Pilot-scale evaluation of sulfite-activated ferrate for water reuse Pilot-scale evaluation of sulfite-activated ferrate for water reuse applications applications

Abstract


Introduction
Global water stress has generated demand for recycling municipal wastewater effluent (i.e., water reuse) (Miller, 2006).Water reuse may be especially advantageous for rural and arid areas which are generally more impacted by water stress (Bauer, 2020).However, water reuse comes with associated public health risks including presence of pathogenic organisms and residual effluent organic matter (EfOM) which may include various organic contaminants of emerging concern (Crockett, 2007;Roberts and Thomas, 2006).Successful water reuse treatment must address these risks.Common approaches for risk mitigation include implementation of ozonation or radical-based advanced oxidation processes (AOPs) (Blackbeard et al., 2016;Gerrity and Snyder, 2011;James et al., 2014).However, Ozone and AOPs require significant auxiliary systems for generation, which may not be appropriate for small or rural systems.High valent iron species (i.e., ferrate (Fe(VI)) have emerged as an alternative oxidant to ozone, and other strong oxidants (Sharma et al., 2015).Implementation of Fe(VI) can transform organic contaminants (Jiang, 2014) and inactivate pathogens (Daer et al., 2021;Schink and Waite, 1980), while offering operational simplicity over other oxidation technologies (Goodwill et al., 2016).Fe(VI) can be produced on-site via an electrochemical (Jiang et al., 2009) or wet chemical process (Thompson et al., 1951), or purchased from commercial suppliers as a stable salt (e.g., K 2 FeO 4 , Monzyk et al., 2013).Fe(VI) has also been shown to produce fewer brominated DBPs when compared to ozonation (Huang et al., 2016;Jiang et al., 2019Jiang et al., , 2016)).
Utilization of FeSAOP may be especially advantageous for water reuse systems as a simple alternative AOP and allow traditional water treatment systems to implement an AOP as needed in response to an urgent water quality concern (Goodwill et al., 2021).However, there has been no continuous flow, larger scale (e.g., pilot) exploration of the activated Fe(VI) process, with existing pilot studies only examining non-activated Fe(VI) in conventional surface water (Goodwill et al., 2016) and wastewater treatment (Jiang et al., 2009).There are several extant research questions blocking full scale water reuse adaptation of FeSAOP including: (1) what are the performance benefits of FeSAOP compared to standard nonactivated Fe(VI) pre-oxidation, (2) what impacts does implementation of FeSAOP have on downstream physicochemical treatment processes, and (3) can a water reuse system using FeSAOP pre-oxidation produce effluent water quality that meets common water quality goals (e.g., the US Safe Drinking Water research gaps in a continuous flow treatment system using synthetic and field-collected wastewater effluent, creating a pathway towards adaptation.

Continuous Flow Apparatus and Instrumentation
A continuous flow experimental apparatus (CFA) was designed, constructed, and operated to replicate a full-scale water treatment process with pre-oxidation, coagulation, clarification, and dual media filtration (Figure 1).A full description of operational specifications and additional images of the CFA are provided in the supporting information (see Text S1 and Figure S1).The CFA was equipped with online instrumentation to continuously monitor major water quality parameters, in addition to benchtop instruments used to periodically analyze grab samples (Table 1).A full description of continuous and grab sampling is also given in Text S1.
Flow was then activated approximately 40 seconds downstream in a static mixer with sodium sulfite (i.e., FeSAOP) at a sub-stoichiometric activation ratio (0.5 µM SO 3 :1.0µM Fe(VI)) , following activation conditions recommended by Spellman et al., 2022.Activation ratio accounted for Fe(VI) decay during the 40 seconds between Fe(VI) dosing and activation.The resulting solution was allowed to react for 30 minutes in a pipe reactor with near plug-flow characteristics before particle removal steps (see Text S1).Filter effluents were subject to chlorination to quantify the regulated total trihalomethane (TTHM) DBP yield.Samples were buffered with borate at pH 7.0, chlorinated at 3 and 20 mg/L Cl 2 and then incubated 20 °C for 72-hours following a published method for examining Fe(VI) pre-oxidation on regulated DBPs (Goodwill et al., 2016).Average values presented in section 3 represent the average of all samples (grab or continuous) collected during the entirety of the eight-hour system runtime.S1.The synthetic effluent was spiked with 2.1 mg/L caffeine, a lowtoxicity target contaminant found in MWW effluents (Shon et al., 2006) that has been examined in bench-scale Fe(VI) experiments (Manoli et al., 2017a;Nie et al., 2020;Pan et al., 2020).

