SoftwareX 11 (2020) 100483

Contents lists available at ScienceDirect

SoftwareX
journal homepage: www.elsevier.com/locate/softx

Original software publication

3D-coupling of Volume-of-Fluid and Lagrangian particle tracking for

spray atomization simulation in OpenFOAM
∗
Martin Heinrich , Rüdiger Schwarze
Institute of Mechanics and Fluid Dynamics, TU Bergakademie Freiberg, Lampadiusstr. 4, 09599 Freiberg, Germany

article info a b s t r a c t

Article history: Simulations of spray breakup are challenging due to a wide range of different length and time scales.
Received 7 February 2020 While Volume-of-Fluid (VoF) is suited for the primary breakup, the secondary breakup and the dilute
Received in revised form 24 March 2020 spray cloud is better resolved using a Lagrangian particle tracking (LPT) method. This publication
Accepted 16 April 2020
proposes a 3D coupling between both models using the CFD software OpenFOAM to resolve the
Keywords: complete process of spray atomization. Combined with adaptive mesh refinement, the computational
VoF cost can be reduced significantly. The simulation of a fuel jet in crossflow demonstrates the accuracy
Lagrangian particle tracking and speedup of the proposed method.
OpenFOAM © 2020 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license
Validation (http://creativecommons.org/licenses/by/4.0/).

Code metadata
Current code version 1.0
Permanent link to code/repository used of this code version https://github.com/ElsevierSoftwareX/SOFTX_2020_30
Code Ocean compute capsule none
Legal Code Licence GNU GPL V3
Code versioning system used git
Software code languages, tools, and services used C++
Compilation requirements, operating environments & dependencies OpenFOAM v1912
If available Link to developer documentation/manual
Support email for questions [email protected]

1. Motivation and significance aerodynamic forces and turbulence grow and finally cause the jet
to rupture into ligaments. (2) During secondary breakup those
Spray atomization is a widely utilized engineering application ligaments disintegrate further into a disperse cloud of smaller
in combustion, metal powder production for additive manufac- droplets. This effect continues until surface tension dominates
turing, or inhalers. It is a very challenging process due to multiple and results in stable spherical droplets [1].
phases, complex physics of jet breakup, and a wide range of scales Interface capturing methods like front-tracking [2], cell and
between large and small structures. Simulation of this kind of marker [3], level-set [4,5], Volume-of-Fluid (VoF) [6], or a com-
flow is vital to gain insight in mechanisms of breakup, optimizing bination such as CLSVOF [7,8] are typically used for modelling
nozzle geometry, and to predict droplet size distributions. Ad- primary breakup. An arbitrarily shaped interface, e.g. liquid jet,
ditionally, experiments are not always feasible, e.g. liquid metal ligaments, or individual droplets, can be captured as long as it
melt atomization at very high temperature. is sufficiently resolved by the computational mesh. As a result, a
The atomization process can be divided into two parts: (1) disperse cloud of small droplets will result in huge computational
Primary atomization and (2) secondary breakup. (1) Primary at- cost for a high grid resolution. On the other hand, Lagrangian
omization is characterized by the breakup of a liquid jet emerging particle tracking (LPT) methods were developed to track large
through a nozzle. Instabilities at the liquid–gas interface due to numbers of individual droplets or particles. The computational
cost is very low as the cell size has to be significantly larger
∗ Corresponding author. than the characteristic particle diameter. Coupled with models
E-mail address: [email protected] (M. Heinrich). for droplet breakup [9–12] and collision [13,14], the Lagrangian

https://doi.org/10.1016/j.softx.2020.100483
2352-7110/© 2020 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
2 M. Heinrich and R. Schwarze / SoftwareX 11 (2020) 100483

