ESOE 5136 2021 May 3 - Handout Lo

Derivation of the solitary wave solution

The derivation shown here is based on the textbook by Mei et al. (2005).

1. 1DH Boussinesq equations in constant depth

The Boussinesq wave theory assumes that both nonlinearity and frequency dispersion are
equally small. Therefore, O() = O(µ2 ) < 1, and anything smaller than O() or O(µ2 ) can
be neglected. In other words, the waves are assumed to be weakly nonlinear and weakly
dispersive.
Accurate to the order of O(, µ2 ), the dimensionless Boussinesq equations (expressed in
terms of the velocity potential) in 1DH constant depth are

ηt + (ηφx )x + φxx = O(µ2 , µ4 ) (1)

and
 µ2
φt + η + φ2x = φxxt + O(µ2 , µ4 ). (2)
2 3
(2) can be rearranged to yield

 2 µ2

2 4
 η = −φt − φx + φxxt + O(µ , µ )


2 3
. (3)
  µ 2
 η = −φ − (φ2 ) + φ
 2 4
t tt t xxtt + O(µ , µ )
2 x 3

(3) is then used to replace ηt in (1):

 µ2
−φtt − (φx )2t + φxxtt + (−φt φx )x + φxx = O(µ2 , µ4 ). (4)
2 3
Noting that

φtt = φxx + O(, µ2 )







1 1


(φt φx )x = φtx φx + φt φxx , ( φ2x )t = φx φxt , ( φ2t )t = φt φtt






 2 2

⇒ −(φt φx )x = −(φx φxt + φxx φt ) = −(φx φxt + φt φtt ) + O(2 , µ2 ) , (5)


    
⇒ − (φx )2t − (φt φx )x = − (φx )2t − (φx )2t − (φt )2t + O(2 , µ2 )



2 2 2 2





 1
= −(φ2x + φ2t )t + O(2 , µ2 )



2
we can simplify (4) to be

µ2 1
φtt − φxx = φxxtt − (φ2x + φ2t )t + O(µ2 , µ4 ), (6)
3 2
which is another form of the Boussinesq equation in 1DH constant depth.

2. Permanent wave solution

In linear wave theories, the most elementary wave solutions have the form eik(x−Ct) , in
which C is the dimensionless wave speed. Since these solutions represent waves propagating

1
ESOE 5136 2021 May 3 - Handout Lo

at constant speed without change of form, they are called permanent waves (定形波) or
stationary waves (定常波).
A natural question to ask is then: do these permanent waves exist in nonlinear wave
theories as well? We recall from previous lectures that nonlinear effects tend to steepen
waves, whereas dispersive effects tend to lengthen waves. Therefore, if permanent waves
exist in a nonlinear and dispersive wave theory, they must correspond to an equilibrium
in which the nonlinear effects and the dispersive effects are in perfect balance.
To investigate the possibility of permanent waves within the Boussinesq wave theory, let
us seek a solution of the form φ = φ(ξ), where ξ = x − Ct, to the rearranged Boussinesq
equation (6).
With ξ = x − Ct, the derivatives become
∂ d ∂ d
(· · · ) = (· · · ) = (· · · )0 , (· · · ) = −C (· · · ) = −C(· · · )0 . (7)
∂x dξ ∂t dξ
The rearranged Boussinesq equation (6) then becomes
µ2 2 0000 1
(C 2 − 1)φ00 = C φ + C(1 + C 2 )(φ02 )0 + O(µ2 , µ4 ). (8)
3 2
Noting the order of magnitude of each term in (8), we learn that in order for the above
equality to hold,
√ a
C 2 − 1 = O(, µ2 ) ⇒ C = 1 + O(, µ2 ), since 1 + a ' 1 + for small a. (9)
2
Substituting (9) into the right hand side of (8) and omitting higher-order terms gives
µ2 0000 3 02 0
(C 2 − 1)φ00 = φ + (φ ) + O(µ2 , µ4 ). (10)
3 2
Integrating (10) once yields
µ2 000 3 02
(C 2 − 1)φ0 + A1 =φ + φ + O(µ2 , µ4 ), (11)
3 2
where A1 is an integration constant. (3) says
η = −φt + O(, µ2 ) = φ0 + O(, µ2 ), (12)
which allows us to write (11) in terms of η as
µ2 00 3 2
(C 2 − 1)η + A1 = η + η + O(µ2 , µ4 ). (13)
3 2
It is understood that terms of O(µ2 , µ4 ) or smaller are omitted in the equations. Therefore,
from now on we will stop writing the O(µ2 , µ4 ) term in the equations.
Multiplying (13) by η 0 gives
µ2 00 0 3 2 0
(C 2 − 1)ηη 0 + A1 η 0 = η η + η η. (14)
3 2
Then, we integrate (14) once to obtain
1 2 µ2 02  3  η2 µ2 02
(C − 1)η 2 + A1 η + A2 = η + η ⇒ − η 3 + (C 2 − 1) + A1 η + A2 = η , (15)
2 6 2 2 2 6
where A2 is another integration constant.
(15) is now a nonlinear ordinary differential equation of the first order (i.e., the first
derivative is the highest derivative). Two types of solutions are known to exist: the
solitary wave solution and the cnoidal wave solution.

