STOKES THEORY FOR WEAKLY NONLINEAR WATER WAVES

Per A. Madsen
Dep. of Mechanical Engineering, Technical University of Denmark,
2800 Kgs Lyngby, Denmark

Version 20-09-2005

Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2 The governing equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
3 The perturbation method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
4 First order solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
5 Second order solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
6 Third order solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

1. Introduction
Linear water wave theory was originally developed by Airy (1845), while the nonlinear
wave theory was developed by Stokes (1847). George Gabriel Stokes (1819-1903) was
an Irish mathematical physicist. He was a contemporary of Kelvin and Maxwell, and
he has become well known to generations of international scientists, mathematicians and
engineers, through his association with various physical laws and mathematical formulae.
In standard textbooks of mathematics, physics and engineering we find Stokes law, Stokes
theorem, Stokes waves, and the Navier-Stokes equations. He spent most of his working
life at the University of Cambridge, occupying the Lucasian Chair of Mathematics from
1849 until his death in 1903. This prestigious chair was once held by Isaac Newton, and
is currently held by Stephen Hawking. Stokes contribution (1847) on higher-order waves
of uniform amplitude set the tone of research for nearly 100 years.

2. The governing equations

We adopt a Cartesian coordinate system with the x-axis located on the still water level
(SWL) and with the z-axis pointing vertically upwards. We emphasize that this is dif-
ferent from the approach discussed in HCR (Svendsen & Jonsson), where the x-axis is
always located at the mean water level (MWL). We shall discuss the consequences for the
solutions, whenever they appear.
The fluid domain is bounded by the horizontal sea bed at z = h and by the free
surface z = (x, t). The governing equations for the fully nonlinear wave problem in 2D
consist of the two linear equations

2 2
+ = 0, for h z , (1)
x2 z 2

= 0, for z = h, (2)
z
and of the two nonlinear equations

t + = 0, for z = , (3)
z x x

1
 STOKES THEORY FOR WEAKLY NONLINEAR WATER WAVES 2

1
+ g + + = 0, for z = . (4)
t 2 x x z z
It is convenient to introduce the particle velocities defined directly at the free surface,
i.e.

e u(x, , t)
u , (5)
x z=


e w(x, , t)
w , (6)
z z=
by which we can reformulate (3) and (4) to obtain

e+u
t w e = 0, for z = , (7)
x
2
e + g + 1 u
e2 = 0, for z = ,
e +w (8)
2
where
e
. (9)
t z=

Finally, we need to invoke the Bernoulli equation in order to determine the pressure
distribution, once the velocity field has been determined. This equation reads

1 2
p(x, z, t) = gz + + u + w2 . (10)
t 2

It is common to split up the pressure in two parts: the hydrostatic part and the excess
pressure p+ due to the wave motion. In this case we obtain

p(x, z, t) = p+ (x, z, t) gz. (11)

We have now established all the governing equations. The challenge is to find analytical
solutions to this nonlinear problem. For this purpose we invoke the perturbation method.

3. The perturbation method

A comprehensive description of perturbation techniques can be found in Nayfeh (1973).
The principle is to assume that some parameter naturally appearing in the equations, in
our case the nonlinearity, is small i.e. 1. Now we can expand the unknown variables
in powers of i.e.
(x, t) = (1) + 2 (2) + 3 (3) + ...., (12)
(x, z, t) = (1) + 2 (2) + 3 (3) + ..... (13)
where the super-scripts indicate the order of the solution. One then substitutes this expan-
sion into the governing equations, expands all terms for small , and collects coecients
of each power of . Since these equations must hold for all values of , each coecient of
must vanish independently because sequences of are linearly independent. This leads
to a hierarchy of equations, which can be solved successively. Note that in the following,
dierentiation will be defined by subscripts for brevity.
 STOKES THEORY FOR WEAKLY NONLINEAR WATER WAVES 3

3.1. A simple example. Before we start on the water wave problem, let us consider
a simple example of the use of perturbation expansions. For this purpose we consider the
following dierential equation:

uxx 2 u uux = 0, where u = u(x).

For small, we expand the unkown variable in powers of i.e.

u(x) = u(1) + 2 u(2) + 3 u(3) + ......,

and insert this expansion in the dierential equation. Collecting terms of equal powers of
now leads to the following hierarchy
2 (1)
u(1)
xx u = 0, O()
2 (2)
u(2)
xx u = u(1) u(1)
x , O(2 )
2 (3)
u(3)
xx u = u ux + u(1) u(2)
(2) (1)
x . O(3 )

Note that the linear operator acting on the unknown functions in each equation is always
the same, while the nonlinear terms on the right hand side change but they are always
determined from the previously established solutions. We call this technique successive
approximations.

