Scale-up modelling and life cycle assessment of electrochemical oxidation in wastewater treatment

1 Scale-up modelling and life cycle assessment of 1 electrochemical oxidation in wastewater treatment 2 Sara Feijoo1,*, Sofía Estévez2, Mohammadreza Kamali1, Raf Dewil1,3, and María Teresa 3 Moreira2 4 1 KU Leuven, Department of Chemical Engineering, Process and Environmental Technology Lab, 2860 Sint-Katelijne-5 Waver, Belgium 6 2 Universidade de Santiago de Compostela, Department of Chemical Engineering, CRETUS, 15782 Santiago de 7 Compostela, Spain 8 3 University of Oxford, Department of Engineering Science, Parks Road, Oxford, OX1 3PJ, United Kingdom 9 * Corresponding author: Sara Feijoo, [email protected] 10 Keywords— Electrochemical Advanced Oxidation Processes (eAOPs), Life Cycle Assessment (LCA), 11 wastewater treatment, scale-up modelling, micropollutants 12 Abstract 13 The need to improve current wastewater treatments to ensure a clean and sustainable water 14 supply is an unquestionable contemporary challenge. It is therefore essential to facilitate 15 knowledge transfer between research institutions and the industry by developing novel 16 technologies to a proof-of-concept stage, demonstrating both treatment efficiency and 17 compliance with environmental criteria. This study has combined process modelling for the 18 design of an electrochemical Advanced Oxidation Process (eAOP) to remove carbamazepine 19 (CBZ) from wastewater with the identification of the environmental impacts associated with its 20 operation. A comprehensive set of scenarios considering several reactor designs and operating 21 conditions provides the assessment framework to identify the influence of different process 22 variables on the environmental profile of the pilot-scale eAOP. The most sustainable treatment 23 corresponds to the operation of a standardised modular reactor in batch mode, especially when 24 the wastewater has a low concentration of scavengers, such as other ions, organics or 25 pollutants. Nevertheless, in all scenarios evaluated, the main environmental hotspot was 26 attributed to the electrical energy consumed by the auxiliary pumps rather than the 27 electrochemical reactor itself. In comparison to other AOPs, our system showed considerably 28 lower impacts in the global warming potential (GWP) category, with a minimum of 7.6 kg CO2 29 eq per g CBZ removed for the most promising scenario. This demonstrates the implementation 30 potential of eAOPs as well as the importance of data from scaled-up experiments, where 31 optimisation should focus on mitigating the impacts of energy-intensive pieces of equipment. 32 Abbreviations: API – Active pharmaceutical ingredient, B – Batch mode, BDD – Boron-doped diamond, C – Continuous mode / 33 Chemical, CBZ – Carbamazepine, CECs – Contaminant(s) of emerging concern, COD – Chemical oxygen demand, CSTR – Continuous 34 stirred tank reactor, E – Energy, eAOP(s) – Electrochemical advanced oxidation process(es), EC – Enhanced conductivity, FB – Fed-35 batch mode, FE – Freshwater eutrophication, FRS – Fossil resource scarcity, FU – Functional unit, GAC – Granular activated carbon, 36 GWP – Global warming potential, H – High content, L – Low content, LCA – Life cycle assessment, M – Multicomponent, ME – 37 Marine eutrophication, NF – Nanofiltration, RL – Regulatory limits, SDG – Sustainable development goal, SPF – Solar photo-Fenton, 38 SW – Synthetic wastewater, TA – Terrestrial acidification, TET – Terrestrial ecotoxicity. 39 2 1 Introduction 40 In 2020, approximately 2 billion people lacked safely managed drinking water, 2.3 billion people 41 suffered from poor hygiene and up to 3.6 billion people did not have access to basic sanitation, 42 which has raised some concern about the accomplishment of Sustainable Development Goal 43 (SDG) No. 6 on Clean Water and Sanitation by 2030 [1]. One of the root causes is the occurrence 44 of contaminants of emerging concern (CECs) in wastewater treatment plant effluents, which 45 represents a major issue not only to human health but also to ecosystems [2–5]. Carbamazepine 46 (CBZ) is one of these contaminants since it is a pharmaceutical poorly removed by conventional 47 biological treatment (i.e., removal efficiency is typically lower than 10%) [6–8]. In fact, due to 48 its widespread consumption and recalcitrant nature, CBZ has been recently found to be the most 49 recurring active pharmaceutical ingredient (API) in river basins worldwide [9]. 50 One of the solutions to this problem is the development novel wastewater treatments that 51 prevent the release of pollutants through their effective degradation, as is the case of Advanced 52 Oxidation Processes (AOPs). AOPs are an extensive family of treatments comprising ozonation, 53 heterogeneous and homogeneous (photo)catalysis, Fenton and Fenton-like processes, and 54 electrochemical-, ultrasound-, microwaveor gamma-radiation treatments as well as any of 55 their combinations [10]. Among these various technologies, electrochemical Advanced 56 Oxidation Processes (eAOPs) have received significant attention in recent years [11–13]. They 57 are commonly used as tertiary wastewater treatments, driving pollutant degradation through 58 direct and indirect oxidation pathways by electrochemically generating highly reactive oxidative 59 species, mainly hydroxyl (•OH) and sulfate (SO4•−) radicals [14–16]. Electrochemical AOPs allow 60 for high degradation efficiencies and reaction rates under mild conditions, while showing no or 61 limited dependence on chemical addition [2, 17]. Other advantages are their versatility, ease of 62 process integration and safe operation [13, 18]. 