Chemical Engineering Journal 334 (2018) 1383–1391

Contents lists available at ScienceDirect

Chemical Engineering Journal

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

Enhancing anaerobic fermentation performance through eccentrically T

stirred mixing: Experimental and modeling methodology
Yuankai Huanga, Farzaneh MahmoodPoor Dehkordya, Yan Lia, Sharareh Emadib,
⁎
Amvrossios Bagtzogloua, Baikun Lia,
a
Department of Civil and Environmental Engineering, University of Connecticut, Storrs, CT 06269, United States
b
Department of Biomedical Engineering, University of Connecticut, Storrs, CT 06269, United States

G RA P H I C A L AB S T R A C T

A R T I C L E I N F O A B S T R A C T

Keywords: Anaerobic hydrogen production (AHP) is a biochemical process capable of producing hydrogen and fatty acids
Anaerobic hydrogen production (AHP) from organic substrates in acidogenic phase. Currently, almost all of anaerobic systems use centrally stirred
Eccentrically stirred mixing mixing where the flow is essentially rotated uniformly in the radial direction. Given the high energy requirement
Mixing eccentricity of mixing and the urgent needs for enhancing the contact between microbial particles and substrates, a new type
Mixing height
of mixing strategy should be developed. This study focused on the enhancement of biogas production and or-
Lyapunov exponent
ganic removal efficiency in AHP systems through a novel chaotic mixing regime. The stirring impeller was
Navier-Stokes equations
operated eccentrically with the stirring center eccentrically located 2 and 4 cm off the center, respectively and at
different heights (5, 10.5 and 16 cm). The eccentrically stirred mixing was expected to enhance the mixing
efficiency and accelerate the contact between biomass and organic substrates in the AHP system. A numerical
finite element model was developed to simulate the flow patterns and the biomass distribution, through which
the Lyapunov exponent λ values were calculated in order to assess the strength of eccentrically stirred mixing at
different mixing eccentricities. The model simulations showed that at the height of 10.5 and 16 cm, the
Lyapunov exponent λ increased as the mixing eccentricity increased. This corresponded well with the experi-
mental results, which showed that the organic removal efficiency and biogas production increased with the
mixing eccentricity. Compared with centrally stirred mixing, eccentrically stirred mixing enhanced the inter-
action of flow and biomass particles and ultimately enhanced the biogas production and organic removal in AHP
systems.

1. Introduction capable of producing hydrogen and organic acids (e.g., acetic acid,
butyric acid) from organic substrates (e.g., glucose and wastewater)
Anaerobic hydrogen production (AHP) is a biochemical process under acidogenic phase. Compared with methane production from

⁎
Corresponding author.
E-mail address: baikun.li@uconn.edu (B. Li).

