Brazilian Journal of Chemical Engineering

https://doi.org/10.1007/s43153-024-00494-4

ORIGINAL PAPER

CFD analysis of multiphase flow in an airlift reactor: superficial velocity

and gas holdup influence on the loop recirculation
Jamille Coelho Coimbra1 · Pedro Henrique Rocha Batista2 · Diego Gaspar Souza Paz2 · Patrick Silva Oliveira3 ·
Diego Martinez Prata4

Received: 5 March 2024 / Revised: 27 May 2024 / Accepted: 18 June 2024

© The Author(s) under exclusive licence to Associação Brasileira de Engenharia Química 2024

Abstract
Several products of great importance in the food and pharmaceutical industries are produced by biotechnological processes,
using high conversion and high specificity chemical reactions carried out by microorganisms. Such processes can be con-
ducted in airlift reactors (ALRs), which are capable of operating multiphase systems. The ALRs are pneumatically agitated,
have a high surface area of contact between the phases and have satisfactory homogenization, favoring the transfer of mass
and energy. However, thus yet, the comprehensive investigation of the relationships among flow patterns, dead zones, gas
holdup, phase velocity profiles, and the loop recirculation has not been extensively investigated. The purpose is to analyze
the dynamics of multiphase flow at a local level, which is scarcely explored in the literature. This will provide a complete
understanding of the recirculation pattern found in the ARL reactor. In this work, Computational Fluid Dynamics (CFD)
technique was used to study the effects of different air inlet velocities (0.27 m/s, 0.36 m/s, 0.45 m/s and 0.54 m/s) and the
average Sauter diameter of the bubbles injected by the sparger (12 mm, 15 mm, 18 mm and 21 mm) on the type of flow
(stagnant, recirculation and preferential path zones), gas holdup, alpha parameter (α) and phase velocity profiles. The model
was validated through experimental data. The simulation results indicated that the higher the air velocity injected by the
sparger, the greater the amount of gas in the riser relative to the downcomer, which results in a lower alpha parameter (use-
ful information for reactor sizing). It was also observed that the velocity profile of the liquid phase is closely related to that
of the gas phase. The operating conditions that provided the most satisfactory results (better mixing time, more effective
homogenization, and better establishment of the recirculation circuit) were obtained for an inlet air velocity of 0.54 m/s and
an average Sauter bubble diameter of 21 mm.

Keywords Airlift reactor · CFD · Biotechnological processes · Modeling · Simulation

List of symbols u Phase velocity (m ­s−1)

ρ Density (kg ­m−3) 𝜇𝛽,eff Effective viscosity (Pa s)
r Volume fraction (dimensionless) 𝜇𝛽 Molecular viscosity (Pa s)
β Phase (liquid/gas) 𝜇𝛽T Dynamic viscosity (Pa s)
𝜇𝛽s,T Shear-induced turbulence viscosity (Pa s)
* Jamille Coelho Coimbra 𝜇𝛽b,T Bubble-induced turbulence viscosity (Pa s)
[email protected] MI,𝛽 Interphase momentum force (N ­m−3)
1 ∇P Pressure gradient (Pa ­m−1)
Departament of Chemical Engineering, Universidade
Federal de Minas Gerais, Belo Horizonte, MG 31270‑901, g Gas phase
Brazil l Liquid phase
2
Department of Chemical Engineering, Universidade Federal 𝜎 Interfacial tension (N ­m−1)
de Viçosa, Viçosa, MG 36570‑000, Brazil Vg Gas volume in the reactor ­(m3)
3
Department of Numerical Simulation – CFD, Universidad Vl Liquid volume in the reactor ­(m3)
Politécnica de Madrid, Calle Pastor, 3, 28003 Madrid, Spain 𝜀g Gas holdup (dimensionless)
4
Department of Chemical and Petroleum Engineering, us Gas velocity in the sparger (m ­s−1)
Universidade Federal Fluminense, Niterói, RJ 24220‑900, ud Gas velocity in the draft tube (m ­s−1)
Brazil

1 3 Vol.:(0123456789)
 Brazilian Journal of Chemical Engineering

Introduction surface velocity and bubble size. Sikula and Markoš (2008)
emphasized that the riser has a diversified geometry as it
Chemical reactions commonly use airlift reactors (ALRs), depends on the desired type of fluid dynamic regime in the
especially in the pharmaceutical and food industries and process. The most common structure, the draft tube, has
other biotechnological processes as well as wastewater the shape of a hollow cylinder without any of its bases. It is
treatment sector (Li et al. 2022; Naidoo et al. 2021; Zhang located in the central region of the reactor, near the sparger
et al. 2022). Among these reactions, the use of microalgae and its function is to direct the gas stream to the top area.
in ALRs to capture C ­ O2 in treating aqueous waste from Recently, the influence of the sparger and its design on the
chemical process industries (Abid et al. 2017) or bio-fuels gas holdup and mass transfer has been demonstrated (Naidoo
production (Li et al. 2022) have stood out, which are both et al. 2021). The gas separator removes from the system,
in alignment with the sustainable development goals pro- fully or partially, the remaining fraction of the incoming
posed by the United Nations in the 2030 Agenda. From this gas stream that reaches the top of the reactor. Therefore, it
equipment, it is possible to obtain a high conversion of the controls the number of bubbles remaining inside the equip-
limiting reagent with high specificity because its operating ment after reaching the top, which will go to the downcomer
principle promotes the mass transfer of oxygen in an intense afterward (Klein et al. 2001).
way, favoring the aerobic metabolism of the cells, in addition According to Vial et al. (2002), the downcomer is the
to increasing the overall control of the process fluid dynam- section responsible for directing the fluid to the base. It can
ics (Sikula and Markoš 2008). Therefore, it avoids problems have two configurations: be integrated into the main part of
such as dead zones and preferential pathways. This char- the reactor, which configures an internal loop airlift reactor
acteristic intensifies the reaction kinetics, which increases (IL-ALR) or be designed separately from it. In this case,
the yield as well as decreases the residence and space times the equipment is called an external loop airlift reactor (EL-
without causing damage by shear to the microorganisms ALR). These configurations cause the gas phase dynamics
responsible for the reactions (Zhang et al. 2017). to differ throughout the system (Vial et al. 2002). Thus, the
The ALRs have different geometries based on the type downcomer is essential for establishing the fluid dynamic
of process that uses them. They are, however, always com- regime of interest. In turn, the base region is the part of the
posed of five main parts (Fig. 1): sparger, riser, gas sepa- reactor in which the fluid from the downcomer meets the
rator, downcomer and base region (Freitas et al. 1999; Lu gas stream introduced by the sparger (Zhang et al. 2017).
et al. 1994). Each one plays a primary role in the operating In it, one flow cycle ends and another begins. According to
principle of the reactor and has great influence on the type Hu and Zhong (2001), the design of the base region is very
of fluid dynamic regime established in the system (Sikula important since it influences the type of flow established and
and Markoš 2008; Zhang et al. 2017). This fluid dynamic the mixing efficiency. In turn, gas hold-up in the downcomer
regime will define the area and time of contact between the increases in response to changes in the volumetric flow of
phases and the mass, energy and momentum transfer (Zhang gas or specific geometric parameters of the ARL, such as
et al. 2017). Therefore, the magnitude of the driving force an increase in the diameter of the draft tube, as stated by
that promotes these phenomena depends on the design of Lestinsky et al. (2012).
each part. The ALRs are relatively simple apparatus that require
The sparger is the apparatus that introduces the gas minimal space both in the industrial plant and in the labo-
stream into the system (Naidoo et al. 2021), which allows ratory scale. Some particular characteristics, such as easy
controlling the feed flow rate and, consequently, the gas operation, low-cost maintenance, low energy consumption,
high mixing performance, and the absence of moving parts
to promote homogeneous agitation and low turbulence, have
made them increasingly employed (Abashar et al. 1998;
Jasim et al. 2022; Li et al. 2022). Due to these character-
istics, multiphase processes frequently employ this type
of reactor, especially regarding reactions involving highly
shear-sensitive microorganisms (Klein et al. 2001; Sikula
and Markoš 2008; Zhang et al. 2017).
In addition, the pressure drop in this equipment is lower
when compared to the Plug Flow Reactor (PFR) and the
Continuous Stirred Tank Reactor (CSTR), which facilitate
the control of the flow rate and reduce the pump power
Fig. 1  The five main parts of an airlift reactor: A (sparger); B (riser); demand (Abashar et al. 1998). This reactor can operate
C (gas separator); D (downcomer); E (base) safely at high pressures, which increases the solubility of