63 When implementing an eAOP, the selection of the electrode material and the precursor species 64 for the oxidative radicals are key factors influencing the overall treatment efficiency and 65 selectivity [14]. Boron-doped diamond (BDD) electrodes are of particular interest for 66 wastewater applications, as they have demonstrated high efficiency in the generation of 67 oxidative species and degradation of several contaminants, as well as high conductivity, 68 stability, O2 overpotential and durability [17, 19, 20]. Despite the significant energy consumption 69 associated with electrochemical treatments [15], the in situ radical generation offered by the 70 BDD material from water molecules [21] and ionic species such as sulfate ions [22, 23] has a 71 high added value for industrial implementation, given that they are already available in 72 wastewater streams [24, 25]. The absence of additional chemicals can minimise not only the 73 overall consumption of raw materials but also the generation of secondary waste streams and 74 hence the associated environmental impacts [10]. Consequently, achieving SDG No. 6 while 75 aiming for sustainable and carbon neutral processes is essential to provide a far-reaching 76 solution. In this regard, the Life Cycle Assessment (LCA) methodology is a useful resource for 77 evaluating the environmental friendliness of novel eAOPs. Nonetheless, as of June 2022, a 78 Scopus search for studies applying LCA methodology to electrochemical oxidation in 79 wastewater treatment retrieved 281 documents, of which only 8 publications specifically 80 included an electro-oxidation system (Table C.1). Among these studies, none were dedicated to 81 the removal of pharmaceuticals, 5 were applied to synthetic or real wastewater matrices, and 82 3 only two considered wastewater volumes at a large scale. Consequently, there is a significant 83 knowledge gap on the environmental implications of eAOPs in wastewater treatment. 84 In our previous work [26], two preliminary design considerations for the implementation of a 85 BDD-based eAOP as a tertiary wastewater treatment were addressed: the effects of the 86 wastewater composition and the reactor mode of operation. A comparative assessment of 8 87 different scenarios, both in terms of CBZ degradation and electrical energy consumption per unit 88 of effective operation time, revealed that the competition reactions taking place due to 89 wastewater components could be mitigated when operating in fed-batch mode, since a 2.1-90 fold increase in CBZ degradation and a 60% reduction in energy consumption were achieved with 91 respect to a conventional batch operation. Similarly, operating in continuous rather than batch 92 mode resulted in significant energy savings (approximately 19%) for similar degradation 93 efficiency. However, in order to further evaluate the advantages and disadvantages of each of 94 the investigated scenarios, it is essential to take into account their potential environmental 95 impact when applied on a larger scale. In fact, most conventional municipal wastewater 96 treatment plants already involve high energy consumption from the grid due to all the machinery 97 involved [27], leading to a significant carbon footprint (i.e., 23–432 kg CO2 per population 98 equivalent) [27–29], of which approximately 70% was attributed to the indirect emissions from 99 energy requirements [29]. 100 To fill the knowledge gap on the environmental performance of pilot-scale eAOPs for the 101 removal of pharmaceuticals from secondary wastewater effluents, this study focused on 102 conducting a techno-environmental analysis including the following: 103 (i) Development of a scale-up model to translate the laboratory results into a pilot-scale 104 operation. To this end, two reactor configurations have been considered: a standardised 105 modular reactor and a vertical plate stirred tank reactor. 106 (ii) Quantification of the environmental profile of the eAOP by LCA methodology under 107 multiple experimental conditions. More specifically, the influence of the reactor 108 configuration, the mode of operation (namely, batch, fed-batch and continuous), the 109 wastewater matrix (considering various compositions of pure and synthetic wastewater, 110 different amounts of oxidising species and the possible presence of additional 111 pollutants) and the potential oversizing effect have been evaluated. 112 2 Methodology 113 2.1 Experimental scenarios 114 The micropollutant degradation experiments were carried out using a BDD electro-oxidation 115 system, as previously described by Feijoo et al. (2022) [26]. Both single and multicomponent 116 systems were primarily aimed at CBZ removal, where a diverse set of concentrations for 117 oxidative and scavenger species were investigated. The reactor operating modes included in the 118 comparative analysis were batch, fed-batch and continuous. As a result, the following 8 119 scenarios were evaluated in the techno-environmental analysis: 120 4 • Scenario of “Regulatory Limits for Sulfates and Nitrates Conducted in Batch Mode (RL-121 B)”: CBZ degradation was carried out in batch mode and in a pure water matrix 122 containing the concentration limits for nitrate (50 mg/L) and sulfate (250 mg/L) ions as 123 defined by their respective EU directives [30, 31]. 124 • Scenario of “Enhanced Conductivity Medium Conducted in Batch Mode (EC-B)”: an 125 extension of the RL-B scenario assumed that nitrate and sulfate concentrations were 126 higher than the regulatory limits, at 100 mg/L and 500 mg/L, respectively. 127 • Scenario of “Enhanced Conductivity Medium in Synthetic Wastewater with Low Organic 128 Load Conducted in Batch Mode (ECSWL-B)”: CBZ degradation was performed in batch 129 mode and in the presence of the enhanced nitrate and sulfate concentrations as in the 130 EC-B scenario. The treated water matrix consisted of a synthetic secondary effluent with 131 low concentrations of other organics and ions (COD: 25.2 mg/L, total N: 5.0 mg/L, total 132 P: 0.5 mg/L, alkalinity: 2.5 mg/L). 133 • Scenario of “Enhanced Conductivity Medium in Synthetic Wastewater with Higher 134 Organic Load Conducted in Batch Mode (ECSWH-B)”: this variation to the ECSWL-B 135 scenario consisted of the degradation of CBZ in a synthetic wastewater matrix with a 136 high ion and organic composition (COD: 50.4 mg/L, Total N: 10.0 mg/L, Total P: 0.9 mg/L, 137 Alkalinity: 4.9 mg/L). 138 • Scenario of “Enhanced Conductivity Medium in Synthetic Wastewater with Low Organic 139 Load Conducted in Fed-Batch Mode (ECSWL-FB)”: this modification to the ECSWL-B 140 scenario consisted of fed-batch operation, where CBZ spikes were added at the beginning 141 of each 60 min cycle for a total of 6 cycles to reuse sulfate and nitrate species already 142 present in the wastewater. 143 • Scenario of “Enhanced Conductivity Medium in Synthetic Wastewater with Low Organic 144 Load Conducted in Continuous Mode (ECSWL-C)”: the ECSWL-B was adapted to a 145 continuous operation, that is, with continuous inlet and outlet flows set to 25 mL/min, 146 leading to an average residence time of 30 min. 147 • Scenario of “Multicomponent System in Synthetic Wastewater with Low Organic Load 148 Conducted in Batch Mode (MSWL-B)”: this variation to the ECSWL-B scenario consisted 149 of the simultaneous degradation of CBZ with additional micropollutants, including 150 caffeine, diclofenac and sulfamethoxazole. 