Documente Academic
Documente Profesional
Documente Cultură
TO ULTRASONICS
20-251-728
Ultrasonics is a branch of acoustics dealing with the generation and use of (generally)
inaudible acoustic waves. There are two broad areas of use, sometimes called as the low- and
high-intensity applications. In low-intensity applications, the intent is to convey information
about or through a system, while in high-intensity applications, the intent is to permanently alter
a system. To some extent, the low- and high-intensity fields are also delineated by a frequency
range and power level. Thus, low-intensity applications typically involve frequencies on the order
of 106 Hz or higher and power levels on the order of milliwatts. High-intensity applications will
typically involve frequencies of 5 to 100 kHz and powers of hundreds to thousands of watts. In
actual fact, the total frequency range of all ultrasonic applications is enormous, ranging from 5 -
10 kHz to as high as 10 GHz. There are also applications, such as sonar, which are exceptions to
the previous categorizations, since intense power levels are involved in conveying information
via underwater sound.
1-2
The main difference between the basic methods of Ultrasonics and the more specialized
ones used in Ultrasonic NDE is in the approach to the elastic medium. In ultrasonics, the material
is usually assumed to be ideal (isotropic, homogeneous, linear, attenuation-free, dispersion-free,
temperature-independent, etc.) in order to study the basic laws of elastic wave propagation in
their simplest form. In ultrasonic NDE, real materials with more complex elastic properties
(anisotropy, inhomogeneity, nonlinearity, attenuation, dispersion, temperature-dependence, etc.)
are considered. The primary purpose of ultrasonic NDE is to understand the wave-material
interaction and assess the sought material properties from the observed deviation in the ultrasonic
response from that of an ideal, defect-free medium. The main topics to be covered in Ultrasonics
and Ultrasonic NDE are listed in the following table.
1-3
Ultrasonics Ultrasonic NDE Wave-Material Interaction
(high-frequency wave (the propagation medium is an (special physical phenomena due
propagation in elastic imperfect medium, i. e., a real to interaction with
media) material) imperfections)
1-4
Elements of ultrasonic waves
x
u( x , t ) = A cos[ω ( t − ) + ϕ], (1.1)
c
where A denotes the amplitude, ω = 2π f is the angular frequency, where f is the cyclic
frequency, ϕ is the phase angle at x = t = 0, and c denotes the propagation (phase) velocity.
Here, u could represent longitudinal or transverse displacement of a string, particle velocity in a
solid, pressure wave amplitude in a gas, or a number other physical quantities. The basic wave
parameters of propagation, the wavelength λ and the period of vibration T are related through
c / λ = 1/ T = f .
u t = t1 c
A t2
t3
-A
λ
1-5
In order to better facilitate algebraic manipulations needed to solve ultrasonic wave
propagation problems, we will use complex notation without explicitly indicating as such.
Equation (1.1) can be written as follows
u( x , t ) = U e± i ( k x − ω t ) , (1.2)
where U is a complex amplitude that includes the phase term, and k is the so-called wave
number. In this notation, only the real part of the complex quantity corresponds to the actual
physical quantity, therefore the + and - sign conventions are equivalent.
It will be found that three basic types of wave may exist in a material, depending on
whether it is solid or fluid and depending on the nature of its boundaries. The wave types are
dilatational, shear, and surface waves. Propagation velocities will depend on the material, and
may range from 102 m/s to 104 m/s. The basic natures of the waves are shown in Figure 1.2.
The dilatational wave (also called longitudinal or pressure wave) may exist in solids, liquids, and
gases, and is the familiar wave of acoustic theory. It is seen that particle motion is in the same
direction as the propagation direction. A shear wave (also called transverse or equivoluminal
wave), on the other hand, may exist only in a solid. It is seen that particle motion is at right
angles, or transverse, to the direction of propagation. These are the only two types of wave that
may exist in an extended media. If a free surface exists on a solid half-space, a surface (or
Rayleigh) wave may also propagate. Such a wave has a complicated particle motion at the
surface, and has an amplitude that rapidly decays away from the surface. A main point to
emphasize is that these waves are all well known from classical acoustic and elasticity theory. No
"mysterious" new waves are associated with ultrasonics.
Finally, the behavior of waves upon encountering surfaces and boundaries is another
fundamental aspect of wave propagation. The simplest situation is depicted in Figure 1.3a, where
a wave encounters a boundary at right angle or normal incidence. The interaction only involves
reflection of some of the wave and transmission of a portion, with the amount of energy in each
part depending on the material characteristics. A more complicated situation may arise,
particularly in solids, when the wave strikes at an angle, or at oblique incidence. What may
occur, as shown in Fig. 1.3b, is that two types of waves are reflected for a single incident wave.
This phenomenon is known as mode conversion, and is illustrated for the case of a pressure wave
generating both pressure and shear waves. Yet another aspect is involved when waves encounter
edges. Complex scattering and diffraction of the waves may occur, similar to optics. This is
meant to be illustrated by Figure 1.3c.
1-6
Longitudinal Wave:
Shear Wave:
Surface Wave:
1-7
a) Incident Wave Reflection
ρ1, c1
ρ2, c2
Transmission
b)
Incident Wave Reflection
θi θr
Liquid
Solid
θd Longitudinal
Transmission
θs Shear
Transmission
c)
Incident Wave Reflection
θi θr
Edge Diffraction
Figure 1.3 Different types of acoustic wave interaction with material discontinuities:
reflection and transmission (a), refraction and mode conversion (b), and
diffraction and scattering (c).
1-8
Historical aspects
Since ultrasonics is a part of acoustics, its development, particularly in the early years, is
to some extent embedded in the broad developments in acoustics. However, the history of
classical acoustics can be traced back to Pythagoras in the 6th century B. C., while investigations
of high-frequency waves did not originate until the 19th century. The era of modern ultrasonics
started about 1917, with Langevin's use of high-frequency acoustic waves and quartz resonators
for submarine detection. Since that time, the field has grown enormously, with applications
found in science, industry, medicine and other areas. The following is meant to identify dates of
some of the major developments in ultrasonics.
1-9
1921 Cady discovered the quartz stabilized oscillator.
1922 Hartmann developed the air-jet ultrasonic generator.
1925 Pierce developed the ultrasonic interferometer.
1927 Wood and Loomis described effects of intense ultrasound.
1928 Pierce developed the magnetostrictive transducer.
1928 Herzfeld and Rice developed molecular theory for dispersion and
absorption of sound in gases.
1929 Sokolov proposed use of ultrasound for flaw detection.
1930 Debye and Sears and Lucas and Biquard discover diffraction of light by
ultrasound.
1937 Sokolov invented an ultrasonic image tube.
1937 Dussik brothers made first attempt at medical imaging with ultrasound.
1938 Pierce and Griffin detect the ultrasonic cries of bats.
1939 Pohlman investigated the therapeutic uses of ultrasonics.
1940 Firestone, in the United States and Sproule, in Britain, discovered
ultrasonic pulse-echo NDT.
1940 Sonar extensively developed and used to detect submarines.
1945 Piezoelectric ceramics discovered.
1948 Start of extensive development of power ultrasonic processes.
1948 Start of extensive study of ultrasonic medical imaging in the United States.
For more details on this subject, see Karl F. Graff, A History of Ultrasonics, in Physical
Acoustics, Volume XV (Academic Press, New York, 1982).
1-10
1.2 Vibrations of a simple oscillator
The vibrational characteristics of the simple oscillator will be reviewed first. This will
provide the opportunity to emphasize certain vibrational characteristics of special interest in
ultrasonics. The simple wave equation will then be reviewed. This basic equation will be found
to be applicable to a wide range of ultrasonic wave propagation problems.
k m
mu + k u = 0 . (1.3)
where u denotes the displacement from the equilibrium position. Putting this in the form
u + ω o2 u = 0 , ωo = k , (1.4)
m
1-11
where ωo is the natural (angular) frequency of the free vibration of an undamped mechanical
resonator. This result may also be expressed in terms of an amplitude and a phase angle as
u = D cos (ω o t − φ ) , (1.6)
where
C
D = B2 + C 2 , and tan φ = . (1.7)
B
If a dashpot is added to the oscillator, as shown in Fig. 1.5, the equation of motion is
simply
mu + d u + k u = 0 . (1.8)
ωo
u + u + ω 2o u = 0 , (1.9)
Q
where ω 2o = k / m as before and the so-called quality factor Q is defined here as an impedance
ratio at the resonance frequency
k m
Figure 1.5 Simple damped mechanical resonator of a spring, a mass and a dashpot.
1-12
X reactive mωo k mk
Q = = = = . (1.10)
Rdissipative d d ωo d
The solution to Eq. 1.9 may take several forms depending on the value of Q. The so-called
“damped periodic” case (Q ≥ ½) is the most applicable to ultrasonics and to most vibration
situations. The solution, in exponential form, is
u = e−ζ t ( A1 ei Ω t + A2 e − i Ω t ) , (1.11)
where
1 ωo
Ω = ωo 1− 2
and ζ = . (1.12)
4Q 2Q
Ω is the natural (angular) frequency of the free vibration of a damped mechanical resonator and
ζ is a decay constant. In terms of sine and cosine functions, the result is
The well-known pattern of free vibration of a damped oscillator is shown in Fig. 1.6. It may be
easily shown that the ratio between amplitudes exactly one period (T ) apart is
u
e -ζ t
1-13
u( t )
= eζ T ≈ eπ/ Q . (1.14)
u( t + T )
This leads to an other more general definition of the quality factor through the “logarithmic
decrement,” ∆, given by
u( t ) π
∆ = ln = ζT = . (1.15)
u( t + T ) Q
Consider now the forced vibrations of simple oscillators. The case of an undamped
oscillator shown in Fig. 1.4 subjected to a harmonic forcing function exp (i ω t ) results in the
governing equation
mu + k u = F o e i ω t (1.16)
or
F
u + ω 2o u = o e i ω t . (1.17)
m
Fo
u = k e iω t . (1.18)
ω2
1 − 2
ωo
Noting that Fo / k is simply the displacement of the spring-mass system under a static force Fo,
the amplitude response may be written as
++u = 1
. (1.19)
δ st
+ ω2
1− 2
ωo
+
This response is shown in Fig. 1.7.
1-14
From the result of Eq. 1.19 and Figure 1.7, it is seen that as ω → ω o , the amplitude
“blows up,” i.e., approaches infinity. This is the phenomenon of undamped resonance. When the
forced vibrations of a damped oscillator previously shown in Fig. 1.5 are considered, the
governing equation of motion becomes
u
δst
1
ω / ωo
0
0 0.5 1 1.5 2 2.5
ωo F
u + u + ω o2 u = o e i ω t . (1.20)
Q m
Fo Fo
iω t
u = m
ωo ω e
= k e iω t . (1.21)
ω 2o −ω +i
2 ω2 ω
Q 1− 2 + i
ωo Q ωo
At the resonance frequency, the vibration amplitude δ m = Q δ st and it is 90o off-phase with
the driving force. In other words, the peak amplitude at the resonance frequency is determined
1-15
solely by the mechanical impedance of the dashpot δ m = Fo / d ω o . After a bit of manipulation,
it is possible to express this in terms of a real amplitude and a phase angle as
δ st
u = e i (ω t − φ )
ω ! 1/ 2
(1.22)
+ 2 "
2 2
1− ωo2
ω 2
Q ω o2 "
#
and
ω
Q ωo
φ = atan . (1.23)
ω2
1− 2
ωo
The behavior of the amplitude and phase are shown in Fig. 1.8 for various degrees of
damping. It is quite evident that for decreasing values of the quality factor (increasing damping)
the sharpness of the resonance decreases. Put another way, for increasing damping, the
bandwidth of the oscillator increases. It is also seen that the change in phase angle is quite rapid
for small frequency changes at low damping, where the phase angle changes much slower for
heavy damping. A physical quantity of great interest in vibrations is the energy dissipation rate,
which offers another way to define the quality-factor of the system. The probably most basic
definition of this parameter is
2 πU s
Q = , (1.24)
Ud
where Us denotes the maximum stored energy and Ud is the energy dissipated per each cycle
(this definition is essentially the same as Eq. 1.15). For the simple oscillator, the maximum
stored energy would be given by
1 m Fo2
Us = k δ 2m = , (1.25)
2 2d2
where δ m is the maximum vibration amplitude at the resonance frequency as obtained from Eq.
1.22. The energy dissipated per cycle may be found by carrying out the integral
1-16
4
u
δ st Q=∞
3 5
3.33
2.5
2 2
0.7 0.5
0
0 0.5 1 1.5 2
ω / ωo
π
Q=∞
φ 5
3.33
2.5
2
π /2
1 0.7 0.5
0 0.5 1 1.5 2
ω / ωo
1-17
H Re{F (t )} Re{u(t )} d t π Fo2
T
Ud = = , (1.26)
o d ωo
where we took advantage of the fact that the driving force and the velocity are in phase at the
resonance frequency. The quality factor can be expressed as
mωo mk
Q = = , (1.27)
d d
which is the same as or previous definition. This relates two different measures of system
damping.
The Q of a system may also be found from the so-called half-power (-3 dB) points of
the resonance curve. These correspond to the two points on the curve where the amplitude is
δ m / 2 . Forgoing the somewhat lengthy manipulations, the result is
ωo fo
Q ≈ = (1.28)
ω1 − ω 2 f 1− f 2
where f1 and f2 are the half-power frequencies. The quantity f1 - f2 = B is the system
bandwidth.
1-18
1.3 Wave propagation
x
u( x , t ) = A cos[ω ( t − )] = A cos(ω t − k x ) , (1.29)
c
where k is the so-called wave number which is related to the reciprocal of wavelength
k = 2π / λ and is introduced mainly for convenience in writing wave propagation expressions.
The phase velocity c is meant to rigorously define the velocity of the wave as the speed with
which two successive points of constant phase move past a certain point. This serves to
distinguish it from other types of velocities associated with waves, such as the group velocity. A
propagating wave may be described by several equivalent expressions. Thus,
u = A cos ( k x − ω t ) = A cos[k ( x − c t )]
= Aei ( k x − ω t ) = Aei k ( x − ct ) . (1.30)
These all may be considered expressions for a rightward propagating harmonic wave. A leftward
wave would simply be given by a sign change, as u = A cos (kx + ω t ).
The relationship between wave propagaton and standing wave vibrations in a system may
be shown by superimposing two oppositely propagating waves. Thus, using simple trigonometric
identities, it may be shown that
A A
u = cos ( k x + ω t ) + cos ( k x − ω t ) = A cos ( k x ) cos(ω t ) . (1.31)
2 2
This latter expression describes the vibrations of a system with well-defined nodes and antinodes,
as shown in Fig. 1.9 (node means a point, line or surface of a vibrating body that is free from
vibratory motion).
1-19
u t = t1
A t2
t3
-A
λ
Several other features of simple waves may be easily shown. For example, the
propagation of a pulse of arbitrary shape is shown in Fig. 1.9a. This may be described
mathematically as
u = f x−ct0 5 (1.32)
where f (x) is an arbitrary function describing the pulse shape. An oscillatory wave packet, such
as shown in Fig. 1.10b, may be described by
u = f ( x − c t )cos[k ( x − c t )] (1.33)
Another simple situation involves attenuation of a harmonic wave with distance, possibly
due to internal frictional losses or wave scattering. This may be described by
1-20
a)
u c
f(x-ct) f ( x - c [ t + dt ] )
b)
u
c
f(x-ct)
cos [ k ( x - c t ) ]
Figure 1.10 Propagation of (a) an arbitrary pulse and (b) a harmonic wave packet.
