Powder Technology 268 (2014) 177190

Contents lists available at ScienceDirect

Powder Technology
journal homepage: www.elsevier.com/locate/powtec

Ultrasonic backscattering method for in-situ characterisation of

concentrated dispersions
Robert Weser a,, Sebastian Woeckel b, Benno Wessely a, Ulrike Steinmann b, Frank Babick a, Michael Stintz a
a
Technische Universitaet Dresden, Institute of Process Engineering and Environmental Technology, 01062 Dresden, Germany
b
Institute for Measuring and Automation Technology (ifak e.V.), Measurement Technology and Power Electronics, 39106 Magdeburg, Germany

a r t i c l e i n f o a b s t r a c t

Article history: An acoustic reection measurement technique is presented, which detects and evaluates the ultrasonic waves
Received 14 April 2014 backscattered at the particles within a liquid dispersion. The technique is based on an iterative pulse-echo method,
Received in revised form 31 July 2014 which gathers a sufciently high number of single reection signals and eventually conducts a statistical analysis
Accepted 14 August 2014
yielding two principal parameters of the backscattering signal: its maximum amplitude and its time-dependent
Available online 23 August 2014
decay. Theoretical considerations and experimental investigations show that maximum amplitude of the backscat-
Keywords:
tering signal correlates with the averaged backscattering coefcient of the particles and that the decay rate of
Dispersion characterisation the backscattering signal can be attributed to sound attenuation, even though the impact of multiple scattering
Concentrated dispersion (i.e. non-linear concentration effects) is much more pronounced as for a transmission setup. Since the developed
Particle characterisation measurement technique is specically designed for the in-situ characterisation of concentrated dispersion, it is
Sound scattering necessary to adequately account for multiple scattering effects when evaluating the backscattering data. For this
Backscattering measurement purpose a semi-empirical model is developed, which combines an established single particle scattering model
Backscattering coefcient with predictions of wave interference for dened microstructure of the dispersion. Experiments on two different
types of dispersion conrm the applicability of this approach for concentrations up to 30 vol.%. The experimental
results additionally show that the maximum backscattering amplitude is highly sensitive to particle size, whereas
the decay rate particularly reects the impact of particle concentration. Finally a scheme is proposed, which allows
for calculating the mean particle size and the volume fraction of the dispersed phase.
2014 Elsevier B.V. All rights reserved.

1. Introduction
The characterisation of dispersions basically intends to determine
both the composition and the state of the disperse system. Besides the
type and mixing ratio of continuous and disperse phase, the state is of
particular importance for the macroscopic dispersion properties. For
instance, the size and the shape of pigment particles within a paint are
decisive for the optical properties (colour, coverage) of the paint. In gen-
eral, particle size and concentration, along with thermodynamic and
stability issues, essentially inuence the state of a dispersion. Hence,
measuring particle size and concentration is compulsory for dispersion
characterisation.
There is a large variety of measurement techniques for the determi-
nation of particle concentration and size or size distribution. Despite
this, only few techniques facilitate the characterisation of concentrated
suspensions and emulsions, e.g. for the purpose of process monitoring.
Corresponding author. Tel.: +49 351 463 35226.
E-mail addresses: robert.weser@tu-dresden.de (R. Weser), sebastian.woeckel@ifak.eu
(S. Woeckel), benno.wessely@tu-dresden.de (B. Wessely), ulrike.steinmann@ifak.eu
(U. Steinmann), frank.babick@tu-dresden.de (F. Babick), michael.stintz@tu-dresden.de
(M. Stintz).
Optical detection principles can usually not be employed in this case,
because of the dispersions' opacity. Using ultrasound can overcome
this limitation. Moreover, ultrasonic measurement techniques are in
general capable to deal with eld application where sampling and prep-
aration (i.e. dilution) should be avoided (in-situ characterisation) [1].
A well-known technique, the ultrasonic attenuation spectroscopy,
uses a transmission arrangement for measuring the portion of sound
which is extinguished by the dispersion. Two acoustic transducers are
arranged face to face with a gap between them. The frequency depen-
dent attenuation (f) is used to determine particle size distribution,
for which purpose the measured attenuation is adapted to the values
predicted by physical model [2,3]. For strongly attenuating media, like
highly concentrated dispersions, the measuring gap has to be dimin-
ished to a few millimetres. Under these circumstances, single coarse
particles or highly viscous media can plug the gap and reduce the in-
situ capability of this method.
Beside the transmitted sound waves, which are evaluated within the
attenuation spectroscopy, the ultrasonic waves scattered at the particles
also contain information about the dispersion [4]. Taking advantage
of this particle-born sound signal, it is possible to arrange an open
measurement setup in reection mode without any gap. In medical
application this approach is used since many years to non-invasively