13
Brazilian Journal of Chemical Engineering

oxygen in the liquid phase (Klein et al. 2001; Sikula and assessed the kind of gas distributor as well as the location
Markoš 2008). Another advantage is that temperature con- and direction of injection. However, thus yet, the compre-
trol can be performed simply and accurately due to the effi- hensive investigation of the relationships among flow pat-
cient homogenization system (Artok and Schobert 2000). terns, dead zones, gas holdup, phase velocity profiles, and
According to Huang et al. (2015), if solid particles are pre- loop recirculation has not been extensively investigated.
sent, such as catalysts, they are not considered a hindrance The bubble diameter and superficial velocity ranges
because the operating parameters are easily modified in of the gas phase examined in this study have not been
ALRs so that the fluid flow rate can be adjusted so that its established in the existing literature, especially for higher
rising velocity is higher than the terminal velocity of the values (Table 1, Appendix). According to Shi et al. (2021),
solid particle making it possible to establish a fluidized bed there is a direct relationship between the rise in superficial
condition when necessary. gas velocity and the increase in bubble size. An increase
Since laboratory and industrial processes are becoming in bubble diameter and gas velocity might result in the
increasingly complex, there is a need to search for more suit- formation of bubble clusters, especially in the upper part
able and efficient reactor models (Zhang et al. 2017). With of the riser, which can disrupt the flow pattern (recircula-
recent computational advances, process modeling and refin- tion regime). This phenomenon remains unclear, probably
ing discretization meshes have become simpler to develop because the literature has not examined the particular com-
(Fletcher 2022). In this context, the Computational Fluid binations of bubble diameter and gas velocity that could
Dynamics (CFD) modeling technique proves to be extremely lead to this effect. Therefore, this paper sheds light on
useful since, by employing the fundamental equations of these issues.
fluid dynamics, kinetics expressions, mass, energy, and Thus, this research proposes a comprehensive study
momentum balances, it is possible to model and predict the of an airlift reactor through CFD modeling tool utilizing
behavior of a multiphase phenomenon (Raman et al. 2018). transport equations of momentum for the multiphase flow
Modeling using CFD allows a local analysis of phase flow, of air and water mixture using a reference design (Sikula
mixing, and holdup. Such information allows for determin- and Markoš 2008) and also experimental data to perform
ing flow patterns (recirculation zones, dead zones) and mix- the CFD model validation. The purpose is to analyze
ing profiles (Fletcher 2022). Because this method performs the dynamics of multiphase flow at a local level, which
the calculations considering each point’s characteristics and is scarcely explored in the literature. This will provide a
how they vary over time, its predictions are very close to complete understanding of the recirculation pattern found
what occurs in reality. This way, it will be possible to better in the ARL reactor. Investigations will be conducted into
understand the functioning of ALRs and optimize existing multiphase flow patterns, including the visualization of
designs. liquid and gas velocity profiles, dead zones, recirculation
A mixing model for recirculating flow in an IL-ALR zones, and design variables such as the alpha parameter
was proposed by Hari and Tan (2002), Wang et al. (2004), and gas volume fraction at varying flow rates and bubble
Cao et al. (2009), Silva et al. (2011), and García et al. sizes.
(2012) assessed gas holdup profiles. Šimčík et al. (2011)
used a constant bubble size to analyze the field flow. More
specifically Ghasemi et al. (2012) examined the transi-
tion of the flow regime from stagnation to recirculation at
various flow rates. The variation in surface gas velocity
affects bubble size, although it was assumed to remain Table 1  Mathematical model representing ALR’s multiphasic flow
constant. From another perspective, Pawar (2017) used
an Eulerian–Lagrangian technique to study bubble size 𝜕(𝜌r)𝛽
+ ∇ ⋅ (𝜌ru)𝛽 = 0;𝛽 = l, g
𝜕t ( ( ))
distributions in the ARL domain. Shi et al. (2021) sug- 𝜕 (r𝛽 𝜌𝛽 u𝛽 )
+ ∇ ⋅ r𝛽 𝜌𝛽 u𝛽 u𝛽 = −r𝛽 ∇P + r𝛽 𝜌𝛽 g + ∇ ⋅ r𝛽 𝜇𝛽,eff ∇u𝛽 + ∇uT𝛽 + MI,𝛽
𝜕t
gested altering the traditional design of the airlift reac- rl + rg = 1
tor by incorporating a contraction–expansion guide vane Vg
𝜀g =
to induce the formation of two recirculation zones and Vg +Vl

enhance mixing. Similarly, using CFD as design tool, Li 𝜇𝛽,eff = 𝜇𝛽 + 𝜇𝛽T

et al. (2020) presented a CFD model to forecast the mass 𝜇gT =
𝜌g 𝜇lT

transfer and mixing time for cell culture. From this per- 𝜌l 𝜎

spective, Ramonet et al. (2022) proposed various designs 𝜇𝛽T = 𝜇𝛽s,T + 𝜇𝛽b,T
for single-stage and multi-stage internal circuit ALRs to MI,g = −MI,l
investigate flow and mixing patterns. In order to enhance MI,l = −MI,g = MD,l + ML,l + MVM,l + MTD,l
mixing and gas–liquid dispersion, Qiao et al. (2023)