1-21
1.4 Longitudinal wave propagation in thin rods
We now will investigate the development of a very basic equation governing the
propagation of waves in many mechanical and electrical systems. In other words, any number of
systems could be used as the starting point for derivation of this equation. The case of a thin
elastic rod will be used here, as shown in Fig. 1.11.
a)
u
x dx
b)
σ + ∂σ
∂
σ x dx
dx
Figure 1.11 A thin rod (a) with coordinate x and displacement u of a section and (b)
the stress acting on a differential element of the rod.
∂σ ∂2 u
− σ A + (σ + dx ) A = ρ A dx , (1.35)
∂x ∂t 2
where A is the cross-sectional area and ρ is the mass density. This reduces to
1-22
∂σ ∂2 u
= ρ 2 . (1.36)
∂x ∂t
σ = Eε (1.37)
∂u
ε = . (1.38)
∂x
(In elementary mechanics, the strain is defined by ε = ∆" / " . In the present development,
∆ u ≈ ∆ " , ∆ x ≈ ", leading to ε = ∆ u / ∆ x . Taking this to differential form would be
ε = d u / d x . Since u = u (x, t), this becomes the partial derivative, ε = ∂ u / ∂ x .)
Substitution of (1.38) in (1.37) and this in (1.36) gives the following equation for waves
in thin rod
∂2 u ∂2 u
E = ρ , (1.39)
∂x2 ∂t 2
where E denotes Young's modulus. This may be put in the more usual form,
∂ 2u 1 ∂ 2u E
= , co = . (1.40)
∂x2 c o2 ∂ t 2 ρ
It may be shown that a propagating harmonic wave represents a solution of the wave
equation. Thus, substitution of
1-23
u = A sin[k ( x − ct )] (1.41)
into (1.40) leads to c = co. In other words, the propagation velocity of the wave must be co. It
may also be shown that the arbitrary pulse from f (x-c o t ) also satisfies the wave equation
u ( x , t ) = u ( x ) sin(ω t ) . (1.42)
∂ 2u ω
+ k2 u = 0 , k= . (1.43)
∂x 2 co
To determine C and D a statement of the conditions (or boundary conditions) at the ends of the
rod must be made. Suppose both ends are completely free then we would have a free-free rod:
At x = 0, " , the stresses are zero. Since σ = E ∂ u / ∂ x , this is equivalent to
∂u
= 0 at x = 0, " . (1.45)
∂x
∂u
= k ( C cos k x − D sin k x ) . (1.46)
∂x
1-24
∂u ∂u
Then, at x = 0 , = 0 gives C = 0. At x = ", = 0 gives
∂x ∂x
This leads to
or
n π co n co
ωn = , fn = (1.49)
" 2"
These are the natural frequencies of the rod, and lead to the following vibrational modes
nπ
u ( x , t ) = D cos( x ) where n = 1, 2,... (1.50)
"
The first few modes are shown in Fig. 1.12. Note that the distributed system has an infinite
number of modes.
x
u
n =1
n=2
n=3
1-25
1.6 Wave types in solids and fluids
Only a relatively few types of waves may exist in continuous infinite media. In an
extended fluid medium, only a pressure (or dilatational or longitudinal) wave may exist, while in
a solid, both pressure and shear (or transverse or equivoluminal) waves may exist. The existence
of a free surface on a solid brings in the possibility of a surface wave. The basic natures of these
waves have been previously shown in Fig. 1.2. Some additional details will now be noted.
Generally, the acoustic wave equation of Equation 1.40 can be written for any types of
polarization in the same form
∂ 2u 1 ∂ 2u
= , (1.51)
∂x2 c2 ∂ t 2
where u here is the general acoustic displacement in an arbitrary polarization direction and c
is the phase velocity that can be determined as follows
stiffness
c = . (1.52)
density
These equations govern all types of one-dimensional wave propagation. Of course, for a
thin rod aligned with the x-direction (σ y = σ z = 0 ),
σx = E εx (1.53)
and the longitudinal wave velocity is given by the previously derived result (Eq. 40)
E
crod = . (1.54)
ρ
In a thin rod, the material is free to move in the lateral direction according to the Poisson effect
as shown in Figure 1.13. In comparison, for a thin plate parallel to the x-y plane (ε y = σ z = 0 ),
the stiffness in the x-direction is increased by the Poisson effect
E
σx = εx , (1.55)
1 − ν2
1-26
where ν is Poisson's ratio, and the longitudinal velocity
E crod
c plate = = ≈ 105
. crod (for ν = 0.3) (1.56)
(1 − ν ) ρ 2
1 − ν2
For a longitudinal pressure wave propagating in an infinite medium shown in Figure 14.a
ε y = ε z = 0, therefore the Poisson effect further increases the stiffness
E E (1 − ν)
σx = ε x = εx (1.57)
2 ν2 (1 + ν)(1 − 2 ν )
1−
1− ν
E (1 − ν) (1 − ν)
cd = = crod ≈ 116
. crod (1.58)
(1 + ν)(1 − 2 ν) ρ (1 + ν)(1 − 2 ν )
is even higher.
0 5
σ x = λ + 2µ ε x = λ + 2µ 0 5 ∂∂ ux . (1.59)
where λ is a Lamé constant and µ is the shear modulus (often denoted by G). In an extended
elastic solid (a so-called "infinite medium"), the pressure wave velocity may be also expressed in
terms of the elastic constants of the medium and the density as
1-27
λ + 2µ
cd = . (1.60)
ρ
As a final note on pressure waves, the propagation velocity of acoustic waves in a gas can
be obtained from the so-called gas equation p = ρ RT . Here T is the (absolute) temperature
and R denotes the gas constant that is the ratio of the universal gas constant and the average
molecular weight. For an adiabatic process p = K ργ so that the bulk modulus
B = ρ ∂p / ∂ρ = γ K ργ . The sound velocity is given again by c d = B / ρ as follows
po γ
cd = = γ RT , (1.61)
ρ
where
σy
y -σx σx
- σy
x u
1-28
An example of transverse wave propagation is the flexural vibration propagating in a thin
rod shown in Figure 1.15. The well-known differential equation governing the bending
deformation of a thin rod is
∂4 v
IE = q, (1.62)
∂x 4
where I is the moment of inertia for the cross-sectional area, v is the transverse displacement,
and q is the distributed load intensity for a unit length. In our case, the load is entirely due to
inertia forces accelerating the beam
∂2 v
q = − ρA , (1.63)
∂t 2
I E k 4 = ρ A ω2 . (1.64)
ω I E ω2
cf = = 4 (1.65)
k ρA
1-29
proportional to the square-root of frequency. For a rectangular bar of height h, the flexural
velocity can be written as
E h2 ω 2
cf = 4 = 0.5373 crod h ω , (1.66)
12 ρ
where crod = E / ρ is the previously discussed longitudinal wave velocity in the thin rod.
Similarly, for a cylindrical rod of diameter d, I / A = d 2 / 16 and c f = 0.5 crod d ω . This
wave mode is also the limiting case of the lowest-order asymmetric Lamb mode in very thin
plates (Lamb waves are elastic waves propagating in a solid plate with free boundaries, which
will be discussed in more detailed in Chapter 5). However, for thin plates, the phase velocity can
be calculated from Equation 1.66 by substituting E / (1 − ν2 ) for E.
Another example of transverse wave propagation is the case of shear waves in infinite
solid media, which is illustrated in Fig. 1.16.
τyx
- τxy τxy
- τyx
x v
1-30
The expressions relating shear stress τ xy , shear strain γ xy and transverse particle displacement
v are
∂v
τ xy = µ γ xy , γ xy = . (1.67)
∂x
∂ 2v 1 ∂ 2v
= . (1.68)
∂x2 c s2 ∂ t 2
µ
cs = . (1.69)
ρ
It may be shown that the ratio of the two wave velocities in an isotropic solid, for which
E = 2 (1 + ν) µ , the ratio of cd and cs depends only on Poisson's ratio:
cd 2−2ν
= . (1.70)
cs 1− 2 ν
Finally, the velocity of surface waves may be noted. There is not a simple formula for this
velocity. However, an approximate expression is given by
0.87 + 112
. ν
cR ≅ cs . (1.71)
1+ ν
Acoustic Impedance
ux ( x , t ) = Aei( k x − ω t ) , (1.72)
1-31
∂ ux
v x ( x, t ) = = − i ω Aei( k x − ω t ) , (1.73)
∂t
and
∂ ux
σ x = (λ + 2 µ ) = ( λ + 2 µ ) Ai k ei ( kx − ωt ) . (1.74)
∂x
The ratio of the pressure (or negative stress) to the particle velocity is called the acoustic
impedance. For a dilatational wave propagating in the positive direction,
σ c 2 ρ Ai k ei ( kx − ωt )
Zd = − x = d = ρ cd . (1.75)
vx i ω Aei( k x − ω t )
The product of density and wave velocity occurs repeatedly in acoustics and ultrasonics and is
called the characteristic acoustic impedance (for a plane wave). As it will be seen later, it will be
the impedance that acoustically differentiates materials, in addition to the moduli and density.
The densities, velocities and acoustic impedances of a number of materials are summarized in
Table 1.1.
1-32
Table 1.1 Densities, acoustic velocities and acoustic impedances of some materials.
Nonmetals
Liquids
1-33
1.7 Wave Dispersion
Dispersion means that the propagation velocity is frequency-dependent. Since the phase
relation between the spectral components of a broadband signal varies with distance, the pulse-
shape gets distorted and generally widens as the propagation length increases. Figure 1.17 shows
schematically the distortion of a unipolar pulse caused by dispersive wave propagation.
input pulse
δc
>0
δω
δc
=0
δω
δc
<0
δω
1-34
Group Velocity
phase
velocity
group
velocity
1-35
Figure 1.19 shows the so-called beating in the case of small (4%) frequency difference
between two harmonic signals.
u1 = cos(ω1 t )
u2 = cos( ω 2 t )
ω + ω2 ω − ω2
u1 + u2 = cos(ω1 t ) + cos(ω 2 t ) = 2 cos( 1 t ) cos( 1 t) (1.76)
2 2
Figure 1.19 Beating in the case of small (4%) frequency difference between two
harmonic signals.
The velocity of the modulation on a wave is called the group velocity. For a wave with
one-dimensional modulation envelope, this is defined in the following manner. A modulated
wave is constructed by taking two waves with slightly different values of ω and k.
u( x , t ) = cos( kx − ω t ) + cos[( k + δ k ) x − (ω + δ ω ) t ]
δk δω δk δω
= 2 cos[( k + ) x − (ω + ) t ] cos( x− t ), (1.77)
2 2 2 2
1-36
where the first high-frequency term is called carrier wave and the second low-frequency term is
the modulation envelope. This shows that the propagation velocity of the carrier is the phase
velocity
δω
ω+
cp = 2 ≈ ω , (1.78)
δk k
k+
2
and the propagation velocity of the modulation envelope is the group velocity
δω ∂ω
cg = → . (1.79)
δk ∂k
The characteristic equation of a certain wave mode provides the relationship between cp
and k. Then, the group velocity can be easily calculated from cp (k) as
∂ cp
cg = c p + k . (1.80)
∂k
Very often the phase velocity is given in the form of c p (ω ) . Then, the group velocity can be
calculated as follows
cp
cg = . (1.81)
ω ∂ cp
1−
cp ∂ω
1-37
Spectral Representation
x x
f (t − ) becomes f (t − ). (1.82)
c c( ω )
Let us assume that f (t ) is known at x=0. Its Fourier transform can be written as
∞
F { f ( t )} = F ( ω ) = H dt f (t ) exp( −i ω t ) .
−∞
(1.83)
The inverse Fourier transform can be used to obtain the signal in the time domain again
1 ∞
F -1{F (ω )} = f ( t ) =
2 π −∞
H
dω F (ω ) exp( i ω t ) . (1.84)
F { f ( t − t p )} = F ( ω ) exp( − i ω t p ) , (1.85)
therefore the frequency spectrum of the signal after dispersive propagation over a distance of x
is
x
F (ω, x ) = F (ω ,0 ) exp[− i ω ] = F (ω,0) exp[− i x k (ω )] . (1.86)
c( ω )
It should be mentioned that essentially the same approach based spatial rather than temporal
frequency representation is often used in two- and three-dimensional wave propagation problems,
too.
1-38
Material versus Geometrical Dispersion
There are two main causes of dispersive wave propagation of ultrasonic fields. First,
inherent material behavior such as relaxation in polymers, which is best described by a
characteristic time constant. The degree of the dispersion is generally rather weak and the
dispersion is dependent on the ratio of this time constant to the time period of the ultrasonic
vibration. Second, geometrical effects such as in the case of dispersive Lamb waves when the
dispersion is determined by a characteristic dimension. In this case, the degree of the dispersion
can be very high and the dispersion is dependent on the ratio of this dimension (e. g., plate
thickness) to the acoustic wavelength.
As an example of inherent material dispersion, Figure 1.20 shows the sound velocity as a
function of frequency in polyethylene [M. O'Donnell et al. J. Acoust Soc. Am. 69, 696 (1981)].
2.8
Velocity [km/s]
2.7
2.6
0 2 4 6 8 10
Frequency [MHz]
1-39
To a much less degree than the flexural mode, the longitudinal mode propagating in a thin
plate also becomes dispersive as the frequency increases due to the Poisson effect. As an example
of geometrical dispersion, Figure 1.21 shows the phase and group velocities of the lowest-order
symmetric Lamb mode as functions of frequency in a thin aluminum plate.
1.5
Normalized Velocity
phase
0.5 group
0
0 1 2 3 4 5 6
Normalized Frequency
Figure 1.21 Phase and group velocities of the lowest-order symmetric Lamb mode as
functions of frequency in a thin aluminum plate (the velocities are
normalized to the shear velocity and the normalized frequency is ks d).
