You are on page 1of 15

MODELING AND SIMULATION OF

NANO-ALUMINUM SYNTHESIS IN A
PLASMA REACTOR

Nelson Settumba and Sean C. Garrick Formatted: Font: 11 pt


Department of Mechanical Engineering
University of Minnesota
Minneapolis, MN USA

ABSTRACT

The synthesis of aluminum (an energetic material) nanoparticles in a plasma Deleted: ,


reactor is simulated. The effects of flow-field mixing on nanoparticle growth Deleted: ,
are investigated via direct numerical simulation. The flow consists of high Deleted: is
temperature argon/aluminum jet impinges on a low-temperature argon jet. To
analyze the influence of fluid dynamic mixing on nanoparticle growth, the
momentum ratio of the two jets is varied. The flow-field is obtained by solving Deleted: are
the compressible Navier-Stokes equations while the evolution of the particle
field is obtained by using a nodal approach to represent the aerosol general
dynamic equation. The results indicate that increasing the momentum of the Deleted:
cooler jet increases dilution of the aluminum jet and increases flow-through Deleted: d
time of nanoparticles (the time required by particles to travel the length of the
domain).

INTRODUCTION

Vapor-phase synthesis is among the preferred methodologies for the


manufacture of nanoparticles. The growth of nanoparticles in such processes is
dependent upon several factors, including nucleation, coagulation,
condensation, evaporation, and transport, to name a few. Thermal plasmas offer
several advantages for nanoparticle synthesis, including virtually complete
dissociation of reactants to their elements, virtually complete chemical
flexibility, and, because of extremely high energy densities, high throughputs.
Particles nucleate in a rapidly quenched environment, which inherently leads to
high nucleation rates and the production of particles having nano-scale
dimensions. A dominant mechanism in many problems is convection, whereby
the material (reactants, precursors, particles, etc.) is in part carried along with
the fluid velocity field of the fluid.1–6 The growth of nanoparticles in turbulent
flow systems is dependent on the time history of particles and the underlying
fluid, thermal, and chemical fields.7 The simulation of nanoparticle synthesis
requires that mathematical and computational models capture the effects of
chemical reactions, nucleation, condensation, evaporation, and coagulation
amongst other physico-chemical phenomena. The effects of physical
parameters, such as Reynolds number and temperature, have been extensively
investigated.8–12 Computational tools which can predict particle properties (size, Deleted: -
shape, composition, morphology, etc.) are very useful in helping to design Deleted: -
nanoparticle production facilities.

This work is towards the synthesis of aluminum nanoparticles in a counter-


flowing plasma reactor. Ultrafine aluminum has been identified as a very
promising ingredient of new metastable intermolecular composites which
exhibit widely tunable energy release characteristics. The main obstacle in the
development of these new materials is the availability of consistent metal
powders with well-defined particle size distribution and protective, oxidation
resistant coating. The counter flow process has the advantage that the cold
counter flowing gas will produce a thermal boundary layer, the location of
which can be controlled by the relative momentum of the jets. In so doing, we
can control the quench rate of the plasma and, therefore, provide very tight
control of the formation and growth processes, which are very sensitive to Deleted: is
temperature. The Navier-Stokes equations and the general dynamic equation is
solved in a model-free manner to provide the fluid, thermal, chemical, and
particle fields as a function of space and time. Additionally, the effect of
counter-flow velocity on the particle size distribution is studied. Deleted: and

FORMULATION

The flows under consideration are governed by the conservation of mass,


momentum, and enthalpy equations:

(1)

(2)

and , (3)
where ρ is the fluid density, ui is the velocity in the i direction, p is the fluid
pressure, τ ij is the stress tensor for a Newtonian fluid, h is the enthalpy, k the
thermal conductivity, and C p is the specific heat.