13
 Brazilian Journal of Chemical Engineering

Materials and methods turbulence of the dispersed phase is significantly smaller due
to its modest density. Thus, the dynamic viscosity of the gas
Mathematical modeling phase ( 𝜇gT ) is limited by the flow of the continuous phase
(𝜇 Tl ) according to Eq. (6). In multiphase flow, the dynamic
The CFD technique uses theoretical methods and mathemat- viscosity of the continuous phase also has a contribution of
ical equations to calculate the transport phenomena within shear-induced turbulence (𝜇𝛽s,T ) (estimated from continuous
the system. The method used is the finite volume method, phase) and bubble-induced turbulence (𝜇𝛽b,T ) (estimated from
where the geometry, previously defined, is divided into dispersed phase) (Eq. 7). The interfacial momentum transfer
smaller parts (control volumes), forming a discrete mesh. ( MI,β ) is given by Eq. (8), which occurs due to viscous stress
The variables describing the fluid motion are then calculated distribution across the interface. The force exerted by the gas
according to this mesh (Jasim et al. 2022). on the liquid is the same as that exerted by the liquid on the
The Navier–Stokes equations were solved using Ansys gas, but in opposite directions. From bubble-driven flows,
CFX software (version 19 R1). The Eulerian-Eulerian the most essential forces to be examined are: draft force, lift
method, in which both phases are considered continuous, force, virtual mass and turbulent dispersion force (Eq. 9)
was used to describe the multiphase behavior following the (Van Baten et al. 2003).
assumptions: In the equations presented, 𝜌 represents density, r repre-
sents volumetric fraction, 𝛽 is the phase ( g for gas and l for
• Mass transfer between phases was disregarded; liquid) and u , the phase velocity. The parameters μβ,eff , μβ ,
• Both phases are treated as interpenetrating; μTβ , μs,T , μb,T and MI,β are associated with an 𝛽 phase and cor-
• The fluids are incompressible; β β
respond to effective viscosity, molecular viscosity, dynamic
• Uniform pressure field;
viscosity, shear-induced turbulence, bubble-induced turbu-
• Only spherical, permanent bubbles of uniform size con-
lence and interphase momentum force, respectively. The
stitute the gas phase;
terms ∇P, 𝜎, Vg, Vl, 𝜀g, us and ud indicate pressure gradient,
• Turbulence in the dispersed phase is insignificant, and
interfacial pressure, the gas volume in the reactor, the liquid
the k – 𝜖 model is used for the continuous phase (with
volume in the reactor, gas holdup, gas velocity in the sparger,
medium intensity);
and gas velocity in the draft tube, respectively.
• Isothermal condition.

The conservation equations for multiphase flow in ALR Model setup

are shown in Table 1 according to Luo and Al-dahhan
(2011). Equations (1) and (2) are the continuity and momen- The setup of different simulation scenarios is present in
tum, known as the Navier Stokes equation for multiphase Table 2. Through a correlation between bubble diameter
flow. The terms on the left-hand side of the Eq. (2) indicate and gas surface velocity, described by Luewisutthichat
local and advective acceleration forces, respectively. The et al. (1997), bubble size in each simulation can be esti-
terms on the right-hand side indicate the pressure, gravita- mated (Table 2).
tional, viscous stress, and interfacial forces, respectively, The interphase transfer model was the particle model and
acting on the fluid element in each phase. Equation (3) rep- the surface tension coefficient used was 0.072 N/m. Drag
resents a constraint on the volume fractions of the two force arises when bubbles travel at various velocities in a
phases occupying the reactor. Equation (4) is the gas holdup, continuous phase. It is the very relevant interfacial force,
which is defined as the volume fraction occupied by the gas. which depends on the drag coefficient considered constant
Equation (5) represents the molecular viscosity ( 𝜇𝛽 ) and in this work (= 0.9). The standard k – 𝜖 model is frequently
dynamic viscosity (𝜇𝛽T ) contributions to effective viscosity utilized in multiphase flow, displaying stability and accu-
(𝜇𝛽,eff ). The dynamic viscosity due to turbulent effect associ- racy (Zhang et al. 2022). The model for bubble-induced
ated with two phases must be considered. However, the

Table 2  Pre-defined simulation Variable Simulation 1 Simulation 2 Simulation 3 Simulation 4

variables
ud (m/s) 0.015 0.020 0.025 0.030
us (m/s) 0.27 0.36 0.45 0.54
Bubble size (mm) 12 15 18 21

13
Brazilian Journal of Chemical Engineering

turbulence was Sato Enhanced Eddy Viscosity. In this Table 3  Simulation boundary conditions
model, the dispersed phase influences the continuous Boundary Condition
phase by making an extra contribution to turbulent viscos-
ity. Further mathematical details of these two models can Sparger inlet Specified (0.27;
0.36; 0.45;
be obtained from Luo and Al-dahhan (2011) and Ma et al.
0.54 m/s)
(2015).
Oulet Degassing
For both the steady-state and transient simulations, the
External ALR Gas:free slip
first-order Upwind scheme as well as the convergence stop Water: non-slip
criterion based on the Root Mean Square (RMS) metric Draft tube Gas:free slip
equivalent to 10–4 were adopted. Specifically, in the tran- Water: non-slip
sient regime setup, the resolution method was the second-
order backward Euler scheme, because of its great numerical
stability (Zhu et al. 2016). The simulations were performed Results and discussion
using a reference geometry (Fig. 2), described by Sikula and
Markoš (2008) and built in the SpaceClaim tool. Spatial independence test
The boundary conditions used in the simulations are pre-
sented in Table 3. It was considered that the reactor con- The number of control volumes and elements comprising
tained only water in the initial condition (t = 0). the mesh discretizes the computational domain, as defined
The ALR adopts the proportions shown in Fig. 3. by the spatial and time independence tests. The comparison
It was then necessary to discretize the geometry for parameter selected to perform the tests was the air superfi-
application of the finite volume method, creating a mesh cial velocity. Three meshes were tested, each with different
to be used in solving the continuity and momentum equa- amounts of control volume and elements. The first was the
tions, which must be properly verified by spatial and time program’s standard mesh with 57,938 elements (mesh 0),
independency tests. A mesh test for spatial independence and the others were more refined meshes with 77,096 ele-
was performed in which 5 triangular meshes with different ments (mesh 1) and 345,063 elements (mesh 2). The test
numbers of control volumes. The elements were created and aimed to verify if refining the mesh any further would alter
compared based on the relative error regarding the values of the results of the simulation. Its results (Fig. 5) indicated that
air superficial velocity at the central position of the reactor mesh 2 (345,063 elements) is adequate for this study since
(= ­(us,1–us,2)/us,1, where 1 and 2 represent, respectively, its values presented an error of less than 1% relative to the
the meshes with lower and higher refinement), defined as previous mesh (mesh 1).
1% at maximums. Timestep tests for time independence,
based on the same criteria, were also performed, comparing Time independence test
timestep values of 0.5, 1 and 2 s. The acceptance criterion
for the mesh independence test followed the methodology The timestep determination was similar to that of the
proposed by Luo et al. (2020). The best mesh is represented mesh and, three tests of 0.5, 1, and 2 s were performed,
by 71,325 control volumes and 345,063 elements, as illus- with the air superficial velocity as the analysis parameter.
trated in Fig. 4. The timestep defined was 2 s. The smaller the timestep, the greater the number of itera-
tions made by the program and, theoretically, the greater