2-
]:[Fe(VI)] of 0.53 (±0.05), an activation ratio previously shown to be beneficial in reuse applications (Spellman et al., 2022).After the pre-oxidation reaction (30 min), the solution was dosed with FeCl 3 at 12 (±1.6)mg/L as Fe to achieve coagulation mainly via the adsorption-destabilization mechanism (see Figure S2) (Johnson and Amirtharajah, 1983), the desired mechanism when operating upflow clarification.Coagulant dose in all runs was adjusted to maintain a streaming current (see SI text S1) after coagulation of 0 (±30).The pH of the flow was not adjusted after coagulation (avg = 5.7).The up-flow clarifier was backwashed using DI water for 10 minutes after 270 minutes of run time to remove build-up of collected Fe particles and extend the media filter run time.

Pilot Runs Using Field-Collected MWW
Two additional comparative runs were conducted on the CFA with activated and nonactivated Fe(VI) pre-oxidation of field collected MWW.The non-chlorinated effluent utilized in these runs was collected at the Mattabassett District Water Pollution Control Facility, directly from the facility's secondary effluent flume (facility details provided in SI Text S2 and Figure S3).Average raw water quality conditions used for both runs are found in Table S2.The raw water was spiked with 9.1 mg/L of caffeine.Although higher than typical concentrations found effluents (Thomas and Foster, 2005), similar elevated caffeine concentrations have been utilized in prior activated-Fe(VI) experiments (e.g., Manoli et al., 2017) allowing for comparison of this work to bench scale experimentation.Solutions were coagulated using Nacrolyte 8100 (Nalco Water, Saint Paul, MN) cationic polymer by dosing until the solution had a circumneutral streaming current value (0±15).Coagulant was switched to polymer for these comparative runs so that all iron particles resulted from Fe(VI) and enable particle comparisons between Fe(VI) and FeSAOP.

Water Quality Improvements
The eight-hour averaged water quality improvements (e.g., changes between filter effluent and raw influent) resulting from continuous flow FeSAOP experiments are presented in Figure 2.  Organic matter (e.g., EfOM, caffeine) and nitrogen (e.g., TDN and NO 3 ) exhibited some level of removal, but to a far lesser extent (<30%).An oxidant dose four times the caffeine concentration (by mol) resulted in a 25% decrease in caffeine during our continuous-flow experiments.These results are in agreement with previous activated Fe(VI) literature which showed similar removal of caffeine in wastewater effluent (Manoli et al., 2017b).The incomplete transformation of caffeine is likely due to oxidant demand driven by EfOM that may react with oxidant species faster than with caffeine (Manoli et al., 2017b (Feng et al., 2018).Removals of caffeine by continuous flow FeSAOP in this study did exceed those by Fe(VI) alone in the aforementioned bench-scale study by (Feng et al., 2018)Higher removals of caffeine (near 100% transformation) absent of competing organic matter has been reported for other Fe(VI) activation methods such as acid-activated (Manoli et al., 2016), likely due to elevated Fe(VI) oxidation potential at lower pH, and silica gel-enhanced activated (Manoli et al., 2017a), where Fe(VI) adsorbs to the gel surface decreasing the kinetics of Fe(VI) decomposition.However, caffeine oxidation by silica-gel activation was also significantly impeded by the presence of competing organic matter (Manoli et al., 2017a).
Furthermore, both activation methods have additional downstream considerations, including significant pH adjustment and large (>30µm) SiO 2 particles, respectively, which may prohibit scale adaptation.Comparatively, caffeine removals were lower than ozonation at pilot and full scale where transformation of caffeine was demonstrated as high as 90%, but with orders of magnitude more O 3 than caffeine in waters with relatively low initial UV254 (<0.050) suggesting fewer competing organic compounds (Broséus et al., 2009).
Although overall transformation of organics (EfOM and caffeine spike) across the system were relatively low, implementation of FeSAOP did alter characteristics of organic matter.The significant change in UV absorbance with a lower total DOC removal suggests the FeSAOP processes was effective at transforming aromatic electron arrangements and double-bonded organics (e.g., humic acids), but did not target remaining straight-chain aliphatic compounds (e.g., caffeine).The shift in DOC structure is also demonstrated by the 60% decrease in specific UV absorbance (SUVA).Influent SUVA exceeded 6.5 implying the raw water carbon was largely made up of high molecular weight, hydrophobic, and aromatic compounds (e.g., caffeine which comprised 39% of influent DOC) while lower effluent SUVA (2.8) suggests remaining DOC had a lower molecular weight and was more aliphatic (Edzwald, 1993).The decreased fraction of aromatic DOC compounds demonstrated with SUVA results are also supported by excitation-emission matrices (EEMs), where there was a notable decrease in fluorescent DOC between the system influent and the filter effluent (Figure S4).