3.2. The water wave problem. The fully nonlinear water wave problem is somewhat
more complicated than the example just given. Therefore we need a few additional tricks
in order to establish the hierarchy of successive approximation. One problem is that we
need the determine the unknown variables directly on the free surface, which is also an
unknown. We fix this problem by using a Taylor expansion from z = 0 i.e. we approximate
e, w,
u e e and e at the free surface by

e (x, , t) (x, z, t) + z (x, z, t) + 1 2 zz (x, z, t) + ...
, (14)
2 z=0

1
e u(x, , t) x (x, z, t) + xz (x, z, t) + 2 xzz (x, z, t) + ...
u , (15)
2 z=0

1
we w(x, , t) z (x, z, t) + zz (x, z, t) + 2 zzz (x, z, t) + ... , (16)
2 z=0

e

1
t (x, z, t) + tz (x, z, t) + 2 tzz (x, z, t) + ... . (17)
t z= 2 z=0

Now we can insert (12)-(13) in (14)-(17), which are then inserted in (7) and (8), and
finally we collect the hierarchy of equations in equal powers of . In principle this leads
to equations on the form

L{(1) } = 0,
L{2 (2) } = N {(1) },
L{3 (3) } = N {(1) + 2 (2) },
L{m (m) } = N {(1) + 2 (2) + .... + m1 (m1) },
 STOKES THEORY FOR WEAKLY NONLINEAR WATER WAVES 4

where L is a linear operator and N is a nonlinear operator. This hierarchy is quite similar
to the one established in the simple example above. Again the linear operator acting on
each solution will be the same, while the nonlinear terms on the right hand side will dier
from equation to equation, but they will always be known expressions based on previous
solutions. In mathematical terms, the first order equation i.e. O(1), is homogeneous, while
all the other equations are inhomogeneous. It is well known from e.g. linear mechanical
oscillations that homogeneous equations represent free oscillations, while inhomogeneous
equations represent forced oscillations with amplitudes and frequencies determined
entirely by the external forcing terms (in our case coming from the nonlinear opera-
tor N ). This is a very important property, and it means that we can always estimate the
form of the individual solutions by calculating the forcing terms on the right hand side of
the equation.
Actually, it turns out that N will consist of quadratic nonlinearities, and we can obtain
a first rough estimate of the right hand sides of these equations simply by taking the square
of the sum of all previous solutions e.g.
2
L{m (m) } (1) + 2 (2) + .... + m1 (m1) .

The purpose of using this very rough approximation, is to get a quick estimate of the
general form of the next order solution. For instance if (1) sin , we find that ((1) )2
sin 2 and hence we should look for second order solutions on the form (2) sin 2. This
trick allows us, in practise, to replace (12)-(13) by more specific forms of the solutions as
demonstrated in the following.

4. First order solution

To obtain the first order solution for progressive waves (i.e. we ignore the possibility of
evanescent solutions), we look for solutions of the form

(1) = A1 cos(t kx), (18)

(1) = B1 cosh k(z + h) sin(t kx). (19)

These solutions satisfy the Laplace equation as well as the kinematic condition at the sea
bed, so we can concentrate on satisfying the surface conditions defined by (7) and (8)
with (14)-(17). Collecting terms of order O(1) leads to the following algebraic problem

k sinh kh A1 0
= .
g cosh kh B1 0

To obtain non-trivial solutions to this homogeneous problem, we need to require that the
determinant is zero. In that case the two equations are no longer independent, and we
obtain the condition
2 gk tanh kh = 0. (20)
This is the linear dispersion relation, which defines the wave number for a given water
depth and a given wave frequency. Once (20) is satisfied, we can solve any one of the
equations in the matrix problem (they will be identical once you invoke (20)) and we can
determine B1 in terms of A1 :
A1
B1 = . (21)
k sinh kh
 STOKES THEORY FOR WEAKLY NONLINEAR WATER WAVES 5

Now we can express the velocity potential in terms of

cosh k(z + h)
(1) = A1 sin(t kx). (22)
k sinh kh
The wave height is defined as the distance from the wave crest to the wave trough i.e.

H = max min , (23)

which, from (18) is simply

H = 2A1 . (24)

4.1. The particle velocities. The resulting particle velocities are determined by
spatial dierentiation of (1) and we obtain
cosh k(z + h)
u(1) (1)
x = A1 cos(t kx), (25)
sinh kh
sinh k(z + h)
w(1) (1)
z = A1 sin(t kx). (26)
sinh kh
4.2. The pressure distribution. By the use of (10), (11) and (22), we determine
the first order excess pressure by
cosh k(z + h)
p+(1) = gA1 cos(t kx). (27)
cosh kh
5. Second order solution
Before we start looking for the second order solution, we roughly estimate its form. As
explained in section 3, we do this by taking the square of (1) , which yields
2 1
(1) = 2 A21 (1 + cos 2) , t kx.
2
Now we shall look for second order solutions of the form

(2) = A2 cos 2 + C2 , (28)

