You are on page 1of 20

Equations of weakly nonlinear acoustics in a strongly

heterogeneous , high speed moving fluid - Paraxial


approximation in generalized ray coordinates

François Coulouvrat

Paris, February 2005 - Mumbai and Bangalore, March 2005


Chapter 1

Equations of high frequency and


weakly nonlinear acoustics in a
heterogeneous, moving fluid

1.1 Euler equations


We presently recall Euler equations satisfied by a perfect fluid, in the absence of viscosity,
heat conduction or any more complex absorption mechanism (such as molecular relaxation).
At a given point x, and at time t, we denote by ρ(x, t) the volumic mass or density ( unit
kg.m−3 ), by p(x, t) the thermodynamic pressure (unit P a), by v(x, t) the flow velocity ( unit
m.s−1 ) and by s its specific entropy (unit J.kg −1 .K −1 ).
With these notations, we can write the balance equations for :
1) Mass
∂ρ
+ ∇.(ρv) = 0. (1.1)
∂t
2) Momentum  
∂v
ρ + (v.∇)v = −∇p + ρg (1.2)
∂t
3) Energy
∂s
+ (v.∇)s = 0. (1.3)
∂t
That system of equations is to be completed by state equations characterizing the mechan-
ical behaviour of the fluid :
p = p (ρ, s) (1.4)

1.2 Acoustical perturbations and the weakly nonlinear


approximation
We now assume the flow motion can be separated into two parts : a steady-state basic flow
which is supposed to be known (indexed by 0 ), and a small acoustical perturbation (indexed
by a ).

1
ρ(x, t) = ρ0 (x) + ρa (x, t) (1.5)
v(x, t) = v0 (x) + va (x, t) (1.6)
p(x, t) = p0 (x) + pa (x, t) (1.7)
s(x, t) = s0 (x) + sa (x, t) (1.8)

Approximation 1 : Acoustical perturbations are assumed to be small.

This ”‘smallness”’ can be measured by the acoustical Mach number Ma = Ua /c0 << 1
where Ua measures the amplitude of the acoustical wave velocity, and c0 is the sound speed
defined by :
 
∂p
c0 (x) = (ρ0 , s0 ) (1.9)
∂ρ s
If the acoustical amplitude is known for the pressure amplitude Pa as it is more frequently
the case, the acoustical Mach number is given by : Ma = Pa /(ρ0 c20 ). Typical acoustical Mach
number for air shock waves such as sonic boom or high intensity ultrasounds in water is of
the order 10−3 which is indeed small. Acoustical Mach numbers of the order of 10−1 can
nevertheless be reached for High Intensity Focused Ultrasounds (HIFU) in some biomedical
devices such as lithotripters.
The equations for the basic flow are of course exactly Eqs. 1.1, 1.2, 1.3, 1.4, but now for
the quantities indexed by 0 . In particular Eq.1.2 allows to express the gravity as follows :
∇p0
g= − (v0 .∇)v0 (1.10)
ρ0
The balance and state equations for the acoustical perturbations now run :
1) Mass
 

+ v0 .∇ ρa + ρ0 ∇.va = −va .∇ρ0 − ρa ∇.v0 − ∇.(ρa va ) (1.11)
∂t
2) Momentum
 
∂ ρa
ρ0 + v0 .∇ va + ∇pa = ∇pa − ρ0 (va .∇)v0 (1.12)
∂t ρ0
∂va
−ρa − ρ0 (va .∇)va − ρa (v0 .∇)va − ρa (va .∇)v0
∂t
3) Energy  

+ v0 .∇ sa = −(va .∇)s0 − (va .∇)sa (1.13)
∂t
4) State equation

1 ∂2p
 2 
1 ∂2p
     
2 ∂p 2 ∂ p
p a − c 0 ρa = sa + ρ + ρa s a + s2 (1.14)
∂s 0 2 ∂ρ2 0 a ∂ρ∂s 0 2 ∂s2 0 a

2
The above equations have been written as follows :
- the left-hand side of each equations involves only linear terms that would remain even
if the basic flow is homogeneous with constant speed. These terms involve time and space
derivatives of acoustical perturbations only;
- the linear terms in the right-hand (two first terms for the mass equation, first line for
momentum equations, first term for the energy equation and for the state equation) are as-
sociated to the basic flow heterogeneity as they involve only space derivatives of the basic
flow;
- in the nonlinear terms in the right-hand side (last term for the mass and energy equations,
second line for the momentum equations, three last terms for the state equation) all cubic and
higher-order nonlinear terms have been omitted, as the present study is limited to weakly
nonlinear waves (small acoustical Mach number). It is noticeable that the last term in the
momentum equation involve both nonlinearities and heterogeneity.
The fact that the linear entropy term at the right-hand side of the state equation is associ-
ated to heterogeneity comes from the fact that, in the linear, homogeneous case, the entropy
perturbation, and therefore that term would be zero,

1.3 The generalized eikonal function and the high fre-


quency approximation
Now we seek the solution of the above equations Eqs. 1.2, 1.12, 1.13, 1.14 under the following
form :

ρa (x, t) = ρˆa (x, τ = t − ψ(x)) (1.15)


va (x, t) = vˆa (x, τ = t − ψ(x))
pa (x, t) = pˆa (x, τ = t − ψ(x))
sa (x, t) = sˆa (x, τ = t − ψ(x))