151 • Scenario of “Multicomponent System in Synthetic Wastewater with Low Organic Load 152 Conducted in Fed-Batch Mode (MSWL-FB)”: this modification to the MSWL-B scenario 153 was conducted in fed-batch mode for 6 cycles of 60 min with multicomponent spikes. 154 2.2 Selected reactor designs 155 After conducting a review of available configurations for pilot-scale BDD electrochemical 156 reactors, it was observed that a large number of studies considered commercial DiaCell® units 157 5 [32–38], filter press flow cells [38–41], multielectrode stacks with a serpentine array [39–41], 158 or a vertical electrode plate arrangement in a stirred tank reactor [39, 42, 43]. In this study, the 159 two designs selected were (i) a standardised modular reactor inspired by the DiaCell® units and 160 (ii) a fully customised vertical plate stirred tank reactor (Fig. 1). Both configurations are 161 commonly reported in the literature, feasible to scale up and significantly different from each 162 other in terms of area, geometry and distance between the electrodes. 163 164 (a) Standardised modular reactor. (b) Vertical plate stirred tank reactor. 165 Figure 1: Schematics of the (a) standardised modular reactor and (b) vertical plate stirred tank reactor. For their 166 fed-batch operation, a dosing pump is added. 167 2.2.1 Standardised modular reactor design 168 The standardised modular reactor configuration was based on the DiaCell® 1001 electrochemical 169 cell [34, 36]. It comprises multiple compartments constituted by two BDD anodes and one 170 stainless steel cathode with an interelectrode distance of 1 mm, leading to a total of 10 anodes 171 and 5 cathodes per cell. Standard shapes for the electrodes are circular, with a surface area of 172 70 cm2 and monopolar connections (Fig. 1a). During its operation, a process tank is loaded with 173 the secondary wastewater to be treated, and if needed, additional chemicals are added. 174 Afterwards, the content of the tank is continuously stirred and fed to the standardised modular 175 reactor, where it is distributed between five compartments in parallel. The system operates in 176 recirculation mode, meaning that the total volume of wastewater remains constant and is 177 recirculated until the desired degradation is attained. Consequently, this reactor design is 178 applicable for batch and fed-batch operations. Finally, the treated effluent is accumulated in 179 the process tank and discharged. 180 2.2.2 Vertical plate stirred tank reactor design 181 The vertical plate stirred tank reactor consists of a set of parallel monopolar electrodes that are 182 fully immersed in the bulk of the reactor (Fig. 1b). The number of electrode pairs as well as their 183 size and arrangement are versatile parameters, and hence, any reactor design can be 184 implemented. To avoid any damage to the electrodes during operation, stirring inside the 185 reactor is promoted by the inlet and the recirculation pump flows. In addition, the electrode 186 channels can contain an inert polymer mesh and other turbulence promoters to improve mass 187 transfer. It is assumed that the current density and voltage are uniformly distributed across the 188 6 cell. This type of setup allows for either a batch, fed-batch or continuous operation with 189 recirculation. 190 2.3 General scale-up considerations 191 Based on the collected experimental data from laboratory experiments in a 1 L electrochemical 192 cell, the scale-up target was to model the steady-state conditions in a 100 L reactor filled up to 193 75% of its capacity and where 90% CBZ degradation can be attained. The scale-up methodology 194 consisted of analysing the experimental results based on the reaction kinetics, electrical 195 consumption and treatment capacity. This enabled mass and energy balances to be performed 196 at the pilot scale, with the required pieces of equipment (i.e., electrochemical cell and associated 197 pumps) designed accordingly. Relevant scale-up considerations are defined in the following 198 subsections. 199 2.3.1 Common design conditions 200 To compare scenarios under the same time reference, all reactor designs were simulated to 201 operate for 1 day (i.e., 24 h). The number of batch and fed-batch experiments during that time 202 to achieve 90% removal of CBZ were calculated considering the effective reaction times observed 203 experimentally. In addition, a total of 25 min was considered per experiment to account for 204 preparation, charge and discharge activities. 205 The starting concentrations of the different chemicals involved were assumed to be the same as 206 in the experiments at the lab scale, given that they are independent of the reactor type and size. 207 Therefore, their total initial mass was directly proportional to the scaled-up reactor volume. For 208 the addition of sulfate and nitrate ions, only the differential concentrations with respect to the 209 regulatory limits were considered as input chemicals in the LCA inventory, given that it is 210 plausible that the regulatory limits may already be found in the influent wastewater. In the case 211 of fed-batch operation, it was assumed that a concentrated stream of 200 mg/L CBZ was used 212 for the spikes to guarantee that volume variations after their addition during 1 day of operation 213 would not yield to more than an overall 10% increase. 214 2.3.2 Mass balance assumptions 215 Since the kinetic constants (k, h−1) were determined from lab-scale experiments, a correction 216 factor was applied to estimate the final CBZ concentrations in the pilot-scale standardised 217 modular reactor. The need for a correction factor in this specific reactor configuration arises 218 from the differences in the number of electrodes and subsequent electroactive areas between 219 the lab-scale reactor used and the scaled-up design. These differences lead to distinct area-to-220 volume ratios, and therefore, the variation in kinetic constants has been estimated accordingly. 221 As shown in Eq. 1, k is related to the mass transfer coefficient (km, m/h), a pseudo-first order 222 kinetic constant related to the activity of inorganic oxidants (ki, h−1), the electroactive area (A, 223 m2) and the reactor volume (V, m3) [34, 44]. Assuming that ki is negligible in our system as 224 oxidants are present in excess and that km remains constant with increasing scale, the observed 225 7 kinetics are affected by the A/V ratio. Consequently, the kinetic rate constants in the batch and 226 fed-batch scaled-up standardised modular reactor (kscale, h−1) were calculated as shown in Eq. 227 2, where A and Ascale are the electroactive areas (m2) at the lab and pilot scales, respectively, 228 and V and Vscale are the volumes (m3) of treated wastewater at the lab and pilot scales, 229 respectively. 