Steady-state Performance of Surrogate Parameters
Data for key continuously monitored effluent parameters are shown in Figure 3. Turbidity was continuously removed at >95% and never exceeded the CA22 2.0 NTU limit for no directcontact reuse (e.g., irrigation, toilet flushing, etc.) nor the 0.3 NTU Treatment Technique requirement in the SDWA (see Figure 3A).The UV254 absorbance was also removed across the system throughout the run at a relatively high rate (e.g., 73%) and never exceeded 0.050 cm -1 (Figure 3B).Effluent turbidity and UV absorbance levels are similar to those achieved in previous Fe(VI) continuous flow experiments on surface waters (Goodwill et al., 2016).packed bed which it likely does not (Yao et al., 1971).S3, respectively.

Disinfection and Byproducts
The regulated total trihalomethanes (i.e., TTHMs) DBP yield was measured after chlorinating filter effluent at a low and high dose of Cl 2 (Figure 5 (Hua et al., 2015;Reckhow et al., 1990).This is demonstrated by the decrease in UV254 absorbance across the treatment train (discussed above), as presence of aromatic and unsaturated bonding in system effluent would have showed more elevated UV absorbance (Edzwald et al., 1985).It is noteworthy that brominated species accounted for nearly all TTHM with limited formation of trichloromethane under both doses.This is in agreement with prior studies that suggest brominated DBPs are of elevated concern in water reuse applications (Spellman et al., 2022).The elevated formation of brominated THMs is attributable that chlorine oxidizing effluent bromide to HOBr, plus existing HOBr formed during FeSAOP (Spellman et al., 2022), which substitutes faster with active sites on residual organics than HOCl, leading to increased yields of brominated THMs (Hua et al., 2006;Symons et al., 1981).
Although brominated DBPs are of concern with FeSAOP in reuse scenarios (Spellman et al., 2022), yields would likely be less than if the same water was ozonated, as Fe(VI)-based oxidation generally produces fewer brominated DBPs than O 3 due to the slower kinetics between Fe(VI) reactive species and Br - (Jiang et al., 2019(Jiang et al., , 2016)).Both pre-oxidation methods resulted in near complete removal (>99%) of PO 4 and fecal coliform pathogens, as well as high removal (>96%) of total residual Fe.Elevated removal of Fe and PO 4, even with the coagulant switched from FeCl 3 to polymer (see Section 2.3), suggests the particles resulting from pre-oxidation are primarily responsible for PO 4 removal via adsorption.

Removal of Select Contaminants
However, the treatment system was ~15% less effective at decreasing turbidity after FeSAOP pre-oxidation when compared to Fe(VI) alone, resulting in an increase in average effluent turbidity from 0.35 to 0.42 NTU.Average effluent particle counts also more than doubled in FeSAOP filter effluent from 7 particles/mL to 17 particles/mL, but were not exceedingly high in either trial.Average effluent particle size also changed significantly with non-activated pre-consistent prior studies which demonstrated FeSAOP shifts particle size distribution towards an elevated number of relatively smaller particles, likely due to the nearly instant particle precipitation mechanism where dimeric Fe hydroxo-species are rapidly supplied in hydrolysis resulting in more amorphic iron particles (Bzdyra et al., 2020;Goodwill et al., 2015).

