Enhancement of volatile fatty acid production using semi-continuous anaerobic food waste fermentation without pH control

Abstract

This study proposed a cost-effective and high-yield volatile fatty acid (VFA) production strategy using anaerobic food waste (FW) fermentation without pH control, which could be recommended for practical scale VFA production and FW treatment. Efficient hydrolysis, high VFA production (867.42 mg COD per g of VS) and VFA/SCOD (88.65%), and slight methane production (20.56 mL per g of VS, only 5.34% of theoretical value) were obtained via semi-continuous anaerobic FW fermentation under the optimized conditions, which were a substrate to inoculum ratio of 5, a temperature of 40 °C, a solid retention time of 7 d and an organic loading rate of 9 g VS per L per d. The highly-efficient VFA production was mainly attributed to the proper pH, which was not adjusted artificially and was maintained at 5.2–6.4 by the neutralization effect of ammonia release. And both the kinetic model and the carbon mass balance analysis showed that the substrate was adequately utilized to produce VFA under the optimized conditions. These results were of great guiding significance for the practical continuous recovery of VFA from FW.

---

1. Introduction

One of the biggest challenges in human development in the 21^st century is meeting the food supply for the increasing population while reducing the negative effect of unavoidable food waste (FW) on the environment.¹ As the largest fraction of municipal waste, the amount of FW is increasing extremely quickly every year in China,² and it has already become the main source of decay, odour, toxic gas, and groundwater contamination, which severely threatens environmental health and security.³ Thus, the efficient, sustainable and eco-friendly treatment and management of FW has become an urgent task. At the same time, FW also has been regarded as an ideal substrate to generate energy due to its highly moist nature, easily biodegradation, well balanced carbon and nutrient content, and organic composition.⁴ As an economical and eco-friendly strategy, anaerobic fermentation has attracted widespread attention for treating FW and meanwhile harvesting biogas and volatile fatty acids (VFAs).⁵ Therein, VFAs are generally regarded as a better product than biogas owing to their high yield and wide application. Based on previous studies, VFAs have been successfully used for nutrient removal enhancement,⁶ biodegradable plastic production,⁷ biogas and biodiesel bioconversion,⁶ polyhydroxyalkanoate biosynthesis⁸ and electricity generation.⁹ Given the high practical application value of VFAs, a study on high-efficiency VFA production from FW is of great significance.

Three steps of hydrolysis, acidogenesis and methanogenesis are involved in anaerobic fermentation, and VFAs are produced in the first two steps. Complex organic substances are solubilised by hydrolysis, and then the dissolved organic components are utilized by acidogenic bacteria for their growth while accumulating VFAs.¹⁰ Thus, to enhance VFA production, the hydrolysis must be accelerated to produce more soluble substrates, while the acidogenesis rate should be promoted to maximise the conversion of these substrates to VFAs, and equally importantly, methanogenesis should be prevented to reduce the consumption of the generated VFAs.¹¹ In addition, anaerobic fermentation is greatly affected by the relevant operational conditions. The specific operation conditions for the hydrolytic and acidogenic stages have been widely studied in batch tests,^9,12 but relevant research in the semi-continuous mode is still scarce although it will better guide future large-scale applications. The key operational parameters for semi-continuous VFA production mainly include temperature (T), solid retention time (SRT), organic loading rate (OLR), and pH etc.¹³ Recently, Jiang et al.⁴ has found that the optimal operating conditions for VFA production via FW anaerobic fermentation were pH 6, T = 35 °C and OLR = 11 g per L d, and Chen et al.⁹ has reported that pH 8, C/N = 22, T =37 °C and a fermentation time of 6 d were optimum to enhance VFA accumulation. However, the studies mentioned above only focused on maximizing VFA production but did not investigate in detail the influences on hydrolysis, acidification and methanation. Besides, these studies improved VFA production by adding a large amount of NaOH/HCl to obtain an alkaline or weakly acidic solution, which caused some unavoidable disadvantages such as higher operation cost and complexity, unmanageable fermentation liquid with a high chemical concentration, and potential negative impacts of chemicals on microorganisms.³ To date, few studies reported that VFAs could be efficiently produced without controlling the pH. Our previous work on batch anaerobic FW fermentation showed that, by adjusting the substrate to inoculum ratio (S/I), the self-formed buffer action system could automatically control the pH within a suitable range (5.7–6.5), which was beneficial for hydrolysis and acidogenesis and inhibited methanogenesis.³ In this way, the operation complexity and production cost could be greatly reduced without the extra addition of acid or alkali for practical application. However, whether this strategy can be successfully carried out in the semi-continuous mode has not been verified, and the effect of the optimum S/I on VFA production under specific optimal conditions has not been investigated. Meanwhile, with the aim of achieving the practical application of continuous VFA production without artificial pH adjustment, it is also necessary to study the optimal conditions of the key influencing factors including T, SRT and OLR in the semi-continuous mode.

