Energy Conversion and Management 88 (2014) 12191227

Contents lists available at ScienceDirect

Energy Conversion and Management

journal homepage: www.elsevier.com/locate/enconman

Integration of fermentative biohydrogen with methanogenesis

from fruitvegetable waste using different pre-treatments
Xuan Jia a,b, Mingxiao Li a,b, Beidou Xi a,b, Chaowei Zhu b,, Yang Yang a, Tianming Xia a, Caihong Song a,
Hongwei Pan a,b
a
Innovation Base of Groundwater and Environmental System Engineering, Chinese Research Academy of Environmental Science, Beijing 100012, China
b
State Key Laboratory of Environmental Criteria and Risk Assessment, Chinese Research Academy of Environmental Sciences, Beijing 100012, China

a r t i c l e i n f o a b s t r a c t

Article history: Fruitvegetable waste was subjected to three different pre-treatments to enrich the two-stage biofuel
Available online 28 February 2014 production potential. The uorescence excitationemission matrix (EEM) spectra coupled with parallel
factor (PARAFAC) analysis and uorescence regional integration analysis were utilised to investigate dis-
Keywords: solved organic matter degradation during two-stage fermentation process. The results showed that com-
Fruitvegetable waste pared with that of alkali and enzyme pre-treatments, the acid pre-treatment resulted in the maximum
Biohydrogen biogas production rates and proportion in the hydrogenogenic stage (10.11 mL/h, 41.2% hydrogen) when
Two-stage biofuel production
combined with the methanogenic process (4.67 mL/h, 76.1% methane). In addition, the analysis of soluble
Pre-treatment
Excitationemission matrix
metabolites composition indicated that both ethanol- and butyric acid-type fermentation processes had
Parallel factor analysis taken place as a result of acid pre-treatment, whereas only butyric acid-type fermentation resulted from
alkali and enzyme pre-treatments. The PARAFAC analysis modelling of the EEM spectra revealed three
uorescent components in the efuents of three fermentation stages and assumed that the projected
characteristic value may be used as a rapidly obtained indicator for substrate degradation and system sta-
bility of a two-stage biofuel production process.
2014 Elsevier Ltd. All rights reserved.

1. Introduction
Fruitvegetable waste (FVW) is the most abundant organic
resource produced from food markets and households. Reports
have estimated that the annual yield of FVW is more than 100 mil-
lion tons in China [1]. However, FVW treatment options are limited
in China due to regulations and the availability of suitable tech-
niques. Concerns over vermin attraction, odour, leachate produc-
tion and greenhouse gas emissions have led to many highly
putrescible waste streams, such as FVW being disposed to landlls
[2]. Due to its high biogas potential, fermentative hydrogen pro-
duction is a promising energy-saving and energy-producing pro-
cess for pre-treating and degrading highly putrescible waste
streams, such as FVW [35].
Sequential hydrogen and methane processes are more efcient
than independent hydrogen processes. In the former, hydrogen can
be recovered during hydrolysis and the decomposition of complex
substrates, such as proteins, carbohydrates and lipids, into smaller
Corresponding author. Address: No. 8, Dayangfang, Beiyuan Road, Beijing
10012, China. Tel./fax: +86 10 84937789.
E-mail address: zhucw@craes.org.cn (C. Zhu).
units. The latter require longer hydraulic detention times and pro-
duce methane [6]. In a two-stage process, the residual substrates
from the rst stage can be reduced at the second-stage to yield
an energy conversion ratio of 89% [7].
Dissolved organic matter (DOM) is the most active component
in the anaerobic fermentation process, and its chemical and struc-
tural characteristics are most likely to affect its biodegradation [8].
As a simple, sensitive and non-destructive technique, uorescence
excitationemission matrix (EEM) spectroscopy is often used to
trace the composition and biogeochemical cycling of DOM [9,10].
However, the EEM spectra of DOMs in fermentation efuent are
composed of various types of overlapping uorophores, which
can be very difcult to interpret. Parallel factor analysis (PARAFAC)
can decompose uorescence signals into underlying uorescent
phenomena and accurately quantify them by uorescence regional
integration (FRI) analysis [11]. Guo et al. [8] observed the changes
in the uorescent components that occur in the DOM of a swine
fermentation slurry through uorescence spectroscopy using a
PARAFAC analysis. Thus, the combination of EEMs and PARAFAC
is a powerful tool in the assessment of DOM dynamics during
the two-stage anaerobic fermentation process. Most of the chal-
lenges involved in developing a two-stage fermentation process

http://dx.doi.org/10.1016/j.enconman.2014.02.015
0196-8904/ 2014 Elsevier Ltd. All rights reserved.
1220 X. Jia et al. / Energy Conversion and Management 88 (2014) 12191227

concern the initial hydrogen-producing stage. Additional pre-treat-
ment processes are necessary to improve the availability of carbo-
hydrates from FVW to hydrogenogens. To date, the effects of
various pre-treatment methods have been systematically investi-
gated [1214]. In contrast, few studies have analysed the metabolic
mechanism of FVW in response to different pre-treatments using a
combination of EEM and PARAFAC analysis.
The objectives of this study are evaluated three pre-treatments
for FVW by the biogas energy productivity, soluble metabolites
characteristic and present an integrated substrate degradation
and stability evaluation method based the structural characteris-
tics of DOM using EEM spectra with PARAFAC analysis. The kinetics
of the two-stage process were analysed to determine important
parameters, such as the maximum hydrogen/methane production
rate, lag time during the process and varied soluble metabolite
composition in response to pre-treatment.
out by varying the pre-treatment methods. Batch experiments
were conducted using pretreated food waste as substrates (initial
loading rate at 15 g VS/L) and performed in triplicate in 500-mL
serum bottles. 25 mL PADRs were added to each bottle. The work-
ing volume was adjusted to 300 mL using distilled water. The bot-
tles headspace was purged with nitrogen gas to provide anaerobic
conditions. The serum bottles were placed in water bath with its
vibrator rotating at 150 rpm at 37 1 C to provide better contact
among substrates. Control bottles were also prepared using the
FVW without any pre-treatment at the same time.
Each experiment included two stages, named hydrogenogenic
stage (87 h) and methanogenic stage (300 h). In the hydrogenogen-
ic stage, the initial pH was adjusted to 6.0 using 2 M NaOH or HCl.
In the methanogenic stage (after 87 h, when no hydrogen produc-
tion), the pH was adjusted to 7.5, and there was no need to adjust
the pH condition during the following experiment process.

