Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Sebastian Lambertz
from Duderstadt
at the Institute for Nonlinear Dynamics
Thesis period:
First referee:
Second referee:
Abstract
The aim of the present study was to examine the impact of the different forces acting
on a rigid sphere moving in a Newtonian fluid. The equation of motion proposed
and improved by Basset [Bas88], Boussinesq [Bou85], Oseen [Ose10], Tchen [Tch47],
Maxey and Riley [MR83] and Gatignol [Gat83] was solved for the quasi-steady and
transient case.
A linear motion system with a travel distance of 4.05 m and a maximal force of
7 600 Newton was build up. With a net weight of 115 kg, we were able to run the
moving platform with a velocity up to 7.5 m/sec and accelerations up to 28 m/sec2 .
In the constant driving mode the velocity deviation was 0.1% at a velocity of 5 m/sec.
With this system we investigated the response of a sphere to sinusoidal forcing.
Therefore we filmed the motion of the millimetre sized spheres in a container with
silicon oil. The data fitted the solution quite well and correspond with the measurements found in the literature. We checked the dependence on the Reynolds number
and found that the solution is still correct when both Reynolds number and Strouhal
number are small but of the same order. For the first time, the break-down of the
solution due to non-linear effects was measured.
The results are especially important for the moving of tracer and inertia particles
in unsteady flows.
Keywords:
Basset history force, memory term, fractional calculus, relaxation, low Re flow, nonlinear effects
iii
Contents
1 Introduction
1.1
1.2
1.3
1.4
1.5
1.1.1
1.2.1
1.2.2
Force measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.3.1
1.3.2
1.3.3
1.3.4
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Velocity measurement
2.2
. . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.4.1
1.4.2
Experimental studies . . . . . . . . . . . . . . . . . . . . . . . 22
1.4.3
1.4.4
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2 Theory
2.1
35
2.1.2
2.2.2
2.3
2.4
Contents
2.5
.
.
.
.
.
.
.
.
.
.
.
.
42
44
45
46
47
51
3 setup - censored
53
4 Results
4.1 Quasi-steady state . . . . . . . . . . . . . . . . . .
4.1.1 Full solution versus Quasi-steady state . . .
4.1.2 Assumptions . . . . . . . . . . . . . . . . . .
4.1.3 Measured trajectories . . . . . . . . . . . . .
4.1.4 Results . . . . . . . . . . . . . . . . . . . . .
4.1.5 Comparison with the data in the literature .
4.2 High Reynolds number regime . . . . . . . . . . . .
4.2.1 Constant frequency . . . . . . . . . . . . . .
4.2.2 Constant amplitude . . . . . . . . . . . . . .
4.2.3 The role of the forcing . . . . . . . . . . . .
4.2.4 High Reynolds number and density variation
4.2.5 Summary and application of the results . . .
4.3 Discussion . . . . . . . . . . . . . . . . . . . . . . .
55
55
55
56
58
60
63
65
67
69
70
75
76
77
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
5 Outlook
81
6 Appendix
6.1 Solution in the inertial frame . . . . . . . . . . . . . .
6.1.1 Analysis of the result in the inertial frame . .
6.2 Solution for the co-moving frame . . . . . . . . . . .
6.2.1 Analysis of the result in the co-moving frame .
6.3 Inverse Laplace Transform of the Basset equation . .
83
83
85
88
90
93
vi
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Nomenclature
Latin Letters
Variable
Meaning
Unit
a
As = A
Ap
(Ap /As )0
AcJ
m
m
m
AcOH
r
[TK80] AcT = fp 1 U 2 2a dU
dt
AcT
r,T emkin
cD
cH
cA
cT emkin
d
D
E
f
F
Fdrag
FHistory
Fadded
Fdisturbance,T chen
m
m
kg m s3 A1
sec1
m sec2
kg m sec2
kg m sec2
kg m sec2
kg m sec2
vii
Nomenclature
Variable
Meaning
Unit
Fdisturbance,CL
kg m sec2
Fdrag
Fn
g
g(t)
Gravity acceleration
Time dependent part of force
m sec2
Fdisturbance,Buevich
Fdisturbance,M R
r
Ga
KBasset
KM ei
Ka
Kr
L
mp
mf
mnonlinear
n
p
P
Ref
ReJourdan
kg m sec2
kg m sec2
1 F a3
m
m sec2.5
Galileo number Ga =
m sec2.5
N m1
N m1
m
kg m3
kg m3
kg m3
kg m1 sec2
A m2 sec
ReJourdan =
r
2a vup
dup
dt
Res
p
f
kg m sec2
dup 2
dt
dup,y
dt
2
Rev
Ro
viii
aAp
Nomenclature
Variable
Meaning
Unit
s
S
Sl
T
u
sec
m sec1
m sec1
m sec1
up
uts
u0
u
umean
Ur,T emkin
v
v
V
x0
m sec1
m sec1
m sec1
m sec1
m3
Greek Letters
Variable
Meaning
Unit
Density ratio =
f
p
q q
2
f
0
Expansion parameter
Electric permittivity of free space
(dynamic) Viscosity
Integration variable
2
9
1
S
A2 sec4 kg1 m3
kg m1 sec1
ix
Nomenclature
Variable
Meaning
Unit
kg sec2
f =
p
kg m3
kg m3
1 , 2
kg m2 sec2
m2 sec1
m sec1
2 f
p
9
2 (p + 12 f )
sec1
1
a2 p + 2 f
2
9
Relaxation time
2
r = 92 a fp
sec1
e
0
00
kg sec1
kg m sec2
rad sec 1
Nomenclature
miscellaneous
Meaning
Symbol
Partial derivative
x f (t) =
Total derivative following the particle
f (t)
f (t)
df (t)
=
+ up
dt
t
x
f (t)
f (t)
Df (t)
=
+v
Dt
t
x
f (t)
x
f (t)
f = t f (t) =
t
a It =
t Ib
1 Zt
(t )(1) f ()d
() a
1 Zb
=
( t)(1) f ()d
() t
m
a Dt f (t) =
dn
n
f (t)
aI
dxn t
Laplace Transformation
F (s) = L {f } (s) =
Z
0
sC
xi
1 Introduction
In this thesis work the motion of particles in unsteady flows is examined theoretically
and experimentally. It is organized as follows. We start with a general introduction
and identify proper parameters to characterize the problem. The previous works
are summarized and discussed. Based on the parameter range and the results of the
previous experiments we chose the configuration of our experiment. The following
theory part contains solutions for the governing equations. Then we go to the setup
and explain the constituent parts of our experiment. The fourth section contains
the results of the experiment and we compare our results with the previous results.
The last part contains a outlook based on the new results.
We start this section with a short introduction and explain why the problem is
important for a wide range of applications. Then we go on to the experimental
findings. We compare the flow around fixed and freely moving spheres and discuss
the implications for our experiment. The various experimental and theoretically
work on this problem is divided into two parts depending on the ansatz. We start
with the force measurements at fixed velocity and go on to the velocity measurements
at controlled forcing. The last part of this section summarizes the previous work
and gives an overview, which part of the parameter space we explore experimentally
and why we do this.
1 Introduction
different from the motion of small pollen. Another important quantity is the density
ratio of the particle and the fluid. The motion of a bacteria with almost the same
density as the surrounding water will be different from the motion of an air bubble
with the same size. Also the properties of the flow are important. A strong wind
distributes pollution particles faster than a light breeze. The problem gets even more
complicated, if we allow the flow to be unsteady. A flow with a time dependent
velocity is the most complicated, but also most interesting problem we want to
discuss in this thesis.
The motion of pure fluids and suspensions has been studied for very long time. For
the first time (in 1823) Navier derived the equations, which were named after him
and Stokes. Navier [Nav23] - like Cauchy in 1823 [Cau23], Poisson in 1829 [Poi28]
and Saint-Venant in 1843 [dSV43] - began with the elasticity theory and applied his
equations to fluids, whereas Stokes (in 1845) [Sto46] derived the equations directly
and pointed out their connection to the elasticity theory. Let the fluid velocity be
denoted by v, the dynamic viscosity by and the fluid density by . The external
forces acting on the fluid are denoted by f . Then the Navier-Stokes equations for
an incompressible fluid are given by:
~v
+ ~v ~v = p + 2~v + f~
t
(1.1)
The discussion of the different terms is easier, if we make the equation non-dimensional.
Let T define a typical time scale of the change of the boundary conditions. The velocities are rescaled with the mean velocity U , and the length is rescaled with a
typical length scale L to get the Navier-Stokes Equation in non-dimensional form.
The prime indicates a dimensionless parameter:
v0 =
v
,
U
p0 =
p
,
U2
0 = L,
t0 =
t
.
T
02 v 0 + forces
+ v 0 0 v 0 = 0 p0 +
0
T t
LU
v 0
1 02 0
Sl
v + forces
+ v 0 0 v 0 = 0 p0 +
0
t
Re
v 0
Ro
+ Re v 0 0 v 0 = Re0 p0 + 02 v 0 + forces
t0
|
{z
unsteady term
{z
convective term
{z
inertial terms
| {z }
pressure term
| {z }
viscous term
1 Introduction
L by the radius of the sphere a and the velocity scale U by the flow velocity v of
the windtunnel. The freely moving sphere generates the same flow field. The fluid
flows around the sphere and one can think of different flow layers that are slightly
deformed by the sphere. The flow is also called laminar. An important observation
is that the flow field is axisymmetric.
Figure 1.1: The flow around a sphere at low Reynolds numbers. Picture by Taneda
[Tan56] at Re = 4.575.
Freely moving sphere When the sphere is not fixed, the problem has additionally
six degrees of freedom, namely the three position and the three velocity coordinates
of the sphere. In the low Reynolds number regime, the asymptotic state of the freely
moving sphere is steady and the wakes of the free and fixed sphere are identical. As
shown by Natarajan and Acrivos [NA93] the wake is axisymmetric until a Reynolds
1 Introduction
number Re = ua
up to 105. We will call this case A. At this critical number,
the wake loses its axisymmetry due to a forward Hopf bifurcation. Jenny, Bochet
and Dusek [JBD03] showed that the additional degrees of freedom influence the
symmetry breaking and the wake of the free and fixed sphere cannot be considered
identical anymore. After the first bifurcation the wake loses its axisymmetry, but
the particle trajectory is still planar. For higher Reynolds numbers there are the
steady regime with planar symmetry (105 < Re < 137, 5, which we call case
B) and the unsteady periodic regime (Re > 150, which is called case C). The loss
of the planar symmetry characterizes the chaotic regime, which appears for higher
Reynolds numbers (Re > 250) and will be called case D. For spheres lighter than
the fluid another regime, the zig-zag regime, occurs as shown by Jenny, Dusek and
Bouchet [JDB+ 04]. The pictures 1.3 and 1.4 show the different flow regimes.
The result found in the paper mentioned before were tested in a series of experiments, which we will not discuss completely. An overview was given by Ern et al.
[ERFM12]. The first experiment with freely moving spheres focusing was undertaken by Veldhuis et al. [VBVWL05] in 2005. They filmed the motion of spheres
settling or rising in water. By imposing a small temperature gradient, they were
able to visualize the wake with the Schlieren technique. The plastic spheres had a
radius in the range of 0.75 to 10 mm and the density ratio = fp was between 0.35
a
and 2. The resulting Reynolds number Re = umean
was between 102.5 and 2311.5.
Veldhuis et al. confirmed the existence of the different flow regimes (case A-D) that
Jenny et al. predicted. They showed that the wakes of a freely moving sphere in the
different cases displays remarkable differences when compared to the wakes behind
spheres held fixed. Also, the wake for a sphere lighter than the fluid differs from
the wake of a heavier sphere, even for the same Reynolds number.
Summary We discussed the different flow fields for fixed and freely moving spheres.
At low Reynolds numbers the flow field will be the same for both cases, but if the
Reynolds number is higher the symmetry braking mechanism is different. For the
freely moving sphere, we identified four different cases, which can be characterized
by the number of symmetries. In the first case (A) the wake of the sphere is both
axisymmetric and steady. Case B is characterized by a loss of axisymmetry, but the
wake has planar symmetry and is steady. Case C is unsteady, but the wake is still
planar. Case D has neither temporal nor spatial symmetries.
This result affects the interpretation of our experiment in the following way: The
(A)
(B)
Figure 1.3: Picture by Veldhuis et al. Flow field visualized with Schlieren rechnique.
The left part of each picture shows the flow in the xz-plane and the right part shows
the flow in the yz-plane. Left: Flow around a sphere in water at Re = 102.5. The
flow field is steady (case A): Right: Flow around a sphere in water at Re = 162.5.
The axisymmetry is broken (case B).
(C)
(D)
Figure 1.4: Picture by Veldhuis et al. Flow field visualized with Schlieren rechnique.
The left part of each picture shows the flow in the xz-plane and the right part shows
the flow in the yz-plane. Left: Flow around a sphere in water at Re = 225. The
axisymmetry is broken and the flow field is oscillating (case C). Right: Flow around
a sphere in water at Re = 985. The flow field is chaotic (case D).
1 Introduction
force measurement is done typically with a sphere fixed in the flow, whereas the
velocity measurement is done on freely moving spheres. Hence, the result might
differ at high Reynolds numbers.
We start with the force measurements at low and high Reynolds numbers. After
this section the velocity measurements will be discussed.
F
aAf
aup
Rev =
2f a2
a2
S=
=
9
9
F =
(1.2)
(1.3)
(1.4)
We reduced the five dimensional parameter space (F, f, A, a, ) to a three dimensional parameter space (F , Rev , S). Now we have to answer the following question:
how does the normalized force F depend on Rev and S?
The first part of this section focuses on the low Reynolds number regime and we
discuss the different contributions to the force on the sphere in the frequency and
time domain. Then we turn to the high Reynolds number case and examine the
influence of this control parameter.