The main objectives of this work were therefore (1) to investigate the optimum operation conditions (S/I, T, SRT and OLR) for VFA production via semi-continuous anaerobic FW fermentation using response surface methodology, and to verify the optimized model; (2) to analyze the effects of the optimum S/I on hydrolysis, acidification and methanation; and (3) to discuss the factors for efficient VFA production under optimum conditions without artificial pH control.

2. Materials and methods

2.1 FW and inoculum

FW, which mainly consisted of rice, noodles, vegetables and meats, was obtained from a cafeteria in Harbin Institute of Technology (Harbin, China). After removing the superficial oil, the FW was crushed using an electrical blender and stored at 4 °C for experimental use. The waste activated sludge (WAS) obtained from the secondary setting tank of the Harbin Wenchang sewage treatment plant (Harbin, China) was used as inoculum. The characteristics of the FW and inoculum used in this study are shown in Table 1.

Table 1 Characteristics of FW and inoculum

Parameters	FW	Inoculum
Total solid (TS) (g L^−1)	66.17 ± 0.81	27.63 ± 0.53
Volatile solid (VS) (g L^−1)	65.15 ± 1.32	15.96 ± 0.74
Total chemical oxygen demand (TCOD) (g L^−1)	104.37 ± 10.40	25.10 ± 3.41
Soluble chemical oxygen demand (SCOD) (g L^−1)	50.64 ± 6.63	1.19 ± 0.37
Total biological oxygen demand (TBOD) (g L^−1)	72.46 ± 5.34	9.67 ± 0.55
Total carbohydrate (g L^−1)	56.72 ± 5.75	3.58 ± 0.58
Total protein (g L^−1)	10.47 ± 1.04	21.14 ± 2.33
Soluble NH4^+–N (mg L^−1)	340.38 ± 9.45	160.76 ± 1.07
pH	5.62 ± 0.10	7.03 ± 0.13

---

2.2 Optimization of operation conditions

Response surface methodology (RSM) is always used to analyze the mutual relationships between the response and independent variables, and obtain the optimum operation conditions.¹⁴ RSM based on a five-level-four-variable central composite design (CCD) by Design Expert 8.0 was used in this study. The variations of S/I (based on the volatile solid content), T, SRT and OLR, and their levels are shown in Table S1 (ESI†). The ratio of VFA to SCOD (VFA/SCOD), indicating the amount of soluble substance that converts into VFAs, is often regarded as an evaluation of successful acidogenesis.⁴ Thus, VFA/SCOD was regarded as a dependent output variable. The response variable was fitted by the response surface regression procedure, and the second order polynomial equation and the analysis of variance (ANOVA) are described in section S1 (ESI).†

Thirty identical fermentation reactors (with a working volume of 500 mL) designed using RSM were operated in the semi-continuous mode (once-a-day draw-off and feeding) for VFA production, and the fermentation equipment is shown in Fig. S1 (ESI).† Each reactor was inoculated with a designated amount of FW and inoculum, and the final VS concentration in all reactors was 17 820 ± 635 mg L⁻¹. During anaerobic fermentation, according to the designated SRT and OLR, everyday a certain amount of fermentation mixture was withdrawn from each reactor, and then the same volume of the fresh FW and inoculum mixture was added. Because most anaerobic microorganisms could not tolerate hostile environments such as alkaline or acidic conditions, the pH in all reactors was not controlled. All the reactors were flushed with nitrogen gas (99.9%) for 5 min to remove oxygen after the daily feeding, and then the reactors were stirred in a water-bath shaker (180 rpm). The liquid samples were taken from the reactors daily for SCOD and VFA concentration analyses. When the SCOD and VFA concentrations remained relatively stable, the reactors were considered to be in a steady state and then the related investigations were carried out.

2.3 Operation of semi-continuous reactors under optimal conditions

Some previous studies have investigated the effect of SRT, OLR, T and pH on VFA production, but less research has been devoted to probing the effect of the S/I ratio especially in the fermentation process without pH control. Thus, three identical semi-continuous reactors were operated to observe the effects of the optimum S/I on hydrolysis efficiency, and VFA and methane production. Reactor #1 solely contained FW, reactor #2 was inoculated with just WAS, and reactor #3 was seeded with FW and inoculum according to the optimal S/I obtained using RSM. In the three reactors, T, SRT and OLR were controlled in accordance with the optimum conditions obtained using RSM, and the final VS concentrations were 17 820 ± 635 mg L⁻¹.