http://dx.doi.org/10.1016/j.powtec.2014.08.026
0032-5910/ 2014 Elsevier B.V. All rights reserved.
178 R. Weser et al. / Powder Technology 268 (2014) 177190
characterise biological tissue [5]. Also in non-destructive material test-
ing (NDT) ultrasound reection measurements are used to detect
defects or cracks within a material matrix [6]. Acoustic backscattering
measurements in ultrasound range are even used to characterise gran-
ular sediments at the sea oor [7]. Although the detection and analysis
of scattered or reected ultrasound exhibit clear advantages regarding
eld application, this method is not established within particle analysis
yet. On the one hand, measuring the acoustic amplitude or power
scattered by small particles (in micrometre range and below) is quite
difcult because of the weak scattering at particles that are small com-
pared to the wavelength of ultrasound, which applies to all particles
smaller 50 m at frequencies below 10 MHz. On the other hand, the
interpretation of the measured scattering signals, based on models for
sound propagation in disperse systems, is challenging because of the
physical complexity due to effects of multiple scattering. There is an
extensive use of ultrasonic backscattering methods within several elds
of application indeed. However, within medical application acoustic
backscatter is especially used for imaging purposes whereas in NDT
scattering mostly originates from single inhomogeneity. Neither of
both applications is aiming at the characteristics of concentrated parti-
cle systems. Hence, their specic methods of evaluating ultrasonic
reection/backscattering are not suited to the purpose of this paper.
This paper introduces an acoustic reection technique to gather the
ultrasonic waves which are reected, respectively backscattered at the
particles. The corresponding backscattering signal contains information
about the scattering behaviour of the dispersion under observation.
Experimental data, presented within this paper, show a high sensitivity
of backscattering signal to particle size and concentration. Backscattered
intensity is compared to backscattering coefcient calculated using an
empirical modelling approach. Therefore, an established single particle
scattering model is combined with an approach that calculates the inter-
ference of scattered waves for dispersion with dened microstructure.
2. Theory
2.1. Scattering of sound by single particles an analytical scattering model
The existence of particles accompanies with the scattering of sound
waves that are emitted into the dispersion (Appendix A). As it is well
known, the scattering behaviour of a particle is determined by the
size-to-wavelength ratio (between particle size x = 2a and sound
wavelength , respectively sound frequency f), expressed in form of
the dimensionless wave number
ka " x= " f " x=cc ;
where cc denotes the sound velocity in the continuous phase [2,8,9].
Furthermore, scattering is affected by the acoustic contrast between
particle and surrounding medium. Based on the acoustic impedance [8]
Z " c;
which can be determined for both phases, the relation of the acoustic
impedances of the disperse1 (index d) and the continuous (index c)
phase is denoted as the acoustic contrast
K Z d =Z c :
For small values of ka (ka b 0.1), visco-inertial and thermal effects ad-
ditionally affect scattering (ECAH and coupled phase theories) [3,10,11].
With vanishing values of ka scattering itself becomes rather weak. In
general, acoustic scattering is inelastic (dissipation of energy). In partic-
ular for small values of ka, the inelastic components dominate, while
1
For the disperse phase the acoustic impedance Zd is calculated using the longitudinal
sound velocity cd,L of the particle material.
the elastic scattering is rather weak. Within this work, examinations of
scattering are focused on dimensionless wave numbers ka N 0.1. In accor-
dance to this, only elastic scattering components will be considered.
A commonly accepted approach to describe elastic scattering was
developed by Faran [12]. In contrast to the ECAH or coupled phase the-
ories, Faran's model does not have any restrictions regarding particle
size. It can be applied to solid and uid particles because longitudinal
as well as transverse waves inside the particle are considered. Anderson
presented an alternative model, which is restricted to uid particles due
to the negligence of transverse waves [13,14]. Based on the material
properties (acoustic contrast K) of the involved phases (particle and
surrounding medium, Table 1) and the size-to-wavelength ratio ka,
Faran calculated the coefcients An (Appendix B).
The scattering of sound can be expressed in terms of the averaged
scattering coefcient H(), which is the averaged differential scattering
cross section per unit volume (Appendix A). For a distributed particle
size, determined by j size classes and a cumulative distribution function
Q(xj) with
X
Q j 1 4
j
and
! " ! "
Q j Q x j Q x j1 ; 5
the scattering coefcient has to multiplied with the number weights
Q0,j in the form
X ## ! "#2
#
H cN " #L ka j ; # " Q 0; j ; 6
j
whereat L(kaj, ) is the angle distribution factor comprising the coef-
cients An and the dependency on scattering angle (multipole expan-
sion, Appendix A). In terms of the volume weighted size distribution
Q3(xj), Eq. (6) has to be rewritten:
X ## ! "#2 Q
6 # 3; j
H "c " #L ka j ; # " : 7
V x
3
j j
Within this formulation the particle concentration is denoted as
volume concentration cV (instead of number concentration cN). Never-
theless, the assumption of a linear inuence of particle concentration
1 only is valid as long as scattering at one particle is not affected by the
scattering at particles in its vicinity (and as long as the scattering events
are incoherent).
2.2. Scattering of sound by concentrated dispersions an empirical
approach for covering non-linear effects at higher particle concentration
2
The acoustic scattering behaviour of dispersions, especially at higher
particle concentration, substantially differs from that of single particles.
With increasing particle concentration, the single particle scattering
approach is no longer valid because of the decreasing average distance
3 between the particles. Accordingly, the scattering at one particle is
affected by the scattering at neighbouring particles. This fact, known
as multiple scattering, has already been studied intensively for sound
waves interacting with collectives of particles [2,9]. Most theoretical ap-
proaches are not valid for particle concentrations above 10 vol.%, or they
are limited to specic material mixtures, due to the physical complexity
of multiple scattering in dispersions [16].
For this reason an alternative approach will be used to compute the
inuence of particle concentration on the scattering behaviour of dis-
persions. Instead of considering the particles as sources of scattering,
the complete dispersion as a volume with time and spatial variable
 R. Weser et al. / Powder Technology 268 (2014) 177190 179

Table 1
Material properties of uids and disperse solid at 25 C.

Material Density Sound velocity, longitudinal cL Sound velocity, transverse cT Acoustic impedance Z
[kg m3] [m s1] [m s1] [kg m2 s1]

Water, distilled 1000a 1500a 1.500 106

Silicone oil (Wacker AK50) 960b 992c 0.952 106
Glass beads (Potters Ballotini 3000 CP) 2500d 5525d 3380d 13.81 106
a
[8].
b
[15].
c
[16].
d
Estimated from data in [8] for crown glass.

acoustic properties is under observation. Thus, scattering is no longer
caused by the particles themselves but by the uctuations of acoustic
properties. These uctuations again are caused by the local uctua-
tion of particle concentration, which leads to a varying acoustic con-
trast due to the different material properties. In consequence of this
macroscopic treatment, the scattering coefcient H from a dispersion
volume V = i Vi is no longer proportional to the number of parti-
cles (Eq. (6)) but to the mean square deviation of Ni within a volume
element Vi [17,18]:
# ! "#
Z Ne 2 " ##L ka ; ##2
i j
H :
Vi
Vi
e 2 can be calculated [19]:
The variance of the number of particles N i
2
Nei Ni " W;
where W is the packing factor. The packing factor is a degree of the free
space between the particles, which decreases with increasing particle
concentration, and therefore it species an increase in spatial order.
According to Green and Twersky, the packing factor can be determined
by [19,20]:
2 Z
Nei 3 the backscattering coefcient
W 1 cN g r 1'd r:
Ni
Vi
The radial distribution function (or pair correlation function) g(r)
gives the probability for nding another particle within the distance r
from the particle under examination. The pair correlation function itself
depends on the effective mechanisms of interaction between the parti-
cles. That is the reason why there is no general formulation for g(r). In
the case of non-overlapping (impenetrable) particles within a uid
(hard sphere model), Twersky has given a closed form solution for the
packing factor as function of volume concentration [21]:
1cV m1
W m cV : 11
1 m1 " cV '
m1
It is easily seen that the packing factor W approaches 1 for vanishing
solid content (cV 0). The parameter m corresponds to the spatial
dimension (packing dimension) used to represent the particles within
the packing. With respect to the number of spatial dimensions needed
to represent the particle shape, values m b 3 seem to be reasonable.
Even greater values (m 3) can be used to address for fractal particles
[18].
Based on the packing factor, the scattering coefcient for concen-
trated dispersions can be calculated based on number weighted size
distribution:
X ## ! "#2
#
H cN " W m cV " #L ka j ; # " Q 0; j ;
j are calculated for 0.1 m b xi 300 m with xi = 0.1 m. The
while for Q3 it becomes
6 X ## ! "#2 Q
# 3; j
H " cV " W m cV " #L ka j ; # " : 13
x
3
j j
The concept of packing factor gives a stochastic parameter that
quanties the non-linear inuence of particle concentration on the
scattering of sound by dispersions, based on a statistical examination
of particles' positions. Moreover, the inuence of particle concentration
is being regarded independently from the scattering behaviour of the
8 single particle. These quite simplifying approximations represent a
pragmatic approach to depict the underlying, complex effects of multi-
ple scattering. Viewed realistically, the packing factor also depends on
the distribution of particle size and shape as well as on the interparticle
distance.
9
2.3. Sensitivity of sound scattering to dispersion's composition
Based on the previous modelling steps, the sensitivity of sound
scattering to composition of the dispersion will be examined next. The
composition of dispersions is mainly affected by the size and the con-
centration of the particles as well as by the material parameters of
both phases. Besides this, the examination of scattering behaviour
strongly depends on the scattering angle . For the sensitivity analysis,
10
$ %
H 180 14
is evaluated due to the central concept of investigating the capability of
dispersion characterisation by means of backscattering measurement.
Furthermore, backscattering coefcient can be visualised much easier
as function of input parameters. According to Eq. (13), the dependency
on particle size and material properties can be investigated separately
from concentration dependency.
2.3.1. Sensitivity to particle size
The inuence of particle size and material properties is captured
within the single particle scattering approach by calculating the coef-
cients An (Appendix A and Appendix B). The size impact can be traced
back to the dimensionless wave number ka (Eq. (1), cf. Appendix B).
Accordingly, Fig. 1 shows the backscattering coefcient for varying
dimensionless wave number ka. The backscattering coefcient rises
with increasing dimensionless wave number, even though not mono-
tonically. Furthermore, a discontinuity can be observed for ka 4.25,
which is caused by interferences between the longitudinal and the
transverse wave inside the particle. Correspondingly, this effect only
appears for solid particles.
For distributed particle size, the backscattering coefcient can be
calculated using Eq. (7). To investigate the inuence of distribution
width, a lognormal distribution with a constant volume weighted medi-
an size x50,3 = 50 m and a varying logarithmic standard deviation
12 0.1 ln 0.5 is chosen. The corresponding size distributions Q3(xi)
180 R. Weser et al. / Powder Technology 268 (2014) 177190