The new function ψ(x) is presently not determined. We call it the generalized eikonal
function. That denomination will be justified later on. It will also be shown later on how to
choose it. Presently, we simply notice it introduces an additional unknown (6 scalar fields).
The quantity τ is called the retarded time. The objective will be to make some additional
approximations to reduce that number, indeed to two only (pressure and generalized eikonal
function), the last one being independant on the frequency. Introduction of the (generalized)
eikonal function is useful in the high frequency case. Indeed, let us assume that the field is a
frequency wave pa = Pa (x)exp(−iωτ ). One has of course the following expressions :

∂f ∂ fˆ
= (1.16)
∂t ∂τ
∂ fˆ
∇f = −∇ψ + ∇fˆ
∂τ

3
For the frequency case, one has ∂ fˆ/τ = −iω fˆ so that the partial derivative with respect
to the retarded time yields a term proportional to frequency. In the high frequency limit, that
term is supposed to be dominant. This is the

Approximation 2 : High frequency assumption

according to which derivatives with respect to retarded time are much larger than those
with respect to space variables (for theˆvariables) :

∂ fˆ
∇ψ >> ∇fˆ (1.17)

∂τ
the ratio between the two being assumed of order ǫ, and also of the same order of magnitude
as nonlinear effects ǫ = O(Ma ). As a consequence, all terms at the right-hand side of Eqs.
1.2, 1.12, 1.13 turn out to be of order ǫ ; the nonlinear terms because of the smallness of
the Mach number, and the heterogeneous terms because of the high frequency approximation
(they are not proportional to frequency). According to that approximation, and now dropping
superscriptsˆto simplify notations, it is possible to get simplified balance equations (the state
equation Eq. 1.14 keeps unchanged). To do this, it must be noticed that derivatives of
acoustical terms with respect to space variables (of order ǫ) in the right-hand side (also of
order ǫ) of Eqs. 1.2, 1.12, 1.13 are indeed of order ǫ2 , and therefore can be dropped (contrarily
to derivatives with retarded time) while keeping the resulting equations consistent. Those are
as follows.
1) Mass
∂ρa ∂va ∂(ρa va )
(1 − v0 .∇ψ) − ρ0 ∇ψ. = −v0 .∇ρa − ρ0 ∇.va − va .∇ρ0 − ρa ∇.v0 + ∇ψ. (1.18)
∂τ ∂τ ∂τ
2) Momentum
∂va ∂pa ρa
ρ0 (1 − v0 .∇ψ) − ∇ψ = −ρ0 (v0 .∇)va − ∇pa + ∇p0 − ρ0 (va .∇)v0 (1.19)
∂τ ∂τ ρ0
∂va ∂va ∂va
−ρa + ρ0 (va .∇ψ) − ρa (va .∇)v0 + ρa (v0 .∇ψ)
∂τ ∂τ ∂τ
3.a) Energy
∂sa ∂sa
(1 − v0 .∇ψ) = −(v0 .∇)sa − va .∇s0 + (va .∇ψ) (1.20)
∂τ ∂τ
That last equation shows that, at first order, the acoustical fluctuations of entropy are of
order ǫ. Therefore, all terms involving acoustical entropy sa on the right-hand side turn out
to be indeed of order ǫ2 and can be neglected so that the above equation simplifies into :
3.b) Energy
∂sa
(1 − v0 .∇ψ) = −(va .∇)s0 (1.21)
∂τ
Similarly, all nonlinear terms involving entropy sa in the right-hand side of the state equa-
tion Eq.1.4 can be omitted. At first order, one has simply the usual relation pa = c20 ρa + O(ǫ).
Replacing this into the right-hand side, the state equation reduces to :

4
4) State equation
∂2p
   
∂p 1
pa − c20 ρa = sa + 4 p2a (1.22)
∂s 0 2c0 ∂ρ2 0

1.4 The usual eikonal function and the ray approxima-


tion
In Eqs. 1.18, 1.19, 1.21 and 1.22, right-hand sides are of a smaller magnitude order that
left-hand sides. Therefore, keeping only the leading terms yields :
1) Mass
1 ∂pa ∂va
2
(1 − v0 .∇ψ) − ρ0 ∇ψ. = O(ǫ) (1.23)
c0 ∂τ ∂τ
2) Momentum
∂va ∂pa
ρ0 (1 − v0 .∇ψ) − ∇ψ = O(ǫ) (1.24)
∂τ ∂τ
where again use has been made that pa = c20 ρa +O(ǫ) according to the state equation. These
two equations provide indeed both a relation between pressure and velocity. A compatibility
relation has to be satisfied by the eikonal function so that they be consistent with one another.
This is obtained by multiplying Eq. 1.23 by (1 − v0 .∇ψ) and Eq. 1.24 by .∇ψ and yields the
well-known eikonal equation :
1
2
(1 − v0 .∇ψ)2 − (∇ψ)2 = 0 (1.25)
c0
Introducing the wavefront unit normal vector n such that :