2. Materials and methods 2.4. Analytical methods

2.1. Materials The TS, VS and pH were determined according to standard

methods [15]. The total gas production was measured by the dis-
FVW was collected from a market in Beijing, China. The charac- placement of saturated brine solutions. The composition of the bio-
teristics of FVW are shown in Table 1. The FVW was composed of gas (H2, CH4 and CO2) in the reactors headspace was analysed
(w/w basis) lettuces (40%), lemon (40%) and grape (20%). The larger using a gas chromatograph (GC) (Perkin Elmer Clarus 500, New Jer-
chunks of FVW were chopped into small pieces measuring approx- sey, USA) equipped with a thermal conductivity detector (TCD) and
imately less than 5 mm in length and width. a 2-m high-porosity polymer bead-packed column.
The volatile fatty acids (VFAs) and ethanol concentration were
determined using a GC equipped with a ame ionisation detector
2.2. Seed microora
(FID) and a 30 m  0.25 mm  0.25 mm fused-silica capillary col-
umn (Agilent DB-VRX). Helium was used as the carrier gas at a ow
The microora (piggery anaerobic digested residues, PADRs)
rate of 1.2 mL/min and a split to a column ow ratio of 10:1. The
were enriched from an anaerobic reactor used for treating pig
injection temperature was 200 C. The oven temperature was ini-
manure. Before being used, the PADRs were diluted by an equal
tially set to 40 C with a holding time of one minute; the temper-
volume of distilled water and then sieved through a 100-mesh
ature was increased to 220 C thereafter at a rate of 9 C per
sieve to remove stones, sand, and other coarse matter. The charac-
minute.
teristics of the PADRs were as follows: a pH of 7.8, a total COD
(TCOD) of 59.09 g/L, a total solids (TS) content of 21.2%, a volatile
2.5. Model analysis
solids (VS) content of 9.95%, a suspended solids (SS) content of
19.1%, a volatile suspended solids (VSS) content of 9.15% and a
The cumulative hydrogen or methane production during the
gravimetric moisture content of 78.8%.
batch experiments followed the modied Gompertz equation
[16,17]:
2.3. Experimental design   
Rm e
H P exp k  t 1 1
Acid and alkali pre-treatments were performed by mixing P
10.0 g dry-weight FVW with 50 mL of dilute HCl (or NaOH) aque-
ous solution at 0.25% (or 1.0% (w/v)) concentration, and mixed where H is the cumulative hydrogen/methane production (mL), P is
for 24 h at 37 1 C in serum bottles. 10.0 g dry-weight FVW were the hydrogen/methane production potential (mL), Rm is the maxi-
mixed with 50 mL cellulase R-10 (Yakult, Japan) aqueous solution mum hydrogen/methane production rate (mL/h), e is 2.72, k is the
at 10 mg/L concentration and soaked for 48 h in serum bottles. lag-phase time (h) and t is the incubation time (h). The correspond-
The bottles were placed in an orbital shaker at 48 1 C since the ing values of P, Rm and k for each batch were estimated using Origin
optimal working condition for cellulase R-10 is: the pH level of 7.5, a scientic graphing and data analysis software program.
4.56.5 and the temperature of 4560 C. The mixture was then
neutralised to pH 6.0 by the addition of 2.0 M NaOH (or HCl) after 2.6. EEM spectra scan
the pre-treatments.
The two-stage anaerobic fermentation of FVW that integrated Samples were collected from the initial fermentation broth after
the fermentation of biohydrogen and methanogenesis was carried different pre-treatments at the end of the hydrogenogenic and
methanogenic stages. The suspensions were centrifuged at
12,000 rpm for 10 min at 4 C and ltered through a 0.45-lm
Table 1 membrane lter. Before uorescence analysis was performed, the
The characteristics of FVW.
total organic carbon (TOC) was measured using a Shimadzu TOC-
Parameters Lettuces Lemon Grape Mixed FVW TNM analyser. All of the sample concentrations were similarly ad-
Moisture content (%) 93.32 83.92 86.61 88.95 justed to make them mutually comparable. The nal TOC content
Ash content (%) 18.34 3.42 1.40 9.84 was approximately 8 mg/L. Fluorescence EEM spectroscopy was
Lignin (%) 36.79 21.57 21.65 33.75 performed on each sample using a Hitachi F-7000 uorescence
Hemicellulose (%) 13.27 17.49 4.71 13.53
spectrophotometer (Hitachi, Tokyo, Japan) at room temperature.
Cellulose (%) 7.46 9.85 0.78 7.04
The slit widths were adjusted to 10 nm for both the excitation
 X. Jia et al. / Energy Conversion and Management 88 (2014) 12191227
and emission monochromators, and the scan speed was set to
1200 nm/min. The emission wavelength was varied from 200 to
500 nm, whereas the excitation wavelength increased from 200
to 450 nm in increments of 5 nm. After the EEM spectra scan were
obtained, a method for handling scattering using interpolation in
the areas affected by rst- and second-order Rayleigh and Raman
scatter was used [18].
2.7. PARAFAC analysis
PARAFAC is a statistical tool used to decompose three-way data
into individual components [19]. In this study, all EEM spectra data
comprise a three-way array X featuring components I  J  K. The
three-way array X can be decomposed into tri-linear terms and a
residual array using the following expression:
X
F
X ijk cif bif akf rijk i 1; . . . ; I; j 1; . . . ; J; k 1; . . . ; K
f the fraction of methane was exactly 89.0% in the methanogenic
In this application, F denes the number of components; Xijk is
the uorescence intensity of sample i at emission wavelength j and
excitation wavelength k; cif is directly proportional to the concen-
tration of the uorophore component f in sample i (dened as
scores); and bjf and akf are estimates of the emission and excitation
spectra, respectively, for uorophore component f (dened as load-
ings). F is the number of components (individual uorophore moi-
eties) and rijk is the error term representing the variability not
accounted for by the model.
The standard PARAFAC algorithm, which is based on the inter-
active least-squares algorithm, minimises the sum of squared