x 10-9 material parameters (Table 1). Afterwards, a recalculation of is done

Backscattering coefficient [sr -1m -1 6 at which each of the material parameters is changed by 5 %, respec-
tively 10 %. The corresponding ratio /0 as function of dimensionless
5 wave number quanties the sensitivity of backscattering coefcient
to material properties. Fig. 3 shows the inuence of density d on the
4 backscattering coefcient. A similar change of c (with opposite sign)
leads to an equivalent change of An due to the fact, that the density
3 ratio d/c is used to calculate the coefcients An (Eq. (9)). For the
suspension and the emulsion the proportionality of backscattering
2 coefcient to density change gets inversed. This seems plausible
because for the suspension d N c, whereas for the emulsion d b c
1 holds true. Furthermore, the inuence of density itself is affected
by the dimensionless wave number ka. In case of glass beads (solid
particles) in water signicant resonance effects can be again observed
0
0 2 4 6 8 10 for ka 4.25 due to interference effects mentioned before. For uid
particles (e.g. silicone oil), where transverse waves do not appear,
Dimensionless wave number ka [-]
there cannot be observed such uctuations. In the case of silicone oil
in water minor changes of density cause a comparatively intense
Fig. 1. Backscattering coefcient for glass beads (3000 CP) in water as a function of
dimensionless wave number ka, calculated using Eq. (8) with = 180, cN = 1/m3 and
change of backscattering coefcient. In general, the lower the density
corresponding material parameters (Table 1). difference between disperse (d) and continuous (c) phase the higher
is the inuence of a change of either of them.
Beside density, acoustic contrast is also affected by sound velocity.
logarithmic standard deviation is varied within the range ln = The inuence of sound velocity is again proportional to the acoustic con-
[0.1; 0.2; 0.3; 0.4; 0.5]. trast (Fig. 4). The lower the sound velocity difference between disperse
The calculation of backscattering coefcient is carried out without (cd,L, cd,T) and continuous phase (cc) the higher is the inuence of a
consideration of concentration dependency. For this reason, the volume change of one of them. As mentioned before, resonance effects occur
concentration within Eq. (7) is set to cV = 1. Fig. 2 shows the equivalent for solid particles. Furthermore, transverse sound velocity cd,T (in the
backscattering coefcient as function of frequency for these ve size dis- case of solid particles) signicantly affects the dimensionless wave
tributions. The spectral (frequency dependent) backscattering coef- number at which resonances occur.
cient indicates a clear dependency on distribution width: the broader Numerical investigations on the material dependency of backscat-
the size distribution the more smoothed is the spectral backscattering tering coefcient indicate a signicant inuence of material properties.
coefcient (f). Polydispersity causes an overlay of the monodisperse The strongest inuence can be attributed to the density ratio d/c and
backscattering coefcients (xmin b x xmax) in accordance to the distri- to the relation between the longitudinal sound velocity cd,L and cc. In
bution function Q3(x). general, the lower the acoustic contrast the stronger is the inuence of
material properties. The uncertainty of material properties can arise
2.3.2. Sensitivity to material properties from several reasons, e.g. additives changing effective material parame-
Apart from the dimensionless wave number ka, backscattering coef- ters or (unknown) temperature dependency of material properties.
cient is signicantly affected by the acoustic contrast K. Within the For a plenty of material, the density is known quite well. That also ap-
formulation of acoustic contrast, i.e. acoustic impedance, sound velocity plies to the sound velocity of uids. In contrast, sound velocities of
as well as density appear (Eq. (2)). For this reason, the sensitivity solid materials are often unknown and can hardly be measured in the
analysis is carried out separately for sound velocity and density. case of disperse materials.
The observed parameters are the sound velocities (cc, cd,L, cd,T) and
the densities (c, d) of the continuous and the disperse phase. Using 2.3.3. Sensitivity to particle concentration
the example of glass beads (3000 CP) and silicone oil (AK50) in water, In addition to the acoustic contrast and the size-to-wavelength ratio,
the backscattering coefcient 0 is initially calculated with nominal the particle concentration has also considerable inuence on the
Backscattering coefficient [sr -1m -1]

100 300
Distribution function Q3(x) [%]

80 250

200
60 ln = 0.1

ln = 0.2 150
40
ln = 0.3
100
ln = 0.4
20
50
ln = 0.5

0 0
10 100 300 0 5 10 15 20 25 30
Particle size x [m] Sound frequency f [MHz]

Fig. 2. Backscattering coefcient for glass beads (3000 CP) in water as a function of frequency f for different size distributions, calculated using Eq. (7) with = 180, cV = 1 and corre-
sponding material parameters (Table 1); left: particle size distributions Q3(x), right: backscattering coefcients (f).
 R. Weser et al. / Powder Technology 268 (2014) 177190 181

2 2
+10% +10%
1.8 1.8
+5% +5%
1.6 1.6
-5% -5%
1.4 -10% -10% 1.4
1.2 1.2
/0 [-]

/0 [-]
1 1
0.8 0.8
0.6 0.6
0.4 0.4
0.2 glass beads(3000CP) / water silicone oil(AK50) / water 0.2
0 0
0 2 4 6 8 10 0 2 4 6 8 10
Dimensionless wave number ka [-] Dimensionless wave number ka [-]

Fig. 3. Relative backscattering coefcient /0 for a percentage change of density d for left: glass beads (3000 CP) and right: silicone oil (AK50) in water; is calculated using Eq. (8) with
= 180 and cN = 1/m3; 0 is calculated with the nominal material parameters (Table 1).

backscattering behaviour. As presented in Eq. (13), the concentration 3. Experiments

dependency is denoted within the product
3.1. Measurement setup and signal extraction

Sm cV " W m cV : 15 Experimental investigations on backscattering behaviour of dis-

The function Sm increases as long as cV b cV,max, reaches its maximum
at cV = cV,max and decreases for cV N cV,max (Fig. 5, left). Furthermore, con-
centration dependency is inuenced by the packing dimension m,
whereupon cV,max as well as the maximum Sm(cV,max) shows a non-
linear dependency on m. If m increases cV,max and Sm(cV,max) decrease.
Besides this, the sensitivity of Sm to particle concentration rises with
an increasing packing dimension (Fig. 5, right).
persions are performed with a self-developed measurement setup
(Fig. 6). One single transducer, operating in pulse-echo mode, emits
the exciting sound waves and also detects the sound waves reected
or backscattered from interfaces or particles. The delay line is used to
protect the active surface of the transducer from the abrasive particle
medium, to separate the excitation and the backscattering signal and
to provide far eld conditions in terms of scattering.
In general, it is quite difcult to detect signals from weak scattering
particles due to the low signal amplitude. For this reason, signal-to-