Fig. 2  Geometry created with

SpaceClaim

13
 Brazilian Journal of Chemical Engineering

Fig. 3  Proportions of the geom-

etry used

Fig. 4  Best mesh after spatial

independency test

the accuracy of the results. However, the simulation time Model validation
increases as the timestep decreases. Therefore, this analy-
sis aims to define the longest timestep that will provide This section presents the validation of the mathematical
results with good accuracy, close to those of the shortest CFD model. The goal is to demonstrate its accuracy after
timestep. For the tests performed in this study, the results defining the appropriate discrete mesh against experimen-
(Fig. 6) for the timestep of 2 s (shortest simulation time) tal data and previously published simulation results (Blazej
were very close to those of 0.5 s (relative error < 1%). 2004a). Therefore, the CFD model may be used with confi-
Thus, this work used a timestep of 2 s. As the ratio of the dence, and the results may be considered credible for further
simulation time (600 s) to the timestep gives the number applications and designs.
of iterations, 300 iterations were performed for each of the To allow a fair comparison the same geometric specifica-
four operational conditions studied. tions reported in Blazej (2004a) were adopted, and the gas

13
Brazilian Journal of Chemical Engineering

coefficient and Favre averaged drag force, respectively. In

turn, the model suggested by Blazej et al. (2004a) neglected
the turbulent dispersion force and considered the Grace
drag equation, respectively. In addition, the discrete mesh
and solver configurations of both works were distinct. By
employing a quantitative and comparative examination of
the proposed model with the experimental data from Blazej
(2004a), the subsequent observations were deduced.
The predicted data exhibited a strong agreement with
the experimental data in all presented simulations (Fig. 7),
showing a Sum of the Squared Error (SSE) (Eq. 10) of
4.51 × ­10–2 (m/s)2, 2.51 × ­10–2 (m/s)2, 9.35 × ­10–4 (m/s)2,
and 1.06 × ­10–4 (m/s)2 for Fig. 7A–D, respectively. The Root
Mean Square Error (RMSE) (Eq. 11) for Fig. 7A–D were
5.68 × ­10–2, 4.23 × ­10–2, 8.17 × ­10–3, and 2.76 × ­10–3, respec-
tively. The data predicted by Blazej (2004a), on the other
hand, did not fit as well, as shown by the higher Root Mean
Fig. 5  Mesh Analysis Graph: mesh 0 (57,938 elements); mesh 1 Square Error (RMSE) values of 6.32 × ­10–2, 1.29 × ­10–1,
(77,096 elements); mesh 2 (345,063 elements) 1.60 × ­10–2, and 1.81 × ­10–2 for Fig. 7A–D, respectively.
Thus, since the CFD proposed model is able to capture
the actual dynamics of the system, it can be considered
validated.
n
∑ )2
(10)
(
SSE = yi − ŷi
i

√
√∑ y − ŷ 2
√ n ( )
RMSE =
√ i i
(11)
i
n

In Eqs. (10) and (11) yi represents the experimental value,

which is considered to be error-free due to the absence of
any information regarding measurement errors, according
to Blazej (2004a). On the other hand, ŷi represents the value
predicted by the model.

Gas holdup behavior

Fig. 6  Timestep analysis graph: 0.5 s (1200 iterations); 1 s (600 itera-
tions); 2 s (300 iterations)
The simulations allowed predicting the fluid dynamic behav-
ior established in the reactor for the transient regime from
velocity applied at the riser inlet was set at 0.018, 0.036, the gas and liquid velocity profiles and the gas volume frac-
0.072, 0.090, 0.108, and 0.135 m/s for a bubble air diameter tion. According to the operating principle of an airlift reac-
of 5 mm and a gas phase fraction of 1. tor, the air bubbles are expected to concentrate in the draft
Based on both simulation results, qualitatively, it is tube region when inserted into the system by the sparger. As
observed that the predicted values from the applied model they rise in this region of the reactor, the liquid is carried to
exhibit better overall agreement with the experimental data the degasser region. There, the removal of not-incorporated
than those results previously reported by Blazej (2004a) gas from the system occurs, and the liquid goes from the
(Fig. 7), especially for predicted profiles of liquid phase downcomer to the base of the reactor, where it again enters
velocity in the downcomer, gas holdup in the riser and the draft tube to establish a recirculation loop (Vial et al.
downcomer, as shown in Fig. 7B, C and D, respectively. 2002). The dependence of the Sauter mean diameter of the
The main difference between the models resides in the inter- bubbles on the velocity with which they are injected into the
phase force equations: specifically the drag force and tur- reactor and on the sparger geometry can be related, as pro-
bulent dispersion ones. The applied model defines the drag posed by Luewisutthichat et al. (1997). According to these

13
 Brazilian Journal of Chemical Engineering

Fig. 7  Experimental data (filled circle) obtained from Blazej (2004a) influenced by the air velocity in the riser from Blazej (2004a) (red
and predicted liquid phase velocity in the riser (A), in the downcomer lines) and from this work (blue lines)
(B), air volume fraction in the riser (C), and in the downcomer (D)

authors, the higher the entrance velocity of the gas phase, differences in gas holdup and not the desired recirculation
the larger the size of the bubbles will be, and this is a factor zones as the superficial air velocity at the inlet increases.
that influences the flow and the mass transfer regimes of This effect can be proven by simulating streamlines at the
the components of interest from the gas phase to the liquid top of the downcomer (Fig. 9), which may show that the
phase. higher the gas inlet velocity, the more likely it is that an
The simulation results corroborate those from García unwelcome small air circulation zone will form. To avoid
et al. (2012), since the air volume fraction is higher in the this type of problem, design parameters such as equipment
riser region near the sparger under all operating conditions geometry, inlet flow rate, and bubble size can be modified
(Fig. 8). Moreover, if the gas assimilation by the liquid is (Zhang et al. 2017).
high and the system operates efficiently, the air volume frac- Driving force of the recirculation is proportional to cross-
tion in the downcomer should be minimal. However, Fig. 8 sectional area ratio (AD/AR) and height-to-diameter ratio
shows a higher air volume fraction at the top of the down- (H/D) (Li et al. 2020). As AD/AR and H/D increase, gas
comer, indicating the formation of small zones with higher holdup correspondingly increases, hence, recirculation is
gas holdup. Stagnation zones may have been produced favored (Li et al. 2020; Merchuk 1990). Choi et al. (2021)
due to the limited area at the top ALR and even the bubble gathered the design specifications of 21 ALRs from the most
grouping. Generally, the ALR top is enlarged, perhaps to relevant studies in the literature. Among other parameters,
improve the degassing of the system. This results in local AD/AR varied from 0.108 to 4.928 while H/D varied from

