146 views

Uploaded by spsingh72

save

- Generalized Thermoelastic Wave Propagation in Circumferential Direction of Transversely Isotropic Cylindrical Curved Plates
- Dynamic Analysis of Me Ms
- Me Scope
- Tubaldi, Dall'Asta, Dezi 15wcee PRSI Bridges
- Cen Et Al (2016) Effect of Mass Ratio on Hydrodynamic Response of a Flexible Cylinder
- Blade Group Resonant Stresses
- 1-s2.0-S0022460X03003833-main
- 1-s2.0-S1878522009001246-main
- Modes Indicate Cracks in Wind Turbine Blades
- 2005_3.pdf
- Zheng-Wang Qu et al- The Huggins band of ozone: Unambiguous electronic and vibrational assignment
- 1112012913Dynamics Response Spectrum Analysis.pdf
- 7-3
- Mdof Frf Curvefit
- GL Ship Vibration 09
- Mathematics Ncert Ch03 Pair of Linear Equations in Two Variables Ex 3.6 Ans Nvb
- Cavity Drag and Flow Oscillations
- Acoustic Logging
- J. Bredenbeck et al- The vibrational spectrum of deuterated phosphaethyne: A quantum mechanical, classical, and semiclassical analysis
- T1073-79.pdf
- mathematics
- 10th Maths
- Windowing Time Data
- Serpentine Mode Converter Design
- Sensor Location Effect on Flexible Robot Control
- A Baseline and Vision of Ultrasonic Guided Wave Inspection Potential
- EMCF-9-F14
- ITE 2702 - CM moratuwa university
- AWB130 Dynamics 04 Spectrum
- ANSYS57_tut_modal
- z-013.pdf
- P-100(Process).pdf
- Ships Hydrodynamics
- Norsok M-501 ed 6 Clarification log 2015 rev 1_finished.pdf
- FPI Rules E-Feb14
- Norsok M-501 FAQ.pdf
- Fatigue Rules
- fea-dundes
- Buckling Strength of Plated Structures
- OSS-302_Offshore Riser Systems_Oct 2003
- CAP437(Feb 2013)
- Rigzone Career
- Os e406_2010 04(Free Fall Lifeboats)
- 21km+2011

You are on page 1of 66

**DEPARTMENT OF OCEAN ENGINEERING
**

Cambridge, MASS. 02139

WAMIT

THEORY MANUAL

by

C.-H. Lee

Report No. 95-2

October 1995

This work was conducted under the Joint Industry Project ‘Wave eﬀects on oﬀshore

structures’. The sponsors have included Chevron, Conoco, David Taylor Model Basin, Det

Norske Veritas, Exxon, Mobil, the National Research Council of Canada, Oﬀshore Tech-

nology Research Center, Norsk Hydro AS, Petrobr´ as, Saga Petroleum, Shell Development

Company, and Statoil AS.

Copyright c Massachusetts Institute of Technology 1995

CONTENTS

1. INTRODUCTION

1.1 Description of the Problem

2. THE FIRST-ORDER PROBLEM

2.1 Boundary Value Problem

2.1.1 Multiple bodies

2.1.2 Body deformations

2.1.3 Bodies near walls

2.2 Integral Equations

2.2.1 Green function

2.2.2 Linear system

2.3 Removal of the Irregular Frequencies

3. THE SECOND-ORDER PROBLEM

3.1 Boundary Value Problem

3.2 Second-order Incident Waves

3.3 Second-order Scattering Waves

4. THE FIRST- AND SECOND-ORDER FORCE

4.1 Force and Moment

4.1.1 Coordinate system

4.1.2 Coordinate transform

4.1.3 Pressure integration

4.1.4 Hydrostatic force and moment of O(1)

4.1.5 Linear force and moment

4.1.5 Second-order force and moment

4.2 Haskind Force and the Second-order Indirect Force

4.3 Equations of Motion

REFERENCES A

REFERENCES B

APPENDICIES

Chapter 1

INTRODUCTION

This Theory Manual describes the theoretical background of the computer program

WAMIT (WaveAnalysisMIT). It provides an overview of the theories and the computa-

tional methodologies in a concise manner. For detailed information on the selected subjects,

some of the published and unpublished articles are included in the Appendix. The use of

the program WAMIT and its capabilities are described in a separate User Manual.

WAMIT is a panel program designed to solve the boundary-value problem for the inter-

action of water-waves with prescribed bodies in ﬁnite- and inﬁnite- water depth. Viscous

eﬀects of the ﬂuid are not considered throughout and thus the ﬂow ﬁeld is potential with-

out circulation. A perturbation series solution of the nonlinear boundary value problem is

postulated with the assumption that the wave amplitude is small compared to the wave

length. It is also assumed that the body stays at its mean position and, if it is not ﬁxed,

the oscillatory amplitude of the body motion is of the same order as the wave amplitude.

The time harmonic solutions corresponding to the ﬁrst- and second-terms of the series

expansion are solved for a given steady-state incident wave ﬁeld. The incident wave ﬁeld

is assumed to be represented by a superposition of the fundamental ﬁrst-order solutions of

particular frequency components in the absence of the body. The boundary value problem

is recast into integral equations using the wave source potential as a Green function. The

integral equation is then solved by a ‘panel’ method for the unknown velocity potential

and/or the source strength on the body surface. Using the latter, the ﬂuid velocity on the

body surface is evaluated.

For incident waves of speciﬁc frequencies and wave headings, the linear problem is solved

ﬁrst. Chapter 2 describes the linear boundary value problem and the solution procedure.

The quadratic interaction of two linear solutions deﬁnes the boundary condition for the

second-order problem. Chapter 3 is devoted to the discussion of the second-order problem.

In Chapter 4, the rigid body motion is discussed in the context of the perturbation series

and the expressions for the hydrodynamic forces are derived. Chapter 5 lists the references

quoted in this manual and Chapter 6 lists the references illustrating the computational

results obtained from WAMIT. The Appendix includes the articles on the subjects which

are not explained in detail in this manual.

Since the following will focus primarily on the hydrodynamic interaction, it is appropri-

1

ate to mention here restrictions on the bodies which may be analyzed using WAMIT. The

bodies may be ﬁxed, constrained or neutrally buoyant and they may be bottom mounted,

submerged or surface piercing. Multibody interactions between any combinations of these

bodies can be analyzed. If only ﬁrst-order quantities are required, the bodies need not be

rigid. Flexible bodies can be analyzed if the oscillatory displacements of the body surface

are described by speciﬁed mode shapes.

An extremely thin submerged body may be diﬃcult to analyze when the small thickness

between the two facing surfaces compared to the other dimensions of the body may render

the integral equation ill-conditioned

1

. However, a thin body ﬂoating on the free surface (in

other words the draft is very small) is not included in this category, since it has only one

large wetted surface. In the extreme case of the latter, a body touching the free surface

may be analyzed.

1.1 Description of the Problem

The ﬂow is assumed to be potential, free of separation or lifting eﬀects, and it is governed

by the velocity potential Φ(x, t) which satisﬁes Laplace’s equation in the ﬂuid domain:

∇

2

Φ = 0 (1.1)

Here t denotes the time and x=(x, y, z) denotes the Cartesian coordinates of a point in

space. x may be expressed as the sum of the component vectors as x = xi +yj +zk. The

undisturbed free surface is the z = 0 plane with the ﬂuid domain z < 0.

The ﬂuid velocity is given by the gradient of the velocity potential

V(x, t) = ∇Φ =

∂Φ

∂x

i +

∂Φ

∂y

j +

∂Φ

∂z

k (1.2)

The pressure follows from Bernoulli’s equation:

p(x, t) = −ρ(

∂Φ

∂t

+

1

2

∇Φ · ∇Φ +gz) (1.3)

where ρ is the density of the ﬂuid and g is the gravitational acceleration.

The velocity potential satisﬁes the nonlinear free-surface condition:

∂Φ

∂t

2

+g

∂Φ

∂z

+ 2∇Φ · ∇

∂Φ

∂t

+

1

2

∇Φ · ∇(∇Φ · ∇Φ) = 0 (1.4)

applied on the exact free surface

ζ(x, y) = −

1

g

(

∂Φ

∂t

+

1

2

∇Φ · ∇Φ)

z=ζ

(1.5)

1

See Martin and Rizzo (1993)

2

With the assumption of a perturbation solution in terms of a small wave slope of the

incident waves, the velocity potential is expanded in a form

Φ(x, t) = Φ

(1)

(x, t) + Φ

(2)

(x, t) +· · · (1.6)

When the body is not ﬁxed the motion amplitude

2

of the body is also expanded in a

perturbation series

ξξ ξ = ξξ ξ

(1)

+ξξ ξ

(2)

+· · · (1.7)

From (1.5) and (1.6) and with aid of the Taylor expansion with respect to the mean

free surface, the free surface elevation also takes the form of the perturbation series,

ζ(x, y) = ζ

(1)

(x, y) +ζ

(2)

(x, y) +· · · (1.8)

with

ζ

(1)

(x, y) = −

1

g

∂Φ

∂t

(1)

(1.9)

ζ

(2)

(x, y) = −

1

g

(

∂Φ

∂t

(2)

+

1

2

∇Φ

(1)

· ∇Φ

(1)

−

1

g

∂Φ

∂t

(1)

∂

2

Φ

(1)

∂z∂t

) (1.10)

In the equations (1.9) and (1.10), the right-hand sides are evaluated on z = 0.

From the equations (1.4), (1.6) and (1.8-10), the free surface boundary conditions for

Φ

(1)

and Φ

(2)

imposed on z = 0 are derived. For a moving body, from the equations (1.6)

and (1.7) and with the Taylor expansion of Φ with respect to the mean body surface, the

body boundary condition is derived. These boundary conditions are discussed in later

chapters.

Given a wave spectrum, it is customary to assume the spectrum is expressed as a linear

superposition of the ﬁrst-order incident waves of diﬀerent frequencies. Thus the total ﬁrst

order potential for the wave-body interaction can be expressed by a sum of components

having circular frequency ω

j

> 0:

Φ

(1)

(x, t) = Re

j

φ

j

(x)e

iω

j

t

(1.11)

Here we introduce the complex velocity potential φ

j

(x), indepedent of time, with the

understanding that the real part of the time harmonic solution is physically relevant.