residuals [20]. In addition, the scores from the rst loading of the
PARAFAC model were obtained. These scores were based on the
signal intensities of the components present in each sample and
expressed as Fmax values. The utility of using PARAFAC with EEM
data is that individual uorophores moieties can be identied
and their relative concentrations quantied in complex mixtures
if the correct number of components are chosen and outliers are
rst identied and then removed from the dataset.
3. Results and discussion
3.1. Performance of two-stage fermentation process
Various pre-treatments of FVW were performed to evaluate
their efciency. Fig. 1 illustrates the inuence of different pre-
treatments on the cumulative biogas production and proportion
of hydrogen and methane in the two-stage fermentation process.
In the hydrogenogenic stage, the three (acid, alkali and enzyme)
pre-treatments applied to the FVW resulted in higher biogas yields
and hydrogen proportions compared to those of the control
(Fig. 1a). Notably, the efciency of the cumulative biogas produc-
tion of the acid pre-treatment was the highest at 730 mL, more
than ve times that of the control (125 mL). The alkali and enzyme
pre-treatments resulted in biogas production volumes of 300 and
580 mL, respectively.
The main constituents of the biogas generated during the
hydrogenogenic stage were hydrogen and carbon dioxide. The re-
sults revealed that the maximum hydrogen proportion was 41.2%
using the acid pre-treatment. Reactors that contained alkaline
and enzyme pre-treatments of FVW achieved hydrogen propor-
tions of 16.4% and 38.1%, respectively.
After 87 h of operation, the pH was adjusted to 7.5 during the
methanogenic stage. In this process, the highest cumulative biogas
production was calculated to have been 1290 mL when the FVW
was pre-treated with acid. A similar biogas production volume
was observed for the alkali treatment (1230 mL), which indicated
1221
that these methods could signicantly improve the biogas produc-
tion beyond that achieved by the control (450 mL) (Fig. 1). The
main constituents of the generated biogas were methane and car-
bon dioxide, with a small amount of hydrogen arising from some
treatments. Fig. 1b presents the effects of various pre-treatments
on the methane proportion when the initial pH was changed to
7.5 during the methanogenic process. The methane fractions in
the three pre-treatment reactors were higher than the methane
fraction of the control (18.4%). The results indicate that the maxi-
mum methane proportion was 89.0% using the alkali treatment
method. The methane proportions in the total biogas were 76.1%
and 80.7% in response to acid and enzyme pre-treatments,
respectively.
When examining the overall performance, the acid pre-treat-
ments in the hydrogenogenic stage resulted in the most efcient
biogas production (730 mL, 41.2% hydrogen) when combined with
2 the methanogenic process (1290 mL, 76.1% methane). However,
process during the two-stage operation after alkali pre-treatment.
This result indicates that the pre-treatments could promote the
fermentation potential of FVW, which was attributed to the in-
creased transfer efciency of organic matter from the particulate
fraction to the soluble fraction in the substance [14].
3.2. Kinetic analysis
A kinetic analysis was performed to illustrate the time-course
regularities of the experimental data (Fig. 2 and Table 2). The re-
sults of this analysis indicate that the cumulative hydrogen and
methane production in response to different pre-treatments condi-
tions could be well described by the modied Gompertz equation
for non-linear numerical estimations with an overall determina-
tion coefcient (R2) of 0.96. Table 2 presents important parameters,
such as the maximum hydrogen/methane production potential (P),
maximum hydrogen/methane production rate (Rm) and lag-phase
time (k).
In the hydrogenogenic stage, the maximum hydrogen produc-
tion potential of each of the three pre-treatments exceeded that
of the control. The acid pre-treatment yielded the maximum
hydrogen production potential of 180 mL, followed by 142
and19 mL for the enzyme and alkali pre-treatments, respectively.
The highest hydrogen production rates of the three individual
pre-treatment methods were 10.11, 3.64 and 5.65 mL/h for the
acid, alkali and enzyme pre-treatments, respectively. This produc-
tion rate was only 0.16 mL/h for the control. The kinetic analysis
illustrates that the hydrogen-producing behaviour performance
was improved in response to the acid pre-treatment. The lag phase
of the acid and alkali pre-treatments varied narrowly from 3.88 to
3.61 h and increased up to 4.54 h for the enzyme pre-treatments. A
minimum lag phase of approximately 0.38 h was observed for the
control. The results indicated that water-soluble byproducts such
as weak acids, furan derivatives and phenolic compounds are also
generated after acid pre-treatment, which negatively affected the
bacteria activity and extend lag phase. The alkali and enzyme
pre-treatments make the lag phase longer due to the change in
ideal growth conditions of hydrogenogenic bacteria [21].
In the methanogenic stage, the acid and alkali pre-treatments
performed similarly, with maximum methane production poten-
tials of 851 and 825 mL, respectively, whereas the production poten-
tial for the enzyme pre-treatment was 332 mL. The maximum
methane production potential was lowest for the control at 9 mL.
Methane production rates of 4.67, 6.63 and 2.63 mL/h were ob-
served for the acid, alkali and enzyme pre-treatments, respectively.
The lowest methane production rate of 0.15 mL/h was observed for
the control. In addition, the lag phases were even longer for the acid
and enzyme pre-treatments compared to the lag phase of the control
1222 X. Jia et al. / Energy Conversion and Management 88 (2014) 12191227