Impact on Organics
Implementation of both pre-oxidation methods significantly altered the characteristics of organic matter during continuous flow experiments.Figure 7 gives the changes in fluorescent organic matter resulting from Fe(VI) and FeSAOP pre-oxidation.Regions presented in each EEM represent different types of organic matter components expected at each excitationemission fluorescence, as originally presented by Chen et al. (2003).The raw field-collected MWW exhibited the highest intensities in Regions II, III and V, indicating significant amounts of aromatic proteins, fulvic acid-like, and humic acid-like organic matter, respectively, which is typical for a MWW effluent EEM (e.g., Zheng et al., 2014).The treatment process with both preoxidation methods were relatively successful at decreasing fluorescence in all regions.The largest change in regional intensity (see Text S3.3) appeared in region V, with a 47% and 54% decrease compared to influent for Fe(VI) and FeSAOP, respectively.This intensity decrease in region V (i.e., humic acids) is expected as humic-like compounds are generally more susceptible to oxidation and coagulation processes, due to the presence of electron rich moieties and anionic polyelectrolytic properties of humic substances (Amy et al., 1 .While both runs were effective at removing humic-like substances, implementation of FeSAOP resulted in a notably larger decrease in intensity volume in region IV and V between excitation 260-280 (56% versus 37% decrease) suggesting improved transformation of organics with FeSAOP compared to Fe(VI).This is attributable to the presence of SO 4 during FeSAOP (Shao et al., 2020), which have previously been shown to have a high affinity for humic substances (McKay et al., 2014;Yang et al., 2015), and are known to improve transformation of electron-rich moieties when compared to Fe(VI) alone (Feng et al., 2018;Spellman et al., 2022).A difference between Fe(VI) and FeSAOP was also noted in TTHMs yield.When chlorinated with 29 mg/L Cl 2 , FeSAOP decreased TTHM yield from 77.4 to 58.8 µg/L, a 24% decrease in TTHMs compared to Fe(VI).THM modeling from Solarik et al. (2000) suggested TTHM yields would be lower in FeSAOP effluent than Fe(VI), but to a greater extent (33% less) than what was observed experimentally (24% less).However, the TTHM yields in both conditions were lower than model predictions in both experiments.Similarly, the lower THM yield with FeSAOP compared to Fe(VI) scaled with differences in both filter effluent UV254 absorbance (see Figure 6) and region V fluorescence (see Figure 7) where FeSAOP outperformed Fe(VI).THM yield was also directly proportional to decreases (influent to effluent) in SUVA, where FeSAOP decreased SUVA 31% while Fe(VI) only decreased SUVA by 5%.These results demonstrate FeSAOP is more effective at oxidizing aromatic and doublebonded (i.e., DBP-forming) compounds (Hua et al., 2015;Reckhow et al., 1990).This improved transformation of THM precursor compounds is likely due to the presence of highly reactive Fe(IV)/Fe(V) and SO 4 as both SO 4 oxidizing DBP precursor organics (Sarathy et al., 2011;Wang et al., 2014).

Operational Considerations
Operational parameters were also considered during both runs where polymer was used as coagulant.Figure 8 compares the development of headloss across the media filter during both runs.The headloss presented is normalized to the filters calculated clean bed headloss (i.e., headless = 0 at 0 min).Headloss increased linearly and modestly, with both runs resulting in < 28 in-H 2 O (2.3 ft) after 8 hours.The final 8-hr headloss in both studies was relatively low compared to maximums typically set for media filters (e.g., 75-120 inches; Davies and Wheatley, 2012;Stoddart and Gagnon, 2015), implying filters could have had notably longer run times.The relatively slow development of headloss is attributable to significant Fe particle removal occurring prior to filtration in the clarifier (i.e., total Fe after clarification <0.60 mg/L).The predicted headloss in each run was modeled according to Ives (1970), assuming the majority of particles collected in the filter were iron-oxides (Figure 8).Experimental headloss data fit well with the modeled headloss having strong (>0.99)correlation in both trials.There were key differences in modeled and measured headloss when comparing the two pre-oxidation methods.
FeSAOP implementation developed headloss at a ~50% faster rate than non-activated (1.6 in-H 2 O/hr and 1.1 in-H 2 O/hr, respectively), a notable operational tradeoff between the examined oxidation methods.This increase in headloss development from FeSAOP is driven by two main factors: (i) the aforementioned relatively smaller FeSAOP average particle size, which would be collected more-efficiently by the media filter (e.g., Figure 4), and (ii) slight variations in the way FeSAOP particles are arranged on the media after collection likely attributable to morphological differences (Tobiason and Vigneswaran, 1994).respectively).Headloss values reported are normalized to the filters calculated clean bed headloss.Solid (Ives, 1970)) (See Text S3.4 and Table S3).
The chemical operating cost of FeSAOP, an important operational consideration for scale adaptation, has not previously been discussed in the activated Fe(VI) literature.While bulk costs of reduced sulfur compounds that could be used for FeSAOP (e.g., sulfite, bisulfite, etc.,) are readily available and approximated to be ~$1.30/gallon of bulk solution (Narragansett Bay Commission, 2022), the bulk cost of potassium ferrate (K 2 FeO 4 ) powders are currently unknown due to current lack of mass production at water treatment scale.However, manufacturers approximate electrochemically-generated K 2 FeO 4 powder (Monzyk et al., 2013) would cost between $3-45 per dry pound (varying with exact method used), calculated with known electric demands and raw chemical costs for production (Ramchandran and Goodwill, 2022).Using these estimates, the chemical operating cost for FeSAOP (not including pumping or mixing) per 100,000 gallons (~3.8x105L) of raw water under operating conditions presented in section 2.2 would be approximately $25 per 100,000 gallons.