2 2
1.8 cd,L cd,T cc 1.8
1.6 1.6
1.4 1.4
1.2 1.2
/0 [-]

/0 [-]
1 1
0.8 0.8
0.6 0.6
0.4 0.4
glass beads(3000CP) / water
0.2 0.2
0 0
0 2 4 6 8 10 0 2 4 6 8 10 0 2 4 6 8 10
2 Dimensionless wave number ka [-] 2
1.8 1.8
1.6 1.6
1.4 1.4
+10%
1.2 1.2
/0 [-]

/0 [-]

+5%
1 1
-5%
0.8 0.8
-10%
0.6 0.6
0.4 0.4
0.2 silicone oil(AK50) / water 0.2
0 0
0 2 4 6 8 10 0 2 4 6 8 10
Dimensionless wave number ka [-] Dimensionless wave number ka [-]

Fig. 4. Relative backscattering coefcient /0 for a percentage change of left: particle's longitudinal sound velocity cd,L, center: particle's transverse sound velocity cd,T and right: sound
velocity of the continuum cc for glass beads (3000 CP, top) and silicone oil (AK50, bottom) in water; is calculated using Eq. (8) with = 180 and cN = 1/m3; 0 is calculated with
the nominal material parameters (Table 1).
182 R. Weser et al. / Powder Technology 268 (2014) 177190

0.16 1

m=1 0.8
0.12 m=2 cV=0,001
m=3
m=4 c =0,01
V
0.6

Sm/S1 [-]
m=5 cV=0,1
Sm [-]

0.08
0.4

0.04 cV=0,2
cV=0,3 0.2
cV=0,4
0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 1 2 3 4 5
Volume concentration cV [-] Packing dimension m [-]

Fig. 5. Concentration inuence on the parameter Sm left: Sm as function of volume concentration cV, right: Sm/S1 as function of the packing dimension m.

noise ratio has to be improved e.g. by using narrow-band transducers
or adapted signal processing techniques. Details on capturing and
processing the backscattering signals, especially on extracting particle
signal components from the raw signal u(t) (e.g. eliminating window
reections), have already been described in a previous paper [22]. The
frequency range for measurements is limited between 4 MHz and
14 MHz with a resolution of 1 MHz due to the narrow-band transducer
characteristics in combination with a pulse shaped excitation.2 This
comparatively low frequency resolution mainly is caused by the fact
that signal processing is performed in both time and frequency domain
(limited timefrequency resolution). Moreover, during signal process-
ing a sufciently high number of single echo signals is evaluated. This
statistical evaluation is necessary to ensure a presentative imaging of
the whole particle system, because various particle constellations and
a high particle number within the measuring volume are captured
[22]. Evaluating a minimum number of N = 1500 single echo signals
yields the backscattering amplitude b(t), which corresponds to the
mean value of all single echo signals. It shows a characteristic maximum
that arises from the dispersion layer directly in front of the delay line
(start of measuring zone, Fig. 7). Amplitude drops off with an increasing
travelling time (Fig. 7, left). As a rst approximation, this decay can be
regarded as exponential, because it is caused by the attenuation of
sound , that acts on both, the emitted and the scattered sound waves
[8]. Taking the logarithm of amplitude b(t), this relationship can be
linearised (Fig. 7, right). Within logarithmic scale two parameters can
be extracted as function of window centre frequency: the maximum
amplitude bmax and the decay of backscattering amplitude . The maxi-
mum value bmax arises from the scattering at the particles in the absence
of attenuation effects.
The acoustic impedance of the dispersion Zdisp is a function of
concentration, especially at high particle concentrations. However, the
inuence of a varying acoustic impedance Zdisp on the reection
strength at the interface between delay line window and dispersion
does not affect the determination of bmax since the window reections
are already removed during pre-processing of raw signals.
with different particle size fractions and different particle concen-
trations. This procedure enables basic investigations regarding the
inuence of i.e. size, concentration and the material properties on back-
scattering behaviour. Suspensions are prepared by using glass beads,
which are dispersed in water. The original disperse material (Potters
Ballotini 3000 CP) is graded into ve size fractions by vanning and siev-
ing (Fig. 8). Emulsions, consisting of silicone oil (Wacker AK50) and
water are produced using a rotor/stator mixer. Three different size dis-
tributions are obtained by the variation of dispersion energy. For
stabilisation of oil droplets (avoid unwanted changes of particle size),
Tween80 is used as an emulsier (1 vol.% of the total oil content).
Size distribution is monitored off-line by means of laser diffraction
(Sympatec Helos). Both disperse materials are mixed in distilled water
at different concentrations up to 30 vol.%. An ultrasonic attenuation
spectrometer, a DT1200 from Dispersion Technologies, Inc. is used to
measure sound attenuation of suspension. This transmission mea-
surement cannot be applied to the emulsion, because an unwanted
change of droplet size is observed (ow forces cause a progressive
emulsication) within the measuring cell of the DT1200. For compari-
son purposes, the attenuation is transfered to its time-dependent
delay
transducer window
line measuring
zone
pulser
particles
u(t)

3.2. Application to suspensions and emulsions

The performance of the developed measurement technique is exam-
ined for dispersions of solid particles (suspensions) as well as such of
liquid particles (emulsions). Both kinds of dispersion are produced
2
During signal processing the backscattering signals are ltered in the frequency do-
main using a Hanning window with varying centre frequency between 4 MHz and
14 MHz (integer MHz values) and 500 kHz bandwidth.
receiver
t
Fig. 6. Measurement setup (schematic) used for experimental studies of ultrasonic backscat-
tering from dispersions. The ultrasonic transducer is pressure sensitive and yields a signal
amplitude (in terms of electrical voltage), which is proportional to the sound pressure p.
Reection signal is recorded as electrical voltage u as function of travelling time t.
 R. Weser et al. / Powder Technology 268 (2014) 177190 183

x 10-6
7 -103

Backscattering amplitude b [a.u.]

Backscattering amplitude b [dB]

measuring zone
6
bmax
-109
5
-115
window

window
4
~e c t
3 -121
2
-127
1
noise floor Tfit
0 -133
0 5 10 15 20 25 0 5 10 15 20 25
Time t [s] Time t [s]

Fig. 7. Backscattering amplitude b as function of travelling time t resulting from statistical evaluation of 1500 single echo signals captured from 30 wt.% glass beads (3000 CP, unclassied)
in water, left: linearly scaled amplitude inuenced by attenuation , right: logarithmically scaled amplitude in decibels.

equivalent t: using the sound velocity of the dispersion c, which is also measured
with the DT1200. During acoustic measurements, the dispersions are
cm=s' " s tempered in a water quench because material properties themselves
t dB=s' " dB=cm' 16
104 " m are strongly depending on temperature. Moreover, the dispersions

100
Distribution function Q3 (x) [%]

80

60
CP_<10m
CP_10-20m
40
CP_20-32m
AK_<20m
CP_32-50m
20 CP_>50m
AK_<43m

unclassified AK_<73m

0
0 20 40 60 80 100 120 140 0 20 40 60 80
Particle size x [m] Particle size x [m]

Fig. 8. Particle size distributions measured with laser diffraction spectrometer Sympatec Helos, left: glass beads, right: silicone oil droplets. Particle size fractions are named according to
their separating (glass beads) or maximum particle size (oil droplets).

x 10-5
1.8 1.4
10 MHz
1.6 4 MHz
6 MHz
12 MHz 1.2
1.4 14 MHz
Maximum bmax [a.u.]