The fluid consists of argon and aluminum vapor, which is considered to be a


trace species. The equation of transport of aluminum vapor, with a nucleation
sink term ω& N for the aluminum vapor mass concentration, is given by the
species conservation equation:

(4)

where Y is the mass fraction of the condensable aluminum vapor and DY is its
diffusion coefficient. The heat release due to nucleation is neglected in this
work since we only deal with low mass conversions, but its inclusion would be
a simple thing.

The transport of nano-scale particles dispersed in the fluid is governed by the


aerosol general dynamic equation (GDE). The GDE describes particle
dynamics under the influence of various physical and chemical phenomena -
convection, diffusion, coagulation, surface growth, nucleation, and other
internal/external forces. The GDE is written in discrete form as a population
balance on each cluster or particle size. From a practical standpoint, such
systems of equations cannot be solved explicitly except for very small particle
sizes, typically less than one thousand molecules. Our methodology uses the
nodal/sectional method to approximate the GDE.12-18 This approach effectively
divides the aerosol population into three classes - monomers, clusters, and Deleted: l
particles. The GDE is therefore solved as a set of Ns transport equations, one for
each section Qk, k = 1,2,. . . , Ns. Monomers of size 0.5nm in diameter populate
section 1, while sections 2 and 3 are populated by clusters of molecules. Deleted: u
Molecular clusters of size 1nm and larger are considered “particles.” The
general transport equation for the concentration of monomers, clusters, and
particles in bin k, Qk is written as:

(5)

Where DQ is the diffusivity given by


(6)

kb is the Boltzman constant, Cc is the Cunningham correction factor, and dp is


the particle diameter.19 The source terms ω& kN and ω& kQ represent the effects of
formation and growth, respectively, and are given by

(7)

where Xijk is given by

(8)

The collision frequency function β ij is that for Brownian collisions in the free-
molecular regime and is given by

(9)

where T is the fluid temperature, vi is the volume of a particle in the ith section,
and ρ p is the particle density. The nucleation of the nanoparticles is assumed
to occur in the nozzle, upstream of the computational domain (see Fig. 1), and
therefore, nanoparticle nucleation, ω& NK , is not computed directly. This has the Formatted: Not Highlight

effect of greatly reducing the compute time for each simulation. We have
performed other simulations of aluminum nanoparticle nucleation and believe
this assumption to be quantitatively correct, as the nucleation rate in the jet is Deleted: both
some 12 orders of magnitude lower than that upstream of the nozzle. Deleted: -

RESULTS

Flow Configuration

The flows simulated consist of the impingement of two, three-dimensional


roundjets, a schematic of which is shown in Fig. 1. The hot jet (flowing left-to-
right) is composed of aluminum vapor diluted in Argon at a temperature of
2,000K and issues with a velocity of 300m/s while the cool jet (flowing right-
to-left) issues at a temperature of 1,000K. The jets are separated by 30cm or 6
jet diameters, 6D. The velocity of the “cool” jet is varied in this investigation.
The spatial coordinates are x=[x, y, z] where x is the streamwise direction and y
and z are the cross-stream directions. The Reynolds number based on the jet
diameter, D, and the velocity of the hot jet is Re = U1D/v = 2,000. A total of
fifteen bins are used to represent/discretize the particle field, ie. Ns = 15,
yielding the coverage of particle volumes spanning 4 orders of magnitude, or a
diameter range of 1 ≤ dp ≤ 25.4nm. Deleted: In

Figure 1: Opposed-jet, plasma flow reactor (computational domain in red) Formatted: Centered
Deleted: .
Formatted: Line spacing: single
In this preliminary investigation, we assume that nanoparticle nucleation occurs
upstream of the computational domain in the nozzle of the plasma reactor and
that the aluminum enters the domain as 1nm diameter particles. The volume
fraction is assumed to be Φ = 1.0 × 10−7 which corresponds to a particle
concentration of 1.78 × 1020 particles per second per m3. Four simulations are
performed. The velocity of the “cool” jet in each simulation is changed in
increments of 25%. That is, the velocities are U2 = −0.25, U1, U2 = −0.5, U1, U2
= −0.75 U1, and U2 = − U1.

