Optimization of hydraulic retention time and organic loading rate for volatile fatty acid production from low strength wastewater in an anaerobic membrane bioreactor

This study aims to investigate the production of volatile fatty acids (VFAs) from low strength wastewater at various hydraulic retention time (HRT) and organic loading rate (OLR) in a continuous anaerobic membrane bioreactor (AnMBR) using glucose as carbon source. This experiment was performed without any selective inhibition of methanogens and the reactor pH was maintained at 7.0 ± 0.1. 48,24,18,12,8 and 6 hr - HRTs were applied and the highest VFA concentration was recorded at 8 hrs with an overall VFA yield of 48.20 ± 1.21 mg VFA/100 mg COD feed . Three different ORLs were applied (350, 550 and 715 mg COD feed ) at the optimum 8 hr-HRT. The acetic and propanoic acid concentration maximums were (1.1845 ± 0.0165 and 0.5160 ± 0.0141 mili-mole/l respectively) at 550 mg COD feed . The isobutyric acid concentration was highest (0.3580 ± 0.0407 mili-mole/l) at 715 mg COD feed indicating butyric-type fermentation at higher organic loading rate.


Introduction
In past decades, anaerobic bioreactors have been utilized to recover value-added chemicals and bioenergy from different waste materials (Khan et al., 2016a;Wang et al., 2018). The recent development in this technology includes coupling a membrane module with conventional an anaerobic digestion system for treating industrial and municipal wastewater (Liu et al., 2018). Although biogas has been considered as the primary resource to be recovered from wastewater, research studies have proven the technical and economic feasibility for recovering VFA and biohydrogen from anaerobic bioreactors (Liu et al., 2018;Xin et al., 2017). the inhibition of the methanogenesis process so that the methanogens cannot consume the VFA during the final stage of the anaerobic process to convert them into biogas. The selective inhibition of methanogenesis is mainly carried out by heat shock and load shock treatment. Additionally, acidic and alkaline pH treatments are also applied for selective inhibition of methanogens (Khan et al., 2016a). These methanogens have been reported to have optimum microbial growth at a pH range of 6.5 to 8.2 (Mao et al., 2015), therefore reducing the pH below 6.5 or above 8.2 can be applied to inhibit the activity of methanogens.
Secondly, VFA can be simultaneously produced with biogas. This production technology involves the anaerobic process where the initial stage of acidogenesis and final stage of methanogenesis are separated through a multiple stage bioreactor design (Li & Yu, 2011;Schievano et al., 2014). In this experiment, the first type of production scheme was used to produce VFA without any selective inhibition of methanogenesis process. The reason for this approach is to improve the industrial application of the AnMBR as following the operating conditions any existing AnMBR model can be tuned to produce VFA without any design alterations.  (2017) states that a general decrease in HRT increases the VFA production through anaerobic acidification (Kuruti et al., 2017). However, the value at which the highest VFA yield and production rate of VFA would be achieved depends on bioreactor design, microbial community, and feed characteristics. In contrast, it has been identified that an initial increase in the loading rate increases the VFA production but at the same time can affect membrane fouling and bioreactor performance in terms of COD and nutrients removal (Khan et al., 2016b;Mao et al., 2015). An optimum value for OLR for any anaerobic process depends on the bioreactor design, and an OLR above the optimum value reduces the production of VFA Although numerous studies were carried out to optimize the production of VFA, most of them utilized anaerobic digestion (AD) process. A very limited number of researches performed to produce VFA from ANMBR where potential membrane fouling is an important area of concern. Additionally, the available studies aimed to optimize VFA production mainly used different inhibition process for methanogenesis. This is the first research which aims to find out the optimum hydraulic retention time and loading rate in the AnMBR treating low strength (synthetic) wastewater. The first stage of this experiment includes AnMBR operation in six different HRTs for 48, 24, 18,12,8 and 6 hrs. In the second stage, the same bioreactor was used with three different organic loading rates using 350, 550 and 715 mg/l COD of synthetic wastewater.