F = < 6U a(
1
|{z}
Stokes drag
|{z}
Basset History
1 2
)eit
|{z}
added mass
(1.5)
1 Introduction
10
Stokes regime
Basset regime
10
10
10
Stokes
History force
Added mass
2
10 1
10
10
S = a2 /9 (nondim)
10
Figure 1.5: The force on an oscillating sphere in dependence of the time. All parameters set to unity.
1
0.8
0.6
0.4
0.2
0
0.2
0.4
0.6
Stokes
History force
Added mass
0.8
1
0
5
6
time (arb. units)
10
Figure 1.6: Scaling of the forces on an oscillating sphere as function of the dimen2
sionless frequency S = a9 . All parameters set to unity.
10
11
1 Introduction
from the fact that additional vorticity created at the solid particle surface due to
relative acceleration diffuses into the flow and creates a temporary perturbation in
the flow field. Like all diffusive effects it produces an 1/2 term in the equation for
oscillatory motion. In the Maxey Riley model, the convective terms are neglected
and the history force is modelled by
6a ()
2
1/2
Z t
0
d(u)
d
(t )1/2
d.
(1.6)
In this notation we used the velocity u in the comoving frame (u = up v). The
correct form of the History force kernel is discussed in several publications and many
different forms of the equation of motion have been proposed. The discussion on
the correct form of the History force started with the empirical formula by Odar
and Hamilton [OH64], who proposed different pre-factors for the history force. Note
that the Kernel of (t ) leads to a scaling of in the frequency range. A further
discussion will be given later in the theory chapter. Mei and Klausner [MLA91]
observed in a numerical study that the history force decays much faster than t1/2 .
In subsequent studies Mei and Adrian [MA92] solved the problem with the matched
asymptotic method and a t2 decay was proposed.
Lovalenti and Brady [LB93b] and [LB93a] used the reciprocal theorem to derive
an expression for the hydrodynamic force on a particle in time-depended motion.
They also found a t2 decay for suddenly accelerated particles from rest, but a
exponential decay for the change from velocity u1 to u2 with u2 > u1 > 0. According
to Mei [Mei94], this behaviour was caused by the linearisation of the Navier-Stokes
equations in the analysis by Lovalenti and Brady.
Kim, Elghobashi and Sirignano [KES98] corrected the Kernel for very heavy par
ticles (particle-to-fluid density ratio = fp < 0.2), but the temporal decay is also
t2 . The equation by Kim et al. is much more complex than the model by Mei.
Recently Loth and Dorgan [LD09] published another proposal for the equation of
motion.
In summary, the different history Kernels are given by:
Fhistory = 6f a
12
Z t
K(t )
du
d
d
(1.7)
KBasset
#1/2
"
KM ei
16(t )
=
a2
(1.8)
#1/2c1
4(t )2
+
f H a2
"
Rev
2
3 #1/c1 c1
and
(1.9)
(1.10)
1
fh = (0.75 + c2 Rev )3
2
(1.11)
Loth and Dorgan used the same formula as Mei, but with different constants:
c1,M ei = 2
c2,M ei = 0.0105
c1,Loth = 2.5
c2,Loth = 0.2
Both models result in an faster decay of the history force for long times, but the
decay remains algebraically.
Note that the term long time behaviour of the Kernel is misleading. Of course, the
correction terms will dominate for long times, but they also scale with the Reynolds
number. If the Reynolds number is very small, the long time behaviour remains
t1/2 and therefore the correction terms should be better denoted Reynolds number corrections.
Disturbance of the flow In the literature, the flow disturbance term denotes the
change of a non-uniform flow due to the presence of a particle inside the fluid.
Starting from the sphere settling under gravity in unsteady, but uniform flow, Tchen
[Tch47] derived an equation for non-uniform flow. He replaced the particle velocity
up by (up v) and observed that the presence of a particle will result in a fluid
acceleration. This fluid acceleration requires a pressure gradient throughout the
fluid. In his equation the term for the disturbance of the flow field
Fdisturbance,T chen = mf luid
v
v
Dv
= mf luid
+ mf luid v
Dt
t
x
was added to the equation of motion. The term v denotes the undisturbed flow field.
Corrsin and Lumley [CL56] and Buevich [Bue66] pointed out inconsistencies in the
work by Tchen. They criticized that Tchen used only the static pressure gradient
13
1 Introduction
and not the full Navier Stokes equations. The corrected term proposed by them is
Fdisturbance,CL = mf luid
Dv
2 v
Dt
and
Fdisturbance,Buevich = mf luid
d
v.
dt
d
denotes a time derivative following the particle. In 1983, Maxey
The derivative dt
and Riley [MR83] published their form of the equation of motion, which is now
generally accepted as the best approximation. Gatignol derived the same formula
independently [Gat83]. They concluded that the flow field disturbance is given by
Dv
= mf luid .
Dt X(t)
Fdisturbance,M R
Thereby the total derivative has to be taken at the particle position X(t). This is
the same result, which Tchen derived by an ad hoc extension of the equation of
motion.
14
d
F = cD a2 |u|u + cA a3 a + cH a2 ()1/2
2
3
0
t
(1.12)
The correction factors for added mass cA , Basset History cH and Stokes drag cD all
depend on the Reynolds number. The proposed equation was questioned later by
Mei and Adrian [MA92]. The errors in the measurement of the added mass term
have most likely compensated the errors in the history force measurement (and vice
versa). With the assumption that the added mass term is constant, Michaelides
proposed a revised formula [MR10].
Karanfilian and Kotas [KK78] performed an experiment similar to that by Odar
and Hamilton. They also fixed a sphere to a rod and measured the force on the
sphere. In their experiment the oscillation direction was vertically and the sphere
velocity was higher. The amplitude of the oscillation was between 10 and 60 mm and
the angular frequency was between 0.566 and 2.262 rad/sec. The experiments were
15
1 Introduction
done in either water ( = 0.918106 m2 /s) or diesel oil ( = 3.68106 m2 /s) and the
was in
sphere radius was 19.02 mm. The resulting Reynolds number Rev = aup,max
a2
the range 50 to 5000. The dimensionless frequency S = 9 was not explicitly given
in the paper, but with the parameters above the available range was 20 < S < 400.
They measured the (total) drag force on the sphere for different velocities and
accelerations. Instead of the dimensionless frequency S, they used the acceleration
du
number AcOH = 2a
and the Reynolds number to characterize the flow. They
u dt
concluded that the drag coefficient is not a function of the Reynolds number alone.
It also depended on the acceleration number and they gave an empirical formula for
the dependence. The problem with this approach is that the acceleration number
is not constant over one cycle, as pointed out by Tsuji, Kato and Tanaka [TKT91].
The definition of the acceleration number leads to a problem for the oscillating
sphere, since the maximum value goes to infinity during one cycle. Unfortunately,
the parameters at the different data points were not reported in their data, so that
we were not able to check the dependence on the frequency and amplitude variation.
Wind tunnel experiments Now we move on to the wind tunnel experiments with
steady flow conditions. The sphere is fixed in the measurement section of a windtunnel and the forces are measured in dependence on the Reynolds number. As
shown before, the Stokes drag will dominate in the low frequency regime. At higher
flow velocities, the flow field itself will become unsteady and it is not possible to
isolate the different contributions. The result is usually given in terms of the total
drag force
Fdrag cD v 2
with the drag coefficient cD , which is defined as follows
cD =
2Fdrag
(total) drag force
= 2 2.
dynamic pressure projected area
v a
The results are summarized in the following plot 1.7 from H. Schlichtings book
Boundary Layer Theory [SG00]. In contrast to the rest of this thesis, the Reynolds
number in this plot is based on the diameter of the sphere. The plot shows that for
24
small Re the value of cD is proportional to Re
. For intermediate Reynolds number,
the cD coefficient is constant and at high Reynolds numbers a sudden decrease of
cD is apparent. A further discussion would lead to far and we conclude that relation
between force and velocity is very complex. The force on a sphere in unsteady flow
16
Figure 1.7: Dependence of the drag coefficient of spheres on the Reynolds number.
Curve 1 and 2 represent the prediction by Stokes and Oseen and curve 3 is the
numerical result by B. Fornberg (1988).
conditions was also measured by Tsuji, Kato and Tanaka [TKT91]. They fixed the
sphere in the windtunnel and blocked the flow periodically behind the sphere to
achieve an unsteady flow field. They measured in the range of 4000 < Rev = au
<
8000 with pulsation frequencies from 2.5 to 8 Hertz. The interpretation of their
results is rather complicated. They used the ansatz by Odar and Hamilton (see
[OH64]) and tried to fit the data to get correction factors for the Stokes drag cS ,
Basset history cH and added mass force cA . Under the assumption that the history
force can be neglected, the total drag coefficient cD is given by
4
cD = cS + (1 + cA )AcOH .
3
(1.13)
AcOH is the acceleration factor Odar and Hamilton [OH64] and given by AcOH =
2a du
( u being the velocity of the sphere relative to the fluid). The equation 1.13
u2 dt
indicates a linear relation between the difference cD cS and the acceleration factor,
if the added mass correction factor cA is constant. For low frequencies this statement
is true. Nevertheless for high frequencies, the data deviated from the linear relation,
which the authors explain by the effect of the history force. They conclude that
their results agree with the experimental results by Odar and Hamilton.
17
1 Introduction
1.3.4 Summary
We discussed force measurements for different Reynolds numbers both in the time
and frequency space. At low Reynolds numbers, the force on the sphere is the sum
of the different forces (Stokes drag, Basset history and added mass force), which
scale different with the frequency. Therefore they can be separated in the frequency
domain for the case of sinusoidal motion. The measurements at higher Reynolds
numbers indicate that the low Reynolds number solution needs correction terms.
The form of the corrections and their justification is unclear.
f
p
au
Rev =
a2
S=
T
=
(1.14)
(1.15)
(1.16)
This choice is not very good, because the velocity is the variable of the problem and
should not be connected to the control parameter. Instead we define a Reynolds
number based on the force F on the particle:
viscous time scale
a2 /
=
force time scale
a/ F a
a Fa
=
Ref =
18
(1.17)
(1.18)
19
1 Introduction
Figure 1.8: The relative strength of the hydrodynamic forces for a settling sphere.
Taken from [SOC07], page 136. The particle Reynolds number was Rep = 0.5.
Sobral, Oliveira and Cunha use to define to density ratio, but the definition is
the same: = f /p . The virtual mass corresponds to the added mass term in
our notation. The time was made non-dimensional with the relaxation time scale
2
r = 92 a fp .
will reach the steady state faster. Due to the high computation costs of the Basset
force, only a qualitative analysis was possible. They conclude that the classical
Basset force is correct for higher Reynolds numbers, if the dimensionless frequency
S is high. The conclusion can be questioned, because Sobral et al. simulated only
the effects of the known forces and included the non-linear corrections for the drag
force. The inclusion of this correction factor is for sure not enough to capture the
non-linear effects at different Reynolds and dimensionless frequencies.
In the case of sinusoidal forcing and therewith oscillatory motion of the sphere,
the first analytical result was given by Coimbra et al. in 2004 [CLL+ 04]. The
authors derived a general solution for the equation of motion (see Coimbra and
Rangel [CR98] or Coimbra and Kobayashi [CK02]) and solved it with a complex wave
ansatz in the laboratory frame. They gave the solution in terms of the dimensionless
2
frequency S = a9 and the density ratio = fp . The velocity of the fluid is given
by Aei S t and the particle responds with a velocity Bei( S t+) . Then the relation
20
e =1+
A
iS( + 2) + + 3 Sei/4
(1.19)
The equation 1.19 contains both the amplitude ratio and the phase difference. The
result is shown in the plots 1.9 and 1.10, which give the dependence on the frequency
S and the density ratio .
2.5
=5
=2
= 0.5
= 0.2
1.5
0.5
0 6
10
10
10
10
S (nondim)
10
10
10
Figure 1.9: The amplitude ratio of fluid and particle velocities as a function of the
21
1 Introduction
=5
=2
= 0.5
= 0.2
[rad]
/8
/8
10
10
10
10
S (nondim)
10
10
10
Figure 1.10: The phase difference between fluid and particle velocity as a function
The parameter space is then defined by the two parameters density ratio = fp and
2
dimensionless frequency S = a9 . Since the forcing on the particle was not reported
in every experiment, we chose the Reynolds number based on the velocity to compare
the experiments. All values have been converted into our notation Re = au
. Figure
1.11 summarizes the parameter space of the different experiments in the literature.
22
10
10
Abbad b
Weinstein
LEsperance
0
10
Abbad a
Coimbra
10
Gonzales
4
10 2
10
10
10
10
23
1 Introduction
While Coimbra et al. performed horizontal shaking of the fluid cell, Addad and
Souhar designed an experiment with vertical shaking direction. The whole experiment was assembled on a plate and sinusoidally moved up and down by a motor.
Again the motion of the particles was recorded with a high speed camera, but this
time via shadow imaging. The maximal vertical vibration frequency was 10 Hz and
the amplitude was fixed to A = 5 mm. They published two papers on the same
apparatus [AS04a] and [AS04b]. The first study covered the motion of rigid spheres
with different densities and radii ranging from 1 to 2.5 mm. By using silicon oil and
glycerine as fluid and teflon and polyamide spheres, they could measure with density
ratios = f /p from 0.43 to 1.11. The viscosities were oil = 2 104 m2 /s and
glycerine = 6.9 104 m2 /s. Using the definition of the Reynolds number given above,
the Reynolds number in this experiment was less than 1.25. The second experiment
was done with both light bubbles and heavy droplets and the Reynolds number in
this experiment was smaller (Re < 0.25). They report four different experiments.