Cumulative biogas production(mL)

800 Control
700 Acid
Alkali
600 Enzyme
500

400

300

200

100

0
45
40
Hydrogen propotion (%)

35
30
25
20
15
10
5
0
0 20 40 60 80 100
Time (h)

(a)
1400
Cumulative biogas production(mL)

1300 Control
1200 Acid
1100 Alkali
1000 Enzyme
900
800
700
600
500
400
300
200
100
0
100
Methane propotion (%)

80

60

40

20

0
0 50 100 150 200 250 300
Time (h)

(b)
Fig. 1. The cumulative biogas production and proportion in response to different treatments (a) hydrogenogenic stage and (b) methanogenic stage.

(83.75 h) at 146.65 and 147.39 h, respectively. Nevertheless, the 3.3. Production of soluble metabolites
lag-phase shorten when bacteria enriched by pre-adaptation and
ultilization to inimical growth conditions after alkali pre-treatment Hydrogen formation is typically accompanied by the generation
during the methanogenic stage [22]. The lag time of the alkali treat- of VFAs and ethanol during anaerobic digestion processes. Hence,
ment (87.53 h) was similar to that of the control. the composition and concentration of the produced soluble metab-
The acid pre-treatment could enhance the hydrogen and olites are useful indicators for monitoring the hydrogen production
methane production potential, but did not clearly improve the process. The investigation of the soluble metabolites at the end of
maximum methane production rate or lag-phase time. The alkali the hydrogenogenic and methanogenic processes (Table 3 and
pre-treatment signicantly increased the maximum methane pro- Fig. 3) showed that the concentrations of VFAs and ethanol were
duction rate, with negligible increases in the methane production a function of the pre-treatments.
potential, indicating that methane was rapidly produced with a The total production of VFAs remained nearly unchanged at
smaller lag-phase time. 2407.7 and 2475.1 mg/L in response to the acid and enzyme
 X. Jia et al. / Energy Conversion and Management 88 (2014) 12191227 1223

200 Control product after the methanogenic stage. Guwy et al. [7] also reported
Acid that the efuent from dark fermentative biohydrogen production is
180
Cumulative hydrogen production (mL)

Alkali
typically high in volatile fatty acids, such as acetate and butyrate.
160 Enzyme
The majority of the soluble metabolites for the alkali and en-
140 zyme pre-treatments consisted of acetate and butyrate, as assessed
120 by GC (which together accounted for 99.4% and 98.7% of the total
soluble metabolites, respectively), indicating that a butyric acid-
100
type fermentation occurred in the hydrogenogenic stage [23]. For
80 the acid pre-treatments, the soluble metabolites consisted of etha-
60 nol, acetate and butyrate, which together accounted for 99.6% of
the total soluble metabolites. This composition indicated that both
40
ethanol- and butyric acid-type fermentation processes had oc-
20 curred in the hydrogenogenic systems. However, butyric acid fer-
0 mentation was the dominant metabolic pathway for the acid
pre-treatments. In addition, the control test resulted in higher con-
0 20 40 60 80 100 centrations of lactic acid (3.84 g/L) than those observe in all pre-
Time (h) treatments after the hydrogenogenic stage, which may threaten
(a) the anaerobic digestion efciency after the hydrogen production
process. During the methanogenic stage, the concentrations of sol-
900 uble metabolites, especially acetate and butyrate acids, decreased
Control
800
dramatically, whereas the propionic acid percentage was greater
Acid
Cumulative methane production (mL)

Alkali than that of the hydrogenogenic stage for all pre-treatments. Ren
700 Enzyme et al. [24] suggested that different pre-treatments would change
600 the metabolic pathway to the butyric acid type, mixed-acids type
or ethanol type.
500 We then attempted to correlate the hydrogen and methane pro-
400 duction with these metabolites ratios. The results indicate that the
hydrogen yield was directly proportional to the BU:AC ratio,
300
whereas the methane yield was proportional to the PR:AC ratio.
200 The maximum hydrogen production rate occurred when the BU:AC
ratio was maximised to approximately 1.6 in response to acid pre-
100
treatments.
0

