Detrimental vs. beneficial influence of ions during solar (SODIS) and photo-Fenton disinfection of E. coli in water_ (Bi)carbonate, chloride, nitrate and nitrite effects

In this work, we studied the effect of inorganic ions occurring in natural waters on E. coli inactivation by solar and photo-Fenton processes, two crucial methods for drinking water treatment in sunny or developing countries. − − − − − HCO , Cl , SO , NO , NO 3 4 2 3 2 and + NH4 were assessed at relevant concentrations for their inhibiting or facilitating role. The inactivation enhancement during solar disinfection (SODIS) was mainly attributed to the generation of HO• radicals produced during by excitation of − − NO , NO 3 2 , while the HO• of photo-Fenton may be transformed into other radical species in presence of ions. Natural organic matter (NOM) was found to enhance both processes but also to hinder most of the enhancing ions, except for − NO2 ; modeling with the APEX software unveiled the inter-relations in the presence of NOM, and the possible inactivation activity by NO2 • . The photoFenton inactivation was more significantly enhanced by ions than SODIS (besides the case of − − NO , NO 3 2 ), but both processes were found robust enough.


Introduction
Waterborne pathogens causing diseases constitute one of the acute health risks associated with urban wastewater discharge and reuse. They have been identified as a major infection risk in streams, rivers and estuaries. The use of solar radiation to disinfect water, more known as the solar disinfection process (SODIS), has been successfully evaluated as a way to eliminate pathogens from waters destined for consumption [1,2]. Unfortunately, SODIS is prone to temperature dependence and has shown possible bacterial regrowth issues [3][4][5]. The attempts to enhance the SODIS efficiency focused on low-cost technological or physicochemical modifications [4,[6][7][8][9], aimed at decreasing the exposure time needed to achieve "permanent" microorganism elimination. This is the rationale for trying to achieve acceleration of the SODIS kinetic performance with the addition of H O 2 2 to raw water, or with the photo-Fenton process [6,[10][11][12][13]. H O 2 2 directly attacks the cellular membrane, increasing its permeability and affecting cell survival. It can also diffuse into the cell and initiate an intra-cellular process of cell death [14,15]. The photo-Fenton process involves the reaction of H O 2 2 with photogenerated + Fe 2 ions, leading to the formation of Reactive Oxygen Species (ROS), such as HO • radicals. The latter are powerful oxidizing species that can achieve inactivation of bacteria and viruses [12,16,17]. Moreover, the photo-Fenton reagents can also trigger intracellular events due to the transport of iron and H O 2 2 inside cells [18].
Natural water sources have an important content of Natural Organic Matter (NOM) and inorganic ions, such as (1) (2) In the above reactions, X is a halogen and the relevant species are halide ions and Reactive Halogen Species (RHS). The generation of RHS could still lead to microorganism disinfection [33], as they retain oxidizing power [29,34]. Although less reactive and more selective than HO • , their reaction mechanism involves pathways such as one-electron oxidation or addition to unsaturated CeC bonds [29,35]. The HO • radicals are scavenged with second-order reaction rate constants in the order of 10 4 M -1 s -1 for − H PO 2 4 and 10 6 M -1 s -1 for − HCO 3 . However, the rate constants can be as high as 10 8 M -1 s -1 for − CO 3 2 and Fe 2+ [36], 10 9 M -1 s -1 for − Cl [37] (but the actual outcome for chloride is pH-dependent as mentioned above), and even 10 10 M -1 s -1 for − NO 2 and − Br [38]. Dissolved ions have also the potential to enhance either SODIS or photo-Fenton. Examples include the generation of HO • by the illumination of − NO 3 / − NO 2 [39,40], or the complexation reactions of Fe 2+ or Fe 3+ with inorganic ligands. The latter can affect the distribution and reactivity of the iron species [22,32,41,42]. Hence, depending on their concentration, speciation or distribution, inorganic ions can have contrasting effects on both SODIS and the photo-Fenton processes.
Moreover, the organic matter in solution (NOM) has a double activity as an antagonist or a facilitator of the photo-chemical processes [43][44][45]. Its presence under sunlight enables a large variety of photochemical reactions that proceed by energy transfer and result in singlet oxygen and radical species generation. Such reactions also yield additional ROS such as superoxide and H O 2 2 [45][46][47]. NOM has the potential to increase the efficiency of both SODIS and photo-Fenton processes, by providing effective ligands that trigger Ligand-to-Metal Charge Transfer (LMCT) processes and produce ligand radicals, ROS and further Fe 2+ [41,45,[48][49][50]. However, as almost every organic compound, NOM has oxidizable moieties that have the potential to significantly scavenge the photo-produced reactive species [51,52]. NOM is also able to absorb sunlight, but the path lengths of radiation in water during SODIS are never too high and the absorption effect is less important compared to other contexts, such as the water column of natural aquatic environments.
The above phenomena can explain the intrinsic inconsistencies of the literature about the roles of ions and organic matter in water disinfection, as well as the lack of a systematic investigation. For these reasons, the main goal of this work is to unveil the effect of a series of inorganic ions, namely  4 , in the absence and in the presence of NOM, on E. coli inactivation by the SODIS and photo-Fenton processes. To attain this goal, for each ion under scrutiny, the related chemical events that could result in bacterial inactivation were reviewed, thereby contextualizing our disinfection experiments with the current understanding of natural-water photochemistry. The bacterial cultivability as well as the effect of ions concentration during solar exposure and photo-Fenton processes was evaluated, i.e., in presence or absence of the ions. A systematic kinetic assessment will describe the critical parameters in defining bacterial inactivation, namely lag phase and inactivation rate, while the potential role of secondary radicals will be elucidated. In some cases, the changes in lag phase and inactivation kinetics could be modeled, to estimate the effectiveness of photoinduced disinfection as a function of the ions' concentration in sunlit natural waters.