(2) = B2 cosh 2 (z + h) sin 2 + U2 x, (29)

where A2 , B2 , C2 , U2 and 2 are to be determined. By inserting (29) into the Laplace
equation, we immediately find that 2 = 2k.
Note that the constant C2 has been added to (28) to take into account that the surface
elevation is measured from the SWL rather than the MWL i.e. if we time average the
surface elevation over a wave period the result is not neccessarily zero. To second order
we find that
ZT
1
(x, t)dt = C2 . (30)
T
0

Again we emphasize that in this note h is measured from the SWL. In contrast the h used
in HCR is measured from MWL i.e. hHCR = hMW L . To second order, the dierence
between the two is given by
hMW L = h + C2 . (31)
 STOKES THEORY FOR WEAKLY NONLINEAR WATER WAVES 6

Note also that the term U2 has been added to (29) to allow for a correction of the
wave-averaged mass flux. We shall return to a discussion of this property later in this
section.
Now we substitute (28) and (29) into (12) and (13), which again are substituted into
(14)-(17), which are finally substituted into (7) and (8). Terms of order O() are collected
and as a result we obtain two algebraic problems: one which involves harmonic terms
proportional to cos 2 and sin 2 i.e.

2 2k sinh 2kh A2 A1 B1 k 2 cosh kh
= 1 2 2 1 , (32)
g 2 cosh 2kh B2 4 B1 k + 2 A1 B1 k sinh kh

and one which involves the determination of the second order setup (or setdown) of the
mean water level, C2 . The latter problem immediately leads to the solution

kA21
C2 = , (33)
2 sinh 2kh
which implies that at second order the wave-averaged water level is lower than the still
water level.
Next, let us return to the matrix problem defined by (32). First, we realize that the
determinant of the left hand matrix is non-zero, because the set (2k, 2) does not satisfy
the linear dispersion relation in contrast to the set (k, ). For this reason we denote the
second order solution bound waves in contrast to the free waves which always satisfy to
linear dispersion relation. The solution to (32) is readily found to be

1 2 coth kh
A2 = A k (2 + cosh 2kh) ,
4 1 sinh2 kh
3 A21
B2 = ,
8 sinh4 kh
which implies that the second order solution (measured from SWL) reads

(2) 1 2 coth kh kA21
= A k (2 + cosh 2kh) cos 2 , (34)
4 1 sinh2 kh 2 sinh 2kh

3 cosh 2k(z + h)
(2) = A21 sin 2 + U2 x, (35)
8 sinh4 kh
where t kx and where U2 remains to be determined.

Exercise 1. Follow the recipe given in this lecture note to derive second order solutions
for A2 , B2 and C2 . HINT: Start your Mathematica notebook with a definition of the
governing equations as functions of the surface variables ue, w, e Then define these vari-
e .
ables as functions of (x, z, t) by using (14)-(17). Then define the form of the first order
solutions for and , insert in the governing equations and collect terms of order O(1).
Determine the first order problem i.e. the dispersion relation and B1 . Now introduce the
form of the second order solution (but include the first order solution), collect terms of
order O(), and solve with respect to A2 , B2 and C2 .
 STOKES THEORY FOR WEAKLY NONLINEAR WATER WAVES 7

0.5

q
1 2 3 4 5 6

- 0.5
Second

- 1
First

Figure 1: First and second order surface elevation. T = 8s, h = 10m, H = 2m.

5.1. The surface elevation. At second order the surface elevation is now given by
(1) + 2 (2) , hence by dropping the marker, we obtain

1 2 coth kh kA21
(x, t) = A1 cos + A1 k (2 + cosh 2kh) cos 2 . (36)
4 sinh2 kh 2 sinh 2kh

The wave height is defined by (23) and we realize that the first order result (24) still
holds to second order i.e. H = 2A1 (i.e. the peak and trough are both shifted upwards by
the same amount).
From Figure 1 we notice that the eect of the second order correction to the surface
elevation is an increase of the wave crest as well as of the wave trough. It is also evident
that the crest of the wave becomes more narrow, while the through becomes wider. Figure
2 shows a case where the wave height is more than twice the one from Figure 1, and we
notice the appearance of secondary undulations in the wave trough. This is an indication
of a violation of the validity of the perturbation theory, and we shall require that secondary
undulations do not occur in the profile. In order to establish a criteria for this requirement,
we use
(x, t) = A1 cos + A2 cos 2 + C2 .
The second derivative of this expression with respect to reads

= A1 cos 4A2 cos 2,

and we shall simply require that at = (i.e. at the trough) the sign of this curvature
should always be positive i.e.
1
A2 < A1 . (37)
4
If this condition is violated the second order theory is no longer valid.
 STOKES THEORY FOR WEAKLY NONLINEAR WATER WAVES 8

h
3

q
1 2 3 4 5 6
- 1 Second

- 2
First

Figure 2: First and second order surface elevation. T = 8s, h = 10m, H = 4.5m.

5.2. The velocity field. The second order particle velocities are determined by spa-
tial dierentiation of (2) and we easily obtain
3 2 cosh 2k(z + h)
u(2) (2)
x = A k cos 2 + U2 , (38)
4 1 sinh4 kh
3 2 sinh 2k(z + h)
w(2) (2)
z = A1 k sin 2, (39)
4 sinh4 kh
where U2 remains to be determined.