approach is suited to resolve the secondary breakup and the method for advection of a sharp interfaces [26]. Surface tension
following disperse cloud of smaller droplets. is implemented in the momentum equation based on the Contin-
Since interface capturing methods and the Lagrangian ap- uum Surface Force (CSF) model introduced by Brackbill [27]. Flow
proach only cover parts of the entire atomization process, a properties are weighted with the phase fraction as following:
combination of both is needed for a complete numerical model.
Special care has to be taken for the coupling since the require-
ρ = αρ1 + (1 − α) ρ2 , (3)
ments for the mesh resolution and the solved governing equa- µ = αµ1 + (1 − α) µ2 , (4)
tions are significantly different for both approaches. One method
with kinematic viscosity µ and density ρ of phase 1 and 2,
for this is using a 2D coupling plane (e.g. Grosshans et al. [15],
respectively.
Spitzenberger et al. [16]). However, characteristics about the jet
breakup are needed to place the coupling plane accordingly. 3D
2.2. Lagrangian particle tracking method
coupling algorithms have the advantage of tracking and convert-
ing liquid droplets in a specified volumetric region. Thus, no prior
The LPT method tracks the movement of Lagrangian parcels
knowledge of the jet breakup is required. Hermann et al. [17,
throughout the computational domain. Each parcel represents
18] applied a level-set method to resolve large ligaments while
a certain number of droplets with identical properties, such as
smaller ones are tracked in 3D, removed from the Eulerian space,
diameter and velocity. They are treated as point masses and
and injected in the Lagrangian framework preserving position,
thus contain no volume. Position and velocity are updated at
velocity, and diameter. 3D methods can be either one-way or
each time step based on differential equations for trajectory and
two-way coupling. For one-way, entities in the Eulerian space are
momentum:
converted to the Lagrangian framework only. A two-way coupling
is capable to reverse this process and convert Lagrangian entities dxp
= up , (5)
back to the Eulerian space. The latter was implemented and used dt
for spray simulations by Edelbauer et al. [19] and Zuzio et al. [8]. dup
mp = F D + F G, (6)
However, Ling et al. [20] reported a small difference between dt
one-way and two-way coupling for their test cases. with parcel velocity up , mass mp , position xp , and acting forces on
One application for such an algorithm is for example a multi- the parcels F for drag and gravitation. Drag force is calculated as
phase jet in cross flow (JICF), which is used for plume dispersion, following:
gas turbine combustor cooling, and fuel injection. Configurations
like that are widely investigated to control and optimize the π D2p
ρg ug − up ⏐ug − up ⏐ ,
( )⏐ ⏐
F D = CD (7)
mixing process, e.g. [21–23]. For this reason, comprehensive ex- 8
perimental data is available to model and simulate this complex where Dp stands for the parcel diameter, ρg for the gas phase