230 𝐶𝐶𝐶𝐶𝐶𝐶 =𝐶𝐶𝐶𝐶𝐶𝐶0∙𝑒𝑒(−𝑘𝑘∙𝑡𝑡)=𝐶𝐶𝐶𝐶𝐶𝐶0∙𝑒𝑒�−�𝐴𝐴 𝑉𝑉∙𝑘𝑘𝑚𝑚+𝑘𝑘𝑖𝑖�∙𝑡𝑡� (1) 𝑘𝑘𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 =𝑘𝑘 ∙𝐴𝐴𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑉𝑉𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 � 𝐴𝐴𝑉𝑉 � (2) Similarly, the modelling for the continuous operation was based on the definition of an ideal 231 continuous stirred tank reactor (CSTR) (Eq. 3), where X is the conversion of the target pollutant 232 obtained experimentally and τ is the residence time (h). After substitution of common terms 233 with Eq. 2, the conversion in the scaled-up standardised modular reactor (Xscale) was obtained 234 from Eq. 4, where F and Fscale are the flow rates (m3/h) of treated wastewater at the lab and 235 pilot scales, respectively. Given that a target of 90% CBZ removal was selected, Eq. 4 was used 236 to retrieve the required flow rate at the pilot scale [34, 45]. 237 𝑘𝑘𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 =𝑋𝑋 1−𝑋𝑋 ∙1 𝜏𝜏 (3) 𝑋𝑋𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 1−𝑋𝑋𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 =𝑋𝑋 1−𝑋𝑋∙𝐴𝐴𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝐹𝐹𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 � 𝐴𝐴𝐹𝐹 � (4) Regarding the vertical plate stirred tank reactor, the experimental kinetic constants were used 238 in the mass balance since the area-to-volume ratio was considered constant. For other 239 wastewater components, it was assumed that they were present in excess and that variations 240 in concentration during the treatment were negligible. In addition, the consumption of NaOH to 241 neutralise acidic outlet streams before discharge was also calculated at the pilot scale and 242 included in the mass balance. 243 2.3.3 Energy balance assumptions 244 The limiting current density (jlim, A/m2) of each treatment was estimated based on the model 245 developed by Panizza et al. (2001) defined in Eq. 5, where F is the Faraday constant (C/mol), km 246 is the average mass transport coefficient in the electrochemical cell (m/s) and COD is the 247 chemical oxygen demand expressed in mol O2/m3 [46]. The mass transport coefficient (km) at 248 the pilot scale was estimated according to the correlations as a function of the flow rate 249 proposed by Anglada et al. (2009) [47], with a maximum value of approximately 1.7·10−5 m/s 250 for a 10 L/min flow. 251 𝑗𝑗𝑠𝑠𝑙𝑙𝑙𝑙 = 4 ∙𝐹𝐹 ∙𝑘𝑘𝑙𝑙∙𝐶𝐶𝐶𝐶𝐶𝐶 (5) As a result, scenarios involving wastewater with low and high concentrations of organics and 252 8 other ionic species showed estimated limiting current densities of 5.2 and 10.3 A/m2, 253 respectively. Given that experiments were performed at higher current densities, it can be 254 concluded that electrochemical oxidation is under mass transport control and that pollutant and 255 COD removal follow an exponential trend. 256 The electrical energy consumption by the pilot-scale pumps (Ppump, kWh), which depends on the 257 supplier catalogue nominal power (Pn), was determined based on the modelled operation time 258 (t, h), as shown in Eq. 6. The operation time for the recirculation, inlet and outlet pumps 259 corresponded to the actual reactor operation, whereas for the pumps dedicated to individual 260 charge, discharge and dosing operations, it was calculated as the time required to transport a 261 scaled-up volume of liquid (Vscale, m3) at a specific flow rate (Fscale, m3/h), as shown in Eq. 7. 262 𝑃𝑃𝑝𝑝𝑝𝑝𝑙𝑙𝑝𝑝 =𝑃𝑃𝑛𝑛∙𝑡𝑡 (6) 𝑡𝑡=𝑉𝑉𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝐹𝐹𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 (7) For batch and fed-batch operations in both reactor configurations, centrifugal pumps used for 263 charge/discharge operations were assumed to be similar to the model KPM 50 by Speroni S.p.A., 264 which has a Pn of 0.37 kW and can operate between 5 and 30 L/min [48]. The selected flow rate 265 for the charge/discharge pumps was 15 L/min to minimise time losses during the 1-day 266 operation. During the electrochemical treatment, the same pump type was considered in both 267 reactors to drive a continuous recirculation at 10 L/min to ensure a Reynolds number higher than 268 500 [49]. 269 The dosing pump for the fed-batch operation in both reactor configurations was assumed to be 270 similar to the Model A peristaltic pump by Redox.me, which has a Pn of 0.04 kW and can operate 271 between 0.07 and 380 mL/min [50]. The selected flow rate was approximately 5 mL/min. 272 For continuous operation, which is only applicable to the vertical plate stirred tank reactor, 273 model KPM 50 was also selected for the recirculation pump, whereas the WT600F-65/KZ25 274 model by Golander Pump was chosen as the inlet/outlet pump, with a Pn of 0.2 kW and a flow 275 rate window between 0.25 and 6 L/min [51]. The selection of these pumps is justified based on 276 an in-depth analysis regarding the scaled-up flow rates needed, since the flow rate directly 277 influences the mass transfer and the residence time inside the reactor, and hence, the overall 278 conversion and the electrical energy consumption. Given that it was desired to ensure a 279 recirculation flow rate with a Reynolds higher than 500, an approximate recirculation ratio of 10 280 was required (Fig. A.1a). Based on that ratio, the effect of the selected influent flow rate on the 281 entire electrochemical system was investigated. As depicted in Fig. A.1b, an influent flow rate 282 of 0.75 L/min was required for an overall 90% conversion, meaning that the recirculation flow 283 rate had to be approximately 8.2 L/min. Both these flow rates can be delivered with the selected 284 pumps, and therefore, their catalogue nominal power allowed for a suitable energy estimation. 285 9 2.4 LCA framework 286 2.4.1 Goal and scope 287 The goal of the Life Cycle Assessment (LCA) study was to evaluate the environmental profile of 288 the pilot-scale electrochemical oxidation of CBZ when several secondary wastewater 289 compositions and reactor configurations were involved. Therefore, attention was paid to the 290 operation stage, and scenarios were evaluated from a gate-to-gate perspective. That is, the 291 operation of the electrochemical reactor and its associated pumps was considered, whereas the 292 impacts related to construction, decommissioning, upstream and downstream processes were 293 excluded. The analysis consisted of an attributional LCA following ISO standards 14040:2006 294 and 14044:2006 [52, 53]. 