2.4 Analytical methods

The analyses of TS, VS and ammonia were conducted in accordance with standard methods.¹⁵ TCOD and SCOD were measured using a COD analyzer (DR 1010, HACH, USA). Protein content was determined using a BCA Protein Assay Kit (P0012, Beyotime Institute of Biotechnology, China). Carbohydrate content was determined using the phenol-sulfuric acid method with glucose as standard. The pH value was measured using a pH probe (Germany WTW Company pH meter). VFAs were analyzed using a gas chromatograph (GC) (HP 7890, Agilgent Technologies, USA) with a flame ionization detector (FID). The injector and detector temperatures were programmed to be 200 and 250 °C, respectively. The oven temperature of the GC began at 80 °C, then increased to 180 °C, and finally was held at 200 °C. The sample injection volume was 1.0 μL, and the carrier gas was nitrogen at the flux of 25 mL min⁻¹. The COD conversion factors of acetic, propionic, butyric and valeric acids were 1.066, 1.512, 1.816 and 2.036, respectively. Methane production was determined using a GC (GC-SC2, Shanghai Analytical Apparatus, Shanghai, China) with a thermal conductivity detector and a 2.0 m stainless steel column packed with TDS-01 (60/80 mesh). Three-dimensional excitation-emission matrix (EEM) fluorescence spectroscopy analysis was conducted using a fluorescence spectrometer (FP-6500, JASCO, Japan). Scanning emission (E_m) spectra from 220 to 650 nm were obtained at 1 nm increments by varying the excitation (E_x) wavelength from 220 to 450 nm at 5 nm increments. The scan speed was 2000 nm min⁻¹, and the slit widths were 5 nm for both excitation and emission monochromators. The EEM plots were generated from the fluorescence spectral data using Origin 8.0. All the soluble parameters were determined after centrifugation (10 000 rpm, 10 min) and filtration (0.45 μm). All tests were carried out in triplicate in parallel and the average values were determined to minimize error.

2.5 Kinetic analysis

The VFA production was related to hydrolysis, acidogenesis and methanogenesis, thus the kinetic model presented here describes the above three stages. The hydrolysis step was described by first-order kinetics, which has been commonly selected as an appropriate model to evaluate the hydrolysis rate of complex fermentation substrates.^16,17 The conversion of hydrolyzate and the transformation of VFAs to methane were described using Michaelis–Menten kinetic models, which are widely applied to evaluate anaerobic fermentation processes.¹⁸ Accordingly, the anaerobic fermentation was presented by the following four equations:¹⁹where k is the hydrolytic rate for particular organic matter (d⁻¹); k_h is the maximum utilization rate for the hydrolysis products (g SCOD d⁻¹); k_v is the maximum utilization rate for VFAs (g VFA d⁻¹); k_s,h; and k_s,v and k_s,m are the saturation constants (g SCOD). The concentrations of particular organic matter (S_p), hydrolysis products (S_h), VFAs (S_v) and methane (S_m) were expressed in g COD L⁻¹. The value of the constants k, k_h, k_v, k_s,h, k_s,v and k_s,m were determined by experimental data and Mathcad 15.0.

3. Results and discussion

3.1 Optimization of fermentation conditions

3.1.1 ANOVA and model fitting. The effects of the four independent variables i.e., S/I (A), SRT (B), OLR (C) and T (D) on VFA/SCOD obtained from anaerobic FW fermentation were investigated using RSM. The CCD experiments with 30 runs were randomly conducted to minimize the negative effect of uncontrolled variables. The experimental and predicted responses of VFA/SCOD are presented in Table S2.† Obviously, the experimental values were close to the predicted values. A second-order polynomial equation for VFA/SCOD (Y) fitted in terms of an actual factor obtained by Design Expert 8.0 is shown in eqn (5).

Y = −579.24 + 94.43A + 26.77B + 23.76C + 10.73D − 0.61AB + 0.35AC + 0.22AD − 0.26BC + 0.05BD + 0.05CD − 9.76A² − 1.58B² − 1.50C² − 0.15D² (5)

ANOVA was used for the regression analysis of the experimental data and the response surfaces, which is shown in Table S3.† And the further analyses and model fitting are described in section S2 (ESI).†

3.1.2 Response surface plots. The four independent variables (S/I, SRT, OLR and T) and the effects of their interactions on VFA/SCOD against any two variables while holding the other factors at the zero level were investigated to determine the optimum operation conditions by plotting 3D response surfaces (Fig. 1). Each response surface plot showed a clear peak, suggesting that the optimum condition was inside the design boundary. The optimum conditions for the four variations obtained using the “point optimization” tool of Design Expert 8.0 for maximum VFA/SCOD were as follows: S/I = 5, SRT = 7 d, OLR = 9 g VS per L per d and T = 40 °C, which were reasonable and similar to those in the previous literature. It has been reported that a SRT shorter than 8 d was favorable for acidogenic bacteria while longer than 8 d was beneficial for methanogens.²⁰ Lim et al.²¹ and Chen et al.¹² showed similar OLRs (9 and 8.31 g VS per L per d, respectively) and variation trends to this paper, i.e. VFA concentrations increased with an OLR increment, whereas a higher OLR led to an unstable operating reactor for the viscous fermentation broth. Jiang et al.⁴ reported that the maximal VFA/SCOD (82.6%) could be achieved at a fermentation temperature of 45 °C, which was in favour of the efficient solubilization and acidification of FW, whereas the VFA/SCOD obtained at 35 °C and 55 °C was only 75.6% and 26.4%, respectively. Improper fermentation temperature could suppress the growth of hydrolytic and acidogenic bacteria and the relevant enzymatic activities. In addition, it is worth noting that the S/I used in this study not only provided microbial biomass but affected the pH in the fermentation reactors. Thus, the optimum S/I obtained in this paper might be different to others, and detailed discussions will be given in the following section.