In (1.11) φ

j

(x) denotes the ﬁrst-order solution in the presence of the incident wave of

frequency ω

j

and the wave heading β

j

. The directional spreading of the incident waves is

not shown explicitly in (1.11). However if it is necessary, one can easily incorporate it by

expressing φ

j

(x) as a sum of the velocity potential components each of which corresponds

to a particular wave heading.

At the second-order, the total velocity potential takes a form

2

Here ξξ ξ includes both the translational and the rotational modes of the rigid body. Frequently ξξ ξ is

used, in a restrictive sense, to denote only the translational modes. In that case αα α is used for the rotational

modes.

3

Φ

(2)

(x, t) = Re

i

j

φ

+

ij

(x)e

i(ω

i

+ω

j

)t

+φ

−

ij

(x)e

i(ω

i

−ω

j

)t

(1.12)

with symmetric conditions

φ

+

ij

= φ

+

ji

and φ

−

ij

= φ

−∗

ji

(1.13)

Here we may assume ω

i

≥ ω

j

≥ 0 without loss of the generality. φ

+

ij

and φ

−

ij

are

referred to as the sum- and diﬀerence-frequency velocity potential with frequencies ω

i

+ω

j

and ω

i

−ω

j

, respectively. These are determined by two linear incident wave components:

one has frequency ω

j

and wave heading β

j

and the other ω

i

and β

i

. If one consider the

directional spreading of the incident waves, φ

±

ij

(x) are represented by a double summation

over wave headings (β

i

, β

j

).

In the next two chapters, we discuss the solution procedures for the complex velocity

potentials φ

j

and φ

±

ij

in (1.11) and (1.12).

4

Chapter 2

THE FIRST-ORDER PROBLEM

In this Chapter, we review the ﬁrst-order boundary value problem and its solution proce-

dure. Here we consider one particular frequency component, ω

j

from the discrete spectrum

(1.12). Thus we omit the subscript j indicating the frequency component in this Chapter.

The velocity potential of the ﬁrst-order incident wave is deﬁned by

φ

I

=

igA

ω

Z(κz)e

−iκ(x cos β+y sin β)

(2.1)

representing a plane progressive wave of the circular frequency ω and the wave heading

angle β. β is an angle of incidence to the positive x − axis. A denotes the complex wave

amplitude.

The function Z represents the depth dependence of the ﬂow and is given by

Z(κz) = e

κz

(2.2)

for inﬁnite water depth, where κ =

ω

2

g

≡ ν. For a ﬂuid of depth h, it is given by

Z(κz) =

cosh(κ(z +h))

cosh(κh)

(2.3)

where the wavenumber κ is the real root of the dispersion relation

κ tanhκh =

ω

2

g

(2.4)

An eﬃcient root ﬁnding algorithm for the computation of κ from equation (2.4) is

explained in Newman (1990).

The scattering velocity potential φ

S

represents the disturbance to the incident wave due

to the presence of the body in its ﬁxed position. Linearity of the problem allows it to be

distinguished from the disturbance due to the the motion of the body which is represented

by the radiation potential φ

R

. Thus the total velocity potential is given by

φ = φ

I

+φ

S

+φ

R

= φ

D

+φ

R

(2.5)

5

where the diﬀraction potential φ

D

is deﬁned to be the sum of φ

I

and φ

S

.

The radiation potential itself is a linear combination of the components corresponding

to the modes of motion such that

φ

R

= iω

6

k=1

ξ

k

φ

k

(2.6)

Here ξ

k

is the complex amplitude of the oscillatory motion in mode k of the six degrees of

freedom, and φ

k

the corresponding unit-amplitude radiation potential (speciﬁcally, unit-

amplitude means the unit-amplitude linear or angular velocity of the rigid body motion).

These modes are referred to as surge, sway, heave, roll, pitch and yaw in the increasing

order of j.

2.1 Boundary Value Problem

The total ﬁrst-order velocity potential φ, along with each of its components, satisﬁes the

Laplace equation in the ﬂuid domain:

∇

2

φ = 0, (2.7)

the linear free-surface condition:

φ

z

−νφ = 0 on z = 0, (2.8)

and a condition on the sea bottom:

∇φ →0 as z →−∞, (2.9a)

or

φ

z

= 0 on z = −h (2.9b)

for inﬁnite- and ﬁnite- water depth, respectively.

In addition, the scattering and the radiation potentials are subject to a radiation con-

dition stating that the wave energy associated with the disturbance due to the body is

carried away from the body in all direction in the far ﬁeld.

Finally, the conditions on the body surface complete the description of the boundary

value problem. They take the form

∂φ

k

∂n

= n

k

(2.10)

and

∂φ

S

∂n

= −

∂φ

I

∂n

(2.11)

6

where (n

1

, n

2

, n

3

) = n and (n

4

, n

5

, n

6

) = x × n. The unit vector n is normal to the

body boundary and it is assumed the normal vector points out of the ﬂuid domain. x is

the position of a point on the body boundary. From (2.5), it follows that

∂φ

D

∂n

= 0 (2.12)

Variations to the canonical problem speciﬁed above include multibody interactions,

generalized modes and the presence of the vertical walls.

1

These topics are reviewed brieﬂy

next, along with the exploitation of the symmetry of the bodies.

2.1.1 Multiple bodies

The decomposition of the radiation potential into components, corresponding to the modes

of the rigid body motion, can be extended to multi-body interaction. This is done by

deﬁning φ

k

as the velocity potential corresponding to a particular mode of one body while

the other bodies are kept stationary. In this way, the total radiation potential consists of

6N components, where N is the number of bodies.

2.1.2 Body deformations

Appication can also be made to analyze the more general modes of motion of the structure,

beyond the rigid body motions. Examples are the bending modes of structural deformation

and the compressible modes of a ﬂexible body. To analyze these modes, the mode shapes are

described in terms of the vector displacement of the body boundary. Let u

k

(x) be the mode

shape given as a function of the points on the body boundary. Then the corresponding

radiation potential φ

k

satisﬁes

∂φ

k

∂n

= u

k

· n (2.13)

on the body boundary. The details of the analysis for generalized modes and its applications

are described in Newman (1994).

2.1.3 Bodies near walls

When walls are present, the velocity potential is subject to the condition

∂φ

∂n

= 0 (2.14)

on the walls. If we deﬁne the incident wave as in (2.1), a reﬂected wave (φ

r

I

) should be

added to φ

I

to describe the wave ﬁeld in the absence of the body. The resultant wave ﬁeld

1

Walls are assumed to coincide with the x = 0 or the y = 0 planes. When there are two walls they

meet perpendicularly at the origin.

7

satisﬁes (2.14) and it is a standing wave with maximum free surface elevation 2|A|. (The

actual free surface elevation depends on the incident wave heading β.) The total velocity

potential is then expressed in the form

φ = φ

I

+φ

r

I

+φ

S

+φ

R

(2.15)

If the body geometry is symmetric with respect to the x = 0 plane, the y = 0 plane or

both, the computational domain for φ

D

, φ

S

and φ

R

can be reduced to half or one quadrant

of the entire domain. In these cases, the velocity potentials are expressed as a sum of the

symmetric and antisymmetric components with respect to the planes of symmetry. On

these planes, the symmetric component satisﬁes

∂φ

∂n

= 0 (2.16)

and the antisymmetric component satisﬁes

φ = 0 (2.17)

2.1.4 Integral Equations

The boundary value problem is solved by the integral equation method. Thus the velocity

potential φ

k

on the body boundary is obtained from the integral equation

2πφ

k

(x) +

__

S

B

dξξ ξφ

k

(ξξ ξ)

∂G(ξξ ξ; x)

∂n

ξξ ξ

=

__

S

B

dξξ ξn

k

G(ξξ ξ; x) (2.18)

where S

B

denotes the body boundary.

The corresponding equation for the total diﬀraction velocity potential φ

D

is (Korsmeyer

et al (1988))

2πφ

D

(x) +

__

S

B

dξξ ξφ

D

(ξξ ξ)

∂G(ξξ ξ; x)

∂n

ξξ ξ

= 4πφ

I

(x) (2.19)

Alternatively, the scattering potential can be obtained from

2πφ

S

(x) +

__

S

B

dξξ ξφ

S

(ξξ ξ)

∂G(ξξ ξ; x)

∂n

ξξ ξ

=

__

S

B

dξξ ξ(−

∂φ

I

∂n

)G(ξξ ξ; x) (2.20)

and the diﬀraction potential follows from (2.5). From a computational point of view,

(2.19) is slightly more eﬃcient than (2.20) due to the simpler form of the right-hand-side.

We need to evaluate the ﬂuid velocity on the body surface and in the ﬂuid domain for

the computation of the second-order wave forces as discussed in Chapter 4. The velocity

may be computed from the spacial derivatives of Green’s integral equations, (2.18) to

(2.20). However when these equations are solved using a low order panel method, they

can not predict the velocity accurately on or close to the body boundary. This is due to

the hypersingular integral arising from the double derivative of the constant strength wave

8

source potential distributed on a quadrilateral panel. For this reason, the ﬂuid velocity is

computed based on the source formulation. The integral equation for the source strength

σ

k

corresponding to the radiation potential φ

k

takes the form (Lee and Newman (1991))

2πσ

k

(x) +

__

S

B

dξξ ξσ

k

(ξξ ξ)

∂G(ξξ ξ; x)

∂n

x

= n

k

(2.21)

and that of σ

S

corresponding to the scattering potential φ

S

2πσ

S

(x) +

__

S

B

dξξ ξσ

S

(ξξ ξ)

∂G(ξξ ξ; x)

∂n

x

= −

∂φ

I

∂n

(2.22)

The ﬂuid velocity on the body boundary or in the ﬂuid domain due to φ

k

or φ

S

is then

obtained from

∇φ(x) = ∇

__

S

B

dξξ ξσ(ξξ ξ)G(ξξ ξ; x) (2.23)

The ﬂuid velocity due to the incident wave is evaluated directly from (2.1).