Chemicals and reagents
The effect of + Na as counter ion is negligible because it is harmless for E. coli and cannot absorb sunlight. Therefore, sodium-based salts were used as sources of the ions under scrutiny. The used salts were NaHCO 3 , NaNO 3 , NaNO 2 , NaCl, Na 2 SO 4 and (NH 4 ) 2 SO 4 (the latter as source of ammonium, vide infra for the rationale of the choice), all supplied by Sigma-Aldrich. Aqueous solutions of salts, in appropriate concentrations, were prepared in Milli-Q water (MQ). FeSO 4 ·7H 2 O and H O 2 2 30 % w/v (Sigma-Aldrich) was used to prepare the stock solutions of the photo-Fenton reagents (1000 ppm each).

Photochemical experiments
Solar irradiation with intensity of 620 W m −2 was simulated by an Atlas SUNTEST Solar simulator. Irradiation experiments at 350 rpm of agitation by magnetic bars placed on stirrer plates and at room temperature were performed, testing the effects on bacteria of − HCO 3 at concentrations from 5 to 100 mg L -1 , − NO 3 from 1 to 50 mg L -1 , − NO 2 from 0.01 to 5 mg L -1 , − Cl from 1 to 500 mg L -1 , − SO 4 2 from 10 to 500 mg L -1 and + NH 4 from 0.1 to 10 mg L -1 . The salts and their concentrations have been selected in accordance to the actual ions' presence in natural waters, according to Table 1.
In order to define the bacterial survival in the presence of the maximum concentrations of these ions, control experiments after 240 min in dark conditions were performed. In photo-Fenton experiments, the concentration of + Fe 2 and H O 2 2 solutions were 1 mg L −1 and 10 mg L −1 respectively.
A depiction of the experimental set-up is given in the supplementary material (Scheme S1). The test took place in Pyrex glass reactors with Milli-Q water at near neutral starting pH. The reactors contained 100 mL of E. coli dispersion with concentration of 10 6 colony forming units per mL (CFU mL −1 ). Before every experiment, reactors were sterilized by autoclaving and after each experiment, reactors were washed with acid to ensure iron removal, with ethanol to remove any other contaminant and finally with deionized water in abundant amounts.