295 2.4.2 Assessment method 296 The LCA was performed using the ReCiPe MidPoint (H) V1.06/World (2010) and EndPoint (H/H) 297 V1.06/World (2010) methods [54] in SimaPro 9.3.0.2. software [55]. The following impact 298 categories were selected as they are representative of energy, toxicity and water effects: global 299 warming potential (GWP), terrestrial acidification (TA), freshwater eutrophication (FE), marine 300 eutrophication (ME), terrestrial ecotoxicity (TET) and fossil resource scarcity (FRS). Additional 301 results on other impact categories can be found in the Supplementary Material, Appendix B. The 302 functional unit (FU) selected was 1 mg of CBZ removed per cubic metre of wastewater treated 303 during one day of operation, and hence, its units are mg/(m3·day). Based on the different reactor 304 configurations modelled at the pilot scale and the estimated inventories (Tables A.4 and A.5), 305 the following environmental analyses were conducted: 306 (i) To elucidate the influence of the reactor operating mode, the results of the scale-up 307 modelling for all experimental scenarios were analysed in terms of chemical and energy 308 requirements per FU in the vertical plate stirred tank reactor. Based on these results, a 309 benchmark on the environmental profiles of the batch, fed-batch and continuous modes 310 was conducted. To this end, the ECSWL-B, ECSWL-FB and ECSWL-C scenarios were 311 compared. The results reported correspond to the midpoint assessment method. 312 (ii) To discern the effect of the wastewater matrix, a benchmark on the environmental profiles 313 when diverse influent compositions are treated in the standardised modular reactor in 314 batch mode was conducted. To this end, the RL-B, EC-B, ECSWL-B, ECSWH-B and MSWL-315 B scenarios were compared. The results reported correspond to the midpoint assessment 316 method. 317 (iii) To determine the influence of the reactor configuration, a benchmark on the environmental 318 profiles of the standardised modular reactor and the vertical plate stirred tank reactor 319 operated in batch mode was conducted. To this end, the RL-B, EC-B, ECSWL-B, ECSWH-320 B and MSWL-B scenarios were compared for both reactor types. The results reported 321 correspond to the midpoint and endpoint assessment methods. 322 (iv) The oversizing effect was analysed for the standardised modular reactor. Given that this 323 16 3.2.2 Benchmark of wastewater compositions 479 To elucidate the environmental effects of the influent wastewater composition, the scenarios 480 operated in batch mode (i.e., RL-B, EC-B, ECSWL-B, ECSWH-B and MSWL-B) were analysed 481 under the standardised modular reactor configuration. Their relative contributions across the 482 selected LCA midpoint categories are shown in Fig. 5. It can be observed that the contributions 483 of the different scenarios are uniform across all impact categories, with the scenarios with the 484 most complex wastewater matrices (i.e., MSWL-B and ECSWH-B) being the predominant ones. 485 The scenarios in pure water (i.e., RL-B and EC-B) corresponded to less than 22% of the impact 486 of the multicomponent system MSWL-B. In addition, between those two, the addition in scenario 487 EC-B of sulfate and nitrate species above the regulatory limits contributed to an overall 488 reduction over the RL-B scenario for all categories, except for terrestrial ecotoxicity (TET), where 489 the relative impact was 0.1% higher. This is due to the enhanced degradation kinetics by 490 increasing the amount of oxidative radical sources, which translates into a reduced operation 491 time and hence a lower energy consumption (Table A.4). Nonetheless, a pure water-based 492 operation differs from what in practice a wastewater treatment plant will be dealing with. 493 Therefore, scenarios ECSWL-B, ECSWH-B and MSWL-B are more interesting from an 494 implementation perspective. From their relative differences, it can be argued that increasing the 495 wastewater matrix complexity also negatively affected the environmental profile of the 496 treatment (i.e., an increase of 186-264% between ECSWL-B and MSWL-B in all categories), given 497 that the more competition reactions taking place, the slower the CBZ degradation and the higher 498 the energy consumption for the same removal target. In fact, the main contributor to the 499 environmental impacts of these three scenarios was the electricity attributed to the recirculation 500 pump (Fig. 6), accounting for 51-74% in the categories of global warming potential (GWP), 501 terrestrial acidification (TA), freshwater eutrophication (FE) and fossil resource scarcity (FRS). 502 Regarding terrestrial ecotoxicity (TET), the contribution of the recirculation pump was slightly 503 lower, although predominant (i.e., 36-49%). Sodium nitrate was the main contributor in the 504 marine eutrophication (ME) category, accounting for 59-62% of the overall impact. 505 3.2.3 Benchmark of reactor configurations 506 Regarding the environmental profile of the different reactor configurations in batch mode, it was 507 observed that the vertical plate stirred tank reactor consistently presented higher LCA midpoint 508 impact values than the standardised modular reactor (approximately 23-54% higher) regardless 509 of the experimental scenario and the category considered (Fig. 7). An increasing trend in impact 510 values was also observed with regard to the wastewater matrix complexity, as previously 511 elucidated. Only the ECSWH-B and MSWL-B scenarios displayed the same impact in the marine 512 eutrophication (ME) category for both reactors, with less than a 3% difference. 513 The contributions of the different categories for the vertical plate stirred tank reactor (Fig. B.6) 514 were analogous to those mentioned above for the standardised modular reactor (Fig. 5). 515 Therefore, the benchmark between both reactor types was conducted from an endpoint 516 perspective, as it led to more accentuated differences among the experimental scenarios in 517 batch mode. As depicted in Fig. 8, the single score indicator allocated to the consumption of 518 chemicals was very similar for both reactors regardless of the scenario under consideration. In 519 addition, its value showcased a mild increase with increasing wastewater complexity. On the 520 17 other hand, higher scores and more evident differences were found regarding energy 521 requirements. For scenarios treating pure water matrices (i.e., RL-B and EC-B), the vertical plate 522 stirred tank reactor scored 30-33% higher than the standardised modular reactor. As more 523 compounds were found in the influent and hence triggered competition reactions that hindered 524 the degradation kinetics, this difference was approximately 44-49% higher for the vertical plate 525 stirred tank reactor, corresponding to scenarios ECSWL-B, ECSWH-B and MSWL-B. 526 527 Figure 5: Environmental benchmark in the selected LCA midpoint categories for the standardised modular reactor 528 across scenarios operated in batch mode as defined in Section 2.1. GWP: global warming potential, TA: terrestrial 529 acidification, FE: freshwater eutrophication, ME: marine eutrophication, TET: terrestrial ecotoxicity, FRS: fossil 530 resource scarcity. 