Integral equations, (2.18) to (2.22) are solved by the panel method. The wetted body

surface is represented by an ensemble of quadrilateral panels (a triangular panel is a special

type of quadrilateral panel where two vertices coalesce). The unknowns are assumed to

be constant over each panel and the integral equation is enforced at the centroid of each

panel (a collocation method). For example, the discrete form of the equation (2.20) takes

the form

2πφ

S

(x

k

) +

NEQN

n=1

φ

S

(x

n

)

_

Sn

dξξ ξ

∂G(ξξ ξ; x

k

)

∂n

ξξ ξ

=

NEQN

n=1

−

∂φ

I

(x

n

)

∂n

_

Sn

dξξ ξG(ξξ ξ; x

k

) (2.24)

where NEQN is the total number of panels (unknowns) and x

k

are the coordinates of

the centroid of the k-th panel.

2.1.5 The Green Function

The Green function G(x; ξξ ξ), which is referred to as the wave source potential, is the velocity

potential at the point x due to a point source of strength −4π located at the point ξξ ξ. It

satisﬁes the free-surface and radiation conditions, and in inﬁnite water depth it is deﬁned

by (Wehausen and Laitone (1960))

G(x; ξξ ξ) =

1

r

+

1

r

+

2ν

π

_

∞

0

dk

e

k(z+ζ)

k −ν

J

0

(kR) (2.25)

r

2

= (x −ξ)

2

+ (y −η)

2

+ (z −ζ)

2

(2.26)

9

r

2

= (x −ξ)

2

+ (y −η)

2

+ (z +ζ)

2

(2.27)

where J

0

(x) is the Bessel function of zero order. In ﬁnite depth, it is deﬁned by

G(x; ξξ ξ) =

1

r

+

1

r

+ 2

_

∞

0

dk

(k +ν) cosh k(z +h) cosh k(ζ +h)

k sinh kh −ν cosh kh

e

−kh

J

0

(kR) (2.28)

(r

)

2

= (x −ξ)

2

+ (y −η)

2

+ (z +ζ + 2h)

2

. (2.29)

In both expressions (2.25) and (2.28), the Fourier k−integration is indented above the

pole on the real axis in order to enforce the radiation condition.

In the equation (2.24), the inﬂuence due to the continuous distribution of the Rank-

ine part of the wave source potential on a quadrilateral panel is evaluated based on the

algorithms described in Newman (1985). The remaining wave part of the Green function

is evaluated based on the algorithms described in Newman (1992). The integration of the

latter part over a panel is carried out using either one or four point Gauss quadrature.

When the ﬁeld and source points are close together and in the vicinity of the free surface,

G takes a form

G(x; ξξ ξ) =

1

r

+

1

r

−2νe

ν(z+ζ)

(log(r

+|z +ζ|) + (γ −log 2) +r

+O(r

log r

)) (2.30)

where γ is the Euler constant. The logarithmic singularity in (2.30) may not be evaluated

accurately using Gauss quadrature in this situation. The algorithm for evaluating the

inﬂuence of the logarithmic singularity distributed over a panel can be found in Newman

and Sclavounos (1988). When the panel is on the free surface, special care is required for

the evaluation of Green function as is discussed in Newman (1993) and Zhu (1994). The

former is attached in the appendix.

When there are walls or when we exploit the symmetry/antisymmetry of the ﬂow, the

velocity potentials are subject to the conditions (2.14), (2.16) or (2.17). In this case, the

image source must be placed at the reﬂected points of x with respect to the planes of

symmetry. The Green function is then modiﬁed to be a sum of (2.15) or (2.28) and its

image sources.

2.1.6 Linear system

The linear system, generated by implementing the solution procedure described above (for

example (2.24)), is solved by an iterative method, a block iterative method or Gauss elim-

ination. The details on the iterative method is described in Lee (1988). The relative

computational eﬃciency between the iterative method and Gauss elimination can be illus-

trated by the ratio of the number of ﬂoating point operations. This ratio, the former to the

latter, is O(

3KM

NEQN

). Where NEQN is the number of unknowns or the dimension of the

linear system and K is the number of iterations. M is the number of the right-hand sides

of the linear system sharing a left-hand side. (As an example, in (2.24) the number of the

10

right-hand-side vectors is the same as the number of the diﬀerent wave headings of φ

I

for a

ﬁxed left-hand-side matrix.) In most applications the iterative method converges in 10-15

iterations and is the most eﬃcient way to solve the linear system. Gauss elimination is not

only slower than the iterative method for large NEQN but also requires suﬃciently large

core memory to store the entire matrix elements, since the LU decomposition is performed

using the core memory. On the other hand, in the iterative method, all or part of the

matrix may be stored on the auxiliary storage devise (hard disk) and then retrieved at

each iteration.

There are some problems for which the iterative method is slowly convergent or non-

convergent due to bad conditioning of the linear system. Some examples are i) barges with

shallow draft, ii) bodies with large ﬂare and iii) multiple bodies separated by small gaps.

Another case where the iterative method may be slowly convergent is in the linear system

for the extended boundary integral equation (see Section 2.3). For these problems, Gauss

elimination or the block iterative method should be used. The block iterative method is

based on the same algorithm as the iterative method, but Gauss elimination is applied

locally for the speciﬁed diagonal blocks. At each stage of the iterations, the back substitu-

tion is performed for the diagonal blocks. This accelerates the rate of convergence as the

dimension of the block increases. The limiting case is the same as Gauss elimination. The

opposite limit is the case when the dimension of the blocks is one, which is identical to the

iterative method. The block iterative method requires core memory to store the elements

of one block for LU decomposition.

2.1.7 Removal of Irregular Frequencies

The integral equations (2.18-22) have nonunique solutions at the common irregular fre-

quencies corresponding to the Dirichlet eigen-frequencies for the closed domain deﬁned

by the body boundary, and the interior free surface, S

i

. The eigenmodes correspond to

the solution with the homogeneous Dirichlet condition on S

B

and the linear free-surface

condition on S

i

. Numerical solutions of these equations are erroneous near the irregular

frequencies. The extended boundary integral equation method is applied to remove the

irregular frequency eﬀect from the velocity potentials in (2.18-20).

The extended boundary integral equations for φ

k

are

2πφ

k

(x) +

__

S

B

dξξ ξφ

k

(ξξ ξ)

∂G(ξξ ξ; x)

∂n

ξξ ξ

+

_

S

f

dξξ ξφ

k

(ξξ ξ)

∂G(ξξ ξ; x)

∂n

ξξ ξ

=

__

S

B

dξξ ξn

k

G(ξξ ξ; x) (2.31a)

−4πφ

k

(x) +

__

S

B

dξξ ξφ

k

(ξξ ξ)

∂G(ξξ ξ; x)

∂n

ξξ ξ

+

_

S

f

dξξ ξφ

k

(ξξ ξ)

∂G(ξξ ξ; x)

∂n

ξξ ξ

=

__

S

B

dξξ ξn

k

G(ξξ ξ; x) (2.31b)

where φ

k

is an artiﬁcial velocity potential deﬁned in the interior domain. Equations

(2.31a) and (2.31b) are for x on S

B

and S

i

, respectively. These two equations are solved

simultaneously for φ

k

on S

B

and φ

k

on S

i

. We discard φ

k

after we solve the equations,

since only φ

k

on S

B

is physically relevant.

11

The equations for the diﬀraction potential φ

D

are given by

2πφ

D

(x) +

__

S

B

dξξ ξφ

D

(ξξ ξ)

∂G(ξξ ξ; x)

∂n

ξξ ξ

+

_

S

f

dξξ ξφ

D

(ξξ ξ)

∂G(ξξ ξ; x)

∂n

ξξ ξ

= 4πφ

I

(x) (2.32a)

−4πφ

D

(x) +

__

S

B

dξξ ξφ

D

(ξξ ξ)

∂G(ξξ ξ; x)

∂n

ξξ ξ

+

_

S

f

dξξ ξφ

D

(ξξ ξ)

∂G(ξξ ξ; x)

∂n

ξξ ξ

= 4πφ

I

(x) (2.32b)

The scattering potential takes the same form as the radiation potential but −

∂φ

I

∂n

replaces n

k

on the right-hand side of the equations (2.31a) and (2.32b).

The extended boundary integral equation for the source formulation of the radiation

or scattering problem takes a form

2πσ(x) +

__

S

B

dξξ ξσ(ξξ ξ)

∂G(ξξ ξ; x)

∂n

ξξ ξ

+

_

S

f

dξξ ξσ

(ξξ ξ)

∂G(ξξ ξ; x)

∂n

ξξ ξ

= g(x) (2.33a)

−4πσ

(x) +

__

S

B

dξξ ξσ(ξξ ξ)

∂G(ξξ ξ; x)

∂n

ξξ ξ

+

_

S

f

dξξ ξσ

(ξξ ξ)

∂G(ξξ ξ; x)

∂n

ξξ ξ

= −V (x) (2.33b)

where g(x) = n

k

(g(x) = −

∂φ

I

∂n

) for the radition (scattering) problem. The proper

condition of the function V (x) is discussed in Lee et al (1995).

The details of the derivation of these equations are described in Kleinman (1982), Lee

et al (1995) and Zhu (1994). Zhu (1994) also makes a comparison with other methods

for the removal of the irregular frequency eﬀect and demonstrates the eﬀectiveness of the

present method.

12

Chapter 3

THE SECOND-ORDER PROBLEM

From the quadratic interaction of the two linear wave components of the frequencies ω

i

and ω

j

in the discrete spectrum (1.11)

1

, we have the second-order waves of the frequency

components ω

i

+ ω

j

and ω

i

− ω

j

in (1.12). In this Chapter we restrict our attention to

the solution of the second-order boundary value problem, i.e. the second-order potential.

Other second-order quantities, such as the quadratic second-order force , obtainable from

the ﬁrst-order solution, are discussed in Chapter 4.