Fig. 1 Three-dimensional response surface plots for VFA/SCOD. Effects of (a) S/I and SRT at OLR = 8 g VS per L per d and T = 40 °C; (b) S/I and OLR at SRT = 7 d and T = 40 °C; (c) S/I and T at SRT = 7 d and OLR = 8 g VS per L per d; (d) SRT and OLR at S/I = 5 and T = 40 °C; (e) OLR and T at S/I = 5 and SRT = 7 d; and (f) SRT and T at S/I = 5 and OLR = 8 g VS per L per d.

3.1.3 Verification of the model. The fermentation experiments under the above optimum conditions were performed in triplicate, and the average value of VFA/SCOD was 88.65%, which was in good agreement with the predicted value (89.79%). The maximum VFA/SCOD was compared with some semi-continuous tests (Table 2), which applied other operation parameters or treatment methods. As seen in Table 2, Zhang et al.²² obtained a higher VFA/SCOD (90.36%), which was mainly attributed to the constant pH control of 6.5 by manually adding NaOH/HCl every day. In this way, the cost of adding chemical agents would hinder its scale-up application. Although the VFA/SCOD (88.65%) obtained in this study ranked a bit lower, the advantage was that the self-adjusting pH enabled this method to be cost-effective. Additionally, it was noted that the optimum conditions used in this study were more advantageous in gaining better VFA production and a higher VFA/SCOD compared with the fermentation conditions in other literature.

Table 2 Comparison of VFA/SCOD with those in other relevant literature

Substrates	Operation conditions	pH	VFA/SCOD	Reference
FW	SRT = 5 d, OLR = 11 g TS per L d, 35 °C	6	72.8%	4
FW	S/I = 4, 30 °C	6	50–70%	5
FW	SRT = 9 d, 35 °C	9	61.0%	8
FW	SRT = 8 d, OLR = 9 g TS per L d, 35 °C	6	63.29%	21
FW	S/I = 9, SRT = 8 d, OLR = 4 g VS per L per d, 35 °C	6.5	90.36%	22
FW	S/I = 5, SRT = 7 d, OLR = 9 g VS per L per d, 40 °C	—	88.65%	This study

---

3.1.4 EEM fluorescence spectroscopy analysis of soluble substrates. The composition of the dissolved organic matter (DOM) plays a significant role in the microbial growth and VFA production. EEM can provide an overall view of the fluorescent properties of the DOM over a selected spectral range using the locations and intensities of the fluorescence peaks.¹⁴ Thus, EEM analysis was introduced to further investigate the component variations of the soluble substrates and clearly reveal the effects of FW hydrolysis and utilization. The EEM fluorescence spectra of the initial supernatant before fermentation and the fermentation liquid after 20 d of operation are shown in Fig. 2. And the variations of the peak locations and intensities are shown in Table S5.† Three main fluorescence peaks marked as Peak A, B and C were observed, which were associated with a tryptophan-like protein (moderate biodegradability), a soluble microbial by-product (high biodegradability) and humic-acid like organics (low biodegradability), respectively.^23,24 It can be seen from Fig. 2 and Table S5† that the specific fluorescent intensities of Peak A, B and C were strengthened after 20 d of operation compared with those in the initial supernatants, especially in 3#, indicating that many fluorescence materials such as protein-like and humic-like components and so on were dissolved out by anaerobic fermentation. The shifts of the peak location are always associated with the variations in soluble organic matter in the fermentation liquid.²⁵ Previous literature reported that a red shift was connected with the solubilization of a complex substrate,^24,26,27 while a blue shift was mainly attributed to the hydrolysis or degradation of organic matter.^25,26 The fluorescent intensities of Peak A and B in 3# were enhanced by 118.3% and 159.6%, respectively, while Peak C only increased by 59.4%, indicating that the substrates in 3# produced an abundance of highly and moderately biodegradable organic matter. Additionally, Peak B in 3# had an obvious blue shift, while Peak A and C appeared to have the tendencies of red shifts, suggesting that 3# not only obtained a higher hydrolysis degree of biodegradable organics but also significantly enhanced the solubilization of substrates with low biodegradability. Therefore, considerable soluble organic fragments were released into the fermentation liquid of 3# including carboxyl, amino, hydroxyl and alkoxy groups,²⁷ which provided enough substrates for more VFA production. In addition, it was noted that all of the fluorescent intensities in 1# and 2# were much lower than those in 3#, and, except Peak B in 1# which showed a blue shift tendency, all of the peaks appeared to have a slight red shift. This phenomenon demonstrated that the substrates in 1# and 2# obtained incomplete dissolution and an insufficient number of biodegradable organics, which thus led to lower hydrolysis efficiencies.