8 MHz
1
Decay [dB/s]

1.2
1 0.8
0.8 0.6
0.6
0.4
0.4
0.2 0.2

0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
Median particle size x50,3 [m] Median particle size x50,3 [m]

Fig. 9. Measurement results for glass bead fractions (3000 CP) in water at 4 vol.% as function of median particle size x50,3 for selected sound frequencies, left: maximum amplitude bmax,
right: decay of backscattering amplitude . Lines between points of measurement only serve as a guide to the eye.
184 R. Weser et al. / Powder Technology 268 (2014) 177190

<10m 10-20m 20-32m 32-50m >50m

1.2 4 5

1.0 4 MHz 8 MHz

12 MHz 4

Decay [dB/s/MHz]
Decay [dB/s/MHz]

3
0.8
3
0.6 2
2
0.4
1
1
0.2

0 0 0
0 0.2 0.4 0.6 0.8 1.0 1.2 0 1 2 3 4 0 1 2 3 4 5
Attenuation t [dB/s/MHz] Attenuation t [dB/s/MHz] Attenuation t [dB/s/MHz]

Fig. 10. Comparison between decay of backscattering amplitude and sound attenuation t for three selected frequencies.

kept stirring during measurements in order to avoid sedimentation
(suspension) or creaming (emulsion) and to ensure a permanent ex-
change of particles within measuring zone. N = 1500 single echo signals
are captured and evaluated as described before.
4. Results
The resulting backscattering parameters bmax and are recorded as a
function of the window centre frequency. In addition to , the variance
of the linear t is plotted as a ratio of t length Tt (Fig. 7). For further
analysis bmax as well as are plotted against particle size (median size
x50,3) and concentration for selected frequencies.
4.1. Glass bead suspension
Fig. 9 shows the results of 4 vol.% glass bead suspensions. The max-
imum backscattering amplitude bmax increases with an increasing parti-
cle size and increasing frequency as well the higher the dimensionless
wave number ka the higher is bmax. In the case of amplitude decay
there is no such unique dependency against particle size: for particle
sizes x50,3 b 10 m decay increases with decreasing particle size, where-
as for x50,3 10 m decay is proportional to both x50,3 and frequency.
An increase of amplitude decay for small particles can be attribut-
ed to an increase of attenuation for ka b 0.2 due to dissipative attenu-
ation mechanisms [2]. In general, is mainly caused by attenuation ,
but cannot be equalised to it. Multiple scattering of sound waves gener-
ate additional components within backscattering signal yielding to a
decay b , especially for higher frequency and for higher attenuation
[22,23]. This fact is conrmed by the results of attenuation measure-
ment (Fig. 10). Furthermore, the comparison between and indicates
an inuence of frequency the higher the frequency the greater is the
difference . Differences between data points and parity line, that is
particularly signicant at 4 MHz, can be attributed to the uncertainty of
sound velocity measurement. The small gap between both transducers
leads to a comparatively large error by determining c. Hence, this
error propagates within the calculation of t.

3 2.5

2.5
2 4MHz
Decay [dB/s]
Decay [dB/s]

2 6MHz
1.5 8MHz
1.5 10MHz
1
1 12MHz
CP_<10m 14MHz
0.5 0.5
CP_10-20m
0 0

3 3.5 5.5
5
2.5 CP_20-32m CP_32-50m 3 CP_>50m 4.5
2.5 4
Decay [dB/s]

Decay [dB/s]

2 3.5
2 3
1.5
1.5 2.5
1 2
1 1.5
0.5 1
0.5
0.5
0 0 0
0 4 8 12 16 20 24 28 32 0 4 8 12 16 20 24 28 32 0 4 8 12 16 20 24 28 32
Volume concentration cV [vol-%] Volume concentration cV [vol-%] Volume concentration cV [vol-%]

Fig. 11. Decay of backscattering amplitude for glass bead (3000 CP) fractions in water as a function of volume concentration cV for selected frequencies. Lines between points of
measurement only serve as a guide to the eye.
 R. Weser et al. / Powder Technology 268 (2014) 177190 185

x 10-6 x 10-6
1.8 4

1.6 CP_<10m 3.5

4MHz

Maximum bmax [a.u.]

Maximum bmax [a.u.]
3
1.4 6MHz
2.5
1.2 8MHz
2 10MHz
1 CP_10-20m
1.5 12MHz
0.8
1 14MHz

0.6 0.5
0.4 0
x 10-6 x 10-5 x 10-5
8 2 3
7 CP_20-32m
1.6 2.5

Maximum bmax [a.u.]

Maximum bmax [a.u.]

6
2
5 1.2
4 1.5
3 0.8
1
2
0.4
0.5
1
0 0 0
0 4 8 12 16 20 24 28 32 0 4 8 12 16 20 24 28 32 0 4 8 12 16 20 24 28 32
Volume concentration cV [vol-%] Volume concentration cV [vol-%] Volume concentration cV [vol-%]

Fig. 12. Maximum of backscattering amplitude bmax for glass bead (3000 CP) fractions in water as function of volume concentration cV for selected frequencies. Lines between points of
measurement only serve as a guide to the eye.

When particle concentration rises, also increases (Fig. 11). Inde-
pendent from frequency, a nearly linear increase of as a function of
volume concentration cV can be discovered. Except for the smallest
particles (fraction CP_b10m), the slope of the curves rises with
frequency. In the case of the size fraction CP_b10m an evaluation for
frequencies below 8 MHz is not feasible due to the low signal-to-noise
ratio caused by the comparatively low scattering crosssection of the par-
ticles. Additionally, the penetration depth of the emitted sound waves is
strongly decreased for particle concentrations above 20 vol.% due to the
growing attenuation. Because of that, the decay of backscattering ampli-
tude cannot be determined appropriately for the smallest glass bead
fraction.
Maximum backscattering amplitude bmax shows a considerable non-
linear behaviour (Fig. 12). The maximum amplitude bmax initially rises
with increasing concentration independently from particle size and fre-
quency. The slope decreases with further increasing concentration,
followed by a decrease of bmax at higher concentrations. Accordingly, a
characteristical maximum occurs at concentration cV,max with cV,max
depending on particle size and frequency.
4.2. Silicone oil emulsion
Fig. 13 shows the results of 4 vol.% silicone oil emulsions. The maxi-
mum backscattering amplitude bmax increases with an increasing droplet
size as well as with increasing frequency the higher the dimensionless
wave number ka the higher is bmax. This relationship is similar to the
glass bead suspensions. However, the maximum backscattering ampli-
tude for oil droplets is signicantly higher (for a comparable particle

x 10-5
2.5 1.4
4 MHz 10 MHz 1.2
2 6 MHz 12 MHz
Maximum bmax [a.u.]

