Numerical simulations of gas-focused micro jets
VII International Conference on Computational Methods for Coupled Problems in Science and Engineering
COUPLED PROBLEMS 2017
M. Papadrakakis, E. Oñate and B. Schrefler (Eds)
NUMERICAL SIMULATIONS OF MICRO JETS PRODUCED WITH A
DOUBLE FLOW FOCUSING NOZZLE
GREGA BELŠAK1, SAŠA BAJT2, KENNETH R. BEYERLEIN3
AND BOŽIDAR ŠARLER1,4
1
Laboratory for Simulation of Materials and Processes
Institute of Metals and Technology
Lepi pot 11, SI-1000 Ljubljana, Slovenia
email: grega.belsak@imt.si
2
Photon Science
Deutsches Elektronen-Synchrotron DESY
Notkestraße 85, 22607 Hamburg, Germany
email: sasa.bajt@desy.de
3
Center for Free-Electron Laser Science
Deutsches Elektronen-Synchrotron DESY
Notkestraße 85, 22607 Hamburg, Germany
email: kenneth.beyerlein@cfel.de
4
Laboratory for Fluid Dynamics and Thermodynamics
Faculty of Mechanical Engineering
University of Ljubljana
Aškerčeva c. 6, SI-1000 Ljubljana, Slovenia
email: bozidar.sarler@fs.uni-lj.si
Key words: Multiphase Flow, Micro Jet, Double Flow Focusing Nozzle
Abstract. Stable and reliable micro jets are important for many applications. Double flow
focused micro jets are a novelty with an important advantage of significantly reduced sample
consumption. Numerical simulations of double flow focused micro jets are a highly complex
task. They represents a great computational challenge due to the multiphase nature of the
problem, strong coupling between the gas and the two liquids and the sub-micron size cells
needed. Simulations were performed with the open source computational fluid dynamics
toolbox called OpenFOAM. Two multiphase solvers were used, one of which was modified in
order to properly describe the interface between the focusing liquid and the gas. In this study
two different incompressible physical models were considered and compared. A model with
no mixing of the two fluids (multiphaseInterFoam solver) and a model where the diffusion of
the two fluids is permitted (modified interMixingFoam solver). The results of simulations for
the two different physical models using the same inlet parameters are presented. Additionally,
a parametric analysis for the mixing case was performed to study the effects of different
parameters on the jet formation. Particularly how the different diffusion values couple with
the jet length, diameter and its stability. Results show a match in jet diameter and jet length
for both models when the same set of parameters is used.
1239
Grega Belšak, Saša Bajt, Kenneth R. Beyerlein and Božidar Šarler.
1
INTRODUCTION
Controlled production of liquid jets by means of a co-flowing immiscible fluid stream can
have diverse technical applications. One of these applications is in the field of X-ray serial
crystallography. Nozzles that create stable, long and fast jets of just a few micrometres in
diameter are needed to deliver protein nanocrystals into the intense X-ray beam. X-rays
scattered off these crystals create diffraction patterns that are recorded on a detector.
Reconstructed diffraction images provide the atomic resolution protein structure. The main
bottleneck in the protein structure determination is the sample preparation, especially for the
membrane proteins which do not like to form larger crystals. Serial femtosecond
crystallography with x-ray free electron lasers (FEL) opened up the possibly to obtain protein
structures also from nanocrystals, which were previously too small for standard X-ray
crystallography. Nevertheless, samples are hard to prepare and the amount of the material is
very limited. It is critical to develop ways of using the minimum amount of sample material.
Delivering such nanocrystals to the X-ray beam in a form of a micro jet proved to have
several advantages [1]. Here, we are particularly interested in understanding the nozzle
geometry that creates these micro jets. In the past such nozzles were prepared manually,
which was time consuming, non-reproducible and limited to simple designs. Ceramic microinjection moulded nozzles were a step forward ensuring reproducibility and faster assembly
[2]. However, because of the high cost of the moulding tools it is desirable to test new designs
before investing in a new moulding tool. Recently, a 3D printing technology enabled printing
of macroscopic nozzles with a very high precision [3]. These nozzles can be used either for
testing a new design or in final application. The development of numerical models presented
here, gives an insight in the fluid dynamics of such systems and should help to improve future
nozzle designs.
2
DOUBLE FLOW FOCUSING NOZZLE DESIGN
Early experiments were performed using gas dynamic virtual nozzles (GDVN) [4]. This
nozzle structure uses two phases to create a stable micro jet:
Figure 1: Graphical representation of the double flow focusing nozzle. Typical values are: R s= 20 µm, Rfl-i=
55 µm, Rfl-o= 62 µm, Rg-i= 175 µm, Rg-o= 245 µm, α =17.5 ͦ, β= 25 ͦ, H1= 70 µm, H2= 85 µm, D= 35 µm.