5.3. The wave-averaged mass flux. We are now ready to introduce a new concept,
the wave-averaged mass flux M , which is defined by
ZT Z
1
M u(x, z, t)dzdt. (40)
T
0 h

At first order, this quantity is zero, but it turns out that this is not neccessarily the case
at second order. To obtain a consistent second order approximation, we first split up this
integral in two contributions and obtain

ZT Z0 Z
1
M= u(x, z, t)dz + u(x, z, t)dz dt.
T
0 h 0

The first integral in the bracket is straight forward to determine, and because of the
time-averaging the only contribution will come from the non-harmonic term U2 . Hence
we get
ZT Z0
1
M1 = (u(1) + u(2) )dz dt = 2 U2 h.
T
0 h
 STOKES THEORY FOR WEAKLY NONLINEAR WATER WAVES 9

The second integral in the bracket calls for a more careful treatment: First, we expand u
in a Taylor series from z = 0 i.e.

u z2 2u
u(x, z, t) ' (u)z=0 + z + + ..... (41)
z z=0 2 z 2 z=0

This leads to
Z
1 u 1 2u
udz ' (u)z=0 + 2 + 3 + ... (42)
2 z z=0 6 z 2 z=0
0

i.e.
Z
udz ' 2 (1) u(1) (x, 0, t) = 2 A21 coth kh cos2 , (43)
0

and after time-averaging over a wave period we get

ZT Z
1 1
M2 = udzdt ' 2 A21 coth kh. (44)
T 2
0 0

Now the total wave averaged mass flux becomes

1
M = 2 hU2 + A21 coth kh , (45)
2

where U2 still remains to be determined.

We note that in a wave flume or perpendicular to a beach, we will require that the
wave-averaged mass flux is zero, so in this case we choose
1 2
U2 = A coth kh. (46)
2h 1
Below the wave trough this leads to a wave-averaged return flow, i.e. a net flow against
the wave direction. In general, however, i.e. in the open ocean, we will typically accept
a wave-averaged mass flux and in this situation we will use U2 = 0, unless there is an
ambient current. It turns out that the choice of U2 will influence the speed of the wave
at third order (see Section 6).

5.4. The wave-averaged Eulerian particle velocities. In the calculation of the

wave-averaged mass flux, we noticed that the non-zero contribution came from M0 i.e.
from (44). This means that it actually comes from the wave motion in the region between
the wave trough and the wave crest. We shall now have a closer look on this region from
an Eulerian perspective i.e. from a fixed coordinate system.
First, we shall determine the wave-averaged particle velocity at a fixed position some-
where between the trough and the crest i.e. A1 z A1 . We introduce the scaled
vertical coordinate z/A1 , which will cover the interval from 1 to 1. During one wave
period, an arbitrary point in this interval will be wet only part of the time and dry for
the rest of the time (see Figure 3). With = A1 cos the wet condition can be expressed
 STOKES THEORY FOR WEAKLY NONLINEAR WATER WAVES 10

0.5
z

q
- 4 - 2 2 4

- 0.5

- 1

Figure 3: Sketch showing the surface elevation as cos and a chosen vertical level
(horizontal line). A point at this level is wet only part of the time (filled curve).

as cos , which implies that the wet period covers arccos arccos . Hence
the Eulerian wave-averaged particle velocity is determined by

Z
arccos
1
u(, ) = u(1) d. (47)
2
arccos

We insert (41) in (47) and obtain

q
A1 coth kh
u(, ) = 1 2 + O(A21 ). (48)

The spatial variation of (48) is illustrated in Figure 4. It goes to zero at the trough and
crest, it has a maximum at z = 0 and it is symmetrical around the SWL. This figure
confirms Figure 10b in HCR p. 159.
Finally, as a check we may vertically integrate (48) over the interval A1 z A1 ,
which yields
ZA1 Z1
1
M1 = u(, )dz = A1 u(, )d = A21 coth kh. (49)
2
A1 1

This is in perfect agreement with (44) as expected.

5.5. Particle motions and the Lagrangian velocity field. The previous subsec-
tion concentrated on the Eulerian velocity field i.e. velocities measured from fixed points
in space. If we are interested in the motion of individual fluid elements, a Lagrangian
description is more appropriate. In the following we shall derive the Lagrangian properties
of the motion from the Eulerian solution already found. The presentation follows Phillips
(1977) chap 3.3 and Mei (1989) p. 424. First let us consider a particle which moves in
 STOKES THEORY FOR WEAKLY NONLINEAR WATER WAVES 11

z
1
0.75
0.5
0.25
u
0.2 0.4 0.6 0.8 1
- 0.25
- 0.5
- 0.75
- 1

Figure 4: The vertical variation (from the trough to the crest) of the time-averaged
Eulerian particle velocity (scaled so the maximum is unity).