Fig. 2 EEM fluorescence spectroscopy analysis of the DOM.

3.1.5 The production and composition of VFAs. VFA/SCOD, VFA production and VFA composition were often used to represent the degree of hydrolysis and acidification.¹¹ It was observed that the VFA/SCOD in 1#, 2# and 3# was 20.32%, 14.65% and 88.65%, and the corresponding VFA production was 314.57, 168.38 and 867.42 mg COD per g of VS, respectively. Moreover, the variation of VFA production with fermentation time was presented in Fig. S2† and the relevant analyses are shown in S3 (ESI).† Obviously, both the VFA/SCOD and VFA production in 3# were much higher than in 1# and 2#. The different acidic abilities were mainly attributed to the different pH values, which directly influenced the hydrolysis and acidification efficiencies. The reason for this will be further investigated in the following section.

The detectable VFAs mainly included acetic, propionic, iso-butyric, n-butyric, iso-valeric and n-valeric acids, and the individual VFA concentrations and percentages are shown in Fig. 3. It can be seen that acetic acid was the most predominant VFA (31.52%, 43.43% and 34.42% for 1#, 2# and 3#, respectively, and the corresponding concentrations were 99.15, 73.13 and 298.556 mg COD per g of VS, respectively), and propionic acid was ranked as the second (22–30%), followed by n-butyric acid, which was similar to the result of Chen et al.⁹ However, Feng et al.²⁸ reported that the top three VFAs were propionic, acetic, and n-butyric acids, in that order. This different observation might be attributed to their higher pH (6.0–9.0), which could promote the conversion of acetic and butyric acids to propionic acid.²⁹ In addition, the SRT of 7 d in this study also gave a positive influence on the acetic acid accumulation but had a negative effect on that of propionic acid, meaning that plenty of propionic acid could be converted to acetic acid by acetogenic bacteria in the continuous fermentation with a SRT of 7 d.¹³ Moreover, the concentrations and percentages of propionic acid increased while those of acetic acid decreased in 1# (sole FW) and 3# (S/I = 5) compared with 2# (sole WAS). This might be due to the higher carbohydrate/protein ratio in FW than in WAS. Carbohydrate acidification mainly produced propionic acid, while acetic acid formed the largest fraction of the products of protein fermentation in FW and WAS treatment.⁹ The higher protein concentration in 2# also contributed to the higher percentages of iso-valeric and n-valeric acids, which were mainly associated with protein fermentation by deamination of single amino acids or oxidation–reduction of a pair of amino acids.³⁰

Fig. 3 Comparison of individual VFAs in 1#, 2# and 3#: (a) concentrations, and (b) percentages.

3.1.6 The production of methane. It is well known that the generated VFAs can be further converted to methane by methanogens during anaerobic fermentation.^3,9,11 As reported by Rittmann et al.,³¹ 1 g of BOD could generate 0.35 L of methane at standard temperature and pressure. The fermentation substrate in 3# had an initial BOD concentration of 19.684 g L⁻¹, and the theoretical methane production should be 3.43 L (0.35 L × 19.684 g L⁻¹ × 0.5 L = 3.43 L) if the BOD was fully utilized. But in fact, the average methane production was 183.19 mL, which only accounted for 5.34% of the theoretical methane production. Therefore, only a small part of the VFAs was consumed by methanogens to produce methane, and more generated VFAs were accumulated. However, for 2#, a relatively higher methane production of 480.16 mL was achieved, while a relatively lower VFA production (168.38 mg COD per g ofVS) was obtained compared with that in 3# (867.42 mg COD per g of VS), indicating that a high methanogen activity in inoculum had a negative effect on VFA production. Therefore, the suppression of methanogens in inoculum was of great significance. It was also noted that 1# represented low VFA and methane production simultaneously, along with a low pH range of 3.7–4.8 during the whole fermentation process (Fig. 4). There were two likely reasons. On the one hand, too much FW accumulation and insufficient microorganisms caused food overloading, which slowed down the hydrolysis and acidogenesis processes.³² On the other hand, the low pH inhibited the growth and activity of microorganisms, and prevented the further dissolution of organic matter and the subsequent correlative reactions.³³

Fig. 4 Carbon balance analysis of the fermentation substrates.

In fact, the methane production obtained in the three reactors was far below the theoretical values (Table 3). In the literature, Luo et al.¹³ and Miron et al.²⁰ reported that a SRT shorter than 8 d was favourable for the acidogenic bacteria, whereas a SRT that exceeded 8 d would enhance the activities of the methanogens. Thus the low methane production could be partly attributed to the opportune SRT of 7 d. And as a main inhibitor in aceticlastic methanogenesis (one of the main methanogenic pathways), the high acetic acid concentrations obtained in this study, especially in 3#, also led to low methane production.² In addition, the methane production was significantly influenced by the pH variations. The pH drop caused by the rapid VFA accumulation could directly lead to the activity loss in acid-sensitive glycolytic enzymes and the inhibition of methanogen, but a pH value that was too low was also unfavorable for acidification bacteria.³⁴ Therefore, the different values for VFA and methane production were also greatly attributed to the different pH levels in the three reactors.