531 532 Figure 8: Single Score Indicators from LCA endpoint analysis for the standardised modular reactor and the vertical 533 plate stirred tank reactor configurations operated in batch mode as defined in Section 2.1. 534 18 535 Figure 6: Environmental contributions in the selected LCA midpoint categories for the standardised modular reactor 536 in scenarios ECSWL-B, ECSWH-B and MSWL-B as defined in Section 2.1. Data labels correspond to the contribution 537 of the hotspot per scenario and category. GWP: global warming potential, TA: terrestrial acidification, FE: 538 freshwater eutrophication, ME: marine eutrophication, TET: terrestrial ecotoxicity, FRS: fossil resource scarcity. 539 540 Figure 7: Environmental benchmark in the selected LCA midpoint categories for the standardised modular reactor 541 and the vertical plate stirred tank reactor across scenarios operated in batch mode as defined in Section 2.1. GWP: 542 global warming potential (kg CO2 eq), TA: terrestrial acidification (kg SO2 eq), FE: freshwater eutrophication (kg P 543 eq), ME: marine eutrophication (kg N eq), TET: terrestrial ecotoxicity (kg 1,2-DCB eq), FRS: fossil resource scarcity 544 (kg oil eq). 545 19 3.2.4 Oversizing effect 546 As for the standardised modular reactor configuration, the concept of oversizing effect is 547 introduced here to account for the mismatch between the number of anodes commercially 548 available and the number of anodes actually needed to achieve a desired CBZ removal rate. This 549 gap occurs because this reactor configuration is composed of fixed stacks of 10 anodes, which 550 in practice will lead the round-up in the number of anodes to the closest tenth multiple. 551 Consequently, the energy consumption and investment costs of electrochemical treatment may 552 also inevitably increase without being effectively exploited. The effect of oversizing was 553 analysed for the RL-B scenario by assuming that the anodes can operate with different effective 554 surface areas (Fig. 9). After calculating the required number of anodes for each effective surface 555 and rounding it to the nearest tenth multiple, the number of reactors required was obtained. 556 Depending on the differences between the theoretically required energy consumption for a given 557 effective anode area and the actual energy consumption due to the required number of reactors 558 (that is, the gap between the straight and the dotted lines), the oversizing may range from 0.95% 559 to 54.3% imbalance. The smallest gap corresponded to 65% effective area, and the largest to 560 10%. In the case of the 90% effective anode area that was considered for scale up, a gap of 17.7% 561 was observed. 562 In order to elucidate the effect of oversizing on the environmental profile of the different reactor 563 configurations, the 17.7% intrinsic excess found in the standardised modular reactor was applied 564 to the vertical plate stirred tank reactor. Under the EC-B scenario, a benchmark in terms of LCA 565 midpoint and endpoint categories for the three resulting reactor configurations was conducted 566 (Fig. 10 and Fig. 11). As shown in Fig. 10, the oversized vertical plate stirred tank reactor 567 presented the largest contribution to all LCA midpoint categories, with a difference of 1-3% over 568 the vertical plate stirred tank reactor and 20-36% over the standardised modular reactor. 569 Regarding the LCA endpoint categories (Fig. 11), the scores for Human Health and Ecosystems 570 categories were the most substantial, with the Resources category reporting scores up to 2 571 orders of magnitude lower. For all endpoint categories, between 63 and 72% of the individual 572 scores were attributed to energy consumption. According to this analysis, the oversized vertical 573 plate stirred tank reactor was again the most impactful configuration, followed closely by its 574 non-oversized version. Similar to the outcomes of the midpoint benchmark, the 17.7% oversizing 575 did not lead to a dramatic increase in endpoint scores, as there was a difference of 576 approximately 2% in the Human Health and Ecosystems categories and a difference of 25% for 577 Resources, although the latter category contributed less to the overall endpoint damage. In both 578 the midpoint and endpoint analyses, the standardised modular reactor was the configuration 579 with the lowest environmental impact, although inherently oversized, mainly due to the lower 580 energy consumption of the electrodes (Table A.4) relative to the vertical plate stirred tank 581 reactor (Table A.5). This reduction originated not only from the different anode areas involved 582 but also from the distances between electrodes (being considerably lower for the modular 583 reactor), which affected the calculation of the potential difference at the pilot scale (Vdiff ), as 584 discussed in Sections 3.1.1 and 3.1.2. 585 20 586 Figure 9: Correlations between energy consumption and the number of reactors with respect to the anode effective 587 area in the RL-B scenario. 588 589 Figure 10: Environmental benchmark in the selected LCA midpoint categories for the EC-B scenario operated under 590 different reactor configurations, including the oversizing effect. GWP: global warming potential, TA: terrestrial 591 acidification, FE: freshwater eutrophication, ME: marine eutrophication, TET: terrestrial ecotoxicity, FRS: fossil 592 resource scarcity. 593 21 594 Figure 11: Environmental scores in the LCA endpoint categories for the EC-B scenario operated under different 595 reactor configurations, including the oversizing effect. 