the vicinity of the mean position (x0 , z0 ). The Lagrangian velocity uL is now equal to the
Eulerian velocity at the mean position plus a correction due to the fact that the particle is
moving in an environment where the velocity field varies. The correction is proportional
to the distance travelled and to the spatial rate of change of the local velocity field. We
express this by a Taylor series expansion from the mean position and obtain
u u
uL (x, z, t) = u(x0 , z0 , t) + (x x0 ) + (z z0 ) + ....., (50)
x z
w w
wL (x, z, t) = w(x0 , z0 , t) + (x x0 ) + (z z0 ) + .....
x z
Next, we utilize that the distance travelled can be determined by time-integration of the
Lagrangian velocity i.e.
Z Z
x x0 = uL dt, and z z0 = wL dt. (51)

At lowest order there is no dierence between the Lagrangian and the Eulerian velocity
and (50) simply yields
(1) cosh k(z0 + h)
uL (x, z, t) = u(1) (x0 , z0 , t) = A1 cos 0 , (52)
sinh kh
(1) sinh k(z0 + h)
wL (x, z, t) = w(1) (x0 , z0 , t) = A1 sin 0 .
sinh kh
At second order (50) and (51) lead to
Z (1) Z (1)
(2) (2) (1) u (1) u
uL (x, z, t) = u (x0 , z0 , t) + uL dt + wL dt , (53)
x z
Z Z
(2) (1) w(1) (1) w(1)
wL (x, z, t) = w(2) (x0 , z0 , t) + uL dt + wL dt .
x z
 STOKES THEORY FOR WEAKLY NONLINEAR WATER WAVES 12

We can now insert the first and second order Eulerian velocities in (53), and after time-
integration and spatial dierentiation we obtain

(2) kA21 cosh 2k(z0 + h)
uL = U2 + 2 cosh 2k(z0 + h) + cos 2 2 + 3 , (54)
4 sinh2 kh sinh2 kh
and
(2) 3 kA21
wL = sinh 2k(z0 + h) sin 2. (55)
4 sinh4 kh
(2)
Notice that while wL = 0, we get

(2) kA21
uL = U2 + cosh 2k(z0 + h).
2 sinh2 kh
The vertical variation of this wave-averaged Lagrangian velocity is shown in HCR, p. 159,
Fig 10a. In the absence of a return flow (i.e. if U2 = 0), it is remarkable that while the
wave-averaged Eulerian velocity is zero below the trough, the Lagrangian is non-zero in
the whole water column.
Finally, we can determine the particle paths by time-integration of uL and wL and to
second order we obtain

cosh k(z0 + h) kA21 cosh 2k(z0 + h)
xL = x0 + A1 sin 0 + 2 + 3 sin 20 +
sinh kh 8 sinh2 kh sinh2 kh

kA21
cosh 2k(z0 + h) + U2 t,
2 sinh2 kh

sinh k(z0 + h) 3 kA21

zL = z0 + A1 cos 0 + sinh 2k(z0 + h) cos 20 .
sinh kh 8 sinh4 kh
As a result the particles will drift in the x-direction (see Figure 5), a mechanism which
can be quite important for e.g. the motion of suspended sediments. Finally, we emphasize
that xL agrees perfectly with HCR, p.158, eq 34, while zL diers due to a printing error
in HCR.

5.6. The pressure distribution. Having determined the velocity field, we can use
the Bernoulli equation to determine the corresponding pressure field. We insert the first
and second order solutions into (10), collect terms of order O() and obtain the second
order excess pressure

gkA21 1 + cosh 2kh 6 cosh 2k(z + h)
p+(2) = cos 2 + cosh 2k(z + h)
2 sinh 2kh 2 sinh2 kh
(56)
Note that this excess pressure is measured from the SWL. When we compare it to the
expression (29) given in HCR p. 155, we find that

+(2) gkA21
p+(2) pMW L = gC2 = . (57)
2 sinh 2kh
This dierence is identical to the dierence in hydrostatic pressures in the two coordinate
systems and simply corresponds to the height dierence between the MWL and SWL
 STOKES THEORY FOR WEAKLY NONLINEAR WATER WAVES 13

0.1

0.05

-0.15 -0.1 -0.05 0.05 0.1 0.15 0.2

-0.05

-0.1

Figure 5: Particle paths at second order with (x0 , z0 )=(0,0).

levels. Consequently, the total pressure is identical in the two coordinate systems, as
expected.
At the sea bottom, the second order pressure becomes

+(2) gkA21 7 + cosh 2kh
p (x, h, t) = cos 2 + 1 ,
2 sinh 2kh 2 sinh2 kh
hence when we time-average this expression over one wave period, we obtain

p+(2) (x, h, t) = gC2 .

Hence the total time-averaged pressure at the sea bottom becomes

p(x, h, t) = gh + gC2 = ghMW L .