∇ψ = αn with α > 0 (1.26)


the eikonal equation reduces to :
1
(1 − αv0 .n)2 − α2 = 0 (1.27)
c20
and solutions can be expressed by ∇ψ = n/(c0 + v0 .n) or ∇ψ = −n/(c0 − v0 .n). For
subsonic basic flows, only the first solution is admissible, so that :
n
∇ψray = (1.28)
c0 + v0 .n
Replacing this into Eq.1.24 finally yields the usual impedance relation between the acousti-
cal velocity and pressure :
pa n
va = (1.29)
ρ0 c 0
Eqs. 1.25, 1.28 and 1.29 are usual ones for geometrical acoustics. Then it is possible to
derive at next order O(ǫ) the usual transport equation governing the pression amplitude in
the geometrical (ray) approximation. However, it is well-known that the eikonal equation is
not a well-posed one. It may be singular, as there can be at some points multiple solutions or

5
on the contrary zero solution. The first case corresponds to caustics, and the second case to
shadow zones. To remove that singularity, the usual way is to introduce diffraction. However,
in the present study, we propose a complementary approach, still by introducing diffraction
(as diffraction is a physical effect that is known to be dominant in some situations) but also
by modifying the eikonal equation so as to remove any singularity. That will lead to a general
paraxial scalar approximation of the equations of weakly nonlinear acoustics in a heterogeneous
and moving medium, without any assumption on the basic flow (such as weakly heterogeneous
or low Mach). Depending on the latitude left for the choice of the generalized eikonal function,
that new formulation will be shown to provide a single, nonlinear and time-domain formulation
for several different model equations such as :
- classical geometrical (ray) approximation,
- Burgers’ equation describing 1D nonlinear wave propagation (including the case of a
heterogeneous, moving medium),
- paraxial approximation (so-called KZ equation or 2D Burgers’ equation) describing dif-
fraction in a homogeneous, non-moving medium,
- generalized KZ equation for a weakly heterogeneous fluid,
- generalized nonlinear Tricomi equation for shadow zone,
- generalized KZ equation for a slowly moving medium (to be checked !)
- model equation for echo-gallery waves (to be checked !)
- weakly nonlinear ray theory (to be checked !).

6
Chapter 2

Paraxial approximation in generalized


ray coordinates

2.1 The generalized eikonal function


The eikonal equation Eq. 1.25 arises from the complete neglect of order ǫ right-hand side terms
in Eqs. 1.18, 1.19, 1.21. Therefore, within the same order of approximation, it is possible to
introduce some latitude in the choice of the generalized eikonal function, provided this one
does not deviate ”‘too much”’ from the usual eikonal function. Therefore, it will be required
that :

ψ = ψray + O(ǫ) (2.1)


or, similarly, that the generalized eikonal function does not satisfy exactly the eikonal
equation, but only approximately:

c20 (∇ψ)2
1− = O(ǫ) (2.2)
(1 − v0 .∇ψ)2
Examples of possible choices for the generalized eikonal function will be illustrated later
on.

2.2 The paraxial approximation


Now let us return to the momentum Eq. 1.19. Considering only first order left-hand side, it
shows that the acoustical velocity is related to the pressure by the relation :
∇ψ pa ∇ψ pa
va = = + O(ǫ) (2.3)
1 − v0 .∇ψ ρ0 Ω ρ0
where we have introduced the notation Ω = 1 − v0 .∇ψ. Eq. 2.3 reduces to the usual
impedance relation 1.29 when choosing the general eikonal function equal to the usual one.
We also introduce the generalized unit vector normal to the wavefront with the same definition
as in Eq. 1.26.

7
Finally, we are introducing diffraction effects by considering the component of the acousti-
cal velocity v⊥ (so-called transverse velocity) that is normal to the wavefront unit vector n,
so that :
∇ψ pa
va = + v⊥ + O(ǫ) (2.4)
Ω ρ0
with v⊥ .n = 0. According to Eq.2.3, this implies necessarily that v⊥ = O(ǫ). In usual
ray theory the transverse velocity is neglected, but it must be taken into account to recover
diffraction effects that play a major role in many circumstances already enumerated (caustics,
shadow zone, scattering...). On the contrary, the O(ǫ) component of the velocity that is tan-
gential to the vector n will remain neglected : that is the

Approximation 3 : Paraxial approximation.

2.3 The weakly nonlinear paraxial wave equation in gen-


eralized ray coordinates
As Eq. 2.4 is exact at first order, and all right-hand side terms in Eqs. 1.18 and 1.19 are of
order O(ǫ), it will not reduce the precision to replace acoustical velocity there by Eq. 2.4. We
denote by M the resulting equation for balance of mass, and by Q the one for momentum. For
similar reasons, it is possible to neglect entropy on the right-hand side of the energy balance
equation Eq. 1.21, which then reduces to:
 
∂sa ∇ψ pa
Ω =− + v⊥ .∇s0 + O(ǫ) (2.5)
∂τ Ω ρ0
Using this and derivating the state equation Eq. 1.22 with respect to retarded time τ , one
eliminates entropy within the state equation :
4) State equation

Ω ∂ 2p ∂p2a
       
∂ pa 1 ∂p ∇ψ pa
Ω ρa − 2 = 2 + v⊥ .∇s0 − 6 + O(ǫ) (2.6)
∂τ c0 c0 ∂s 0 Ω ρ0 2c0 ∂ρ2 0 ∂τ