596 4 Treatment selection and literature comparison 597 When evaluating the most sustainable mode of operation, the first finding of this work was that 598 continuous operation showed the highest chemical consumption, given that its nitrate and 599 sulfate requirements were above the regulatory limits, and thus, these ionic species needed to 600 be continuously added to the wastewater influent. On the other hand, the fed-batch operation 601 could minimise such chemical consumption with a 74-93% reduction, while the batch 602 performance ranged in between regardless of the wastewater treated. Regarding electrical 603 energy consumption, the fed-batch operation was the one with the highest energy demand, 604 especially when treating wastewater with multiple pollutants. This was due not only to the 605 energy required by the operation of the recirculation pump but also to its lower wastewater 606 volume capacity and negatively affected degradation efficiency by competition kinetics. Energy-607 wise, both continuous and batch operations treating complex wastewater matrices were 608 comparable even if the energy consumption was distributed differently across pump types. 609 However, when the influent wastewater contained a lower content of other ions and organics, 610 the batch operation stood out as the most environmentally friendly solution across all LCA 611 impact categories, resulting from its lower energy consumption and regardless of the chemical 612 additions. Therefore, from an environmental perspective, the decision-making process regarding 613 the mode of operation comes down to the energy-related impacts, where batch is the most 614 sustainable option, followed by continuous and fed-batch. If the addition of chemicals were to 615 be avoided by operating with wastewater influents with enough sulfate and nitrate 616 compositions, the batch operation would still be the leading condition, and the overall treatment 617 would also be improved from an economic point of view. 618 Regarding the effect of the influent wastewater composition, it was found that targeting single 619 vs multiple contaminant wastewater matrices significantly affected the environmental profile 620 of the treatment, with an increase between 186-264% across the LCA impact categories 621 considered. Here, again, these impacts derived from the negatively affected degradation 622 efficiencies, which resulted in longer reaction times and higher energy demands by the different 623 22 pumps involved. 624 In this study, two different reactor types were scaled up and compared in environmental terms: 625 a standardised modular reactor and a vertical plate stirred tank reactor. The latter presented 626 the advantage that it allowed for a fully customisable design, as it was not restricted to the 627 anode geometry, size or number of commercially available cell modules. Nonetheless, the 628 vertical plate stirred tank reactor scored higher for the majority of LCA midpoint and endpoint 629 categories, with differences up to 54% and 49%, respectively, resulting from the increased energy 630 requirements by the electrodes and the recirculation pump. Based on complementary analysis 631 regarding the oversizing effect, it was found that even if 17.7% was oversized, the modular 632 reactor was the most attractive configuration. 633 Consequently, the most environmentally friendly configuration for the electrochemical oxidation 634 of CBZ through BDD anodes corresponded to batch operation in a standardised modular reactor, 635 preferably when the influent wastewater matrix had a low content of scavengers, such as other 636 ions, organics or pollutants. The associated environmental impacts at the midpoint for these 637 conditions are detailed in Table B.1. Despite the lack of comparable references on LCA applied 638 to pilot-scale eAOPs for CBZ removal, a preliminary comparison of our most promising 639 configuration in single and multicomponent systems (i.e., the ECSWL-B, ECSWH-B and MSWL-640 B scenarios in the standardised modular reactor) with respect to other pilot-scale treatments in 641 terms of GWP is presented here. Given the diversity of functional units, the reported values have 642 been extrapolated to a common reference of kg CO2 eq per g CBZ removed (Fig. 12). Considering 643 that the average CBZ intake for adults is 600 mg/day [60] and that approximately 72% is 644 absorbed by the human body [8], the CO2 emissions associated with the removal of the daily 645 CBZ discharge per patient have been correlated to the equivalent distance covered by an 646 average passenger car for the same environmental impact (considering that an average of 107.5 647 g CO2/km was emitted in 2020 for new passenger cars registered in Europe [61]). The underlying 648 calculations can be found in Table B.2, although it should be noted that the comparison between 649 the studies should not be taken unquestionably, as there are considerable differences in their 650 LCA scopes. 651 23 652 Figure 12: Reported GWP impacts (in kg CO2 eq per g CBZ removed) for several wastewater treatments. GWP results 653 are also linked to the equivalent distance (in km) travelled by an average passenger car for the same emissions. 654 Treatments including the impacts associated with the infrastructure are denoted with *. GAC: granular activated 655 carbon, NF: nanofiltration, SPF: solar photo-Fenton. 656 Pesqueira et al. (2021) conducted an LCA on pilot-scale solar-based treatments, including solar 657 photolysis and TiO2 photocatalysis (with and without H2O2 addition) and near-neutral photo-658 Fenton [62]. Their LCA was based on the chemical and energy consumption in the photoreactor 659 and, at a later stage, the impact for its construction was also considered (indicated with * in 660 Fig. 12). Solar photolysis exhibited the lowest associated GWP (i.e., 5 kg CO2 eq per g CBZ 661 removed excluding infrastructure), although it was argued that the applicability of the process 662 was hindered by the lower mineralisation efficiencies attained. On the other hand, solar photo-663 Fenton presented the highest impact (i.e., 57 kg CO2 eq per g CBZ removed excluding 664 infrastructure) due to the need for acidification, neutralisation and iron removal steps. As a 665 result, solar TiO2-P25 treatment without H2O2 was presented as the most suitable alternative, 666 considering that the catalyst should be reused at least 5 times [62]. In absolute terms, this 667 treatment resulted in a GWP of 22 kg CO2 eq per g CBZ removed, which increased by 15.5% when 668 considering infrastructure impacts. In our study, the scenario that would outperform TiO2-P25 669 photocatalysis would be the ECSWL-B scenario, with 7.6 kg CO2 eq per g CBZ removed. By 670 increasing the complexity of the wastewater matrix, the ECSWH-B and MSWL-B scenarios 671 showed impacts of up to 18.9 and 26.5 kg CO2 eq per g CBZ removed, respectively. However, the 672 analysis by Pesqueira et al. (2021) did not include the impacts associated with the operation of 673 other equipment such as pumps, which are the main sources of electricity consumption and thus 674 24 GWP. Consequently, it can be argued that our electrochemical setup entails a significantly lower 675 environmental impact than solar photo-Fenton for the three scenarios selected, while a more 676 comprehensive analysis of the electrical energy consumption by solar TiO2-P25 photocatalysis 677 is needed. Nonetheless, the electrochemical treatment has the added value of not requiring a 678 catalyst, and therefore, avoiding the need to optimise the reuse, regeneration, operating costs 679 and environmental impacts of the catalyst material. 