Bacterial strain and growth media
The bacterial strain used in this study was E. coli K12, a non-pathogenic wild-type strain, which can be handled with little genetic Table 1 Mean ionic composition of water sources frequently used for Solar disinfection (SODIS) [53][54][55][56][57]. ND = not determined.

Ions
River Water (mg L −1 ) Lake Water (mg L −1 ) Harvested rainwater (mg L −1 ) In order to prepare the bacterial stock solution, a colony of bacteria was extracted from the master plate and inoculated into 5 mL of Luria-Bertani (LB) Broth. Specifically, LB consisted of 10 g L −1 tryptone, 10 g L −1 NaCl and 5 g L −1 yeast extract in Milli-Q water. The saline solution was a sterile NaCl/KCl solution (8 g L −1 NaCl and 0.8 g L −1 KCl at pH 7-7.5). After a strong mixing by a vortex machine for 1 or 2 min, it was incubated inside a 37°C temperature-controlled room for 8 h and constantly agitated by circular movement at 750 rpm. After 8 h, 2.5 mL of sample were diluted in 250 mL of LB Broth and incubated for 15 h in the same room to ensure that the stationary physiological phase was reached. A 25 mL aliquot of this bacterial sample was separated during the stationary growth phase by centrifugation and was washed 3 times with saline solution. Washing took place in a 4°C centrifuge (Hermle Z 323 K, Renggli Laboratory Systems), at 5000 rpm for 15 min the first time and 5 min the remaining two, with 10 mL of saline solution. After the final wash, 25 mL of clean saline solution was added to the bacterial pellet. This procedure resulted in a bacterial dispersion of approximately 10 9 CFU mL −1 .

Sampling and bacterial enumeration
Samples of 1 mL were taken from the body of the reactor under stirring and placed in sterile plastic Eppendorf vials, to ensure their sterile preservation. In order to obtain information about the disinfection kinetics, sampling was performed at time intervals of 0, 30, 60, 90, 120, 180 and 240 min for SODIS, and at time intervals of 0, 20, 40, 60, 90 and 120 min for photo-Fenton. For reproducibility, each experiment was carried out at least in duplicate (biological/chemical replicates) in double series (statistical replicates) and using 2 or 3 serial dilutions (technical replicates), to achieve measurable bacterial count on the plates; the optimal colony counts in this method are among 15-150. Total inactivation was considered achieved when no bacteria colony was observed any longer in the plates after treatment. The spread plate technique was performed on PCA, contained in plastic sterile Petri dishes, by injecting drop-by-drop 100 μL of samples. The detection limit was 1 CFU mL -1 for undiluted samples and 10 CFU mL -1 for diluted ones [58]. The incubation period was 18−24 hours at 37°C.

