Fouling mechanisms and reduced chemical potential of ceramic membranes combined with ozone

Ceramic membrane systems offer potential benefits in water recycling as they are robust and long lasting. For example, it has been reported that ceramic membrane installations have been operating with no loss in flux since 1998 (Dow et al. 2013). Due to their robustness, ceramic membranes have been explored in conjunction with ozone. Operating with a small ozone residual at the membrane surface (∼1 mg/L) when treating waste water, yields much higher fluxes than polymer membranes (Clement et al. 2009; Lehman & Liu 2009; Dow et al. 2013). Because of these higher fluxes, a smaller membrane area is needed to treat the same volume of water. With this, and improvements in the design of the ceramic membrane modules, ceramic membranes are now considered cost competitive to their polymer counterparts.

While ceramic membrane technology is potentially more economically viable due to this performance improvement, the use of ozone in contact with the ceramic membrane material is unique in water treatment and can offer other potential benefits. For example, it has been shown that a residual concentration of ozone at ∼1 mg/L flowing to the membrane improved disinfection of the backwashed effluent, e.g. with respect to E. coli present in the secondary treated effluent feed (Dow et al. 2013). The ozone clearly has an effect when flowing continuously through the fouling layer and membrane material that increases water disinfection. Despite the practical findings of increased flux and enhanced disinfection, the mechanisms leading to them remained unconfirmed. Ozonization is a well-known water treatment process, commonly used to improve aesthetics – e.g., remove color or to destroy micro-contaminants – by reacting with organic matter. However, on contact with aluminum oxide (the material used to construct many ceramic membranes), ozone is converted catalytically to form highly reactive hydroxyl radicals. There are thus various potential explanations behind the high fluxes that arise from the presence of ozone residuals, relating, among other things, to the altered organic chemistry and/or ozone reaction with the membrane material.

In an attempt to isolate the effects from the altered organic chemistries and the residual ozone in contact with the membrane, an ‘ozone quenching’ experiment was set up. The reducing agent sodium metabisulfite (Na₂S₂O₅) was injected immediately upstream of the ceramic membrane with the purpose of removing the ∼1 mg/L ozone residual at the membrane surface. By observing the response of the transmembrane pressure (TMP) over time, before and during quenching, and then after quenching had stopped, the effect of ozone on the membrane surface could be observed and confirmed.

High fluxes are also known to occur with the application of ozone, although frequent chemical backwashes are still required. This trial showed that sodium hypochlorite (NaClO) was the most significant cleaning chemical used (Dow et al. 2013). While its scale of use is small compared to the volume of water treated (8 L of 13% hypochlorite solution per ML of water treated), exploring the potential for its reduction at high fluxes is of interest. A test involving fewer NaClO backwashes was carried out to observe the extent of ozone's beneficial effects in assisting with membrane cleaning.

Ozone was dosed in a separate system external to the pilot plant, by injecting the gas directly into a Statiflo static mixer as shown in Figure 2. The injection flow rate and gas phase ozone concentration were adjusted to ensure that an ozone residual of 0.8 mg-O₃/L or more was measured in the water immediately prior to the membrane. During some trials, polyaluminium chloride coagulant was dosed (PACl – 23% as Al₂O₃) directly into the line upstream of the membrane to give 3 mg-Al³⁺/L. The minimum dose was determined by ‘jar tests’, as described in a previous paper (Dow et al. 2013).

Chemically enhanced backwashing (CEB) was carried out with both oxidant and acidic solutions. For the oxidant CEB, 100 mg/L of NaClO, was made in the backwash tank at room temperature and backwashed through the membrane at 2 bar pressure after every five regular backwashes. For the acid CEB, a solution of pH 2 hydrochloric acid (HCl) was made in the backwash tank and backwashed through the membrane in a similar way to the oxidant CEB, but only after seven oxidant CEBs.

The unique feature of this test was dosing a reducing agent immediately prior to the membrane, to quench residual ozone and remove any effects associated with ozone's interaction with the membrane and/or fouling layers. As a result, only the ozone-reacted organics were present at the membrane. This was done by dosing sodium metabisulfite, Na₂S₂O_5, just upstream of the membrane – see Figure 2. The ozone residual was measured prior to the membrane by the indigo method, and confirmed the quenching effect.

Na₂S₂O₅ was dosed upstream of sampling point 3 (Figure 2). The quenching effect of Na₂S₂O₅ was confirmed when ozone was substantially removed (<0.08 mg-O₃/L detected) at this point when ozone was injected. The indigo method's limit of detection in this configuration was 0.001 mg-O₃/L. Tests were conducted without and with coagulant dosing, but when coagulant was used, ozone quenching was incomplete. However the ozonated feed without coagulation was suitable for confirmation of the effect being studied.

A flux of 50 L/m²/h was chosen for the quench test, as it was suitable for operation with and without ozone, according to the trial outcomes (Dow et al. 2013). The volume between the injector and sample point 3 was estimated at 34.6 L, so, at this flux, the ozone contact time with the feed water prior to the membrane was about 85 seconds.

For the reduced chemical cleaning test, the CEB frequency was reduced to 10% of normal at the highest flux achieved (182 L/m²/h), which involved both ozone and coagulant dosing. A test with no oxidant CEB was also conducted to observe the effect of chemical free operation on performance.