0 50 100
Fig. 2. The cumulative hydrogen and methane production in response to different
pre-treatments as a function of time (a) hydrogenogenic stage and (b) methano-
genic stage.
pre-treatments, respectively. These values were higher than those
observed for other treatments after the hydrogenogenic stage.
However, the production of VFAs decreased dramatically at the
methanogenic stage.
The study of the effects of the FVW pre-treatments revealed
that the treatments inuenced not only the total production of
VFAs but the percentages of individual VFAs as well. In our study,
acetic, propionic, butyric, pentanoic, hexanoic, lactic acids and eth-
anol were observed in all experiments. As shown in Table 3, a very
small amount of ethanol was detected in the control and acid pre-
treated samples. The analysis of the composition of VFAs in the two
stages showed that propionic and acetic acid ranked rst in the
hydrogenogenic stage, whereas propionic acid was the dominant
150 200 250 300 3.4. FRI analysis of EEM spectra
Time (h)
The EEM uorescence spectra of the DOM and its corresponding
fractions resulting from different pre-treatments of FVW are
shown in Fig. 4. All of the spectra correspond to the presence of dif-
ferent uorophores as characterised by their excitation/emission
(Ex/Em) wavelength pairs. To better understand the EEM uores-
cence characteristics of these DOM samples, FRI method as de-
scribed by Chen et al. [25] was used to analyse the ve Ex/Em
regions. In general, peaks at shorter excitation wavelengths
(200250 nm) and shorter emission wavelengths (<80 nm) are re-
lated to simple aromatic proteins, such as tyrosine and tryptophan
(Regions I and II). Peaks located at excitation wavelengths between
200 and 250 nm and emission wavelengths of 380 nm or higher
represent fulvic acid-like substances (Region III). Peaks at interme-
diate excitation wavelengths (250280 nm) and shorter emission
wavelengths (<380 nm) are related to soluble microbial by-prod-
uct-like materials (Region IV). Peaks at longer excitation wave-
lengths (>280 nm) and longer emission wavelengths (>380 nm)
are related to humic acid-like organics (Region V) [26].

Table 2
Kinetic coefcients for different pre-treatments.

Stages Pre-treatments P (mL) Rm (mL/h) k (h) R2

H2 Control 0.78 0.01 0.16 0.03 0.38 0.67 0.969
Acid 180.69 1.73 10.11 0.46 3.88 0.43 0.997
Alkali 19.76 0.33 3.64 0.54 3.61 0.46 0.980
Enzyme 142.05 3.32 5.65 0.51 4.54 1.12 0.986
CH4 Control 9.45 0.12 0.15 0.01 83.75 2.27 0.995
Acid 851.14 87.01 4.67 0.25 146.65 4.00 0.991
Alkali 825.87 16.43 6.63 0.34 87.53 2.85 0.996
Enzyme 332.29 9.81 2.63 0.11 147.39 2.24 0.997

Note: values represent average STD. P: maximum hydrogen/methane production potentials; Rm: maximum hydrogen/methane production rate; k: lag time.
1224 X. Jia et al. / Energy Conversion and Management 88 (2014) 12191227

Table 3
The soluble metabolite concentrations produced after the different pre-treatments.

Stages Pre-treatments Soluble metabolite concentrations (mg/L)

ET AC PR BU PE HE VFAs LA pH
H2 Control 12 102 6.7 0 0 20.3 129 3840 5.11
Acid 4.26 925 8.5 1470 0 0 2403.5 376 4.80
Alkali 0 1340 12 820 0 0 2172 320 5.09
Enzyme 0 662 13.6 1780 19.5 0 2475.1 379 5.80
CH4 Control 13.1 719 0 1500 94.2 197 2725.2 361 7.07
Acid 0 76.2 199 84.5 94 71.6 565.8 0 7.58
Alkali 0 78.5 202 0 0 72 352.5 0 7.61
Enzyme 0 120 242 0 0 68.8 466.8 0 7.46

Data were the average of triplicates with standard deviations (n = 3). ET: ethanol; AC: acetic acid; PR: propionic acid; BU: butyric acid; PE: pentanoic acid; HE: hexanoic acid;
LA: lactic acid.

HE PE BU PR AC FA Rmax
100 12

90
10
80

70
Individual VFA (%)

8
60

Rm (mL/h)
50 6

40
4
30

20
2
10

0 0
Control-HS Control-MS Acid-HS Acid-MS Alkali-HS Alkali-MS Enzyme-HS Enzyme-MS

Pretreatments

Fig. 3. Percentage of individual VFA accounting for the total VFAs and Rm in response to different pre-treatments with the combined hydrogen/methane production process.