According to the state law for the basic flow, the term ∇s0 can be replaced by
 
∂p
∇s0 = ∇p0 − c20 ∇ρ0 (2.7)
∂s 0
so that one has :

Ω ∂ 2p ∂p2a
     
∂ pa 1 ∇ψ pa 2
Ω ρa − 2 = 2 + v⊥ .(∇p0 − c0 ∇ρ0 ) − 6 + O(ǫ) (2.8)
∂τ c0 c0 Ω ρ0 2c0 ∂ρ2 0 ∂τ

That equation will allow us to eliminate density from the mass equation M .
Therefore the system reduces to the mass M and momentum Q equations and involve only
the acoustical pressure pa and the transverse velocity v⊥ . A first relation between the two is

8
obtained by computed the combination (M + Q.∇ψ/Ω)/(1ρ0 Ω) which eliminates between the
two the velocity va at the left-hand side. After some lengthy algebra undetailed here, there
remains :

c2 (∇ψ)2
 
1 ∂pa β ∂ 1
1− 0 2 p2a = − [∇.v⊥ + b0 .v⊥ ] (2.9)

2a0 .∇pa + (∇.a0 )pa + − 2 4
ρ0 c20 Ω ∂τ ρ0 c0 ∂τ Ω

where the following notations have been introduced :

ρ0 ∂ 2 p
 
β =1+ 2 (2.10)
2c0 ∂ρ2 0
∇ψ v0
a0 = + (2.11)
ρ0 Ω2 ρ0 c20 Ω
∇p0 1
b0 = 2
+ [(∇v0 ).∇.ψ − (v0 .∇)∇ψ] (2.12)
ρc Ω
 0 0 
∇p0 1
b1 = P. − [(∇ψ.∇)v0 + (v0 .∇)∇ψ] (2.13)
ρ0 c20 Ω

Here β is the usual parameter of nonlinear acoustics, equal to (γ + 1)/2 = 1.2 for air and
to 3.5 for water;
P is the projection operator in the plane normal to the wavefront unit vector n : P = I−n⊗n.
Please note that, in Eq. 2.9 the term :

c20 (∇ψ)2 ∂pa


 
1
1−
ρ0 c20 Ω2 ∂τ
comes from the left hand side of mass and momentum equations, but that, according to
the relation Eq. 2.2, it is nevertheless of order ǫ and therfore of the same order of magnitude
as the other terms. It is noted also that other terms O(ǫ) emanating from the right-hand side
of the combination turn out to be multiplied by the same quantity

c20 (∇ψ)2
 
1
1− ,
ρ0 c20 Ω2
but are therefore of order ǫ2 and hence negligible.
The second relation between the acoustical pressure pa and the transverse velocity v⊥
is simply given by the projection P.Q of the momentum equation in the transverse plane.
Obviously, the left-hand side will involve only the transverse velocity. At the right-hand side,
terms involving v⊥ when substituting Eq. 2.4 are omitted as they are of a higher order of
magnitude. This yields :
∂v⊥
ρ0 Ω = −P.∇pa + pa b1 (2.14)
∂τ
Finally, a single, scalar equation involving only pressure can be obtained by derivating Eq.
2.9 with respect to τ , and eliminating ∂v⊥ /∂τ with Eq. 2.14. This finally yields the thought

9
after equation describing weakly nonlinear paraxial approximation for the pressure field in a
moving and heterogeneous fluid and in generalized ray coordinates :

c20 (∇ψ)2 ∂ 2 pa β ∂2
   
∂pa ∂pa 1 2

2a0 . ∇ + (∇.a0 ) + 1 − − p a = (2.15)
∂τ ∂τ ρ0 c20 Ω2 ∂τ 2 ρ20 c40 ∂τ 2
       
∇.(P.∇pa ) 1 1 b0 − b1 1 b1 b0 .b1
+ (P.∇pa ). ∇ + − pa ∇. +
ρ0 Ω2 Ω ρ0 Ω ρ0 Ω2 Ω ρ0 Ω ρ0 Ω2
In the above equation, on the first line, the two first terms are associated to the usual
transport equation for geometrical acoustics, the next one measures the deviation of the gen-
eralized eikonal equation from the usual one, and the final one is the usual nonlinear quadratic
term. In the second line, all terms are associated to diffraction. The first one involves the
transverse laplacian of the pressure field in the plane tangential to the wavefront. Other terms
are additional terms appearing only due to heterogeneity.

10
Chapter 3

Applications : some particular cases

We will now show that Eq. 2.15 indeed provides a single formalism for a wide range of
particular cases. We have listed from the known literature at least 7 different particular cases
of applications, all included within the general theory previously detailed.