13
Brazilian Journal of Chemical Engineering

1.567 to 27.027. In this work, the reference ALR have AD/

AR and H/D equal to 0.95 and 12.3, respectively. Thus, it
is probable that the design of the ALR encourages good
recirculation, which can be further improved by raising the
H/D ratio, although the scale required may restrict these
adjustments.
The parabolic and symmetric holdup profiles for differ-
ent air inlet flow rates and bubble sizes are presented in
Fig. 10. The gas–liquid distributions in the draft tube and
downcomer under different ranges of superficial gas veloci-
ties demonstrated similar profiles. The effects of virtual mass
force and wall reaction force explain this phenomenon, as
they cause the bubble plume to be more concentrated in
the center of the draft tube than in its walls (García et al.
2012). Near the injection (lines 1 and 2), as the liquid veloc-
ity increases, a symmetrical gas holdup profile is observed,
indicating that the gas bubbles rise uniformly. As you move
further away from the sparger injection, the flow becomes
more heterogeneous, possibly because the bubbles start to
collide with the draft tube walls (lines 3, 4, and 5). It was
able to identify the fully established (uniform gas–liquid
dispersion) at a height of 30 cm for all ranges of superficial
air velocity (Fig. 10).
An increase in the gas flow rate and velocity showed a
rise in the air volume fraction (Eq. 4), especially in the riser,
Fig. 8  Air volume fraction contours: A for an air inlet velocity of due to the greater convective effect compared to the down-
0.27 m/s and particle diameter of 12 mm; B for an air inlet velocity of
comer (Figs. 11 and 12). After running Simulation 4 (with
0.36 m/s and particle diameter of 15 mm; C for an air inlet velocity of
0.45 m/s and particle diameter of 18 mm; D for an air inlet velocity of an air speed of 0.54 m/s and an average Sauter diameter of
0.54 m/s and particle diameter of 21 mm 21 mm), the holdup profile showed better mixing because
there was more air in the draft tube (Fig. 10). This result
leads to a considerable difference in the specific mass of the
solution in the riser (which has more air) and downcomer

Fig. 9  Streamlines at the top

of the downcomer: A for an air
inlet velocity of 0.27 m/s and
particle diameter of 12 mm;
B for an air inlet velocity of
0.36 m/s and particle diameter
of 15 mm; C for an air inlet
velocity of 0.45 m/s and particle
diameter of 18 mm; D for an air
inlet velocity of 0.54 m/s and
particle diameter of 21 mm

13
 Brazilian Journal of Chemical Engineering

Fig. 10  Graphs of the air volume fraction at different positions in the 15 mm; C for an air inlet velocity of 0.45 m/s and particle diameter of
reactor: A for an air inlet velocity of 0.27 m/s and particle diameter of 18 mm; D for an air inlet velocity of 0.54 m/s and particle diameter
12 mm; B for an air inlet velocity of 0.36 m/s and particle diameter of of 21 mm

(which has less air), which supports the loop recirculation through the liquid as well as the greater the deceleration of
regime (Blažej et al. 2004a,b). the dispersed phase. Thus, for the same flow rate, the higher
According to Blažej et al. (2004a), the gas holdup pro- the reactor height, the more homogeneous the flow pattern
file in the riser is directly proportional to the superficial gas in the draft tube and the more uniform the mixing in this
velocity, because the slip due to relative velocity between 2 domain (Blažej et al. 2004b). Since the degasser removes
phases remains constant even with higher gas flow. Similar most unincorporated air, the gas velocity in the downcomer
profiles, close to linear, are expected in the downcomer, but region should be low because its volume fraction is minimal
with a less pronounced rise in gas holdup as the gas surface and the air has lost a significant amount of momentum dur-
velocity increases (Blažej et al. 2004a). ing the flow. It has already been seen that there is a certain
amount of air in the upper region of the downcomer due
Gas phase velocity profile behavior to the formation of unwanted recirculation zones (Figs. 8
and 9). However, due to the relatively low air velocity in
The velocity profiles (Figs. 13 and 14) show great depend- this region (up to 0.010 m/s) compared to the draft tube
ence on the position, and the higher it is, the greater the (up to 0.22 m/s), the presence of these unwanted recircu-
loss of momentum by the gas during its ascending flow lation zones does not significantly impact the gas velocity

13
Brazilian Journal of Chemical Engineering

Fig. 11  Graph of the average air volume fraction in the riser for dif-
ferent velocity values

Fig. 13  Air superficial velocity contours: A for an air inlet velocity of

0.27 m/s and particle diameter of 12 mm; B for an air inlet velocity of
0.36 m/s and particle diameter of 15 mm; C for an air inlet velocity of
0.45 m/s and particle diameter of 18 mm; D for an air inlet velocity of
0.54 m/s and particle diameter of 21 mm

efficiency is achieved due to the looped flow generated by

the pumping of air through the sparger as well as the exist-
ence of the draft tube and the downcomer, which promote
the directed flow of the fluids (Vial et al. 2002). Thus, the
air bubbles drag the stationary liquid phase to the top of the
equipment, where it flows through the downcomer to the
Fig. 12  Graph of the average air volume fraction in the downcomer base, and a new cycle begins. The superficial velocity of the
for different velocity values air is higher in the region of the riser near the sparger. The
liquid tends to behave similarly.
distribution. Blažej et al. (2004a) achieved velocity values The simulation results (Fig. 15) showed that the central
of 0.02 to 0.9 in the riser and less than 0.01 m/s in the stag- region of the draft tube, just above the sparger, was where
nant zones of the downcomer. As the gas superficial velocity the aqueous phase presented the highest velocity. However,
increases (above 0.05 m/s) in the downcomer, the holdup at higher elevations, the flow revealed a more dispersed
also increases, and more smaller bubbles are dragged and behavior. Moreover, by the operating principle of the reac-
recirculate into the ALR (Blažej et al. 2004a). Thus, they are tor, it is expected that the water velocity in the downcomer
not visible in Fig. 13. The distribution for the air superficial is high enough to establish the recirculation loop but is lower
velocity corroborates those from García et al. (2012). than the velocity in the draft tube since this is where the
liquid receives the greatest momentum quantity (Fig. 16).
Liquid phase velocity profile behavior Figures 17 and 18 show the streamlines for the upper and
lower regions of the riser, respectively. Through them, it is
The use of ALRs is justified, among other factors, by the possible to observe that the fluid dynamic regime established
homogenization efficiency of the fluid dynamic regime in the reactor is the one of interest since, upon reaching the
established inside the equipment without moving parts. This top of the riser, most of the water is directed to the base

13
 Brazilian Journal of Chemical Engineering

Fig. 14  Graphs of the air superficial velocity at different positions in diameter of 15 mm; C for an air inlet velocity of 0.45 m/s and particle
the reactor: A for an air inlet velocity of 0.27 m/s and particle diam- diameter of 18 mm; D for an air inlet velocity of 0.54 m/s and particle
eter of 12 mm; B for an air inlet velocity of 0.36 m/s and particle diameter of 21 mm

region by the downcomer (Fig. 17). Upon reaching the base diameter of 21 mm) was the one in which the liquid velocity
of the reactor, the water enters the lower part of the riser to profile showed better mixing because the loop recirculation
start a new recirculation cycle due to the influence of the was most intense (Figs. 13, 14, 15, 16).
gas being injected by the sparger (Fig. 18). Figure 17 also
illustrates the formation of liquid stagnant zones at the top Alpha parameter analysis
of the downcomer tube, as can be observed for the gas phase
in Fig. 9. One way to assess whether an ALR operates properly is
The aqueous phase and the gas phase velocity profiles are through the alpha parameter given by the ratio between the
tied. It is expected that the liquid velocity in both the riser gas holdup in the downcomer and riser and the gas holdup
and downcomer will be higher for higher air velocity values was obtained by the average value of the contour of each
since it will transfer a greater quantity of momentum to the subdomain (riser and downcomer regions). Through it, it
water as it flows through it. This phenomenon was demon- is possible to know where the air bubbles are in greater
strated by the simulation findings in this research (Figs. 19 quantity and verify whether the fluid dynamic regime estab-
and 20). Analogously to the observed for the gas phase, lished in the reactor follows the design, proposing a scale-up
simulation 4 (air velocity of 0.54 m/s and average Sauter afterward (Blažej et al. 2004b). As already mentioned, for