The uorescence intensity and peak wavelength ranges showed
marked differences after all pre-treatments: the fulvic acid-like
matter peaks (Regions III) nearly disappeared, but the uorescence
intensity of Region V gradually decreased, whereas the uores-
cence intensities of Regions I and II increased simultaneously. This
result indicates that the refractory humic acid-like organics (Re-
gion V) in the substrate were converted to biodegradable pro-
tein-like matter in response to the pre-treatments of FVW.
Compared with that observed during the initial fermentation
stage, the uorescence intensity of aromatic proteins (Regions I
and II) and by-product-like components (Region IV) gradually in-
creased, with uorescence spectral features dominated by biologi-
cal uorophores during the hydrogenogenic stage, which indicate
that the microbial activity increased during this anaerobic fermen-
tation process. A study of the EEM spectra illustrated that the sub-
strates metabolism produced biodegradable protein-like matter
and lower recalcitrant substances, which had the potential for
anaerobic digestion methane production [8].
During the methanogenic stage, the emission wavelength of the
maximum peak position shifted from 330 to 400 nm (red shift).
According to Milori et al. [27], longer emission wavelengths are
indicators of an increased amount of conjugated aromatic moieties,
higher molecular weight and a more complex structure of the
DOM. In addition, the strength of the uorescence peaks decreased
dramatically and species enriched in response to the acid and en-
zyme pre-treatments during this period. However, the alkali pre-
treatment did not induce obvious changes in the uorescence
intensity corresponding to the hydrogen and methane production
potentials during the two-stage fermentation process. Therefore,
the biodegradable DOM was utilised rapidly and completely during
the hydrogenogenic stage by the acid and enzyme pre-treatments,
and the lowest hard-degradation organic matter content was ob-
tained. This result suggests that the metabolite production primar-
ily consisted of biodegradable DOM (tyrosine-like, tryptophan-like
and by-product-like materials) and slightly recalcitrant DOM (ful-
vic acid-like and humic acid-like substances) during the two-stage
fermentation process.
3.5. Fluorescence-PARAFAC analysis of DOM
The DOM in the fermentation efuent from the two-stage
anaerobic fermentation process consisted of a large number of
uorescence compounds, and the uorescence peaks attributed
to different compounds may have overlapped. PARAFAC analysis,
a multivariate chemometric method that can decompose EEM
spectra into different independent uorescent components, pro-
vided accurate quantitative information on the distribution of
components in the DOM fractions [28].
Three separate uorescent components were identied in the
DOM by the PARAFAC model (Fig. 5), and the positions of the uo-
rescence maxima are shown in Table 4. The uorescence intensity
of each component is represented by Fmax.
Component 1 was centred at a maximum Ex/Em wavelength
pair of 220/340 nm and related to a typical protein-like substance,
such as tryptophan, which has been previously observed in fer-
mentation efuents [29]. Recent evidence supports the notion that
component 1 is likely derived from phenolic compounds, which
have Ex/Em wavelength pairs similar to those of component 1
and originate from the degradation of lignin [30,31]. The protein-
like substances are represented by component 1, which occupied
 X. Jia et al. / Energy Conversion and Management 88 (2014) 12191227 1225

450 1400 450 450 1000

900
Control-IN 1200
400 400
Control-HS 2000
400 Control-MS 800
1000 700

350 1500 350 600

350

Ex (nm)
Ex (nm)

Ex (nm)
800
500
600 300 1000 300 400
300
300
400
250 250 500 250 200
200
100

200 0 200 0 200 0

250 300 350 400 450 250 300 350 400 450 250 300 350 400 450
Em (nm) Em (nm) Em (nm)
450 4000 450 4500 450 2200

4000 2000
3500
Acid-IN Acid-HS Acid-MS 1800
400 400 3500 400
3000 1600
3000
2500 1400
350 350 350
Ex (nm)

Ex (nm)
Ex (nm)

2500 1200
2000
2000 1000
300 300 300
1500 800
1500
600
1000 1000
250 250 250 400
500 500 200

200 0 200 0 200 0

250 300 350 400 450 250 300 350 400 450 250 300 350 400 450
Em (nm) Em (nm) Em (nm)
450 3500 450 4000 450 4500

3500 4000
Alkali-IN 3000
Alkali-HS Alkali-MS
400 400 400 3500
3000
2500
3000
350 350 2500 350
Ex (nm)

2000
Ex (nm)

Ex (nm)
2500
2000
1500 2000
300 300 300
1500
1500
1000
1000
250 250 250 1000
500
500 500

200 0 200 0 200 0

250 300 350 400 450 250 300 350 400 450 250 300 350 400 450
Em (nm) Em (nm) Em (nm)
450 4000 450 5000 450 1800

3500 4500 1600

Enzyme-IN Enzyme-HS Enzyme-MS
400 400 4000 400 1400
3000
3500
1200
350 2500 350 350
3000
Ex (nm)

Ex (nm)

Ex (nm)

1000
2000 2500
800
300 300 2000 300
1500
600
1500
1000
250 250 1000 250 400

500 500 200

200 0 200 0 200 0

250 300 350 400 450 250 300 350 400 450 250 300 350 400 450
Em (nm) Em (nm) Em (nm)

Fig. 4. EEM spectra of DOM in response to different pre-treatments during three stages of anaerobic fermentation. IN: the initial sample after pre-treatments; HS: the efuent
end of hydrogenogenic stage; MS: the efuent end of methanogenic stage.

a dominant Fmax value in the DOM during the hydrogenogenic
stage in response to the three pre-treatments. However, compo-
nent 1 could not be detected in the original FVW (control). This re-
sult indicates that the protein-like substance of component 1
mainly originated from the freshly produced DOM due to biological
and microbial activities related to the pre-treatments methods.
Component 2 is composed of two excitation maxima at 225 and
275 nm (Table 4), with one emission peak centred at 330 nm. Two
components had excitation/emission characteristics similar to
those of uorescent protein-like compounds and presumably a
combination of uorophores containing uorescent proteins tryp-
tophan and tyrosine: free amino acids or amino acids bound in pro-
tein molecular structures [32]. This nding may reect that both
protein-like uorescence components exhibit the spectral proper-
ties of tyrosine and tryptophan bound into larger structures of
organic molecules rather than the pure compounds during the
two-stage fermentation process. The Fmax value of component 2
showed an decreasing trend during the biodegradation of FVW in
the two-stage production process except for alkali pre-treatment.
However, the Fmax value of component 2 was lower than that of
component 1.
Component 3 showed three uorescence peaks, with the maxi-
mum Ex/Em wavelength pairs centred at 205/425, 230/425 and
320/425 nm. According to Wu et al. [33], the three peaks in compo-
nent 3 are short-wavelength humic-like peaks that originate from
fulvic-like substances. The presence of component 3 during the
three fermentation stages was associated with allochthonous
DOM derived from terrestrial sources [31,32]. The relative concen-
tration of fulvic-like acid substance in component 3 was the lowest
among these three components, with slightly higher concentra-
tions observed after methane production than that of other stages
except for control.
1226 X. Jia et al. / Energy Conversion and Management 88 (2014) 12191227

Component 1 Component 2
400 400 0.12

0.25 0.1

350 350

0.2 0.08

Ex. (nm)
Ex. (nm)

300 0.15 300 0.06

0.1 0.04

250 250
0.05 0.02

200 0 200 0
250 300 350 400 450 250 300 350 400 450
Em. (nm) Em. (nm)

Component 3
400 0.06

0.05
350
0.04
Ex. (nm)

300 0.03

0.02
250
0.01

200 0
250 300 350 400 450 500 550
Em. (nm)

Fig. 5. The EEM spectra of the three components identied by PARAFAC analysis and maximum uorescence intensities (Fmax) in different fermentation stages.