https://doi.org/10.1016/j.cej.2017.11.088
Received 23 September 2017; Received in revised form 14 November 2017; Accepted 17 November 2017
Available online 21 November 2017
1385-8947/ Published by Elsevier B.V.
Y. Huang et al.
anaerobic digestion (AD), AHP offers unique advantages including high
reaction rate, high operational stability, and clean energy production
with hydrogen gas as the major biogas component [1]. Previous re-
searchers have found that several engineering parameters (e.g. pH
[2,3], temperature [4,5] and substrate content [5,6]) directly affected
hydrogen production (e.g. high hydrogen yield and suppression of the
methanogens at pH of 4.8–5.5 and temperature of 60 °C [2]). In fact, the
mixing of anaerobic systems is critical for reaction rate and energy
consumption [7]. Normally, sufficient mixing ensures organic sub-
strates efficiently distributed in anaerobic systems to react with biomass
particles, leading to fast heat transfer, release of trapped bubbles, and
avoid biomass precipitation [8]. Slow mixing (e.g. under 160 rpm) has
been used to facilitate sludge granulation, although complete-mixing
AD systems can cause a washout of granular sludge [9,10]. Traditional
aerobic/anaerobic systems have historically used centrally stirred
mixing (henceforth called centric mixing) where the flow is essentially
rotated uniformly. Although this mixing pattern is well adopted and
easily designed and operated, it might not optimize the interaction
between substrates and biomass particles, especially in AHP systems
where biogas production can complicate the flow field. As one of the
major energy consumption factors in anaerobic systems, it is critical to
better understand the mixing strategy and develop new energy-saving
and efficient mixing patterns.
Mixing in fluids is accomplished slowly by molecular diffusion in
laminar flows [11], which is greatly accelerated by turbulence and is
associated with large velocity gradients, especially in the small scales
and high shear rates. For these reasons, turbulence is not suitable for
mixing in biological situations where the high shear rates could damage
bacterial cells and tear the granular sludge. In other situations such as
highly viscous flows, the flow remains laminar. A novel and funda-
mental mixing process rooted in dynamical systems theory has been
proposed [12] and termed chaotic advection. It was theorized that
chaotic mixing would significantly increase the exposed surface area.
Flow changes direction continuously under the chaotic mixing scheme,
which is expected to greatly enhance organic removal by maximizing
contact between biomass particles and substrates. Even though turbu-
lent mixing is common in high Reynolds number flows where the large
amount of energy drives particles (e.g., bacterial cells, flocs) in random
directions and produces a homogeneous fluid, chaotic mixing can also
occur at low Reynolds numbers when time-periodic oscillations with
fixed amplitudes are imposed [13]. Chaotic advection causes simple
non-turbulent flows to exhibit complicated particle trajectories that
result in enhanced mixing, and has been applied to low Reynolds
number plasma mixing without damaging bacterial cells, to enhancing
heat transfer, and to creating long-chain polymers. Our group has ap-
plied this concept to a variety of wastewater and groundwater appli-
cations and scales [14–17]. Furthermore, parallel-competitive reaction
systems have been simulated in a chaotic flow by solving the differ-
ential convection–diffusion-reaction and it was shown convincingly
that besides the relative rates of reaction, the nature of the chaotic flow
is a determining factor in the evolution of the aqueous concentrations
[18]. The groundwork was presented in a previous study [19] by sol-
ving numerically the aforementioned differential equations. An ex-
cellent state-of-the-art review of the numerous applications and theo-
retical background of chaotic advection was presented this year [20].
Kinetic mixing models have been studied to elucidate the effect of
mixing speed/intensity on mass dispersion, biogas production, sub-
strate removal efficiency and mass viscosity. Idealized models demon-
strating chaotic advection include the point vortex model, tendril-whorl
flow, the pulsed source-sink system, and the three-vortex system. Until
now, mixing in AD systems has been limited to centrally located im-
pellers for the convenience of system configuration and operation [21].
There have been only a few studies of chaotic mixing (off center) on
mass distribution and reaction [22] and no chaotic mixing has been
applied in the field of AHP. In fact, mass distributions in AHP systems
would vary with the mixing position. A flow visualization technique
1384
Chemical Engineering Journal 334 (2018) 1383–1391
was developed by observing the fluorescent green dye in a vessel to
elucidate the mixing characteristics [23]. The biofermentor can be di-
vided into two sections – active and inactive volumes, with active vo-
lume being formed as a cavern around the mixer. It is expected that the
size of the active volume would change with the mixing position in AHP
biofermentors, leading to variable mass distribution, gas bubble dis-
persion, biogas production, substrate precipitation and liquid fermen-
tation production. Previous studies had found that in the journal-
bearing system (a system similar to the AHP experimental design in this
study) consisting of two eccentric cylinders, the annular region was
filled with a viscous liquid and the cylinders were rotated alternately to
produce chaotic advection [24]. Furthermore, the role of diffusion and
transient velocities was studied in the dispersal of passive scalars pro-
duced in a low Reynolds number journal-bearing flow [25].
The breakthrough of this study was to elucidate the impact of
chaotic mixing on the enhancement of biogas production and organic
removal in an AHP biofermentor, with the objective to explore the re-
lationships between the chaotic mixing theme and energy output of
AHP process. There were four tasks in this study. First, a lab-scale AHP
biofermentor was set up to use glucose, a typical carbon source for
hydrogen production bacteria, and operated in different chaotic mixing
conditions. Second, the important parameters in AHP such as gas pro-
duction, liquid components and chemical oxygen demand (COD) re-
moval efficiency were investigated under different mixing impeller
positions. Third, a numerical finite element model was developed to
explore the flow and particle motion in the AHP system and determine
the relationship between mixing and energy output. Finally, the sig-
nificance of applying chaotic mixing for anaerobic biofermentors was
explored in the context of scientific and industrial perspectives.
2. Materials and methods
2.1. Anaerobic hydrogen producing (AHP) biofermentor setup and
operation
A BIOFLO 110 Fermentor (New Brunswick Scientific, NJ) (effective
volume of 2 L, height: 25 cm, radius: 6.5 cm) was used as the AHP
system. The vessel was sealed for air impermeability. An agitator with
an impeller was installed for mixing. The radius of the impeller was
2 cm and the width of the impeller was 1.75 cm with 6 blades uniformly
distributed around the circumference (Fig. 1). Before the tests, 20.625 g
glucose was fully dissolved in 200 ml distilled water and added to the
vessel as the carbon source, followed by filling with 2 L nutrient solu-
tion containing 4.0 g NH4HCO3, 2.6 g NaH2PO4, 200 mg MgSO4·7H2O,
20 mg KCl, 20 mg Na2MoO4·2H2O, 20 mg CaCl2·2H2O, 15 mg
MnSO4·7H2O and 5.6 mg FeCl2. The chemical oxygen demand (COD) of
the mixed solution was 10 g/L. The vessel was then sparged with ni-
trogen gas (99.995% high purity, Airgas Co., CT) for 20 min to achieve
the anaerobic environment. Organic soil (22 g) collected from a
greenhouse at the University of Connecticut was used as the inoculum
for anaerobic bacteria to obtain the biomass particle concentration of
10 g/L in the AHP system.
Previous studies had found that the disturbance of shock loading
was better absorbed by the digester at low speed mixing conditions
(80 rpm) than at high speed mixing conditions (200 rpm) [26], since the
bacteria in the anaerobic digestion process are more likely to live in a
balanced micro-ecosystem [27], and the low speed mixing conditions
helped to maintain this good process in digestors. Furthermore, the
energy dissipated by the system at low mixing was only 1/16 as the
high mixing condition [26]. In this study, the mixing rate of the im-
peller was set at 50 rpm (the lowest mixing setup for the BIOFLO bio-
fermentor). The temperature of the AHP biofermentor was kept at 38 °C
by using a heating pad wrapped around the vessel.
Y. Huang et al. Chemical Engineering Journal 334 (2018) 1383–1391

Fig. 1. The mixing positions in the AHP biofermentor (a)

and the impeller (b).