Two experiments were done with air bubbles rising in silicon oil ( = 5 104 m2 /s)
and glycerin ( = 7 104 m2 /s). The other two experiment included settling glycerine droplets in different silicon oils ( = 5 and 2 104 m2 /s). The density ratio
ranged from 0.77 to 974. Unfortunately, Abbad and Souhar presented their data in
a different way, so the comparison with the data by Coimbra is not possible directly.
They plotted the data in dependence of the penetration depth of vorticity , which
is related to the dimensionless frequency by S = 92 12 . The velocity was normalized
1
by the Stokes terminal velocity uts = 2ga2 ( fp 1) 9
. The graphical extraction
of their data leads to graph 1.12. They found the same result as Coimbra et al,
namely that the History forces cannot be neglected for small Reynolds numbers and
in this range, the form of the equation needs no corrections. The deviation in the
graph 1.12 result almost completely from the graphical extraction, though Abbad
and Souhar estimated their error less than 5 percent.
Another experiment of the same kind was done by Weinstein, Kassoy, and Bell
[WKB08]. They focused on the effects of the history forces on the accuracy of Coriolis flow meters. Therefore they measured the moving of bubble or solid spheres in
a fluid column, which was moved by a shaker table horizontally. They performed
all experiments with an amplitude of 0.35 mm and a frequency of 20 Hz. Additionally all experiments were done with the same kind of oil, leading to a very small
density variation. To achieve a wide range of different Reynolds and dimensionless
frequencies S, different aluminium, steel and POM spheres (0.8 to 4.8 mm) and
24
0.35
0.3
0.2
A /A
0.25
Abbad 1
Abbad 2
with History force
without History force
0.15
0.1
0.05
0
0
0.2
0.4
0.6
0.8
25
1 Introduction
data shows a discrepancy from the theoretical prediction at intermediate frequencies. This error can be explained by the effect of the Basset history force. Also the
data for high frequencies are below the prediction, which is most likely caused by
the added mass term. Nevertheless at low frequencies the Stokes drag dominates
the other forces and the agreement between their data and the linear model is very
well.
Garbin et al. [GDO+ 09] were able to measure the history force on bubbles at
frequencies in the MHz range. They propelled a pair of air bubbles (diameter
2m) with ultrasound at f = 2.25 MHz. Then they recorded the motion of the
microbubbles with the ultra-high speed camera Brandaris 128 at frame rates near
15 106 frames/second. The Reynolds number was below 5. Their model included a
system of two coupled equations of motion. The oscillation of one bubble generates
a fluid field and the other bubble will be accelerated in this field. From the motion
of one bubble the forces on the other bubble could be derived. They compared their
results with the prediction for the history force for bubbles proposed by Magnaudet
and Legendre [ML98] and found that the History force has a crucial influence on
the dynamics. They concluded that the disregard of the Basset History force will
lead to an underestimation of the total viscous dissipation in the case of a bubble
moving in uniform unsteady flow. The bubbles have to be separated far enough, so
that the assumption of a uniform flow holds.
The experiments is the frequency domain all verified the equation of motion including the Basset History force. In the next section, we look at the experiments in
the time domain.
1.4.2.2 Experiments with settling spheres
As mentioned, the solution by Coimbra was only obtained in the domain of the
dimensionless frequency S and not in the time domain. Coimbra and Rangel [CR98]
also gave a general solution, but only in term of complex functions. When the
forcing is given by a step function, the solution can be formulated in terms of real
functions, as shown by Vodopyanov, Petrov and Shunderyuk [VPS10], who obtained
the equation for the settling of a sphere in the time domain. They checked their
solution in an experiment, in which they recorded the settling of steel spheres in
silicone oil. The diameter of the spheres was between 0.237 and 0.396 mm, whereas
the viscosity of the silicon was between 6.7 and 8.9 104 m2 /s. Therewith the
Reynolds number based on the terminal velocity and radius ranged from 0.0753
26
f
and the Reynolds number Re = a F a . The parameter space (, Re)
, S = a
p
9
was explored by different experiments. We already introduced the experiment by
Veldhuis et al. [VBVWL05]. In the literature often a combination of the density
ratio and the Reynolds number is used to describe the problem. This new parameter
is called the Galileo number and is given by:
Ga =
a2 /
= q p
=
a/ |( f 1)|F a
(1.20)
|( f 1)|F a3
q
p
(1.21)
For the case of the settling sphere, the force F is given by gravitational acceleration
g and the parameter space is characterized by different flow regimes (see first part of
the Introduction 1.2.2). The experimental studies cover the whole parameter space,
as shown in graph 1.13. Unfortunately, Veldhuis et al. did not measure the velocity
of the sphere, so we can not use their experiment for our discussion.
27
1 Introduction
|( p 1)|F a3
density ratio define the parameter space for the motion of spheres under gravity.
The different flow regimes are marked in the phase diagram. Picture taken from
[VBVWL05].
f
We need experimental data, which were taken for defined forces and known flow
fields. If the flow field is unknown, we can not specify the fictitious forces due to the
acceleration of the fluid and the total force acting on the particle is unclear. There
are two types of experiments that fits our restrictions. The first type of experiment
was already introduced. The measurement of a settling sphere in still fluid provides
data for the velocity at a known force (gravity) and velocity field (still fluid). The
other type of experiment is drag measurement in shock tubes. Due to the shock
wave, the velocity of the fluid is increased and this acceleration of the fluid leads
to an fictitious force on the particle. This experiment can also be used, hence the
flow field is known from the measurement of the shock wave and the force can be
computed from the acceleration.
This section focuses on the experiments with high Reynolds number flow, in which
the velocity was measured. We start with the experiments with steady forces and
go on to the unsteady case.
Settling spheres An experiment at higher Reynolds numbers was performed by
Mordant and Pinton [MP00]. The settling motion of different spheres was measured
by the Doppler shift of acoustic waves. Using glass, steel and tungsten spheres with
radii ranging from 0.25 to 3 mm and water as fluid, they achieved Reynolds numbers
28
Thereafter, they solved the Maxey Riley equation with the Basset force for the
History force numerically and compared the result with their measurements. For
short times, the settling motion was described correctly, while the memory term
decreased faster than t1/2 for long times. They proposed an experimental fit for
the decay. Their conclusion should be taken with caution, because the decay of
the Basset History force is not measured at constant Reynolds number. A possible
decay of the Basset term couldqresult from the increase of the Reynolds number.
|( p 1)|F a3
from 15.46 to 1911.86. The advantage the Galileo number is twofold. This number is constant for the whole experiment and it can be computed before. The
Reynolds number based on the velocity increases during the experiment and has to
be computed with an implicit equation (see [MP00]). With the Galileo number, we
can easily identify the different flow regime in comparison with the experiment by
[VBVWL05]. For example, Mordant and Pinton compared the motion of different
(see the graphs in 1.14, which were
particle at the same Reynolds number Rev = au
taken from their paper). The bigger sphere shows transitory oscillations, whereas
the trajectory of the smaller sphere does not have these oscillations. This observation
becomes clear, when the definition Ga is considered. Although both experiments
have the same Reynolds number based on the velocity, the Galileo number, which
is a Reynolds number based on the force, is different. This is why, the trajectories
of the two experiments differ.
f
Shock tube experiments As explained before, we can also use shock tube experiments for our discussion, since the force on the sphere and the flow field are
known. Temkin and Kim [TK80] and Temkin and Metha [TM82] used weak shock
tube experiments to measure the velocity of moving particles at a given force. The
force acting on the particle is given by the fictitious forces due to the acceleration of
the fluid. Temkin and Kim [TK80] measured the drag coefficients for high Reynolds
numbers in a shock tube. The particles were micrometer sized droplets of different
densities and the Reynolds number based on the relative velocity of fluid and particle
was between 1.6 and 38.5. The total drag coefficient for steady flows is known from
other experiments, but the interesting aspect of this experiment was the velocity
change. Due to the shock wave, the velocity of the fluid is raised very fast and the
acceleration of the fluid is big. Since the experiment is done in a short time after
29
1 Introduction
(A)
(B)
curve shows the velocity of a 1 mm tungsten sphere and the solid curve represents
a 2 mm glass sphere. Picture (B): Velocity measurements for Rev = au
630.
The dashed curve shows the velocity of a 1 mm steel sphere and the solid curve
represents a 1.5 mm glass sphere.
the shock wave, the unsteady effects can not be neglected. Temkin and Kim used
the Acceleration number
!
2a dUr
p
1
(1.22)
Ac =
f
Ur2 dt
to characterize the unsteadiness of the flow. The variable Ur is the fluid velocity in
the frame moving with the particle. They found that the drag coefficient cT emkin
depends both on the Reynolds and Acceleration number, but the influence can be
simplified:
cT emkin = cD + f (Ac),
where cD is the steady drag coefficient that captures the dependence on the Reynolds
number and f (Ac) represents the functional dependence on the acceleration number. For their experimental data they found a linear function.
The results by Temkin and Kim were tested in a subsequent study by Temkin
and Metha [TM82]. Again, micrometer sized particle were injected into a shock
tube and the particles were accelerated by weak shock waves. The maximal flow
velocity was 10.1 m/sec and the Reynolds number based on the relative velocity of
fluid and particle was between 4.5 and 57.5. The data were taken for both accelerating and decelerating flows. The acceleration number was defined as in equation
30
fit the experimental data with different models, which did not include the Basset
history and added mass force. Also the unsteady effects due to the change in the
fluid velocity were ignored. The discrepancy between these models and the experiment was quite big and they were not able to explain this result. But especially
the omission of the unsteady effects leads to the big discrepancy, which was also the
case for the experiment by Temkin and Kim [TK80].
An overview of other results from wind tunnel measurement can be found in the
paper by Jourdan et al. [JHI+ 07]. They also report their shock tube experiment
with different plastic spheres (density from 25 to 1550 kg m3 ) and different gases.
The Reynolds number was defined as follows:
ReJourdan =
v
u
2a v up u
t
dup
dt
dup
dt
!2
dup,y
+
g
dt
!2
The velocity up,y is the velocity of the particle in the direction of gravity, which was
perpendicular to the direction of the shock wave. With this definition, the Reynolds
number was in the interval 471 < ReJourdan < 68500. The acceleration number
defined as
2a
dup
Ac =
(v up )2 dt
was between 2.5105 and 3.5102 . They found that the measured drag coefficient
deviates from the steady drag coefficient. They argue that the influence of the
unsteady terms can be neglected due to the small value of the acceleration number
31
1 Introduction
Ac. According to them, this can be seen from the comparison with the experiment
by Karanfilian and Kotas [KK78] (discussion in section 1.3.3). This argument should
be questioned because of two facts: the acceleration number is not constant during
the experiment and for this reason it is not a good choice for the measurement of
the unsteadiness of the flow. Also the data from the experiment by Karanfilian and
Kotas are not very precise due to flow disturbance caused by the rod and the walls.
Unfortunately, the data given in the references are not enough to define the force
variation consistently. In contrast to the settling sphere experiments, the force in
the shock tube experiments will decrease linearly with time.
1.4.4 Summary
In this section we discussed the velocity measurement at a given force on the particles. We started with the solution of the equation for low Reynolds numbers and
went on to the experiments. We saw that the Reynolds number should be defined
based on the forces instead of the velocity in the case of force measurements. The
equation of motion with the Basset History force was verified in a big number of
different experiments. For the high Reynolds number case the experiments showed
that the equation of motion needs corrections. In the case of the particle settling
under gravity, the symmetry braking mechanism can explain the break down of the
low Reynolds number solution, but the implication for the correction of the equation of motion are unclear. The shock tube experiments also could not be used to
improve the equation of motion, since the parameters to describe the unsteadiness
of the flow have not been reported.
The next section gives a comparison between the previous experiments and the
parameter space of our experiment.
1.5 Motivation
The motion of particles in an unsteady flow is only partially understood so far. We
have identified the three control parameters: the density ratio , the dimensionless
frequency S and the Reynolds number Re based on the velocity (for force measurements) respectively based on the force (for velocity measurements). For the velocity
measurements the Galileo number is also a suitable choice. The problems can be divided into two main questions: what is the correct form of the forces and how do the
32
1.5 Motivation
forces change due to a variation of the control parameters, especially the Reynolds
number? At a very small Reynolds number, the appearance of the Stokes drag and
added mass force are quite well understood and the correct terms are experimentally
verified. The correct form of Basset history force is also experimentally verified, but
the interpretation of the physical background is under discussion.
The following plots summarize the previous experiments and the parameter range
of this experiment. When the Reynolds number is very small, only the density ratio
and the frequency characterize the problem. The plot 1.15 shows the parameters
and S, at which previous experiments have been done and which parameterspace
our experiment covers. There are experiments for a wide range of density ratios,
so we decided to expand the frequency range towards smaller values. This range is
important, because the temporal decay of the Basset History force for long times
(small frequencies) is unclear.
The data base for the Reynolds number and Galileo number dependence is very
small. We discussed some experiments, but the parameter space , S, Re has not
been examined in detail. The plot 1.16 shows our available parameter space. The
experiment of Veldhuis et al. only checked the dependence on and Ga, while S
was zero.
Our experimental approach is comparable with the experiments by Coimbra et
al. and Weinstein et al. We oscillate a fluid column with a certain frequency
and amplitude and measure the response of a sphere to this external forcing. This
oscillation is conducted by a linear motor system, which enables us to explore a wide
parameter range in the most interesting parameter subspace (Re/Ga, S). There are
two goals of the experiment: The range of experimental data at low Galileo numbers
will be extended to lower frequencies. The second goal of the experiment is the
analysis of the Reynolds number dependence. By increasing the amplitude of the
linear motor the dependence of the velocity response on the Reynolds number will
be checked.