8 MHz 14 MHz
1
Decay [dB/s]

1.5 0.8

1 0.6

0.4
0.5
0.2

0 0
0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35
Median particle size x50,3 [m] Median particle size x50,3 [m]

Fig. 13. Measurement results for silicone oil droplets (AK50) in water at 4 vol.% as a function of median particle size x50,3 for selected sound frequencies, left: maximum amplitude bmax,
right: decay of backscattering amplitude . Lines between points of measurement only serve as a guide to the eye.
186 R. Weser et al. / Powder Technology 268 (2014) 177190
4MHz 6MHz 8MHz
1 1.5
AK_<20m AK_<40m
0.8
Decay [dB/s]
0.6
0.4
0.2
0 0
0 5 10 15 20 25 30 0 5 10 15
Volume concentration cV [vol-%] Volume concentration cV [vol-%]
Fig. 14. Decay of backscattering amplitude for silicone oil droplet (AK50) fractions in water as a function of volume concentration cV for selected frequencies. Lines between points of
measurement only serve as a guide to the eye.
size), although the acoustic contrast K of oil/water is less than that of
glass/water. This is due to the fact that in the case of oil/water Zd b Zc
(K = 0.635), whereas for glass/water Zd N Zc (K = 9.207) holds true. In
other words: an acoustic contrast K b 1 causes stronger scattering than
an equivalent contrast with K N 1.
Amplitude decay also rises with increasing droplet size and fre-
quency (Fig. 13). The higher the frequency the higher is the increase
of as function of droplet size. Unlike the results of the glass bead sus-
pensions, amplitude decay does not show a continuous increase with
increasing particle concentration (Fig. 14). For the smallest droplet
size fraction (AK_b20m), nearly is proportional to volume concentra-
tion cV. For greater oil droplets remains static for concentrations above
15 vol.% or even decreases for cV N 20 vol.%. As mentioned before, the
decay of backscattering amplitude (caused by attenuation) is affected
by multiple scattered waves. Because of the comparatively strong scat-
tering at the oil droplets, multiple scattering becomes rather signicant
in this case. The higher the concentration is the stronger is the inuence
of multiple scattering.
Maximum backscattering amplitude bmax as function of particle con-
centration cV also shows a considerable non-linear behaviour with a
characteristical maximum (Fig. 15). Curve shapes for AK_b20m frac-
tion slightly differ from those for greater droplet sizes at frequencies
above 8 MHz.
5. Discussion
Since size, concentration and material properties are known, it is
possible to compare the measurand bmax to the calculated backscatter-
ing coefcient in order to evaluate the measurement results. The back-
scattering coefcient quanties the backscattered intensity in relation
to the intensity incident on the measurement volume. In general,
sound intensity I is proportional to the square of sound pressure p [8].
Thus, for comparative purposes backscattering amplitude has to be
squared, because b2max p2 I holds true.3
Within Fig. 16 the measurement results (b2max) are compared with
the calculated backscattering coefcient (using Eq. (7) with size distri-
butions in Fig. 8 and material properties in Table 1) for the ve different
glass bead fractions. Only data at 2 vol.% solid content is used in order to
exclude non-linear concentration effects.
Thus, a qualitative good agreement between the maximum back-
scattered intensity (b2max) and the backscattering coefcient can be
3
The proportionality between b2max and p2 is based on the proportionality b u p
(s. chapter 3.1).
10MHz 12MHz 14MHz
2
AK_<70m
1.6
Decay [dB/s]
1
1.2
0.8
0.5
0.4
0
20 25 30 0 5 10 15 20 25 30
Volume concentration cV [vol-%]
stated as long as particle concentration is low and multiple scattering
effects can be ignored. The underlying single particle scattering model
(Faran) is capable to cover the inuences of particle size, sound frequen-
cy and material parameters. Hence, sufcient knowledge of the material
parameters is required. Qualitative differences within curve progres-
sions in Fig. 16 can be attributed to a deviation from spherical particle
shape (which is assumed in scattering theory) as well as to an uncer-
tainty of material parameters in the case of the dispersed solid. Missing
knowledge of incident sound intensity, which is necessary to normalise
b2max, leads to the quantitative difference between b2max and . The mea-
surement volume depends on the spatial dimensions of the emitted
sound eld, which again depend on the transducer's and the delay line's ge-
ometry. The corresponding effective sound eld was not determined, nei-
ther by calculation nor by measurement. So, it is not available for
calculation purposes. The uncertainty with regard to the measurement vol-
ume is also relevant for the qualitative difference between b2max and .
If multiple scattering cannot be ignored anymore, the backscattering
coefcient has to be calculated according to Eq. (13). The underlying
concept of a packing factor W predicts a non-linear inuence of particle
concentration (Fig. 5). Similar trends are also found for the maximum
p
backscattered amplitude (bmax), which is comparable to (Figs. 12
and 15). Within Fig. 17 both the calculated () and the measured
(b2max) backscattered intensity equivalents are compared. For calcula-
tions, packing dimension m is treated as a free parameter, which initially
is determined by tting Sm to measurement data b2max(cV) (minimum
square deviation t). The values of m scatter in the range of 2.1 to 3.5.
For the glass bead suspensions, data points are approximately
located along a straight line (Fig. 17, left, grey line), except the data of
the CP_N50m fraction. When emulsion data are compared with
suspension data, a steeper slope for the regression line can be stated
(Fig. 17, right, black line). This second straight line much better approx-
imates the data from the CP_N50m fraction and from the oil droplets.
Oil droplets can be regarded as ideal spherical particles, whereas the
solid glass particles more or less differ from spherical shape. Especially
the small glass beads (b50 m) rather are plate shaped (which was
veried by SEM micrograph). Accordingly, the calculation (based on
the assumption of spherical particles) leads to a higher scattering inten-
sity compared to the measured one, because the mean effective scatter-
ing cross section is smaller compared to that of the equivalent sphere.
It is also possible that the closed form solution of the packing factor W
(Eq. (11)) no longer is valid for larger particles, respectively larger
values of ka. For a more substantiated analysis, data from ideal spheri-
cally shaped particles with well-known material parameters have to
be evaluated.
As already discussed, the decay of backscattering amplitude results
from the attenuation of sound waves travelling through the dispersion.
 R. Weser et al. / Powder Technology 268 (2014) 177190 187

4MHz 6MHz 8MHz 10MHz 12MHz 14MHz

x 10-6 x 10-5 x 10-5

6 1.8 3
AK_<20m AK_<70m
1.6
5 2.5
AK_<40m 1.4

Maximum bmax [a.u.]

Maximum bmax [a.u.]

4 1.2 2
1
3 1.5
0.8
2 0.6 1
0.4
1 0.5
0.2
0 0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30 0 5 10 15 20 25 30
Volume concentration cV [vol-%] Volume concentration cV [vol-%] Volume concentration cV [vol-%]

Fig. 15. Maximum of backscattering amplitude bmax for different silicone oil droplet (AK50) fractions in water as a function of volume concentration cV for selected frequencies. Lines be-
tween points of measurement only serve as a guide to the eye.