Data treatment and APEX modeling
In order to model the bacterial response under the solar light and photo-Fenton stress, a 60-min and a 30-min lag period was considered for SODIS and photo-Fenton processes, respectively. After this period, log-linear kinetics were fitted by the GInaFiT freeware add-on for Microsoft Excel [59]. For the kinetic modeling, a log-linear equation with delay was used; its calculation was possible for all cases and it provided the exact time of delay and the subsequent log-linear kinetics of inactivation. The "Shoulder log-linear model" was formulated as shown in Eq. (1) [59].
log [1 (exp(k* t ) 1)*exp( k*t)] 10 10 0 s 10 s (6) where: N is the bacterial population at any given time (CFU mL -1 ). N 0 is the initial bacterial population (CFU mL -1 ). t is the investigated time (s). t s is the length of the shoulder period or threshold time to observe inactivation (s).
k is the rate constant of the inactivation (s −1 ). In some cases, it was possible to apply photochemical modeling to get better insight into the functioning of some SODIS systems leading to bacterial inactivation. Eq. (6) or its non-logarithmic equivalent was the basis for modeling. At the moment, photochemical modeling can only take into account exogenous inactivation of E. coli by reactive transient species (HO • , O 1 2 , NOM 3 * ), thereby excluding the endogenous process driven by light only [47,60]. In particular, the value of (exogenous) t s was determined according to the following phenomenological equation, where t s is expressed in min and [HO • ] in mol L -1 [47]: The value of the steady-state [HO • ] was assessed by means of the APEX software (Aquatic Photochemistry of Environmentally-occurring Xenobiotics) [61] which was also used to determine k as the rate con-  [61]. APEX calculates the absorption of radiation by the photosensitizers (NOM, nitrate and nitrite) on the basis of competition for sunlight irradiance, in a Lambert-Beer approach [62]. In the model, NOM is a photochemical source of HO • , O 1 2 , and NOM 3 * as well as a HO • sink, while nitrate and nitrite are both HO • sources. The model results apply to well-mixed waters and provide average values over the whole water column, with contributions from the well-illuminated surface layer and from darker water in the lower depths [63].
It should be underlined that APEX modeling of bacterial inactivation still has several gaps, which presently limit the comparison with experimental results. The main limitations are the following: (i) it is presently not possible to model photo-Fenton, thus APEX is currently limited to pure SODIS; (ii) the model takes into account exogenous inactivation only, thereby neglecting the very important endogenous process. By so doing, model results are bound to underestimate actual photoinactivation; (iii) the model uses summertime sunlight as the radiation source in place of the experimental lamp, which is very useful to simulate field SODIS but further limits comparison with laboratory results; (iv) it is presently not possible to assess the scavenging of HO • by the bacteria. This issue prevents a proper treatment of the systems that contain only bacteria + − NO 3 or bacteria + − NO 2 , in the absence of additional HO • scavengers such as NOM. Despite these limitations, in some cases the model can provide a semi-quantitative insight into the photoinduced processes, which may give interesting indications as far as the inactivation pathways are concerned.

Chemical and analytical methods
According to the standard methods for water analysis [64], Cl concentrations of 0−3000 mg L −1 ). The pH evolution during treatments was recorded using a pH-meter manufactured by Mettler Toledo. Fig. 1 summarizes the bicarbonate-related disinfection tests and the effect of the bicarbonate ion in both SODIS and the photo-Fenton process. Fig. 1(a) shows the pseudo-first order kinetic constant values calculated for both processes, at all the tested − HCO 3 concentrations, while Fig. 1(b) shows the most significant disinfection graphs.