2.2. Mixing positions in the AHP biofermentor test kit (HACH, CO) with a spectrophotometer (DR 2800, HACH, CO).
The pH was on line monitored using a pH meter connected to the AHP
After the glucose, nutrient solution and soil were added into the biofermentor. The constituents of the liquid samples were analyzed
AHP biofermentor, the liquid level height was 19.5 cm. Three mixing using a gas chromatograph (GC) (Agilent 6950, CA) equipped with an
heights (5, 10.5 and 16 cm from the bottom) and three mixing eccen- injector (7683B series) and flame ionization detector. The biogas
tricities (0, 2, 4 cm from the center of the AHP biofermentor) were samples collected from the AHP biofermentor using a 200 μl gas tight
examined individually to determine the effects of eccentrically stirred syringe (Fisher Scientific, MA) were analyzed with a GC (Agilent 6950,
(henceforth called eccentric) mixing to the AHP process (Fig. 1). The CA) equipped with a packed column and thermal conductivity detector.
matrix of the mixing height and eccentricity is listed in Table 1. All nine All the liquid and biogas samples were measured in duplicate.
setups were examined in triplicate. Statistical analyses were conducted as follows: ANOVA was used to
determine the significance of the effects the height and eccentricity
have on COD removal and biogas production. The T-test was used to
2.3. Sample collection from the AHP biofermentor during batch-mode tests
quantify the significance of the slope in the relationships between the
chaotic mixing coefficient (Lyapunov exponent λ) and AHP perfor-
Based on three mixing heights and three mixing eccentricities, a
mance.
total of nine groups of experiments were conducted (Table 1) in the
batch-mode individually for 45 h until no more gas was produced in the
headspace and no COD or pH value changed in the liquid phase. The 2.5. Chaotic mixing modeling and Lyapunov exponent
liquid sample (volume: 10 mL) was collected every 5 h using a sampling
pump connected with the AHP biofermentor, and filtered using a filter COMSOL Multiphysics (COMSOL, Inc., Burlington, MA, USA), a fi-
member (pore size: 4.5 μm, Millipore, MA) to remove the suspended nite element method based software package, was used to simulate the
solids and bacteria and then diluted 10 times using distilled water be- chaotic mixing under different impeller locations in the AHP bio-
fore analysis. The biogas sample was collected from the headspace fermentor. Specifically, sensitive dependence on initial conditions is an
using an airtight bottle (volume: 250 mL), The volume of the biogas important characteristic of chaotic motion, and close neighboring tra-
produced during each batch test was measured by a volumetric flow jectories are found to diverge. Furthermore, the rate of divergence is
meter (Cole Parmer, IL). exponential in chaotic systems, while regular trajectories are found to
separate only linearly in time. The number that distinguishes between
various types of orbits is termed as the Lyapunov exponent. A positive
2.4. Analytical method
Lyapunov exponent is the fundamental criterion for the system to be
considered chaotic.
The COD concentration was measured using a COD TNT plus vial
The Lyapunov exponent can be defined using the following ex-
ample. Assume two points in space are separated by a small distance ε0.
Table 1
Matrix of mixing height and eccentricity tested in the AHP biofermentor. Each of these points generates an orbit in space, so the separation of
two orbits is a function of time. The Lyapunov exponent measures the
Mixing position rpm Input Input pH Temperature (°C) dependence on initial conditions, termed as the separation ε (x0, y0, t).
COD (g/
The mean exponential rate of divergence of the two initially close orbits
Eccentricity (cm) Height (cm) L)
can be calculated using Eq. (1):
0 5 50 9983.33 7.54 38
2 5 50 9718.33 7.22 38 1 |ε (x 0,y0 ,t )|
λ = lim lim
4 5 50 10286.67 7.37 38 t →∞|ε0 → 0| t |ε0 | (1)
0 10.5 50 9523.33 7.55 38
2 10.5 50 9400 7.44 38 where λ is the Lyapunov exponent. For a discrete-time dynamical
4 10.5 50 9363.33 7.61 38 system exhibiting chaotic advection ‖εn ‖ ∼ ‖ε0 ‖e λn . This implies that a
0 16 50 9396.67 7.33 38
semi-logarithmic plot of the separation distance between a pair of
2 16 50 9926.67 7.41 38
4 16 50 9898 7.36 38 particles released very close to each other as a function of time should
appear linear; at least until boundary effects from a reactor (e.g., the

1385
Y. Huang et al. Chemical Engineering Journal 334 (2018) 1383–1391

edge, the bottom, or the impeller) become important. In this study, such
type of analysis was used to verify the chaotic status of the flow fields in
the AHP biofermentor.
In order to verify whether the AHP biofermentor was operated
under laminar flow conditions, the Reynolds number was calculated
(Eq. (2)):
R e = RiΩ(R o−Ri)/ ν (2)
where Ri and Ro, are the impeller and AHP biofermentor outer radii of
the apparatus, respectively, ν is the kinematic viscosity of the liquid and
Ω is the relative angular velocity with which the impeller rotates with
respect to the cylinder. For the AHP biofermentor operated at 38 °C, the
density of glucose solution was measured as 1007 ± 4.47 kg/m3, the
impeller and biofermentor outer radii of the apparatus, Ri and Ro, were
2 cm and 6.5 cm, respectively, and the kinematic viscosity of the liquid
was 1.032 × 10−6 m2/s resulting in a Reynolds number in the order of
700, which is indeed within the laminar flow regime.

3. Results and discussion

3.1. Effect of chaotic mixing conditions on liquid fermentation products

The main liquid fermentation products were acetic acid and butyric
acid under all mixing locations (Table 2), while the concentration of
ethanol, propionate and lactic acid were neglected (< 0.1 mM). No
methane was detected in the biogas content, indicating that only
acidogenesis and butyrate acetogenesis occurred in the AHP process.
Previous study had shown that hydrogen gas was produced at pH of
4.5–6.0, while acetoclastic methanogenesis occurred at pH > 5.7 with
acetic acid and butyric acid as the main liquid products [28]. In this
study, the pH value of the AHP biofermentor dropped rapidly from
∼7.4 to ∼4.8 within 5 h of the batch tests, and the biogas production
started after 6 h, indicating that the growth of methanogens was in-
hibited at low pH.
With glucose as the model substrate, acidogenesis and acetogenesis
started with the conversion of glucose to pyruvate through the glyco-
lytic pathway. Acetic acid, butyric acid or ethanol were produced (Eqs.
Fig. 2. The variation of COD removal efficiency with HAc ratio, mixing height and mixing
(3) and (4) [29]), and highly associated with hydrogen production
eccentricity in the AHP biofermentor (a: HAc ratio; b: mixing height; c: mixing eccen-
(4 mol H2/mole glucose for producing acetic acid and 2 mol H2/g glu- tricity).
cose for producing butyric acid).
C6 H12 O6 + 2H2 O→ 2CH3 COOH + 2CO2 + 4H2 (3) y z y
C x Hy Oz + ( x+ − )O2 → xCO2 + H2 O
4 2 2 (8)
C6 H12 O6 → CH3 CH2 CH2 COOH + 2CO2 + 2H2 (4)
Two other fermentation pathways can also be conducted by According to the mass balance equation of the COD of carbohydrate
Clostridium articum and Clostridium barkeri, leading to the production of (Eq. (8)), the COD concentrations of 1 mol/L acetic acid and butyric
propionate, ethanol, lactic acid and carbon dioxide (Eqs. (5)–(7), [29]). acid are 2 mol/L and 5 mol/L (Eq. (8)), respectively. Thereby, more
COD was generated through butyrate production process than acetate
C6 H12 O6 + 2H2 → 2CH3 CH2 COOH + 2H2 O (5) production process (5 mol/L COD comparing to 4 mol/L COD with
C6 H12 O6 → 2CH3 CH2 OH + 2CO2 (6) 1 mol/L glucose, Eqs. (3)–(5)). There was a positive relationship be-
tween the ratio of acetic acid and COD removal efficiency (Table 2,
C6 H12 O6 → 2CH3 CHOHCOOH + 2CO2 (7) Fig. 2a), demonstrating that higher ratio of acetic acid led to lower COD

Table 2
The liquid products and biogas products of the AHP biofermentor under nine mixing conditions.