1-40
Part 2 MATHEMATICAL
FORMALISMS FOR
ACOUSTIC WAVE
PROPAGATION
Notation:
position vector x ( = e1x1 + e 2 x2 + e3 x3 )
displacement vector u
strain matrix ε
stress matrix τ
stiffness tensor C
2-1
ONE-DIMENSIONAL PROBLEM
b) there are three spatial coordinates but the field parameters change only in
∂ ∂
one dimension = = 0
∂x2 ∂x3
Displacement-strain relation
ξ( x ) = x + u( x )
∂u( x )
ξ( x + dx ) = x + dx + u( x + dx ) ≈ x + dx + u( x ) + dx
∂x
[ξ( x + dx ) − ξ( x ) ] − dx ∂u
ε = ≈
dx ∂x
Stress-strain relation
τ = Cε
Equation of motion
∂ 2u ρ ∂ 2u
=
∂x 2 C ∂t 2
Wave field
x C
u = f ( t − ) , where c =
c ρ
2-2
THREE-DIMENSIONAL PROBLEM
2-3
VECTOR NOTATION
∂ ∂ ∂
∇ ≡ e1 + e2 + e3
∂x1 ∂x2 ∂x3
∂ϕ ∂ϕ ∂ϕ
∇ϕ ≡ grad ϕ ≡ e1 + e2 + e3
∂x1 ∂x2 ∂x3
∂ψ1 ∂ψ 2 ∂ψ 3
∇ ⋅ ψ ≡ div ψ ≡ + +
∂x1 ∂x2 ∂x3
∂ψ 3 ∂ψ 2 ∂ψ ∂ψ 3 ∂ψ ∂ψ1
∇ × ψ ≡ curl ψ ≡ e1 ( − ) + e2 ( 1 − ) + e3 ( 2 − )
∂x2 ∂x3 ∂x3 ∂x1 ∂x1 ∂x2
Laplace operator:
∂2 ∂2 ∂2
∇2 = + +
∂x12 ∂x22 ∂x32
∂ 2ϕ ∂ 2ϕ ∂2ϕ
∇2ϕ ≡ div grad ϕ ≡ + +
∂x12 ∂x22 ∂x32
2-4
INDICIAL NOTATION
x = e1x1 + e 2 x2 + e3 x3
Free index
vector: u = e1 u1 + e 2 u2 + e3 u3 = e i ui or simply ui
∂ui
scalar = ui, j
∂x j
∂u ∂u ∂u ∂u
vector = e1 1 + e 2 2 + e3 3 = e i ui, j or simply ui, j
∂x j ∂x j ∂x j ∂x j
Gradient of a scalar
∂f ∂f ∂f
grad f = e1 + e2 + e3 = e i f,i or simply f,i
∂x1 ∂x2 ∂x3
Divergence of a vector
2-5
DIFFERENTIAL EQUATIONS IN THREE DIMENSIONS
Displacement-strain relation
vector notation ε = ∇s u
∂ui ∂u j
differential notation ε ij = ½ ( + )
∂x j ∂xi
∂u1 ∂u2
ε12 = ε 21 = ½ ( + )
∂x2 ∂x1
∂u2 ∂u3
ε 23 = ε32 = ½ ( + )
∂x3 ∂x2
∂u3 ∂u1
ε31 = ε13 = ½ ( + )
∂x1 ∂x3
vector notation τ = C :ε
2-6
SYMMETRY CONSIDERATIONS
2-7
ABBREVIATED NOTATION
τ11 !" C11 C12 C13 C14 C15 C16 !" ε11 !"
ττ22 "" CC12 C22 C23 C24 C25 C26
"" εε22 ""
τ3323 "" = C1413 C23
C24
C33 C34
C34 C44
C35 C36
C45
" 33
C46 " 2ε 23 "
"
τ31 "" C15 C25 C35 C45 C55 C56 " 2ε31 "
" "
τ12 # C16 C26 C36 C46 C56 C66 # 2ε12 #
2-8
Stress-displacement relation
∂u ∂u ∂u ∂u1
τ11 = λ ( 1 + 2 + 3 ) + 2µ
∂x1 ∂x2 ∂x3 ∂x1
∂u ∂u ∂u ∂u
τ 22 = λ ( 1 + 2 + 3 ) + 2µ 2
∂x1 ∂x2 ∂x3 ∂x2
∂u ∂u ∂u ∂u
τ33 = λ ( 1 + 2 + 3 ) + 2µ 3
∂x1 ∂x2 ∂x3 ∂x3
∂u ∂u
τ12 = τ 21 = µ ( 1 + 2 )
∂x2 ∂x1
∂u ∂u
τ 23 = τ32 = µ ( 2 + 3 )
∂x3 ∂x2
∂u ∂u
τ31 = τ13 = µ ( 3 + 1 )
∂x1 ∂x3
Balance of momentum
vector notation ∇ ⋅ τ + f = ρ
u
differential notation
∂τ11 ∂τ ∂τ ∂ 2u
+ 12 + 13 + f1 = ρ 21
∂x1 ∂x2 ∂x3 ∂t
∂τ 21 ∂τ 22 ∂τ ∂ 2u
+ + 23 + f2 = ρ 22
∂x1 ∂x2 ∂x3 ∂t
∂τ 31 ∂τ 32 ∂τ ∂ 2u
+ + 33 + f3 = ρ 23
∂x1 ∂x2 ∂x3 ∂t
2-9
THREE-DIMENSIONAL WAVE EQUATION
vector notation ∇ ⋅ C: ∇s u = ρ
u − f
( λ + µ ) ∇ ∇ ⋅ u + µ ∇2 u = ρ
u− f
2-10
PLANE WAVE SOLUTIONS, CHRISTOFFEL'S EQUATION
u = A p ei ( k x − ω t)
amplitude A
angular frequency ω
polarization unit vector p
wave vector k = dk
propagation unit vector d
wave number k = k12 + k22 + k32
ω
sound velocity c =
k
In details,
u1 = A p1 e− i ω t ei k ( d1x1 + d2 x2 + d3x3 )
u2 = A p2 e− i ω t ei k ( d1x1 + d2 x2 + d3x3 )
u3 = A p3 e− i ω t ei k ( d1x1 + d2 x2 + d3x3 )
Nontrivial (non-zero) solution requires that the determinant be zero, which provides the
characteristic equation for any propagation direction d. The solutions of this characteristic
equation are frequency-independent (nondispersive) longitudinal (dilatational) and shear waves.
2-11
Since the material is isotropic, d = e1 ( d1 = 1, d2 = d3 = 0 ) can be assumed without loss of
generality.
λ + 2µ
cd = and p2 = p3 = 0
ρ
2-12
Potential decomposition of the displacement vector, Helmoltz Theorem:
u = ∇ϕ + ∇ × ψ
∂ϕ ∂ψ 3 ∂ψ 2
u1 = + −
∂x1 ∂x2 ∂x3
∂ϕ ∂ψ 3 ∂ψ1
u2 = − +
∂x2 ∂x1 ∂x3
∂ϕ ∂ψ 2 ∂ψ1
u3 = + −
∂x3 ∂x1 ∂x2
For an isotropic solid, the stress components can be written in terms of the displacement
potentials as follows
∂ 2ϕ ∂ 2ϕ ∂ 2ϕ ∂ 2ϕ ∂2ψ 3 ∂2ψ 2
τ11 = λ ( + + ) + 2 µ ( + − )
∂x12 ∂x22 ∂x32 ∂x12 ∂x1∂x2 ∂x1∂x3
∂ 2ϕ ∂ 2ϕ ∂ 2ϕ ∂ 2ϕ ∂2ψ 3 ∂2ψ1
τ 22 = λ ( 2 + 2 + 2 ) + 2 µ ( 2 − + )
∂x1 ∂x2 ∂x3 ∂x2 ∂x1∂x2 ∂x2∂x3
∂2ϕ ∂ 2ϕ ∂ 2ϕ ∂ 2ϕ ∂2 ψ 2 ∂2ψ1
τ33 = λ ( + + ) + 2 µ ( + − )
∂x12 ∂x22 ∂x32 ∂x32 ∂x1∂x3 ∂x2∂x3
∂ 2ϕ ∂2 ψ 3 ∂2ψ 2 ∂2ψ 3 ∂2 ψ1
τ12 = τ 21 = µ ( 2 + − − + )
∂x1∂x2 ∂x22 ∂x2∂x3 ∂x12 ∂x1∂x3
∂ 2ϕ ∂2 ψ1 ∂2ψ 3 ∂2ψ1 ∂2ψ 2
τ 23 = τ32 = µ ( 2 + − − + )
∂x2∂x3 ∂x32 ∂x1∂x3 ∂x22 ∂x1∂x2
∂ 2ϕ ∂2 ψ 2 ∂2ψ 3 ∂2ψ 2 ∂2ψ1
τ31 = τ13 = µ ( 2 − + + − )
∂x1∂x3 ∂x32 ∂x2∂x3 ∂x12 ∂x1∂x2
For an isotropic material, the balance of momentum equation was previously written as (without
body forces)
(λ + µ ) ∇ ∇ ⋅ u + µ ∇2 u = ρ &&
u.
2-13
After substituting the scalar and vector potentials,
+ ρ∇ × ψ
( λ + µ ) ∇ ∇ ⋅ ∇ϕ + µ ∇2 ∇ϕ + ( λ + µ ) ∇ ∇ ⋅ ∇ × ψ + µ ∇2 ∇ × ψ = ρ ∇ϕ .
] + ∇ × [ρψ].
∇[( λ + 2µ ) ∇2ϕ ] + ∇ × [µ ∇2ψ] = ∇[ρϕ
] or
∇[( λ + 2µ ) ∇2ϕ ] = ∇[ρϕ ,
( λ + 2µ ) ∇2ϕ = ρϕ
∇ × [µ ∇2 ψ] = ∇ × [ρψ] or .
µ ∇ 2 ψ = ρψ
1 ∂2ϕ 1 ∂2ψ
∇ ϕ =
2
and ∇ ψ =
2
,
cd2 ∂t 2 cs2 ∂t 2
λ + 2µ µ
cd = and cs = .
ρ ρ
displacement u1 = i k ϕ
u2 = u3 = 0
stress τ11 = − ( λ + 2µ ) k 2 ϕ
τ 22 = τ33 = τ12 = τ 23 = τ31 = 0
2-14
DILATATIONAL WAVES
λ
In a fluid, µ = 0 ( cs = 0 ) and the compressional wave velocity is cd = .
ρ
2-15
Dilatational wave in a thin plate
ε12 = ε 23 = ε31 = 0
τ11 = τ 22 = τ33 = (3 λ + 2 µ ) ε11
p = − B ( ε11 + ε 22 + ε33 )
2
B = λ+ µ
3
2-16
Part 3 REFLECTION AND
TRANSMISSION OF
ULTRASONIC WAVES
Nearly all applications of ultrasonics involve the interaction of waves with boundaries.
Nondestructive testing, medical imaging and sonar are ready examples of this. Even basic studies
of material properties, usually involving the attenuation of waves, require in the final analysis
accounting for boundary interactions.
The simplest situation of reflection and transmission occurs when waves are impinging
normal to the surface. In Fig. 3.1, the case of a longitudinal wave incident on the interface
between two media is shown. This situation may be described mathematically in terms of three
propagating waves
ρ1 , c1
ρ2, c2
Transmission
3-1
ui = Ai sin ( k1 x − ω t ) , (3.1)
ut = At sin( k 2 x − ω t ) . (3.3)
The amplitude of the reflected and transmitted waves may be found by noting that the
displacements and stresses must be the same (continuous) at the interface. Thus, for x = 0, it is
required that
ui + ur = ut and τi + τr = τt . (3.4)
Ar ρ c 2 − ρ1 c1
Rd = = 2 (3.5)
Ai ρ1 c1 + ρ2 c 2
and
At 2 ρ1 c1
Td = = . (3.6)
Ai ρ1 c1 + ρ2 c 2
This gives the ratio of the displacement amplitude. More commonly, the stress (or pressure)
amplitudes are given. Thus,
τr ρ c − ρ1 c1
Rs = = 2 2 = Rd (3.7)
τi ρ1 c1 + ρ2 c 2
and
τt 2 ρ2 c2
Ts = = ≠ Td , (3.8)
τi ρ1 c1 + ρ2 c 2
where R and T are known as the reflection and transmission coefficients. It is seen that these
results are in terms of the respective acoustic impedances of the materials.
3-2
Illustration of the reflection and transmission at various interface combinations are worth
considering. For steel-water, we have ρs cs = 46.5 × 106 kg / m 2 s and
ρw cw = 15 . × 106 kg / m 2 s . From Eqs. 3.7 and 3.8, one obtains R = − 0.938 and
T = 0.063 . The interpretation of this result is that the amplitude of the reflected stress wave is
0.938 (or 93.8%) that of the incident amplitude. The negative sign indicates that the reflected
wave is 180o out of phase with the incident wave. Thus, when the incident wave is compressive,
the reflected wave is tensile and vice-versa. The transmitted pressure amplitude is but 6.3% of
the incident amplitude. This reflection/transmission situation is shown in Fig. 3.2a. For water-
steel, by using the previous values for ρs cs,, ρw, and cw, we obtain from Eqs. 3.7 and 3.8
R = 0.938 and T = 1938 . . Thus, the reflected wave amplitude is nearly the same as the
incident amplitude, where the transmitted (stress) amplitude is nearly twice the incident
amplitude, as shown in Fig. 3.2b.
a) steel water
pi
pt
pr
b) water steel
pi pt
pr
Figure 3.2 Sound pressure values in the case of reflection from (a) a steel-water and
(b) a water-steel interface at normal incidence.
The preceding result may appear strange, as though conservation of energy were being violated.
However, both wave amplitude and wave velocity determine the time rate of flow of energy (i. e.,
power) at the interface. In terms of power, there should be a net balance. That is,
3-3
Pr + Pt = Pi (3.9)
should be satisfied. The power per unit area (i. e., intensity) will be given by I = − τv , where
τ = C11 ∂u / ∂x and v = ∂u / ∂t . Using Eqs. 3.1-3.3 to calculate Pi, Pr, Pt and substituting in the
power balance expression shows it to be satisfied.
indicating a simple phase reversal of the incident wave. Suppose media 2 is infinitely rigid, so
that ρ2 c 2 → ∞. Then from Eqs. 3.5 through 3.8 one obtains
The case of shear waves normally incident on a boundary may also be considered.
However, a small subtlety arises. If the two media are bonded together, then conditions at the
interface would be that v i + v r = v t , τ i + τ r = τ t , and expressions of the form Eqs. 3.5 trough
3.8 would be obtained, with all velocities merely being changed to shear wave velocities.
However, the more common cases of a fluid-solid interface or of two solids separated by a thin
film of lubricant would prevent transmission of shear waves across the interface.
The case of waves normally incident on a layer sandwiched between two media is the
next step of complexity and represents a situation frequently arising in ultrasonics. Reflection at
and transmission through an elastic layer exhibit strong frequency dependence associated with
resonances in the layer. One of the simplest approach to describe this problem is applying the
impedance-translation theorem to the layer [See for example, L. M. Brekhovskikh, Waves in
Layered Media (Academic, New York, 1980) pp. 23-26]. The impedance-translation theorem
says that the input impedance Zinput of a layer can be calculated from the loading impedance
Zload presented by the medium behind the layer and the acoustic impedance Zo of the layer
itself as follows:
3-4
Zload − i Zo tan( ko d )
Zinput = Zo . (3.10)
Zo − i Zload tan( ko d )
Although this theorem is well known and widely used in several area, such as electrical
engineering, it is very instructional to derive it from the boundary conditions prevailing at the
two interfaces. Let us write the stress distribution in the layer in the following general form:
which is the sum of a forward and backward propagating plane wave. A+ and A- are the
complex amplitudes of the two waves and we omitted the common exp(-i ω t ) term. The
velocity distribution is given by
∂τ/∂x 1
v( x ) = − = − [ A+ exp( i ko x ) − A− exp( − i ko x )] . (3.12)
i ω ρo Zo
τ( 0 ) A + A−
Zinput = − = Zo + , (3.13)
v ( 0) A+ − A−
where the ratio of the complex amplitudes A+ and A− can be determined from the condition
that
τ( d ) A ei ko d + A− e− i ko d
Zload = − = Zo + i k d . (3.14)
v(d ) A+ e o − A− e− i ko d
A+ Z e− i ko d + Zo e − i ko d
= load i k d (3.15)
A− Zload e o − Zo ei ko d
3-5
The reflection coefficient of the layer can be easily obtained from (3.7) as
Zinput − Z1
R = (3.17)
Zinput + Z1
from Zload = Z2 . In the simplest case of Z2 = Z1, the reflection coefficient turns out to be
while the transmission coefficient can be calculated from the law of energy conservation as
2
T = (1 − R ) . (3.19)
From Equations 3.18 and 3.19, the moduli of the reflection and transmission coefficients
can be written as follows
ξ sin( ko d )
R = (3.20)
ξ2 sin 2 ( ko d ) + 1
and
1
T = , (3.21)
ξ sin ( ko d ) + 1
2 2
The general situation is shown in Fig. 3.2a, where repeated reflections occur within the
layer until a steady reflection, transmission state is reached. Not only do the material impedances
enter, the ratio of layer thickness to acoustic wavelength (d/ λ o ) strongly influences the result,
too. The particular cases of steel and Plexiglas plates in water are shown in Fig. 3.3b.