Ions influence on SODIS and photo-Fenton processes
The results show that the effect of − HCO 3 on SODIS and the photo-Fenton process is not totally straightforward. First, there is very limited effect on the first-order disinfection rate constants, except for − HCO 3 at 10 mg L -1 in the case of photo-Fenton. There is some more effect on the lag times, although the variation is not linear with − HCO 3 concentration. In terms of overall disinfection, the inactivation of bacteria was faster with 10 mg L -1 HCO 3 compared to no bicarbonate, and slower with 100 mg L -1 HCO 3 -( Fig. 1(b)). Carbonate and bicarbonate ions are the main inorganic carbon forms in water; most of the − HCO 3 and − CO 3 2 ions originate from the dissolution of carbonate minerals, the decomposition of organic matter, the respiration of aquatic animals and the exchanges in the carbon cycle [65]. Due to its high solubility, − HCO 3 is widely distributed in natural waters (see Table 1 The disinfection lag time of E. coli has been shown to mainly depend on irradiation and the attack by HO • [60,67]. Moreover, bacteria could be susceptible to the combined effects of irradiation, basic pH and the presence of oxidizing species. Therefore, pH changes in case of the addition of bicarbonate at high concentration, combined with the consumption of photogenerated HO • , could have contrasting effects on the disinfection process and produce non-linear phenomena. In addition, at elevated pH one has enhanced Fe 3+ precipitation that has the potential to hamper the photo-Fenton process. While − CO 3 • is a less selective oxidant than HO • , it has much longer lifetime in aqueous solution and can, therefore, diffuse over a much larger range. The germicidal action of HO • is limited by its inability to react with cell components other than the membrane, while the longer − CO 3 • lifetime could enable additional disinfection pathways. Indeed, shows the most significant disinfection graphs. Interestingly, the addition of chloride enhanced disinfection in both cases (SODIS and photo-Fenton), although most likely for different reasons. The pH of both reaction systems was near-neutral, and it showed no substantial shift during either SODIS or photo-Fenton treatments, compared to the experiments in the absence of − Cl . A decrease in − Cl concentration during both processes was observed, as shown in Fig. 2(c).
As far as the disinfection enhancement observed during SODIS is concerned, there are a couple of instances in the literature that report a similar phenomenon. The effect is most likely correlated with the membrane-− Cl interactions, which are hypothesized to increase membrane permeability [37,71]. In contrast, in order to explain the increase in efficiency during the photo-Fenton process, one should consider that the interaction between − Cl and + Lastly, termination reactions can lead to the formation of free chlorine that actively participates in the disinfection process (Eqs. [24][25][26]. Moreover, it can explain the decrease of − Cl observed during the photo-Fenton process. Other termination reactions include the Cl • recombination or the HOCl reaction with H O 2 2 [29] (Eqs. [27][28].     [76].
It is interesting to observe that the photochemical production of HO • by nitrate and nitrite has the potential to both shorten the lag time and accelerate the disinfection kinetics in the post-lag, exponential phase (Fig. 4(c)). This finding is in agreement with literature reports, according to which HO • is one of the transient species involved in post-lag bacterial inactivation and, at the same time, a major actor in inducing cell-membrane damage that exposes the cell to the action of oxidants. Therefore, elevated HO • causes the lag time to become shorter [47,60]. However, it should be considered that nitrite is a HO • sink as well as a source [77], and this fact may have interesting implications for the inactivation process (vide infra) (Eqs. [30][31].
Generally, − NO 3 and − NO 2 are naturally occurring ions that are part of the nitrogen cycle. These ions can reach both surface water and groundwater because of agricultural activity: in fact, fertilizers contain inorganic nitrogen and wastes contain organic nitrogen, which is first decomposed to give ammonia and then oxidized to give − NO 2 and, finally − NO 3 . It is not surprising that − NO 3 and − NO 2 play a significant role in photochemical processes, although their significance in the photo-Fenton process has been questioned [73]. Our findings suggest that − NO 3 and − NO 2 at environmental concentrations have a real potential to enhance bacterial disinfection, even under photo-Fenton conditions.

NH 4 effects
Ammonia is a common, naturally occurring substance. The main local problem of NH 3 released into the air is the unpleasant odor, which is detectable even at low concentrations. The harm caused by + NH 4 in water bodies is more serious, because it is very toxic to aquatic organisms. Fig. 5 summarizes the experiments carried out upon addition of ammonium sulfate, to simulate an excess of + NH 4 in water. The sulfate counter-ion was chosen because of its limited effects on bacterial disinfection (see Section 3.3). Fig. 5(a) reports the pseudo-first order kinetic constant values calculated for both SODIS and the photo-Fenton processes at all tested + NH 4 concentrations, while Fig. 5(b) shows the most significant disinfection graphs. The addition of + NH 4 had practically no effect in the case of SODIS, while a significant enhancement could be seen with photo-Fenton. The pH of both systems, which is affected by the initial + NH 4 concentration used, did not show any particular change during either process (Fig. 5c). Furthermore, as shown in However, the photo-Fenton process revealed a decrease in + NH 4 concentration (Fig. 5c). The most likely explanation for this finding is that ammonia photo-oxidation by HO • generates − NO 2 and − NO 3 ions [78,79]. In a summarized way, the reaction pathway reads as follows (Eqs. 32-34): We note here that although the oxidation of ammonia may be mediated by O 2 , the relevant reaction rates would be very slow [32]. The formation of photochemically active nitrate and nitrite upon ammonia oxidation might explain the observed disinfection enhancement, which is likely due to the production of HO • upon photolysis of 2 counter-ion to bacterial disinfection at the used concentration values of (NH 4 ) 2 SO 4 is negligible (see Fig. 3).