Characteristics of sludge and feed solution
The AnMBR used for this reaction had mixed liquor seed sludge from two different water treatment plants in Sydney, Australia (Cronulla water treatment plant and Central Park water treatment plant). At the beginning of each experiment, the sludge in the reactor had a mixed liquor suspended solid concentration of 10 g/l. The system was purged with nitrogen to get any unexpected air and oxygen out with diffused aeration tubes. The sludge mixture (30:70 ratios from Cronulla and Central Park respectively) was acclimatized in the reactor for 90 days until a constant COD and nutrient removal was obtained. Characteristics for mixed sludge and feed solution have been listed in Table 1.  respectively, by keeping the HRT fixed at 8 hrs. Glucose, NaNO3, and KH2PO4 were used as the main sources of carbon, nitrogen, and phosphorus respectively. The C: N: P ratio was kept constant at 100:5:1. Each trial for HRT and OLR involved 21 days of AnMBR operation in continuous mode. All relevant reactor operation conditions have been listed in Table 2. Throughout the period of stage 1 and 2 of bioreactor operation, the was maintained at 7.0 ± 0.1 and the ambient temperature in the laboratory was kept constant at 22 ± 1 °C pH fixed at 7.0. Referring the information provided in the introduction section (optimum microbial growth of methanogens at pH 6.5 -8), no inhibition process was applied to suppress methanogenic activity.

Solvent extraction
For preparing the sample, reactor effluent was collected and acidified to pH 2.0 to avoid any further biodegradation by the microorganisms. In order to get rid of the possible suspended matter, the acidified sample was centrifuged to 3500 rpm for 30 minutes. The next step involved taking a 4 ml sample and the addition of 1 g NaCl followed by the extraction with MTBE (2 ml). Once emulsion was formed, the sample was again centrifuged at 4000 rpm for 5 minutes.
The additional step of centrifuging was performed to break the emulsion. Once the separate organic phase was observed, it was collected with a syringe. The whole extraction process with MTBE was repeated for another time to make sure no organic content was left after extraction. Finally, the extracts were added together and taken to a separate test tube. The removed anhydrous Na2SO4 was added to make sure no water was resent before the sample was subjected to GC-MS. After dehydration, the sample was filtered using a 0.22 μm syringe filter to ensure the removal of suspended particles from GC-MS. The composition of fatty acid components was measured based on the retention times and mass spectra of peaks on the chromatograms derived from the extracted sample and the standard VFA solution (Banel & Zygmunt, 2011).

Quantification of VFA from GC-MS
Individual VFA concentrations were measured by gas chromatogram mass spectrometry method (GC-MS TQ8040, Shimadzu, Japan). For each measurement, the open tubular analytical column was used (VF-WAXms, Agilent, U.S). Helium was used as career gas with a flow rate of 2.05 mL/min. The temperature program started at 50˚C and was held for 5 min before ramping to 250˚C at 10˚C/min and was then held for 10 min. Electron impact ion source was set at 230˚C while the injection port and transfer line temperatures were held at 230˚C. Mass spectrometer (MS) operated in a selected ion monitoring (SIM) mode and in a full scan mode (m/z 15-550). Ions for detection of individual VFA in SIM mode were selected using the mass spectra of standards generated in SCAN mode.