3-6
a)
ρ1, c1
ρo , co
ρ2, c2
Transmission
b)
1
0.8
Transmission Coefficient
Plexiglas
0.6
0.4
0.2
Steel
0
0 0.25 0.5 0.75 1 1.25
Thickness / Wavelength
Figure 3.3 (a) Schematic diagram of reflection at and transmission through a layered
medium and (b) specific cases of steel and Plexiglas plates in water.
3-7
Equation 3.20 can be used to answer one of the basic questions of ultrasonic
nondestructive evaluation concerning the reflectivity of thin cracks in solids. As an example,
Figure 3.4 shows the reflectivities of air-filled and water-filled cracks in steel as functions of the
frequency-thickness product [J. Krautkramer and H. Krautkramer, Ultrasonic Testing of
Materials (Springer, Berlin, 1977) p. 29]. For very thin cracks,
lim R = ξ ko d , (3.22)
d →0
i. e., the reflectivity is proportional to the product of impedance mismatch, frequency, and layer
thickness.
1
air gap water-filled crack
0.8
Reflection Coefficient
0.6
0.4
0.2
0
-10 -8 -6 -4 -2 0
log {Frequency x Thickness [MHz mm]}
Figure 3.4 The reflectivities of air-filled and water-filled cracks in steel as functions
of the frequency-thickness product.
3-8
One of the most important consequence of the impedance-translation theorem of Eq. 3.10
is the impedance matching capability of a single layer. When the layer thickness is an odd
multiple of the quarter-wavelength in the layer material, i. e., d = ( 2 n + 1) λ / 4 , the input and
load impedances are related through
Zo2
Zinput = . (3.23)
Zload
This means that perfect matching (total transmission and zero reflection) can be achieved even
between widely different impedances if a quarter-wavelength matching layer of Zo = Z1 Z2
acoustic impedance is applied at the interface. Let us denote the center frequency where the layer
thickness equals to one quarter-wavelength by fo. In the vicinity of this center frequency,
π fo − f
sin( ko d ) ≈ 1, and cos( ko d ) ≈ ∆ , where ∆ = , (3.24)
2 fo
r −1 r −1
R ≈ ≈ i∆ , (3.25)
2 r 2 r
r +1 − i
∆
where r = Z2 / Z1 denotes the impedance ratio between the two media to be matched. The
energy transmission coefficient through the matching layer can be approximated as
( r − 1) 2
Tenergy ≈ 1 − ∆ 2 . (3.26)
4r
Figure 3.5 shows the energy transmission coefficient through a quarter-wavelength matching
layer between quartz (typical transducer element) and water.
Of course, good matching is limited to the vicinity of the center frequency. The relative
bandwidth (inverse quality factor) can be approximated as
1 f − f1 4 2r 18
. r
= 2 ≈ ≈ , (3.27)
Q fo π ( r − 1) r −1
3-9
where f1 and f2 are the half-power (-6 dB) points. In the previously given example of quartz
coupled to water, the relative bandwidth is reasonably wide at 69 %. In the case of larger
impedance differences, the bandwidth where good transmission occurs is much lower. For
example, Figure 3.6 shows the energy transmission coefficient through a quarter-wavelength
matching layer between steel and water where the relative bandwidth is only 33 %.
It can be also seen from Equation 3.10 that whenever the layer thickness is equal to an
integer multiple of the half-wavelength, i. e., d = n λ / 2 , the input impedance is equal to the
load impedance and the presence of the layer does not affect the transmission and reflection
coefficients of the interface between the two surrounding media.
3-10
1
0.8
Energy Transmission
exact
0.6
0.4 unmatched
0.2
approximate
0
0 0.25 0.5
Thickness / Wavelength
0.8
Energy Transmission
exact
0.6
approximate
0.4
0.2 unmatched
0
0 0.25 0.5
Thickness / Wavelength
3-11
3.2 Reflection/transmission - oblique incidence
General case: In the most general case, either a longitudinal or a shear incident wave interacts
with a solid-solid interface. This situation is shown in Figure 3.7.
y
Is θsi θs1 Rs
θdi θd1
Id Rd
solid 1 z
solid 2
θs2
θd2 Td
Ts
3-12
From Snell's Law,
The particle displacement amplitudes of the incident, reflected, and transmitted longitudinal
waves are I d , Rd , and Td , respectively. Similarly, the particle displacement amplitudes of the
incident, reflected, and transmitted shear waves are I s , Rs , and Ts . Only two stress components
are relevant to the boundary conditions:
∂uz ∂u y
τ yy = λ + (λ + 2 µ ) (3.29)
∂z ∂y
and
∂u y ∂uz
τ zy = µ ( + ), (3.30)
∂z ∂y
u(y2) − u(y1) !" 0! − u(yd1) + u(yd 2) − u(ys1) + u(ys2) !" u(yi) !"
u(z2) − u(z1) "" = 0"" − uz( d1) + u(zd 2) − uz( s1) + u(zs2) "" = u(zi) "" ,
τ((yy22)) − τ((yy11)) "" 00""
or (3.31)
− τ((yydd11)) + τ((yydd 22)) − τ((yyss11)) + τ((yyss22)) "" τ((yyii)) ""
τ zy − τ zy # # − τ zy + τ zy − τ zy + τ zy # τ zy #
where the incident wave can be either longitudinal (Id = 1, Is = 0) or shear (Is = 1, Id = 0).
Equation 3.31 can be written by using the displacement amplitudes as follows
3-13
depending on whether longitudinal or shear wave incidence is considered. The matrix elements
aij, bi, and ci can be easily calculated from simple geometrical considerations:
− Zs1 cos 2θ si #
s1
cd1
di
#
The reflection and transmission coefficients can be determined by applying the well-known
Cramer's rule.
where a ( i ) is the matrix obtained by replacing the ith column of a by either b or c vectors
depending on whether longitudinal or shear incidence is used. It will be possible to give only a
few specific results, with just the general behavior outlined for most cases. Figure 3.8a-d shows
the schematic diagrams of reflection and transmission of waves for various combinations of
materials. In these figures and in the following, the first index of the reflection and transmission
coefficients indicate the type of the incident wave. For example, Rsd is the dilatational
reflection coefficient for shear wave incidence and Tdd is the dilatational transmission
coefficient for dilatational wave incidence.
3-14
a) fluid-vacuum b) fluid-fluid (cd2 > cd1)
Id Rdd Id Rdd
θi θr θi θr
fluid fluid 1
vacuum fluid 2
Tdd
θd2
c) solid-vacuum d) solid-vacuum
(longitudinal incidence) (shear incidence)
Rds Is Rss
θs θi θr
θi θr θd
Id Rdd Rsd
solid solid
vacuum vacuum
3-15
e) fluid-solid
Id Rdd
θi θr
fluid
solid
θd Tdd
θs Tds
f) solid-fluid g) solid-fluid
(longitudinal incidence) (shear incidence)
θs1 R ds Is θi θr = θs1 R ss
θi
θr = θd1 θd1
Id Rdd Rsd
solid solid
fluid fluid
θd2 θd2
Tdd Tsd
3-16
h) solid-solid i) solid-solid
(longitudinal incidence) (shear incidence)
θs1 R ds Is θi θr = θs1 R ss
θi
θr = θd1 θd1
Id Rdd Rsd
solid 1 solid 1
solid 2 solid 2
θs2 θs2
Fluid-vacuum: The case of a pressure wave incident on a fluid-vacuum interface is shown in Fig.
3.8a. This is the simplest reflection case and results in
R dd ≡ 1 , θr = θi . (3.36)
Fluid-fluid: Two fluid media in contact result in a reflected and a transmitted (or refracted) wave.
The relationships governing the angles are (Snell's law):
sin θ d 2 sin θi
θr = θi , = . (3.37)
cd 2 c d1
3-17
cd 2
sin θ d 2 = sin θi . (3.38)
c d1
and if
In Figure 3.8b, the dashed line in media 2 is at the incidence angle. When medium 2 is "slower"
than medium 1, the refracted wave is bent or steered toward the normal. When medium 2 is
"faster" than medium 1 (which is the case illustrated), the wave is bent toward the surface.
It may be also seen from (3.31) that for a given ratio of cd 2 / cd1, there will exist an
angle θic (critical angle) for which sin θ d 2 → 1, θ d 2 → 90°. Thus , the refracted wave is
parallel to the interface. For any angle beyond the critical angle, total reflection of the incident
wave occurs.
Solid-vacuum: Since the incident wave is in a solid medium, it may be either pressure (Fig. 3.8c)
or shear (Fig. 3.8d). The most remarkable feature of this and other cases involving solid media is
the phenomenon of mode conversion. What occurs is that a P-wave generates both a P-wave and
an S-wave upon reflection. Similar mode conversion can also occur in the case of an incident
shear wave. The angular relations are as follows:
sin θ s sin θi
P-wave incident: θr ( = θ d ) = θi , = (3.41)
cs cd
sin θ d sin θi
S-wave incident: θr ( = θ s ) = θi , = . (3.42)
cd cs
For the case of an incident shear wave, it may be seen that there will again exist a critical value
of the incidence angle for which sin θ d → 1, or θ d → 90° . This will inevitably occur since
3-18
c d > c s always. For an incident angle beyond the critical angle, total reflection occurs for the
shear wave and the reflected P-wave completely disappears.
We shall use this simple case as an example to demonstrate how to obtain the reflection
and transmission coefficients from the previously given general formulas of Eqs. 3.31-3.35. In
the case of a free solid surface, the boundary conditions require that both normal and transverse
stress disappear at the surface but do not put any limitation of the particle displacement. At the
same time, there are no transmitted shear or longitudinal waves to take into account at the
boundary. Consequently, the first two rows and the second and fourth columns of the a matrix
given in Eq. 3.33 can be eliminated and we get a 2-by-2 matrix. For longitudinal incidence:
By using Cramer's rule, the longitudinal-to-longitudinal reflection coefficient of the free solid
surface can be obtained as
c2
cos2 2θ s − s2 sin 2θ s sin 2θ d
cd
Rdd = − , (3.44)
cs2
cos 2θ s + 2 sin 2θ s sin 2θ d
2
cd
which depends on the Poisson ratio of the solid only and Rdd ( 0° ) = Rdd (90° ) = − 1. Naturally,
the longitudinal-to-shear as well as the shear-to-longitudinal and the shear-to-shear wave
reflection coefficients can be calculated in the same way.
Specific mathematical formulas have been obtained for the amplitudes of the reflected
waves, with the results presented in a number of ways. Basically, the reflection behavior is
dependent on incidence angle and material properties (specifically, Poisson's ratio). One
illustration of this is given in Fig. 3.9 for the case of a solid-vacuum interface. In the case of
shear wave incidence (Fig. 3.9b), the critical angle occurs where the R s s becomes -1. Beyond
the critical angle, Rss ≡ −1 and the reflected dilatational wave is evanescent. Another useful
representation of this type of data is by a polar plot shown in Fig. 3.10 for ν = 0.3.
3-19
1.2
longitudinal-to-
longitudinal
Reflection Coefficient 1
0.8
0.6
0.4
0.2 longitudinal-to-
shear
0
0 10 20 30 40 50 60 70 80 90
Angle of Incidence [deg]
1.2
shear-to-shear
1
Reflection Coefficient
0.8
0.6
0.4
0.2
shear-to-longitudinal
0
0 5 10 15 20 25 30 35
Angle of Incidence [deg]
Figure 3.9 Absolute values of the longitudinal and shear wave reflection coefficients
from a solid-vacuum interface as functions of the angle of incidence for
two different Poisson's ratios (solid lines: ν = 0.3, dashed lines: ν = 0.35).
3-20
a)
15o 0o 15o
30o 30o
o
45 45 o
60o 60o
75o 75o
90o 90o
b)
15o 0o 15o
30o 30o
o
45 45 o
60o 60o
75o 75o
90o 90o
Figure 3.10 Polar diagram of the longitudinal (solid line) and shear (dashed line) wave
reflection coefficients from a solid-vacuum interface in the case of (a)
longitudinal and (b) shear incident waves ( ν = 0.3).
3-21
Displacement, stress, intensity, and power coefficients
The reflection and transmission coefficients determined from Eq. (3.32) denote
displacement ratios (without explicitely indicating it). In many cases, it is necessary to express
the relative strength of the reflected and transmitted waves in terms of stress, intensity, or power.
The stress coefficients can be obtained from the corresponding displacement coefficients by
accounting for the impedance differences as follows:
where Γ stands for either R (j = 1) or T (j = 2), and α and β are either d or s. For
propagating modes, the intensity coefficients then can be easily calculated as a product of the
corresponding displacement and stress coefficients:
Finally, for propagating modes the power coefficients can be obtained from the corresponding
intensity coefficients by accounting for the different refraction angles as follows:
It should be mentioned that the power coefficients are identically zero for evanescent waves,
which do not carry energy away from the interface. The law of energy conservation can be
written as follows:
Rα( power
d
)
+ Rα( power
s
)
+ Tα( dpower ) + Tα( dpower ) ≡ 1. (3.48)
( power ) ( power )
Γαβ ≡ Γβα . (3.49)
3-22
Fluid-solid: This case shown in Fig. 3.8d is of great interest when the fluid is liquid. Figure
3.11 shows the energy reflection and transmission coefficients for two cases of particular
importance, i. e., for aluminum and steel immersed in water. Generally the transmission is much
higher from water into aluminum than into steel, which is caused by the more than eight times
higher density of the latter. Note that there is a wide range of incidence angle above the second
(shear) critical angle for which the water wave is completely reflected. Of course there is no
transmission into the shear component at normal incidence. The longitudinal transmission
disappears above the first critical angle which is approximately 13.5° for aluminum At this
angle, the transmitted longitudinal wave propagates along the surface. At higher angles, the
longitudinal transmitted wave is evanescent and does not carry energy away the interface.