The second-order velocity potential Φ

(2)

(x, t) is subject to the free surface condition

∂

2

Φ

(2)

∂t

2

+g

∂Φ

(2)

∂z

= Q

F

(x, y; t) (3.1)

and the body boundary condition

∂Φ

(2)

∂n

= Q

B

(x; t) (3.2)

where the inhomogeneous right-hand-side of the second-order free-surface condition

(3.1) deﬁnes the quadratic forcing function (Newman (1977, Section 6.4))

Q

F

=

1

g

∂Φ

(1)

∂t

∂

∂z

(

∂

2

Φ

(1)

∂t

2

+g

∂Φ

(1)

∂z

) −

∂

∂t

(∇Φ

(1)

· ∇Φ

(1)

) (3.3)

Q

F

is to be evaluated on z = 0. The forcing function on the body boundary is given

by (Ogilvie (1983))

Q

B

= −

∂Φ

(2)

I

∂n

+n ·

∂H

∂t

x + (αα α

(1)

×n) · (

∂(ξξ ξ

(1)

+αα α

(1)

×x)

∂t

−∇Φ

(1)

)

−n · ((ξξ ξ

(1)

+αα α

(1)

×x) · ∇)∇Φ

(1)

+

6

k=1

∂ξ

(2)

k

∂t

n

k

(3.4)

1

When body is not ﬁxed, the ﬁrst-order motion aﬀects the second-order solution as shown in the

equation (3.4) below.

13

and it is to be evaluated on the mean body boundary. H in (3.4) is the quadratic

components of the coordinate transformation matrix due to the rotation of the body and it

takes a form (3.14) and (3.15), for sum- and diﬀerency-frequency problems. In (3.4), Φ

(2)

I

is

referred to as the second-order incident wave potential. This is the second-order potential

in the absence of the body and will be discussed in Section 3.2. n

k

is the appropriate

component of n for the translational modes and of a component of (x×n) for the rotational

modes.

In the following evaluations of second-order products of two ﬁrst-order oscillatory quan-

tities, use is made of the relation

Re(Ae

iω

i

t

)Re(Be

iω

j

t

) =

1

2

Re[(Ae

iω

i

t

)(Be

iω

j

t

+B

∗

e

−iω

j

t

)] (3.5)

where (

∗

) denotes the complex conjugate.

In accordance with the deﬁnition of the second-order potential (1.12), Q is expressed

as

Q(x, t) = Re

i

j

_

Q

+

ij

(x)e

i(ω

i

+ω

j

)t

+Q

−

ij

(x)e

i(ω

i

−ω

j

)t

_

(3.6)

With the symmetry condition,

Q

+

ij

= Q

+

ji

and Q

−

ij

= Q

−∗

ji

(3.7)

The other second-order quantities such as the motion amplitude ξξ ξ

(2)

= ξ

±

ij

are expressed

in a form of (3.6) with the symmetry condition (3.7). With this in mind, we will omit the

subscript ij hereafter.

From combining (1.11) and (3.3), we have the expressions for the complex amplitudes

of the free-surface forcing functions. Sum- and diﬀerence- frequency forcings on z = 0 are

given by

Q

+

F

=

i

4g

ω

i

φ

i

(−ω

2

j

∂φ

j

∂z

+g

∂

2

φ

j

∂z

2

) +

i

4g

ω

j

φ

j

(−ω

2

i

∂φ

i

∂z

+g

∂

2

φ

i

∂z

2

)

−

1

2

i(ω

i

+ω

j

)∇φ

i

· ∇φ

j

(3.8)

and

Q

−

F

=

i

4g

ω

i

φ

i

(−ω

2

j

∂φ

∗

j

∂z

+g

∂

2

φ

∗

j

∂z

2

) −

i

4g

ω

j

φ

∗

j

(−ω

2

i

∂φ

i

∂z

+g

∂

2

φ

i

∂z

2

)

−

1

2

i(ω

i

−ω

j

)∇φ

i

· ∇φ

∗

j

(3.9)

For future reference, we note that the ﬁrst-order potential φ

i

and φ

j

in (3.8-9), consist

of the incident wave potential (φ

I

) and the body disturbances (φ

B

= φ

S

+ φ

R

). Thus

we may decompose Q

±

F

into the quadratic interactions of the two incident wave potentials

14

(Q

±

II

), the incident and the body disturbance waves (Q

±

IB

), and two body disturbance waves

(Q

±

BB

) such that

Q

±

F

= Q

±

II

+Q

±

IB

+Q

±

BB

(3.10)

Next we consider the sum- and diﬀerence-frequency forcing on the body boundary.

They are given by

Q

+

B

= −

∂φ

+

I

∂n

+

i(ω

i

+ω

j

)

2

n · H

+

x

+

i

4

[(αα α

i

×n) · (ω

j

(ξξ ξ

j

+αα α

j

×x) −∇φ

j

) + (αα α

j

×n) · (ω

i

(ξξ ξ

i

+αα α

i

×x) −∇φ

i

)]

−

1

4

n · [((ξξ ξ

i

+αα α

i

×x) · ∇)∇φ

j

+ ((ξξ ξ

j

+αα α

j

×x) · ∇)∇φ

i

] (3.11)

and

Q

−

B

= −

∂φ

−

I

∂n

+

i(ω

i

−ω

j

)

2

n · H

−

x

+

i

4

[(αα α

i

×n) · (ω

j

(ξξ ξ

∗

j

+αα α

∗

j

×x) −∇φ

∗

j

) + (αα α

∗

j

×n) · (ω

i

(ξξ ξ

i

+αα α

i

×x) −∇φ

i

)]

−

1

4

n · [((ξξ ξ

i

+αα α

i

×x) · ∇)∇φ

∗

j

+ ((ξξ ξ

∗

j

+αα α

∗

j

×x) · ∇)∇φ

i

] (3.12)

The sum- and diﬀerence-frequency components of the last term of (3.4),

Q

±

B

= i(ω

i

±ω

j

)

k

ξ

±

n

k

(3.13)

are omitted from (3.11-12), since they are not a quadratic function of the ﬁrst-order so-

lution. These are proportional to the second-order motion and can be treated separately

from the rest of body forcing as is discussed below.

The matrices H

±

which account for the rotational motion of the body are given by

H

+

=

1

2

_

_

_

−(α

2i

α

2j

+α

3i

α

3j

) 0 0

α

1i

α

2j

+α

1j

α

2i

−α

1i

α

1j

+α

3i

α

3j

0

α

1i

α

3j

+α

1j

α

3i

α

2i

α

3j

+α

2j

α

3i

−α

1i

α

1j

+α

2i

α

2j

_

_

_ (3.14)

H

−

=

1

2

_

_

_

−(α

2i

α

∗

2j

+α

3i

α

∗

3j

) 0 0

α

1i

α

∗

2j

+α

∗

1j

α

2i

−α

1i

α

∗

1j

+α

3i

α

∗

3j

0

α

1i

α

∗

3j

+α

∗

1j

α

3i

α

2i

α

∗

3j

+α

∗

2j

α

3i

−α

1i

α

∗

1j

+α

2i

α

∗

2j

_

_

_ (3.15)

As in the ﬁrst-order problem, it is convenient to decompose the total second-order

potential into three components: the second order incident wave potential (φ

±

I

), the second-

order scattering wave potential (φ

±

S

) and the second-order radiation potential (φ

±

R

). We

deﬁne φ

±

I

as the potential that satisﬁes the second-order free surface condition in absence

of the body. φ

±

R

describes the disturbance due the second-order motion of the body and

15

is linearly proportional to the motion amplitude. It can be decomposed into the mode

dependent components

φ

±

R

= i(ω

i

±ω

j

)

6

k=1

ξ

k

φ

k

(3.16)

where the unit-amplitude radiation velocity potential φ

k

is deﬁned in the same way as φ

k

in Chapter 2. The rest of the second-order potential is deﬁned to be φ

±

S

, whether the

body is ﬁxed or moving. This decomposition is in accordance with the convention for the

ﬁrst-order problem where the wave exciting force is the pressure force due to the sum of

the incident and scattering waves.

3.1 Boundary Value Problem

The total second-order potential φ

±

satisﬁes Laplace’s equation in z < 0 and the condition

on the sea bottom (equations (2.7) and (2.9a-9b)). The boundary condition on the free

surface, on the body boundary and at the far ﬁeld are speciﬁed below for each component

potential.

The incident wave potential is subject to

−(ω

i

±ω

j

)

2

φ

±

I

+g

∂φ

±

I

∂z

= Q

±

II

on z = 0 (3.17)

The unit amplitude radiation potential is subject to

−(ω

i

±ω

j

)

2

φ

k

+g

∂φ

k

∂z

= 0 on z = 0 (3.18)

and

∂φ

k

∂n

= n

k

(3.19)

on the mean body boundary S

B

.

Finally, the scattering potential is subject to

−(ω

i

±ω

j

)

2

φ

±

S

+g

∂φ

±

S

∂z

= Q

±

IB

+Q

±

BB

on z = 0 (3.20)

and

∂φ

±

S

∂n

= Q

±

B

(3.21)

At the far ﬁeld φ

±

R

satisﬁes the same radiation condition as that for φ

R

of the ﬁrst-order.

For φ

±

S

, we apply the ‘weak radiation condition’ suggested by Molin (1979). Thus based

on the asymptotic behaviours of G and Q

±

F

(in other words, from the behaviours of φ

i

and

φ

j

), the surface integral at the far ﬁeld can be shown to vanish with the result shown in

(3.28). The detailed analysis is in Molin (1979).

16

3.2 The Second-order Incident Waves

Each of the two ﬁrst-order incident waves is deﬁned by the amplitude A, frequency ω, and

vector wavenumber K with Cartesian components (κ cos β, κ sin β, 0). Here β is the angle

of incidence relative to the x−axis. The ﬁrst-order incident wave potential (2.1) can be

written in the form

φ

I

=

igA

ω

Z(κz)e

−iK·x

(3.22)

On z = 0, the ﬁrst- and the second-derivatives of Z in vertical direction are

_

∂Z(κz)

∂z

_

z=0

= ν and

_

∂

2

Z(κz)

∂z

2

_

z=0

= κ

2

(3.23)

Combining (3.22) with (3.8-9) gives

Q

+

II

= −

1

2

ig

2

A

i

A

j

exp(−i(K

i

+K

j

) · x)

__

κ

2

j

−ν

2

j

2ω

j

_

+

_

κ

2

i

−ν

2

i

2ω

i

_

+

(ω

i

+ω

j

)

ω

i

ω

j

(K

i

· K

j

−ν

i

ν

j

)

_

(3.24)

and

Q

−

II

=

1

2

ig

2

A

i

A

∗

j

exp(−i(K

i

−K

j

) · x)

__

κ

2

j

−ν

2

j

2ω

j

_

−

_

κ

2

i

−ν

2

i

2ω

i

_

−

(ω

i

−ω

j

)

ω

i

ω

j

_

K

i

· K

j

+ν

i

ν

j

__

(3.25)

Solutions for the second-order incident-wave potential follow from (3.17) in the form

φ

±

I

=

Q

±

II

(x, y)Z(κ

±

ij

z)

−(ω

i

±ω

j

)

2

+gκ

±

ij

tanhκ

±

ij

h

(3.26)

where

κ

±

ij

= |K

i

±K

j

| (3.27)

Note that in inﬁnite water depth, Q

II

= 0 if β

i

= β

j

. Thus the sum frequency incident

wave potential φ

+

I

vanishes in unidirectional regular and irregular waves.