configuration and thus validate the proposed algorithm. density, ug for the gas velocity at the parcel position xp taken
In this paper, a 3D coupling algorithm between a VoF method from the Eulerian flow field, and drag coefficient CD . The drag
and LPT is presented for spray simulations implemented within coefficient is calculated based on the Schiller–Naumann drag
the open source CFD library OpenFOAM v1912 [24]. The whole model [28], which is considered a standard model for sparsely
computational domain is used to track liquid droplets and convert distributed spherical parcels:
them to Lagrangian parcels. The corresponding equations and
1 + 0.15 Re0p.687
{
24
( )
algorithms are presented as well as the conversion criteria for Rep
if Rep ≤ 1000
CD = (8)
the transition between the Eulerian and the Lagrangian space. 0.44 if Rep > 1000,
A fuel jet in cross flow is the validation case for the proposed
algorithm and shows a high accuracy as well as significantly with parcel Reynolds number Rep . Gravitational and buoyancy
reduced computational cost compared to a conventional VoF-only forces are computed as one combined force:
approach. The algorithm is also fully parallelized allowing for ρg
( )
large computational domains. F G = mp g 1 − . (9)
ρp
2. Software description Parcel breakup due to aerodynamic forces is modelled using
the Reitz–Diwakar secondary breakup model [11] and parcel–
2.1. Volume of fluid method parcel collisions and subsequent coalescence is computed with
the collision algorithm from Nordin [14].
In this paper the VoF method is employed to track the in-
terface between two incompressible fluids in an Eulerian frame- 2.3. Coupling algorithm
work [25]. The fluids are treated as a single continuum with one
set of governing equations for mass and momentum. The phase The proposed algorithm is the link between the Volume-of-
fraction α distinguishes between the two phases: Fluid method for tracking the free surface of a two-phase sepa-
rated flow with the Eulerian approach and the Lagrangian particle
0 phase 1,
{
tracking for dilute dispersed flows. It consists of three steps:
α= 0<α<1 interface, (1)
(1) Identify individual droplets in the phase fraction field α ; (2)
1 phase 2.
Calculate their corresponding properties; (3) If the droplets meet
It describes the cell volume occupied by a certain phase and the transformation criteria, convert them by injecting parcels
indicates the spatial distribution of the two phases. The phase into the Lagrangian framework and delete corresponding liquid
fraction itself is solved using one governing equation: droplets from phase field.
In the first step of this algorithm, individual three-dimensional
∂α ∂ (α ui )
+ = 0, (2) droplets in the phase fraction field are identified using a three-
∂t ∂ xi dimensional four-connectivity spatial Connected Component La-
with the flow velocity ui and time t. The interface between belling (CCL) algorithm in the complete computational domain as
the two phases is reconstructed with the geometric isoAdvector shown in Fig. 1. At first, a threshold αt is applied to mark cells as
 M. Heinrich and R. Schwarze / SoftwareX 11 (2020) 100483 3