Solid-fluid: Two cases may arise here, that of an incident P-wave, and an incident S-wave, as
shown in Figs. 3.8f and g. Again, this is of great interest when the fluid is a liquid. The specific
case of an aluminum/water interface is shown in Fig. 3.12. It is very important to realize that,
according to the Reciprocity Theorem, the energy transmission coefficients are the same for the
fluid-solid and the corresponding solid-fluid interfaces. For example, at 18° angle of incidence,
the energy transmission coefficient from water into aluminum is 45.7 %, and the refraction
angle of the shear wave is 39.4°. At the same 39.4° angle of incidence, the energy transmission
coefficient of the aluminum-water interface for an incident shear wave is also 45.7 % and the
compressional wave in the water is refracted at 18°.
Solid-solid: This situation is complicated by the fact that the two media may be solidly bonded
together, or there may be a lubricated interface. The nature of distinction between the two
situations is that for a bonded interface, both the normal and shear stresses must match at the
interface. For a lubricated, or so-called slip, interface, it is only the normal stress and normal
displacement that must match. It should be noted that the angles of reflection and transmission
are not changed by the interface condition, only the various wave amplitudes are affected. The
specific case of a Plexiglas/aluminum combination is shown in Fig. 3.12. The behavior in Fig.
3.12 is somewhat complicated because of the four possible secondary waves generated by mode
conversion. The first critical angle, θi = 25.6°, is the angle at which the transmitted (or refracted)
longitudinal wave disappears. The second critical angle, θi = 62.4°, corresponds to the angle of
total reflection of the incident wave. Above this angle, all incident energy is reflected either as a
longitudinal wave or as a mode-converted shear wave. Between 25.6° and 62.4°, one has a
method of generating (just) shear waves in steel. This technique of assuring a single transmitted
wave by working above the first critical angle is often used in everyday NDE.
3-23
a)
1
Energy Reflection and Transmission
0.8 reflection
Coefficients
0.6
longitudinal
0.4
transmission shear
transmission
0.2
0
0 5 10 15 20 25 30
Angle of Incidence [deg]
b)
1
Energy Reflection and Transmission
0.8 reflection
Coefficients
0.6
0.4
shear
longitudinal transmission
0.2 transmission
0
0 5 10 15 20 25 30
Angle of Incidence [deg]
Figure 3.11 Energy reflection and transmission coefficients for (a) aluminum and
(b) steel immersed in water.
3-24
a)
1
0.9
Energy Reflection Coefficients
0.8
0.7 shear
0.6
longitudinal
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Angle of Incidence [deg]
b)
0.6
Energy Transmission Coefficients
0.5
longitudinal shear
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
Angle of Incidence [deg]
Figure 3.12 Energy (a) reflection and (b) transmission coefficients for a
Plexiglas/aluminum interface in the case of longitudinal incidence.
3-25
Rayleigh wave
Another interesting application of Eq. (3.43) is to derive the characteristic equation of the
free vibration (as opposed to forced vibrations) of a solid-vacuum interface (free surface). The
free vibration has to satisfy the homogeneous equation of
cs2
cos 2θ s + 2 sin 2θ s sin 2θ d = 0 .
2 (3.51)
cd
Later we shall show that this is the characteristic equation of the well-known Rayleigh wave with
sound velocity cR lower than both shear and longitudinal wave velocities in the unbounded
solid:
sin θ s sin θ d 1
= = . (3.52)
cs cd cR
Relative velocities:
cs 1− 2 ν
ξ = (= ) (3.53)
cd 2(1 − ν )
cR
η = (3.54)
cs
Exact Rayleigh equation:
η6 − 8 η4 + 8 (3 − 2 ξ 2 ) η2 − 16 (1 − ξ2 ) = 0 (3.55)
3-26
Approximate expression
0.87 + 112
. ν
η ≈ (3.56)
1+ ν
Longitudinal incident wave: Suppose the P-wave is incident at 50° (θ d = θi = 50° ). Then the
reflected P-wave has an amplitude of 0.35% and negative phase (ν = 0.3). θs can be readily
calculated from
cs
sin θs = sin θi = 0.55 sin 50° = 0.41 which yields θs = 24.2°
cd
and the shear wave reflection coefficient is -1.08 (since the shear wave impedance is much lower
than the longitudinal one, the displacement amplitude of the reflected shear wave can be higher
than the amplitude of the incident longitudinal wave) It is worthwhile to mention, that the phase
of the reflected shear wave includes an arbitrary 180° component depending on the choice of the
coordinate system.
S wave incident. Suppose θs = θi = 20° again. Then the shear-to-shear reflection coefficient is
R s ≈ 0.69 and θd can be determined from
cd
sin θd = sin θs → θd = 38.5° .
cs
The shear-to-longitudinal reflection coefficient is R d ≈ − 0.53. Note that the critical angle of
incidence of the shear wave is 32.3° and beyond that angle, there is no reflected pressure wave.
3-27
Part 4 RAYLEIGH WAVES
after I. A. Viktorov, Rayleigh and Lamb Waves (Plenum, New York, 1967) pp. 1-6.
In 1885, the English scientist Lord Rayleigh demonstrated theoretically that waves can be
propagated over the free surface of an elastic half-space. The amplitude of these surface waves
decays rapidly with depth, i. e., the vibration is limited to a shallow layer of approximately one
wavelength below the surface.
x2 x1
x3
surface wave
x1
x3
4-1
Displacement potential u = ∇ϕ + ∇ × ψ (4.1)
Wave equations:
1 ∂ 2ϕ 1 ∂2 ψ
∇ ϕ =
2
= − kd2 ϕ and ∇ ψ =
2
= − ks2 ψ (4.2)
cd2 ∂t 2
cs2 ∂t 2
Wave velocities:
λ + 2µ µ
cd = and cs = (4.3)
ρ ρ
∂
u2 = 0 ⇒ = 0 , ψ1 = ψ 3 = 0 , ψ 2 = ψ (4.4)
∂x2
∂ 2ϕ ∂ 2ϕ
+ + kd2 ϕ = 0 (4.5)
∂x1
2
∂x32
∂2ψ ∂2ψ
+ + ks2 ψ = 0 (4.6)
∂x1
2
∂x3
2
∂ϕ ∂ψ
transverse particle displacement: u1 = − (4.7)
∂x1 ∂x3
∂ϕ ∂ψ
normal particle displacement: u3 = + (4.8)
∂x3 ∂x1
4-2
∂ 2ϕ ∂ 2ϕ ∂2ψ ∂ 2ϕ
τ11 = λ ( 2 + 2 ) + 2 µ ( 2 − ) (4.9)
∂x1 ∂x3 ∂x1 ∂x1∂x3
∂ 2ϕ ∂ 2ϕ ∂ 2ϕ ∂2ψ
τ33 = λ ( + ) + 2µ( + ) (4.10)
∂x12 ∂x3
2
∂x32 ∂x1∂x3
shear stress:
∂ 2ϕ ∂2 ψ ∂ 2 ψ
τ31 = τ13 = µ ( 2 + − ) (4.11)
∂x1∂x3 ∂x12 ∂x32
ϕ = F ( x3 ) ei ( k R x1 − ωt ) and ψ = G( x3 ) ei ( k R x1 − ωt ) (4.12)
∂ 2ϕ ∂ 2ϕ ∂2 F
+ + kd2 ϕ = 0 ⇒ = ( k R2 − kd2 ) F (4.13)
∂x12 ∂x32 ∂x32
2− 2 2
− x3 k R kd − x3 k R − k s2
F ( x3 ) = Ae and G( x3 ) = B e (4.15)
ϕ = Ae− κ d x3 ei ( k R x1 − ωt ) (4.16)
4-3
ψ = B e− κ s x3 ei ( k R x1 − ωt ) (4.17)
The surface wave solution is a combination of coupled partial (longitudinal and shear) waves.
The B/ A ratio is determined by the condition that the surface is traction free. Both partial waves
are evanescent, i. e., k R > ks > kd
The boundary conditions require that both normal and transverse stresses be zero on the surface
at x3 = 0
∂ 2ϕ ∂ 2ϕ ∂ 2ϕ ∂2ψ
τ33 = λ ( + ) + 2 µ ( + )
∂x12 ∂x32 ∂x32 ∂x1∂x3
= ( λ + 2µ ) κ 2d ϕ − λ k R2 ϕ − 2 µ i κ s k R ψ (4.19)
∂ 2ϕ ∂2 ψ ∂ 2 ψ
τ31 = τ13 = µ ( 2 + − )
∂x1∂x3 ∂x12 ∂x32
= − 2 µ i κ d k R ϕ − µ ( κ 2s + k R2 ) ψ (4.20)
( κ 2s + k R2 )2 − 4 κ s κ d k R2 = 0 (4.23)
4-4
Relative bulk and surface wave velocities:
cs 1− 2 ν c
ξ = (= ) and η = R (4.24)
cd 2(1 − ν ) cs
η6 − 8 η4 + 8 (3 − 2 ξ 2 ) η2 − 16 (1 − ξ2 ) = 0 (4.25)
Approximate expression
0.87 + 112
. ν
η ≈ (4.26)
1+ ν
0.96
0.95
Rayleigh Velocity / Shear Velocity
0.94
0.93
0.92
0.91
0.9
exact
0.89
approximation
0.88
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Shear Velocity / Longitudinal Velocity
B κ 2 + k R2 2iκ k
= s = − 2 d 2R . (4.27)
A 2 i κ s kR κ s + kR
4-5
By substituting the Rayleigh wave number from the characteristic equation, we get a simplified
ratio
B κd
= −i . (4.28)
A κs
The displacement potentials (without the common ei ( k R x1 − ωt ) term) can be written as follows:
ϕ = A e− κ d x3 and ψ = − i ζ A e− κ s x3 , (4.29)
where A is the arbitrary (scalar) potential amplitude and ζ = κ d / κ s denotes the magnitude
of the amplitude ratio between the vector and scalar potentials, which is a negative pure
imaginary number. Like the characteristic velocity ratio η = c R / cs = ks / k R between the
Rayleigh and shear waves, the amplitude ratio ζ is also a function of Poisson's ratio only and it
value is approximately 1.5.
∂φ ∂ψ
u1 = − = i A ( k R e − κ d x3 − ζ κ s e − κ s x3 ) (4.30)
∂x1 ∂x3
and
∂φ ∂ψ
u3 = + = − A ( κ d e− κ d x3 − ζ k R e− κ s x3 ). (4.31)
∂x3 ∂x1
The 90° phase difference between the normal and tangential displacements indicates that the
particle trajectories are ellipses. The aspect ratio of the elliptical particle trajectory changes with
depth. Interestingly, the ratio of the normal and transverse displacement components on the
surface turns out to be the same as the amplitude ratio between the vector and scalar potentials:
4-6
u3 κ − ζ kR
= i d = − iζ . (4.32)
u1 x = 0 kR − ζ κ s
3
Even more surprisingly, the same ratio also shows up between the shear τ13 and normal τ33
τ13 = − i A µ 2 k R κ d ( e− κ d x3 − e− κ s x3 ) (4.33)
and
τ33 = A µ ( k R2 + κ 2s )( e− κ d x3 − e− κ s x3 ). (4.34)
Both τ13 and τ33 have contributions from both longitudinal and shear partial wave
both stresses the amplitudes of the longitudinal and shear components must be equal so that the
combined stress could be zero on the free surface. As a result, they must have identical functional
dependence on x3 , i.e., in any plane parallel to the surface the ratio between the shear and
normal stresses is constant. What is more surprising that this ratio turns out to be again the same
τ13 i2k κ
= − 2 R d2 = − i ζ . (4.35)
τ33 kR + κ s
4-7
1.2
ν = 0.3
0.8
0.6
0.4
0.2
0 transverse
-0.2
0 0.5 1 1.5 2
Depth / Rayleigh Wavelength
1.2
ν = 0.3
1
τ11
0.8
Normalized Stress
0.6
τ 13
0.4
0.2 τ 33
-0.2
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Depth / Rayleigh Wavelength
4-8
Propagating versus evanescent waves
i( k surf x1 + 2
kbulk 2
− k surf x3 ) − i ω t
u( x1, x3 ) = D e e (4.37)
cbulk
sin θbulk = (4.38)
csurf
x1
cs,d
x3
2
− k surf 2
− kbulk x3 i( k surf x1 − ω t ) i( k surf x1 − ω t )
u( x1, x3 ) = D e e = D e − x3 / δ e (4.39)
x1
x3
4-9
Part 5 LAMB WAVES
after I. A. Viktorov, Rayleigh and Lamb Waves (Plenum, New York, 1967)
Lamb waves are elastic perturbations propagating in a solid plate (or layer) with free
boundaries, for which displacements occur both in the direction of wave propagation and
perpendicularly to the plane of the plate. Examples of the resulting deformations are shown in
Figure 1. The coordinate system is shown in Figure 2.
a) symmetric
b) asymmetric
Figure 5.1 Lowest-order symmetric (a) and asymmetric (b) Lamb waves in a plate.
x3
x1
2d
x2
wave direction
5-1
The general form of the scalar potential is
where k3 = kd2 − k12 = i κ d . Similarly for the vector potential can be written as
and
Boundary Conditions
normal stress:
∂ 2ϕ ∂ 2ϕ ∂ 2ϕ ∂ 2ψ
τ33 = λ ( + ) + 2 μ ( + ) (5.6)
∂x12 ∂x32 ∂x32 ∂x1∂x3
shear stress:
5-2
∂ 2ϕ ∂ 2ψ ∂ 2ψ
τ31 = τ13 = μ (2 + − 2) (5.7)
∂x1∂x3 ∂x12 ∂x3
Let us consider only one of the boundary conditions, e. g., τ33 ( x3 = d ) = 0 , in detail.
The normal stress on the upper surface of the plate can be written with the shear and longitudinal
velocities in the plate as
cd2 ∂ 2ϕ ∂ 2ϕ ∂ 2ψ ∂ 2ϕ
τ33 / μ = ( + ) + 2 ( − ) . (5.8)
cs2 ∂x12 ∂x32 ∂x1∂x3 ∂x12
The required partial derivatives are the following (without the common exp[i k x1 ] term):
∂ 2ϕ
= − k 2 [ As cosh( κ d x3 ) + Aa sinh( κ d x3 )] (5.9)
∂x12
∂ 2ϕ
= κ 2d [ As cosh( κ d x3 ) + Aa sinh( κ d x3 )] (5.10)
∂x32
∂ 2ψ
= i k κ s [ Bs cosh( κ s x3 ) + Ba sinh( κ s x3 )] (5.11)
∂x1 ∂x3
The other three boundary conditions can be expressed in the same way to obtain the following
system of linear homogeneous equations for As , Aa , Bs , and Ba .