33
1 Introduction
10
10
Abbad b
Weinstein
LEsperance
own data
10
Abbad a
Coimbra
10
Gonzales
4
10 3
10
10
10
10
10
S
Figure 1.15: Data sets in the different experiments.
0.25
0.2
0.15
0.1
own data
0.05
0 1
10
Veldhuis
0
10
10
10
Ga = (f2 A d3)1/2/
10
10
34
2 Theory
In this section we solve analytically the equation of motion for the sphere moving
in a fluid. The force on the sphere is specified and an expression for the velocity in
dependence on the force is derived. In the last chapter one expression was already
given for the case of an inertial frame and we solve the equation again in the frame
moving with the fluid. We start with the quasi-steady state motion and both the
fluid and the particle motion are modelled with a complex wave ansatz. Afterwards,
we solve the full equation. A short introduction to the fractional calculus is given,
since this is very useful for the solution. The last part summarizes the results of
this chapter.
2
the velocity gradient we can rewrite the second condition as a 1. The following
equation was obtained for the motion:
mp
dup
Dv
1
d(up v)
= (mp mf )g + mf
6af (up v) mf
dt
Dt
2
dt
6a ()
2
1/2
d(up v)
d
d
0 (t )1/2
Z t
35
2 Theory
We used the short notation for the mass of the particle mp =
4 a3
mf = 3f .
4p a3
3
dv
dt
D
v
Dt
= 0.
The drag, added mass and history force do not change, though up (t)v(t)
has to be replaced by u(t).
As a result, the equation of motion changes to:
Z t
du
du
4a3
1
du
2
1/2
d
mp
= (p f )
v 6af u mf
6a () f
d.
dt
3
2
dt
0 (t )1/2
)
v
(mp + mf )
p
f
2
dt
3
0 (t )1/2
r
du
1
du
9 Zt
9
d
(p + f )
+
f
d + 2 f u = (p f )v
(2.1)
1/2
2
dt
2a
2a
0 (t )
Equation 2.1 was first derived by Basset in 1888 [Bas88] for the case of a sphere
settling under gravity. Therefore, the equation of motion 2.1 will be referred to as
the comoving Basset equation or just Basset equation.
36
Sl
St Rev
1 2
u
+ uu = p +
u + forcing
t
Rev
u
+ Rev uu = p0 + 2 u + forcing
t
The unsteady term is kept, whereas the non-linear convective terms are neglected. Hence, the following restriction has to be satisfied
Sl Rev > Rev
Sl > 1
a
a
=
>1
y
9u
9Ap
a
>1
9Ap
Exemplary, this condition can be found in [CK02].
Now we move to the conditions, under which the solution of the equation is valid:
quasi-steady state:We have to assume that the motion of both the particle
and the fluid can be described by a complex wave ansatz. This is only true
after the transient state. In 2.5.3 a solution including the transient state is
given.
known initial conditions: We assume that the velocity at t = 0 is zero.
Also the velocity for t < 0 has to be zero.
no wall effects: In order to solve this equation, it is assumed that the walls
have no hydrodynamic effect on the sphere.
37
2 Theory
= Ap ei ieit
A detailed derivation can be found in the appendix. The other terms are straight
forward and with v = As e(it) and = fp we get:
9
9
1
(it+)
(it+)
+
iAp e
+
Ap e(it+) = (1 )As ie(it)
(1 + )Ap ie
2
2a
2a
The exponential term can be dropped and the terms are rearranged to get the input
on the left side.
Ap e
1
9
i(1 + ) +
2
2
9
i + 2
2
a
2a
= (1 )As i
which gives
(1 21 )i
Ap i
q
e =
As
i(1 + 12 ) + 92 a2 i +
=
(2.2)
2a2
2iS(1 )
iS(2 + ) + 3 iS +
(2.3)
a2
.
9
the solution for different density ratios = fp and different values for the dimensionless frequency S, which also captures the dependence on the particle diameter
a and the fluid viscosity .
38
history force
(2.4)
|{z}
Stokes drag
For a very low frequency S 1 all terms containing the dimensionless frequency
S have a very small influence, thus the motion is largely determined by the Stokes
drag. For intermediate frequencies all three terms contribute. For very high frequencies S 1 the added mass term will dominate.
This is well in agreement with the results found by Lawrence and Weinbaum for
the motion of spheroids, see [LW88]. Also LEsprance et al. [LCTR05] found a
corresponding result. Another interesting observation is the different behaviour of
the phase difference. For very high frequencies the result will be proportional to i
and therefore the phase difference will be /2. For very low frequencies, the real
part will dominate and the phase difference goes to 0.
Influence of the density ratio From the formula 2.4 the influence of the density
ratio = fp cannot be easily seen, thus it has to be plotted. Figure 2.1 shows the
p
effect on the amplitude ratio A
. The dashed lines show the result without History
As
force, whereas the solid lines show the results including the History force. With
the density ratio deviating from unity, the amplitude increases. Both small and
large result in increased amplitude, whereas the effect for light particle ( > 1)
is greater than for heavy particles ( < 1). For high frequencies the amplitudes are
the same with or without the Basset History force. The effect of the History force is
a shift towards higher frequencies. This means if the History force is included, the
amplitude increases slower with increasing frequencies.
The result for the phase is shown in figure 2.2. Again the result for small and
large frequencies coincide. For small frequencies the particle and fluid are in phase,
whereas the phase difference goes to /2 for high frequencies. We start the discussion
with the curves without Basset History force. The effect of is only a shift of the
frequency, at which the phase difference jumps to /2. The form of the curve
remains unchanged. If the Basset History force is considered, the slope of the curves
is smaller. The phase difference starts to deviate from 0 at smaller frequencies and
39
2 Theory
1.4
1.2
1
=5
=2
= 0.5
= 0.2
0.8
0.6
0.4
0.2
0 6
10
10
10
10
S (nondim)
10
10
10
Figure 2.1: Amplitude ratios for different density ratios. The solid (dashed) lines
show the result including (excluding) the History force.
reaches /2 later. Also the form of the curves for different is changed. The phase
difference increases faster for heavy particles ( < 1) than for light particles ( >).
[rad]
/2
=5
=2
= 0.5
= 0.2
/4
0 6
10
10
10
10
S (nondim)
10
10
10
Figure 2.2: Phase for different density ratios. The solid (dashed) lines show the
result including (excluding) the History force.
40
d
Several mathematician have given definitions for the operator dt
1/2 , but the discussion will be restricted to those definitions that are used in this text. A more detailed
discussion can be found e.g. in the textbooks by S. Das [Das11] or R. Hilfer [He00].
Riemann-Liouville First, the Cauchy formula for the n-th integration (a function
f should be locally integrable on (a, ), n-times) is defined
Z
|
...
Z t
{z a}
f ()d =
Z t
1
(t )(n1) f ()d = I n
(n 1)! a
n times
by introducing the gamma function (n) = (n1)! we can extend the formula above
to > 0:
a It
1 Zt
=
(t )(1) f ()d
() a
(2.5)
41
2 Theory
The notation of the three indexes for the operator is essential. The sub-index before
the operator gives the lower bound, whereas the sub-index after the operator gives
the upper bound. The function has to depend on one of these two sub-indexes. Two
cases can be defined:
a It
1 Zt
(t )(1) f ()d
=
() a
t Ib
right hand
and
1 Zb
( t)(1) f ()d
() t
left hand
The operator I is generally considered as Riemann-Liouville fractional integral. Let us consider a number m = n with 0 < < 1 and n being an integer
number. The fractional derivative for the right hand definition is then given by
m
a Dt f (t)
dn
n
f (t)
a It
n
dx
(2.6)
The left hand definition is straight-forward, however we will not use it. In fact, we
will use only the right hand one-half derivative from zero to t:
1/2
d1
1/2
f (t)
0I
dx1 t
Z t
1
f (1) ()(t )1/2 d
=
(1 1/2) 0
0 Dt f (t) =
df ()
1Zt
d1/2 f (t)
dt
d
=
0 (t )1/2
dt1/2
(2.7)
In the third step we used the permutability of integration and derivation. Also we
d1/2
defined the meaning of the operator dt
1/2 in the last step.
42
Z
0
f (t )g()d}
= fe(s)ge(s)
The convolution in the time domain is replaced by a multiplication in the s-domain.
As stated, the Riemann-Liouville definition is used only. We follow the procedure given by Gorenflo and Mainardi [GM08] to define the Laplace transformation.
1
t 2
and use the convoluTherefore we start with the definition of a function = (
1
)
2
tion theorem to observe
1 (t) f (t) =
2
Z t
0
(t ) 2
f ()d
( 12 )
The integral kernel has to vanish for t 6 0 and we get for the half derivative
1 Zt
1 Zt
1/2
21
(t
)
f
()d
=
(t )1/21 f ()d = 0 It f (t)
( 12 ) 0
( 12 ) 0
In the last step we used the Riemann-Liouville definition given in 2.4.1 and we found
a convolution of two functions, namely and f (t), which give the half-derivative of
f (t). With the Laplace transform1 of t1/2 :
1/2
L{t
}=
1 1/2
s
2
we observe a very useful formula for the Laplace transformation of the fractional
integral:
1/2
43
2 Theory
observe
L{
d 1/2
1/2
f (t)} = L{0 Dt f (t)} = s1/2 fe(s) s1/2 f (0)
0I
dt t
(2.8)
F {f (x)} = f () =
f (x)e(ix ) dx0
2
has the useful property that the n-th derivation (n integer number) of the function
f (t) is transformed into a multiplication in the Fourier space:
(
dn
f (x) = (i)n f()
dxn
)
This statement can be generalized for fractional derivatives, as shown by Das [Das11].
The Fourier transformation of the fractional derivative leads to the following expression:
F () 0 D f (t)
The equation shows that a fractional order derivative leads to an decay in the frequency domain with the order of the fractional derivative. The connection to the
decay of the kernel in the Basset History force is given by the definition of the
fractional derivative.
1/2
0 Dt f (t) =
df ()
d1/2 f (t)
1Zt
dt
d
=
0 (t )1/2
dt1/2
(2.9)
We can conclude that a decay of the Basset History kernel with t1/2 leads to a
decay with 1/2 in the frequency domain.
44
du 9
f
d1/2 u
f
9
+
u=
+
1
dt
2 a p + 2 f dt1/2
2a2 p + 12 f
du0 9
d1/2 u0
f
9
f
+
+ 2
u0 =
1
0
1/2
01/2
dt
2 a p + 2 f
dt
2a p + 12 f
p f
v
p + 12 f
p f 0
F q
p + 12 f
du0
d1/2 u0
+
+ u0 = q 0
dt0
dt01/2
In the last step we introduced the parameter =
primes and perform the Laplace transformation:
p
9
.
2 p + 12 f
(2.10)
Now lets drop the
du
d1/2 u
L
+ 1/2 + u = L {q}
dt
dt
(
e
s ue(s) + s1/2 ue(s) + 1 ue(s) u(0) u(0) s1/21 = q(s)
ue(s) =
e
u(0) (1 + s1/21 )
q(s)
+
s + s1/2 + 1
s + s1/2 + 1
e
The initial condition u(0) is time-independent, whereas the function q(s)
can be
time-dependent. We have to find the following functions (convolution theorem for
Laplace transformation):
Z t
0
q(t ) u ()d
(2.11)
with
(
u0 (t) = L
(1 + s1/21 )
s + s1/2 + 1
and
u (t) = L1
1
s + s1/2 + 1
(2.12)
45
2 Theory
To get the solution for the motion of the sphere, we have to perform the inverse
Laplace transformation for these two functions. Before we do this calculation, we
discuss the physical meaning of these functions. The function u0 (t) describes the
decay of the initial condition u(0). If the initial conditions is zero, the convolution
of the function u (t) and the (non dimensional) forcing q(t) gives the velocity. This
means the function u (t) is the response function for this system. Next, we make
a short excursion to discuss the meaning of the response function in terms of the
dielectric response. Thereafter we go back to our problem and solve the equation
2.11.
dP
d P
+ 2 + P = 0
1
dt
dt
(2.13)
Z t
0
E(t )e ()d.
(2.14)
The symbol 0 defines the electric permittivity of free space. In the Laplace space
the response function is given by (see Hilfer [Hil02]):
e (s) =
46
1
.
1 + (2 s) + 1 s
Summary We saw that the Basset equation 2.10 can be used to model the dielectric
response of glass forming materials. With a small number of fitting parameters, the
behaviour of the material for a wide range of frequencies and temperatures can be
modelled. This result shows the connection between the dielectric susceptibility e
and the response function u defined in 2.12. We conclude that the motion of the
particle is the response to the external forcing. The dependence of the response
function on the control parameters is our main interest.
(2.15)
(2.16)
47
2 Theory
Figure 2.5 shows the result for the integration with = 1. The response function
1
u (nondim.)
0.8
0.6
0.4
0.2
0
0
4
6
time (nondim.)
10
Z t
0
Z t
0
q 0 (t ) u0 ()d
q 0 (t ) u0 ()d
The response function has no units and the unit of the normalized force is recovered
with the formula q 0 (t0 ) = g(t0 ). The following result is obtained:
1
u=
g (t ) u0 ()d
0
p f
Z t
F 1
p + 12 f
=
g (t ) u0 ()d
0
Z t
p f
1
F
g
(t
)
u0 ()d
=
1
0 p + 2 f
Z t
(2.17)
To perform the integration we have to specify the forcing on the particle. The next
part shows the result for a sinusoidal forcing with the amplitude A.
Example: Sinusoidal forcing We specify the time dependent part of the forcing
as g(t) = sin(t) and the normalized force amplitude as F = A 2 . This ansatz
48
Z t
0
!