Numerical Specifications

The governing equations are solved in conservative form via a Deleted:


predictor/corrector-type finite-difference scheme.20,21 The scheme is second- Figure 2: Nanoparticle concentration
contours superimposed on vorticity
order accurate in time and fourth-order accurate in space. Computations are magnitude iso-surfaces illustrate
performed on a grid consisting of 320 × 180 × 180 points. The grid spacing is flow/particle dynamics¶
clustered in the x, y, and z directions to provide better resolution in the ¶
boundary layers r/D = ±0.5 and near the stagnation region of the two jets. The Deleted: is
final resolution and grid-clustering parameters were arrived at by increasing the
resolution until the results were free of oscillations.

Flow Dynamics
Instantaneous contours of the concentration of 3nm diameter particles
superimposed on an iso-surface of vorticity magnitude are shown in Fig. 2. The Deleted: is
vorticity is the curl of the velocity, ωz = ∇ × V , and is a measure of the local
rate of rotation. The location of the cool jet can be inferred by the absence of
3nm diameter particles. The figure shows that the two jets meet near x/D = 1.5.
The presence of the vortex rings illustrate the complex flow patterns that result
from the hot and cool jets impinging.

Figure 2: Nanoparticle concentration contours superimposed on vorticity


magnitude iso-surfaces illustrate flow/particle dynamics
Deleted: ¶
The effects of the jet impingement are better seen by following a “fluid
element” as it travels through the flow domain. (In a steady-flow, this would
essentially be along a stream or path line.) The evolution of the normalized Deleted: essentially
particle mass-fraction is shown in Fig. 3.

Figure 3: Evolution of normalized nanoparticle mass across flow reactor


The normalized mass-fraction is the local mass-fraction divided by the mass-
fraction issuing at the center of the hot jet. It is essentially a conserved scalar
which varies from zero to unity and can be used to illustrate the effects of
convection and diffusion on the nanoparticle field. The plot shows that, in all Deleted: the
flows, the mass fraction along the path line decreases as the hot jet travels
across the reactor. This decrease represents the effect of laminar diffusion
(molecular mixing). At a certain point, however, the mass fraction drops rather
precipitously. The point where this transition occurs is a good indication of the
impingement zone or stagnation region of the two jets. This point is near x/D =
2.7 in the 25% counter-flow, x/D = 2.45 in the 50% counter-flow, x/D = 2.15 in
the 75% counter flow, and x/D = 1.75 in the 100% counter-flow. This trend
indicates that an effect of the counter flow is to dilute the particle-laden hot jet. Deleted: ¶

An indication of particle growth dynamics is the evolution of the total particle
number. As the nanoparticles collide, coagulate, and grow, they reduce in
number. Coagulation is the dominant growth mechanism and, as indicated in
Eq. (7), the rate of coagulation is proportional to the square of the number
concentration. Figure 4 shows the change in particle number concentration for
the four different flows.


Figure 3: Evolution of normalized
nanoparticle mass across flow reactor¶
¶ ... [1]
Deleted:

Figure 4: Evolution of nanoparticle number-concentration across flow reactor

The number concentration exhibits the same trend as the particle mass fraction.
That is, it decreases slowly until the stagnation region. The rate of decrease is
faster due to the added effect of coagulation. As the fluid travels downstream,
in addition to the effects of dilution, the particles collide and grow. As they
grow, their number is reduced, satisfying mass conservation, and particles
move from lower- to higher-numbered bins. Figure 4 indicates that the U2/U1 =
−1 flow exhibits the greatest reduction in particle number. The total number of
particles is an order of magnitude less than the U2/U1 = −0.25 flow.