13
Brazilian Journal of Chemical Engineering

Fig. 15  Graphs of the water velocity at different positions in the reac- of 15 mm; C for an air inlet velocity of 0.45 m/s and particle diameter
tor: A for an air inlet velocity of 0.27 m/s and particle diameter of of 18 mm; D for an air inlet velocity of 0.54 m/s and particle diameter
12 mm; B for an air inlet velocity of 0.36 m/s and particle diameter of 21 mm

properly operation, most of the bubbles must be located in varying from 0.015 to 0.073 m/s. The authors discovered
the draft tube. The apparatus function is precisely to direct that one of the most effective methods for determining the
them to the top of the reactor so that they can flow through flow regime is through alpha evaluation. They observed an
the liquid phase and drag it to create a homogenization flow. upward trend in alpha for the air bubble stagnation regime
The bubbles not assimilated by the aqueous phase are then and a downward trend in alpha for the bubble recirculation
removed from the system by the degasser, and the rest of the regime inside the riser.
fluid is directed to the base region by the downcomer (Vial According to Blažej et al. (2004b), another way to ver-
et al. 2002). Thus, the alpha parameter tends to decrease as ify if the fluid dynamic regime is behaving properly in the
the air velocity increases (Fig. 21) because, although the design of ALRs is through the profile between the gas vol-
gas holdup in the downcomer increases with air velocity ume fraction in the downcomer and the gas volume frac-
(Fig. 12), the holdup rise in the draft tube region is greater tion in the riser. For the reactor dimensions and operating
(Fig. 11), indicating possible increased recirculation from conditions studied in this work, the results (Fig. 22) showed
the riser to the downcomer. Ghasemi and Hosseini (2012) that the curve undergoes small changes in its slope, which
evaluated the internal loop of an ALR for surface air velocity is mainly due to alterations in the fluid dynamic regime as

13
 Brazilian Journal of Chemical Engineering

the gas inlet velocity, average Sauter bubble diameter, and

gas holdup increase. This behavior (Fig. 22) is consistent
with the findings of Blažej et al. (2004b), wherein a moder-
ately non-linear dependence was shown between the riser’s
and downcomer’s gas holdup for varying air flow rates in
low-capacity ALRs (less than 40 L). As scale increases,
the linear trend becomes more noticeable, suggesting that
the impact of the existing two-phase flow regime reduced
(Blažej et al. 2004b).

Formation of the gas holdup profile and the liquid

and gas velocity profiles in the transient regime

Disturbances in the system can cause variations that lead

the regime into a transient state. A better understanding
of the fluid dynamics in this state can help with process
control, and, therefore, a simulation in the transient regime
was carried out to verify the behavior of gas holdup and
phase velocity profiles. Figures 23, 24, and 25 show that,
the formation of the holdup and velocity profiles occurred
quickly since, after the first 10 s, the contours indicated that
these profiles were already fully established. This is due to
aspects such as the air velocity in the inlet and the sparger
geometry. For lower velocities and larger reactors, the fluid
dynamic regime would take longer to be fully established in
the system, which should be taken into account in the design
of ALRs since the scale-up of this equipment should avoid
Fig. 16  Water velocity contours: A for an air inlet velocity of bottlenecks such as stagnation zones and failure to establish
0.27 m/s and particle diameter of 12 mm; B for an air inlet velocity of the recirculation loop.
0.36 m/s and particle diameter of 15 mm; C for an air inlet velocity of
0.45 m/s and particle diameter of 18 mm; D for an air inlet velocity of
0.54 m/s and particle diameter of 21 mm

Fig. 17  Streamlines at the upper

region of the riser: A for an air
inlet velocity of 0.27 m/s and
particle diameter of 12 mm;
B for an air inlet velocity of
0.36 m/s and particle diameter
of 15 mm; C for an air inlet
velocity of 0.45 m/s and particle
diameter of 18 mm; D for an air
inlet velocity of 0.54 m/s and
particle diameter of 21 mm

13
Brazilian Journal of Chemical Engineering

Fig. 18  Streamlines at the lower

region of the riser: A for an air
inlet velocity of 0.27 m/s and
particle diameter of 12 mm;
B for an air inlet velocity of
0.36 m/s and particle diameter
of 15 mm; C for an air inlet
velocity of 0.45 m/s and particle
diameter of 18 mm; D for an air
inlet velocity of 0.54 m/s and
particle diameter of 21 mm

Fig. 19  Graph of average water velocity in the riser as a function of

air velocity in the riser inlet Fig. 20  Graph of average water velocity in the downcomer as a func-
tion of air velocity in the riser inlet

Conclusions findings were reached after examining the effects of air

superficial velocity, water velocity, gas holdup, the bubble
The CFD model was developed to analyze the behavior diameter size, and alpha parameter on the internal circuit
of multiphase flow for different compartments of an air- circulation and hydrodynamic profiles.
lift reactor under unsteady state conditions. In order to The gas holdup was higher in the draft tube region clos-
further Sikula and Markoš (2008) paper, which did not est to the sparger, and it increased in both the riser and
assess local fluid dynamics analyses; flow visualization downcomer by raising the air inlet velocity. The gas phase’s
was widely employed to display the recirculation and velocity profile is similar to its volume fraction. It has a
mixing profiles. The mathematical CFD model was vali- high dependence on position, and the closer to the top of the
dated, since it achieved a high level of agreement with reactor, the lower the velocity due to the loss of momentum
the experimental data, as evidenced by the low root mean as it flows through the liquid. In turn, the liquid phase’s
square error criterion (RMSE < 5.68 × ­10–2). The following velocity profile is similar to the gas phase in that the draft
tube region next to the sparger concentrates the highest

13
 Brazilian Journal of Chemical Engineering

Fig. 23  Formation of the gas holdup profile in the transient regime

Fig. 21  Graph of alpha parameter as a function of air velocity in the
riser inlet

Fig. 24  Formation of the gas superficial velocity profile in the tran-

sient regime

Fig. 22  Graph of downcomer air volume fraction by riser air volume

fraction

velocities. However, the water flow at higher elevations was

more dispersed.
In this work, the full formation of the holdup and veloc-
ity profiles took place rapidly (around 10 s). The alpha
parameter decreased as the air velocity increased, indicat-
ing intensification of fluid recirculation from the riser to
the downcomer. For different air flow rates, a slightly non-
linear dependence between the gas holdup of the riser and
downcomer was obtained, indicating the influence between
air and liquid phase on the flow. It should be noted that the
operating conditions that provided the most satisfactory
results, i.e., better mixing time, and better establishment of
the recirculation loop, were those of simulation 4 (inlet air Fig. 25  Formation of the liquid velocity profile in the transient
regime