Mixing position Input COD (g/L) Output COD removal efficiency (%)

Eccentricity (cm) Height (cm) HAc (mM) HBu (mg/L) Glucose (Mm) HAc fraction (%) COD (mg/L)

0 5 9983.33 19.8 28.4 5.5 36.87 7038.33 29.50

2 5 9718.33 20.1 27.5 3.5 39.33 6618.33 31.90
4 5 10286.67 21.5 28.4 3.9 39.96 6879.33 33.12
0 10.5 9523.33 20.5 26.4 4.7 39.73 6706.67 29.58
2 10.5 9400 20.5 26.1 2.7 41.58 6193.33 34.11
4 10.5 9363.33 21.5 26 1.2 44.15 5783.33 38.23
0 16 9396.67 20.1 26.5 3.7 39.96 6335 32.58
2 16 9926.67 22.3 27.1 2.4 43.05 6311.67 36.42
4 16 9898 23.9 25.9 0.4 47.61 5317.83 46.27

1386
Y. Huang et al. Chemical Engineering Journal 334 (2018) 1383–1391

concentration of effluent. Table 3

Theoretically, if all glucose in the influent (∼10,000 mg/L) were Hydrogen production in the AHP biofermentor under nine mixing conditions.
converted into acetic acid, the final COD concentration would be
Mixing position Input COD Output (gas)
∼6666.67 mg/L and the COD removal efficiency was ∼33.33%. (mg/L)
However, half of the COD removal efficiencies among nine mixing lo- Eccentricity (cm) Height (cm) H2 (mM) SHY (mole H2/
cations tested were higher than 33.33% (Table 2). Average COD re- mole glucose)
moval efficiency in this study (34.62%, Table 2) was slightly higher
0 5 9983.33 104.5 2.01
than previous studies (30% at OLR of 10 COD/L·d [30]), which was 2 5 9718.33 102.8 2.03
caused by the lower OLR rate in this study (4.6–5 g COD/L·d). COD 4 5 10286.67 107.9 2.01
removal efficiency had been found to increase as the organic loading 0 10.5 9523.33 103.5 2.09
rate (OLR, g/L·d) decreased in the AHP process [31]. In fact, the results 2 10.5 9400 103 2.10
4 10.5 9363.33 103.4 2.12
in this study corresponded well to the reported COD removal efficiency 0 16 9396.67 101.9 2.08
at similar OLR (36% at OLR of 4.7 COD g/L·d [32]). Furthermore, since 2 16 9926.67 109.6 2.12
acetic acid is the substrate for hydrogen production (Eq. (9)) [33], the 4 16 9898 110 2.13
consumption of acetic acid in AHP can also lower the COD concentra-
tion in the AHP effluent.
CH3 COOH + 2H2 O→ 2CO2 + 4H2 (9)
Mixing height and mixing eccentricity affected the COD removal
efficiency and the acetic acid ratios in liquid products. The maximum
COD removal efficiency (46.26%) was achieved at the chaotic mixing
position of α = 4 cm and h = 16 cm, while the minimum COD removal
efficiency (29.52%) occurred at the symmetric mixing position of α = 0
and h = 5 cm (Table 2). Previous study had found that in laminar fluid
flow, the mixture quality in biofermentors was severely degraded due to
the presence of isolated mixing regions at the symmetric scenario (the
impeller in the center) [34], since the structure of the global unstable
manifolds would never fill the entire flow domain [35]. In this study,
insufficient mixing under the symmetric mixing (the impellers in the
center of the AHP biofermentor) reduced the contact of the biomass
particles with glucose, and resulted in lower glucose degradation rate
and lower COD removal efficiency (Table 2, Fig. 2b and c).
At the chaotic mixing conditions (eccentricities of 2 cm and 4 cm),
the COD removal efficiencies increased with the mixing heights
(Fig. 2b), while no pattern was found at symmetric mixing conditions
(R2 < 0.8 at eccentricities of 0 cm, Fig. 2b). The possible reason was
that stronger chaotic movements of the fluid particles were formed
around the impeller, and more eddies and vortices were formed at
higher mixing positions [22]. In contrast, the chaotic movements in the
symmetric mixing were relatively weaker and the particles have
minimal interaction with each other [36], so that the COD removal in
the symmetric mixing conditions did not change substantially with the
mixing height. Furthermore, the COD removal efficiencies increased
with the mixing eccentricities at all three mixing heights (5, 10.5 and
16 cm) (Fig. 2c), which clearly validated the chaotic mixing enhanced Fig. 3. The variation of SHY values with mixing height (a) and mixing eccentricity (b) in
biochemical reactions. A two-way ANOVA using the mean values of the the AHP biofermentor.
experiments was conducted and showed that COD removal efficiency
depended significantly on both of the mixing eccentricity (p value: certain amount of hydrogen gas could be consumed for propionate
.0486) and the height (p value: .0887). Previous studies had suggested production (Eq. (5), although it did not occur much in this study). The
that the relative eccentricity (α/D, α: distance of the impeller to the specific hydrogen yields (SHY, mole H2/mole glucose) was much lower
center of the biofermentor, D: diameter of the biofermentor) of the than the stoichiometric hydrogen yield (Table 3). The maximum gas
impeller should be between 24% and 48% to obtain the best mixing production per mole glucose (2.13 mol H2/mole glucose) was slightly
quality [22]. In this study, the mixing eccentricity of 4 cm was 30.77%, lower than previous studies (2.04–2.72 mol H2/mole glucose [30],
well in the range previously reported. 2.32 mol H2/mole glucose [37]), which may be attributed to the lack of
pH control and pre-heat shocked soil inocula in this study, since the
3.2. Effect of chaotic mixing conditions on biogas production heated shocked soil was found to selectively favor the growth of
acidogenic bacteria [38].
Hydrogen production in the AHP biofermentor was lower than the Mixing height and mixing eccentricities clearly affected hydrogen
theoretical hydrogen yield (4 mol for each mole of glucose, Eq. (3)). production. The maximum SHY (2.13 mol H2/mole glucose) was
Based on the glucose amount (∼20.6 g) in the feeding solution and the achieved at the chaotic mixing position of eccentricity α = 4 cm and
reaction temperature (38 °C, 311.15 K), the maximum hydrogen pro- height h = 16 cm, while the minimum SHY (2.01 mol H2/mole glucose)
duction in this study should be 11.69 L. In the AHP process, the hy- occurred at the symmetric mixing position of α = 0 and h = 5 cm
drogen production was highly associated with OLR and hydrogen par- (Table 3). Like the COD removal efficiencies (Fig. 2b), the stronger
tial pressure. Glucose might be degraded through other pathways such chaotic movements in the eccentric mixing conditions (mixing eccen-
as butyrate acetogenesis, leading to a lower hydrogen gas production tricities of 2 cm and 4 cm) enhanced the SHY values with the mixing
(half of acidogenesis). Moreover, as the partial pressure increases,