This is simply the weight of the water column from the sea bottom to the mean water
level (MWL). Again this result agrees with HCR, except that in HCR p. 156 it is stated
that "the time mean of the pressure on the bottom is the same with and without the
waves". This statement is NOT correct, because the presence of the waves (and their
amplitude) will influence the magnitude of C2 and hence the magnitude of hMW L .

6. Third order solution

At second order we discovered the new concepts of wave setdown (i.e. C2 ), net mass flux
(M ) and a possible return flow (U2 ). At third order, we shall also discover completely
new concepts, such as amplitude dispersion and a correction to the wave height.
First, however, we need to roughly estimate the expected form of the third order
solution. As explained in section 3, we do this by taking the square of (1) + 2 (2) , and
 STOKES THEORY FOR WEAKLY NONLINEAR WATER WAVES 14

collecting terms of order 3 . This yields

2
(1) + 2 (2) 3 (A1 A2 cos 3 + A1 A2 cos ), where t kx. (58)

Hence we can expect to look for third order solutions of two types, namely cos 3 as well as
cos . The former is to be expected, but the latter is somewhat surprising as it resembles
the first order solution.

6.1. A first naive attempt. Based on the forcing shown in (58), we shall make our
first attempt to determine a third order solution based on the following form

(3) = A3 cos 3 + A13 cos + C3 , (59)

(3) = B3 cosh 3 (z + h) sin 3 + B13 cosh 1 (z + h) sin + U3 x, (60)

where A3 , A13 , B3 , B13 , C3 , U3 , 1 and 3 are to be determined. At first, by inserting (59)
and (60) into the Laplace equation, we immediately find that 3 = 3k and 1 = k.
Next, we substitute (59) and (60) into (12) and (13), which again are substituted into
(14)-(17), which are finally substituted into (7) and (8). Terms of order O(3 ) are collected
and as a result we immediately find that C3 0. Hence, there is no contribution to the
setdown from third order. The remainder consists of two algebraic problems, one which
is proportional to cos , sin and one which is proportional to cos 3, sin 3. We start by
considering the subproblem proportional to cos , sin which leads to

k sinh kh A13 F3
= , (61)
g cosh kh B13 F4

where

1
F3 = A1 U2 k + A2 B1 k 2 + B1 C2 k 2 cosh kh + (62)
2
3
A1 B2 k2 cosh 2kh + A21 B1 k 3 sinh kh,
8

3 2 2 1
F4 = B1 U2 k + A1 B1 k cosh kh + A2 B1 k + B1 C2 k sinh kh + (63)
8 2

1
A1 B12 k 3 + 2A1 B2 k sinh 2kh + B1 B2 k 2 cosh 3kh.
2

We recognize the left hand matrix, which is identical to the first order problem. Unfor-
tunately this is crucial, because we have already required that the determinant of this
matrix be zero (leading to the linear dispersion relation). This means that the third
order driving terms (on the right hand side) are going to resonate with the first order
homogeneous problem. The result is so-called secular solutions which are unbounded in
time. This makes the stated problem unsolvable, and we have to come up with a more
sophisticated perturbation expansion.
 STOKES THEORY FOR WEAKLY NONLINEAR WATER WAVES 15

6.2. Elimination of secular terms. A classical remedy to the problem of secular

terms is to expand the frequency (or the wave number) in powers of . Hence we use the
expansion
= (1) + (2) + 2 (3) + ..... (64)
Since we have already managed to derive bounded second order solutions without this
remedy, we can directly set (2) = 0, and instead we use

= 1 1 + 2 13 + .... . (65)

Now we insert this idea into the first, second and third order solutions, which are re-
expressed as

= A1 cos + 2 (A2 cos 2 + C2 ) + 3 A3 cos 3, (66)

= B1 cosh k(z + h) sin + 2 (B2 cosh 2k(z + h) sin 2 + U2 x) + (67)

3 (B3 cosh 3k(z + h) sin 3 + B13 cosh k(z + h) sin + U3 x) ,

where
t kx = 1 (1 + 2 13 )t kx. (68)
Note that the first and second order solutions derived in Sections 4 and 5 are still valid,
so the unknowns in the third order problem are A3 , B3 , B13 , U3 and 13 . Note also that
we have kept the B13 sin term in (67), while we have dropped the A13 cos term from
our first attempt. Actually this is a choice and we could just as well have kept A13 and
dropped B13 .
We substitute (66) and (67) into (12) and (13), which again are substituted into (14)-
(17), which are finally substituted into (7) and (8). Terms of order O(3 ) are collected
and as a result we obtain two algebraic problems, one which is proportional to cos ,
sin and one which is proportional to cos 3, sin 3. We start considering the subproblem
proportional to cos , sin which leads to

A1 1 k sinh kh 13 F3
= , (69)
B1 1 cosh kh 1 cosh kh B13 F4

where F3 and F4 are already defined by (62) and (63). This time the determinant of the
matrix is non-zero and the problem is solvable. The solution to this problem yields