3.3 The Second-order Scattering Waves

The solution for the second-order scattering potential, φ

±

S

, is obtained from Green’s integral

equation

17

2πφ

±

S

(x) +

_ _

S

B

φ

±

S

(ξξ ξ)

∂G(x; ξξ ξ)

∂n

ξ

dS =

_ _

S

B

Q

±

B

(ξξ ξ)G(x; ξξ ξ)dS

+

1

g

_ _

S

F

Q

±

F

(ξξ ξ)G(x; ξξ ξ)dS (3.28)

where G is the wave source potential deﬁned in (2.25) and (2.28).

After the solution of (3.28), φ

±

S

on the free surface and in the ﬂuid domain can be

obtained from

4πφ

±

S

(x) +

_ _

S

B

φ

±

S

(ξξ ξ)

∂G(x; ξξ ξ)

∂n

ξ

dS =

_ _

S

B

Q

±

B

(ξξ ξ)G(x; ξξ ξ)dS

+

1

g

_ _

S

F

Q

±

F

(ξξ ξ)G(x; ξξ ξ)dS (3.29)

The integral equation (3.28) is solved by the panel method. The left-hand side of (3.28)

is identical to the integral equation for the ﬁrst-order potential and thus the discrete form

of it is the same as that of (2.24) (but with G corresponding to the same frequency ω and

wavenumber as φ

±

S

).

The evaluation of the integrals on the right hand side is described below. We consider

the integral over S

B

ﬁrst. For a ﬁxed body, the evaluation Q

±

B

(x) is simple, since it contains

only the normal velocity due to φ

±

I

which are derived in Section 3.2. When the body is

not ﬁxed, we ﬁrst modify the terms involving the double spacial derivative of the ﬁrst-

order velocity potential by applying Stokes’s theorem to avoid the inaccurate numerical

evaluation of the second-order derivatives based on the lower order panel method (Lee and

Zhu (1993)).

__

S

B

dSG{n · [(ξξ ξ +αα α ×x) · ∇]∇φ} =

__

S

B

dS[n · (ξξ ξ +αα α ×x)](∇φ · ∇G)

+

__

S

B

dSG{n · [(∇φ · ∇)(ξξ ξ +αα α ×x)]} −

__

S

B

dS

∂φ

∂n

[(ξξ ξ +αα α ×x) · ∇G]

+

_

WL

dl · G[∇φ ×(ξξ ξ +αα α ×x)] (3.30)

After substituting (3.30) into Q

±

B

, the integral over S

B

is represented by a sum of the

integral over each panel assuming Q

±

B

(x) is constant on each panel. Q

±

B

(x) is evaluated

on the centroids of the panels. The waterline is approximated by line segments (consisting

of the sides of the panels adjacent to the free surface) and Q

±

B

(x) is evaluated on the

midpoints of the segments.

The free-surface integral which is displayed as the last term in (3.28) will be considered

next in which the ‘quadratic forcing function’ Q

±

F

is given by (3.8-9). For the purposes

of numerical evaluation, the free-surface integral is evaluated separately in two domain

divided by a partition circle of radius ρ = b. b issuﬃciently large to neglect the eﬀect of

the evanescent waves outside the circle. Here ρ =

√

x

2

+y

2

. Following the methodology of

18

Kim and Yue(1989), the integration in the inner domain ρ < b is carried out numerically.

In the outer domain (b ≤ ρ) both the Green function and the asymptotics of the ﬁrst-

order potentials are expanded in Fourier-Bessel series. After integrating the trigonometric

functions with respect to the angular coordinate, the free-surface integrals are reduced

to the sum of the line integrals with respect to the radial coordinate ρ. The method of

integration on the inner domain is described below and that on the outer domain follows

in the next Section.

In the inner domain, in order to avoid the evaluation of second-order derivatives of the

ﬁrst-order potentials in (3.8-9), the surface integral is transformed, using Gauss theorem,

to a form involving only the ﬁrst-order derivatives and line integrals around the waterline

(WL) and the partition circle (PC) as is shown below.

__

S

F

φ

i

∂

2

φ

j

∂z

2

GdS = −

__

WL+PC

φ

i

(∇φ

j

· n)Gdl

+

__

S

F

[(∇φ

i

· ∇φ

j

)G+φ

i

(∇φ

j

· ∇G)]dS (3.31)

where S

F

denotes only the inner domain of the free surface. The divergence ∇ and the

normal vector n must be interpreted in the two dimensional sense on the z = 0 plane.

The line integrals over WL and PC are carried out in a similar way to the line integral

in (3.30). To achieve a better computational eﬃciency for the remaining surface integral,

we divide the inner domain further into two parts separated by a circle of radius ‘a

with

a < b. a is suﬃciently large to enclose the body. Inside this circle, the free surface is

discretized with quadrilateral panels, in an analogous manner to the body surface. Thus

the integration on this region is the sum of the integral over each panels assuming the

constant Q

±

F

on each panel evaluated at the centroid. The domain outside the circle of

radius a to the partition circle forms an annulus. The integral on this domain is carried out

by Gauss-Chebyshev quadrature in the azimuthal direction and Gauss-Legendre quadrature

in the radial direction as discussed in Lee and Zhu (1993).

When the ﬁeld point x on the body and the source point ξξ ξ on the annulus are close, the

Rankine singularity of G(x; ξξ ξ) renders Gauss-Chebyshev quadrature ineﬃcient. Discussion

on the selection of the optimum a is provided in Chapter 11 of the WAMIT User Manual

(1995). The analysis on the integral over the far-ﬁeld free surface is described in detail in

Newman (1991) which is included in the Appendix

19

Chapter 4

THE FIRST- AND

SECOND-ORDER FORCES

The expressions for the ﬁrst- and second-order forces are derived from direct integration of

the ﬂuid pressure over a body boundary in Section 4.1. By making use of Green’s theorem,

part of the forces can be obtained without solving the scattering potential. The ﬁrst-order

Haskind exciting force and the second-order force via indirect approach are discussed in

Section 4.2. The equations of motion for the ﬁrst- and second-order problem are derived

in Section 4.3.

4.1 Hydro-static and -dynamic Force and Moment

4.1.1 Coordinate system

We consider three coordinate systems. X = (X, Y, Z) is a global coordinate system and

with Z = 0 the undisturbed free-surface. The positive Z axis points upward. x = (x, y, z)

is a body-ﬁxed coordinate system and z also is positive upward when the body is at rest.

We introduce a third coordinate system which is ﬁxed in space and coincides with “x at

rest”. This coordinate system is denoted by ˆ x = (ˆ x, ˆ y, ˆ z). Note that X = (X, Y, Z) and

ˆ x = (ˆ x, ˆ y, ˆ z) are inertial reference frames but x = (x, y, z) is not. The origin of ˆ x may be

displaced from the free-surface and Z

o

denotes the Z-coordinate of the origin of ˆ x.

4.1.2 Coordinate transform

The position vectors in the ˆ x− and x− coordinate systems are related to each other by

linear transformation

ˆ x = ξξ ξ +T

t

x (4.1)

20

and the normal vector by

ˆ n = T

t

n (4.2)

In (4.1-2), T

t

is the transpose of T = T

3

T

2

T

1

and T

1

, T

2

and T

3

take forms

T

1

=

_

_

_

1 0 0

0 cos α

1

sin α

1

0 −sin α

1

cos α

1

_

_

_

T

2

=

_

_

_

cos α

2

0 −sin α

2

0 1 0

sin α

2

0 cos α

2

_

_

_ (4.3)

T

3

=

_

_

_

cos α

3

sin α

3

0

−sin α

3

cos α

3

0

0 0 1

_

_

_

ξξ ξ = (ξ

1

, ξ

2

, ξ

3

) and αα α = (α

1

, α

2

, α

3

) are the translational and rotational displacements

of x(-coordinate system) with respect to ˆ x, respectively. They also represent the mo-

tion amplitudes of the body in the order of surge-sway-heave and roll-pitch-yaw. Further

discussion on T can be found in Ogilvie (1983).

4.1.3 Pressure integration

In the following derivation, it is understood that the pressure, the velocity potential and

the body motion amplitudes are functions of time, though the time t does not appear

explicitly.

The total pressure at ˆ x is given by Bernoulli’s equation

P(ˆ x) = −ρ[Φ

t

(ˆ x) +

1

2

∇Φ(ˆ x) · ∇Φ(ˆ x) +g(ˆ z +Z

o

)] (4.4)

The hydro-static and -dynamic force and moment are obtained from the integration of

the pressure over the instantaneous wetted surface.

F =

__

ˆ

S

B

P(ˆ x)ˆ ndS (4.5)

and

M =

__

ˆ

S

B

P(ˆ x)(ˆ x × ˆ n)dS (4.6)

where

ˆ

S

B

denotes the instantaneous wetted body boundary.