Hence, an increase of concentration coincides with an increase of positioning of the sensor. Since sound waves penetrate highly concen-
(Figs. 11 and 14). In case of the glass bead suspensions, is in good trated dispersions, this method is predestined for in-situ application,
agreement with the attenuation for ka b 1, whereas for ka N 1 the because measurements can be performed directly in the dispersion
difference grows with increasing dimensionless wave without any sampling and conditioning, e.g. dilution.
number due to the growing inuence of multiple scattering (Fig. 10). A statistical evaluation yields two backscattering parameters: the
The amplitude decay is proportional to concentration and the corre- maximum value bmax and the decay of backscattering amplitude .
sponding slope itself is proportional to frequency (Fig. 11). In principle, Both parameters could be shown to be sensitive to the composition
this relationship also is valid for the AK_b20m silicone oil emulsion, and state of the dispersion, especially to particle size and concentration.
whereas for greater droplet sizes (AK_b43m and AK_b73m) Thereby, bmax shows an increased sensitivity compared to the particle
does not increase signicantly for concentrations above 15 vol-% size, whereas rather is sensitive to particle concentration. The maxi-
(Fig. 14). This observation can be attributed to a rising inuence mum backscattering amplitude bmax is affected by the acoustic scatter-
of multiple scattering [22]. In general, the difference between attenua- ing strength of the dispersion that again is affected by the scattering
tion and amplitude decay increases with a growing inuence of crosssection of the single particles as well as by the concentration of
multiple scattering, i.e. with an increase of the particles' scattering particles within the measurement volume. For a given material mixture
cross section. (acoustic contrast) and a selected frequency range of measurement,
bmax mainly is affected by particle size. Furthermore, bmax shows a char-
acteristically non-linear dependency on particle concentration arising
6. Conclusion and perspectives from multiple scattering effects. Based on an empirical approach, the
consequences of multiple scattering can be covered in principle.
The presented acoustic backscattering technique covers both captur- Hence, it is possible to calculate the backscattering coefcient for con-
ing and evaluating of ultrasonic signals, that are reected by a disper- centrated dispersions. Theoretical investigations show a high sensitivity
sion. The open measurement setup allows an almost arbitrarily of backscattering coefcient to particle size, concentration and material

x102 x10-10
5 1,8
Squared maximum amplitude bmax [a.u.]
Backscattering coefficient [m-1 sr-1]

CP_<10m 1,6
4 CP_10-20m
2

1,4
CP_20-32m
CP_32-50m 1,2
3
CP_>50m 1,0
0,8
2
0,6

1 0,4
0,2
0 0
2 4 6 8 10 12 14 2 4 6 8 10 12 14
Window centre frequency fF [MHz] Window centre frequency fF [MHz]

Fig. 16. Comparison between backscattering coefcient and squared maximum amplitude for glass bead suspensions as a function of frequency at 2 vol.% solid content, left: backscattering
coefcient calculated using Eq. (7) with corresponding material parameters (Table 1) and right: squared maximum amplitude b2max.
188 R. Weser et al. / Powder Technology 268 (2014) 177190

x10-10 x10-10
7 7

Squared maximum value bmax [a.u.]

CP_<10m 3000CP / water
6 CP_10-20m AK50 / water 6
CP_20-32m
2

5 5
CP_32-50m
4 CP_>50m 4

3 3

2 2

1 1

0 0
0 2 4 6 8 10 12 14 16 18 0 2 4 6 8 10 12 14 16 18
Backscattering coefficient 10 2 [m-1sr -1] Backscattering coefficient 10 2 [m-1sr -1]

Fig. 17. Squared maximum of backscattering amplitude b2max compared to backscattering coefcient , left: for the ve glass bead fractions and right: for glass beads and silicone oil droplets
(with all fractions in each case). is calculated using Eq. (13) with corresponding material parameters (Table 1).

parameters. Thus, material parameters should be well-known when respectively of penetration depth. In contrast to conventional trans-
evaluating scattering behaviour. mission setups, is not only affected by attenuation but also affected
For verication purposes, the square of maximum backscattering by multiple scattering components overlaying the single scattering
amplitude (as backscattered intensity equivalent) is compared to back- signals. Accordingly, cannot be generally equalised to attenuation
scattering coefcient. A good qualitative agreement between both , because b holds true. Moreover, for a device-independent deter-
values can be found. Beside scattered intensity, the decay of backscatter- mination of the sound eld geometry (especially the beam spread
ing amplitude depicts another important measurement parameter angle) also has to be taken into account. Otherwise the distance-
that can be regarded as an attenuation equivalent. Attenuation causes dependent decrease of emitted sound intensity, resulting from beam
a decrease of signal amplitude as function of travelling time, spread angle, yields to an overestimation of .
As a matter of principle, the sensitivity of both backscattering
parameters to particle size and concentration can be used for a model-
based determination of particle size (Fig. 18, x). The maximum backscat-
backscattering measurement
tering amplitude bmax holds relevant information regarding particle
bmax(f) (f) size. If material properties are known, b2max can be calibrated to the
calibrate backscattering coefcient in order to calibrate the measurement
backscattering intensity setup, i.e. the measurement volume. Uncertainty regarding the concen-
2 tration inuence furthermore can be reduced by evaluating (cV).
bmax(f) (f)
Moreover, the spectral information of backscattering parameters
! " en-
ables prediction regarding particle size distribution width Q 3 . There-
correlate fore, further experimental and theoretical investigations are necessary.
no measurands
material properties
with state variable(s) Alternatively, changes of acoustic measurement parameters can be sim-
are known?
or reference data ply correlated with particle size information originating from reference
yes measurements (e.g. laser diffraction). This approach is of particular
importance for process monitoring and even can be realised with the
initialize
packing dimension actual development stage of the measurement method.
m=3 With regard to process application, the construction of the backscat-
m tering probe (delay line with transducer) has to be improved in order to
tolerate harsh requirements (e.g. high pressure and temperature). For
determine estimate
this purpose, the water lled delay line should be replaced by a solid
packing dimension particle size
(f)=func(x, Q3, f, cV, m)
one made of glass ceramic, e.g. sapphire. Moreover, the emitted sound
m=func(x, f)
intensity should be increased in order to further improve signal-to-
x (Q3) noise ratio of backscattering signal. This could be achieved by using
estimate high intensity transducers or special signal waveforms for transducer
concentration excitation.
(f)=func(cV, x, f)

cV Acknowledgement

particle size x (Q3) The authors would like to thank for the nancial support on the part
concentration cV of the AiF (grant number 16681 BR/2). The research project 16681 BR/2
of the research association Forschungsgesellschaft fuer Messtechnik,
Fig. 18. Flow chart for an evaluation of backscattering parameters bmax and in order to Sensorik und Medizintechnik e.V. Dresden was sponsored by the AiF
estimate a mean particle size x and/or a concentration cV . in context of the program for supporting the industrial research and
 R. Weser et al. / Powder Technology 268 (2014) 177190 189

development founded by the Federal Ministry of Economics and Tech- Thus, the scattering coefcient is given as differential scattering
nology, based on a resolution of the German Parliament. cross section per unit volume [25]:

Appendix A. Scattering of sound by a sphere dC sca 2

H cN " cN " jLj : A:8
d
Due to an incident plane wave (pinc) impinging a particle of radius a, The number concentration cN, which is dened as the number of
a scattered wave (psca) propagating radially away from the particle. particles N per volume V and therefore indicates the volume dependen-
cy of H. Eq. (8) is derived based on the assumption, that the scattering
portions of the N particles superimpose each other without interaction
and interference (in other words: linear).

