Recommendations for future research

In this PhD thesis, insights have been gained in reduction and adsorption processes for TrOC degradation. Still, more research is necessary, especially in order to gain more mechanistic insights in the processes applied in this work. In the following, some recommendations for future research are formulated.

 As mentioned in paragraph 1.1.2, the exact factors determining the effect of solution pH on photolytic removal of TrOCs could not be elucidated. In order to be able to better predict photolytic removal of other, new, TrOCs, it is imperative that advances are made in this area. Establishing further mechanistic insights in this area would not only be useful for explaining and predicting TrOC degradation in UV253.7 nm reactors, but could possibly

also be useful for photolytic TrOC degradation in the environment, through visible solar radiation on surface water. This study was conducted on a mixture of TrOCs, therefore it is possible that the photolytic removal is indirectly affected by different TrOCs present in the solution. If reactive species are formed from some types of photolytically excited DOM, then TrOCs themselves could possibly create reactive entities as well, of which the reactivity and generation rate could be pH dependent. To confirm this, separate TrOCs should be investigated in parallel to TrOC mixtures.

 The exact mechanisms governing TrOC removal in environmentally relevant water matrices can be investigated further in detail, by examining TrOC degradation using scavengers for the different reactive entities produced from DOM and inorganic solutes. The role of singlet oxygen can be further investigated by using NaN3 as a 1O2 quencher, by

using rose bengal (a pigment) as a photosensitizer for generation of 1_O

2, and by conducting

experiments in D2O (in which 1O2 has a longer lifetime compared to H2O) [403]. The role

of 3_DOM* _{can be further investigated using sorbic acid, which acts as a} 3_DOM* _quencher

[404]. Experiments should be performed using different types of DOM, and in solutions containing single TrOCs and TrOC mixtures, in order to gain mechanistic insights in the process. This could ultimately lead towards a better prediction of the behaviour of different TrOCs in UV-based processes, such as photolysis, AOPs and ARPs.

 A more extensive degradation is found with advanced reduction compared to catalytic reduction, however it was not possible to determine whether photolysis or reducing radicals are responsible for this further degradation. Since in the UV/sulfite ARP these two processes always occur in parallel, the phenomenon responsible for degradation should be examined with radiation chemistry. Through radiolysis of water, and in the presence of a

[317]. This would enable to make a discrimination between UV25.7 nm induced photolytic

degradation, and 𝑒𝑎𝑞− induced degradation.

 The UV/sulfite ARP was found to be less efficient in terms of energy input compared to the UV/H2O2 AOP. In order to make ARPs more competitive with AOPs, methods to

achieve higher yields of reducing radicals should be investigated. This may be achieved by using different lamps having a different emitting wavelength, and/or by using different electron-donating anions than sulfite.

Because SO32- shows an absorption peak around 210 nm [166], ideally a lamp emitting light

close to 210 nm should be used. This may be achieved using a zinc lamp, emitting light at 213.9 nm [358], however, one problem associated with zinc lamps, is that their lifetime is very limited. After 2 months of continuous operation, the output has declined to 50% of the initial output [405]. Other possible UV sources are excimer lamps, e.g. a KrCl excimer lamp which emits near-monochromatic UV light at 222 nm, or a KrBr excimer lamp emitting at 207 nm [406]. A potential problem associated with applying UV irradiation at lower wavelength, is that the deeper into the UV-C region, the more chance on formation of chlorinated by-products. Chloride ions have a quantum yield of 0.4 at 193 nm [154], forming Cl2 gas, which hydrolyses into HOCl. Thus UV irradiation at shorter wavelengths

increases the chance on formation of chlorinated by-products such as trihalomethanes and haloacetic acids. Note that in reducing environment, this may be less of an issue, as hydrated electrons are efficient dehalogenating species, therefore chlorinated by-products, may be instantaneously dehalogenated. Thus, the formation of chlorinated by-products in reducing environment will depend on the equilibrium between formation (resulting from Cl- _{oxidation upon deep-UV irradiation) and destruction (dehalogenation with 𝑒}

𝑎𝑞− ).

While SO32- is one of the most interesting electron-donating agents, as it is a chemical which

is already approved for use in drinking water treatment [32] and given the fact that the end product is sulfate [171], other electron-donating anios could be used as well. Examples of other chemicals which produce 𝑒𝑎𝑞− are Fe2+ (at wavelengths below 240 nm [407], which

may be useful in ground water), and iodide (showing absorption peaks at wavelengths around 195 and 225 nm [157]). Note that iodide oxidises to I2 upon UV irradiation, which

may in turn react with iodide to form trijodide (I3-). The latter is an oxidant, and thus an

electron scavenger. Qu et al. (2010) found that the optimum KI concentration is around 0.3 mM, while higher KI doses are less interesting due to a higher formation of I3-.

 Because in the UV253.7 nm/sulfite ARP, germicidal irradiation is applied, it can be expected

fluences applied for TrOC removal compared to the fluence necessary for disinfection, this can indeed be expected. If ARP were to be used in drinking water treatment, the expected disinfection may as well decrease the required NaOCl dose in post-chlorination. It is recommended to investigate and validate this experimentally.