Table 4 microbial activities and hydrogen/methane production potentials.

Positions of the uorescence maxima of the three components. The amount of component 3 was slightly increased by degradation
Fluorescent Excitation Emission Description and metabolic processes during the two-stage anaerobic fermenta-
components wavelength wavelength tion for three pre-treatments. In general, the PARAFAC analysis
(nm) (nm) provided additional quantitative information with which to de-
Component 220 340 Protein-like (tryptophan- scribe the distribution of the three components in the DOMs for
1 like) the fermentation efuents during different stages and evaluate
Component 225/275 330 Protein-like (tryptophan-
the effect of different pre-treatments for FVW.
2 like and tyrosine-like)
Component 205/235/345 455 Humic-like
3
4. Conclusion

PARAFAC analysis provided additional quantitative information
with which to describe the distribution of the three components in
each sample. As shown in Fig. 5, the humic-like substances of com-
ponent 3 exhibited a minimum Fmax value in all samples, whereas
the protein-like substances of component 1 and 2 exhibited a max-
imum value. Furthermore, component 1 was more abundant than
component 2 and 3, except for the control test, which is consistent
with the Fmax values.
As discussed above, protein-like substances can be concluded to
be the major components in the examined DOM samples, which
were easily biodegraded during the two-stage fermentation pro-
cess. After pre-treatments for FVW, the Fmax value of component
1 was enhanced dramatically, which mainly originated from the
freshly produced DOM due to hydrolysis of substrate. When com-
paring the three pre-treatments for FVW, a prominent degradation
was observed for protein-like component 1 and 2 after the transi-
tion from the hydrogenogenic to the methanogenic stage for the
acid and enzyme pre-treatments, which resulted from increased
Among the three pre-treatments, the acid pre-treatment re-
sulted in the most efcient biogas production and potential in
the hydrogenogenic stage (10.11 mL/h, 41.2% hydrogen) when
combined with the methanogenic process (4.67 mL/h, 76.1% meth-
ane). However, the fraction of methane was exactly 89.0% and the
maximum methane production rate was 6.63 mL/h in the metha-
nogenic process after the alkali pre-treatment. The analysis of the
composition of soluble metabolites in the two stages showed that
both ethanol- and butyric acid-type fermentation processes had
occurred as a result of the acid pre-treatment, whereas the alkali
and enzyme pre-treatments resulted in butyric acid-type fermen-
tation processes in the hydrogenogenic stage. Furthermore, the
DOM in the two-stage fermentation efuent of FVW was character-
ised by uorescence EEMs in conjunction with PARAFAC analysis in
response to different pre-treatments. This approach can trace pho-
tochemical and microbial reactions to estimate the relative con-
centrations of each uorescent component. In all, three
uorescent components were identied using the PARAFAC model.
 X. Jia et al. / Energy Conversion and Management 88 (2014) 12191227
The biodegradable DOM was rapidly and completely utilised in the
hydrogenogenic stage after the acid and enzyme pre-treatments,
and recalcitrant organic matter content was minimised. The dom-
inance of protein-like uorophores during the anaerobic fermenta-
tion process has already been established. The results of this study
indicate that uorescence may be used as a rapidly obtained indi-
cator of the organic matter degradation and stability of a two-stage
fermentation system after different pre-treatments for FVW.
Acknowledgement
This work is supported by the National Key Technology R&D
Program (2012BAJ21B02).
References
[1] Yang YQ, Shen DS, Li N, Xu D, Long YY, Lu XY. Co-digestion of kitchen waste and
fruitvegetable waste by two-phase anaerobic digestion. Environ Sci Pollut Res
2013;20(4):216271.
[2] Lin J, Zuo J, Gan LL, Li P, Liu FL, Wang KJ, et al. Effects of mixture ratio on
anaerobic co-digestion with fruit and vegetable waste and food waste of China.
J Environ Sci 2011;23(8):14038.
[3] Hawkes FR, Dinsdale R, Hawkes DL, Hussy I. Sustainable fermentative
hydrogen production: challenges for process optimization. Int J Hydrogen
Energy 2002;27(1112):133947.
[4] Zhao MX, Ruan WQ. Improving hydrogen generation from kitchen wastes by
microbial acetate tolerance response. Energy Convers Manage
2014;77:41923.
[5] Bouallagui H, Lahdheb H, Ben Romdan E, Rachdi B, Hamdi M. Improvement of
fruit and vegetable waste anaerobic digestion performance and stability with
co-substrates addition. J Environ Manage 2009;90:18449.
[6] Zhao MX, Ruan WQ. Biogas performance from co-digestion of Taihu algae and
kitchen wastes. Energy Convers Manage 2013;75:214.
[7] Guwy AJ, Dinsdale RM, Kim JR. Fermentative biohydrogen production systems
integration. Bioresource Technol 2011;102(18):853442.