The pressure on the exact body surface (ˆ x ∈

ˆ

S

B

) may be approximated by Taylor

expansion with respect to the mean body surface (x ∈ S

B

)

P(ˆ x) = P(x) + [ξξ ξ + (T

t

−I)x] · ∇P(x) +... (4.7)

21

Substituting (4.7) into (4.1-2), we have

ˆ x = x +ξξ ξ

(1)

+αα α

(1)

×x +Hx +ξξ ξ

(2)

+αα α

(2)

×x +O(A

3

) (4.8)

ˆ n = n +αα α

(1)

×n +Hn +αα α

(2)

×n +O(A

3

) (4.9)

The cross product of (4.8) and (4.9) then takes a form

ˆ x × ˆ n = x ×n +ξξ ξ

(1)

×n +αα α

(1)

×(x ×n) +ξξ ξ

(1)

×(αα α

(1)

×n)

+ H(x ×n) +ξξ ξ

(2)

×n +αα α

(2)

×(x ×n) +O(A

3

) (4.10)

From (4-4), (4-7) and (4.11), the pressure is expressed by the values on the mean body

position

P(ˆ x) = −ρ{g(z +Z

o

) + [Φ

(1)

t

(x) +g(ξ

(1)

3

+α

(1)

1

y −α

(1)

2

x)]

+ [

1

2

∇Φ

(1)

(x) · ∇Φ

(1)

(x) + (ξξ ξ

(1)

+αα α

(1)

×x) · ∇Φ

(1)

t

(x) +gHx · ∇z]

+ [Φ

(2)

t

(x) +g(ξ

(2)

3

+α

(2)

1

y −α

(2)

2

x)]} +O(A

3

) (4.11)

H in (4.7-11) is the second-order component of T

t

and it takes a form

H =

_

_

_

_

−

1

2

((α

(1)

2

)

2

+ (α

(1)

3

)

2

) 0 0

α

(1)

1

α

(1)

2

−

1

2

((α

(1)

1

)

2

+ (α

(1)

3

)

2

) 0

α

(1)

1

α

(1)

3

α

(1)

2

α

(1)

3

−

1

2

((α

(1)

1

)

2

+ (α

(1)

2

)

2

)

_

_

_

_

(4.12)

Substituting the expansions (4.7-11) into (4.5-6), we have the series expansion of the

integrals (4.5-6).

4.1.4 Hydrostatic force and moment of O(1)

These are the buoyancy force and moment when the body is at rest and are expressed by

F = −ρg

_ _

S

B

(z +Z

o

)ndS

(4.13)

M = −ρg

_ _

S

B

(z +Z

o

)(x ×n)dS = −ρg

_ _

S

B

[(z +Z

o

)x] ×ndS

The following relations are invoked frequently in this Chapter.

−

_ _

S

ψndS =

_ _ _

V

∇ψdV

−

_ _

S

n ×ψdS =

_ _ _

V

∇×ψdV (4.14)

∇×(z +Z

o

)x = k ×x

22

where S is a closed surface consisting of S

B

and the waterplane area A

wp

. V denotes

the volume of the body.

Using the relations (4.14), the force and moment are expressed in familiar forms

F = ρgV k

(4.15)

M = ρgV (y

b

i −x

b

j)

where x

b

and y

b

are the x and y coordinates of the center of buoyancy.

In (4.15), i, j, and k are the unit vectors in ˆ x coordinate system.

4.1.5 Linear force and moment

The linear force and moment are obtained from

F

(1)

= − ρ

_ _

S

B

nΦ

(1)

t

dS

− ρg

_ _

S

B

(αα α

(1)

×n)(z +Z

o

)dS

− ρ

_ _

S

B

ng(ξ

(1)

3

+α

(1)

1

y −α

(1)

2

x)dS

(4.16)

M

(1)

= − ρ

_ _

S

B

(x ×n)Φ

(1)

t

dS

− ρg

_ _

S

B

(x ×n)(ξ

(1)

3

+α

(1)

1

y −α

(1)

2

x)dS

− ρg

_ _

S

B

(ξξ ξ

(1)

×n)(z +Z

o

)dS

− ρg

_ _

S

B

[αα α

(1)

×(x ×n)](z +Z

o

)dS

In (4.16), the ﬁrst terms are the hydrodynamic force and moment and the rest are the

hydrostatic. Following the decomposition (2.5), we consider component potentials such

that Φ

(1)

= Φ

(1)

I

+ Φ

(1)

S

+ Φ

(1)

R

= Φ

(1)

D

+ Φ

(1)

R

. The hydrodynamic force and moment are

divided into two components: the “wave exciting force” due to Φ

(1)

D

and the force due to

Φ

(1)

R

expressed in terms of the added mass and damping coeﬃcients.

The integrals of the hydrostatic pressure can be simpliﬁed by applying (4.14) and their

variations with the results

F

(1)

= − ρ

_ _

S

B

Φ

(1)

t

ndS −ρgA

wp

(ξ

(1)

3

+α

(1)

1

y

f

−α

(1)

2

x

f

)k

23

(4.17)

M

(1)

= − ρ

_ _

S

B

(x ×n)Φ

(1)

t

dS

− ρg[−V ξ

(1)

2

+A

wp

y

f

ξ

(1)

3

+ (V z

b

+L

22

)α

(1)

1

−L

12

α

(1)

2

−V x

b

α

(1)

3

]i

− ρg[ V ξ

(1)

1

−A

wp

x

f

ξ

(1)

3

−L

12

α

(1)

1

+ (V z

b

+L

11

)α

(1)

2

−V y

b

α

(1)

3

]j

where L

ij

is the second moment over the waterplane area. For example, L

12

=

_ _

Awp

xydS.

x

f

and y

f

are the coordinates of the center of ﬂoatation.

Following (1.10), the force can be represented by a discrete spectrum (the moment takes

an identical form and is omitted here)

F

(1)

= Re

j

F

(1)

j

e

iω

j

t

(4.18)

4.1.6 Second-order force and moment

The second-order force and moment are obtained from

F

(2)

= − ρg

_ _

S

B

(z +Z

o

)HndS −ρ

_ _

S

B

(αα α

(1)

×n)[Φ

(1)

t

+g(ξ

(1)

3

+α

(1)

1

y −α

(1)

2

x)]dS

− ρ

_ _

S

B

[

1

2

∇Φ

(1)

· ∇Φ

(1)

+ (ξξ ξ

(1)

+αα α

(1)

×x) · ∇Φ

(1)

t

]ndS −ρg

_ _

S

B

(Hx · k)ndS

+

1

2

ρg

_

WL

[η

(1)

−(ξ

(1)

3

+α

(1)

1

y −α

(1)

2

x)]

2

_

1 −n

2

z

dl

− ρgA

wp

(ξ

(2)

3

+α

(2)

1

y

f

−α

(2)

2

x

f

)k

− ρ

_ _

S

B

Φ

(2)

t

ndS

(4.19)

M

(2)

=

1

2

ρg

_

wl

[η

(1)

−(ξ

(1)

3

+α

(1)

1

y −α

(1)

2

x)]

2

_

1 −n

2

z

(x ×n)dl

− ρ

_ _

S

B

[

1

2

∇Φ

(1)

· ∇Φ

(1)

+ (ξξ ξ

(1)

+αα α

(1)

×x) · ∇Φ

(1)

t

](x ×n)dS

− ρ

_ _

S

B

(ξξ ξ

(1)

×n)[Φ

(1)

t

+g(ξ

(1)

3

+α

(1)

1

y −α

(1)

2

x)]dS

− ρg

_ _

S

B

ξξ ξ

(1)

×(αα α

(1)

×n)(z +Z

o

)dS

− ρ

_ _

S

B

αα α

(1)

×(x ×n)[Φ

(1)

t

+g(ξ

(1)

3

+α

(1)

1

y −α

(1)

2

x)]dS

− ρg

_ _

S

B

(z +Z

o

)H(x ×n)dS −ρg

_ _

S

B

(Hx · k)(x ×n)dS

− ρgi[−V ξ

(2)

2

+A

wp

y

f

ξ

(2)

3

+ (V z

b

+L

22

)α

(2)

1

−L

12

α

(2)

2

−V x

b

α

(2)

3

]

− ρgj[ V ξ

(2)

1

−A

wp

x

f

ξ

(2)

3

−L

12

α

(2)

1

+ (V z

b

+L

11

)α

(2)

2

−V y

b

α

(2)

3

]

24

− ρ

_ _

S

B

(x ×n)Φ

(2)

t

dS

We apply (4.14) and their variations to the hydrostatic pressure integrals of (4.19) and

make use of the following relations to simplify (4.19).

−ρ

_ _

S

B

(αα α

(1)

×n)[Φ

(1)

t

+ g(ξ

(1)

3

+α

(1)

1

y −α

(1)

2

x)]dS

= αα α ×F

(1)

+ρgαα α ×

_ _

S

B

(αα α

(1)

×n)(z +Z

o

)dS

= αα α ×F

(1)

+ρgV [−α

(1)

1

α

(1)

3

i −α

(1)

2

α

(1)

3

j + ((α

(1)

1

)

2

+ (α

(1)

2

)

2

)k]

(4.20)

−ρ

_ _

S

B

(ξξ ξ

(1)

×n)[Φ

(1)

t

+ g(ξ

(1)

3

+α

(1)

1

y −α

(1)

2

x)]dS

= ξξ ξ

(1)

×F

(1)

+ρg

_ _

S

B

ξξ ξ

(1)

×(αα α

(1)

×n)(z +Z

o

)dS

−ρ

_ _

S

B

αα α

(1)

×(x ×n) [Φ

(1)

t

+g(ξ

(1)

3

+α

(1)

1

y −α

(1)

2

x)]dS

= αα α

(1)

×M

(1)

+ρgαα α

(1)

×{

_ _

S

B

(ξξ ξ

(1)

×n)(z +Z

o

)dS +

_ _

S

B

[αα α

(1)

×(x ×n)](z +Z

o

)dS}

= αα α

(1)

×M

(1)

−ρgV αα α

(1)

×(ξξ ξ

(1)

× k) −ρgV αα α

(1)

×[αα α

(1)

×(y

b

i −x

b

j)]

The second-order force or moment due to Φ

(2)

is decomposed into a part due to Φ

(2)

I

+

Φ

(2)

S

and the other part due to Φ

(2)

R

as in the ﬁrst-order. Then the force and moment (4.19)

take forms

F

(2)

= F

q

+F

p

− ρgA

wp

(ξ

(2)

3

+α

(2)

1

y

f

−α

(2)

2

x

f

)k

− ρ

_ _

S

B

n

∂Φ

(2)