Fig. 1. Algorithm for detecting the droplets in the phase fraction field α with a threshold of αt = 0.2. The interface between both faces is shown as solid blue line
(left).

liquid, if their phase fraction is larger than αt . Otherwise they are values larger than 1 indicate a non-spherical droplet.
defined as gas and not considered for the further steps in this Sphericity is calculated as following:
algorithm. The limit should be in the range of 10−2 . Otherwise 2rmax
mass is not conserved in the coupling process since some cells at Ψp = , (14)
dp
the interface are not captured as liquid. Next, the algorithm loops
over all cells and marks each liquid cell with their unique cell ID. with the maximum distance from centre of mass to the
This way, the labelled droplets are unique and not ambiguous in droplet free surface rmax and the diameter of a perfect
case of parallel computations. Finally, the smallest ID from one sphere with the same mass as the given droplet dp . Ef-
liquid droplet is propagated towards all neighbouring liquid cells. fects of non-spherical droplets, such as earlier breakup, in-
In this sense, neighbours are cells which share at least one face creased drag, or oscillating deformation, cannot be resolved
with a given cell. As a result, all VoF droplets are identified and accurately using Lagrangian parcels. Therefore, those
labelled based on the phase fraction field α . droplets remain in the VoF frame and are not converted.
The second step loops over all cells to calculate the individual
If those criteria are met, the VoF droplets are converted into
properties for each droplet. For a given droplet these equations
Lagrangian parcels and injected into the computational domain.
are as following:
At the same time, the corresponding liquid droplets are set to
α = 0 and thus deleted from the phase fraction field. Fig. 2 shows
∑
Vp = αi Vi , (10)
i
the flow chart of the presented algorithm.
1 ∑
xp = αi Vi xi , (11) 3. Validation
Vp
i
1 ∑ The proposed algorithm is validated against experimental
up = αi Vi ui , (12)
Vp measurements from Gopala et al. [29] and Sekar et al. [30] for a
i
√ fuel jet in cross flow. This application is often used in gas turbine
3 6Vp engine combustors. Therefore, the breakup mechanisms of the
dp = . (13)
π liquid jet and distribution of droplets are thoroughly investigated
Index i denotes all cells belonging to an droplet. Phase fraction and suitable for validation. Fig. 3 a) shows a schematic sketch of
α is considered in the calculation for the droplet volume Vp , the computational domain, which is based on the experiments. It
position xp , and velocity up . Therefore, cells at the interface with consists of a rectangular air channel with the size of 80 × 80 × 160
α < 1 will account only with their effective liquid phase mass. D, where D represents the nozzle diameter. This only represents a
The equivalent diameter dp is identical to the diameter of a sphere part of the experimental channel size. For this reason, symmetry
with the same volume Vp . plane boundary conditions are used at the top, front, and back
In the third step, the transformation criteria of each droplet are plane. A fully developed air flow profile is defined at the inlet
checked. These criteria are difficult to define based on physical with a pressure of 5 bar and a temperature of 300 K. The liquid
properties due to the complex and highly disturbed flow. For nozzle has a diameter of D = 0.485 mm. It is placed downstream
example, many instabilities occur atomization processes such the gas inlet and consist of a plenum chamber and the nozzle exit.
as Kelvin-Helmholtz, Rayleigh–Taylor, or Plateau-Rayleigh [8]. A uniform velocity is defined at the fuel inlet. The momentum
Therefore, geometrical properties are used as criteria as they are flux ratio, defined as momentum ratio between fuel and air, is
easy to define and calculate: 10, Reynolds number ratio is 144, and the Weber number is 1500.
The fluid properties and operating conditions are summarized in
(1) The droplet’s diameter must be smaller than a certain Table 1.
threshold dmin . This is due to two reasons: (1) Weber and The computational domain is meshed using an unstructured,
Reynolds-number of large droplets is higher and thus the hexahedral mesh with an initial cell count of 4.2 × 106 cells.
droplet may deform significantly. This effect is difficult to Six refinement levels are defined (see Fig. 3, b). The growth ratio
capture within a Lagrangian framework and better resolved between each grid level is 2, starting from the finest level 6 with a
using the VoF approach. (2) Small droplets are poorly re- grid cell size of 15 µm to the coarsest level 1 with about 500 µm.
solved by the VoF approach. Thus, droplets with a diameter As a result, 30 cells are located within the smallest nozzle di-
smaller than five times the characteristic cell length should ameter. Adaptive mesh refinement is employed to automatically
be converted as well. refining the free surface between fuel and air phase.
(2) The droplet’s sphericity Ψp must be smaller than 2. Spheric- The validation case is simulated using the presented VoF-
ity compares the shape of an droplet to a perfectly round LPT coupling algorithm implemented in OpenFOAM v1912. The
sphere. A value of 1 resembles a perfect sphere, whereas incompressible multiphase solver interIsoFoam [26] is utilized
4 M. Heinrich and R. Schwarze / SoftwareX 11 (2020) 100483

Table 1
Fluid properties and operating conditions.
Property Unit Fuel Air
Velocity m/s 32 116.6
Density kg/m3 780.6 5.88
Kinematic viscosity Pa s 7 × 10−4 1.85 × 10−5
Surface tension N/m 0.07
Operating pressure bar 5
Temperature K 300
Momentum flux ratio – 10
Weber number – 1500

no need to solve additional transport equations for turbulence.