⎡ As ⎤ ⎡0⎤
⎢A ⎥ ⎢0⎥
a ⎢ a⎥
= ⎢ ⎥ , (5.13)
⎢ Bs ⎥ ⎢0⎥
⎢ ⎥ ⎢ ⎥
⎣ Ba ⎦ ⎣0⎦
5-3
where matrix a can be given as
(5.14)
It is readily seen that the above system is satisfied if the following to subsystems are satisfied
⎡ ( k 2 + κ 2s ) cosh( κ d d ) 2 i k κ s cosh( κ s d ) ⎤ ⎡ As ⎤ ⎡ 0⎤
⎢ ⎥ ⎢ B ⎥ = ⎢ 0⎥ (5.15)
⎢⎣ 2 i k κ d sinh( κ d d ) −( k 2 + κ 2s ) sinh( κ s d ) ⎥⎦ ⎣ s⎦ ⎣ ⎦
and
⎡ ( k 2 + κ 2s )sinh( κ d d ) 2 i k κ s sinh( κ s d ) ⎤ ⎡ Aa ⎤ ⎡ 0⎤
⎢ ⎥ ⎢ B ⎥ = ⎢ 0⎥ . (5.16)
⎢⎣ 2 i k κ d cosh( κ d d ) −( k 2 + κ 2s ) cosh( κ s d ) ⎦⎥ ⎣ a⎦ ⎣ ⎦
The subsystems have nontrivial solutions only when their determinants are equal to zero. This
leads to the following two characteristic equations, determining the eigenvalues of the wave
number k.
for asymmetric modes. These characteristic equations (Eqs. 5.17 and 18) give the wave number
(or phase velocity) as a function of frequency for the symmetric and asymmetric families of
Lamb modes, respectively. The group velocity can be calculated from the following definition
5-4
∂ω
cg = . (5.19)
∂k
The characteristic equation of a certain wave mode provides the relationship between cp and k.
then the group velocity can be easily calculated from cp ( k ) as
∂ cp
cg = c p + k . (5.20)
∂k
Very often the phase velocity is given in the form of cp ( ω) . Then, the group velocity can be
calculated as follows
cp
cg = . (5.21)
ω ∂ cp
1−
cp ∂ ω
Equations 5.17 and 18 give the wave number as a function of frequency for different Lamb
modes. At the same time, the boundary conditions (Eqs. 5.13 and 14) determine the
As / Bs and Aa / Ba ratios, i. e., the relative part of the longitudinal and shear partial waves in
each Lamb mode (eigenvector). For a symmetric Lamb mode, the scalar and vector potentials
can be expressed from Eqs. 5.3 and 5.4 as follows
and
2 i k κ d sinh( κ d d )
ψ = As sinh( κ s x3 ) exp(i k x1 ) . (5.23)
( k 2 + κ 2s ) sinh( κ s d )
Similarly, for an asymmetric Lamb mode, the scalar and vector potentials are
5-5
and
2 i k κ d cosh( κ d d )
ψ = Aa cosh( κ s x3 ) exp(i k x1 ) . (5.25)
(k 2 + κ 2s ) cosh( κ s d )
Finally, both displacement and stress components can be readily calculated from the following
well-known relationships:
∂ϕ ∂ψ
u1 = − , (5.26)
∂x1 ∂x3
∂ϕ ∂ψ
u3 = + , (5.27)
∂x3 ∂x1
∂ 2ϕ ∂ 2ϕ ∂ 2ϕ ∂ 2ψ
τ11 = λ ( + ) + 2 μ ( − ) , (5.28)
∂x12 ∂x32 ∂x12 ∂x1∂x3
∂ 2ϕ ∂ 2ϕ ∂ 2ϕ ∂ 2ψ
τ33 = λ ( + ) + 2 μ ( + ) , (5.29)
∂x12 ∂x32 ∂x32 ∂x1∂x3
∂ 2ϕ ∂ 2ψ ∂ 2ψ
τ31 = τ13 = μ (2 + − 2) . (5.30)
∂x1∂x3 ∂x12 ∂x3
5-6
Dispersion Curves for Phase and Group Velocities
6
a1 s1 s2 a2
Normalized Phase Velocity
3
s0
2
1 a0
0
0 1 2 3 4 5 6
Normalized Frequency
2
s0
Normalized Group Velocity
s1
a1
1 a0
0
0 1 2 3 4 5 6
Normalized Frequency
Figure 5.3 Normalized (a) phase and (b) group velocity of the lowest-order Lamb
modes in an aluminum plate (the velocities are normalized to the shear
velocity and the normalized frequency is ks d ).
5-7
Asymptotic Behavior
E
lim c p = and cg = c p . (5.31)
ω→ 0 ρ (1 − ν2 )
E
lim c p = 4 ωd and cg = 2 c p . (5.32)
ω→ 0 3ρ (1 − ν2 )
At high frequencies, the velocities of both zero-order modes approach the Rayleigh velocity. All
other modes appear at a certain cut-off frequency and approach the shear velocity at very high
frequencies.
Cut-Off Frequencies
λ
For symmetric modes, 2 d = (2 n − 1) d or 2 d = n λ s , where n = 1, 2, ... (5.33)
2
λ
For asymmetric modes, 2 d = (2 n − 1) s or 2 d = n λ d , where n = 1, 2, ... (5.34)
2
For symmetric modes, the axial or longitudinal displacement (parallel to the plane of the plate) is
an even function of the through-thickness coordinate while the lateral or transverse (normal to
the plane of the plate) displacement is an odd function:
5-8
Part 6 ACOUSTIC RADIATORS
Spherical Waves
A radiation source whose nature is completely opposite to the infinite plane radiator is the
point source, capable of emitting spherical waves. Consideration of this type of source is useful
because superposition of such source permits finite sized radiators to be analyzed. The procedure
is to consider the acoustical problem of waves from an oscillating spherical cavity of radius a.
The velocity amplitude of the cavity oscillation is vo. The situation is shown in Figure 1a. With
the assumption that the cavity radius is small with respect to the acoustic wavelength (a<<λ)
and that the radius, or the distance of the point of observation from the source, is large relative to
the wavelength (r>>λ), the cavity shrinks to a point source, as shown in Figure 1b.
a) b)
vo r
Figure 6.1 Spherical waves from (a) an oscillating cavity and (b) from an equivalent
point source.
Both pressure and velocity distributions have to be of the form of a spherical wave:
ei ( k r − ω t )
p( r ) = po (6.1)
r
6-1
and
ei ( k r − ω t )
v (r ) = v o . (6.2)
r
∂p ∂v
=−ρ , (6.3)
∂r ∂t
where
∂p 1 ei ( k r − ω t )
= (i k − ) po (6.4)
∂r r r
and
∂v ei ( k r − ω t )
= − i ω vo . (6.5)
∂t r
By substituting Eqs. 6.4 and 6.5 into Eq. 6.3, the specific acoustic impedance of a spherical wave
can be written as follows
p(r ) p iωρ
Zsph = = o = . (6.6)
v (r ) vo 1
ik −
r
At large distance from the source, the specific acoustic impedance approaches that of a plane
wave:
while near to the source, it approaches the pure imaginary impedance of a vibrating mass:
6-2
The acoustic field radiated by a vibrating cavity of very small radius (a << λ ) can be
calculated from the known velocity va of the radiator at its surface.
ei ( k a − ω t )
p( a) = po = − i ω ρ a v a e−i ω t , (6.9)
a
po = − i ω ρ a2 v a e −i k a , (6.10)
or approximately
po ≈ − i ω ρ a 2 v a , (6.11)
since ka << 1. This result is usually expressed with the volume velocity of the source So
So = 4 π a 2 v a (6.12)
as
ω ρ So
po ≈ . (6.13)
4πi
p( r2 ) r
= 1 ei k ( r2 − r1 ) , where r2 > r1 (6.14)
p( r1 ) r2
p( r2 ) r
= 1 (6.15)
p( r1 ) r2
6-3
Bipoles and higher-order radiators
The above result for the simple point source of radiation has far-reaching applications in
acoustic wave analysis. By combining sources of various strengths and in various geometrical
arrangements, it becomes possible to form acoustic fields of complex shape. As a start, suppose a
simple source, having a radiation field as shown in Figure 6.2a, is placed on a rigid baffle (or
reflector). The result is a hemispherical field having a field strength equivalent to a double
source, 2So, in a free field, as shown in Figure 6.2b. More complex field shapes are shown in
Figure 6.3. The geometry of the pattern is characterized by source separation distances relative to
the acoustic wavelength (such as by d / λ for a dipole).
a) b)
0°
So 2 So mirror
90°
Figure 6.2 Radiation field of a spherical radiator (a) in free field and (b) on
the surface of a rigid mirror (baffle).
The radiation field of a bipole, i. e., a pair of two identical monopoles separated by a short
distance d, can be given as follows
ei( k r1 − ω t ) ei( k r2 − ω t )
p( r ) = po + po . (6.16)
r1 r2
6-4
where θ denotes the polar angle between the point of observation and the normal of the dipole
radiator. In the slowly changing denominator of Eq. 6.16, r1 and r2 can be further
approximated simply by r to get
e i( k r − ω t )
p( r ) = 2 po D( θ ) , (6.18)
r
where
ei k sin θ d / 2 + e− i k sin θ d / 2
D( θ ) = = cos ( k sin θ d / 2 ) (6.19)
2
denotes the so-called directivity pattern or function of the bipole. Figure 6.3 shows the directivity
pattern of a bipole at different frequencies. At low frequencies, the radiation pattern is essentially
omnidirectional like that of a monopole. At higher frequencies, an increasing number of side-
lobes appear at distinct directions where there is constructive interference between the two point
sources while minima occur at directions corresponding to destructive interference.
Similar approach can be used to study the radiation field of higher-order radiators, too.
For example, the radiation field of a tripole of full diameter d (the separation between
neighboring elements is d / 2) is
e i( k r − ω t )
p( r ) = 3 po D( θ ) , (6.20)
r
where
ei k sin θ d / 2 + e− i k sin θ d / 2 + 1 1
D( θ ) = = [2 cos ( k sin θ d / 2) + 1] (6.21)
3 3
Figure 6.4 shows the directivity pattern of a tripole at different frequencies. As always, at low
frequencies, the radiation pattern is essentially omnidirectional. At higher frequencies,
6-5
d / λ = 0.2 15o 0o 15o
30o 30o
45 o 45 o
60o 60o
75o 75o
90o 90o
d
60o 60o
75o 75o
90o 90o
60o 60o
75o 75o
90o 90o
6-6
d / λ = 0.2 15o 0o 15o
30o 30o
45 o 45 o
60o 60o
75o 75o
90o 90o
d
60o 60o
75o 75o
90o 90o
60o 60o
75o 75o
90o 90o
6-7
the radiation pattern becomes more and more directional as the main lobe becomes narrower and
the side-lobes become weaker. These two main features of the directivity pattern are affected by
two different parameters. The width of the sidelobe is mainly determined by the full diameter of
the radiator (in our case it is denoted by d). The relative strength of the side-lobes is determined
by the number of individual point sources constituting the radiator. An aggregate of discrete
point sources of identical strength might have very narrow side-lobes depending on the overall
size of the radiator, but the intensity of these side-lobes lying in the diffraction directions is as
high as that of the main lobe. We shall see that in the case of continuously distributed radiators,
the radiation pattern is very weak beyond the main lobe.
Rayleigh Integral
Radiation from point sources also finds important use in Huygens' Principle. This states
that every point on a wave surface in turn acts as a simple source. This permits successive steps
of a wave front to be reconstructed from a previous state. Superposition of elementary point
sources can be used to determine the radiation field of continuous radiating surfaces of arbitrary
velocity distributions. Let us assume that a source is located in the x,y plane of a Cartesian
coordinate system. The source velocity distribution is v o ( x , y ) within the source aperture of A
and zero everywhere else, i. e., the radiator is working in a baffle. The field of a point source
acting in a baffle was shown to be
ω ρ So ei ( k r − ω t )
p(r ) = , (6.22)
2πi r
where So denotes the volume velocity of the source. An infinitesimally small elementary
radiator of dA = dx dy area in the x,y plane produces a differential pressure
ω ρ v o dA ei ( k r − ω t )
dp(r ) = . (6.23)
2πi r
By superposition, the total field of the radiator can be obtained by integration over the surface of
the radiator (without the e− i ω t term)
6-8
ωρ ei k r'
p(r ) = z
dx' dy' v o ( x' , y' ) , (6.24)
2πi A r'
where x' and y' are the coordinates of the elementary radiator and r' denotes the coordinate of
the point of observation with respect to this elementary radiator
and
r = x2 + y2 + z2 . (6.26)
Equation 24 is called the Rayleigh Integral and is widely used to study the radiation field
of distributed acoustical sources. In complicated cases, the integral is solved by numerical means.
In simple cases, the integral can be carried out analytically by using certain approximations. Two
main regions can be distinguished depending on the distance of the point of observation from the
plane of the radiator. Close to the transducer, in the so-called near-field or Fresnel region, the
radiator produces a more or less collimated beam which can be approximated as a perturbed
plane wave. Far from the transducer, in the so-called far-field or Fraunhofer region, the radiator
produces a diverging spherical wave and it can be approximated as point source with a given
radiation pattern.
Piston Source
The single most important radiation source in ultrasonics is that of the circular piston
radiator. It is representative of many ultrasonic transducers, and the concepts and results are
applicable to many more complex problems. These include such results as the far-field/near field
transition and the directivity pattern, as well as beam spreading. The original analysis of this
important problem was done by Lord Rayleigh in his famous work entitled Theory of Sound. The
coordinate system used to calculate the acoustic field of a piston radiator is shown in Figure 6.5.
6-9
y
x P(r, θ )
r'
dA
A r θ
z
Figure 6.5 Coordinate system used to calculate the acoustic field of a piston radiator
In the case of a circular piston radiator, the Rayleigh integral can be simplified to a one-
dimensional integral of r', which can be solved numerically. Figure 6 shows the acoustic field of
a circular piston radiator at a/ λ = 10. Detailed pictures of the beam profile are shown in Figure
6.7 for four different distances.
z = 20 a
z = 10a = N
z=a
6-10
z = 0.01 N = 01
.a
z = 01
. N = a
z = N = 10 a
z = 2 N = 20 a
-2 -1 0 1 2
ξ /a
Figure 6.7 Detailed pictures of the beam profile of a circular piston radiator at four
different distances for a / λ = 10.
6-11
Even in this simple case, the Rayleigh integral cannot be evaluated analytically without
certain approximations. In the far-field, the so-called Fraunhofer approximation can be used for
the distance between the elementary radiators and the observation point. In the Fraunhofer
approximation, Equation 6.25 can be written as
x y
r' = ( x − x ' )2 + ( y − y ' )2 + z 2 ≈ r − x' − y' , (6.27)
r r
and r' can be substituted into the rapidly changing exponential term in the Rayleigh integral
while in the slowly changing denominator it is further approximated simply by r:
x y
ω ρ ei k r − i k ( x' + y' )
p(r ) = z
dx' dy' v o ( x' , y' ) e r r . (6.28)
2 πi r A
It is easy to recognize that the radiation field is a spherical wave weighted by a directivity pattern
depending on the direction of observation only:
ei k r x y
p(r ) = − i ω ρ F( , ) , (6.29)
r r r
where
x y
H
x y 1 −i k ( x' + y' )
F( , ) = dx ' dy ' v o ( x ', y ' ) e r r . (6.30)
r r 2π A
so that
x y x y
F ( , ) = v~o ( k , k ) . (6.32)
r r r r
6-12
r ' ≈ r − ξ'sin θ cos α. (6.33)
Due to the axial symmetry of the problem, the two-dimensional Fourier transform simplifies to a
Hankel transform:
2π
1 a
H H H
a
v~o ( kr ) = dξ' ξ' v o ( ξ' ) dα e− i kr ξ'cos( α ) = dξ' ξ' v o ( ξ' ) J o ( kr ξ' ) , (6.34)
2π 0 0 0
where Jo is the zeroth-order Bessel function of the first kind. For even velocity distribution
within the aperture of the piston, Eq. 6.34 can be solved by using
ζ
H dζ ζ Jo (ζ)
0
= ζ J1(ζ ) (6.35)
J (k a)
v~o ( kr ) = a 2 v o 1 r . (6.36)
kr a
ei k r
p( r , θ ) = po D( θ ) , (6.37)
r
where
2 J1 ( a k sin θ )
po = − ½ i ω ρ a 2 v o and D(θ ) = . (6.38)
a k sin θ
The acoustic field is a spherical wave modified by the directivity function representing
the angular dependence. A plot of the directivity function D(ζ ) = D( a k sin θ ) is shown in
Figure 6.8. Note that the first zero at ζ ≈ 383
. defines the main lobe of the beam. The
directivity function is determined by the a k = π d / λ product, i. e., the ratio between the
transducer diameter d = 2a and the wavelength. Figure 6.9 shows a polar representation of the
directivity function at three different transducer diameter-to wavelength ratios.