 As mentioned in paragraph 1.2.1, a more thorough investigation involving more solutes – and preferably at low, environmentally relevant concentrations – is highly recommended to further determine the effects of an oxidative and reductive pre-treatment on activated carbon adsorption of transformation products. To achieve this, a full identification and quantification of transformation products is necessary, as this would enable to investigate more TrOCs, including the ones with low solubility. This is important, as many relevant TrOCs are only weakly soluble in water. However, establishing a full identification and quantification of the transformation products is not a straightforward or simple task. As such, another recommendation for future research, which may be more straightforward, is the determination of the overall toxicity of oxidation and reduction products. If a decrease in toxicity is found for one of the degradation techniques, a full identification and quantification of the transformation product mixture may not be necessary. For ozonation, there have already been some reports of increased toxicity after ozonation, which resulted in the claim for the need of a more extended removal of transformation products, since ozonation transformation products should not be considered as viable end products [287, 288]. For reductive transformation products, however, no studies have been performed on toxicity of reduction products. Overall, a thorough investigation on toxicity of both oxidative and reductive transformation products should be performed. Furthermore, using chromatography techniques, transformation product mixtures can be split in different fractions. These separated fractions (each one containing a different part of the reaction products) can also be subjected to toxicity tests. If a higher toxicity is found in one fraction, the exact compound responsible for the toxicity can be further narrowed down, and identified. This procedure would enable to identify the most toxic transformation products, after which its presence and generation in different reactors can be monitored.

Identifying unknown transformation products represents a major analytical challenge, as no prior information is known about which compounds are present. Screening for non- target (unknown) compounds can be performed through acquiring full-scan accurate mass spectra of several samples (e.g. at different oxidant doses), and identifying peaks of interest (e.g. where peak intensity increases with higher oxidant dose). (An) elemental formula(s)

predefined mass accuracy decreases with increasing resolving power), and plausible structures corresponding to the elemental formula can be derived through comparison with libraries. Specialized software and algorithms are available for automatic peak selection and comparison of elemental formulas with libraries [408]. The number of derived possible structures needs to be reduced, through comparing measured fragmentation patterns with known fragmentation ions in databases. Non-target screening can be performed using Q- Exactive, Q-TOF or LTQ-Orbitrap, as these hybrid instruments allow to both perform precursor ion fragmentation, as well as the acquisition of accurate mass spectra. Ultimately, in non-target screening, HRMS data should be accompanied with other technologies such as hydrogen-1 and carbon-13 nuclear magnetic resonance to elucidate the structural traits of the unknown molecule, in order to truly identify the unknown compound [408].

 Because in ARP dehalogenation is an efficient degradation pathway, the effect of ARP should be investigated on biodegradation. This can be tested using both sand filtration and biological activated carbon filtration.

 A general conclusion, being made throughout this entire work, is that no single technology can be used for an efficient degradation of all TrOCs. As is known from literature, AOPs are less effective at dehalogenation reactions, and as found in this study, also ARPs are less efficient for removal of some TrOCs (e.g. diglyme being completely recalcitrant, and other TrOCs showing slow reaction). It may thus be expected that a combination of both AOPs and ARPs could result in an efficient removal of most – if not all – TrOCs present in the mix. In fact, the technology to combine both oxidation and reduction in a single process exist, under the name advanced oxidation and reduction processes (AO/RPs). In AO/RPs, both hydroxyl radicals, hydrated electrons and hydrogen atoms are formed, each of which can react with its preferred target moieties in TrOCs. These oxidative and reducing radicals are formed through the homolysis of water. Examples of AO/RPs are radiolysis of water through γ-irradiation or E-beam irradiation, but also photolysis of water through UV irradiaton at wavelengths lower than 190 nm (i.e. through vacuum-UV irradiation) [149, 151, 327]. To date, only limited attention has gone to TrOC degradation with AO/RPs, however, since both oxidative and reductive radicals are being produced simultaneously, the technology seems promising and should be investigated further.

 Since the cost of MIPs is very high (minimum producton cost of € 900/kg), it is unlikely that MIPs will find an application in water treatment for TrOC removal. A more cost- worthy application of MIPs could be recovery of platinum group metals or rare earth elements. As a rough calculation: with a platinum price of 36 000 € per kg in August 2016

[409], and assuming a MIP loading of 5 g Pt/kg MIP, one adsorption-desorption cycle of 1 kg MIP adsorbent could potentially recover 180 € of Pt. Obviously, this is just a very rough first estimate, which needs to be investigated properly. MIPs have already been applied successfully for adsorption of ions, such as cadmium(II), chromium(III) and yttrium(III), showing selective binding between MIP and ions [410–412]. As such, application of MIPs could be more cost-effective if the adsorbate is valuable, and thus worthy to be recovered. MIP adsorbents, tailor-made for recovery of rare earth metals could be applied in, for example, treatment of landfill leachate, as these valuable metals can leach from electronic waste. Since different ions will be present, special attention must go to developing the MIP with high selectivity, which would enable to recover Pt or other rare earth metals from complex matrices, even at low concentrations.