3.1 Geometrical acoustics


In the case where the selected general eikonal function is a usual one, and when diffraction
and nonlinear effects are discarded, Eq. 2.15 reduces to the usual transport equation for
geometrical approximation :

2a0 .∇pa + (∇.a0 )pa = 0 (3.1)


We define an acoustical ray as a line that is constantly tangent to the a0 vector. Denoting
by t the unit vector tangent to a given ray and by l the curvilinear abscissa, the transport
equation gets along a given ray :
∂pa
2 ka0 k + (∇.a0 )pa = 0 (3.2)
∂l
For a given ray, we introduce the concept of infinitesimal raytube, eg a bundle of adjacent
rays neighbouring the considered principal ray. Using the Stokes theorem, we can integrate
∇.a0 between the curvilinear abscissa l and l + dl. Doing this, as the lateral boundaries of the
ray tube area will provide a zero contribution, one gets :
Z Z Z Z Z
(∇.a0 )dV = a0 .NdS = ka0 | S(l + dl) − ka0 k S(l) (3.3)

and taking the limit dl → 0 :

1 d(kaok S)
∇.a0 = (3.4)
S dl
Consequently, Eq.3.2 reduces to the invariant :
d p 
ka0 k Spa = 0 (3.5)
dl

11
and can simply be integrated along a given ray, giving rise to the so-called Blokhintsev
invariant :
F (τ )
pa (τ, l) = p (3.6)
ka0 k S
showing that the pressure time waveform does not change along a given ray, only its
amplitude is modulated according to the ray-tube infinitesimal area. Caustics are geometrical
surfaces where this area vanishes S = 0 and hence location of points where the geometrical
acoustics gets singular.

3.2 Whitham theory : nonlinear effects along linear


rays
For sonic boom application, nonlinear effects turn out essential to take into account. Still
neglecting diffraction but now including nonlinearities yields along a ray :

∂pa 1 d(kaok S) β ∂
p2a

2 ka0 k + pa = 2 4 (3.7)
∂l S dl ρ0 c0 ∂τ
Defining :

q(τ, l)
pa (τ, l) = p (3.8)
ka0 k S
Eq 3.7 reduces to :
∂q β ∂q
= q (3.9)
∂l ρ20 c40 a0 S 1/2 ∂τ
3/2

and introducing the age variable σ along a given ray with :


Z l
β
σ= dl (3.10)
2 4 3/2 1/2
0 ρ0 c 0 a 0 S

it ultimatelty leads to the inviscid Burgers’ equation :


∂q ∂q
=q (3.11)
∂σ ∂τ
Whitham (1952) derived that equation for sonic boom application in a homogeneous
medium. Generalization to the moving and heterogeneous case come from the work of Guiraud
(1965) and Hayes, Haefeli and Kulsrud (1969). That modelling remains nowadays the basic one
for computing primary sonic boom at the ground level (Maglieri and Plotkin, 1995, Plotkin,
2001).

12
3.3 Diffraction of nonlinear waves in a homogeneous
medium : the KZ equation
Let us now consider the case of a homogeneous and nonmoving fluid, but now taking into
account diffraction. A particular exact solution of the eikonal equation is the plane wavefront
(in a particular direction chosen as the main axis x) :
x
ψ= (3.12)
c0
so that in this case one simply has :

a0 = ex /(ρ0 c0 )
Ω=1
b0 = b1 = 0
∂pa ∂pa
P.∇pa = ey + ez
∂y ∂z
and Eq. 2.15 reduces to the well-known KZ equation :

2 ∂ 2 pa β ∂2 2
 ∂ 2 pa ∂ 2 pa
− p = + (3.13)
c0 ∂τ ∂x ρ0 c40 ∂τ 2 a ∂y 2 ∂z 2
It can be made dimensionless using the following scaling :

τ̄ = ωτ
x̄ = βkM x
ȳ = y/a
z̄ = z/a
p¯a = pa /P0

where ω is a characteristic (angular) frequency, k = ω/c0 is the associated wave number,


a is a characteristic transverse dimension (such as for instance the dimension of a bounded
emittor), P0 is a characteristic amplitude (for pressure) and M = P0 /ρ0 c20 is the associated
acoustical Mach number. With this choice, one gets :

∂ 2 p¯a
 2
∂ p¯a ∂ 2 p¯a
  
∂ ∂ pa
¯
− p¯a =µ + (3.14)
∂ τ̄ ∂ x̄ ∂ τ̄ ∂ τ̄ ∂ ȳ 2 ∂ z̄ 2
where µ = L/R measures the ratio of nonlinear to diffraction effects, L = 1/(βkM ) is
the shock formation distance associated to nonlinear effects and R = 2ka2 is the Rayleigh
distance separating the nearfield to the farfield of a source of dimension a. In case there is
no geometrical transverse characteristic dimension, one simply has √ to choose µ = 1, which
implies that the transverse length scale a is to be chosen a = 1/(k βM ).
Eq. 3.13 has been obtained by Zabolotskaya and Khokhlov in 1969 for modelling finite
amplitude sound beams and is known as the KZ (or sometimes ZK) equation. In 1970,

13
Kuznetsov added viscosity effects, the resulting equation being known as KZK equation. It
has been studied extensively in the 70’s as a model for parametric arrays, and then in the
80’s for High Intensity (weakly) Focused Ultrasound (HIFU) (see the books of Novikov and
Timoshenko, 19??, and of Hamilton and Blackstock, 1998). Other applications cover Fresnel
diffraction of shock waves (Coulouvrat and Marchiano, 2003), focusing of shock waves at a
caustic cusp (Cramer and Seebass, 1978, Coulouvrat 2000, Piacsek 2002, Marchiano, Coulou-
vrat and Thomas 2005), nonregular reflexion of weak shock waves at near grazing incidence
(Hunter, 200?).
The physical meaning of the high-frequency and paraxial approximation of the KZ equa-
tion is outlined by examinating the dispersion relation associated to the linear version of it.
Returning to physical variables t and x and omitting nonlinear terms, Eq. 3.13 gets:

2 ∂ 2 pa 2 ∂ 2 pa ∂ 2 pa ∂ 2 pa
+ = + (3.15)
c20 ∂t2 c0 ∂t∂x ∂y 2 ∂z 2
The dispersion relation is sought for a plane wave under the form pa = Aexp (i (k.x − ωt))
where the wavevector is separated into its longitudinal and transverse component k = kx ex +
k⊥ , which yields :
 
ω k2⊥ 
kx = 1 −  2  (3.16)
c0
2 cω0
which indeed is the second order Taylor expansion of the exact dispersion relation of the
wave equation :
s 
2
ω
kx = ± − k2⊥ (3.17)
c0
Note that the dispersion relation of the linear KZ or paraxial equation Eq. 3.16 is simply
the osculating parabola to the exact dispersion relation Eq. 3.17 which is a circle (or a
sphere at 3D). This is why the paraxial approximation is very frequently called parabolic
approximation, though the equation is a hyperbolic PDE, not a parabolic one ! Starting from
the work of Tappert (1977) the linear parabolic/paraxial approximation has been heavily used
for simulating long-range propagation of acoustical waves in the ocean and in the atmosphere.

3.4 Diffraction of nonlinear waves in a weakly hetero-


geneous medium : the generalized KZ equation
Let us now consider the case of a weakly heterogeneous (but nonmoving) fluid. We assume the
sound speed is given by c0 (x) =< c0 > +∆c0 (x), where the actual sound speed differs from
the mean sound speed < c0 > only by a small but fluctuating quantity ∆c0 (x). In this case,
we use for the first time the latitude that the generalized eikonal function does not satisfy
exactly the usual eikonal equation. Here we choose it simply associated to a plane wavefront
in the homogeneous case :

14
x
ψ= (3.18)
< c0 >
In this case Eq. 2.15 reduces to the so-called generalized KZ equation :

2 ∂ 2 pa c20
 2
β ∂2  ∂ 2 pa ∂ 2 pa

1 ∂ pa 2
+ 2 1− − p = + (3.19)
c0 ∂τ ∂x c0 < c0 >2 ∂τ 2 ρ0 c40 ∂τ 2 a ∂y 2 ∂z 2
Here, at the first order, c0 (x) can be replaced in all terms by < c0 > except in the term
1 − c20 / < c0 >2 where the fluctuations must be taken into account. This finally leads to :

2 ∂ 2 pa 2∆c0 (x) ∂ 2 pa β ∂2 2
 ∂ 2 pa ∂ 2 pa
− − p = + (3.20)
< c0 > ∂τ ∂x < c0 >3 ∂τ 2 ρ0 < c0 >4 ∂τ 2 a ∂y 2 ∂z 2
That equation has been used by Blanc-Benon et al. (2002) to simulate numerically sonic
boom propagation near the ground (where it is relatively grazing) in the turbulent atmospheric
boundary layer for which temperature fluctuations are sufficiently small (a few degrees) to
satisfy the approximation of a weakly heterogeneous medium.

3.5 Linear sound speed profile : the generalized nonlin-


ear Tricomi equation
At two dimensions x and z, in the particular case where the sound speed is stratified, and for
propagation sufficiently near the ground so that only a linear stratification can be assumed :
c0 (x) = c0 (0) + zdc0 (0)/dz the above equation reduces to the generalized nonlinear Tricomi
equation derived by Coulouvrat (1997, 2002) for modelling sonic boom penetration into the
shadow zone (in the case dc0 (0)/dz < 0). Indeed, identifying in Eq. 3.20 c0 (0) with < c0 >
and ∆c0 (x) with zdc0 (0)/dz, one gets :

2 ∂ 2 pa 2z ∂ 2 pa β ∂2 2
 ∂ 2 pa
+ − p a = (3.21)
c0 (0) ∂τ ∂x Rc0 (0)2 ∂τ 2 ρ0 c0 (0)4 ∂τ 2 ∂z 2
with R = −c0 (0)/dc0 (0)/dz is the radius of curvature of grazing rays. The case for which
dc0 (0)/dz > 0 (never studied yet to our knowledge) would permit to investigate the case
of nonlinear guided waves (for instance in a downward refracting atmosphere because of a
temperature inversion).