2
1240
Grega Belšak, Saša Bajt, Kenneth R. Beyerlein and Božidar Šarler.
a liquid sample fluid (nanocrystals dispersed in water) and a focusing gas (helium). In such
set-up two capillaries are inserted into the nozzle. The central capillary that ends almost at
the nozzle orifice is used to deliver sample liquid, while the gas is delivered through the
second capillary that ends further upstream of the nozzle tip. The high pressure gas focuses
the liquid into a micro jet when flowing through a small nozzle orifice. This approach
typically requires sample fluid flow rates of around 20-40 µl/min.
In order to reduce the sample fluid consumption a novel double flow focusing nozzle
(DFFN), depicted in figure 1, was developed [5,6]. This approach uses an additional fluid
(alcohol) to further focus the sample fluid. The main advantage of using alcohol is its lower
surface tension in comparison to water. It acts as a sheath liquid encapsulating the water jet,
resulting in extension of the jet length by mitigating its breakup. In this way the sample
fluid flow rate can be reduced to around 5 µl/min.
3
GOVERNING EQUATIONS
The multiphase model of isothermal and incompressible flow is governed by the sets of
momentum and mass conservation equations for each of the phases i:
( i i ui )
i i u i u i i p i iu i i i g F s ,i
t
i
ui i 0
t
(1)
(2)
where velocity, phase fraction, density, viscosity and surface tension force for phase i are
given by u i , i , i , i , F i , respectively and g is gravity. The interface compression method
[7] is implemented by adding an additional compression term to the mass conservation
equation in order to compress the volume fraction field and maintain a sharp interface
between the phases.
i
ui i u c i 1 i 0
t
(3)
Compression velocity u c is applied normally to the interface.
4
NUMERICAL PROCEDURE
Numerical simulation of DFFN micro jets is a highly complex task. Multiphase nature of
the problem, strong coupling between the gas and the liquids, the sub-micron size cells
needed for high resolution and proper capturing of the flow all represent a great
computational challenge. Because of the microscopic nature of the nozzle structure and the
physical properties of the fluids used, the Reynolds number is low and therefore the flow is
considered laminar. The fluids are considered to be of a Newtonian nature. The chapter is
divided into three specific parts, each one describing in details the performed work.
3
1241
Grega Belšak, Saša Bajt, Kenneth R. Beyerlein and Božidar Šarler.
4.1 Pre-processing
The computer model of a DFFN was prepared with FreeCAD, an open-source parametric
3D CAD modeler [8]. Non-axial symmetry of the nozzle structure (circular inner capillary
inserted into square middle capillary inserted into circular outer capillary), was treated as
axially symmetric (circular inner capillary inserted into circular middle capillary inserted into
circular outer capillary) while keeping the cross-sectional area of the channel equal as seen in
figure 2 . In this way a three dimensional problem was reduced to a two dimensional one, thus
greatly reducing the calculation time and making simulations of micro jets feasible.
Figure 2: Transformation of the real geometry of the nozzle to axis symmetry.
For the preparation of the high quality mesh the utility called snappyHexMesh was used,
which is a part of the open source computational fluid dynamics (CFD) toolbox called
OpenFoam [9]. A sample mesh can be seen in figure 3. For the simulations to be run in a
reasonable time (up to few days) on a modern computer with approximately 30 cores the
number of cells needed to be kept as low as possible. This proved to be a difficult task for two
reasons. The first reason is the desired high resolution in the jet region. Experiments show
that the typical jet diameter for a DFFN is between 3 and 5 µm. At least 10 cells are needed to
properly describe the fluid flow and the four interfaces between the two liquid phases. This
constrains the maximal cell size to 0.5 µm. Therefore, a cell size of 0.15 µm was chosen in
this study. To keep the computing time reasonable we used the finest mesh only in the area
where the jet was expected to form and a coarser mesh elsewhere. The second reason is that
the vacuum chamber, the area where the jet leaves the nozzle, needs to be large enough (few
millimeters in length). This is because we are setting an artificial condition ( p 0) on the
outlet boundary of the vacuum chamber. In order to avoid the numerical errors and to prevent
any interference of this artificial boundary condition on the jet formation, the size of the
computational domain needs to be few millimeters. Those two constraints led to a mesh with
the finest cell size of 0.15 µm with ~ 225 000 cells.