Deleted: ... [2]


Mean Size and Standard Deviation

We will present two moments of the particle size distribution along a streamline
as the hot jet travels down the reactor and encounters the cooler jet - the mean
particle diameter and the geometric standard deviation of the particle size Deleted: ,
distribution. The impact of the cooler jet is two-fold. Vorticity fields generated Deleted:
as the two jets impinge create circulation zones where the two jets mix, both Deleted: s
macroscopically and microscopically. The growth and temperature dynamics
may be elucidated by showing the mean particle size and temperature along Deleted: a
several path lines. A number of representative particle trajectories are shown in
Fig. 5. The colors correspond to particle temperature and the size of the spheres
to mean diameter. The diameter of the sphere is proportional to the actual mean
diameter and, therefore, only particle growth can be inferred from the image.

Figure 5: Representation of the cooling and growth of nanoparticles


in the flow reactor

Figure 5 shows that the particle temperature remains constant at roughly


2,000K until the hot jet encounters the counter-flowing cool jet. At that point,
the particle temperature and growth-rate drop rapidly. The arrest of particle Deleted: s
growth occurs because the size-independent portion of the collision kernel
appearing in Eq. (7) is proportional to the square-root of temperature T. As the
temperature is reduced, all collisions are reduced and coagulation decreases.
This is further compounded by the effects of the aforementioned “dilution.”
This qualitatively shows that an effect of the counter-flowing cool jet is to
arrest particle growth via temperature reduction.
A more quantitative view is shown in Fig. 6. The mean diameter is a measure
of the overall particle size and is quantified by dm = (6vm/π)1/3, where vm is the
mean volume given by

(10)

The growth rate (slope of the lines) is maximum near the inlet where there is a
high concentration of 1nm diameter particles. The growth of the particles is
identical in all flows until roughly x/D = 0.85. At this point, the particles in the
lower counter-flow simulations grow more slowly. The figure suggests that the
change in growth is a result of a change in residence time. Other than the
“kink” appearing near x/D = 0.95, the growth is identical until the stagnation
region. The results also reveal that, though there are differences in the growth
Deleted:
dynamics, the final mean size is roughly the same in all flows. Increasing the ¶
counter-flow can, therefore, be viewed as a mechanism for arresting growth Figure 6: Evolution of mean particle
earlier. diameter across flow reactor¶

Figure 6: Evolution of mean particle diameter across flow reactor

The geometric standard deviation (GSD), σ g, a measure of the width of the


particle size distribution and, therefore, the polydispersity of the particle field is
given by19,22

(11)
where
Deleted: ¶

(12)

The evolution of the GSD in jet-flows has been reported by Miller and
Garrick.23 In the “unperturbed” jet-core, where there are no concentration
gradients, the GSD increases from the initial value of σ = 1 to the self-
preserving value for nodal method utilizing successive doubling of volume, σ
= 1.5.7,24–26 However, at the interface of the two streams, where strong
concentration gradients are present, the diffusive transport acts to increase the
GSD. Vortical structures then act to distribute these regions throughout the flow
domain. A higher mixing rate acts to homogenize the particle field and, as a
result, the concentration gradients smear out more quickly.27 The evolution of
the GSD along a fluid path-line is shown in Fig. 7.
Figure 7: Evolution of geometric standard
deviation across flow reactor¶
Formatted: Justified
Deleted: have
Deleted: .
Deleted:
Deleted:
Deleted: H

Figure 7: Evolution of geometric standard deviation across flow reactor

The initial value of the GSD is σ = 1. It then increases as the particles collide,
coagulate, and increase the polydispersity. The plot shows that, in all flows, the
particle size distribution has not reached the self-preserving value. Instead, a
maximum value of σ = 1.38 is reached. Then, the particle size distribution Deleted: t
begins to narrow. The GSD decreases where the large structures dominate the
flow. This effect has been observed in planar jets and is thought to be the result
of large-scale vortices homogenizing the particle field. In the work of Miller Deleted: from
Deleted: T
and Garrick,23 the GSD stops increasing as soon as the large-scale structures
form.
SUMMARY AND CONCLUSIONS