3.6 Focus at fold caustics : the nonlinear Tricomi equa-


tion
In case the propagation term ∂ 2 /∂τ ∂x can be omitted in Eq. 3.21, one gets the nonlinear
Tricomi equation :

2z ∂ 2 pa β ∂2 2
 ∂ 2 pa
− p = (3.22)
Rc0 (0)2 ∂τ 2 ρ0 c0 (0)4 ∂τ 2 a ∂z 2

15
That equation has been derived by Guiraud (1965) as modelling sonic boom focusing on a
fold caustic (the simplest of the caustic in the classification of catastrophe theory, see Thom,
1972, Berry, 1976, Arnold, 19??). Other derivations in different cases were proposed by Hayes
(1968), Cole and Kevorkian (19??), Rosales and Tabak (1997) and Auger (2001).
Indeed, here, the proposed equation is directly applicable to the caustics created by re-
fraction in a stratified medium. Because of temperature stratification, rays will turn up at
some altitude (here chosen as the origin 0). Therefore the line z = 0 is a caustic. If the
caustic is created by a line source, or by a supersonic aircraft in steady flight, the problem is
indeed invariant with variable x and one gets Eq. 3.22. It is to be noticed that, contrarily
to all previous ones, that equation is not hyperbolic but of mixed elliptic / hyperbolic type.
In the linear case, it is hyperbolic for z > 0, where two rays exist, and elliptic (no rays, this
is the shadow zone of the caustic) on the opposite side. In the nonlinear case, the sonic line
separating the hyperbolic and elliptic regions is dependant on the pressure.

3.7 Other potential applications


Two other models could be related to the general theory presented here.
- To model whispering gallery waves that propagate in a homogeneous medium but along a
curved surface, Babič and Buldyrev (1991) indeed introduce a generalized eikonal function
which follows the curvature of the boundary. Therefore, it is not an exact solution of the
eikonal equation. Consequently they get a (linear) equation similar to Eq. 3.21 in which
cartesian variables x and z are replaced by s and η where s is the curvilinear abscissa along
the curved rigid boundary, and η is the orthogonal coordinate. Using these variables Eq. 3.21
generalizes this to the nonlinear case. Indeed, this is not surprizing as it is well-known that
whispering gallery waves are equivalent to guided waves in a doq wnward refracting medium.
In one case, multiple relexions associated to the waveguide are produced by the concavity of
the boundary. In the second case, it is produced by the temperature inversion.
- Weakly NonLinear Ray Theory (Prasad, 2001) introduces a nonlinear eikonal function that is
dependant on the amplitude of the wave. In this case, the deviation of the eikonal eikonal from
the usual linear one would balance the nonlinear terms of Eq. 2.15. For all other examples
presented previously, the eikonal function, though it satisfies a nonlinear equation, was chosen
independant of the acoustical pressure , and as a result the transport equation is nonlinear.
In WNLRT, the viewpoint is reversed, the eikonal equation also depends on pressure, but
as a counterpart the transport equation now is linear. That theory has the advantage to
automatically prevent the formation of caustics : as rays converge, the amplitude increases,
and therefore rays are more and more deviated. This phenomenon is known as self-defocusing
and has been observed experimentally (Sturtevant and Kulkarny, 1976) for moderately weak
shock waves (acoustical Mach number of the order 0.1). For very weak shock waves, on the
contrary, the linear pattern of caustics remains. Therefore, the present theory may unify both
approach, by introducing WNLRT choice for the eikonal function, but still keeping diffraction
effects. However, this approach still remains to be worked out.

16
Chapter 4

References

4.1 Recommended Introduction Books


BABIČ, V. M., BULDYREV, V. S. (1991), Short-Wavelength Diffraction Theory. Asymptotic
Methods, Springer-Verlag (Berlin)
HAMILTON, M. F., and BLACKSTOCK, D. T.(1998), Nonlinear acoustics, Academic Press
(San Diego)
OSTASHEV, V. E. (1997), Acoustics in moving inhomogeneous media, E & FN Spon (Lon-
don)
PIERCE, A. D. (1989), Acoustics : an introduction to its physical principles and applications,
Acoustical Society of America
PRASAD, Ph. (2001), Nonlinear hyperbolic waves in multi-dimensions, Chapman & Hall
RUDENKO, O. V., SOLUYAN, S. I. (1977), Theoretical Foundations of Nonlinear Acoustics,
Consultants Bureau, Plenum (New York)
WHITHAM, G. B. (1974), Linear and Nonlinear Waves, Wiley (New York)

4.2 Other references


AUGER, Th. (2001), ”Modélisation et simulation numérique de la focalisation d’ondes de
choc acoustiques en milieu en mouvement. Application à la focalisation du bang sonique en
accélération”, thèse de l’Université Pierre et Marie Curie (Paris 6) (in French).
AUGER, Th., COULOUVRAT, F. (2002), ”Numerical simulation of sonic boom focusing”,
AIAA J. 40, 1726-1734
BAKHVALOV, N. S., ZHILEIKIN, Ya. M., ZABOLOTSKAYA, E. A. (1987), Nonlinear The-
ory of Sound Beams, American Institute of Physics (New York)
BERRY, M. V. (1976), ”Waves and Thom’s theorem”, Adv. Phys. 25, 1-26
BLANC-BENON, Ph., LIPKENS, B., DALLOIS, L., HAMILTON, M. F., BLACKSTOCK,
D. T. (2002), ”Propagation of finite amplitude sound through turbulence: Modeling with geo-
metrical acoustics and the parabolic approximation”, J. Acoust. Soc. Am., 111, 487-498
BLOKHINTZEV, D. I. (1946), Acoustics of a Nonhomogeneous Moving Medium, Leningrad,
trans. NACA TM 1399, National Advisory Committee for Aeronautics, Washington, espe-