1387
Y. Huang et al.
heights, while no pattern was found at the centric mixing conditions
(Fig. 3a). The experimental data of hydrogen production was confirmed
by the COD removal efficiencies, since the COD removal efficiency was
positively correlated with hydrogen production (Eqs. (3) and (4)).
It should be noted that hydrogen production increased with the
eccentricity at heights of 10.5 and 16 cm, but not related with eccen-
tricity at the height of 5 cm (Fig. 3b). The main reason was that the
boundary effects were strong at the height of 5 cm and affected the
overall flow patterns. Furthermore, based on the modeling results, the
flow pattern was faster around the impeller (as shown in Fig. 4a–c). At
the higher mixing position (the height of the impellers: 16 cm), more
power was dissipated on the liquid surface due to the faster flow speed,
which helped releasing the trapped bubbles from the liquid phase to the
headspace level (Fig. 1Sa). In contrast, at the low mixing position (the
height of the impellers: 5 cm), the driving force of releasing the trapped
bubbles from the liquid phase was weak, and thus more bubbles were
1388
Chemical Engineering Journal 334 (2018) 1383–1391
Fig. 4. Results from COMSOL simulation with impeller at
height of 16 cm for three mixing eccentricities in 3D view
and 2D bird’s view. a) for α = 0 cm, b) for α = 2 cm, c) for
α = 4 cm. The color of the slice plot shows the velocity
magnitude (m/s) and the red lines show the streamlines
associated with this flow simulation.
observed on the inner wall of the biofermentor (Fig. 1Sb). Previous
studies had found that when the impeller was placed at a relative high
position, the lower part of the biofermentor was not disturbed [38], and
the bubbles (gas produced in the biofermentor) would not be affected.
However, hydrogen production would be strongly affected in this study
when the impeller was put at a lower position, since gas bubbles were
trapped in the liquid contents and not contributed to biogas production
(as shown in Fig. 1Sb), and thus interfering with the correlation be-
tween the SHY values and eccentricity. Like the COD removal effi-
ciency, a two-way ANOVA using the mean values of the experiments
was also conducted to study the dependence of SHY on the mixing
eccentricity and height. SHY was found to depend significantly only on
the mixing height (p value: .0026) and not depend significantly on the
mixing eccentricity (p value: .165). But this result could easily change
with the addition of few extra data points.
Y. Huang et al. Chemical Engineering Journal 334 (2018) 1383–1391

3.3. Flow simulations of chaotic mixing in AHP

A numerical finite element model was developed using COMSOL

Multiphysics to simulate flow and particle motion in the AHP bio-
fermentor. The flow was simulated by solving the Navier-Stokes equa-
tions using the Arbitrary Lagrangian-Eulerian (ALE) technique [39]
(Eqs. (10) and (11)):
∂ρ ∂x
− . ∇ρ + ∇ . (ρu ) = 0
∂T ∂T (10)

∂u ∂x
ρ ⎛ − . ∇u ⎞ + ρ (u. ∇) u = ∇ . [−pI + τ ] + F
⎝ ∂T ∂T ⎠ (11)