Z
0 12
x
sin( 21 )
1
p f
0t
2
x
A
sin
(t
)
e
dx0 d
1
0 21
0
2
0
2
0
0
p + 21 f
(1 x ) + x + 2(1 x )x cos( 2 )
(2.18)
q
1.10
= 1.768, A = 1 m, = 1 seconds
Figure 2.6 shows the result for = 92 1.10+0.50.9659
and = 2. The graph shows an oscillation, which is superposed with a small drift
towards smaller values. The drift is caused by the fractional term and decreases
fast.
0.3
vel [m/s]
0.2
0.1
0.1
0.2
0
10
time [sec]
Figure 2.6: Result for the transient state
Comparison with quasi-steady state Since we already solved the equation for the
quasi-steady state, we can compare this result with the result from the full solution.
49
2 Theory
First, we rewrite the full solution in terms of and S. We have
9
2
2
9
( + 12 ) 2
S= 2
9
=
2
and get = 0.9659
= 0.878 and S = 11.768
= 2.183. These parameters lead to an
1.1
9
Ap
amplitude ratio of As = 0.0547 and a phase shift of = 0.8784 rad. Figure 2.7
shows the full solution and the quasi-steady solution. The amplitude goes almost
to the value of the quasi-steady within the first oscillation. The amplitude of the
full solution is slightly smaller, which might be caused by round-off errors in the
numerical integration. Nevertheless the amplitude stays constant in both cases.
The phase difference shows also shows the a different behaviour. Due to the initial
conditions the phase differs at the beginning. After a few oscillations the phase of
the two solutions coincide.
0.3
Full solution
Quasi steady
vel [m/s]
0.2
0.1
0.1
0.2
0
time [sec]
50
10
2.5.4 Summary
We solved the equation of motion for the quasi-steady state and the transient case.
The quasi-steady solution results in a harmonic motion of the particle and the
amplitude and phase of this motion depend on the density and the dimensionless
frequency S. The full solution is a rather complicated sum of a part depending on
the initial conditions and a part, which can be identified as response to the forcing.
The resulting trajectory is a harmonic motion with additional shift at the beginning.
We found that the amplitude of the two solutions reaches the steady value within
the first oscillation. The values for the amplitude slightly differ for the two solutions.
The phase difference of the two solutions also differs at the beginning, but the phase
of the full solution and the quasi-steady solution coincide after a few oscillations.
51
3 setup - censored
53
4 Results
The section is organized as follows: The first section covers the classical problem
of an oscillating sphere in quasi-steady state. We discuss our results and compare
them to the previous work by Lesperance [LCTR05], Abbad [AS04a] and Weinstein
[WKB08]. The second section focuses on the influence of the Reynolds number Ref
and the dimensionless frequency S. We will discuss in which parameter range the
results of the quasi-steady state solution are valid. Thereby all the discussion is in
the comoving frame.
55
4 Results
4.2 shows the position versus time plot. Again, the results differ at the beginning,
but the difference between the two results become constant after a few oscillations.
We can conclude that the sphere reaches a quasi-steady oscillation within the first
oscillation. If the shift in the position is removed, the quasi-steady and full solution
agree within the accuracy of the experimental data. We will therefore use the
quasi-steady solution to fit the data and discuss the error due to this assumption
afterwards.
Full solution
Quasi steady
x 10
1.5
1
vel [m/s]
0.5
0
0.5
1
1.5
2
0
time [sec]
10
Figure 4.1: Particle velocity from the full and quasi-steady solution for S = 0.0279.
Parameters: As = 5mm a = 2 mm, = 100 106 m2 sec1 and = 2 rad sec1 .
4.1.2 Assumptions
The steady state solution given in Chapter 2 is only valid, if the assumptions (steady
state, known initial condition and no wall effects) and constraints on dimensionless
parameters (Sl > 1 and Rep < 1) are fulfilled. Before we look at the results, we
have to check, if these assumptions are violated.
The steady state assumption is fulfilled, since we cut off at least one oscillation
before fitting the data. This is in good agreement with the full solution, which
showed that the sphere reaches the quasi-steady state within the first oscillation
56
biggest sphere and 50 for the smallest sphere. Since the non-inertial terms will scale
as 2 = Rep (see Veysey [VIG07]), this number should be small. The value of 2
is in the interval 4.40 105 to 6.46 103 for the experimental data shown in this
section. Therefore the walls will have a very small effect on the oscillating motion.
The conditions on the Reynolds and Strouhal number can be checked easily from the
Full solution
Quasi steady
x 10
ampl [m]
5
4
3
2
1
0
0
time [sec]
10
Figure 4.2: Particle position from the full and quasi-steady solution for S = 0.0279.
Parameters: As = 5mm a = 2 mm, = 100 106 m2 sec1 and = rad sec1 .
57
4 Results
data. The Reynolds number should be much smaller than unity and the Strouhal
number should be greater than unity. The particle Reynolds number takes values
= 9Aa p
between 2.9 104 and 4.3 102 . The Strouhal number defined as Sl = a
9u
is between 1.1 to 7.6. Both conditions are fulfilled.
58
measured
fit
0.2
0.15
0.1
0.1
displacement (mm)
displacement (mm)
0.15
0.05
0
0.05
0.1
0
0.05
0.1
0.15
0.15
0.2
4
0.05
7
t (s)
10
0.2
8
measured
fit
8.5
(a)
9.5
t (s)
10
10.5
11
11.5
(b)
measured
fit
0.15
0.05
0
0.05
0.1
a: run 1, f = 1 Hz
0.15
0.2
9
b: run 2, f = 1.05 Hz
10
11
t (s)
12
13
14
c: run 3, f = 1.05 Hz
The other parameters are the same.
59
(c)
displacement (mm)
0.1
4 Results
4.1.4 Results
The result for the quasi-steady state experiment is shown in figures 6.1 and 6.2 for
the different spheres and sled frequencies. The amplitude of the sled motion was
different, but always so small that the conditions for the solution were fulfilled.
The results for the amplitude are very close to the predicted curve with the Basset
History force. We can confirm that the Basset History force is needed to describe
the behaviour of the particles correctly. Especially for the high values of S the
importance of the Basset History force is evident. The data for the phase scatter
much more than the data for the amplitude. We already explained the difficulties
in the measurement of the phase. Nevertheless the data support the importance of
the Basset History force.
From the full solution we learned that the quasi-steady solution gives a good
prediction for the phase, whereas the amplitude in the full solution is a bit smaller.
The error due to the quasi-steady assumption is much smaller than the variation of
the data points. We can conclude that quasi-steady solution describes the results of
the experiment correctly.
0.04
0.035
Ap/As (nondim)
0.03
0.025
3 mm
4 mm
5 mm
with Basset History force
without Baset History force
0.02
0.015
0.01
0.005
0
10
10
60
/4
[rad]
/8
3 mm
4 mm
5 mm
with Basset History force
without Baset History force
0.02
0.04
0.06
0.08
S
0.1
0.12
0.14
0.16
61
4 Results
0.04
Ap/As (nondim)
0.03
0.02
0.01
0.01
0.02
0.04
0.06
0.08
S
0.1
0.12
0.14
0.16
Figure 4.6: Result for the amplitude ratio with error bars.
/4
[rad]
/8
0
with Basset History force
without Baset History force
data
0.02
0.04
0.06
0.08
S
0.1
0.12
0.14
0.16
Figure 4.7: Result for the phase difference with errors bars.
62
radius [mm]
0.5, 0.75
0.4, 1.0, 1.6
1.6, 2.4
25
25
1.98
1.98
1,5
2
2,5
[m2 sec1 ]
(0.84 0.95) 104
(0.78 0.88) 104
(0.83 0.84) 104
6.9 104
2.0 104
0.4 104
0.4 104
1.0 104
1.0 104
1.0 104
f [Hz]
20
20
20
0 10
0 10
20 70
20 70
0.14 5
0.2 5
0.2 5
0.112
0.3070
0.606
0.5714
0.8517
0.56
0.214
0.8473
0.8473
0.8473
Rev
0 0.5
0 0.5
0 0.5
0 1.25
0 1.25
05
05
4
4.1 10 6.2 102
2.5 103 1.7 102
2.7 103 1.4 101
63
4 Results
0.9
0.8
Weinstein 1
Weinstein 2
Weinstein 3
Abbad 1
Abbad 2
LEsperance 1
LEsperance 2
Own data 3mm
Own data 4mm
Own data 5mm
with History force
without History force
0.7
Ap/As
0.6
0.5
0.4
0.3
0.2
0.1
0
0
3
S
64
10
Ap/As
10
10
10
10 3
10
10
10
S
10
10
Figure 4.9: Comparison with the literature data (logarithmic). Same labels as in 4.8
The correct equation of motion is also important for experiments and applications.
The equation might be used to calculate the motion of small particles in a low
Reynolds number flow, for example small pollen at low wind speeds or sand particles
in water. Also the motion of tracer particles in experiments is an important question.
If the Reynolds number based on the particle radius and relative velocity is small
enough, this equation can be used to calculate the relative motion of the particle
and the fluid. The effect of a higher Reynolds number will be discussed in the next
section.
Now, we increase the Reynolds number and look at the full parameter space , S and
Ref . Before we go into the details of the results, we look again at the dimensionless
Navier Stokes equations, plug in the force F = As 2 sin(t) and use the time scale
65
4 Results
T = 9 , the velocity scale U = Ap and the length scale L = a to get:
U u0 U 2 0
U 02 0
0 0
u
u
=
u + As 2 sin(t)
+
T t0
L
L
u0 Ref 0
1
9As
u 0 u0 = 02 u0 +
sin(t)
+
0
t
S
S
Ap
u0
1
1
1
sin(t)
+ u0 0 u0 = 02 u0 +
0
t
Sl
S
Fn
Ap
, which we
In the last step we defined the new dimensionless parameter F n = 9A
s
call the force number. In the introduction we related every force to the viscous
forces, whereas now we use the unsteady term as criterion. The Strouhal number
Sl = ReS f = 9Aa p was already introduced before. It is the ratio of the unsteady terms
and the convective terms. For values of Sl 1, the convective terms dominate
the unsteady terms. The dimensionless frequency S was also discussed before. For
S 1 the viscous terms are very important in comparison with the unsteady terms.
If the frequency S is constant, we can replace Sl by Ref .
Ap
The force number F n = 9A
can be seen as the ratio of two Reynolds numbers
s
based on the sled and particle amplitude:
aAp /
Ref
Ap
Ap a/
=
=
=
9As
9As a/
9aAs /
Res
inertial f./viscous f.
=
external f./viscous f.
Fn =
We can say that the force number relates the external force to the inertial force.
For F n 1 the external forces are much bigger than the inertial forces. Another
related parameter is the force time ratio f , which is defined as follows:
external force time
L/U
=
unsteady force time
T
a/As
a
=
=
1/
As
f =
The force time ratio relates the time scales of the unsteady term with the time
scale of the external forcing. For f 1 the external forces can be neglected in
comparison with the unsteady terms.
Now we face the problem that we have to compare the relative importance of four
terms (unsteady, convective, viscous and external forcing). We approach the topic
66
1.6
data S = 0.0112
data S = 0.0873
theory
1.4
1.2
A /A (normalized, nondim)
1.8
1
0.8
0.05
0.1
Ref
0.15
0.2
0.25
Figure 4.10: Normalized amplitude ratio in dependence of the Reynolds number Ref .
The amplitude ratio was normalized with the theoretical value (Ap /As )0 .
We measured the ratio of the particle and sled motion and calculated the the-
67
4 Results
oretical values with the linear theory presented in the last chapter. If the linear
p
normalized with the
theory (derived for Ref 1) is correct, the amplitude ratio A
As
Ap
theoretical value ( As )0 should be unity.
Figure 4.10 shows the normalized amplitude ratio in dependence on the Reynolds
number Ref for two different values of S. The data points labelled with S = 0.0112
are measured with 4 mm spheres and vibration frequencies of 0.4 Hz. The oscillation
amplitudes increased from 25 mm to 150 mm. The second data set was taken at a
frequency S = 0.0873. These data points were taken with a 5 mm sphere at f = 2
Hz and the amplitude increased from 5 mm to 30 mm. The data shows that the
linear solution is okay for small values of Ref , but if the Reynolds number increases,
the experimental data and the prediction deviate. For Re 0.2, the deviation is
Re
1
= S f was
more than 60% of the expected value. The inverse Strouhal number Sl
between 0.877 and 1.707.
0.04
0.035
Ap/As ( nondim)
0.03
0.025
0.02
0.015
0.01
0.005
0
0.005
0.01
0
0.02
0.04
0.06
0.08
0.1
Figure 4.11: Normalized amplitude ratio in dependence of the dimensionless frequency S. The data points have been shifted for better visibility. The data points
in each cluster have the same value for S.
For better comparison with the other experiments, we plot the errors in the S
vs amplitude ratio diagram (see Figure 4.11). The error bars were calculated with
the same assumptions as before (see section 4.1.4). Note that the data were shifted
for better visibility. Both data sets show that the error for the amplitude is almost
68
Ap/As (nondim)
0.1
0.08
3 mm
4 mm
5 mm
with Basset History force
without Basset History force
0.06
0.44
0.4
0.38
0.22
0.23
0.17
0.12
0.04
0.06
0.02
0.1
0.08
0.03
0.006
0.005
0.007
0.01
0.017
0.028
0.04
0.018
0.01
10
S
69
4 Results
0.14
0.12
Ap/As (nondim)
0.1
0.08
0.06
0.04
0.02
0
0.02
0
0.05
0.1
0.15
0.2
0.25
Figure 4.13: Same data as 4.12, but linear plot and added error bars.