Figure 3: Representation of a mesh used in the simulations
4
1242
Grega Belšak, Saša Bajt, Kenneth R. Beyerlein and Božidar Šarler.
4.2 Processing
Numerical simulations of DFFN were performed with OpenFoam (version 16.10), which
has a variety of solvers available to use for many different kinds of fluid flow problems. In
this study two different models were considered and hence two different solvers were used.
First, a “multiphaseInterFoam” solver was used for a multiphase model where all the fluids
are considered incompressible and there is no diffusion between the phases i.e. a non-mixing
incompressible model. Second, a modified “InterMixingFoam” solver was used, which
describes a set of three incompressible fluids two of which are miscible, i.e. a mixing
incompressible model. In the later model diffusion between the sample fluid and the focusing
fluid is permitted. As aforementioned the code in this solver had to be slightly modified to
properly describe the interface between the focusing fluid and the gas. The inlet parameters
and the physical properties of the fluids at room temperature (Table 1) were chosen to
resemble the experimental values [6] and were the same in both models.
Table 1: Operating conditions and physical properties used in simulations. Values were obtained from NIST
Chemistry Webbook Database
Density [kg/m3]
Dynamic viscosity [kg/ms]
Volumetric flow rate [µL/min]
Mass flow rate [mg/min]
Surface tension (water-gas) [N/m]
Surface tension (alcohol-gas) [N/m]
Surface
tension
(water-alcohol)
[N/m]
Diffusion (water-alcohol) [N/m2]
sample liquid
focusing liquid
focusing gas
WATER
1000
1.9*10-5
5
/
ALCOHOL
789
1.12*10-3
10
/
0.0728
0.0223
0.0505
HELIUM
0.33
10-3
/
21.6
mixing case
non-mixing case
10-9
0
4.3 Post-processing
Post-processing of the simulations was performed with ParaView [10], an open source,
multi-platform data analysis and visualization application. A code was written to
automatically extract the jet length, diameter and concentration profile, discussed and
presented in the results section. When setting up the simulation case and choosing the velocity
inlet boundary conditions for the fluids a uniform axial flow (constant velocity profile) was
chosen. There was a concern that this non-physical constant profile would affect the
simulations. However, results demonstrated that this is not the case if the capillary is of
sufficient length (above 100 µm) and the jet is monitored long enough (t > 0.3 ms). Under
these conditions the initial constant profile changes to parabolic profile. Full development of
this profile along with stabilization of the recirculation zones was used as the benchmark of a
steady-state solution.
5
1243
Grega Belšak, Saša Bajt, Kenneth R. Beyerlein and Božidar Šarler.
5
RESULTS
5.1 Mixing model.
The physical model, which permits the mixing between the phases, is explored here. In this
three phase system the sample fluid and the focusing fluid were miscible with a diffusion
constant of 10-9 m2/s, which is a typical value for water-ethanol system. None of these liquids
were allowed to mix with the third, gaseous phase. This model explores how diffusion affects
the jet length, the diameter and the concentration profile and allows for a comparison with the
experimental data [6]. In the mixing model it is challenging to distinguish between the natural
(real) and the numerical (artificial) diffusion. The artificial diffusion arises from the spatial
and temporal discretization of a continuous problem and therefore highly depends on the cell
size and the time step. The following discretization parameters were chosen in order to keep
the numerical diffusion an order of magnitude lower than the natural diffusion (10-10 m2/s) and
to prevent it to interfere with the results. In the region of the domain where the diffusion is
present, the maximal cell size was set to 0.15 µm. The time step was controlled by setting the
Courant number to the value of one, which also ensured stability of the simulation. Results
are presented in figure 4.
Figure 4: A snapshot of the simulation of the mixing model at time 0.4 ms. The extracted parameters are jet
diameter dj = 4.5 µm and average jet length Lj = 94.2 µm.
One of the main results of this study is the dependence of the concentration profile, which
is measured at the nozzle orifice perpendicularly to the jet axis, to the varying parameters.
Figure 5 shows the water concentration profile through the jet for two diffusion values. In the
jet only two phases are present: water and alcohol. The total sum of both concentration phases
is equal to one. It can be observed that along the jet axis (jet radius zero) the water
concentration is at the highest, but still not equal to one, indicating the presence of alcohol
along the jet axis. When moving towards the edges of the jet the water concentration
decreases, since the alcohol concentration increases. Increasing the diffusion coefficient by an
order of magnitude (green line) reduces the concentration of water around the jet axis. This is
expected, since higher diffusion coefficient means more alcohol is mixed inside the water.
6
1244
Grega Belšak, Saša Bajt, Kenneth R. Beyerlein and Božidar Šarler.