Table 3 The comparison of the methane production in the three reactors

Parameters	1#	2#	3#
Methane production (mL per g of VS)	34.77	53.89	20.56
Practical methane yield (mL)	309.80	480.16	183.19
Theoretical methane yield (L)	3.47	1.89	3.43
Practical/theoretical (%)	8.93%	25.41%	5.34%

---

3.1.7 Kinetics analysis of hydrolysis, acidification and methanation. The optimized kinetic parameters in the three fermentation stages are shown in Table 4. The hydrolysis rate (k) of 0.0667 d⁻¹ in 3# was higher than the value (0.058 d⁻¹) reported by Xu et al.,¹⁶ who used a similar S/I to this study, indicating that the optimal conditions used in the study were more beneficial for improving the hydrolysis efficiency. Firstly, the suitable OLR of 9 g VS per L per d provided a stable fermentation performance,³⁵ and the opportune fermentation T of 40 °C could effectively enhance the activities of the hydrolytic bacteria and relevant enzymes, while at the same time the SRT of 7 d provided sufficient time for the degradation of particular substrates into dissolved forms and the conversion of complex VS into simple compounds.²¹ Moreover, it was noticed that the hydrolysis rate (k) and the maximum utilization rate of the hydrolysis products (k_h) in 3# were higher than the values obtained for 1# and 2#, suggesting that the S/I of 5 played a significant role both in enhancing the hydrolysis of particular organic matter and in the conversion of soluble hydrolysis products to VFAs. The opportune microbial biomass provided by the S/I of 5 prevented the excessive accumulation of FW and made plenty of solid organics further dissolve into liquid. Additionally, the appropriate pH range for the hydrolytic–acidogenic bacteria and relevant enzymes caused by the S/I of 5 was also closely related to the higher hydrolysis and acidification efficiencies. It was also noted that relatively low maximum utilization rates of the VFAs (k_v) were found in all three reactors especially in 3#, which was in accordance with the practical methane production (Table 3), and the reasons for this have been discussed in detail in 3.1.6. Overall, the high VFA production in 3# could be well explained by the higher k and k_h values, and the lower k_v value.

Table 4 Kinetic parameters of anaerobic fermentation

Parameters	1#	2#	3#
k (d^−1)	0.0329	0.0161	0.0667
kh (g SCOD d^−1)	4.25	2.68	11.65
kv (g VFA d^−1)	0.76	1.19	0.37

---

3.1.8 Carbon mass balance analysis. Based on the experimental results, the carbon mass balance analyses were given as a COD mass to further understand the carbon conversion efficiencies, and the variations and contributions of different organic materials could also be clearly revealed. In general, the carbon in FW would be converted into VFAs, protein, carbohydrate, lactic acid, alcohol, methane and carbon dioxide by anaerobic fermentation.¹¹ To simplify the calculation, the carbon content in the soluble protein, carbohydrate, VFAs and methane were determined. The other carbon-containing compounds, such as lactic acid, alcohol and carbon dioxide were referred to as “other”. The carbon mass balance analyses of the three reactors after 20 d of operation were conducted, and the results are shown in Fig. 4.

The COD masses represented by carbohydrate, “other” and VFAs were dominant in 1#, 2# and 3#, respectively. And the COD masses of “other” formed considerable proportions both in 1# and 2# (31.35% and 62.77% of the total carbon, respectively), and were simultaneously accompanied by the relatively higher methane masses and lower VFA masses compared with those in 3#. Additionally, the carbohydrate mass in 1# and protein mass in 2# also showed greater percentages than in 3#, and were 34.23% and 11.31% of the total carbon, respectively. These data clearly indicated that the fermentation substrates in 1# and 2# were not adequately utilized to produce VFAs. However, the COD masses of “other” and methane in 3# were only 0.57 g and 0.32 g, which was consistent with the negligible methane production reported above. And the COD mass from carbohydrates and proteins only accounted for a small part of the total COD mass compared with the VFA mass. Meanwhile the VFA mass increased significantly and reached the highest value (7.73 g) in 3# compared with 1# (2.83 g) and 2# (1.50 g). Therefore, the optimum S/I of 5 combined with the other optimized conditions (SRT = 7 d, OLR = 9 g VS per L per d and T = 40 °C) was conducive for the sufficient utilization of particular organic matter and the conversion of hydrolysis products into VFAs.