Disinfection was also effective, using E. coli present in the secondary treated effluent as the challenge particle (Dow et al. 2013). E. coli was not detected in the permeate (limit of detection 10 orgs/100 mL), whereas 3 × 10⁴ orgs/100 mL were measured in the feed. Without ozone, a 1.2 to 1.3 log increase in E. coli was measured from feed to backwash. This was expected due to the concentrating effect of operating the membrane at 93% recovery (a 14-fold concentration factor of solids rejected by the membrane). However, when ozone was added, the concentration of E. coli present in the backwash was reduced by 0.5 to 0.7 log, which was evidence of inactivation of the organisms caught on the membrane surface due to the continuous stream of ozonated water flowing over them for periods up to the entire filtration cycle time. Despite this, a direct measure of the effect of ozone was needed to explore its effect on membrane fouling.

The effect of ozone on the membrane is thus confirmed, where Na₂S₂O₅ dosing quenched the residual ozone after 85 seconds of contact with the incoming water. Not only does ozone react with the organics in the wastewater, it also operates specifically on the membrane. However, this result does not show whether the ozone continues to react with the accumulated cake layer on the membrane surface, or reacts with the membrane itself, which is the subject of current studies. If the organics have already reacted with ozone and their ozone demand is reduced by the time they accumulate on the membrane, it seems likely that the remaining ozone in the aluminum oxide membrane material will be converted to hydroxyl radicals. At a flux of 200 L/m²/h, water flows at an average speed of 20 cm/h through the membrane. Ignoring porosity and tortuosity, for a membrane and substrate of 0.5 cm thickness, the water spends 1.5 minutes inside the ceramic's porous structure before leaving as permeate. This is sufficiently long for the catalyzed reactions between ozone and aluminum oxide that are believed to produce the highly reactive hydroxyl radicals. Recent studies however have shown that ozone decomposition to form hydroxyl radicals on aluminum oxide is not favored. Ozone is likely, instead, to react catalytically with organic species adsorbed on the membrane surface (Nawrocki & Fijołek 2013). Indeed, catalyst bed studies of aluminum oxide have shown double the decomposition rate of natural organic matter compared to that arising from ozone alone (Kasprzyk-Hordern et al. 2006). The effect is now being attributed to ozone reactions with organics within membrane pores (Zhang et al. 2013), which is a subject of ongoing research.

Another test was conducted to determine whether NaClO consumption during oxidant CEBs could be reduced. According to our trial, the most heavily used chemical was the coagulant dosed at 3 mg-Al³⁺/L, requiring 22 L of the chemical (23% poly-aluminum chloride as Al₂O₃) per ML treated (Dow et al. 2013). However, NaClO was the most heavily consumed cleaning chemical in the trial, requiring 8 L of 13% hypochlorite per ML treated.

It seems likely that NaClO use can be reduced. While frequent NaClO CEBs were effective at maintaining a stable TMP rise, reducing their frequency leads to potential instability and subsequent rapid fouling. If chemical consumption was a key issue, then the amount of the major cleaning chemical, NaClO, could be reduced significantly, if triggered when needed (i.e. when a rapid rise in TMP is observed). This makes it suitable for highly variable feed waters like secondary treated effluents, where more intense fouling events occur occasionally. In these cases, ceramic membrane systems can maintain high fluxes and low chemical use, but will require additional oxidant CEBs at times. Flux can also be reduced to achieve more stable performance (Dow et al. 2013), but the result is lower plant productivity.

The trial of ceramic membranes combined with ozone has confirmed that high flux performance is achievable in water recycling applications. This is key to achieving cost competitive performance compared to polymer membranes. However the process is unique in water treatment, where ozone interacts with the ceramic membrane to produce high fluxes, as well as increasing disinfection and reducing chemical use.

Quenching ozone before it reached the membrane confirmed that an ozone residual on the membrane surface has a controlling effect on TMP rise. While it is believed that ozone will react to form hydroxyl radicals, recent research has shown that ozone is likely to react catalytically with organics adsorbed on the ceramic membrane surface instead. These findings are assisting efforts to explore the fouling mechanisms for the unique combination of processes studied, in which ozone not only alters the chemistry of the organics but also inhibits TMP rise over time. This means that ozone consumed in the fouling layer, or within the membrane material, keeps the membrane clean and enables high flux operation. Further work is in hand to study the effect(s) of interactions between ozone and the fouling layer and membrane material.

The project also explored the potential for reducing the amount of NaClO consumption, the most significant cleaning chemical used in the pilot study. Its consumption was reduced by 90% from the initial rate, with promising results in terms of TMP stability over time. However, when TMP started to rise and as no CEB was used, the rise continued unimpeded. This led to the conclusion that reduced chemical use is possible, but only with close TMP monitoring. This may be especially valuable in cases where reduced chemical use is important – e.g. remote locations. However, the amount consumed, if stable operation is preferred (the standard setup for the trial), is not particularly significant. Thus, while ozone can reduce chemical use, its cleaning effect is better used to achieve higher fluxes. Reduced chemical use can be traded for higher fluxes, but lower fluxes mean larger plant size.

This project was financially supported by the Australian Water Recycling Centre of Excellence, funded by the Commonwealth of Australia. Funding from Melbourne Water, South East Water and Water Research Australia is also gratefully acknowledged.

The authors also wish to acknowledge the involvement of Pam Kerry (South East Water), and Alastair McNeil and Hazel Ho (Black & Veatch).