (Claes et al. 2019) are positioned near the origin. This implies that these reactors have less productivity (less STY modified ) and less energy efficiency (less PSTY modified ). The less values of STY modified of CAR and PBR are attributed to fluid velocity and characteristic length, where the higher characteristic length and lower fluid velocity resulted in the lower values of STY modified . On the other hand, PSTY modified depends on the STY modified and lamp power required to illuminate the volume of reactor. Also, the reactors which are operated in recirculation have higher modified parameters compared to continuously operated reactors. STR has lower characteristic length and higher fluid velocity compared to CAR implying higher modified parameters. Furthermore, the position of MR (Visan et al. 2014) in the right bottom in the graph (high STY modified and less PSTYmodified ) reveals that the fluid velocity also has significant influence on modified parameters even though the characteristic length is least in MR (Visan et al. 2014). In the case of CGTR, both high productivity and high energy efficiency are achieved due to its high fluid velocity and less characteristic length and operated in recirculation process. Therefore, this study showed that CGTR was the most efficient in terms of both performance and energy efficiency.
The interpretations of modified parameters revealed that the maximum modified space time yield (STY modified ) 225 s −1 and maximum modified photocatalytic space time yield (PSTY modified ) 9.375 s −1 /kW/m 3 was obtained for plug flow reactor with lesser characteristic length when compared to continuously stirred tank reactor (CAR) and arbitrary flow reactor (STR) which has higher characteristic length. The treated throughput (1 L) and fluid velocity (0.15 m/s) are kept constant in CAR, STR and CGTR.

Conclusions
In the performance of photocatalysis, the major influencing factors are light source, photocatalyst material and photocatalytic reactor design. The parameter that plays an important role on the pollutant removal efficiency is design of photocatalytic reactor. Besides the shape of reactor, the other factors such as inlet pipe, liquid pump, hydrodynamic regime (continuously stirred tank reactor or plug flow reactor), depth of substrate, distance between light source and fluid also contribute to the overall performance efficiency of the photocatalytic process. Even though the use of LEDs could increase the energy efficiency of the photocatalytic reactor, the LED orientation and depth of pollutant could not be a trade-off to attain high performance and energy efficiency.
In this research, an attempt is made to correlate and compare the hydrodynamic flow regime of photocatalytic reactors and the photocatalytic performance of Acridine orange removal in three photocatalytic reactors with different geometrical designs.
The CFD modelling and residence time distribution concluded that flow behaviour of CGTR depicted ideal plug flow model. CAR depicted the profile of ideal continuously stirred tank reactor, which is characterised by perfect back mixing and STR depicted the profile in between plug flow reactor and continuously stirred tank reactor, which is characterised as arbitrary flow. The computational fluid dynamics modelling of reactors depicted the velocity profile where Reynolds number (Re) is 300 for CGTR indicating laminar flow (plug flow regime), Re for CAR is 9300 indicating turbulent flow (continuously stirred tank reactor), and Re for STR is 2625 indicating transient flow.
Furthermore, when performance of reactors is evaluated, it was observed CGTR was the best configuration compared to STR and CAR. The highest modified space time yield (STY modified ) 225 s −1 and highest modified photocatalytic space time yield (PSTY modified ) 9.375 s −1 /kW/m 3 was obtained for plug flow reactor when compared to continuously stirred tank reactor (CAR) and arbitrary flow reactor (STR). Therefore, the study proves that plug flow reactor designed with higher throughput (CGTR) performed better when compared to other microreactors, micro-simulated reactors with lesser throughput (MR and PBR) and continuously stirred tank reactor with higher throughput (CAR) and arbitrary flow reactor higher throughput (STR). For characteristic length inclusive comparison of performance, the PSTY modified (x-axis) and STY modified (y-axis) was plotted for the reactors, which proved that CGTR was best performing in terms of dye removal (productivity) and energy efficiency. This study throws light on two aspects: (i) future research works may focus on choosing the characteristic length for the photocatalytic reactors as an important attribute for performance optimization studies; (ii) micro-depth reactor design concepts might be explored to attain plug flow regime without compromising volume throughput and energy efficiency.
Author contribution 1. Kaviya Piriyah Sundar (Corresponding author) had contributed in execution of research work and writing the manuscript.
2. Dr. Kanmani Sellapa guided the complete research work and reviewed the manuscript.
3. Dr. Mahalakshmi Nainangkuppam Venkatesan had contributed in the geometrical modelling of reactors in the software.
All the authors agreed with the content and that all gave explicit consent to submit the manuscript.
Funding The research work was funded under Inspire fellowship by the Department of Science and Technology, Ministry of Science and Technology, Government of India.
Data availability This work is a continuation of the previous research work for substantial establishment of engineering concepts for reactor design.

Declarations
Ethical approval The authors have complied with the Ethical Standards of the Journal.
Consent to participate Not applicable for this study.

Consent for publication
We have obtained consent from the institution where the work has been carried out, before the work is submitted.

Competing interests
The authors declare no competing interests.