3.2 Discussion

The pH plays a crucial role in VFA production due to its effects not only on the anaerobic bacteria community structures, activities and growth rates, but also on the metabolic pathways.³² VFA accumulation would affect the pH in the fermentation liquid during anaerobic fermentation, and the pH variation directly influenced the hydrolysis, acidification and methane generation. Veeken et al.³⁶ reported that the hydrolysis of bio-waste was inhibited at pH 5.0, and Wang et al.¹¹ observed that the activity of acidogenic bacteria was suppressed at pH values below 4.0. Some previous studies reported that the optimal pH for effective hydrolysis and suppressive methanogenesis was 5.0–6.5,^11,16 while a higher pH (exceeding 6.5) could bring about high methanogen activity.³⁷ As shown in Fig. 5, fast drops of pH in the starting phase caused by the rapid acidification of the substrate were observed in all reactors, which meanwhile generated a certain inhibitory effect on the methanogens.² This was followed by a slight increase and then a stabilization due to the neutralization of the fresh substrate and ammonia release. The pH values in 1#, 2# and 3# were not adjusted artificially during the whole fermentation process, and were maintained in ranges of 3.7–4.8, 6.1–6.9, and 5.2–6.4, respectively. This phenomenon was mainly attributed to the buffering capacity of the system itself caused by the released ammonia.³⁸

Fig. 5 The variations of pH and ammonia.

The variation of ammonia concentration during anaerobic fermentation are also shown in Fig. 5. It can be seen that there was a significant release of ammonia together with the production of VFAs, and the ammonia concentration was 240–580, 740–1310, and 820–1190 mg L⁻¹ in 1#, 2# and 3#, respectively. The difference of ammonia concentration in the different reactors was in line with the composition of the substrate, which was produced by the biological degradation of proteins and amino acids.¹¹ The high ammonia level in 2# was attributed to the fact that WAS was a material rich in protein as opposed to FW (1#), which was principally composed of carbohydrate. It should be noted that the variation of pH was closely connected with the variation of ammonia concentration. The release of ammonia provided a high alkalinity, which could compensate for the decreased pH.²⁵ Particularly, during the whole fermentation, the pH in 3# fluctuated between 5.2 and 6.4, a range that was mentioned above to be optimum for hydrolysis and acidification, and inhibiting for methanogenesis, resulting in a high-efficiency hydrolysis, higher VFA production and minimum methane production compared with 1# and 2# (Table 3). The favorable pH range in 3# was mainly attributed to the optimum S/I of 5, which provided an opportune amount of ammonia to form a high buffering capacity. At the same time, it was noted from Table 3 together with Fig. 5 that the pH in 1# was less than 5.0, and the particulate substrates could not be thoroughly dissolved out and hydrolyzed, which resulted in low VFA and methane production. In addition, the vast release of ammonia in 2# resulted in a relatively higher pH range (6.1–6.9), which encouraged the production of methane through methanation and presented a low VFA production. Therefore, it can be concluded that the optimum proportion of FW and WAS can provide a suitable pH range in the semi-continuous fermentation reactor, which is favorable for hydrolysis and acidification and inhibiting for methane generation, thus contributing to a higher VFA production.

4. Conclusions

A high VFA production was obtained from semi-continuous anaerobic FW fermentation without the artificial control of pH , which provided a cost-effective VFA production strategy for practical continuous production. The optimum conditions were S/I = 5, T = 40 °C, SRT = 7 d and OLR = 9 g VS per L per d, and a high VFA/SCOD (88.65%) and VFA production (867.42 mg COD per g of VS) were achieved. The high VFA production was mainly attributed to the advantageous pH (5.2–6.4) obtained by the buffering capacity of the fermentation system, which was favorable for the hydrolysis and acidogenesis of substrates, and inhibiting for methanogenesis. The main constituents of the VFAs were acetic, propionic and n-butyric acids.

Acknowledgements

This work was financially supported by the National Key Technology Support Program (2014BAD02B03) and National Nature Science Foundation of China (51121062). The authors also gratefully acknowledge the financial support by State Key Laboratory of Urban Water Resource and Environment (2014TS06), the Department of Education Fund for Doctoral Tutor (20122302110054), the Harbin Institute of Technology Fund for young top-notch talent teachers (AUGA5710052514).