R

∂t

dS

(4.21)

M

(2)

= M

q

+M

p

− ρgi[−V ξ

(2)

2

+A

wp

y

f

ξ

(2)

3

+ (V z

b

+L

22

)α

(2)

1

−L

12

α

(2)

2

−V x

b

α

(2)

3

]

− ρgj[V ξ

(2)

1

−A

wp

x

f

ξ

(2)

3

−L

12

α

(2)

1

+ (V z

b

+L

11

)α

(2)

2

−V y

b

α

(2)

3

]

− ρ

_ _

S

B

(x ×n)

∂Φ

(2)

R

∂t

dS

where the subscript q denotes the force and moment due to the quadratic interaction of

the ﬁrst-order solution and p due to the second-order potential. The quadratic force and

moment are deﬁned by

25

F

q

=

1

2

ρg

_

wl

[η

(1)

−(ξ

(1)

3

+α

(1)

1

y −α

(1)

2

x)]

2

_

1 −n

2

z

ndl

− ρ

_ _

S

B

[

1

2

∇Φ

(1)

· ∇Φ

(1)

+ (ξξ ξ

(1)

+αα α

(1)

×x) · ∇Φ

(1)

t

]ndS

+ αα α

(1)

×F

(1)

−ρgA

wp

[α

(1)

1

α

(1)

3

x

f

+α

(1)

2

α

(1)

3

y

f

+

1

2

((α

(1)

1

)

2

+ (α

(1)

2

)

2

)Z

o

]k

(4.22)

M

q

=

1

2

ρg

_

wl

[η

(1)

−(ξ

(1)

3

+α

(1)

1

y −α

(1)

2

x)]

2

_

1 −n

2

z

(x ×n)dl

− ρ

_ _

S

B

[

1

2

∇Φ

(1)

· ∇Φ

(1)

+ (ξξ ξ

(1)

+αα α

(1)

×x) · ∇Φ

(1)

t

](x ×n)dS

+ ξξ ξ

(1)

×F

(1)

+αα α

(1)

×M

(1)

+ ρg(−V ξ

(1)

1

α

(1)

3

+V α

(1)

1

α

(1)

2

x

b

−V α

(1)

2

α

(1)

3

z

b

−

1

2

V ((a

(1)

1

)

2

−(a

(1)

3

)

2

)y

b

−α

(1)

1

α

(1)

3

L

12

−α

(1)

2

α

(1)

3

L

22

−

1

2

((α

(1)

1

)

2

+ (α

(1)

2

)

2

)Z

o

A

wp

y

f

)i

+ ρg(−V ξ

(1)

2

α

(1)

3

+V α

(1)

1

α

(1)

3

z

b

+

1

2

V ((a

(1)

2

)

2

−(a

(1)

3

)

2

)x

b

+α

(1)

1

α

(1)

3

L

11

+α

(1)

2

α

(1)

3

L

12

+

1

2

((α

(1)

1

)

2

+ (α

(1)

2

)

2

)Z

o

A

wp

x

f

)j

+ ρg(V ξ

(1)

1

α

(1)

1

+V ξ

(1)

2

α

(1)

2

+V α

(1)

2

α

(1)

3

x

b

−V α

(1)

1

α

(1)

3

y

b

)k

The second-order potential force and moment are given by

F

p

= −ρ

_ _

S

B

∂(Φ

(2)

I

+ Φ

(2)

S

)

∂t

ndS

(4.23)

M

p

= −ρ

_ _

S

B

(x ×n)

∂(Φ

(2)

I

+ Φ

(2)

S

)

∂t

dS

Adopting a form analogous to (1.12) the second-order force F

(2)

can be expressed as (the

moment takes an identical form),

F

(2)

= Re

i

j

F

+

ij

e

i(ω

i

+ω

j

)t

+F

−

ij

e

i(ω

i

−ω

j

)t

(4.24)

4.1.7 Haskind Exciting Force and Indirect Second-order Force

The linear wave exciting force and the second-order potential force F

p

may be evaluated not

from the scattering solution but from the appropriate component of the radiation solution.

The ﬁrst-order force evaluated in this way is referred to as the Haskind exciting force and

the second-order force as the indirect force.

26

4.1.8 Haskind exciting force

We consider a complex amplitude of a spectral component of the ﬁrst-order wave exciting

force. The force for the mode j is given by

F

j

= −iωρ

_ _

S

B

n

j

(φ

I

+φ

S

)dS (4.25)

From Green’s theorem, applied to φ

S

and a component of the radiation potential φ

j

,

we have

0 =

_ _

S

B

(

∂φ

j

∂n

φ

S

−

∂φ

S

∂n

φ

j

)dS (4.26)

It can be shown that integrals over the free surface, on the bottom and at the far ﬁeld

vanish due to the boundary conditions for φ

S

and φ

j

on these surfaces. (Newman (1977,

Section 6.18)) Thus they do not appear in (4.26).

Upon substituting (4.26) into (4.25) with the conditions (2.10) and (2.11), the Haskind

relation follows in the form

F

j

= −iωρ

_ _

S

B

(n

j

φ

I

−

∂φ

I

∂n

φ

j

)dS (4.27)

4.1.9 Indirect second-order force

We consider a complex amplitude of a spectral component of the second-order potential

force F

p

. The force for the mode j is given by

F

±

pj

= −i(ω

i

±ω

j

)ρ

_ _

S

B

n

j

(φ

±

I

+φ

±

S

)dS (4.28)

From Green’s theorem, applied to φ

±

S

and φ

±

j

, we have

0 =

_ _

S

B

(

∂φ

±

j

∂n

φ

±

S

−

∂φ

±

S

∂n

φ

±

j

)dS +

_ _

S

F

(

∂φ

±

j

∂z

φ

±

S

−

∂φ

±

S

∂z

φ

±

j

)dS (4.29)

It can be shown that the integrals on the sea bottom and at the far ﬁeld vanish due to

the boundary conditions. Upon substituting (3.18-21) into (4.29), we have

_ _

S

B

n

j

φ

±

S

dS =

_ _

S

B

Q

±

B

φ

±

j

dS +

1

g

_ _

S

F

(Q

±

IB

+Q

±

BB

)φ

±

j

dS (4.30)

Thus the force can be obtained from

F

±

pj

= −i(ω

i

±ω

j

)ρ[

_ _

S

B

(n

j

φ

±

I

+Q

±

B

φ

±

j

)dS +

1

g

_ _

S

F

(Q

±

IB

+Q

±

BB

)φ

±

j

dS] (4.31)

The radiation potential φ

±

j

with the frequency ω

i

±ω

j

is often referred to as an assisting

potential.

27

4.2 Equations of Motion

The translational and rotational motions of the body are governed by

mˆ x

gtt

= F

T

(4.32)

and

L

t

= M

T

(4.33)

In (4.32), ˆ x

gtt

is the acceleration of the center of mass of the body and F

T

is the total

force due to the ﬂuid pressure, the body mass and the external force. In (4.33), M

T

is the

total moment about the ˆ x

g

and L

t

is the time rate of change of the angular momentum in

the ˆ x coordinate system.

It is convenient to express (4.33) in the body-ﬁxed coordinate system, x, in which the

moment of inertia is time invariant. Following Ogilvie(1983),

I

g

ωω ω

t

+ωω ω ×I

g

ωω ω = TM

T

(4.34)

where I

g

is the moment of inertia about the center of mass, ωω ω

t

(ωω ω) the angular acceler-

ation (velocity) and T the tranform matrix (4.3).

If we are interested in the motion about the origin of the body coordinate system, x

o

,

which may diﬀer from the center of mass, the following relations are to be substitued into

(4.34).

I

g

ωω ω = Iωω ω −mx

g

×(ωω ω ×x

g

) (4.35)

and

M

T

= M

o

− ˆ x

g

×F

T

(4.36)

where I and M

o

are the moment of inertia and moment about o.

(4.32) and (4.34) with (4.35-6), describe the motion of the rigid body.

To derive the ﬁrst- and the second-order equations of motion, we express ˆ x

g

, F

T

, TM

T

and ωω ω in perturbation series.

ˆ x

g

= x

g

+ξξ ξ +αα α ×x

g

+Hx

g

+O(A

3

) (4.37)

F = F

(1)

+ F

(1)

B

+F

(1)

E

+F

(2)

+F

(2)

B

+F

(2)

E

+O(A

3

) (4.38)

TM

T

= M

(1)

−x

g

×F

(1)

+M

(2)

−x

g

×F

(2)

−ξξ ξ

(1)

×F

(1)

−αα α

(1)

×M

(1)

+ x

g

×(αα α

(1)

×F

(1)

) + Γ

(1)

+ Γ

(2)

+O(A

3

) (4.39)

In (4.38-39), the subscript

B

denotes the gravitational force on the body mass,

E

external

force which is assumed to be expandable in a perturbation series. The force without

28

subscript is the hydro-static and -dynamic force derived in Sections 4.1 and 4.2. Γ

(1)

and

Γ

(2)

are the ﬁrst- and second-order components of

Γ = T[M

(1)

B

+M

(2)

B

+M

(1)

E

+M

(2)

E

− ˆ x

g

×(F

(1)

E

+F

(2)

E

) +O(A

3

)] (4.40)

Equilibrium, when the body is at rest, leads to the conditions

F

(0)

+F

(0)

B

+ F

(0)

E

= 0 and M

(0)

+M

(0)

B

+M

(0)

E

= 0 (4.41)

The moment M

E

consists of a pure torque M

τ

and the moment due to the external

force

M

E

= M

τ

+ ˆ x

e

×F

E

(4.42)

The angular velocity in the ˆ x coordinate system can be obtained by linear superposition

of the contributions from time derivatives of Euler-angles.