A21 k 2 k k
13 = 4 (8 + cosh 4kh) + C2 + U2 , (70)
16 sinh kh sinh 2kh 1
and

A31 k 1 A1 1 coth 2kh
B13 = (13 + 24 cosh 2kh + cosh 4kh) + C2 . (71)
64 sinh5 kh sinh kh
6.3. Discussion of the amplitude dispersion. The third order expression for
now reads
= 1 1 + 2 13 , (72)
where 1 is determined by (20) and 13 by (70). At first, it may seem rather strange
that we have expanded the frequency in powers of . After all, the frequency appears
 STOKES THEORY FOR WEAKLY NONLINEAR WATER WAVES 16

to be a fixed quantity once the wave period has been given. This is indeed true and in
reality, what we have done is to expand the dispersion relation, which determines the
wave number for a given depth and frequency. This becomes clear when we insert the
expressions for 1 and 13 into (72), which yields

p A21 k 2 k k
= gk tanh kh 1 + 4 (8 + cosh 4kh) + C2 + U2 . (73)
16 sinh kh sinh 2kh 1
Now this equation should be applied in the following way: Insert the relevant solutions
for C2 and U2 and solve the equation with respect to k for given values of h, A1 and
= 2/T . Clearly the celerity of the wave defined as c /k will now be a function of
water depth, wave frequency and wave amplitude.
From (73) it is obvious that the dispersion relation and therefore also the celerity of
the wave will depend on the value of U2 i.e. on whether we assume that the wave-averaged
mass flux is zero or not. This issue was already discussed by Stokes (1880) and later by
e.g. by Fenton (1985, 1990). Stokes used a so-called "first definition" of the celerity,
which asumes zero return current i.e. U2 = 0 , and a "second definition" which assumes
zero mass flux i.e. M = 0 and U2 6= 0. In the latter case we actually have a wave
propagating on an ambient opposing current, and therefore it is physically plausible that
this should modify the speed of the wave. On the other, as stated by Fenton, if we do not
know whether the mass flux M or the return current U2 is zero, then there is no point in
applying a higher order theory as it is no longer well defined.

6.4. Verification of the amplitude dispersion. Now let us verify 13 against ex-
pressions from the literature. At first, we compare with Mei (1984), who also used the
SWL as the reference datum. On page 619 eq. 3.3 he gives the expression

MEI A21 k 2 2
k
13 = 4 8 + cosh 4kh 2 tanh kh + U2 ,
16 sinh kh 1
where U2 is the arbitrary return current. By inserting our result for C2 we obtain a perfect
agreement with C.C. Mei.
Next, we compare with Svendsen & Jonsson (1974), who used the MWL as the refer-
ence datum. On page 161 eq. 41 they give the expression
2
p A21 k2 A1 k coth kd
HCR = gk tanh kd 1 + (8 + cosh 4kd) , (74)
16 sinh4 kd 2d
where d hMW L and where the last term is caused by the return current. In order to
compare (73) and (74), we first need to utilize that hMW L = h + 2 C2 and consequently
a Taylor expansion of the leading order term yields

p p C2 k
gk tanh k(h + 2 C2 ) gk tanh kh 1 + 2 + O(3 ) .
sinh 2kh
Obviously this expansion exactly accounts for the C2 term appearing in (73). In the
higher order terms we can directly replace d by h, and finally we note that the last term
in (74) is identical to U2 k/ 1 . Hence we have found a perfect agreement with Svendsen
& Jonsson.
Finally it should be mentioned that also Skjelbreia & Hendrickson (1960) give ex-
pressions measured from the MWL as reference level, but they assume that U2 = 0.
Consequently, their results are similar to Svendsen & Jonsson but without the last term
in (74).
 STOKES THEORY FOR WEAKLY NONLINEAR WATER WAVES 17

6.5. The third order problem propotional to cos3. Having removed all the
secular terms proportional to cos and sin , the remaining problem is to determine A3
and B3 so that the cos 3 and sin 3 terms are matched. Collecting these terms we obtain

3 1 3k sinh 3kh A3 F1
= , (75)
g 3 1 cosh 3kh B3 F2

where
3 3
F1 = A2 B1 k2 cosh kh + 3A1 B2 k2 cosh 2kh + A21 B1 k 3 sinh kh,
2 8
1 1
F2 = B1 B2 k2 cosh kh + A2 B1 k 1 sinh kh + 2A1 B2 k 1 sinh 2kh + A21 B1 k2 1 cosh kh.
2 8

Again, we realize that the determinant of the left hand matrix is non-zero, because the
set (3k, 3 1 ) does not satisfy the linear dispersion relation in contrast to the set (k, 1 ).
For this reason also the third order solution consists of bound waves in contrast to the
free waves which always satisfy to linear dispersion relation. The solution to (75) reads

3 A31 k 2
A3 = (14 + 15 cosh 2kh + 6 cosh 4kh + cosh 6kh) , (76)
256 sinh6 kh
1 A31 k 1
B3 = (11 + 2 cosh 2kh) . (77)
64 sinh7 kh
Together with (70) and (71) this finally defines the third order solution for the surface
elevation and the velocity potential.