References

1. B. Grizzetti, U. Pretato, L. Lassaletta, G. Billen and J. Garnier, Environ. Sci. Policy, 2013, 33, 186–195.
2. Z. Y. Xu, M. X. Zhao, H. F. Miao, Z. X. Huang, S. M. Gao and W. Q. Ruan, Bioresour. Technol., 2014, 163, 186–192.
3. W. Q. Guo, Q. L. Wu, S. S. Yang, H. C. Luo, S. M. Peng and N. Q. Ren, RSC Adv., 2014, 4, 53321–53326.
4. J. G. Jiang, Y. J. Zhang, K. M. Li, C. Wang, X. Gong and M. L. Li, Bioresour. Technol., 2013, 143, 525–530.
5. J. Yin, K. Wang, Y. Q. Yang, D. S. Shen, M. Z. Wang and H. Mo, Bioresour. Technol., 2014, 171, 323–329.
6. P. Fontanille, V. Kumar, G. Christophe, R. Nouaille and C. Larroche, Bioresour. Technol., 2012, 114, 443–449.
7. Y. Jiang, Y. Chen and X. Zheng, Environ. Sci. Technol., 2009, 43, 7734–7741.
8. H. Chen, H. J. Meng, Z. C. Nie and M. M. Zhang, Bioresour. Technol., 2013, 128, 533–538.
9. Y. G. Chen, J. Y. Luo, Y. Y. Yan and L. Y. Feng, Appl. Energy, 2013, 102, 1197–1204.
10. H. U. Cho, Y. M. Kim, Y. N. Choi, H. G. Kim and J. M. Park, Bioresour. Technol., 2015, 191, 475–480.
11. K. Wang, J. Yin, D. S. Shen and N. Li, Bioresour. Technol., 2014, 161, 395–401.
12. H. Chen and H. Y. Wu, Bioresour. Technol., 2010, 101, 5487–5493.
13. J. Y. Luo, L. Y. Feng, W. Zhang, X. Li, H. Chen, D. B. Wang and Y. G. Chen, Appl. Energy, 2014, 113, 51–58.
14. X. C. Feng, W. Q. Guo, C. Chen, S. S. Yang, W. B. Jin and N. Q. Ren, Bioresour. Technol., 2013, 143, 642–646.
15. APHA, Standard methods for the examination of water and wastewater, American Public Health Association, Washington, DC, 20th edn, 2005.
16. S. Y. Xu, O. P. Karthikeyan, S. Ammaiyappan and W. C. W. Jonathan, Bioresour. Technol., 2012, 126, 425–430.
17. V. A. Vavilin, B. Fernandez, J. Palatsi and X. Flotats, Waste Manag., 2008, 28, 939–951.
18. P. Zhang, Y. G. Chen and Q. Zhou, Bioresour. Technol., 2010, 101, 6902–6909.
19. R. Borja, A. Martin, E. Sanchez, B. Rincon and F. Raposo, Process Biochem., 2005, 40, 1841–1847.
20. Y. Miron, G. Zeeman, J. B. van Lier and G. Lettinga, Water Res., 2000, 34, 1705–1713.
21. S. J. Lim, B. J. Kim, C. M. Jeong, J. d. r. Choi, Y. H. Ahn and H. N. Chang, Bioresour. Technol., 2008, 99, 7866–7874.
22. M. M. Zhang, H. Y. Wu and H. Chen, Process Saf. Environ. Prot., 2014, 92, 171–178.
23. B. Yu, Z. Y. Lou, D. L. Zhang, A. D. Shan, H. P. Yuan, N. W. Zhu and K. H. Zhang, Bioresour. Technol., 2015, 179, 291–298.
24. G. P. Sheng and H. Q. Yu, Water Res., 2006, 40, 1233–1239.
25. X. F. Huang, C. M. Shen, J. Liu and L. J. Lu, Chem. Eng. J., 2015, 264, 280–290.
26. L. Pang, J. R. Ni and X. Y. Tang, Biochem. Eng. J., 2014, 86, 49–56.
27. N. Senesi, Anal. Chim. Acta, 1990, 232, 77–106.
28. L. Y. Feng, Y. G. Chen and X. Zhang, Environ. Sci. Technol., 2009, 43, 4373–4380.
29. F. R. Hawkes, R. Dinsdale, D. L. Hawkes and I. Hussy, Int. J. Hydrogen Energy, 2002, 27, 1339–1347.
30. D. J. Batstone, J. Keller, I. Angelidaki, S. Kalyuzhnyi, S. Pavlostathis, A. Rozzi, W. Sanders, H. Siegrist and V. Vavilin, Water Sci. Technol., 2002, 45, 65–73.
31. B. E. Rittmann and P. L. McCarty, Environmental Biotechnology: Principles and Applications, McGraw-Hill Book Co, New York, 2001.
32. S. Dechrugsa, D. Kantachote and S. Chaiprapat, Bioresour. Technol., 2013, 146, 101–108.
33. J. M. Pan, X. Chen, K. C. Sheng, Y. H. Yu, C. X. Zhang and Y. B. Ying, Hydrogen Energ., 2013, 38, 12747–12754.
34. H. Bouallagui, Y. Touhami, R. Ben Cheikh and M. Hamdi, Process Biochem., 2005, 40, 989–995.
35. M. Jabeen, Zeshan, S. Yousaf, M. R. Haider and R. N. Malik, Int. Biodeterior. Biodegrad., 2015, 102, 149–153.
36. A. Veeken, S. Kalyuzhnyi, H. Scharff and B. Hamelers, J. Environ. Eng., 2000, 126, 1076–1081.
37. H. Y. Yuan, Y. G. Chen, H. X. Zhang, S. Jiang, Q. Zhou and G. W. Gu, Environ. Sci. Technol., 2006, 40, 2025–2029.
38. L. S. Cadavid-Rodríguez and N. J. Horan, Water Res., 2014, 60, 242–249.

---

Footnote

† Electronic supplementary information (ESI) available: This file contains the statistical analysis of RSM, ANOVA, model fitting and the relevant data and tables. See DOI: 10.1039/c5ra21162j

---