Direct numerical simulations of nanoparticle coagulation in compressible round


jets were performed. A jet with velocity of U1 issues from a nozzle of diameter
D to impinge on a jet with velocity U2, issuing from a diameter D. The Formatted: Not Superscript/
Reynolds number based on the jet diameter and the velocity of the high-speed Subscript
stream is Re = 4000. The fluid field was obtained using a Navier-Stokes solver Deleted: ,
in a model-free manner and a nodal approach was employed in obtaining the
particle field. The methodology solves for the evolution of the concentration of
particles of various sizes in an Eulerian manner. As a result, the entire fluid-
particle field was obtained as a function of space, time, and size. The governing
transport equations are solved using a hybrid MacCormack based compact
difference scheme which is second order accurate in time and fourth order
accurate in space. The particles considered are spherical non-charged
nanoparticles and the phenomenon of Brownian coagulation is assumed as the
sole instrument influencing the particle growth. This study investigates the
effects of the fluid field on nanoparticle coagulation and growth. Four flows are
simulated corresponding to counter-flow velocities of 25%, 50%, 75%, and
100%. The particle field was studied along fluid path lines.

Results indicate that increasing the counter-flow velocity increased residence


time, increased the dilution of the particle field, and increased the growth rate; Deleted: ,
therefore, establishing counter-flow as a viable means for altering the particle
size distribution. The methodology is sufficiently general to allow for the
inclusion of many physical phenomena, such as nucleation, condensation,
evaporation, thermophoresis, and other fluid-particle interactions. Each
simulation required 2,400 CPU hours on an IBM SP. These results provide a
better insight into the role of fluid fields on nanoparticle growth and
coagulation. They may provide an avenue to further improve the high-rate
synthesis of nano-structured materials. Our future activities include utilizing a
homogeneous nucleation model for the aluminum nanoparticles upstream of the
reactor “chamber,” as well as performing calculations for the same conditions
as those in the experiments being performed at the Center for NanoEnergetic
Research.
ACKNOWLEDGMENTS

This research is supported through the DURINT program under grant DAAD
19-01-1-0503 by the Army Research Office. Computational resources are
provided by the Minnesota Supercomputing Institute for Digital Simulation and
Advanced Computation.
REFERENCES Deleted: References