6-13
1.0
0.8
0.6
2 J1 ( ζ )
0.4
ζ
0.2
-0.2
0 2 4 6 8 10 12 14 16 18 20
6-14
15o 0o 15o
30o 30o
45 o 45o
60o 60o
75o 75o
90o 90o
d / λ = 0.5
15o 0o 15o
30o 30o
45 o 45o
60o 60o
75o 75o
90o 90o
d / λ = 1.2
15o 0o 15o
30o 30o
45 o 45o
60o 60o
75o 75o
90o 90o
d / λ= 3
Figure 6.9 Polar diagrams of the directivity function of a circular piston radiator.
6-15
Note that for d/ λ = 0.5, the field is nearly hemispherical, as for a point source. For increasing
d / λ , the beam becomes more and more directive. The half-angle of divergence θd can be
calculated either from the first zero of the directivity function
a k sin θ d ≈ 383
. (6.39)
or, more realistically, from the half-intensity (-3dB) point of the beam
a k sin θ d ≈ 162.
. (6.40)
A typical immersion type ultrasonic transducer used in NDE has 1/2" diameter and works in
water at 10 MHz. This means that the ak product is as high as 250, therefore, for all practical
purposes, sin θ d ≈ θ d can be assumed. From Eq. 6.40, the half-angle of divergence is
162
. λ
θd ≈ = 0.26 , (6.41)
ak a
We now investigate the near field (or Fresnel zone) of a radiator. Exact analytical solution
of the Rayleigh integral (Eq. 6.24) is not available except along the axis (r = z ) of the radiator.
There r '2 = ξ'2 + z 2 and the pressure distribution can be calculated from
a ei k r'
p( z ) = − i ω ρ v o dξ' ξ'
z . (6.42)
0 r'
Changing the integral variable from ξ' to r ' = z 2 + ξ'2 and using ξ' dξ' = r ' dr ' , we get
the following simpler form
a2 + z 2
p( z ) = − i ω ρv o z dr' ei k r' , (6.43)
z
p( z ) = p p ei k z [1 − ei k ( a + z − z ) ] ,
2 2
(6.44)
6-16
where we denoted the "plane wave" approximation of the pressure amplitude by p p = ρ c v o
Equation 6.44 represents the complex amplitude of the pressure field along the axis of a piston
radiator. The magnitude of the pressure can be obtained from
p( z ) = p p 2 − 2 cos[ k ( a2 + z 2 − z )] . (6.45)
. We can re-write Eq. 6.45 into a simpler form by using the following trigonometric identity
α
2 − 2 cos α = 2 sin . (6.46)
2
Finally,
p( z ) = 2 p p sin[ ½ k ( a 2 + z 2 − z )] . (6.47)
A mathematical analysis of the locations of the maxima and minima along the axis for the
circular piston radiator gives the following:
4 a 2 − λ2 ( 2 n − 1)2
z max = ( n = 1, 2,...) (6.48)
4 λ ( 2 n − 1)
and
a 2 − λ2 n 2
z min = ( n = 1, 2,...) . (6.49)
2λ n
2
Figure 6.10 shows the normalized intensity distribution p( z ) / p p for a/ λ = 10 ratio
of piston radius to wavelength. The near-field/far-field transition occurs at z / a = 10. Beyond this
distance the intensity decreases as 1 / r 2 . From Eq. 6.47,
a2
lim p( r ) = p p k lim ( a 2 + r 2 − r ) = p p k , (6.50)
r→∞ r→∞ 2r
6-17
N
p( r ) ≈ p p π . (6.51)
r
It should be noted that the last maximum in Eq. 6.48 occurs at n = 1 and is usually
referred to as the near-field/far-field transition:
4 a 2 − λ2 a2
z max ( n = 1) = N = ≈ for a >> λ . (6.52)
4λ λ
Beyond this last maximum, the amplitude of the wave drops as 1 / r , and the intensity as 1/ r 2 .
This is the far-field behavior previously obtained from the Fraunhofer approximation. Thus, N
marks the transition between the near-field (Fresnel zone) and far-field (Fraunhofer region) of the
radiation.
4
Normalized Intensity
3
1/ r2
0
0 0.5 1 1.5 2
z/N
Figure 6.10 Normalized intensity distribution of a piston radiator along its axis for
a/ λ = 10 (N = 10 a).
6-18
The near-field/far-field distance N is one of the most important parameters determining
the radiated field. Above, we gave a definition for N as the distance of the farthest maximum
along the axis of the radiator and derived a simple formula for it from the exact expression for
the acoustic field. It is very instructional to consider a much simpler approach to estimate the
near-field/far-field distance from conditions on constructive interference between elementary
radiators within the aperture of the piston radiator. In the near-field, both constructive and
destructive interference can occur between different elements on the surface of the radiator since
their distances from the point of observation can differ by many wavelengths. The closest point
on the surface of the radiator from a point of observation lying on the axis is obviously the center
of the piston and this distance is denoted by r. The farthest point is one on the circumference of
the aperture and the corresponding distance is r 2 + a 2 . Consequently, the largest difference is
limited by
a2
∆r ' ≤ r2 + a2 − r ≈ . (6.53)
2r
Destructive interference starts when the difference is larger than λ / 2 (π in phase). Therefore,
from Eq. 6.53, we again get N ≈ a 2 / λ, the same as our previous result given in Eq. 6.52. In
connection with our previous example, we mentioned that a typical immersion type ultrasonic
transducer used in NDE has 1/2" diameter and works in water at 10 MHz. This means that the
near-field/far-field transition is at approximately 280 mm
6-19
diameter at the near-field/far-field transition. Beyond this point, the beam starts to diverge and
reaches its original diameter at z = 2.5 N.
2
"searchlight" -10 dB contour
model
1 simplified model
ξ θ-10 dB
0
a
-1
-2
0 1 2 3 4
z
N
6-20
Radiation Impedance and Radiated Power
The mechanical impedance presented by the total reaction force of the medium on the
vibrating aperture is called the radiation impedance Zr and it is defined as the ratio between the
total force F acting on the piston and the source velocity vo:
F = H dx dy p( x, y,0) .
A
(6.54)
F ωρ ei k r'
Zr = = z
dx dy dx' dy' z . (6.55)
vo 2πi A A r'
The fourfold integral in Eq. 6.55 can be reduced to tabulated functions of a single variable for a
circular piston of radius a with a series of mathematical manipulations. The final result is
where
2 J1( 2 k a ) 2 H1( 2 k a )
R( ka ) = 1 − and X (k a) = , (6.57)
2k a 2k a
Here J1 denotes the first-order Bessel function of the first kind and H1 is the first-order Struve
function of the first kind. Figure 6.12 shows the R( ak ) and X ( ak ) functions at small
arguments. Above ak = 10, the real part of the radiation impedance becomes simply
Rr ≈ Z∞ = ρ c a 2 π as R( a k ) approaches unity. At a k < 2 ( or a < λ / 3) , R( ak ) drops to
zero. Rr is essentially the product of the specific acoustic impedance of the medium and the
radiating area. The third term R(ak) introduces a frequency-dependent correction which is
usually negligible at the very high frequencies used in ultrasonic NDE applications.
6-21
At very low frequencies,
( a k )2 8ak
lim Zr = Z∞ [ − i ] (6.58)
ak → 0 2 3π
and the imaginary part reduces to a simple mass load presented by the evanescent vibration in the
medium
1.2
R(ak)
1
0.8
R(ak), X(ak)
0.6
0.4
0.2 X(ak)
0
0 5 10 15 20
ak
Figure 6.12 The real and imaginary parts of the normalized radiation impedance,
R( ak ), at small arguments.
8
lim Im{Zr} = − i ω ρ a3 . (6.59)
aω → 0 3
6-22
Although this case is far from typical in ultrasonic NDE, the low-frequency results are of great
importance in understanding the phenomenon of decreasing scattering from small cracks at
relatively low frequencies.
The total radiated power of a radiator can be most easily calculated from the far-field
distribution where asymptotic approximations are readily available. From Eqs. 6.37 and 38,
ei k r
lim p( r , θ ) = po D( θ ) , (6.60)
r→∞ r
where
2 J1 ( a k sin θ )
po = − ½ i ω ρ a 2 v o and D(θ ) = . (6.61)
a k sin θ
The total radiated power Pr can be obtained from integrating the far-field intensity
po2
I (θ ) = 2
D2 (θ ) (6.62)
2ρc r
π/2 π po2 π / 2
0
H
Pr = 2 π d θ r 2 sin θ I (θ ) =
ρc 0
H
d θ sin θ D 2 ( θ ) , (6.63)
which is, of course, independent of r. Rewriting Eq. 6.63 with the source velocity yields
ρ c a 4 k 2 π v o2 π / 2
Pr =
4 0
H
d θ sin θ D 2 (θ ) . (6.64)
R( ξ ) =
ξ2 π / 2
H
d θ sin θ
2 J1 ( ξ sin θ )
2
= 1−
!"
2 J1 ( 2ξ )
#
. (6.65)
2 0 ξ sin θ 2ξ
6-23
Of course, Eq. 6.65 shows that the total radiated power can be also expressed with the real part
Rr of the complex radiation impedance
1
Pr = Rr vo2 . (6.66)
2
Diffraction Correction
In practical NDE problems, it is the highest importance to properly account for the spread
of the acoustic beam in the detected signals. The loss of reflection from an ideal mirror caused by
beam spread is called the diffraction loss. Knowledge of this loss is of great importance so that
we can take it into account in the evaluation of the data by applying a diffraction correction.
According to our previous results the acoustic field radiated by a circular piston transducer
behaves more or less as a well-collimated beam up to the near-field/far-field transition and
approaches a diverging spherical wave beyond that. Consequently, the diffraction loss must be
fairly small when z << N, and must be proportional to z in the far-field.
Let us assume that two identical transducers are used in a pitch-catch mode of operation
and the distance between them is denoted by z. Naturally, the same experiment can be carried
out also in a pulse-echo operation with a normally aligned mirror, but then the distance between
the transducer and the mirror is only z / 2. The measured signal is proportional to the integrated
pressure over the receiver's aperture
Fr ( z ) = H dx dy
Ar
p( x , y , z ) , (6.67)
x y
ω ρ ei k r − i k ( x' + y' )
p(r ) = z
dx' dy' v o ( x' , y' ) e r r . (6.68)
2πi r A
Fr ( z )
DL ( z ) = (6.69)
Fr ( z = 0)
6-24
For a circular piston, the diffraction correction can be calculated from [P. H. Rogers and A. L.
Van Buren "An exact expression for the Lommel diffraction correction," J. Acoust. Soc. Am. 55,
724 (1974)]:
DL ( s) = 1 − e i 2 π / s [ J0 (2 π / s) − i J1(2 π / s )] , (6.70)
where s denotes the normalized distance z / N . Figure 6.13 shows the magnitude of the Lommel
diffraction correction for a circular piston radiator. In the far-field the diffraction correction
approaches
lim DL (s ) = − i π / s , (6.71)
s→∞
which is also plotted in Fig. 6.13 with dotted line. Roughly we can assume that there is no beam
spread at all up to the near-field/far-field transition and that the beam diverges like a spherical
wave beyond that point. This crude assumption is acceptable very close to the transducer or at
very large distances, but causes approximately 3 dB error in the vicinity of the transition region.
1
far-field asymptote
0.8
Diffraction Correction
0.6
0.4
0.2
0
0 2 4 6 8 10
z/N
Figure 6.13 Magnitude of the Lommel diffraction correction for a circular piston
radiator.
6-25
Point Source in an Infinite Solid Medium, Stoke's Problem
ud
r
F
O P
The resulting radial displacement ud at the point of observation P at large r distance can be
written in the following form
udo i kd r
ud = e . (6.72)
r
us
O r
P
F
uso i k s r
us = e (6.73)
r
6-26
Arbitrary Orientation
us θ
ud
P r O
ei k d r
ud ( r , θ ) = udo cos θ (6.74)
r
and
ei k s r
us ( r , θ ) = uso sin θ . (6.75)
r
As the above derivation illustrates, the derivation of the directivity functions of a point source in
an infinite elastic medium is very easy from simple symmetry requirements. A general solution
of the acoustic field (Green function) is much more involved [see, e. g., J. D. Achenbach, Wave
Propagation in Elastic Solids (North Holland, Amsterdam, 1984) pp. 96-100]. The far-field
asymptotic solution gives the following simple formulas for the amplitude of the longitudinal and
shear spherical waves
F F
udo = and uso = . (6.76)
4 π (λ + 2 µ ) 4πµ
6-27
Directivity Pattern of a Point Source in an Infinite Solid Medium
longitudinal
shear
Longitudinal (solid line) and shear (dashed line) wave directivity patterns of a point source acting
in an infinite isotropic solid.
6-28
Point Source in on Infinite Solid Half-Space, Lamb's Problem
Rs
θs
θ θ = θd
ud F θd
Rd
P r P r Id
solid
O vacuum O
un
F
ud ( r , θ ) u (r, θ)
= n . (6.77)
Fnormal Fradial
The normal surface displacement is the sum of the normal displacement components produced by
the incident longitudinal, reflected longitudinal and reflected shear waves:
where sin θ s = ( cs / cd )sin θ and the incident displacement contains the dependence on radius
ei k d r
ui ( r ) = udo . (6.79)
r
6-29
1
Dd (θ ) = [cos θ − Rd (θ )cos θ − Rs (θ )sin θ s ] , (6.80)
2
Naturally the shear wave radiation function can be also calculated in the same way.
θ = θs
θ θd
ud θs
P Rs
P r r Is Rd
solid F
O vacuum O
un
F
Again, the normal surface displacement is the sum of the normal displacement components
produced by the incident shear, reflected longitudinal and reflected shear waves. According to the
sign convention of Figure 3.7,
ei k s r
ui ( r ) = uso . (6.82)
r
The shear directivity function is zero both at θ = 0 and θ = π / 2 , therefore we did not use
normalization at all.
6-30
Directivity Pattern of a Normal Point Force on a Solid Half-Space
a)
15o 0o 15o
30o 30o
45 o 45 o
60o 60o
75o 75o
90o 90o
b)
15o 0o 15o
30o 30o
45 o 45 o
60o 60o
75o 75o
90o 90o
Longitudinal (a) and shear(b) directivity patterns of a normal point force acting on an aluminum
half-space [Hutchins et al., J. Acoust. Soc. Am. 70, 1362 (1981)].