To collapse the data, we normalize the amplitude ratio with the predicted value
(Ap /As )0 from the last chapter and plot this versus the inverse Strouhal number
Sl1 (see figure 4.14). The normalized amplitude ratio gives the deviation from
the (linear) theory and figure 4.14 shows that this deviation is proportional to the
inverse Strouhal number. Since the Strouhal number is the ratio of the convective
to the unsteady terms, we can conclude that the data deviation is proportional to
the strength of the convective terms (in comparison with the unsteady terms).
70
10
0
0
0.5
1
1.5
2
1/Slf= a Ap / *1/S
2.5
Figure 4.14: Amplitude ratio in dependence of the inverse Strouhal number Sl1 =
Ref
p
= aA
. The amplitude ratio was normalized with the theoretical value
S
S
(Ap /As )0 .
71
4 Results
0.2
3 mm
5 mm
0.15
0.1
0.05
0.05
180
190
200
210
220
230
t* (non dim)
240
250
260
Figure 4.15: Particle trajectories for different force number f n, but same Ref and
S.
that is by a factor of two too small.
72
0.5
8
7
Ref (nondim)
0.4
6
0.3
5
4
0.2
3
0.1
0
0
2
1
0.05
0.1
0.15
S ( nondim)
0.2
0.25
10
0.25
0.2
0.15
0.1
0.05
0
0
a/ As
0.5
1
1.5
2
1/Sl = a A / *1/S
f
2.5
Figure 4.17: Normalized amplitude ratio in dependence on the inverse Strouhal number Sl1 . The colours indicates the strength of the forcing given by Aas .
73
4 Results
0.6
displacement (mm)
0.4
0.2
0
0.2
0.4
6.5
7.5
t (s)
8.5
74
Intensity [dB]
100
1 st
2 nd
3 rd
200
300
sled frequency
400
0
10
20
f [Hz]
30
40
50
Figure 4.19: Spectrum of the particle motion shown in 4.18. The dashed line shows
the sled frequency.
The strength of the deviation is controlled by the force time ratio f . If f 1, the
external forces can be neglected and the deviation from the linear theory increases
slowly with the strength of the convective terms (increasing Sl1 ). If the external
forces can not be neglected (f 1), the linear theory is valid only if the convective
terms can be neglected. If both the external force and the convective terms need to
be included, the deviation from the theory increases very fast.
In the next section we check the dependence on the density of the particle and
then we summarize and discuss our results.
ratio = fp introduces another dimension to the parameter space (S, Ref , ). Figure 4.20 shows the results for 3 mm steel and plastic spheres. All measurements
were done the same amplitude (5 mm). The numbers next to the data points are
the Reynolds numbers for that measurement. Figure 4.20 shows that the solution is
valid in the same frequency range for plastic and steel spheres. Given an amplitude
of 5 mm, the solution will be valid until Scrit 0.06. Another observation is that
the solution for the steel sphere is valid for higher Reynolds numbers Ref than for
the plastic spheres.
75
4 Results
From our experiment we can conclude that spheres with a diameter of 3 mm will
follow the prediction until a dimensionless frequency of S 0.06. Figure 4.21 shows
the errors for this experiment. In comparison with the previous experiments the
relative error for the amplitude ratio is smaller, since the density of the steel spheres
is much bigger and therefore easier to measure. Also the amplitude of the motion
is bigger, which leads to better signal to noise ratio. Nevertheless the plot shows
a large error bar, but the relative error is small. The error for the dimensionless
frequency is the same as for the plastic spheres.
0.8
0.7
0.6
Ap/As (nondim)
0.5
0.4
steel
plastic
steel with Basset History force
steel without Basset History force
plastic with History force
plastic without History force
1.15
0.94
0.76
0.48
0.3
0.34
0.2
0.1
0
0.22
0.005
0.06
0.01
0.12
0.007
0.1
0.2
0.02
0.03
0.04
0.05
S
0.06
0.07
0.08
Figure 4.20: Validity of the quasi-steady solution in dependence of the density ratio
and Reynolds number. Plot for the amplitude
76
4.3 Discussion
0.8
0.6
0.4
0.2
0
0.2
0.02
0.03
0.04
0.05
S
0.06
0.07
0.08
Figure 4.21: Same data as in Figure 4.20, but with error bars.
sufficient to characterize a flow in nature or experiments.
An application of our results might be possible for the response of rain droplets
to horizontal wind. The size of the droplet is comparable to the size of our particles.
Even though the Reynolds number based on the droplets radius and the relative
velocity (in the horizontal direction) might be small, the particle might be displaced
substantially by the wind, if the force time ratio is high. In general, a flow with
higher force time ratio leads to bigger particle amplitudes for the same Reynolds
number. This is explicitly important for experiments with tracer particles. The
experimentalist should ensure that both force time ratio and Reynolds numbers are
small, otherwise the particle will not follow the fluid. This leads to a problem, since
the fluid velocity might not be measurable directly.
Another conclusion is that the density seems not to be a control parameter.
Spheres with the same diameter, but different densities showed a clear deviation
from the theory at the same critical values. This effect is counter-intuitive and
needs further experimental verification.
4.3 Discussion
In the motivation, we identified two main questions: what is the correct equation
of motion and how does this equation depend on the control parameters? The first
question is answered with the experiments at low Reynolds numbers, whereas the
77
4 Results
second question was answered with high Reynolds number experiments.
Low Reynolds number experiments In the introduction the different approaches
for the equation of motion have been discussed. There are at least five different
proposals for the correct form of the equation. The previous experiments showed
that the equation of motion given by Maxey and Riley with the Basset Kernel for the
History force is correct and we also used this equation. The data base was expanded
towards lower frequencies and we saw that the quasi-steady solution is also valid for
smaller frequencies. The correct form of the Basset History term was questioned
by several authors, but from our experiments we can conclude that Basset History
term is correct. The solution was tested over a wide range of frequencies and there
is no evidence for a break down of the solution at very high or low frequencies.
Nevertheless the equation is only valid for small Reynolds numbers, as shown in the
second part.
Besides, the full solution was compared to the quasi-steady solution, but the
difference was very small for our experiments. We can assume that the sphere
reaches the quasi-steady motion within the first oscillation.
High Reynolds number experiments We concluded that the Basset History force
is correct for small Reynolds number, but the correct form of the equation for higher
Reynolds numbers remained unclear. Several approaches based on numerical solutions of the Navier Stokes equations have been proposed, but the physical justifications for these approaches were not given.
Our experiments showed the dimensionless frequency S and the Reynolds number
Ref are not enough to describe the motion of the particle. These parameter describe
only the dependence on the unsteady and convective terms in relation to the viscous
terms, but we saw that the external forcing is also important. The external forcing
can be measured by the force time scale f , which has to be small, so that the
linear prediction is valid. If the external force is large, the trajectories show a clear
non-harmonic shape.
We concluded that the linear theory is only valid, if the Reynolds number is small
(Ref < 0.1) and the force number time scale is also small (f < 0.04). This result
is very important, since the role of the forcing was not emphasized in the literature
before. As long as these two parameters are small enough, the theory is valid for all
values of the dimensionless frequency S.
78
4.3 Discussion
Measurement errors At this point, we should also take a critical look at our
experimental procedure. We estimated and discussed the errors before (see section
4.1.4).
In the case of the low Reynolds number experiments, the error bars for the dimensionless frequency S were much smaller than the error for the amplitude ratio. This
result is consistent with the experiment. In the low Reynolds number regime, the
amplitude of the particle motion was very small and the overall error due to fitting
was of the order 1%. Besides, the error due to the density measurement, which was
much bigger, was added. The error due to the density could be reduced with better
measurements and every experiment should be done with the same sphere, since the
density of the spheres varied, too. Nevertheless, even with a known density ratio,
the amplitude ratio is still affected by the error due to the measurement of the particle trajectory. One way to improve the data quality is a better illumination. We
were able to improve the data quality (see section 4.1.3), but maybe an even better
measurement is possible with a brighter lamp and better mirrors and lenses.
The high Reynolds number experiments are characterized by larger errors for the
dimensionless frequencies S. As discussed before this error increased both with S and
a. Since the frequency is already known very precisely, only a better measurement
for the particle radius and the viscosity could improve our data quality. The error
for the amplitude ratio is also larger. Since the error for the density stays constant,
the additional errors is caused by the amplitude error. A bigger amplitude leads to
a larger error, since the error scales with the amplitude. Nevertheless, the relative
error is better for higher amplitude because of the better signal to noise ratio.
Beside these errors, we have to discuss the systematical errors. As mentioned
before, the lenses are slightly defocused after one measurement day. Most likely this
comes from the vibration of the sled. The error due to the defocusing was estimated
from the videos. Compared to the other errors, this error can be neglected. Another
important question is the vibration of the camera itself. For very high frequencies
of the sled motion, the whole table vibrates and the camera (including the lenses
and mirrors) might vibrate as well. Unfortunately, we were not able to measure this
error, since no suitable measurement procedure is available. Targets on the sled will
vibrate and targets standing outside the measurement section cannot be focused.
We had to assume that the camera did not vibrate, but we cannot finally evaluate
whether this assumption is fulfilled.
Furthermore, we assumed that the whole fluid column follows exactly the sled
79
4 Results
motion. At every measurement run, we assured that no air is in the container, but
the internal motion of the fluid could not be measured. This assumption needs also
further attention.
Beside the correction of the mentioned sources of errors, the experiment has several
possibilities for further improvements. The next section gives a short overview of
possible modifications and interesting setups.
80
5 Outlook
The correct description of motion of a particle inside a fluid remains unclear. The
parameters of the problem can be grouped into three aspects: the material properties of fluid and particle, the time variation of the boundary conditions and the
forcing. We started with the definition of dimensionless parameters to characterize
the different aspects and defined the three parameters density ratio , dimensionless
frequency S and the Reynolds number Re. For velocity measurement the definition
includes the velocity, whereas the definition was based on the forcing for force measurements.
The first two experiments focused on the dependence on S and . In the literature
several experiments were reported and we expended the parameter space for the
dimensionless frequency towards lower frequencies. Our experiment, as well as all
experiments conducted before, showed that the equation of motion should include
the Basset History force with the classical Basset Kernel. Future experiments could
use more viscous fluids to expand the data range even further. Also the approach
by Garbin et al. [GDO+ 09] is very interesting. The ultrasonic excitation of bubbles
could be used to expend the data range towards higher frequencies.
The transient solution was tested, but the effect was small in the range of the
parameters explored. Therefore the parameter range of the experiment should be
increased towards higher frequencies. Also experiments with slower decreasing response functions could be undertaken. However, due to the fractional contribution
of the Basset History force all response functions are self-similar and the influence
of the transient part is weak.
In the third part of the results, we reported experiments at higher Reynolds numbers. We measured the deformation of the particle trajectory due to non linear
effects and the parameter space (, S and Re) was explored for the first time. Subsequent studies should be done to explore the parameter space completely.
From this starting point, a wide range of different aspects can be investigated.
The theory was restricted to Newtonian fluids, spherical particles and still fluid. The
81
5 Outlook
fluid could be replaced by a non Newtonian fluid or some additives could be used.
The resulting changes of the particle trajectory might lead to a further understanding
of these material properties.
Also the experiment could be repeated with non spherical particles. This will
lead the new kinds of memory force terms, as shown by Lawrence and Weinbaum
[LW88]. A numerical study with a slightly eccentric fluid spheroid moving in an
viscous medium was reported by Abbad, Souhar and Cabillina [ASC06]. The effect
of the shape of the particle is very important for industrial applications, for example
the mixing or the sedimentation of ingredients (see for example Dietrich [Die82]).
The sedimentation of particles leads to another interesting question: the interaction
between particles. The experiment could be performed with more than one particle
to examine the interaction. Also the interaction between the walls and the particles
are a recent research topic (for example Ten Cate et al. [TCNDVdA02]).
Our experiment was only performed with heavy particles, but the motion of bubbles is also very interesting. As noted in the introduction, the wake of a bubble
shows different characteristics. Another question is the possible deformation of the
bubble due to the unsteady flow. As noted by Garbin et al. [GDO+ 09], the motion
of microsized bubbles in blood is an important question for medical applications.
The physics of bubble motion is also interesting because of the interface between
bubble and fluid. The bubble can be deformed or vary the size because of the fluid
motion, but it is also possible that the bubble acts as rigid particle (see Abbad
and Souhar [AS04b] respectively Garbin et al. [GDO+ 09]). Another aspect is the
texture of the particle. Porous particles might show a different behaviour.
Of course, the sphere could also be replaced by a cylinder. On this topic experiments can already be found in the literature (see for example Sarpkaya [Sar86]). The
periodic forcing of the cylindric steel legs is a serious problem in the construction
of offshore platforms. The experimental setup might be interesting to achieve high
Strouhal and Reynolds numbers at the same time.
Last but not least, the fluid could be stirred during the experiment to see the
influence of a non uniform flow.