where ∂ρ and ∂u are the derivatives of density and velocity, respec-

∂T ∂T
tively, with respect to mesh reference time (T), x is a function of angular
∂r
velocity and time, u = v + ∂t , in which v is the velocity vector in the
rotating coordinate system, and r is the position vector, p is pressure, τ
Fig. 5. Average particle separation distance analysis for impeller at height of 16 cm and at is shear stress, and F is external forces. Non-overlapped domains for the
different eccentricities 0 cm, 2 cm, and 4 cm away from the center. The Lyapunov ex- rotating and fixed parts were defined to simulate the flow field. A flow
ponent λ is calculated over the first 5–8 s of the process. continuity boundary condition was applied on the common boundaries
of fixed and rotating domains. A rotating interior wall boundary con-
Table 4
dition was applied on the walls of the impeller. In order to simulate a
Lyapunov exponent (1/s) calculated as a function of height and eccentricity. free surface condition on the top boundary of the AHP biofermentor, a
symmetry boundary condition was applied to avoid any shear stress
Lyapunov λ α = 0 cm α = 2 cm α = 4 cm development. Because the velocity of the fluid at a solid wall is zero
H = 5 cm 0.37511 0.22465 0.17411
relative to the wall, no-slip boundary conditions were assumed at the
H = 10.5 cm 0.22757 0.30097 0.42960 walls. The simulations employed an unstructured triangular mesh
H = 16 cm 0.18662 0.35799 0.43704 consisting of 357,303 domain elements, 20,758 boundary elements,
and 1717 edge elements.
To simplify the simulation, the velocity field was computed using
the frozen rotor assumption, which was a special case of a steady state
study. To study particles’ motion and their distances relative to each
other, the effect of different flow fields created by different impeller
mixing locations (as shown in Table 1) were explored. Specifically,
several pairs (10–12) of massless particles were released and scattered
all over the AHP biofermentor. The particles in each pair were initially
placed 1–3 mm apart and the separation distance of the particles in
individual pairs was calculated for the first 60 s of the simulation. To
ensure that the boundaries of fixed and rotating domains should not
impact the particles’ motion, the particle-continuity boundary condi-
tion was applied on the common boundaries of fixed and rotating do-
mains.
The streamlines and velocity field for three different locations of the
impeller were demonstrated (Fig. 4). The control experiment whereby
the impeller was located at the center was characterized by a more
consistent three-dimensional (3D) helical and two-dimensional (2D)
circular flow pattern (Fig. 4a). The streamlines from the control ex-
periment were not completely disordered and a circular pattern was
observed. In contrast, a more complex flow pattern was observed when
the impeller was eccentrically located at 2 cm and 4 cm off the center
(Fig. 4b and c), respectively. In this case the flow pattern was dis-
ordered, showing that the eccentricity of the impeller created a more
complex and possibly chaotic flow in comparison to the control ex-
periment (impeller in the center).
The particle separation distance for impeller at height of 16 cm and
at different eccentricities α (0 cm, 2 cm, and 4 cm) away from the center
was illustrated (Fig. 5). The Lyapunov exponent λ was calculated over
the first 5–8 s of the process, since after that point in time, boundary
effects were experienced due to the finite size of the biofermentor with
particles bouncing off or getting stuck at the walls. For instance, in the
control experiment, two of the particles got stuck to the biofermentor
wall after 7.6 s, so tracking their separation distance after that point
Fig. 6. Relationships between Lyapunov exponent λ and COD removal efficiency (a) and was impossible. The natural logarithm of particle separation distance
gas production (b).
with respect to time is plotted for the impeller located at 0 cm, 2 cm,
and 4 cm away from the center and at height of 16 cm (Fig. 5), re-
spectively. Based on the analysis, as the eccentricity increased, the