[8] Guo XJ, He XS, Zhang H, Deng Y, Chen L, Jiang JY. Characterization of dissolved
organic matter extracted from fermentation efuent of swine manure slurry
using spectroscopic techniques and parallel factor analysis (PARAFAC).
Microchem J 2012;102:11522.
[9] He XS, Xi BD, Wei ZM, Jiang YH, Geng CM, Yang Y, et al. Physicochemical and
spectroscopic characteristics of dissolved organic matter extracted from
municipal solid waste (MSW) and their inuence on the landll biological
stability. Bioresource Technol 2011;102(3):23227.
[10] He XS, Xi BD, Wei ZM, Jiang YH, Yang Y, An D, et al. Fluorescence excitation
emission matrix spectroscopy with regional integration analysis for
characterizing composition and transformation of dissolved organic matter
in landll leachates. J Hazard Mater 2011;190(1):2939.
[11] Paredes C, Prez-Murcia MD. Dissolved organic matter fractions formed during
composting of winery and distillery residues: evaluation of the process by
uorescence excitationemission matrix. Chemosphere 2007;68:3019.
[12] Blackburn J, Liang Y, Das D. Biohydrogen from complex carbohydrate wastes as
feedstocksdcellulose degraders from a unique series enrichment. Int J
Hydrogen Energy 2009;34:742834.
1227
[13] Elbeshbishy E, Hafez H, Dhar BR, Nakhla G. Single and combined effect of
various pretreatment methods for biohydrogen production from food waste.
Int J Hydrogen Energy 2011;36(17):1137987.
[14] Ozkan L, Erguder TH, Demirer GN. Effects of pretreatment methods on
solubilization of beet-pulp and bio-hydrogen production yield. Int J Hydrogen
Energy 2011;36(1):3829.
[15] APHA. Standard methods for examination of water and wastewater. 19th
ed. Washington DC., USA: American Public Health Association; 1998. p. 1220.
[16] Zwietering MH, Jongenburger I, Rombouts FM, Vant Riet K. Modeling of the
bacterial growth curve. Appl Environ Microb 1990;56(6):187581.
[17] Kishor Jugal, Singh SP. Kinetics of anaerobic digestion. Energy Convers Manage
1988;28(4):3357.
[18] Shao ZH, He PJ, Zhang DQ, Shao LM. Characterization of water-extractable
organic matter during the biostabilization of municipal solid waste. J Hazard
Mater 2009;164:11917.
[19] Bro R. PARAFAC. Tutorial and applications. Chemometr Intell Lab Syst
1997;38(2):14971.
[20] Zhuo JF, Guo WD, Deng X, Zhang ZY, Xu J, Huang LF. Fluorescence excitation
emission matrix spectroscopy of CDOM from Yundang Lagoon and its
indication for organic pollution. Spectro Spectr Analysis 2010;30(6):153944.
[21] Cao GL, Ren NQ, Wang WQ, Xu JF, Liu BF. Effect of lignocellulose-derived
inhibitors on growth and hydrogen production by Thermoanaerobacterium
thermosaccharolyticum W16. Int J Hydrogen Energy 2012;35:1347580.
[22] Tobin PR, Maria JO, Anu K, Bernard MM. The effect of the growth environment
on the lag phase of Listeria monocytogenes. Int J Food Microbiol
1998;44:8392.
[23] Mu Y, Yu HQ, Wang G. Evaluation of three methods for enriching H2-producing
cultures from anaerobic sludge. Enzyme Microb Tech 2007;40:94753.
[24] Ren NQ, Guo WQ, Wang XJ, Xiang WS, Liu BF, Wang XZ. Effects of different
pretreatment methods on fermentation types and dominant bacteria for
hydrogen production. Int J Hydrogen Energy 2008;33:431824.
[25] Chen W, Westerhoff P, Leenheer JA, Booksh K. Fluorescence excitation
emission matrix regional integration to quantify spectra for dissolved organic
matter. Environ Sci Technol 2003;37(24):570110.
[26] Artinger R, Buckau C, Geyer S, Fritz P, Wolf M, Kim JI. Characterization of
groundwater humic substances: inuence of sedimentary organic carbon.
Appl Geochem 2000;15:97116.
[27] Milori DMBP, Martin-Neto L, Bayer C, Mielniczuk J, Bagnato VS. Humication
degree of soil humic acids determined by uorescence spectroscopy. Soil Sci
2002;167(11):73949.
[28] He XS, Xi BD, Li X, Pan HW, An D, Bai SG, et al. Fluorescence excitation
emission matrix spectra coupled with parallel factor and regional integration
analysis to characterize organic matter humication. Chemosphere
2013;93(9):220815.
[29] Yamashita Y, Tanoue E. Chemical characterization of protein-like uorophores
in DOM in relation to aromatic amino acids. Mar Chem 2003;82(3):25571.
[30] Maie N, Scully NM, Pisani O, Jaffe R. Composition of a protein-like uorophore
of dissolved organic matter in coastal wetland and estuarine ecosystems.
Water Res 2007;41(3):56370.
[31] Hernes PJ, Bergamaschi BA, Eckard RS, Spencer RGM. Fluorescence-based
proxies for lignin in freshwater dissolved organic matter. J Geophys Res:
Biogeosci 2009;114(G4).
[32] Kowalczuk P, Durako MJ, Young H, Kahn AE, Cooper WJ, Gonsior M.
Characterization of dissolved organic matter uorescence in the South
Atlantic Bight with use of PARAFAC model: interannual variability. Mar
Chem 2009;113(3):18296.
[33] Wu HY, Zhou ZY, Zhang YX, Chen T, Wang HT, Lu WJ. Fluorescence-based rapid
assessment of the biological stability of landlled municipal solid waste.
Bioresource Technol 2012;110:17483.