Effect of ions on SODIS and photo-Fenton in the presence of Natural Organic Matter (NOM)
The presence of Natural Organic Matter (NOM) in water is ubiquitous, and it is the product of both autochthonous and allochthonous processes [80,81]. NOM can act as a filter for sunlight and, because it absorbs throughout the UV-vis spectrum, it can inhibit the inactivation of E. coli. However, UV light absorption by NOM produces the corresponding triplet states ( 3 NOM*), the deactivation of which occurs in several ways that include the reaction with oxygen to form singlet oxygen (see Eqs. 35 and 36). The photoinduced formation of transient species as a function of NOM type, oxygen and NOM concentration was recently systematically investigated, and the main pathways are as follows [45,47] Singlet oxygen can react with water contaminants or bacteria forming peroxidation products, thereby contributing to photochemical decontamination. In addition, if both iron and NOM occur in water at the same time, complex species like [Fe-NOM] are generated. Compared to NOM, these complexes show higher light absorption and quantum yields, enabling ligand-to-metal charge transfer as shown in Eq. 37. These reactions contribute to bacteria inactivation [82].
Figs. 6 and 7 present an overview of the experimental results obtained when concentrations of ions showing a significant (either positive or negative) effect on disinfection ("optimal" concentrations, as determined in the previous section) were added in the presence or absence of organic matter, for both SODIS (Fig. 6) and photo-Fenton processes (Fig. 7). In both series of experiments, the Suwanee River NOM (SRNOM, 2 mg L -1 ) that was used as model is expected to actively participate in E. coli inactivation. During SODIS a fraction of light can be filtered by SRNOM, inducing its excitation and transient species generation. If the SRNOM amount is sufficiently low and water is not deep (i.e., the optical path is short, as in the present case), the overall system is optically thin and there is limited competition for irradiance between SRNOM and bacteria. Indeed, the fact that bacterial inactivation showed an increase in kinetics in the presence of the organic material suggests that the photosensitization effect of SRNOM was more important than its light-screening role. Interestingly, whichever the added ions, no process was significantly faster than plain solar/NOM. Most of the ions did not induce further effects, including • that may be less reactive compared to 3 NOM*. Actually, − CO 3 • is effectively scavenged by ground-state NOM, differently from 3 NOM* that mainly reacts with dissolved oxygen [50]; moreover, the interaction between 3 NOM* and − HCO 3 may partially proceed via physical quenching, without generation of The photo-Fenton process was affected by the presence of organic matter in a similar way as SODIS. Firstly, Fe-NOM complexes could be formed in NOM-added photo-Fenton systems; their photolysis enhances the Fe 3+ /Fe 2+ recycling and, consequently, the disinfection kinetics as well. NOM is able to scavenge all radicals studied before, namely As a provisional conclusion based on the above findings, we can report that, with minor exceptions, SODIS and photo-Fenton can proceed faster in the presence of organic matter and relatively high amounts of ions. In an effort to generalize our findings about SODIS and photo-Fenton disinfection, the influence of ions and organic matter will be now qualitatively and quantitatively analyzed, regarding the aspects of durability of SODIS and photo-Fenton in the presence/absence of ions, the kinetic modeling of bacterial inactivation (with the aid of APEX software) and the intracellular vs. extracellular pathways to inactivation. In the previous sections, the effects of each ion during SODIS and photo-Fenton were considered. Over the range of concentrations that are expected to be found in natural waters subjected to SODIS, most of the ions showed a variation in their profile of enhancement or antagonism towards the treatment process. If one considers the two disinfection options, namely SODIS and photo-Fenton, for waters with an unknown ionic composition, a valid question would be: which process is safer to be applied as a function of its vulnerability to ions that may be present in water? In order to answer this question, a common response variable was chosen for both processes and all ions, i.e., the time necessary to achieve 4logU reduction (T 99.99% ) of the bacteria. The results are summarized in Fig. 8. Fig. 8 presents the change of T 99.99% over the original value achieved by SODIS or photo-Fenton without ions (i.e., 204 and 88 min, respectively: the original 4-logU times can be found in the supplementary material, Table S1). This normalized change was calculated separately for each process and for each anion (Eq. 38): Where, for each process, T 99.99% was the time necessary for 4 logU reduction and ′ T 99.99% the corresponding time resulting from the addition of ions.
It can be easily deduced that as an overall trend, all ions improved both processes (i.e., decreased their normalized T99.99 %), except for − HCO 3 during photo-Fenton. As far as the two processes are compared, some ions seem to affect SODIS to a higher extent, namely − Cl ,  Table S2. Four of the tested ions acted beneficially for SODIS and photo-Fenton, and two hindered the process; only − SO 4 2 and − Cl presented a shift in their influence. Therefore, we can suggest that both SODIS and the photo-Fenton process can work potentially well in natural water, since most of the ions have positive, or at least not negative, effect within their typical concentration ranges. Moreover, the ions show predictable behavior as a function of their concentration in water. Among the types of water that can be used for drinking purposes, groundwater usually has the highest number and concentration of ionic species, as well as higher pH (alkaline). However, as Fig. 8 suggests, when − Cl , − SO 4 2 and + NH 4 are already encountered at high concentrations, they can have negative (antagonistic) effects during SODIS or photo-Fenton if their concentration is further increased (note that in most cases, even at high ionic concentrations, the treatment was still faster compared to the case of ultrapure water; however, ultra-pure water does not occur in the natural environment). For instance, in the case of SODIS, Similarly, in the photo-Fenton process, one has inhibition in the presence of 10 mg L -1 − HCO 3 or higher, and above 100 mg L -1 − Cl . However, 10 mg L -1 − HCO 3 is usually near the lower limit, while 100 mg L -1 − Cl is usually higher than the actual values (except for some groundwaters). Therefore, − HCO 3 can be considered as an antagonistic ion and − Cl as a synergistic one during the photo-Fenton process. This means that, in the typical concentration ranges of the two ions, photo-Fenton is accelerated by increasing − Cl and slowed down by increasing  From the previous parts it can be concluded that in the presence of ions in water, bacterial inactivation is highly dependent on their concentration and, most importantly, the process would be mainly governed by events that take place in the solution bulk (as opposed to events affecting the bacterial cells or membranes). As such, an attempt to model the bacterial inactivation by solar-mediated processes could be performed with the APEX software. The SODIS systems containing NOM, NOM + − NO 2 and NOM + − NO 3 are amenable to photochemical modeling, as far as endogenous inactivation alone is concerned. The model predicts considerably slower kinetics compared to laboratory experiments (compare the model trends of Fig. 9a with the experimental trends for comparable systems, reported in Fig. 6), for several reasons that are listed in Section 2.5. However, the relative kinetics (i.e., what is faster and what is slower) should be preserved despite these differences [60]. In this context, it is interesting to observe that the model results in Fig. 9a predict the NOM + − NO 3 system to produce faster inactivation compared to NOM + − NO 2 , in clear disagreement with the experimental data. The reason is that the steady-state [HO • ] would be higher in the presence of 30 mg L -1 nitrate compared to 5 mg L -1 nitrite, as shown in Fig. 9b. Indeed, although − NO 2 is more photoactive than − NO 3 and undergoes photolysis to a higher extent (the HO • formation rate is predicted to be higher with • curve of Fig. 9b, while no plateau is observed in the case of nitrate. However, the reaction between HO • and − NO 2 yields a further transient species (NO 2 • ) that might also be involved in the bacterial inactivation process [77]: it is predicted to be five orders of magnitude more concentrated. Therefore, there is potential for NO 2 • to contribute to E. coli inactivation, which could explain why the nitrite-containing system was more effective than the nitrate one, despite the lower predicted [HO ] • .