The convective term in the momentum equation is discretized
with a second order Linear Upwind Stabilized Transport (LUST)
scheme. Discretization of the remaining spatial terms (Lapla-
cian, gradients, and interpolation) use second order Gauss linear
schemes. Temporal discretization is first order with a temporal
resolution of 8 × 10−9 s, which results in a maximum Courant
number of about 0.15.
Fig. 4 verifies the coupling algorithm by comparing the jet
in crossflow simulation with and without VoF-LPT coupling. The
flow was simulated for 0.1 µs without any coupling using the VoF
solver interIsoFoam. From this point forward, two simulations
were run for 250 time steps: one without (Fig. 4, left) and one
with coupling (right). The results show how the droplets removed
from the phase field α (grey surface) are converted into the
Lagrangian framework (green spheres). They travel with the flow
and match the position of VoF droplets in the simulation without
coupling. The conversion of VoF droplets to Lagrangian parcels
removes the interface of the phase fraction and thus the need
for a high mesh resolution. Therefore, the mesh size is already
reduced by 5% during this short time frame. This effect gets more
Fig. 2. Flow chart of the presented VoF-LPT coupling algorithm (blue box) within
the framework of OpenFOAM. (For interpretation of the references to colour in
noticeable as the simulation continues and the spray is more
this figure legend, the reader is referred to the web version of this article.) dispersed.
The numerical model is validated in two ways. At first, the
Sauter mean diameter d32 of the parcel cloud is evaluated at
for solving the phase fraction α . This solver uses the isoAdvec- two postprocessing planes located 30 and 60 times the noz-
zle diameter downstream the nozzle exit. These planes have
tor VoF method in combination with the PIMPLE algorithm, a
a width and height of 1 mm and 20 mm, respectively. Each
combination of PISO [31] and SIMPLE [32], for pressure–velocity
Lagrangian parcel crossing those planes will get registered and
coupling. Turbulent effects are modelled using Implicit Large used for calculating d32 . VoF droplets do not have to be taken
Eddy Simulation (ILES) [33]. Large flow scales are resolved while into account here since all are converted to Lagrangian parcels at
smaller ones are modelled by the spatial discretization of the this point of the flow. The comparison between experiment and
convection terms in the momentum equation. Therefore, there is presented numerical model is shown in Fig. 5 for both planes.

Fig. 3. Computational domain (left) and close-up of hexahedral mesh around the nozzle exit (right).
 M. Heinrich and R. Schwarze / SoftwareX 11 (2020) 100483 5

Fig. 4. Comparison between Volume-of-Fluid only simulation (left) and coupling between Volume-of-Fluid and Lagrangian particle tracking at the same time step.
(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 5. Comparison of Sauter mean diameter d32 for experiment (with standard deviation) and simulation as function of wall distance y at first (left) and second
(right) postprocessing plane.

The Sauter mean diameter is plotted against the wall distance

y with a standard deviation for the experimental measurements
of ±5 µm. The results reveal a good accuracy of the proposed
coupling algorithm. The d32 diameter is well within the standard
deviation of the experiment and increases nearly linearly with
wall distance. However, the local maximum of d32 at about y=4-
5 mm is not resolved accurately. This could be attributed to the
coarse boundary mesh resolution in this test case.
Penetration jet trajectory defines the outermost edge of the jet.
Comparing the edge with the experimental measurements (see
Fig. 6) reveals a very good agreement. This indicates a seamless
transition from VoF droplets to Lagrangian parcels since primary
breakup captured by VoF is concluded at about z/D=4-5. From
this point forward, Lagrangian parcels dominate the dispersed
multiphase flow field.
One advantage of the VoF-LPT coupling is a significant sav-
ing in computational cost. Converting VoF to Lagrangian parcels
avoids the need for a high mesh resolution downstream the jet
breakup. This reduces the overall mesh size considerably and thus Fig. 6. Comparison of penetration jet trajectory between experiment and
simulation.
computational cost as shown in Fig. 7. The black line shows the
CPU time for each time step for a VoF-only simulation. It increases
significantly over time as the spray is more spread out and thus
the number of cells (20.1 × 106 ) rises due to adaptive mesh result from automatic load balancing of the mesh between the
refinement of the free surface. The sudden drops in CPU time processor partitions during runtime. Using a VoF-LPT coupling
6 M. Heinrich and R. Schwarze / SoftwareX 11 (2020) 100483

Acknowledgments

The work was supported by the Deutsche Forschungsgemein-

schaft (DFG, German Research Foundation) - Projektnummer
54473466 - SFB 799. This support is sincerely acknowledged by
the authors.