Conclusions
In general, continuous flow evaluations demonstrate FeSAOP as a viable pre-oxidation technology in a water reuse setting, resulting in several improved water quality outcomes, as supported by the data presented herein.The following conclusions will help create a pathway for FeSAOP adaptation at scale: Reuse systems with FeSAOP pre-oxidation can produce water quality meeting most CA22 and SDWA requirements, such as effluent turbidities <0.14 NTU FeSAOP does not have appreciable detriment to downstream processes and may improve physicochemical treatment effectiveness.However, FeSAOP does result in 50% faster development of headloss across media filters, a notable tradeoff compared to Fe(VI) Operating with FeSAOP pre-oxidation generally produced higher-quality water than the system using traditional Fe(VI) FeSAOP effectively transforms aromatic and double-bonded EfOM compounds.
However, effective EfOM transformation hinders oxidation of target compounds (i.e., caffeine), similar to comparable strong oxidants.
PO 4 is effectively removed via adsorption onto particles resulting from both FeSAOP and Fe(VI) pre-oxidation methods FeSAOP transforms DBP precursor organic matter and lowers yields by 25% when compared to Fe(VI) alone.However, formation of nitrogenous-DBPs in waters with elevated NO 3 requires additional research as there may be implications for water reuse systems.

Figure 1 :
Figure 1: (A) Process flow diagram and (B) image of the continuous-flow pilot water treatment apparatus

FeSAOP
pre-oxidation followed by coagulation, clarification, and filtration generally led to improved overall water quality relative to raw water.Greater than 95% of turbidity, PO 4 , and total Fe were removed across the treatment system.High performance of the clarifier and dual media filter was demonstrated by the low turbidity and effluent particle counts (discussed further below).PO 4 was removed below detection in filter effluent, with grab sampling between clarification and filtration indicating almost all removal occurred during clarification.This PO 4 removal is likely from adsorption onto Fe particles present in the oxidant reactor and clarifier that originated from pre-oxidation and Fe-based coagulation.This also supported by the similarly high (~99%) removal of total Fe (Fe from Fe(VI) and FeCl 3 ).Although coagulation and preoxidation introduced 19 mg/L Fe, effluent continuously reported < 0.2 mg/L Fe.

Figure 2 :
Figure 2: Activated Fe(VI) performance with synthetic MWW related to key filter effluent water quality

Figure 4 :
Figure 4: Bars represent the relative particle size distribution (i.e., fraction of total particles) from grab

Figure 6
Figure 6 compares the relative performance removing select contaminants by FeSAOP and

Figure 6 :
Figure 6: Relative performance of key filter effluent water quality parameters comparing Non-Activated

Figure 7 :
Figure 7: Comparison of organic matter fluorescence (i.e., EEM) in field wastewater and filter effluent

Figure 8 :
Figure 8: Development of headloss (presented in inches of H 2 O) across the dual-media filter after (A) Non-Activated Fe(VI) and (B) FeSAOP pre-oxidation and the filter effluent turbidity (panels C and D,

Table 1 :
Summary of continuous (cont.)andgrab sample monitoring of pilot system.A detailed description of parameters below is provided in Text S1.