Appendix B. Determination of coefcients An according to

Faran's model

By evaluating the boundary conditions at the particle-uid interface,

Fig. A1. Schematic diagram of sound scattering by a spherical particle with convention on Faran found the coefcients An in the form [12]:
parameters for incident pinc and scattered psca wave as well as on scattering angle .
i " tan n
An ; B:1
1 i " tan n
Both waves, the incident (A.1) and the scattered (A.2) one, can be
expressed as synthesis of series of spherical Bessel and Hankel functions where the complex phase shift tan(n) is given by [26]:
(jn, hn):
tan n ka tan n kL a; kT a
c

X tan n tan n ka d
B:2
n tan n ka tan n kL a; kT a
c

pinc r; p0 i " 2n 1 " P n cos " jn kr ; A:1 d

n0
with

X n
psca r; p0 i " 2n 1 " P n cos " An " hn kr: A:2 jn ka
n0 tan n ka ; B:3
nn ka

The Legendre polynomial Pn of order n, in terms of the scattered 0

ka " jn ka
wave, accounts for the angle distribution of scattering amplitude. tan n ka ; B:4
jn ka
Furthermore, the scattered wave psca contains the coefcients An, de-
pending on the underlying scattering theory as function of the wave 0
ka " nn ka
number ka as well as of the material properties. tan n ka : B:5
nn ka
For an innite large distance from the particle (r ) the resulting
amplitude (p = pinc + psca) becomes [24]
As well as the spherical Bessel function jn, the spherical Neumann
ikr
function nn is a linear combination of the Hankel functions hn. Primes
e (jn, nn) of the Bessel and Neumann functions denote differentiation
pr; p0 " Lka; ; A:3
r with respect to the argument (ka). The material properties of the phases
involved are used to determine tan n, which is given by
with the angle distribution factor
tan n kL a; kT a B:6

i X 2 3
Lka; " 2n 1 " P n cos " An : A:4
k n0 tan k a n2 n
n L
2
k a2 6
6 1 tan kL a 2
n n1k 2a tan n kT a
T
7
7
T 6 ! " 7
2 4n2 nk a 2 tan k a2 2
n n 1 tan n kT a 5
Beside the amplitude, the scattering coefcient is a further, frequently 2
T
2 n L
1 tan n kL a n n1
k a T
2
tan n kT a
used parameter for quantifying scattering. Based on the scattering cross 2

section of a spherical scatterer [24]

with
Z

2 4 X 2 xf
C sca 2 jLj d 2n 1 " jAn j ; A:5 kL a B:7
0 k n0 cd;L

that quotes the relation between the totally scattered intensity and the being the dimensionless wave number of longitudinal wave and
intensity of the incident wave, the differential scattering cross section
xf
is dened as the portion of scattered intensity dCsca relative to the solid kT a B:8
cd;T
angle d [24]:

being the dimensionless wave number of compression wave inside the

dC sca particle [12].
2
jLj A:6
d
References

with [1] P. Hauptmann, N. Hoppe, A. Puettmer, Application of ultrasonic sensors in the process
industry, Meas. Sci. Technol. 13 (2002) R73R83.
[2] R.E. Challis, M.J.W. Povey, M.L. Mather, A.K. Holmes, Ultrasound techniques for char-
acterizing colloidal dispersions, Rep. Prog. Phys. 68 (2005) 15411637.
d 2 sind: A:7 [3] A.S. Dukhin, P.J. Goetz, Ultrasound for Characterizing Colloids, Elsevier, 2002.
190 R. Weser et al. / Powder Technology 268 (2014) 177190
[4] A. Volker, M. de Kroon, Characterization of suspensions by ultrasonic reection mea-
surements, Ultrasonics 36 (1998) 283289.
[5] K.K. Shung, G.A. Thieme, Ultrasonic Scattering in Biological Tissues, CRC Press, 1993.
[6] S.E. Kruger, J.M.A. Rebello, J. Charlier, Broadband ultrasonic backscattering applied to
nondestructive characterization of materials, IEEE Trans. Ultrason. Ferroelectr. Freq.
Control 51 (2004) 831837.
[7] P.D. Thorne, D.M. Hanes, A review of acoustic measurement of small-scale sediment
processes, Cont. Shelf Res. 22 (2002) 603632.
[8] R. Millner, Wissensspeicher Ultraschalltechnik, VEB Fachbuchverlag, Leipzig,
1987.
[9] A.K. Hipp, G. Storti, M. Morbidelli, Acoustic characterization of concentrated suspen-
sions and emulsions. 1. Model analysis, Langmuir 18 (2002) 391404.
[10] P.S. Epstein, R.R. Carhart, The absorption of sound in suspensions and emulsions. I.
Water fog in air, J. Acoust. Soc. Am. 25 (1953) 553565.
[11] J.R. Allegra, S.A. Hawley, Attenuation of sound in suspensions and emulsions: theory
and experiments, J. Acoust. Soc. Am. 51 (1972) 15451564.
[12] J.J. Faran, Sound scattering by solid cylinders and spheres, J. Acoust. Soc. Am. 23
(1951) 405418.
[13] V.C. Anderson, Sound scattering from a uid sphere, J. Acoust. Soc. Am. 22 (1950)
426431.
[14] C. Feuillade, C.S. Clay, Anderson (1950) revisited, J. Acoust. Soc. Am. 106 (1999)
553564.
[15] A.G. Wacker Chemie, Data sheet Wacker AK50 silicone oil, 2012.
[16] A. Richter, Ultraschalldaempfungsspektroskopie grobdisperser Systeme, (PhD Thesis)
Dresden University of Technology, 2008.
[17] B. Angelsen, A theoretical study of the scattering of ultrasound from blood, IEEE
Trans. Biomed. Eng. 27 (1980) 6167.
[18] P. Bascom, R. Cobbold, On a fractal packing approach for understanding ultrasonic
backscattering from blood, J. Acoust. Soc. Am. 98 (1995) 30403049.
[19] H. Green, The Molecular Theory of Fluids, Dover Publications Inc., 1969
[20] V. Twersky, Transparency of pair-correlated, random distributions of small scatterers,
with applications to the cornea, J. Acoust. Soc. Am. 65 (1975) 524530.
[21] V. Twersky, Acoustic bulk parameters in distributions of pair-correlated scatterers,
J. Acoust. Soc. Am. 64 (1978) 17101719.
[22] R. Weser, S. Woeckel, B. Wessely, U. Hempel, Particle characterisation in highly
concentrated dispersions using ultrasonic backscattering method, Ultrasonics 53
(2013) 706716.
[23] R. Weser, S. Woeckel, U. Hempel, B. Wessely, J. Auge, Particle characterization in
highly concentrated suspensions by ultrasound scattering method, Sensors Actuators
A 202 (2013) 3036.
[24] P.M. Morse, K.U. Ingard, Theoretical Acoustics, Princeton University Press, Princeton,
New Jersey, 1986.
[25] R. Sigelmann, J. Reid, Analysis and measurement of ultrasound backscattering from
an ensemble of scatterers excited by sine-wave bursts, J. Acoust. Soc. Am. 53 (1973)
13511355.
[26] A.E. Hay, D.G. Mercer, On the theory of sound scattering and viscous absorption in
aqueous suspensions at medium and short wavelengths, J. Acoust. Soc. Am. 78
(1985) 17611771.