Results in figure 5 show a noticeable difference in the concentration profile for two
diffusion values. This indicates that the observed diffused interface between the water and the
alcohol is a result of natural and not numerical diffusion. It is interesting to note that changes
of the diffusion coefficient do not affect the jet diameter.
Figure 5: Water concentration profile of a jet measured at the nozzle orifice in the perpendicular direction to
jet axis. Two different diffusion values are considered.
5.2 Non-mixing model.
Additionally, a multiphase model consisting of three incompressible fluids and no
diffusion between the phases was explored. Results for this immiscible case are presented in
figure 6.
Figure 6: A snapshot of the simulation of the non-mixing model at time 0.4 ms. The extracted parameters are jet
diameter dj = 4.8 µm and average jet length Lj = 97.5 µm.
7
1245
Grega Belšak, Saša Bajt, Kenneth R. Beyerlein and Božidar Šarler.
5.3 Model comparison
Results of simulations for miscible and immiscible model are presented in table 2.
Calculations were performed on the same mesh under identical operating conditions and
physical parameters. Findings indicate that diffusion does not affect the jet diameter and only
slightly affects the jet length. Interesting thing to note is that in the non-mixing model small
water droplets are forming inside the alcohol jet.
Table 2: Comparison of extracted parameters
Mixing model
Non-mixing model
average jet LENGTH [µm]
jet DIAMETER [µm]
94.2 ± 0.3
97.5 ± 0.3
4.5 ± 0.3
4.8 ± 0.3
The surprising result is the recirculation zone in the meniscus of the jet. In an immiscible
model a stable sample fluid recirculation zone is established. On the other hand in the
miscible model there is no sample fluid recirculation, but only a small focusing fluid
recirculation in the outermost layers of the jet as shown in figure 7.
Figure 7: Comparison of recirculation zones. Left panel non-mixing model with recirculation zone. Right
panel mixing model where only small recirculation occurs in the outer most layers of the focusing liquid
phase.
We believe that with different operating conditions of the gas (higher gas speeds inside the
nozzle) the recirculation zone would become even stronger and would also appear in the
miscible case.
The numerical results published in [6] differ from the ones obtained here which we
attribute to different initial conditions of the gas. In the previous work we assumed lower
helium mass flow rate and inserted the gas into the nozzle under higher pressure. As a result
the maximal gas velocity developed inside the nozzle orifice was around 65 m/s. Under the
present operating conditions, the maximal gas velocity reaches a value of around 350 m/s,
resulting in a thinner and shorter jet.
8
1246
Grega Belšak, Saša Bajt, Kenneth R. Beyerlein and Božidar Šarler.
6
CONCLUSIONS
The scope of this work included numerical simulations of DFFNs. All the fluids in these
simulations were considered incompressible. Experimental data [6] indicate chocked flow for
the gas flowing through such DFFN into a vacuum chamber. Correct description requires a
model with incompressible sample and focusing liquid, and compressible focusing gas. It is
conceivable that the simulations, where the compressibility is taken into account, would result
in different values of the jet diameter and length under the same initial conditions. The length
of the jet is expected to change (shorten) when compressibility is added, because we would be
able to describe the expansion of the high pressure gas into the low pressure vacuum chamber.
This would result in higher gas velocities inside the vacuum chamber. Although not supported
with full simulation, we predict that this, along with the changed gas stream shape will affect
the jet length and stability. Jet diameter will also be affected by a decrease in pressure and
density of gas and increased gas velocity at the nozzle orifice. Future work will include
upgraded, more realistic models to address these issues.
7
ACKNOWLEDGEMENTS
We would like to thank Henry N. Chapman (CFEL, Univ. Hamburg, CUI), Dominik
Oberthuer (CFEL), Juraj Knoška, and Max O. Wiedorn (CFEL, Univ. Hamburg) for fruitful
discussions, and Luigi Adriano (DESY) for technical support. This work was supported by
grant of Slovenian Grant Agency (ARRS) J2-7384, Program Group P0-0501-0782 and by
Helmholtz Association through project oriented funds.