2
ESOE 5136 2021 May 3 - Handout Lo

孤立 波) solution
3. Solitary wave (孤
Based on Scott Russell’s observations, we know solitary waves have the following proper-
ties:

(a) The wave is an elevation wave; i.e., η(ξ) > 0.

(b) The wave smoothly decays to zero; i.e., η(ξ) → 0, η 0 (ξ) → 0, and η 00 (ξ) → 0, for
|ξ| → ∞.

In hindsight (事後看來), we know solitary waves have the shape

η(ξ) = sech2 (Kξ), (16)

where K is the effective wave number. Therefore, we can also plot the sech2 function
(figure 1) to confirm the above observations.
1

0.8

0.6

0.4

0.2

0
-5 -4 -3 -2 -1 0 1 2 3 4 5
K

Figure 1: Plot of the solitary wave shape given in (16).

Imposing the solitary wave shape requirements, i.e.,

η(ξ) > 0, and η(ξ) → 0, η 0 (ξ) → 0, η 00 (ξ) → 0, for |ξ| → ∞, (17)

we find A1 = 0 from (13) and A2 = 0 from (15). The Boussinesq equation (15) simplifies
to
µ2 02  η2 3 3  C2 − 1
η = − η 3 +(C 2 −1) ⇒ η 02 = 2 (C 2 −1)η 2 − 2 η 3 = 3η 2 ( 2 )( −η). (18)
6 2 2 µ µ µ 

For real-valued solutions, η 02 must be non-negative; that is, the right hand side of (18)
must be non-negative. Keeping in mind the requirements in (17), we then have

C2 − 1
( − η) ≥ 0 ⇒ C 2 ≥ 1 ⇒ C ≥ 1. (19)

We see that the nonlinear wave speed C must be equal to or greater than the linear wave
speed of 1.
At the wave crest, η = 1 and η 0 = 0, (18) requires

C2 − 1 √
( − 1) = 0 ⇒ C 2 = 1 +  ⇒ C = 1 + , (20)

which is the (nonlinear) wave speed of the solitary wave! In dimensional form, the wave
speed of the solitary wave is
C ∗ = g(h + A),
p
(21)

3
ESOE 5136 2021 May 3 - Handout Lo

where h is the water depth and A is the wave amplitude. We see that the larger the wave
amplitude A is, the faster the solitary wave propagates. We note that for a solitary wave,
its wave amplitude A is the same as its wave height H.
To determine the shape of the solitary wave, we substitute (20) into (18) to obtain

 dη (3)1/2
η 02 = 3η 2 ( )(1 − η) ⇒ η 0
= = η(1 − η)1/2 . (22)
µ2 dξ µ

The ordinary differential equation (22) is separable and can be solved as

√
(3)1/2
Z Z
1 −1
p 3
√ dη = dξ ⇒ −2 tanh 1−η = (ξ − ξ0 ), (23)
η 1−η µ µ

in which ξ0 is an integration constant. Rearranging (23) gives

√  √3
−1
p 3 p 
tanh 1−η =− (ξ − ξ0 ) ⇒ 1 − η = tanh − (ξ − ξ0 )
2µ 2µ
, (24)
 √3   √3 
2 2
⇒ η = 1 − tanh − (ξ − ξ0 ) = 1 − tanh (ξ − ξ0 )
2µ 2µ

since tanh2 (x) is an even function.

Lastly, using the property
sech2 (x) = 1 − tanh2 (x), (25)
and setting ξ0 = 0 so that the wave crest corresponds to ξ = 0, we write (24) as
r 3  r 3 
2 2
η(ξ) = sech ξ = η(x, t) = sech (x − Ct) , (26)
4µ2 4µ2

in which the dimensionless wave speed C is given by (20). (26) is the leading-order solitary
wave solution – accurate to O(, µ2 ).
With dimensions, the solitary wave solution (26) becomes
r 3A 
η ∗ (x∗ , t∗ ) = Asech2 (x ∗
− C ∗ ∗
t ) , (27)
4h3
in which the dimensional wave speed C ∗ is given by (21).

橢圓函 數波) solution

4. Cnoidal wave (橢
Cnoidal waves are periodic solutions to (15). They are called “cnoidal” waves because the
wave shapes are captured by the Jacobi elliptic function (雅可比橢圓函數) “cn”. We will
not discussed cnoidal waves in class. The derivation can be found in Mei et al. (2005).

References
Mei, C., M. Stiassnie, and D. K.-P. Yue (2005), Theory and Applications of Ocean Surface
Waves. World Scientific.