6.6. The wave height relation. The wave height is defined by (23), which at third
order leads to H = 2 (A1 + A3 ) i.e. we get

3 A21 k2
H = 2A1 1 + (14 + 15 cosh 2kh + 6 cosh 4kh + cosh 6kh) . (78)
256 sinh6 kh

This equation needs to be combined with the nonlinear dispersion relation (73) in order
to determine k and A1 for given values of h, T and H.

6.7. The velocity field. The velocity field is detemined by dierentiating the velocity
potential given by (67). This yields

u(z, ) = kB1 cosh k(z + h) cos 2 2kB2 cosh 2k(z + h) cos 2 + 2 U2

3 3kB3 cosh 3k(z + h) cos 3 3 kB13 cosh k(z + h) cos + 3 U3 , (79)

w(z, ) = kB1 sinh k(z + h) sin 2 2kB2 sinh 2k(z + h) sin 2

3 3kB3 sinh 3k(z + h) sin 3 3 kB13 sinh k(z + h) sin , (80)

Now we can determine the wave averaged mass flux of order 3 by using (42) i.e.
Z
1 u 1 2u
udz ' (u)z=0 + 2 + 3 + ...
2 z z=0 6 z 2 z=0
0
 STOKES THEORY FOR WEAKLY NONLINEAR WATER WAVES 18

We insert (79) and (66) into this expression and collect terms at order 3 , which yields
Z
1
udz = 3 k cosh kh A2 B1 (cos + cos 3) + B1 C2 cos
2
0

3 2 2 3 1
k A1 B1 sinh kh cos + cos 3
8 8
3 kA1 B2 cosh 2kh (cos + cos 3) .

The time-averaged value of this result is obviously zero, i.e. there is no contribution from
third order to the wave-averaged mass flux. Hence we can set the third order return flow
U3 to zero.

Exercise 2. Continue the previous exercise to third order, remove the secular terms by
using (66), (67) and (68) and derive solutions for 13 , B13 , A3 and B3 .

References
[1] Airy, G.B. (1845). Tides and waves. Encycl. Metrop. London
[2] Dean, R.G. and Dalrymple, R.A. (1984). Water wave mechanics for engineers and
scientists.Volume 2, Advanced Series on Ocean Engineering, World Scientific, Sin-
gapore.
[3] Dingemans, M.W. (1997). Water wave propagation over uneven bottoms. Part 1 -
Linear wave propagation. Part 2 - Nonlinear wave propagation. Volume 13, Ad-
vanced Series on Ocean Engineering, World Scientific, Singapore.
[4] Fenton, J.D. (1985). A fifth-order Stokes theory for steady waves. J. Waterways Port,
Coastal and Ocean Eng., ASCE, 111, 216-234.
[5] Fenton, J.D. (1990). Nonlinear wave theories. The Sea, Vol. 9, Chapter 1, Ocean
Engineering Science, John Wiley, New York.
[6] Goda, Y. (2000). Random seas and design of maritime structures. Volume 15, Ad-
vanced Series on Ocean Engineering, World Scientific, Singapore.
[7] LeBlond, P.H. and Mysak, L.A. (1980). Waves in the Ocean. Elsevier Oceanography
Series, Amsterdam.
[8] Lighthill, J. (1979). Waves in Fluids. Cambridge University Press, Cambridge.
[9] Mei, C.C. (1989). The applied dynamics of ocean surface waves. Volume 1, Advanced
Series on Ocean Engineering. World Scientific, Singapore.
[10] Nayfeh, A. (1973). Perturbation methods. Pure and applied mathematics. John Wiley,
New York.
[11] Phillips, O.M. (1977). The dynamics of the upper ocean, 2nd ed. Cambridge Univer-
sity Press, London.
[12] Skjelbreia, L. and Hendrickson, J. (1960). Fifth order gravity wave theory. Chapter
10, p. 184-196, 7th Int. Conf. on Coastal Engineering, ASCE.
 STOKES THEORY FOR WEAKLY NONLINEAR WATER WAVES 19

[13] Stoker, J.J. (1957). Water waves: The mathematical theory with applications. Wiley
Classics Library, New York.
[14] Stokes, G.G. (1847). On the theory of oscillatory waves. Trans. Cambridge Philos.
Soc. 8, 441-455. Reprinted in Math. Phys. Papers Cambridge University Press, Lon-
don, vol 1, 314-326.
[15] Stokes, G.G. (1880). On the theory of oscillatory waves. Math. Phys. Papers Cam-
bridge University Press, London, vol 1, 202-203.
[16] Svendsen, I.A. and Jonsson, I. (1976). Hydrodynamics of Coastal Regions, Den private
ingenirfond, DTU, Lyngby.
[17] Whitham, G.B. (1973). Linear and nonlinear waves. Pure and Applied Mathematics.
John Wiley, Singapore.