82
6 Appendix
0 = p v f u + f (v u)
+
(v
u)
+
dt
1
2
2 a2
2 a2
t t1
Z t
p + 1/2f
9
9 1
v(t
1 ) u(t
1)
v 3/2 u +
(v
u)
+
0=
dt
1
f
2 a2
2 a
t t1
Z t
v(t
1 ) u(t
1)
1
9
9 1
dt1
0 = (1 + )v 3/2 u +
(v u) +
2
2
2a
2 a
t t1
"
e dt1
i
= ... = eit i
t t1
83
6 Appendix
9
1
0 = (1 + )Ap ieit+i0 3/2 As ieit +
(Ap eit+i0 As eit )
2
2 a2
v(t
1 ) u(t
1)
9 1 Z t
dt1
+
2 a
t t1
1
9
9
9
1
9
9
1
i0
i0
3/2 As i +
iAs = ( )Ap ie +
iAp ei0
As +
Ap e +
2
2
2a
2 a
2
2a
2 a
Now we rewrite the result in terms of a complex number:
3/2 i + 29 a2 + 92 a i
Ap i0
e =
As
(1 + 12 )i + 92 a2 + 92 a i
(1 + 12 )i (1 + 21 )i
3/2 i + 29 a2 + 92 a i
+
=
(1 + 12 )i + 92 a2 + 92 a i (1 + 12 )i + 29 a2 + 92 a i
=1+
=1+
3/2i (1 + 21 )i
(1 + 12 )i + 92 a2 + 92
2(1 1)i
(21 + 1)i + 9 a2 + 9
84
a2
9
a2
9
a2
,
9
we show
2( 1)iS
Ap i0
e =1+
As
(2 + )iS + + 32 (1 + i) S
=1+
=1+
2( 1)iS
(2 + )iS + +
3 ei 4 2 S
2
2
2( 1)iS
(2 + )iS + + 3 Sei 4
a + ib
ac + bd
bc ad
= 2
+i 2
2
c + id
c +d
c + d2
is given by:
r = |z| =
v
u
u
t
ac + bd
c2 + d2
!2
bc ad
+ 2
c + d2
!2
e =1+
As
iS( + 2) + 3 S ei 4 +
Ap i
2iS( 1)
e =1+
As
iS( + 2) + 3 S 22 (1 + i) +
85
6 Appendix
iS( + 2) + 3 S 22 (1 + i) +
2iS( 1)
Ap i
e =
+
As
iS( + 2) + 3 S 22 (1 + i) + iS( + 2) + 3 S 22 (1 + i) +
Ap i 2iS( 1) + iS( + 2) + 3 S 22 + i3 S 22 +
e =
As
iS( + 2) + + 3 S 22 + i3 S 22
2
2
Ap i 3 S 2 + + i 2S( 1) + S( + 2) + 3 S 2
e =
As
+ 3 S 22 + i S( + 2) + 3 S 22
3 S 22 + + i 3S + 3 S 22
Ap i
e =
As
+ 3 S 22 + i S( + 2) + 3 S 22
2
+
wherea = 3 S
2
2
b = 3S + 3 S
2
2
c = + 3 S
2
2
d = S( + 2) + 3 S
2
|z| =
2
(3 S 2
v
u
u
t
ac + bd
c2 + d2
!2
bc ad
+ 2
c + d2
!2
2
+ )( + 3 S 22 ) + (3S + 3 S 22 )(S( + 2) + 3 S 22 )
( + 3 S 22 )2 + (S( + 2) + 3 S 22 )2
2
2
2
2 2 1/2
(3S + 3 S 2 )( + 3 S 2 ) (3 S 2 + )(S( + 2) + 3 S 2 )
+
( + 3 S 22 )2 + (S( + 2) + 3 S 22 )2
86
=
=
<
bcad
c2 +d2
ac+bd
c2 +d2
bc ad
ac + bd
(3S + 3 S 22 )( + 3 S 22 ) (3 S 22 + )(S( + 2) + 3 S 22 )
=
(3 S 22 + )( + 3 S 22 ) + (3S + 3 S 22 )(S( + 2) + 3 S 22 )
(For the sake of simplicity we gave only the result for < > 0. We used the matlab
function atan2 in order to obtain the correct result.)
Figure 6.1 and 6.2 show the results for the amplitude ratio and the phase difference
in the inertial frame of reference.
2.5
=5
=2
= 0.5
= 0.2
1.5
0.5
0 6
10
10
10
10
S (nondim)
10
10
10
Figure 6.1: Amplitude ratios for different density ratios. Inertial frame of reference.
87
6 Appendix
=5
=2
= 0.5
= 0.2
/8
[rad]
/8
10
10
10
10
S (nondim)
10
10
10
Figure 6.2: Phase difference for different density ratios. Inertial frame of reference.
du
0
=
As 2 eit 6f U 6f
+
f
dt
f
V
V
2 dt
p V
with V = 4/3a3 ,
88
0 =
a2
,
f
2f
a2
= S 1/2
2
9
and
S=
a2
.
9
p +
f
2 f
du
p f
a
U
a
+
=
As 2 eit 6f
6
f
3
3
dt
f
4/3a
4/3a
p +
f
2 f
U
du
p f
9
9
=
As 2 eit 2 f U 2 f
+
dt
f
2a
2a
p + 21 f
f
= 1 +
1
2
a2
2f
4
2
2f
a2
a4
f2
du
dt
du
dt
and get:
p = f 1/2f
p f
f 1/2f f
=
= 3/2
f
f
Insertion leads to:
q
9S
9
9
9S
du
du
= ( 3/2) As 2 eit 2 f U 2 f
U+ 2
dt
2a
2a
2
dt
In the next step we sort terms according to order u(t)
q
9S
9f 2 du 9
9
+
+
+
f
f
2a2
dt
2a2
2a2
9S
U = ( 3/2) As 2 eit
2
9S
9f 2
9
9
2 2 it+
+
A
i
e
+
+
f
p
f
2a2
2a2
2a2
9S
9f 2
9
9
+
Ap i2 ei + 2 f + 2 f
2
2a
2a
2a
Ap i
e =
As
+
9f
2a2
9S
2
9S
Ap ieit+ = ( 3/2) As 2 eit
2
s
9S
Ap iei = ( 3/2) As
2
( 3/2)
i2 +
2a2 f
2a2 f
9S
2
89
6 Appendix
The last step is to write the result in terms of S and :
Ap i
e =
As
Ap i
e =
As
a 9
( 3/2) 9
a2
9S
9
i2 + 2af2 2 i2 + i 2a92 f +
2a2 f
9S
2
( 3/2)S
i2 +
9
a2
9f
9 2a2
a2
9S 2
i
2
a 9
a 9
+ i 9
+ i 9
2a2 f
2a2 f
2
9S
2
Ap i
( 3/2)S
q
q
e =
1
9S
As
i2 S + i2 9S
+
i
+
i
2
2
2
Ap i
( 3/2)S
q
e =
1
As
(1
+
i)
+
i iS + 9S
2
2
i( 3/2)S
Ap i
q
e =
i 4 2
As
iS + 9S
e
+
2
2
1
2
i(1 + 12 3/2)S
Ap i
e =
As
(1 + 12 )iS + 32 S ei 4 + 12
2iS(1 1)
Ap i
e =
As
iS(21 + 1) + 3 S ei 4 + 1
2iS(1 1)
Ap i
e =
As
iS(21 + 1) + 3 S ei 4 + 1
Ap i
2iS( 1)
e =
As
iS( + 2) + 3 S ei 4 +
with S =
a2
9
and =
f
p
We do the same steps as in the last section and get for the norm of the result:
r = |z| =
90
v
u
u
t
ac + bd
c2 + d 2
!2
bc ad
+ 2
c + d2
!2
e =
As
iS( + 2) + 3 S ei 4 +
i2S( 1)
Ap i
e =
As
iS( + 2) + 3 S 22 (1 + i) +
a=0
b = 2S( 1)
2
c = + 3 S
2
2
d = S( + 2) + 3 S
2
2
(2S( 1))(S( + 2) + 3 S 22 )
+
( + 3 S 22 )2 + (S( + 2) + 3 S 22 )2
2
)
2
(2S( 1))( + 3 S
( + 3 S 22 )2 + (S( + 2) + 3 S 22 )2
2 1/2
bcad
c2 +d2
ac+bd
c2 +d2
bc ad
bc 0
=
ac + bd
0 + bd
+
3
S
c
2
= arctan( ) = arctan
d
S( + 2) + 3 S 22
(For the sake of simplicity we gave only the result for < > 0. We used the matlab
function atan2 in order to obtain the correct result.)
Figure 6.3 and 6.4 show the results for the amplitude ratio and the phase difference
in the comoving frame .
91
6 Appendix
1.4
1.2
1
=5
=2
= 0.5
= 0.2
0.8
0.6
0.4
0.2
0 6
10
10
10
10
S (nondim)
10
10
10
Figure 6.3: Amplitude ratios for different density ratios. Comoving frame.
[rad]
/2
=5
=2
= 0.5
= 0.2
/4
0 6
10
10
10
10
S (nondim)
10
10
10
Figure 6.4: Phase difference for different density ratios. Comoving frame.
92
u0 (t) = L
1 + s1/21
s + s1/2 + 1
and
u (t) = L1
1
s + s1/2 + 1
We solve the integral by the application of complex integration. We show the steps
for the first integral u (t) explicitly, the solution for the second integral can be
derived analogous.
First, we explicitly write down the Laplace integral we want to solve:
1
)
s + s1/2 + 1
1
1 Z +i st
e
ds
=
2i i
s + s1/2 + 1
1 Z
1
=
est
ds
2i H() s + s1/2 + 1
u (t) = L1 (
+i
. . . ds by the Hankel integral
Thereby we have replaced the Bromwich integral i
R
H() . . . ds. This was possible, because the integral kernel has no zeros and we can
deform the Bromwich path (line at from + on imaginary axis, with
greater than real part of zeros) to a Hankel path (loop around origin with radius
0). The sketch defines our integration path: The result will be given by the
sum of the three parts defined by:
1
u (t) =
2i
Z
A
Z
AB
Z
B
The first integration is done along the line s = xei with x going from (as
s goes from ). By defining the function F (s) = s+s11/2 +1 the resulting
integral is:
Z
A
est F (s)ds =
Z
est F (s)ds =
Z
Z
ee
i xt
F (xei )ei dx
93
6 Appendix
Im
Bromwich path
deformation
of integral
path
Hankel path
Re
Figure 6.5: Integration path: the direction is marked by the small arrows
est F (s)ds =
= i
0
=0
94
R
B
Z
Z
1
+
+
2i A
AB
B
Z
Z
1
1 Z xt
e F (xei )dx
i Z
1 xt
1
i
dx
e Im
|
s=xe
s + s1/2 + 1
Z
In the last step the real part is zero, because we integrate first from to and
then back, so both contributions cancel. Now we evaluate the Imaginary part I:
1
i
I = Im
|
1/2 + 1 s=xe
s
+
= Im
1
xei + x 2 [cos( 21 ) + i sin( 12 )] + 1
Under consideration of the rule for the imaginary part of a fraction Im(z) = Im( a+bi
)=
c+d
bcad
we get
c2 +d2
1
I=
x 2 sin( 21 )
1
Finally, the limit 0 is taken and the result is obtained: The following result is
obtained
1 Z xt
e Idx
u (t) =
0
1
x 2 sin( 12 )
1 Z xt
=
e
dx.
1
0
(1 x)2 + 2 x + 2(1 x)x 2 cos( 12 )
(6.1)
(6.2)
95
Bibliography
[AHP88] T.R. Auton, J.C.R. Hunt, and M. PrudHomme. The force exerted
on a body in inviscid unsteady nonuniform rotational flow. Journal
of Fluid Mechanics, 197(1):241257, 1988.
[All00] H.S. Allen. The motion of a sphere in a viscous fluid. Philosophical
Magazine Series 5, 50(306):519534, 1900.
[Arn11] H.D. Arnold. Limitations imposed by slip and inertia terms upon
stokess law for the motion of spheres through liquids. Physical
Review Series I, 32:233233, 1911.
[AS04a] M. Abbad and M. Souhar. Effects of the history force on an oscillating rigid sphere at low reynolds number. Experiments in fluids,
36(5):775782, 2004.
[AS04b] M. Abbad and M. Souhar. Experimental investigation on the history
force acting on oscillating fluid spheres at low reynolds number.
Physics of Fluids, 16:3808, 2004.
[ASC06] M. Abbad, M. Souhar, and O. Caballina. Note on the memory force
on a slightly eccentric fluid spheroid in unsteady creeping flows.
Physics of Fluids, 18:013301, 2006.
[Bas88] A.B. Basset. On the motion of a sphere in a viscous liquid. Philosophical Transactions of the Royal Society of London. A, 179:4363,
1888.
[BE11] E. Bodenschatz and M. Eckert. Prandtl and the gottingen school.
A Voyage Through Turbulence, page 40, 2011.
[Bou85] V.J. Boussinesq. Sur la rsistance quoppose un liquide indfini en
repos. In Comptes rendus de lAcadmie des sciences, volume 100,
page 9357, 1885.
97
Bibliography
[Bre05] C.E. Brennen. Fundamentals of multiphase flow. Cambridge Univ
Press, 2005.
[Bue66] Y.A. Buevich. Motion resistance of a particle suspended in a turbulent medium. Fluid Dynamics, 1(6):119119, 1966.
[Cau23] B.A.L. Cauchy. Recherches sur lquilibre et le mouvement intrieur
des corps solides ou fluides, lastiques ou non lastiques. 1823.
[CK02] C.F.M. Coimbra and M.H. Kobayashi. On the viscous motion of a
small particle in a rotating cylinder. Journal of Fluid Mechanics,
469:257286, 2002.
[CL56] S. Corrsin and J. Lumley. On the equation of motion for a particle
in turbulent fluid. Applied Scientific Research, 6(2):114116, 1956.
[CLL+ 04] CFM Coimbra, D. LEsperance, RA Lambert, JD Trolinger, and RH
Rangel. An experimental study on stationary history effects in highfrequency stokes flows. Journal of Fluid Mechanics, 504:353363,
2004.
[CR98] C.F.M. Coimbra and R.H. Rangel. General solution of the particle momentum equation in unsteady stokes flows. Journal of Fluid
Mechanics, 370(1):5372, 1998.
[Das11] S. Das. Functional fractional calculus. Springer Verlag, 2011.
[Deb34] P. Debye. Part i. dielectric constant. energy absorption in dielectrics
with polar molecules. Trans. Faraday Soc., 30:679684, 1934.
[Die82] W.E. Dietrich. Settling velocity of natural particles. Water Resources Research, 18(6):16151626, 1982.