References

[1] Gorokhovski M, Herrmann M, atomization Modelingprimary. Modeling

primary atomization. Annu Rev Fluid Mech 2008;40(1):343–66.
[2] Tryggvason G, Bunner B, Esmaeeli A, Juric D, Al-Rawahi N, Tauber W, et
al. A front-tracking method for the computations of multiphase flow. J
Comput Phys 2001;169(2):708–59.
[3] Harlow FH, Welch JE. Numerical calculation of time-dependent vis-
cous incompressible flow of fluid with free surface. Phys Fluids
1965;8(12):2182–9.
[4] Sussmann M, Smereka P, Osher S. A level set approach for com-
puting solutions to incompressible two-phase flow. J Comput Phys
Fig. 7. Computational cost of the VoF-LPT coupling simulation compared to a
1994;114(1):146–59.
VoF-only simulation. (For interpretation of the references to colour in this figure
[5] Desjardins O, Moureau V, Pitsch H. An accurate conservative level
legend, the reader is referred to the web version of this article.)
set/ghost fluid method for simulating turbulent atomization. J Comput Phys
2008;227(18):8395–416.
[6] Ling Y, Zaleski S, Scardovelli R. Multiscale simulation of atomization with
small droplets represented by a Lagrangian point-particle model. Int J
keeps the cell count low (7.7 × 106 ) and thus the CPU time per
Multiph Flow 2015;76:122–43.
time step (blue line). As a result, the coupling algorithm is faster [7] Sussmann M, Puckett EG. A coupled level set and volume-of-fluid method
by a factor of up to 3 compared to a classical VoF-only approach. for computing 3D and axisymmetric incompressible two-phase flows. J
Comput Phys 2000;162(2):301–37.
[8] Zuzio D, Estivalezes J-L, DiPierro B. An improved multiscale Eulerian–
4. Impact and conclusion Lagrangian method for simulation of atomization process. Comput & Fluids
2017;176:285–301.
[9] Tanner F. Liquid jet atomization and droplet breakup modeling of
A 3D coupling algorithm between Volume-of-Fluid and La- non-evaporating diesel fuel sprays. SAE Trans J Eng 1997;106:127–40.
grangian particle tracking is presented for the open-source CFD- [10] Pilch M, Erdman C. Use of breakup time data and velocity history data
to predict the maximum size of stable fragments for acceleration induced
library OpenFOAM v1912. It is based on a four-way CCL method
breakup of a liquid drop. Int J Multiphase Flows 1987;13(6):741–57.
and droplet sphericity and diameter are defined as conversion [11] Reitz R. Modelling atomization processes in highpressure vaporizing
criteria. An experimental fuel jet in cross flow is used for vali- sprays. Atom Spray Technol 1987;3(4):309–37.
dation. The results reveal a very good agreement with the mea- [12] Schmehl R, Maier G, Witting S. CFD analysis of fuel atomization, secondary
droplet breakup and spray dispersion in the premix duct of a LPP
surements in terms of Sauter mean diameter and penetration combustor. In: Eighth international conference on liquid atomization and
jet length. Furthermore, the coupling algorithm shows a signif- spray systems. Pasadena, California, USA; 2000.
icant improvement in computing time compared to a VoF-only [13] O’Rourke P. Collective drop effects on vaporizing liquid sprays. Department
of Mechanical and Aerospace Engineering, Princeton University; 1981.
simulation.
[14] Nordin N. Complex chemistry modeling of diesel spray combustion. Ph.D.
Conventional spray atomization simulations with OpenFOAM thesis, Chalmers University of Technology; 2001.
were either limited to the primary breakup using VoF due to [15] Grosshans H, Szasz R-Z, Fuchs L. Development of an efficient statistical
computational cost or to the secondary breakup with LPT in volumes of fluid–Lagrangian particle tracking coupling method. Internat J
Numer Methods Fluids 2014;74(12):898–918.
combination with empirical atomization models. With the help [16] Spitzenberger A, Neumann S, Heinrich M, Schwarze R. Particle detection