13
Brazilian Journal of Chemical Engineering

velocity of 0.54 m/s and average Sauter diameter of the bub- Declarations
bles of 21 mm).
The gas holdup profiles at different positions and the Conflict of interest There is no conflict of financial or personal inter-
est.
alpha parameter were the most effective characteristics
for identifying the recirculation regime, according to CFD Ethical statements Not applicable.
results. However, it is still a challenge to maintain recircula-
tion and homogeneous flow in all sections of the AIR, which
encourages further works in this area, especially in ALR
References
design optimization.
Abashar ME, Narsingh U, Rouillard AE, Judd R (1998) Hydrodynamic
flow regimes, gas holdup, and liquid circulation in airlift reactors.
Appendix Ind Eng Chem Res 37:1251–1259
Abid A, Saidane F, Hamdi M (2017) Feasibility of carbon dioxide
sequestration by Spongiochloris sp microalgae during petroleum
See below Table 4 here. wastewater treatment in airlift bioreactor. Bioresour Technol
234:297–302
Artok L, Schobert HH (2000) Reaction of carboxylic acids under coal
liquefaction conditions. 2. Under hydrogen atmosphere. J Anal
Appl Pyrolysis 54:235–246
Bannari R, Hilali Y, Essadki A, Bannari A (2019) Computational fluid
dynamic for improving design and performance of an external
loop airlift reactor used in electrochemical wastewater treatment.
Table 4  Typical airlift reactor operating conditions, according to lit- SN Appl Sci. https://​doi.​org/​10.​1007/​s42452-​019-​1523-z
erature Blažej M, Cartland Glover GM, Generalis SC, Markoš J (2004a) Gas–
liquid simulation of an airlift bubble column reactor. Chem Eng
Reference works Bubble Gas velocity (m/s) Process 43:137–144
diameter Blažej M, Kiša M, Markoš J (2004b) Scale influence on the hydro-
(mm) dynamics of an internal loop airlift reactor. Chem Eng Process
43:1519–1527
Huang et al. (2010) 5–6 us= 0.02–0.006
Cao C, Zhao L, Xu D, Geng Q, Guo Q (2009) Investigation into bub-
Luo et al. (2011) 3 us= 0.38 ble size distribution and transient evolution in the sparger region
Ebrahimifakhar et al. 5 us= 0.02–0.12 of gas−liquid external loop airlift reactors. Ind Eng Chem Res
(2011) 48:5824–5832
Šimčík et al. (2011) 5 ud = 0.010–0.075 Chen Z, Shen W, Zhang J (2023) Computational fluid dynamics study
of gas-liquid mass transfer in concentric tube airlift reactor with
Lestinsky et al. (2012) 5.8–11.1 ud = 0.025–0.270
helical baffle. Chem Eng Res Des 198:121–128
Ghasemi et al. (2012) 2 us= 0.015–0.073 Choi KH (2021) Prediction of gas holdup in various types of airlift
Garcia et al. (2012) 4 us= 0.013 reactors. Korean J Chem Eng 38:1781–1790
Huang et al. (2016) 3 us= 0.009–0.019 Ebrahimifakhar M, Mohsenzadeh E, Moradi S, Moraveji M, Salimi
M (2011) CFD simulation of the hydrodynamics in an internal
Wadaugsorn et al. (2016) 5 us= 0.02–0.10
air-lift reactor with two different configurations. Front Chem Sci
Pawar (2017) 3–10 us= 0.01–0.05 Eng 5:455–462
Liew et al. (2017) 3.40 us= 0.034 Fletcher DF (2022) The future of computational fluid dynamics (CFD)
Moudoud et al. (2018) 3 us= 0.025–0.070 simulation in the chemical process industries. Chem Eng Res Des
187:299–305
Bannari et al. (2019) 0.01–0.25 us= 0.060–0.064
Freitas C, Fialová M, Zahradnik J, Teixeira JA (1999) Hydrodynamic
Li et al. (2020) 1 us= 0.02–0.04 model for three-phase internal- and external-loop airlift reactors.
Shi et al. (2020) 4 us= 0.001–0.004 Chem Eng Sci 54:5253–5258
Kouzbour et al. (2020) 5 us= 0.01–0.08 García S, Paternina E, Pupo OR, Bula A, Acuña F (2012) CFD simula-
tion of multiphase flow in an airlift column photobioreactor for
Shi et al. (2021) 0.4–16.13 us= 0.005–0.090
the cultivation of microalgae. In: ASME 2012 6th International
Teli et al. (2021) 3–11 us= 0.002–0.017 Conference on Energy Sustainability, Parts A and B. American
Ramonet et al. (2022) 4.5 us= 0.01–0.04 Society of Mechanical Engineers, July 23–26, 2012, San Diego,
Puiman et al. (2022) 1–7 us= 0.096–0.280 CA, USA, pp. 1253–1262
Ghasemi H, Hosseini SH (2012) Investigation of hydrodynamics and
Zhang (2022) 1–15 us= 0.062–0.123
transition regime in an internal loop airlift reactor using CFD.
Qiao et al. (2023) 4.5 us= 0.002–0.008 Braz J Chem Eng 29:821–833
Chen et al. (2023) 3.5–4.5 us= 0.002–0.003 Hari M, Tan RBH (2002) A dynamical systems approach to mixing in
Li et al. (2023) 3 us= 0.01 circulating flows. Chem Eng Technol 25:811–818
Hu W-W, Zhong J-J (2001) Effect of bottom clearance on performance
Xu et al. (2024) 0.5–0.63 us= 0.01–0.04
of airlift bioreactor in high-density culture of panax notoginseng
Khan et al. (2024) 2.1–6.4 us= 1.7 × ­10–4–8.1 × ­10–4 cells. J Biosci Bioeng 92:389–392
This work 12–21 us= 0.018–0.540