17
cially pp. 35-40. The propagation of Sound in an Inhomogeneous and Moving Medium, I, J.
Acoust. Soc. Am., 18, 322-328
BURGERS, J. M. (1948), ” A mathematical model illustrating the theory of turbulence ”,
Adv. Appl. Mech., 1, 171-199
COULOUVRAT, F. (1992), ” On the equations of nonlinear acoustics ”, J. Acoustique, 321-
359
COULOUVRAT, F. (1997), ”Théorie géométrique non linéaire de la diffraction en zone
d’ombre”, C. R. Acad. Sci. Paris, 325, Série IIb, 69-75 (in French, abridged english ver-
sion)
COULOUVRAT, F. (2000), ”Focusing of weak acoustic shock waves at a caustic cusp”, Wave
Motion, 32, 233-245
COULOUVRAT, F. (2002), ”Sonic boom in the shadow zone : a geometrical theory of dif-
fraction,” J. Acoust. Soc. Am. 111, 499-508.
COULOUVRAT, F. and MARCHIANO, R. (2003), ” Nonlinear Fresnel diffraction of weak
shock waves ”, J. Acoust. Soc. Am., 114, 1749-1757
CRAMER, M. S., and SEEBASS, A. R. (1978), ”Focusing of weak shock waves at an arête,”
J. Fluid. Mech., 88, 209-222
ESCLANGON, E. (1925), L’acoustique des canons et des projectiles, Imprimerie Nationale
(Paris) (in French)
GUIRAUD, J.-P. (1965), ”Acoustique géométrique, bruit balistique des avions supersoniques
et focalisation,” J. Mécanique 4, 215-267 (in French).
HAYES, W. D. (1968), ”Similarity rules for nonlinear acoustic propagation through a caustic,”
Second Conference on Sonic Boom Research, NASA SP-180, 165-171.
HAYES, W. D. (1968), ”Energy invariant for geometric acoustics in moving medium”, Phys.
Fluids, 11, 1654
HAYES, W. D., HAEFELI, R. C., KULSRUD, H. E. (1969), ” Sonic boom propagation in a
stratified atmosphere with computer program ”, NASA CR-1299
KRAVTSOV, Yu. A. and ORLOV, Yu. I. (1993), Caustics, catastrophes and wave fields
(Springer-Verlag, Berlin), pp. 8-33
KUZNETSOV, V. P. (1970), ”Equations of nonlinear acoustics,” Sov. Phys. Acoust., 16,
467-470
MAGLIERI, D. J., PLOTKIN, K. J. (1995), ”Sonic Boom”, in Aeroacoustics of Flight Vehi-
cles, Vol. 1 (Noise Sources), ed. Hubbard H. H., Acoustical Society of America, 519-561
MARCHIANO, R., COULOUVRAT, F., GRENON, R. (2003), ”Numerical simulation of shock
wave focusing at fold caustics, with application to sonic boom,” J. Acoust. Soc. Am., 114,
1758-1771
MARCHIANO, R., COULOUVRAT, F., THOMAS, J.-L. (2005), ” Nonlinear focusing of
acoustic shock waves at a caustic cusp ”, J. Acoust. Soc. Am., 117, 566-577
MENDOUSSE, J. S. (1953), ” Nonlinear dissipation distorsion of progressive sound waves at
moderate amplitude ”, J. Acoust. Soc. Am., 25, 51-54
MILNE, E. A. (1921), ”Sound waves in the atmosphere”, Phil. Mag., 42, 96-114
NOVIKOV, B. K., RUDENKO, O. V., TIMOSHENKO, V. I. (1987), Nonlinear Underwater
Acoustics, Translation Series, American Institute of Physics, INC (New York)
PECHUZAL, G., KEVORKIAN, J. (1977), ” Supersonic-transonic flow generated by a thin
airfoil in a stratified atmosphere ”, SIAM J. Appl. Math., 33, 8-33

18
PIACSEK, A. A. (2002), ”Atmospheric turbulence conditions leading to focused and folded
sonic boom wave fronts,” J. Acoust. Soc. Am., 111, 520-529
PLOTKIN, K. J. (2002), ”State of the art of sonic boom modeling”, J. Acoust. Soc. Am.,
111, 530-536
ROSALES, R. R., TABAK, G. E. (1997), ” Caustics of weak shock waves ”, Phys. Fluids, 10,
206-222
STURTEVANT , B. and KULKARNY, V. A. (1976), ”The focusing of weak shock waves,” J.
Fluid Mech. 73, 651-671.
TABAK, G. E., ROSALES, R. R. (1994), ” Focusing of weak shock waves and the Von Neu-
mann paradox of oblique shock reflection ”, Phys. Fluids, 6, 1874-1892
THOM, R. (1972), ”Stabilité structurelle et morphogenèse” (Benjamin, Reading), pp. 72-107
(in French)
WHITHAM, G. B. (1952), ”The flow pattern of a supersonic projectile”, Comm. Pure Appl.
Math., 5, 301-348
WHITHAM, G. B. (1956), ”On the propagation of weak shock waves”, J. Fluid Mech., 1,
290-318
ZABOLOTSKAYA, E. A. and KHOKHLOV, R. V. (1969), ”Quasi-plane waves in the non-
linear acoustics of confined beams,” Sov. Phys. Acoust. 15, 35-40.

19

You might also like