AnMBR performance in COD and nutrients removal
For each stage of the AnMBR operation, bioreactor performance was analyzed in terms of COD and nutrient removal efficiency. Reactor effluents were added every 4 days during each trial for HRT and OLR. Figure 2 displays the efficiencies for COD, nitrate, and phosphorus removal at different HRTs and loading rates.  Figure 2, it is evident that the COD removal performance was steady at approximately 70 % throughout the trials for both stages of operation. NO3removal performance was maximum at the longest HRT (48 hrs) referring to the condition where the microorganisms were allowed enough time to undertake the denitrification process. As HRTs became shorter, a slight decrease in the NO3removal was observed. The reason may be associated with the fact that the contact time between the feed wastewater and the denitrifying bacteria was lowered and shorter HRTs. Additionally, as the influent COD was kept constant at 550 ± 10 mg/l, the organic loading rate was also increased at shorter HRTs. As the denitrification process involved processing the high organic loading, the minor decrease in the nitrate removal was expected (Wang et al., 2018). However, a minimum removal efficiency of 93.2% indicates efficient denitrification process in the bioreactor. As expected, the anaerobic process had a steady PO4 3-, the removal efficiency from 0.9 to 4.6% throughout the experiment.
During the second stage, at loading rates of 350 and 550 mg COD/l, the lowest COD removal efficiency was steady at about 70.9 ± 1.1%. NO3removal was observed above 98.2 ± 1.7% with a maximum removal efficiency of 99.4 ± 0.1%. PO4 3removal was steady within the range of 0.9 ± 0.2% to 1.8 ± 0.1%. Instead, at a loading rate of 715 ± 10 mg COD/l, the COD removal efficiency dropped to 65.1 ± 2.2% and consequent NO3removal efficiency dropped at 91.9 ± 0.5%. The deterioration in the general AnMBR may be associated to multiple facts

Membrane fouling of AnMBR
Properties of the membrane, sludge characteristics, wastewater properties, and operating conditions are the key factors that control the membrane fouling in a bioreactor (Guo et al., 2012). For this experiment, the same type of membrane, sludge, and synthetic feedwater were used throughout the experiment except for changes in the HRT (stage 1) and OLR (stage 2). Instead of discussing the mechanism of membrane fouling, this section includes discussions on how different hydraulic retention times and organic loading rates change the fouling pattern in the AnMBRs. MLVSS was fixed at 10 g/L at different HRTs and membrane fouling was measured in terms of Trans Membrane Pressure (TMP). All data have been plotted in Figure 3  hrs. This is because at this loading rate the soluble organics and the nutrients present in the feed water started to foul the membrane surface. The manufacturers' recommendation was to change/clean the membrane module once the TMP exceeds 30 Kpa, therefore during these 21 hrs of operation membrane cleaning was not performed.
At shorter HRTs (8 and 6 hrs) a steady increase in TMP development was observed until day 11 (up to 13.6 Kpa). The steady increase was followed by a rapid TMP development for both