1389
Y. Huang et al.
Table 5
The comparison of energy input and output under centric mixing and eccentric mixing in AHP biofermentors.
Mixing positions Energy Lyapunov exponent
inputa λ efficiency (%)
Eccentricity Height
Centric mixing 1 0.228
Eccentric mixing (α = 2, h = 10.5) 1 0.301
Eccentric mixing (α = 4, h = 16) 1 0.437
a
Because the energy inputs were almost same at all mixing conditions, they were considered to be “1”.
b
Energy output in H2 and HAc was the fraction of H2 COD to total COD.
c
Energy output in HAc was the fraction of HAc COD to total COD.
d
Energy wasted was the fraction of unreacted glucose COD to total COD.
Lyapunov exponent became more positive, which was an indication of
chaotic flow.
The particle separation distance analysis was conducted for all nine
impeller heights and eccentricity values (Table 4). The Lyapunov ex-
ponent λ was the largest for the middle and top impeller positions and
at the maximum eccentricity α. For those two height positions there
exist a monotonic increase in λ as a function of α. This, however, did
not hold true for the bottom impeller position (the height of 5 cm), the
boundary effects were significant and affected the overall flow patterns.
In addition, for a specific eccentricity α the height effect appeared to be
inconclusive. A two-way ANOVA showed that the effect of impeller
eccentricity was significant (p value: .0447), while the effect of impeller
height was not significant (p value: .375) (Table 4).
3.4. Correlation of chaotic mixing simulation with AHP performance
The simulation results demonstrated well the correlation between
the mass distributions with the chaotic mixing positions in the AHP
biofermentor. At the height of 10.5 cm and 16 cm (minimal boundary
effects), the Lyapunov exponent increased with the mixing eccentricity,
representing a more eccentric mixing condition. Due to the boundary
effects from the bottom of the AHP biofermentor, the chaotic mixing at
the height of 5 cm exhibited the opposite behavior. The Lyapunov ex-
ponent also increased with the mixing height at eccentric mixing con-
ditions. These simulation results corresponded nicely to experimental
results. At the height of 10.5 cm and 16 cm, the COD removal efficiency
and hydrogen production increased with the mixing eccentricity, while
it did not relate with the mixing eccentricity at the height of 5 cm. At
eccentricities of 2 cm and 4 cm, the COD removal efficiency and hy-
drogen production also increased with the mixing heights, but did not
relate with the mixing heights for the centric mixing condition. It was
assumed that the flow mixing increased with the mixing eccentricity
and height, leading to more eccentric mixing conditions (higher
Lyapunov exponent λ value) for sufficient mixing of biomass particles
in the AHP biofermentor, so that the glucose and organic acids could be
utilized efficiently. This assumption was validated by the strong posi-
tive correlations between λ (the simulated results) and COD removal
efficiency and gas production (AHP experimental results, with the
strong boundary effects at the height of 5 cm being excluded) (Fig. 6). A
T-test on significance of slope was conducted and the result showed that
COD removal efficiency as a function of λ was significant (p value:
.02 < p < .05) whereas SHY as a function of λ was only borderline
significant (p value: .1 < p < .2).
The impact of eccentric mixing conditions on the AHP process was
explored in this study to gain a better understanding of the potential for
enhanced bioenergy production in biofermentors. The correlation be-
tween the Lyapunov exponent λ (a proxy for the intensity of chaotic
mixing) and energy output (biogas and liquid products) in AHP pro-
cesses was established under different mixing positions (Table 5). At the
same energy inputs (mixing rpm, mixing duration, and influent COD
concentration), a more eccentrically stirred mixing condition (asso-
ciated with a larger Lyapunov exponent λ value) led to a higher COD
Chemical Engineering Journal 334 (2018) 1383–1391
COD removal Specific hydrogen Energy output in Energy output in Energy
yields H2b HAcc wastedd
29.13 1.873 0.156 0.126 0.092
34.07 1.945 0.162 0.144 0.056
46.24 1.951 0.163 0.158 0.002
removal efficiency, high specific hydrogen yield (SHY, mole H2/mole
glucose) and higher energy outputs (biogas phase and liquid product
phase) and less wasted energy (unreacted glucose). Under the laminar
fluid flow, the mixture quality increased at the eccentrically stirred
mixing positions, allowing for a more sufficient contact between glu-
cose and biomass particles, leading to more H2 and HAc output in
higher λ value mixing conditions. Because the acetic acid fermentation
pathway has the highest theoretical hydrogen output (4 mol H2/mole
glucose, Eq. (3)), the energy output of acetic acid favored the energy
output of hydrogen gas. The energy output of acetic acid increased
more with λ than hydrogen gas (Table 5), since a portion of hydrogen
gas was consumed by other chemical reactions (Eq. (5)), and hydrogen
gas trapped in the liquid solution of the AHP biofermentor.
However, the eccentric mixing positions in this study were con-
ducted in a batch-mode, lab-scale biofermentor (volume: 2 L, radius of
biofermentor: 6.5 cm, radius of impeller: 2 cm). The impeller to bio-
fermentor ratio was about 0.3 in this study, which is larger than a ty-
pical bioreactor in real-world wastewater treatment plants (ratio: 0.18
[40]). Higher ratio of the impeller would make the boundary effects
more significant and lower the efficiency of eccentric mixing. There-
fore, better results are anticipated in a larger biofermentor. However,
the effectiveness of eccentric mixing for scale-up AHP systems needs to
be validated in future studies. Furthermore, the biomass particles (or-
ganic soil) used in this study were greatly benefited from the chaotic
mixing, while bacterial cells (colloids) and soluble reagents normally
used in anaerobic systems might be less affected by the chaotic mixing.
4. Conclusions
The use of eccentric mixing to enhance mass distribution and
bioenergy outputs in AHP biofermentors was for the first time, com-
prehensively studied through both experimental and modeling metho-
dology. Eccentric mixing regimes were achieved under different mixing
height and eccentricity. A numerical finite element model showed that
the Lyapunov exponent λ values (strength of chaotic mixing) increased
as a function of eccentricity at heights of 10.5 cm and 16 cm, and in-
creased as a function of height at eccentricities of 2 cm and 4 cm. Due to
boundary effects, the Lyapunov exponent λ values were irregular at the
bottom position (the height of 5 cm). The experimental results of the
batch mode AHP tests showed that COD removal efficiency and hy-
drogen gas production had a strong positive correlation with the λ
values. Mass distribution became more sufficient and the AHP reactions
proceeded more efficiently under chaotic mixing conditions. This lab-
scale study clearly demonstrated that eccentrically stirred mixing ac-
celerated the interaction between biomass particles and organic sub-
strates and ultimately enhanced the AHP processes.
Acknowledgement
The graduate student (Yuankai Huang) is supported by the China
Scholarship Council (CSC) Doctorate Program.
1390