680 Gallego-Schmid et al. (2019) evaluated several pilot-scale solar photo-Fenton (SPF) processes 681 in combination with nanofiltration (NF). Their results showed that the NF unit helped to reduce 682 the environmental impact of acidic and neutral SPF by 38-43% by enhancing the treatment 683 efficiency. In terms of SPF performance, neutral SPF was hampered by the impact associated 684 with the use of an iron complexing agent, making it less environmentally friendly than 685 conventional acid treatment [63]. The higher impacts achieved by the neutral SPF (i.e., 167.3 kg 686 CO2 eq per g CBZ removed) compared to those of Pesqueira et al. (2021) could be attributed to 687 the higher number of target pollutants (and thus process efficiency affected), the higher number 688 of consumables (including reagents and iron complexing agents) and the inclusion of transport 689 and dismantling in the scope. As confirmed through this study, our electrochemical treatment 690 stands out as more sustainable, as GWP impacts are between 2.2 and 22 times lower. Zepon 691 Tarpani and Azapagic (2018) also investigated the environmental impact of SPF, ozonation and 692 other conventional wastewater treatments, such as granular activated carbon (GAC) and 693 nanofiltration (NF). In terms of CBZ removal, their four treatments showed considerably higher 694 GWP impacts than any previous work (i.e., between 189.1 and 312.9 kg CO2 eq per g CBZ 695 removed), presumably also due to the larger scope of the LCA [64]. 696 In relation to previous literature on LCA applied specifically to electrochemical oxidation, the 697 studies from Chatzisymeon et al. (2013) and Li et al. (2022) are available, although they were 698 applied to olive mill wastewater treatment and PFAS removal from groundwater, respectively 699 [65, 66]. In both studies, it was concluded that the environmental impact of the electrochemical 700 treatment was primarily determined by the electrical energy consumption, which was also 701 observed in our study. Their absolute GWP values reached 160 and 0.205 CO2 eq per cubic metre 702 of treated wastewater, respectively. Under the standardised modular reactor configuration, the 703 ECSWL-B, ECSWH-B and MSWL-B scenarios presented GWP impacts of 7.6·10−3, 1.9·10−2 and 704 2.7·10−2 kg CO2 eq per 1 mg CBZ removed per cubic metre of treated wastewater during one day 705 of operation. Comparison between the three assessments is certainly hampered by the different 706 target pollutants, wastewater origins, functional units and LCA scopes considered, as reflected 707 in the different orders of magnitude of the results obtained. Therefore, future LCA studies on 708 electrochemical oxidation applied to the removal of pharmaceuticals are necessary to 709 consolidate the environmental profile of eAOPs. 710 5 Conclusions 711 This study has demonstrated the importance of scaling up laboratory results for a more 712 comprehensive evaluation of an electrochemical treatment, since most of the environmental 713 impacts in a modelled scaled-up pilot operation were found to be related to the electrical energy 714 25 consumed by complementary pumps and not the electrochemical reactor itself. Consequently, 715 optimising the energy requirements of all pieces of equipment is crucial to aim towards 716 sustainable and carbon neutral wastewater treatment. In this way, the efforts made to achieve 717 SDG No. 6 of Clean Water and Sanitation are not jeopardised by increasing the levels of CO2 and 718 other greenhouse gases in the atmosphere. 719 From an environmental point of view, this work has shown that the most promising eAOP for the 720 removal of CBZ is carried out in a standardised modular reactor operated in batch mode, 721 preferably when the complexity of the influent wastewater is as low as possible. Under these 722 conditions, our eAOP has been shown to outperform previously reported AOPs, such as ozonation 723 and solar Photo-Fenton, in terms of GWP (i.e., ranging from 10% to 96% less kg CO2 eq per g CBZ 724 removed). However, further LCA studies on similar eAOPs are required to confirm their suitability 725 for future applications, especially if the scope of the LCA can be extended to a full plant 726 operation. 727 Finally, it should be noted that this techno-environmental analysis is based on steady-state 728 modelling and would therefore benefit from validation studies. To corroborate the robustness 729 and effectiveness of the eAOP for real wastewater treatment, experiments should be replicated 730 on a larger scale, and dynamic modelling aspects, such as possible alterations of process 731 variables (e.g., flow rate, current density and wastewater composition) and deterioration of 732 equipment over time (e.g., fouling of electrodes) should be evaluated. In addition, a toxicity 733 assessment of the treated effluent is recommended to ensure safe and viable operation. 734 Acknowledgements 735 This research received funding from the European Union’s EU Framework Programme for 736 Research and Innovation H2020 under Grant Agreement No 861369 (MSCA-ETN InnovEOX), from 737 the KU Leuven Industrial Research Council under grant number C24E/19/040 (SO4ELECTRIC), 738 and from the HP-Nanobio project (PID2019-111163RB-I00), granted by Spanish Ministry of 739 Science and Innovation. S. Estévez thanks the Spanish Ministry of Science, Innovation and 740 Universities for financial support (Grant reference PRE2020-092074). 741 Competing interests 742 The authors declare no competing interests. 743 Supplementary Material 744 The Supplementary Material includes additional results regarding the scale-up modelling and 745 environmental analyses as well as a literature review. 746 32 229–234. doi:10.1016/j.jclepro.2013.05.013. 965 [66] G. Li, J. Dunlap, Y. Wang, Q. Huang, K. Li, Environmental Life Cycle Assessment (LCA) of 966 Treating PFASs with Ion Exchange and Electrochemical Oxidation Technology, ACS ES&T Water 967 2 (9) (2022) 1555–1564. doi:10.1021/acsestwater.2c00196. 968