of the presented algorithm it is possible to simulate the com- in VOF-simulations with openfoam. SoftwareX 2020;11:1–7.
plete atomization process of liquid sprays. Primary breakup and [17] Hermann M. On simulating primary atomization using refined level set
grid method. Atomization Spray 2011;21(4):283–301.
larger ligaments are resolved using the VoF model while smaller
[18] Hermann M. A parallel eulerian interface tracking/Lagrangian point particle
droplets and the dilute spray cloud are modelled using LPT. The multi-scale coupling procedure. J Comput Phys 2010;229(3):745–59.
conversion from VoF to LPT is handled automatically by the cou- [19] Edelbauer W, Birkhold F, Rankel T, Pavlovic Z, Kolar P. Simulation of the
pling algorithm based on droplet diameter and sphericity. Since liquid break-up at an adblue injector with the volume-of-fluid method
followed by off-line coupled Lagrangian particle tracking. Comput & Fluids
the coupling algorithm scans the whole computational domain
2017;157:294–311.
for droplets, no coupling plane has to be defined. [20] Ling Y, Zaleski S, Scardovelli R. Multiscale simulation of atomization with
small droplets represented by a Lagrangian point-particle model. Int J
Multiph Flow 2015;76:122–45.
Declaration of competing interest [21] Prakash S, Sinha A, Raghunandan B, Ravikrishna R. Jet in cross flow - Influ-
ence of liquid jet entry conditions. In: Triennial international conference
on liquid atomization and spray systems. Tainan, Taiwan; 2015.
The authors declare that they have no known competing finan- [22] Ruiz A, Lacaze G, Oefelein J. Flow topologies and turbulence scales in a
cial interests or personal relationships that could have appeared jet-in-cross-flow. Phys Fluids 2015;27(4):1–41.
to influence the work reported in this paper. [23] Octau C, Dbouk T, Watremez M, Meresse D, Lippert M, Schiffler J, et
al. Liquid–solid two-phase jet in a turbulent crossflow: Experiments and
simulations. Chem Eng Res Des 2020;155:156–71.
CRediT authorship contribution statement [24] Openfoam: The open source CFD toolbox. 2019, www.openfoam.com.
[25] Hirt C, Nichols B. Volume of fluid (VOF) method for the dynamics of free
boundaries. J Comput Phys 1981;39(1):201–25.
Martin Heinrich: Conceptualization, Methodology, Software, [26] Roenby J, Larsen B, Bredmose H, Jasak H. A new volume-of-fluid method
in openfoam. In: 7th International conference on computational methods
Validation, Formal analysis, Investigation, Writing - original draft,
in marine engineering. Nantes, France; 2017.
Visualization. Rüdiger Schwarze: Project administration, Funding [27] Brackbill J, Kothe D, Zemach C. A continuum method for modeling surface
acquisition. tension. J Comput Phys 1992;100(2):335–54.
 M. Heinrich and R. Schwarze / SoftwareX 11 (2020) 100483 7

[28] Schiller L, Naumann Z. A drag coefficient correlation. VDI Zeitung [31] Issa R. Solution of the implicitly discretized fluid flow equations by
1935;77:318–20. operator-splitting. J Comput Phys 1986;62(1):40–65.
[29] Gopala Y, Zhang P, Bibik O, Lubarsky E, Zinn B. Liquid fuel jet in crossflow [32] Caretto L, Gosman A, Patankar S, Spalding D. Two calculation procedures
- trajectory correlations based on the column breakup point. In: 48th AIAA for steady, three-dimensional flows with recirculation. In Proceedings
aerospace sciences meeting. Orlando, Florida; 2010. of the third international conference on numerical methods in fluid
[30] Sekar J, Rao A, Pillutla S, Danis A, Hsieh S-Y. Liquid jet in cross flow mechanics; 1972.
modeling. In Proceedings of ASME turbo expo 2014: turbine technical [33] Grinstein F, Margolin L, Rider W. Implicit large eddy simulation: computing
conference and exposition. Düsseldorf, Germany; 2014. turbulent fluid dynamics. Cambridge: Cambridge University Press; 2007.