Extracellular vs. Intracellular mechanisms of bacterial inactivation during SODIS and photo-Fenton processes
In many of the processes considered so far, bacteria are the terminal acceptor (target) of a transient species that leads to their inactivation. The addition of ions in the bulk suggests that the participation of the transient species to photo-chemical events happens in the bulk as well. With a few possible exceptions, the short lifetimes of the transients suggest that the occurring damage is located at the cell wall, whose eventual rupture leads to bacterial death. However, there is growing evidence that bacterial inactivation by the photo-Fenton or the persulfate process can also affect the intracellular domain of microorganisms [15,74]. Here we present an overview of the pathways that are induced when certain ions are present in the extracellular environment of bacteria, and how their presence might affect bacterial inactivation mechanisms.
Among the mentioned ions,  transferred by the bicarbonate transporters family or the Sbt (Sbt-sodium-dependent bicarbonate transporter) homologues [90] (see a summary of the transporters in supplementary Fig. S8). If we account for the photo-Fenton process taking place inside the cell, (and involving "natural" intracellular, non-bulk added, Fe and H 2 O 2 ), which also leads to the generation of HO • , then the aforementioned ions at high concentration may cause a new intracellular oxidative balance: As such, a summary of the intracellular and extracellular mechanisms that lead to bacterial inactivation is provided in Fig. 10.
The depicted actions are separated horizontally as SODIS (10a,b) and photo-Fenton (10c,d) while their effect, positive or negative is denoted vertically (positive: 10a,c and negative: 10b,d). The overview of the involved actions is as follows, stating from SODIS (numbers below correspond to those in Fig. 10 that naturally occur in the intra-cellular compartments, even without external addition of Fenton reagents [15]).
For the photo-Fenton process, the following actions can be highlighted. Action 4: The presence of + Fe 2 and H O 2 2 in the solution ensures the transport of both species into the cell and, consequently, the enhancement of intracellular photo-Fenton. Action 5: The above process (4) leads to the generation of HO • that attacks the cell, while light regenerates + Fe 3 to + Fe 2 . Action 6: The HO • generated as per the above discussion can react with − HCO 3 to produce − CO 3 • . Action 7: Furthermore, in the presence of − NO 3 / − NO 2 , additional HO • production ensues that enhances bacterial inactivation. On the other hand, the normal HO • production (Action 8) is disrupted by high amounts of − HCO 3 that acts as scavenger, thereby exerting a negative effect on both the HO • occurrence (Action 9 and Action 10) and the subsequent bacterial inactivation.