VFA concentration at different HRT
The major components of VFA include acetic acid, propanoic acid, butyric acid, and valeric acid. The components are mainly produced in the acidogenic phase of anaerobic digestion.  From the experiment, it has been observed that the acetic acid concentration was nearly doubled (from 0.4922 ± 0.0134 to 0.8321 ± 0.0160 mili-mole/L) when the HRT was reduced from 48 hrs to 24 hrs. Production of acetic acid was increasing gradually when HRT was shortened and the maximum concentration was achieved at 8 hrs (1.1845 ± 0.0165 mili-mole /l). The change from 48hrs to 24 hrs indicates a shift in microbial activity from methanogenesis to acidogenesis. A gradual increase in acetic acid concentration was observed when the HRT was reduced further to 18,12 and 8 hrs. These trials with shorter HRT involved higher organic loading rates as the COD of influent wastewater was kept constant at COD of 550 ± 10 mg/l. Although the increase in acetic acid concentration had the highest degree of increase during the first change from 48hrs to 24 hrs, the following increase in the trend for acetic acid was associated with the high amount of organics and nutrients loading the bioreactor at a fixed MLVSS of 10.1 ± 0.1.
A further decrease in the HRT (from 8 to 6 hrs) reveals a drop in acetic acid concentration to 1.0095 ± 0.008 mili-mole/l. Although the initial decrease in HRT supported acetic acid production, an HRT below 8 hrs indicates an imbalance in the initial hydrolysis and acidogenesis process. More explicitly, the high amount of organics fed into the reactor at this HRT had a faster rate of initial hydrolysis whereas the acidogenic bacteria could not perform their action by consuming amino acids, sugar and other fatty acids that are produced through the initial hydrolysis process in anaerobic digestion.  (Wang et al., 1999). For this study, the reactor pH was maintained to 7.0 ± 0.1 by adjusting the pH of the feed solution. Therefore, the possibility of propanoic acid accumulation was very small during this experiment.
The remaining components present in our analysis included isovaleric, n-valeric, iso-caproic, n-caproic and heptatonic acid. Although no particular trend was observed in their concentration, an overall decrease in their concentration was observed when the HRT was shortened. The results may be associated with the fact that shorter HRT encouraged the production of major VFA components like acetic, butyric and propanoic acid. Consequently, VFA production shifted towards acetic, butyric and propanoic acid at shorter HRTs.   (1.1845 ± 0.0165 mili-mole/l) along with an increase for propanoic acid (from 0.27070 ± 0.03 to 0.5160 ± 0.0104 mili-mole/l). This initial increase in the influent COD increased the supply of organics and nutrients to the microbes that were performing acidogenesis.
However, at this loading rate the butyric acid concertation dropped 0.2393 ± 0.0406 to 0.2284 ± 0.0023 mili-mole/l. The subsequent drop in butyric acid concentration may be associated with the fact that it was degraded to acetic acid by acetic acid producing bacteria (Shen et al., 2018). Another reason for this drop in the butyric acid concentration is linked to the fact that, the system was not supported with the optimum pH (5.5 to 6.5) for butyrate type fermentation (Kuruti et al., 2017). An increase in the propanoic acid concentration was observed at this loading rate due to the reason that it was not consumed by any other acidogenic bacteria or methanogenic archaea (Khan et al., 2016a).
Finally, for 715 mg/l COD at the influent, there was a decrease in acetic acid and propionic acid concentration (from 1.1881 ± 0.0081 mili-mole/l to 1.1385 ± 0.0081 and 0.5160 ± 0.03 to 0.4167 ± 0.03 mili-mole/l respectively). In addition to this decrease, an overall drop in AnMBR performance was also observed (COD removal rate dropped to 65.1 ± 2.2% and the NO3removal rate dropped to 91.9 ± 0.5%). In contrast, an increase in the trend of isobutyric acid concentration was observed (0.2284 ± 0.0117 to 0.3580 ± 0.0407 mili-mole/l) in this condition. A possible reason may be at this loading rate VFA accumulation in the reactor triggered a momentary drop of the reactor pH below 6.5 that encouraged butyrate type fermentation. In summary, the high organic loading rate can be referred to as a trade-off between AnMBR performance and maximizing butyric acid production.
A possible future improvement opportunity can be operating the bioreactor by altering the pH condition into the acidic zone (5.5 to 6.5). Where the acetic acid and propanoic acid production can be maximized at this pH range (Begum et al., 2018), it would be interesting to see the possibility of overall VFA concentration exceeding the values that are obtained in the loading rate of 550 mg COD /l.

Advantages of VFA production from continuous AnMBR
The current study utilizes a continuous AnMBR to produce VFA from low-strength synthetic wastewater. Over the past few years, there have been a lot of experiments to extract VFA using anaerobic digestion, but most of these researches involve anaerobic digestion process in Hence, a continuous AnMBR can offer many advantages over the currently available VFA production processes. Firstly, VFA production in the continuous mode of operation makes it more applicable for wastewater treatment whereas the batch mode of operation is more practical for anaerobic digestion of organic waste. Secondly, this study has achieved a maximum VFA yield of 48.20 ± 1.21% (mg VFA/mg COD in feed solution) without the inhibition of methanogenic activity. Therefore, the result can be used in the simultaneous production of VFA and methane can increase the amount of revenue earned from the AnMBR. Thirdly, acidification is a major operational problem in AnMBR that are primarily caused by VFA accumulation (Khan et al., 2016b). Throughout this experiment, the reactor pH was maintained at 7.0 ± 0.1. Recovering VFA from this process offered an operational benefit by reducing the chance of rapid acidification. Finally, the membrane fouling profile during VFA production under different operating conditions has not yet completely discovered. Therefore, the findings from this research study could be beneficial to reduce/eliminate membrane fouling in the future.

Conclusion
The experimental results concluded that the highest individual VFA concentration was