[DN74] G. Doetsch and W. Nader. Introduction to the Theory and Application of the Laplace Transformation. Springer Berlin, 1974.
[dSV43] B. de Saint-Venant. Note a joindre au memoire sur la dynamique
des fluides. Comp. Rend, 17:1240, 1843.
[ERFM12] P. Ern, F. Risso, D. Fabre, and J. Magnaudet. Wake-induced oscillatory paths of freely rising or falling bodies. Annual Review of
Fluid Mechanics, 44(1), 2012.
98
Bibliography
[Gat83] R. Gatignol. The faxen formulae for a rigid particle in an unsteady
non-uniform stokes flow. J. Mc. Thor. Appl, 1(2):143160, 1983.
[GDO+ 09] V. Garbin, B. Dollet, M. Overvelde, D. Cojoc, E. Di Fabrizio, L. van
Wijngaarden, A. Prosperetti, N. de Jong, D. Lohse, and M. Versluis. History force on coated microbubbles propelled by ultrasound.
Physics of Fluids, 21:092003, 2009.
[GHG00] I Gonzlez, T.L Hoffmann, and J.A. Gallego. Precise measurement
of particle entrainment in a standing-wave acoustic field between 20
and 3500 hz. Journal of Aerosol Science, 31(12):14611468, December 2000.
[GM08] R. Gorenflo and F. Mainardi.
Fractional calculus: integral
and differential equations of fractional order. Arxiv preprint
arXiv:0805.3823, 2008.
[He00] R. Hilfer (ed.). Applications of fractional calculus in physics, volume
463. World Scientific Singapore, 2000.
[Hil02] R. Hilfer. Experimental evidence for fractional time evolution in
glass forming materials. Chemical physics, 284(1-2):399408, 2002.
[HK94] T. L. Hoffmann and G. H. Koopmann. A new technique for visualization of acoustic particle agglomeration. Review of scientific
instruments, 65(5):15271536, 1994.
[JBD03] M. Jenny, G. Bouchet, and J. Duek. Nonvertical ascension or fall
of a free sphere in a newtonian fluid. Physics of Fluids, 15:L9, 2003.
[JDB+ 04] M. Jenny, J. Dusek, G. Bouchet, et al. Instabilities and transition
of a sphere falling or ascending freely in a newtonian fluid. Journal
of Fluid Mechanics, 508(1):201239, 2004.
[JHI+ 07] G. Jourdan, L. Houas, O. Igra, J. L Estivalezes, C. Devals, and E. E.
Meshkov. Drag coefficient of a sphere in a non-stationary flow: new
results. Proceedings of the Royal Society A: Mathematical, Physical
and Engineering Science, 463(2088):33233345, 2007.
99
Bibliography
[KES98] I. Kim, S. Elghobashi, and W.A. Sirignano. On the equation for
spherical-particle motion: effect of reynolds and acceleration numbers. Journal of Fluid Mechanics, 367:221253, 1998.
[KK78] S.K. Karanfilian and T.J. Kotas. Drag on a sphere in unsteady
motion in a liquid at rest. Journal of Fluid Mechanics, 87:8596,
1978.
[LB93a] P.M. Lovalenti and J.F. Brady. The force on a sphere in a uniform flow with small-amplitude oscillations at finite reynolds number. Journal of Fluid Mechanics, 256:607607, 1993.
[LB93b] P.M. Lovalenti and J.F. Brady. The hydrodynamic force on a
rigid particle undergoing arbitrary time-dependent motion at small
reynolds number. Journal of Fluid Mechanics, 256(1):561605, 1993.
[LCTR05] D. LEsprance, C.F.M. Coimbra, J.D. Trolinger, and R.H. Rangel.
Experimental verification of fractional history effects on the viscous dynamics of small spherical particles. Experiments in fluids,
38(1):112116, 2005.
[LD09] E. Loth and A.J. Dorgan. An equation of motion for particles of
finite reynolds number and size. Environmental fluid mechanics,
9(2):187206, 2009.
[Lie27] H. Liebster. ber den widerstand von kugeln. Annalen der Physik,
387(4):541562, 1927.
[Lin99] E. R. Lindgren. The motion of a sphere in an incompressible viscous fluid at reynolds numbers considerably less than one. Physica
Scripta, 60(2):97110, August 1999.
[LS24] H. Liebster and L. Schiller. Kinematographische messungen der
fallbewegung von kugeln in zaher flussigkeit, auch in nachster nahe
einer wand. Physikalische Zeitschrift, 25:670, 1924.
[LW86] C.J. Lawrence and S. Weinbaum. The force on an axisymmetric
body in linearized, time-dependent motion: a new memory term.
Journal of Fluid Mechanics, 171(1):209218, 1986.
100
Bibliography
[LW88] C.J. Lawrence and S. Weinbaum. The unsteady force on a body
at low reynolds number; the axisymmetric motion of a spheroid.
Journal of Fluid Mechanics, 189(1):463489, 1988.
[MA92] R. Mei and R.J. Adrian. Flow past a sphere with an oscillation in the
free-stream velocity and unsteady drag at finite reynolds number.
Journal of Fluid Mechanics, 237(1):323341, 1992.
[Mei94] R. Mei. Flow due to an oscillating sphere and an expression for
unsteady drag on the sphere at finite reynolds number. Journal of
Fluid Mechanics, 270(1):133174, 1994.
[MK92] R. Mei and J.F. Klausner. Unsteady force on a spherical bubble
at finite reynolds number with small fluctuations in the free-stream
velocity. Physics of Fluids A: Fluid Dynamics, 4:63, 1992.
[ML98] J. Magnaudet and D. Legendre. The viscous drag force on a spherical
bubble with a time-dependent radius. Physics of fluids, 10:550554,
1998.
[MLA91] R. Mei, C. Lawrence, and R. Adrian. Unsteady drag on a sphere
at finite reynolds number with small fluctuations in the free-stream
velocity. Journal of Fluid Mechanics, 233:613631, 1991.
[MP00] N. Mordant and J.F. Pinton. Velocity measurement of a settling
sphere. The European Physical Journal B-Condensed Matter and
Complex Systems, 18(2):343352, 2000.
[MR83] M.R. Maxey and J.J. Riley. Equation of motion for a small rigid
sphere in a nonuniform flow. Physics of Fluids, 26:883, 1983.
[MR10] E.E. Michaelides and A. Roig. A reinterpretation of the odar and
hamilton data on the unsteady equation of motion of particles.
AIChE Journal, 2010.
[NA93] R. Natarajan and A. Acrivos. The instability of the steady flow
past spheres and disks. Journal of Fluid Mechanics, 254(1):323344,
1993.
101
Bibliography
[Nav23] C. Navier. Mmoire sur les lois du mouvement des fluides. Mmoires
de lAcadmie Royale des Sciences de lInstitut de France, 6:389440,
1823.
[OH64] F. Odar and W.S. Hamilton. Forces on a sphere accelerating in a
viscous fluid. Cambridge University Press, 1964.
[Ose10] C.W. Oseen. ber die stokessche formel und ber eine verwandte
aufgabe in der hydrodynamik. Arkiv fr matematik, astronomi och
fysik, 6:29, 1910.
[PKM95] W.C. Park, J.F. Klausner, and R. Mei. Unsteady forces on spherical
bubbles. Experiments in fluids, 19(3):167172, 1995.
[Poi28] S. D. Poisson. Mmoire sur lquilibre et mouvement des corps lastiques. LAcadmie des sciences, 1828.
[Sar86] T. Sarpkaya. Force on a circular cylinder in viscous oscillatory flow
at low KeuleganCarpenter numbers. Journal of Fluid Mechanics,
165:6171, 1986.
[SBB+ 09] G. Schlffel, M. Bastide, S. Bachmann, C. Mundt, and F. Seiler. Investigation of the acceleration of aluminum particles behind a shock
wave using instantaneous laser doppler velocimetry. Shock Waves,
19(2):125134, 2009.
[Sch28] J. Schmiedel. Experimentelle Untersuchungen ber die Fallbewegung
von Kugeln und Scheiben in reibenden Flssigkeiten. Universitt
Leipzig, 1928.
[SG00] H. Schlichting and K. Gersten. Boundary-layer theory. Springer
Verlag, 2000.
[SOC07] Y.D. Sobral, T.F. Oliveira, and F.R. Cunha. On the unsteady forces
during the motion of a sedimenting particle. Powder Technology,
178(2):129141, 2007.
[Sto46] G.G. Stokes. Report on recent researches in hydrodynamics. Brit.
Ass. Rep, 1:120, 1846.
102
Bibliography
[Sto51] G.G. Stokes. On the effect of the internal friction of fluids on the
motion of pendulums. Transactions of the Cambridge Philosophical
Society, 1851.
[Tan56] S. Taneda. Experimental investigation of the wake behind a sphere
at low reynolds numbers. Journal of the Physical Society of Japan,
11:11041108, 1956.
[Tch47] C.M. Tchen. Mean value and correlation problems connected with
the motion of small particles suspended in a turbulent fluid. PhD
thesis, TU Delft, 1947.
[TCNDVdA02] A. Ten Cate, C.H. Nieuwstad, J.J. Derksen, and H.E.A. Van den
Akker. Particle imaging velocimetry experiments and latticeboltzmann simulations on a single sphere settling under gravity.
Physics of Fluids, 14:4012, 2002.
[TK80] S. Temkin and S. S Kim. Droplet motion induced by weak shock
waves. Journal of Fluid Mechanics, 96:133157, 1980.
[TKT91] Y. Tsuji, N. Kato, and T. Tanaka. Experiments on the unsteady
drag and wake of a sphere at high reynolds numbers. International
journal of multiphase flow, 17(3):343354, 1991.
[TM82] S. Temkin and H. K. Mehta. Droplet drag in an accelerating and
decelerating flow. Journal of Fluid Mechanics, 116(1):297313, 1982.
[Tri88] D.J. Tritton. Physical fluid dynamics. Oxford, Clarendon Press,
1988.
[VBVWL05] C. Veldhuis, A. Biesheuvel, L. Van Wijngaarden, and D. Lohse. Motion and wake structure of spherical particles. Nonlinearity, 18:C1,
2005.
[VIG07] John Veysey II and N. Goldenfeld. Simple viscous flows: From
boundary layers to the renormalization group. Reviews of Modern
Physics, 79(3):883, 2007.
[VPS10] I.S. Vodopyanov, A.G. Petrov, and M.M. Shunderyuk. Unsteady
sedimentation of a spherical solid particle in a viscous fluid. Journal
of Fluid Dynamics, 45(2):254263, 2010.
103
Bibliography
[WB07] L. Wakaba and S. Balachandar. On the added mass force at finite
reynolds and acceleration numbers. Theoretical and Computational
Fluid Dynamics, 21(2):147153, 2007.
[Wie22] C. Wieselsberger.
Weitere feststellungen ber die gesetze
des flssigkeits-und luftwiderstandes. Physikalische Zeitschrift,
23:219224, 1922.
[Wil15] W.E. Williams. On the motion of a sphere in a viscous fluid. Philosophical Magazine, 29:526, 1915.
[WKB08] J.A. Weinstein, D.R. Kassoy, and M.J. Bell. Experimental study
of oscillatory motion of particles and bubbles with applications to
coriolis flow meters. Physics of Fluids, 20:103306, 2008.
104
Thanks
First, I would like to thank Prof. Dr. Eberhard Bodenschatz, my advisor, for the
opportunity to conduct this research in his group. He gave me the idea for the
experiment and inspired me with trust to set up the experiment. Also I would like
to thank Prof. Dr. Andreas Tilgner for being the co-referee.
I am also very grateful to Dr. Haitao Xu for his assistance throughout the whole
thesis project. His experience was a great support at all times. The discussions
with him about the experimental setup, the conduction of the experiments and the
results were always fruitful. Especially, I am grateful to him for writing independent
analysis codes, that were much faster and preciser than my codes. These programs
simplified the data analysis substantially. I also would like to thank Walter Pauls,
Mathieu Gibert and Ewe Wei Saw for discussions on the setup and the results.
The mechanical setup of the motor was done by Andreas Gehrke, Nils Winkelmann
and Daniel Wollborn under the supervision of Udo Schminke. I am very grateful
to them for their work and their patience with my numerous design changes that
caused lots of work for them. The electrical setup of the motor was done by Andreas
Renner and Ortwin Kurre, who I would like to to thank as well. Without them the
different signals might still interfere and stop the motor. The initial operation of
the motor and the improvement of the motor performance took a large part of the
thesis period. I am thankful to the Bosch Rexroth service team, especially Michael
Eggert, Wolfgang Walter and Mathias Wahler for their help. I am grateful to the
Heidenhain technical staff and Rainer Schtig from Rheinwerkzeug for their support.
Another company I would like to thank is HEGLA, which gave us the opportunity
to check the performance of their system. The comparison lead to new ideas and
was very useful for the project.
The setup of the experiment itself involved also almost everyone from our group.
Special thanks got to Joachim Hesse, Andre Heil, Andreas Kopp, Dr. Artur Kubitzek and Wolfgang Schubert. Last but not least, I want to thank my brother
Christian and Haitao Xu for correcting the various drafts of the thesis.
105
Eidesstattliche Erklrung:
Hiermit erklre ich, dass ich diese Abschlussarbeit selbstndig verfasst habe, keine
anderen als die angegebenen Quellen und Hilfsmittel benutzt habe und alle Stellen,
die wrtlich oder sinngem aus verffentlichten Schriften entnommen wurden, als
solche kenntlich gemacht habe. Darberhinaus erklre ich, dass diese Abschlussarbeit nicht, auch nicht auszugsweise, im Rahmen einer nichtbestandenen Prfung an
dieser oder einer anderen Hochschule eingereicht wurde.
Gttingen, January 18, 2013
(Sebastian Lambertz)