Conclusions
In this work, the effect of inorganic ions and natural organic matter occurring in aqueous matrices on the efficacy of E. coli removal by the SODIS and photo-Fenton processes was systematically studied. The investigated concentration values varied extensively to cover commonly encountered concentrations in surface waters, rainwater and groundwater, which constitute the most commonly used matrices in solarmediated disinfection.
From the obtained results, we can confer that not all ions have the same impact, and their effect is subject to the concentration values and the process applied (SODIS or photo-Fenton). More specifically, − HCO 3 was found to produce a small enhancement in inactivation kinetics in some conditions but, at environmentally relevant concentrations, it will always be an antagonistic factor. On the contrary, − NO 3 / − NO 2 and + NH 4 , can be expected to aid either SODIS or photo-Fenton disinfection. Disinfection by both processes will be faster in the presence of − Cl and − SO 4 2 than in their absence but, starting from the typical concentration values found in surface waters, a further increase of − Cl will enhance SODIS while disrupting the photo-Fenton process, and the opposite will occur with − SO 4 2 . Nevertheless, despite the various levels tested in this study, natural waters that contain a mixture of these ions and NOM present in various cases a negative impact. Furthermore, although NOM was beneficial for both SODIS and photo-Fenton, its presence can be an inhibiting factor for the secondary oxidants and radicals generated by the ionic species during the photo-assisted processes (except for − NO 2 ). However, a wider investigation will be necessary to locate the possible tipping point(s) in the interaction between NOM and the ions.
Finally, from the aforementioned results we can conclude that SODIS and photo-Fenton are quite robust processes: they are certainly suitable for the disinfection of natural waters, although their effectiveness could be hampered when treating some types of groundwater. Indeed, in most cases the added ions at typical concentration values in surface waters produced a decrease in the time required to inactivate 99.99 % of E. coli (exception: − HCO 3 ). In addition, even in highly antagonistic conditions, photo-Fenton was always faster than the SODIS process. This fact makes photo-Fenton an attractive solution that must be further evaluated in sunny or developing countries, to see whether it really is an effective measure at household or community level to achieve natural water disinfection.
CRediT authorship contribution statement

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.