1
Kennedy, IM (1985) Flow Field Effects on Nucleation in a Reacting Mixing Layer,
Phys. Fluids A., 28:3515–3524.
2
Shuen, JS, Solomon, ASP., Zhang, QF., and Faeth, GM, Structure of Particle-Laden
Jets: Measurements and Predictions, AIAA J., 23:396–404.
3
Hansell, D, Kennedy, IM, and Kollmann, W (1992) A Simulation of Particle
Dispersion in a Turbulent Jet, Int. J. Multiphase Flow, 18:559–576.
4
Mashayek, F, Jaberi, FA, Miller, RS, and Givi, P (1997) Dispersion and Polydispersity
of Droplets in Stationary Isotropic Turbulence, Int. J. Multiphase Flow, 23:337–355.
5
Jenkins, TP and Kennedy, IM (2000) Measurements of Aerosol Product in an
Axisymmetric Co-flow Jet, Experiments in Fluids, 29:532–544.
6
Jenkins, TP and Kennedy, IM (2000) Probability Density Functions of Product
Concentration in a Turbulent Reacting Coflow Jet of NH3 and HCl, Phys. Fluids,
12:3050–3059.
7
Xiong, Y and Pratsinis, SE (1991) Gas Phase Production of Particles in Reactive
Turbulent Flows, J. Aerosol Sci., 22:637–655.
8
Wagner, C, Huttl, TJ, and Friedrich, R (2001) Low-Reynolds-number Effects Derived
from Direct Numerical Simulations of Turbulent Pipe Flow, Computers and Fluids,
30:581–590.
9
DeGraaff, DB, Webster, DR, and Eaton, JK (1999) The Effect of Reynolds Number on
Boundary Layer Turbulence, Experimental Thermal and Fluid Science, 18:341–346.
10
Heinz, S and Roekaerts, D (2001) Reynolds Number Effects on Mixing and Reaction
in a Turbulent Pipe Flow, Chem. Eng. Sci., 56:3197–3210.
11
Hobbs, DM and Muzzio, FJ (1998) Reynolds Number Effects on Laminar Mixing in
the Kenics Static Mixer, Chem. Eng. Journal, 70:93–104.
12
Modem, S, Garrick, SC, Zachariah, MR, and Lehtinen, KEJ (2002) Direct Numerical
Simulation of Nanoparticle Coagulation in a Temporal Mixing Layer, in Proceedings of
the 29th Symp. (Int.) on Combustion, The Combustion Institute, Pittsburgh, PA.
13
Gelbard, F and Seinfeld, JH. (1980) Simulation of Multicomponent Aerosol
Dynamics, J. Colloid Interface Sci., 78:485–501.
14
Gelbard, F, Tambour, Y, and Seinfeld, JH (1980) Sectional Representations for
Simulating Aerosol Dynamics, J. Colloid Interface Sci., 76:541–556.
15
Biswas, P,Wu, CY, Zachariah, MR, and McMillin, B (1997) Characterization of Iron
Oxide-Silica Nanocomposites in Flames: Part II: Comparison of Discrete- Sectional
Model Predictions to Experimental Data, J. Mat. Res., 12:714–723.
16
Lehtinen, KEJ and Zachariah, MR (2001) Self-preserving Theory for the Volume
Distribution of Particles Undergoing Brownian Coagulation, J. Colloid Interf. Sci,
242:314–318.
17
Lehtinen, KEJ and Zachariah, MR (2002) Energy Accumulation in Nanoparticle
Collision and Coalescence Processes, J. Aerosol Sci., 33:357–368.
18
Modem, S and Garrick, SC (2003) Nanoparticle Concentrations in Temporal Mixing
Layers: Mean and Size-selected Images, J. Visualization, 6:333–342.
19
Friedlander, SK (1977) Smoke, Dust and Haze: Fundamentals of Aerosol Behavior,
John Wiley and Sons, New York, NY.
20
Carpenter, MH (1990) A High-Order Compact Numerical Algorithm for Supersonic
Flows, in Morton, K. W., editor, Twelfth International Conference on Numerical
Methods in Fluid Dynamics, Lecture Notes in Physics, Vol. 371, pp. 254–258,
Springer-Verlag, New York, NY.
21
Kennedy, CA and Carpenter, MH (1994) Several New Numerical Methods for
Compressible Shear-Layer Simulations, Appl. Num. Math., 14:397–433.
22
Fuchs, NA (1964) The Mechanics of Aerosols, Pergamon, Oxford, England.
23
Miller, S and Garrick, SC (2004) Nanoparticle Coagulation in a Planar Jet, Aerosol
Sci. Technol., 38:79–89.
24
Xiong, Y and Pratsinis, SE (1993) Formation of Agglomerate Particles by
Coagulation and Sintering - Part I. A Two-dimensional Solution of the Population
Balance Equation, J. Aerosol Sci., 24:283–300.
25
Pratsinis, SE (1998) Flame Aerosol Synthesis of Ceramic Powders, Prog. Energy
Combust. Sci., 24:197–219.
26
Pyykönen, J and Jokiniemi, J (2000) Computational Fluid Dynamics Based Sectional
Aerosol Modelling Schemes, J. Aerosol Sci., 31:531–550.
27
Ottino, JM, (1989) The Kinematics of Mixing: Stretching, Chaos, and Transport,
Cambridge University Press, Cambridge, UK.
Page 7: [1] Deleted Sandi Richter 1/15/2007 4:28:00 PM

Figure 3: Evolution of normalized nanoparticle mass across flow reactor

Figure 4: Evolution of nanoparticle number-concentration across flow reactor

Page 7: [2] Deleted Sandi Richter 1/15/2007 4:31:00 PM


Figure 5: Representation of the cooling and growth of nanoparticles
in the flow reactor

You might also like