8
REFERENCES
[1] Chapman, H.N., Fromme, P., Barty, A., White, T.A., Kirian, R.A., Aquila, A., Hunter,
M.S., Schulz, J., DePonte, D.P., Weierstall, U., Doak, R.B., Maia, F.R.N.C., Martin, A.V.,
Schlichting, I., Lomb, L., Coppola, N., Shoeman, R.L., Epp, S.W., Hartmann, R., Rolles,
D., Rudenko, A., Foucar, L., Kimmel, N., Weidenspointner, G., Holl, P., Liang, M.,
Barthelmess, M., Caleman, C., Boutet, S., Bogan, M.J., Krzywinski, J., Bostedt, C., Bajt,
S., Gumprecht, L., Rudek, B., Erk, B., Schmidt, C., Hömke, A., Reich, C., Pietschner, D.,
Strüder, L., Hauser, G., Gorke, H., Ullrich, J., Herrmann, S., Schaller, G., Schopper, F.,
Soltau, H., Kühnel, K., Messerschmidt, M., Bozek, J.D., Hau-Riege, S.P., Frank, M.,
Hampton, C.Y., Sierra, R.G., Starodub, D., Williams, G.J., Hajdu, J., Timneanu, N.,
Seibert, M.M., Andreasson, J., Rocker, A., Jönsson, O., Svenda, M., Stern, S., Nass, K.,
Andritschke, R., Schröter, C., Krasniqi, F., Bott, M., Schmidt, K.E., Wang, X., Grotjohann,
I., Holton, J.M., Barends, T.R.M., Neutze, R., Marchesini, S., Fromme, R., Schorb, S.,
Rupp, D., Adolph, M., Gorkhover, T., Andersson, I., Hirsemann, H., Potdevin, G.,
Graafsma, H., B. and Spence, J.C.H. Femtosecond X-ray protein nanocrystallography.
Nature 470, 73–77, doi: 10.1038/nature09750 (2011).
[2] Beyerlein, K.R., Adriano, L., Heymann, M., Kirian, R., Knoška, J., Wilde, F., Chapman,
H.N., and Bajt, S. Ceramic micro-injection molded nozzles for serial femtosecond
crystallography sample delivery. Review of Scientific Instruments, 86(12), (2015)
p.125104. Available at: http://dx.doi.org/10.1063/1.4936843
9
1247
Grega Belšak, Saša Bajt, Kenneth R. Beyerlein and Božidar Šarler.
[3] Nelson, G., Kirian, R.A., Weierstall, U., Zatsepin, N.A., Faragó, T., Baumbach, T., Wilde,
F., Niesler, F.B.P., Zimmer, B., Ishigami, I., Hikita, M., Bajt, S., Yeh, S., Rousseau, D.L.,
Chapman, H.N., Spence, J.C.H., and Heymann, M. Three-dimensional-printed gas
dynamic virtual nozzles for x-ray laser sample delivery. Opt. Express 24, 11515-11530
(2016)
[4] DePonte, D.P., Weierstall, U., Schmidt, K., Warner, J., Starodub, D., Spence, J.C.H., and
Doak, R.B. Gas dynamic virtual nozzle for generation of microscopic droplet streams. J.
Phys. D 41, 195505 (2008).
[5] Ganán-Calvo, A. M., González-Prieto, R., Riesco-Chueca, P., Herrada, M. A. and FloresMosquera, M. Focusing capillary jets close to the continuum limit. Nature Physics 3, 737–
742, doi: 10.1038/nphys710 (2007).
[6] Oberthuer, D., Knoška J., Wiedorn, M.O., Beyerlein, K.R., Bushnell, D.A., Kovaleva,
E.G., Heymann, M., Gumprecht, L., Kirian, R.A., Barty, A., Mariani, V., Tolstikova, A.,
Adriano, L., Awel, S., Barthelmess, M., Dörner, K., Xavier, P.L., Yefanov, O., James,
D.R., Nelson, G., Wang, D., Calvey, G., Chen, Y., Schmidt, A., Szczepek, M.,
Frielingsdorf, S., Lenz, O., Snell, E., Robinson, P.J., Šarler, B., Belšak, G., Maček, M.,
Wilde, F., Aquila, A., Boutet, S., Liang, M., Hunter, M.S., Scheerer, P., Lipscomb, J.D.,
Weierstall, U., Kornberg, R.D., Spence, J.C.H., Pollack, L., Chapman, H.N. and Bajt, S.
Double-flow focused liquid injector for efficient serial femtosecond crystallography. Sci.
Rep. 7, 44628; doi: 10.1038/srep44628 (2017)
[7] Weller, H.G., A new approach to VOF-based interface capturing methods for
incompressible and compressible flow. Tech.Rep., OpenCFD, 2008.
[8] FreeCAD website: https://www.freecadweb.org/
[9] OpenFoam © 2004-2017 OpenCFD Ltd (ESI Group) website: http://www.openfoam.com/
[10]
Ahrens, J., Geveci, B., Law, C. ParaView: An End-User Tool for Large Data
Visualization. Visualization Handbook, Elsevier, 2005, ISBN-13: 978-0123875822
10
1248