13
 Brazilian Journal of Chemical Engineering

Huang Q, Yang C, Yu G, Mao ZS (2010) CFD simulation of hydrody- industrial external-loop syngas-to-ethanol fermentation. Chem Eng
namics and mass transfer in an internal airlift loop reactor using a Sci. https://​doi.​org/​10.​1016/j.​ces.​2022.​117770
steady two-fluid model. Chem Eng Sci 65:5527–5536 Qiao N, Yue S, Cheng J, Wang C, Wang X, Shi Y, Guo J, Yu D (2023) A
Huang Q, Zhang W, Yang C (2015) Modeling transport phenomena and gas distributor capable of multiple injection directions to improve
reactions in a pilot slurry airlift loop reactor for direct coal liquefac- the gas–liquid dispersion performance in the airlift loop reactor.
tion. Chem Eng Sci 135:441–451 Biochem Eng J 190:108770
Huang J, Ying J, Fan F, Yang Q, Wang J, Li Y (2016) Development of Raman RK, Dewang Y, Raghuwanshi J (2018) A review on applications
a novel multi-column airlift photobioreactor with easy scalability of computational fluid dynamics. Int J LNCT 2:137–143
by means of computational fluid dynamics simulations and experi- Ramonet F, Haddadi B, Jordan C, Harasek M (2022) Modelling and
ments. Biores Technol 222:399–407 design of optimal internal loop air-lift reactor configurations through
Jasim M, Mohammed T, Sabri L (2022) Air-lift reactor’s characterization computational fluid dynamics. Chem Eng Trans 94:817–822
via Computational Fluid Dynamic (CFD): Review. Eng Technol J Shi Y, Wu S, Ren H, Jin M, Wang L, Qiao N, Yu D (2020) Computational
40:484–497 fluid dynamics and factor analysis of a novel swirling demulsified
Khan H, Kováts P, Zähringer K, Rzehak R (2024) Experimental and airlift loop reactor for the treatment of refined soybean oil wastewa-
numerical investigation of a counter-current flow bubble column. ter. Biores Technol. https://​doi.​org/​10.​1016/j.​biort​ech.​2019.​122316
Chem Eng Sci. https://​doi.​org/​10.​1016/j.​ces.​2023.​119503 Shi J, Guo K, Wang Z, Zheng L, Liu H, Xiang W, Liu C, Li X (2021)
Klein J, Godo Š, Dolgoš O, Markoš J (2001) Effect of a gas–liquid separa- Computational fluid dynamics simulation of hydrodynamics in a
tor on the hydrodynamics and circulation flow regimes in internal- two-stage internal loop airlift reactor with contraction-expansion
loop airlift reactors. J Chem Technol Biotechnol 76:516–524 guide vane. ACS Omega 6:6981–6995
Kouzbour S, Stiriba Y, Gourich B, Vial C (2020) CFD simulation and Sikula I, Markoš J (2008) Modeling of enzymatic reaction in an airlift
analysis of reactive flow for dissolved manganese removal from reactor using an axial dispersion model. Chem Pap 62:10–17
drinking water by aeration process using an airlift reactor. J Water Silva MK, Davila MA, Mori M (2011) CFD modelling of a bubble col-
Process Eng. https://​doi.​org/​10.​1016/j.​jwpe.​2020.​101352 umn with an external loop in the heterogeneous regime. Can J Chem
Lestinsky P, Vayrynen P, Vecer M, Wichterle K (2012) Hydrodynamics Eng 89:671–681
of airlift reactor with internal circulation loop: experiment vs. CFD Šimčík M, Mota A, Ruzicka MC, Vicente A, Teixeira J (2011) CFD
simulation. Procedia Eng 42:892–907 simulation and experimental measurement of gas holdup and liquid
Li X, Zhang G, Zhao X, Zhou J, Du G, Chen J (2020) A conceptual air-lift interstitial velocity in internal loop airlift reactor. Chem Eng Sci
reactor design for large scale animal cell cultivation in the context of 66:3268–3279
in vitro meat production. Chem Eng Sci 211:115269 Teli SM, Mathpati CS (2021) Experimental and numerical study of gas-
Li L, Xu X, Wang W, Lau R, Wang C-H (2022) Hydrodynamics and mass liquid flow in a sectionalized external-loop airlift reactor. Chinese
transfer of concentric-tube internal loop airlift reactors: a review. J Chem Eng 32:39–60
Bioresour Technol 359:127451 Van Baten JM, Ellenberger J, Krishna R (2003) Hydrodynamics of inter-
Li L, Mohd Shafie ZMH, Huang T, Lau R, Wang CH (2023) Multiphysics nal air-lift reactors: experiments versus CFD simulations. Chem Eng
simulations of concentric-tube internal loop airlift photobioreactors Process 42:733–742
for microalgae cultivation. Chem Eng J. https://​doi.​org/​10.​1016/j.​ Vial Ch, Poncin S, Wild G, Midoux N (2002) Experimental and theoreti-
cej.​2023.​141342 cal analysis of the hydrodynamics in the riser of an external loop
Liew SY, Gimbun J (2017) CFD simulation on the hydrodynamics in airlift reactor. Chem Eng Sci 57:4745–4762
gas-liquid airlift reactor. Chem Prod Process Model. https://​doi.​org/​ Wadaugsorn K, Limtrakul S, Vatanatham T, Ramachandran PA (2016)
10.​1515/​cppm-​2017-​0030 Hydrodynamic behaviors and mixing characteristics in an internal
Lu W-J, Hwang S-J, Chang C-M (1994) Liquid mixing in internal loop loop airlift reactor based on CFD simulation. Chem Eng Res Des
airlift reactors. Ind Eng Chem Res 33:2180–2186 113:125–139
Luewisutthichat W, Tsutsumi A, Yoshida K (1997) Bubble characteristics Wang T, Wang J, Jin Y (2004) Experimental study and CFD simulation
in multi-phase flow systems: bubble sizes and size distributions. J of hydrodynamic behaviours in an external loop airlift slurry reactor.
Chem Eng Jpn 30:461–466 Can J Chem Eng 82:1183–1190
Luo H-P, Al-Dahhan MH (2011) Verification and validation of CFD Xu X, Zhang Y (2024) Hydrodynamics and mass transfer in an airlift loop
simulations for local flow dynamics in a draft tube airlift bioreactor. reactor: comparison between using two kinds of spargers. Processes.
Chem Eng Sci 5:907–923 https://​doi.​org/​10.​3390/​pr120​10035
Luo J, Chen L, Min T, Shan F, Kang Q, Tao W (2020) Macroscopic Zhang T, We C, Ren Y, Feng C, Wu H (2017) Advances in airlift reac-
transport properties of Gyroid structures based on pore-scale studies: tors: modified design and optimization of operation conditions. Rev
Permeability, diffusivity and thermal conductivity. Int J Heat Mass Chem Eng 33:163–182
Transf 146:118837 Zhang C, Liu Y, Jiao W, Qi G, Guo J (2022) A turbulent mass diffusiv-
Ma T, Ziegenhein T, Lucas D, Fröhlich J (2015) Large eddy simulations ity model for the simulation of the biodegradation of toluene in an
of the gas–liquid flow in a rectangular bubble column. Nucl Eng internal loop airlift reactor. Process Saf Environ Prot 165:646–657
Des 299:146–153 Zhu A, Xu Y, Downar T (2016) Stability analysis of the Backward Euler
Merchuk J (1990) Why use air-lift bioreactors? Trends Biotechnol time discretization for the pin-resolved transport transient reactor
8:66–71 calculation. Ann Nucl Energy 87:252–266
Moudoud N, Rihani R, Bentahar F, Legrand J (2018) Global hydrody-
namic of hybrid external loop airlift reactor: experiments and CFD Publisher's Note Springer Nature remains neutral with regard to
modelling. Chem Eng Process 129:118–130 jurisdictional claims in published maps and institutional affiliations.
Naidoo N, Pauck WJ, Carsky M (2021) Effects of sparger design on the
gas holdup and mass transfer in a pilot scale external loop airlift Springer Nature or its licensor (e.g. a society or other partner) holds
reactor. S Afr J Chem Eng 37:127–134 exclusive rights to this article under a publishing agreement with the
Pawar SB (2017) CFD analysis of flow regimes in airlift reactor using author(s) or other rightsholder(s); author self-archiving of the accepted
Eulerian-Lagrangian approach. Can J Chem Eng 95:420–431 manuscript version of this article is solely governed by the terms of
Puiman L, Abrahamson B, van der Lans RGJM, Haringa C, Noorman such publishing agreement and applicable law.
HJ, Picioreanu C (2022) Alleviating mass transfer limitations in

13