Y. Huang et al.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in the
online version, at http://dx.doi.org/10.1016/j.cej.2017.11.088.
References
[1] A. Ghimire, L. Frunzo, F. Pirozzi, et al., A review on dark fermentative biohydrogen
production from organic biomass: process parameters and use of by-products, Appl.
Energy 144 (2015) 73–95.
[2] J. Zhong, D.K. Stevens, C.L. Hansen, Optimization of anaerobic hydrogen and me-
thane production from dairy processing waste using a two-stage digestion in in-
duced bed reactors (IBR), Int. J. Hydrogen Energy 40 (45) (2015) 15470–15476.
[3] Y. Chen, J.J. Cheng, K.S. Creamer, Inhibition of anaerobic digestion process: a re-
view, Bioresour. Technol. 99 (10) (2008) 4044–4064.
[4] Y. Xue, H. Liu, S. Chen, et al., Effects of thermal hydrolysis on organic matter so-
lubilization and anaerobic digestion of high solid sludge, Chem. Eng. J. 264 (2015)
174–180.
[5] F. Passos, J. Carretero, I. Ferrer, Comparing pretreatment methods for improving
microalgae anaerobic digestion: thermal, hydrothermal, microwave and ultrasound,
Chem. Eng. J. 279 (2015) 667–672.
[6] X. Fonoll, S. Astals, J. Dosta, et al., Anaerobic co-digestion of sewage sludge and
fruit wastes: evaluation of the transitory states when the co-substrate is changed,
Chem. Eng. J. 262 (2015) 1268–1274.
[7] S. Ghanimeh, M. El Fadel, P. Saikaly, Mixing effect on thermophilic anaerobic di-
gestion of source-sorted organic fraction of municipal solid waste, Bioresour.
Technol. 117 (2012) 63–71.
[8] P. Kaparaju, I. Buendia, L. Ellegaard, et al., Effects of mixing on methane production
during thermophilic anaerobic digestion of manure: lab-scale and pilot-scale stu-
dies, Bioresour. Technol. 99 (11) (2008) 4919–4928.
[9] Z.P. Zhang, K.Y. Show, J.H. Tay, et al., Rapid formation of hydrogen-producing
granules in an anaerobic continuous stirred tank reactor induced by acid incuba-
tion, Biotechnol. Bioeng. 96 (6) (2007) 1040–1050.
[10] Y. Hou, D. Peng, B. Wang, et al., Effects of stirring strategies on the sludge gran-
ulation in anaerobic CSTR reactor, Desalin. Water Treat. 52 (34–36) (2014)
6348–6355.
[11] A.D. Stroock, S.K.W. Dertinger, A. Ajdari, et al., Chaotic mixer for microchannels,
Science 295 (5555) (2002) 647–651.
[12] H. Aref, Stirring by chaotic advection, J. Fluid Mech. 143 (1984) 1–21.
[13] P. Hall, D.T. Papageorgiou, The onset of chaos in a class of Navier-Stokes solutions,
J. Fluid Mech. 393 (1999) 59–87.
[14] A.C. Bagtzoglou, N. Assaf-Anid, R. Chevray, Effect of chaotic mixing on enhanced
biological growth and implications for wastewater treatment: a test case with
Saccharomyces cerevisiae, J. Hazard. Mater. 136 (1) (2006) 130–136.
[15] A.C. Bagtzoglou, P.M. Oates, Chaotic advection and enhanced groundwater re-
mediation, J. Mater. Civil Eng. 19 (1) (2007) 75–83.
[16] A.C. Bagtzoglou, A. Novikov, Chaotic behavior and pollution dispersion char-
acteristics in engineered tidal embayments: a numerical investigation, JAWRA J.
Am. Water Resour. Assoc. 43 (1) (2007) 207–219.
[17] P. Zhang, S.L. DeVries, A. Dathe, et al., Enhanced mixing and plume containment in
porous media under time-dependent oscillatory flow, Environ. Sci. Technol. 43 (16)
(2009) 6283–6288.
[18] J.M. Zalc, F.J. Muzzio, Parallel-competitive reactions in a two-dimensional chaotic
flow, Chem. Eng. Sci. 54 (8) (1999) 1053–1069.
[19] F.J. Muzzio, M. Liu, Chemical reactions in chaotic flows, Chem. Eng. J. Biochem.
1391
Chemical Engineering Journal 334 (2018) 1383–1391
Eng. J. 64 (1) (1996) 117–127.
[20] H. Aref, J.R. Blake, M. Budišić, et al., Frontiers of chaotic advection, Rev. Mod.
Phys. 89 (2) (2017) 025007.
[21] C. Ratanatamskul, T. Saleart, Effects of sludge recirculation rate and mixing time on
performance of a prototype single-stage anaerobic digester for conversion of food
wastes to biogas and energy recovery, Environ. Sci. Pollut. Res. 23 (8) (2016)
7092–7098.
[22] H. Ameur, Effect of the shaft eccentricity and rotational direction on the mixing
characteristics in cylindrical tank reactors, Chin. J. Chem. Eng. 24 (12) (2016)
1647–1654.
[23] S.C. Low, N. Eshtiaghi, P. Slatter, et al., Mixing characteristics of sludge simulant in
a model anaerobic digester, Bioprocess Biosyst. Eng. 39 (3) (2016) 473–483.
[24] Y. Fan, R.I. Tanner, N. Phan-Thien, A numerical study of viscoelastic effects in
chaotic mixing between eccentric cylinders, J. Fluid Mech. 412 (2000) 197–225.
[25] P. Dutta, R. Chevray, Inertial effects in chaotic mixing with diffusion, J. Fluid Mech.
285 (1995) 1–16.
[26] X. Gomez, M.J. Cuetos, J. Cara, et al., Anaerobic co-digestion of primary sludge and
the fruit and vegetable fraction of the municipal solid wastes: conditions for mixing
and evaluation of the organic loading rate, Renewable Energy 31 (12) (2006)
2017–2024.
[27] G. Lettinga, The anaerobic treatment approach towards a more sustainable and
robust environmental protection, Water Sci. Technol. 52 (1–2) (2005) 1–11.
[28] C.H. Ting, D.J. Lee, Production of hydrogen and methane from wastewater sludge
using anaerobic fermentation, Int. J. Hydrogen Energy 32 (6) (2007) 677–682.
[29] C. Li, H.H.P. Fang, Fermentative hydrogen production from wastewater and solid
wastes by mixed cultures, Crit. Rev. Environ. Sci. Technol. 37 (1) (2007) 1–39.
[30] Y. Sharma, B. Li, Optimizing energy harvest in wastewater treatment by combining
anaerobic hydrogen producing biofermentor (HPB) and microbial fuel cell (MFC),
Int. J. Hydrogen Energy 35 (8) (2010) 3789–3797.
[31] B. Rincón, R. Borja, J.M. González, et al., Influence of organic loading rate and
hydraulic retention time on the performance, stability and microbial communities
of one-stage anaerobic digestion of two-phase olive mill solid residue, Biochem.
Eng. J. 40 (2) (2008) 253–261.
[32] S.V. Mohan, V.L. Babu, P.N. Sarma, Anaerobic biohydrogen production from dairy
wastewater treatment in sequencing batch reactor (AnSBR): effect of organic
loading rate, Enzyme Microb. Technol. 41 (4) (2007) 506–515.
[33] A.C. Basagiannis, X.E. Verykios, Catalytic steam reforming of acetic acid for hy-
drogen production, Int. J. Hydrogen Energy 32 (15) (2007) 3343–3355.
[34] C. Galletti, E. Brunazzi, On the main flow features and instabilities in an unbaffled
vessel agitated with an eccentrically located impeller, Chem. Eng. Sci. 63 (18)
(2008) 4494–4505.
[35] M.M. Alvarez, J.M. Zalc, T. Shinbrot, et al., Mechanisms of mixing and creation of
structure in laminar stirred tanks, AIChE J. 48 (10) (2002) 2135–2148.
[36] K.C. Ng, E.Y.K. Ng, Laminar mixing performances of baffling, shaft eccentricity and
unsteady mixing in a cylindrical vessel, Chem. Eng. Sci. 104 (2013) 960–974.
[37] Y. Sharma, B. Li, Optimizing hydrogen production from organic wastewater treat-
ment in batch reactors through experimental and kinetic analysis, Int. J. Hydrogen
Energy 34 (15) (2009) 6171–6180.
[38] J. Wang, W. Wan, Comparison of different pretreatment methods for enriching
hydrogen-producing bacteria from digested sludge, Int. J. Hydrogen Energy 33 (12)
(2008) 2934–2941.
[39] F. Duarte, R. Gormaz, S. Natesan, Arbitrary Lagrangian-Eulerian method for Navier-
Stokes equations with moving boundaries, Comput. Methods Appl. Mech. Eng. 193
(45) (2004) 4819–4836.
[40] I. Angelidaki, L. Ellegaard, B.K. Ahring, Applications of the Anaerobic Digestion
Process[M]//Biomethanation II, Springer, Berlin Heidelberg, 2003, pp. 1–33.