6-31
Part 7 TRANSDUCERS
Electro-Mechanical Transformer
I v
V 1:n F
The total (electrical plus mechanical) power flowing into the transducer is
Ptotal = 1
2
Re{Fv *} + 1 Re{VI *}
2
= 1
2
Re{z11v v *} + 1
2
Re{z12I v *} + 1
2
Re{z21v I *} + 1
2
Re{z22I I *} (7.2)
In a reciprocal transducer, some of this total power is dissipated either as electrical 1 Re{z22I I *}
2
or mechanical 1 Re{z11v v*} loss, but the transformation itself from electrical to mechanical or
2
mechanical to electrical energy is lossless 1
2
Re{z12I v *} + 1
2
Re{z21v I *} ≡ 0 therefore
z12 = − z21 *.
Of course, there are twelve different descriptions for the same electromechanical
transformer. For example,
... etc.
7-1
Piezoelectricity
At high frequencies, it is not sufficient to excite the radiating body just at one single point
since it would deform and thus not every portion would oscillate with equal phase and amplitude.
At sufficiently high frequencies, virtually none of the component elements are really rigid,
instead all elements behave as transmission lines exhibiting certain resonances due to their
limited dimensions. In the simplest case, a piezoelectric transducer consists of a layer of
piezoelectric material with thin metal electrodes on both sides. If an alternating electrical voltage
is applied to these electrodes, the thickness of the layer will vary according to the variation of the
electrical field. It is particularly important that the piezoelectric effect is reversible and thus can
be employed also for the transduction of alternating mechanical displacements or forces into
corresponding electrical signals, i. e., for the detection of ultrasound.
The best known piezoelectric material is quartz (silicon dioxide, SiO2) which in nature is
found for example in rock crystals and ordinary sand. It is the material in which the piezoelectric
effect was discovered by J. and P. Curie in 1880. Quartz has low electrical and mechanical
losses, it is easily cut and ground, very resistant to chemical agents and can be used at high
temperatures. Several important piezoelectric materials are basically ferroelectric. A ferroelectric
material exhibits an internal dielectric moment even without an electric field applied to it. This
moment is of uniform orientation within certain local regions called domains. Since adjacent
domains are polarized randomly, the material as a whole appears unpolarized. Above a certain
temperature, the so-called Curie temperature, the material loses its ferroelectricity. From a
phenomenological point of view, ferroelectricity is very similar to ferromagnetism. Ferroelectric
materials have very high dielectric constants (electric permitivity). When polarized, ferroelectric
7-2
materials are strongly piezoelectric. The most important ferroelectric materials are barium
titanate (BaTiO2) and lead zirconate titanate (Pb(Zr,Ti)O3) called PZT. These are ceramics
manufactured from powder base materials into arbitrary shapes. Generally, the sintered material
is poled by the application of an electric DC field of about 20 kV/cm at elevated temperature.
During this process the domains with polarization in the field direction grow at the expense of
the others. By cooling down the material in the presence of the strong polarizing electrical field,
the dielectric moment is frozen into the ceramics. Poled ferroelectric materials are subject to
some aging, i. e., their polarization decreases depending on the mechanical stresses, electrical
fields, and temperatures to which they are exposed. New and promising materials for ultrasonic
transducers are piezoelectric polymers such as polyvinylidene fluoride (PVDF) which are
available in the form of thin foils.
+ + + + + + +
- - - - - - -
+ +
+
- Si
- - Si - - Si -
O
+ - + + O + + O +
- -
+ + + + + + +
- - - - - - -
7-3
Piezoelectric Equations
Below, the piezoelectric effect will be explained as employed in the most important and
most common type of ultrasonic transducer, a single plate vibrating in its thickness or expansion
mode. Let us consider a disk of thickness b which is cut in a suitable way from a piezoelectric
crystal. Its transverse dimensions are supposed to be very large compared to its thickness. This
arrangement combines the function of an electro-mechanical transformer, a mechanical resonator
and a piston radiator.
σ= F
A
V E b
D = eS . (7.5)
The inverse piezoelectric effect occurs in the plate when an electrical voltage is applied to
its electrodes producing an electrical field E. When the plate is clamped, the electrical field
produces a given normal elastic stress in the plate
σ = −eE, (7.6)
7-4
where e is the so-called piezoelectric constant. If the plate is not completely clamped, the elastic
stress σ will cause a variation of the thickness depending upon the induced stress and the
mechanical load. It is a characteristic of the reversible piezoelectric effect that the same constant
e enters both equations. For example, in x-cut quartz, e ≈ 0159 . N / Vm .
D = εS E + e S (7.7)
σ = − e E + KE S . (7.8)
Here, εS is the dielectric constant of the material under constant strain (e. g., the plate is
clamped in such a way that its thickness is forced to remain constant). Similarly, KE is the
elastic constant of the material in the case of a constant electric field (e. g., the electrodes are
short-circuited or connected to a source of constant voltage).
e
S = E (7.9)
KE
and
7-5
e2
D = (ε S + ) E. (7.10)
KE
e2
εσ = ε S + (7.11)
KE
is larger than that of the clamped plate. The electrical energy stored in the piezoelectric material
per unit volume is
e2
½ D E = ½ εS E 2
+ E2. (7.12)
2 KE
The second term on the right side of Equation 7.12 represents the interaction energy due to the
piezoelectric effect. If we divide it with the total stored energy, we obtain the so-called coupling
factor:
e2 E e2
k2 = = . (7.13)
KE D K E εσ
The coupling factor is a dimensionless combination of the piezoelectric, dielectric and elastic
properties. It can be used to re-write Equation 7.11 in the following simple form
ε S = ε σ (1 − k 2 ) (7.14)
By similar reasoning, one finds for a piezoelectric plate with open-circuit electrical
electrodes (D = 0) that
e2
σ = ( KE + ) S = KD S , (7.15)
εS
e2
k2 = (7.16)
KD εS
7-6
and
K E = K D (1 − k 2 ) (7.17)
This means than the mechanical stiffness of a piezoelectric plate decreases when the electrodes
are short-circuited or kept at a constant voltage. The mutual dependence of the electrical and
mechanical properties of a piezoelectric material (Equations 7.14 and 17) may be quite strong
since the coupling factor of some materials is as high as 0.5 or even more (for quartz, the
coupling factor is approximately 0.1).
Equations 7.7 and 8 describe the connection between electrical and mechanical quantities
in a piezoelectric material of infinite lateral dimensions. A one-dimensional elastic wave
propagating through this plate satisfies the following wave equation
∂σ ∂ 2u
= ρ 2 . (7.18)
∂x ∂t
∂σ ∂ 2u ∂E
= KE 2 − e (7.19)
∂x ∂x ∂x
∂2u ∂E ∂ 2u
KE − e = ρ (7.20)
∂x 2 ∂x ∂t 2
∂E e ∂S
− = (7.21)
∂x ε S ∂x
7-7
and
e2 ∂ 2u ∂2u
( KE + ) = ρ 2 . (7.22)
ε S ∂x 2 ∂t
e2
KE +
KD εS KE
c = = = . (7.23)
ρ ρ (1 − k 2 ) ρ
This simple result is the direct consequence of the homogeneity of the wave equation.
Since there are no piezoelectric forces generated in the bulk of the material, the sources of the
generated sound wave are localized on the faces of the plate. The tensile and compressive
stresses acting on the opposite faces of the plate are of magnitude σ e = − e E . The transmission
line equivalent circuit of a piezoelectric disk are shown in Figure 7.3.
Zo
σe σe
7-8
Z1, Z2 and Zo denote the characteristic impedance of the front and rear medium and the
piezoelectric material, respectively. In the simplest case the front and rear media are identical
(Z1 = Z2). The stress amplitudes produced by the exciting electrical pulse in the piezoelectric
material and the front medium are po and p1, respectively.
σ e ( t ) = p1( t ) − po ( t ) , (7.24)
where σ e = − eV / b . The particle velocities of both waves are obtained from the sound
pressures by dividing them with the corresponding characteristic impedances. At the surface of
the plate, they must be equal
p p
− o = 1. (7.25)
Zo Z1
Zo Z1
po ( t ) = − σe (t ) and p1( t ) = σe (t ) . (7.26)
Zo + Z1 Zo + Z1
The radiated pulse is however further complicated by the fact that the pressure waves traveling in
both directions in the plate are repeatedly reflected at the boundaries. At each reflection, their
amplitudes are reduced by the reflection coefficient
Z1 − Zo
R = , (7.27)
Z1 + Zo
which is usually negative (i. e., the characteristic impedance of the piezoelectric material is
higher than that of the medium it radiates into). Figure 7.4 shows the schematic diagram of the
reverberating acoustic field within the piezoelectric plate. The (pressure) transmission coefficient
from the plate into the surrounding medium is T = 1 + R . Thus the total pressure in the front
medium is
7-9
where τ = b / c is the one-way propagation time through the piezoelectric plate. Figure 7.4
shows the schematic diagram of multiple reflections within a piezoelectric plate.
b = cτ
T R 4 po
R 4 po T R3 po
R 4po
R3 po T R2 po
R3 po
R2 po T R po
R2 po
R po R po T po
p1
po po
Z1 Zo Z1
Z1
pr ( t ) = {σ e ( t ) − (1 − R )[σ e ( t − τ ) + R σ e ( t − 2 τ ) + R2 σ e ( t − 3 τ ) + ...]} .
Zo + Z1
(7.29)
7-10
In the case of simple harmonic excitation of
e Vo i ω t eV
σ e (t ) = − e = σ o ei ω t , where σ o = − o (7.30)
b b
Z1
pr ( t ) = σ o {1 − (1 − R ) e− i ω τ [1 + R e− i ω τ + R2 e −2 i ω τ + ...]}
Zo + Z1
Z1 (1 − R ) e− i ω τ Z1 1 − e− i ω τ
= σ o [1 − ] = σo . (7.31)
Zo + Z1 1 − R e−i ω τ Zo + Z1 1 − R e− i ω τ
σ o Z1 (1 − e− i ω τ )
pr =
( Z1 + Zo ) − ( Z1 − Zo ) e− i ω τ
1 − e−i ω τ σo
= σo = . (7.32)
Z Z
(1 − e− i ω τ ) + o (1 + e− i ω τ ) 1 − i o cot( 12 ω τ )
Z1 Z1
Figure 7.5 shows the frequency-dependence of the normalized intensity of the radiated
fieldpr2 / σ o2 for different ξ = Zo / Z1 impedance ratios. Sharp resonances occur in the transfer
function whenever f τ = n − ½ ( n = 1, 2, 3, ...) , i. e., the plate thickness b is an odd
multiple of the half-wavelength. The bandwidth of these resonance peaks is determined by the
impedance ratio between the piezoelectric material and the surrounding medium. Under the
condition of Z1 << Zo , which is often fulfilled, we obtain
2 cd Z1
B = . (7.33)
π b Zo
B 4 Z1
= , (7.34)
f1 π Zo
7-11
it becomes apparent that the relative bandwidth is determined by the relative impedance
mismatch between the two materials.
Normalized
Intensity
pr2
ξ=3
σo2
ξ = 10
ξ = 30
0
0 0.5 1 1.5 2 2.5 3
Normalized Frequency, f τ
Because of the axial resonance of the piezoelectric plate, the peak sensitivity is always
unit, but the bandwidth is proportional the acoustic impedance of the surrounding medium, i. e.,
it might be very low for low-impedance materials like water, which is often used as coupling
medium. The resonance can be eliminated or at least greatly reduced by a highly attenuating
backing of similar acoustic impedance Z2 to that of the piezoelectric material. However, in that
case, Equation 7.28 reduces to
Z1 Zo
pr ( t ) = p1( t ) + Tpo ( t − τ ) = σe (t ) − T σe (t − τ ) . (7.35)
Z1 + Zo Zo + Z2
Z1 Z1
pr ( t ) = σ o (1 − e− i ω τ ) = 2 i σ o e− i ω τ/ 2 sin( 12 ω τ ) (7.36)
Zo + Z1 Zo + Z1
7-12
In the case of perfect loading on the back side, the peaks of the transducer's response are again at
f τ = n − ½ ( n = 1, 2, 3, ...) while the bandwidth is constant at B = 1 / ( 2 τ) regardless of the
degree of acoustic mismatch between the piezoelectric element and the material the transducer
radiates into (the relative bandwidth B / f1 = 1 , i. e., 100 %). At the same time, the peak
sensitivity is only
pr 2 Z1 Z
= ≈ 2 1 . (7.37)
σo Zo + Z1 Zo
7-13
Piezoelectric Ultrasonic Transducers
Figure 7.6 shows the schematic diagram of a piezoelectric ultrasonic transducer. The
active element is a piezoelectric or ferroelectric disk. It converts electrical energy such as an
excitation pulse from a driver into ultrasonic energy. It also works in reverse and converts
ultrasonic energy into electrical energy. The most commonly used materials are polarized
ceramics which can be cut in a variety of manners to produce longitudinal or shear waves. The
backing is usually a highly attenuating, high-density material that is used to control the vibration
of the transducer by absorbing energy radiating from the back face of the piezoelectric element.
When the acoustic impedance of the backing matches that of the active element, the result will be
a heavily damped transducer that features a broad bandwidth but may be lower in sensitivity. If
there is a mismatch in acoustic impedance between the piezoelectric element and the backing,
more sound energy will be reflected forward. This results in a transducer that is narrower in
bandwidth, but has higher sensitivity.
connector
electrical housing
network
backing
electrical lead
piezoelectric
electrodes disk
matching layer &
wear plate
In the case of contact transducers, the main purpose of the face plate is to protect the
active element from the environment. In angle-beam, delay-line, and, especially, in immersion
transducers, the face plate has the additional purpose of serving as an acoustic transformer
7-14
between the high impedance of the active element and the coupling medium of lower impedance
(angle-beam transducers usually use plastic wedges, delay-line transducers have glass buffers,
while immersion transducers use ordinary water for coupling). The impedance transformation is
accomplished by selecting a matching layer that is one quarter-wavelength thick and of the
desired acoustic impedance (geometrical mean of the two impedances to be matched). Often a
simple electrical network or terminator is also used to match the electrical impedance of the
transducer to those of the driver (in transmission) and/or the amplifier (in reception).
Figure 7.7 shows the typical echo signals and frequency responses of commercial (Ultran)
immersion transducers. The first transducer (Model 1113) is a 5-MHz narrow band transducer
with 23 % relative (-6 dB) bandwidth and ≈1.4 µs (7 cycles) waveform duration. The second
transducer (Model 1626) is a 10-MHz broadband transducer with 96 % relative bandwidth and
≈0.15 µs (1.5 cycles) waveform duration.
7-15
a) Ser. No. 1113/110916
0.3 -10
0.2
Amplitude [dB]
Amplitude [V]
0.1 -20
0
-0.1 -30
-0.2
-0.3 -40
45 50 55 0 2 4 6 8 10
Time [µs] Frequency [MHz]
0.1 -20
-25
Amplitude [dB]
Amplitude [V]
-30
0 -35
-40
-45
-0.1 -50
49 50 51 0 5 10 15 20
Time [µs] Frequency [MHz]
Figure 7.7 Echo signals and frequency responses of typical (Ultran) immersion
transducers.
7-16