ˆ ωω ω =

_

_

_

α

1t

0

0

_

_

_ +T

t

1

_

_

_

0

α

2t

0

_

_

_ +T

t

1

T

t

2

_

_

_

0

0

α

3t

_

_

_ (4.43)

and angular velocity in the x coordinate system, ωω ω, is obtained from

ωω ω = T

t

ˆ ωω ω (4.44)

Thus we have at the ﬁrst-order

ωω ω

(1)

=

_

_

_

α

(1)

1t

α

(1)

2t

α

(1)

3t

_

_

_ (4.45)

and at the second-order

ωω ω

(2)

=

_

_

_

α

(2)

1t

+α

(1)

2t

α

(1)

3

α

(2)

2t

−α

(1)

1t

α

(1)

3

α

(2)

3t

+α

(1)

1t

α

(1)

2

_

_

_ (4.46)

Upon substitution of (4.35-46) into (4.32) and (4.34), the equations of motion at each

order follows. These are given by

[(M +M

E

+A){

ξξ ξ

(1)

tt

αα α

(1)

tt

} + (B +B

E

){

ξξ ξ

(1)

t

αα α

(1)

t

} + (C +C

E

){

ξξ ξ

(1)

αα α

(1)

}] = {

F

(1)

exc

M

(1)

exc

}(4.47)

and

[(M +M

E

+A){

ξξ ξ

(2)

tt

αα α

(2)

tt

} + (B +B

E

){

ξξ ξ

(2)

t

αα α

(2)

t

} + (C +C

E

){

ξξ ξ

(2)

αα α

(2)

}] = {

F

(2)

exc

M

(2)

exc

}(4.48)

29

where A and B are the added mass and damping coeﬃcients and they depend on the

frequency. M

E

, B

E

and C

E

are the external force components which are proportional to

the unknown body acceleration, velocity and the motion amplitude, respectively. They

may or may not be diﬀerent for the ﬁrst and the second-order motions. M and C are the

same in (4.47) and (4.48) and they are given by

M =

_

¸

¸

¸

¸

¸

¸

¸

¸

_

m 0 0 0 mz

g

−my

g

0 m 0 −mz

g

0 mx

g

0 0 m my

g

−mx

g

0

0 −mz

g

my

g

I

11

I

12

I

13

mz

g

0 mx

g

I

21

I

22

I

23

−my

g

mx

g

0 I

31

I

32

I

33

_

¸

¸

¸

¸

¸

¸

¸

¸

_

(4.49)

C = g

_

¸

¸

¸

¸

¸

¸

¸

¸

_

0 0 0 0 0 0

0 0 0 0 0 0

0 0 ρA

wp

ρA

wp

y

f

−ρA

wp

x

f

0

0 0 ρA

wp

y

f

ρ(V z

b

+L

22

) −mz

g

−ρL

12

−ρV x

b

+mx

g

0 0 −ρA

wp

x

f

−ρL

12

ρ(V z

b

+L

11

) −mz

g

−ρV y

b

+my

g

0 0 0 0 0 0

_

¸

¸

¸

¸

¸

¸

¸

¸

_

(4.50)

Finally the force and moment on the right-hand side of (4.47) are

F

(1)

exc

= −ρ

_ _

S

B

Φ

D(1)

t

ndS + (ΣF

(1)

E

)

ex

(4.51)

M

(1)

exc

= −ρ

_ _

S

B

(x ×n)Φ

D(1)

t

dS + (ΣM

(1)

τ

+ Σx

e

×F

(1)

E

)

ex

and those on the right-hand side of (4.48) are

F

(2)

exc

= F

q

+F

p

−mH

tt

x

g

+ (ΣF

(2)

E

)

ex

(4.52)

M

(2)

exc

= M

q

+M

p

−ξξ ξ

(1)

×F

(1)

−αα α

(1)

×M

(1)

+ ρg[V ξ

(1)

1

α

(1)

3

i +V ξ

(1)

2

α

(1)

3

j −(V ξ

(1)

1

α

(1)

1

−V ξ

(1)

2

α

(1)

2

)k]

+ mg[(

1

2

((α

(1)

1

)

2

−(α

(1)

3

)

2

)y

g

−α

(1)

1

α

(1)

2

x

g

+α

2

α

3

z

g

)i

− (

1

2

((α

(1)

2

)

2

−(α

(1)

3

)

2

)x

g

+α

1

α

3

z

g

)j + (α

1

α

3

y

g

−α

2

α

3

x

g

)k]

+ x

g

×(αα α

(1)

×F

(1)

) −mx

g

×H

tt

x

g

−I

g

αα α

q

tt

− αα α

(1)

t

×I

g

αα α

(1)

t

+ Σ[Hx

e

×F

(0)

E

−αα α

(1)

×((αα α

(1)

×x

e

) ×F

(0)

E

)]

+ {Σ[M

(2)

τ

+x

e

×F

(2)

E

−αα α

(1)

×M

(1)

τ

−(x

e

−x

g

) ×(αα α

(1)

×F

(1)

E

)]}

ex

30

The subscript

ex

denotes the part of the force and moment which is not linearly propor-

tional to the motion amplitude, the velocity or the acceleration at each order. In (4.52),

αα α

q

tt

is a vector and its components are the time derivatives of the second terms of the vector

elements of (4.46).

31

REFERENCES A

Kim, M.-H. & Yue, D. K. P. 1989 ‘The complete second-order diﬀraction solution for an

axisymmetric body Part 1. Monochromatic incident waves,’ J. Fluid Mech. 200, 235–264.

Kleinman, R. E. 1982 ‘On the mathematical theory of the motion of ﬂoating bodies - an

update,’ DTNSRDC Rep 82/074.

Korsmeyer, F. T., Lee, C.-H., Newman, J. N., & Sclavounos, P. D., 1988 ‘The analysis

of wave interactions with tension leg platforms,’ OMAE Conference, Houston.

Lee, C.-H. 1988 ‘Numerical methods for the solution of three-dimensional integral equa-

tions in wave-body interactions,’ Ph.D. Thesis, Dept. of Ocean Engineering, MIT.

Lee, C.-H. & Newman, J. N. 1991 ‘First- and second-order wave eﬀects on a submerged

spheroid,’ J. Ship Research 35, No 3, 183-190.

Lee, C.-H., Newman J.N. & Zhu X. 1995 “An extended boundary integral equation for

the removal of the irregular frequency eﬀect”, Submitted for publication .

Lee, C.-H., & Zhu, X. 1993 ‘Second-order diﬀraction and radiation solutions on ﬂoating

bodies,’ 8-th International Workshop on Water Waves and Floating Bodies, St. John’s,

Newfoundland.

Martin & Rizzo 1993 ‘Boundary integral equations for bodies of small, but ﬁnite, thickness’

8-th International Workshop on Water Waves and Floating Bodies, St. John’s, Newfound-

land.

Molin, B. 1979 ‘Second order diﬀraction loads upon three dimensional bodies,’ Applied

Ocean Research 1, 197-202.

Newman, J. N. 1977 ‘Marine Hydrodynamics’. MIT Press.

Newman, J. N. 1985 ‘Algorithms for the Free-Surface Green Function’. Journal of Engi-

neering Mathematics, Vol. 19, pp. 57-67.

Newman, J. N. 1990 “Numerical solutions of the water-wave dispersion relation,” Applied

Ocean Research, 12, 14-18.

Newman, J. N. 1991 ‘WAMIT V3.1S (Second-Order Module) - Theoretical Background,’

unpublished report included in April 1991 Steering Committee Report.

Newman, J. N. 1992 ‘The approximation of free-surface Green functions,’ in ‘Wave As-

ymptotics,’ P. A. Martin & G. R. Wickham, editors, Cambridge University Press.

32

Newman, J. N. 1993 ‘Note on bodies and panels in the free surface,’ unpublished report.

Newman, J. N. 1994 ‘Wave eﬀects on deformable bodies,’ Applied Ocean Research, 16,

1, 47-59.

Newman, J. N. & Sclavounos, P. D., 1988 ‘The computation of wave loads on large

oﬀshore structures,’ 5th International Conference on the Behaviour of Oﬀshore Structures

(BOSS ‘88) Trondheim.

Ogilvie, T. F. 1983 ‘Second-order hydrodynamic eﬀects on ocean platforms,’ International

Workshop on Ship and Platform Motions, Berkeley.

WAMIT User Manual 1999 WAMIT Inc.

Wehausen, J. V. and Laitone, E. V., 1960 ‘Surface waves,’ Handbuch der Physik, 9,

446-778, Springer, Berlin.

Zhu, X. 1994 ‘Irregular Frequency Removal from the Boundary Integral Equation for the

Wave-body Problem’, Master Thesis, Dept. of Ocean Eng., MIT.

33

REFERENCES B

Korsmeyer, F. T., Lee, C.-H., Newman, J. N., & Sclavounos, P. D., ‘The analysis of wave

interactions with tension leg platforms,’ OMAE Conference, Houston, 1988.

Newman, J. N. & Sclavounos, P. D., ‘The computation of wave loads on large oﬀshore

structures,’ 5th International Conference on the Behaviour of Oﬀshore Structures BOSS

‘88 Trondheim, Norway, 1988.

Newman, J. N. ‘The numerical towing tank, fact or ﬁction?’ (Eleventh Georg Weinblum

Memorial Lecture), Ship Technology Research/Schiﬀstechnik, 36, 4, 155-168, 1989.

Emmerhoﬀ, O. J., ‘Analysis and computation of wave eﬀects on large volume structures,’

thesis prepared at M.I.T. for siv. ing. degree from NTH, Trondheim, 1989.

Lee, C.-H. & Newman, J. N. ‘First- and second-order wave eﬀects on a submerged spher-

oid,’ J. Ship Research, 1991.

Lee C.-H., Second-order wave loads on a stationary body,’ 6th International Workshop on

Water Waves and Floating Bodies, Woods Hole, 1991.

Lee C.-H., Newman, J. N., Kim, M. H., & Yue, D. K. P. ‘The computation of second-order

wave loads,’ OMAE Conference, Stavanger, 1991.

Lee, C.-H. & Newman, J. N. ‘Sensitivity of wave loads to the discretization of bodies’

BOSS ‘92, London 1992.

Newman, J. N. ‘Panel methods in marine hydrodynamics,’ Keynote Lecture, Eleventh

Australasian Fluid Mechanics Conference, Hobart, Australia 1992.

Lee, C.-H., & Zhu, X. ‘Second-order diﬀraction and radiation solutions on ﬂoating bodies,’

8-th International Workshop on Water Waves and Floating Bodies, St. John’s, Newfound-

land 1993.

Newman, J. N. ‘Deformable ﬂoating bodies,’ Eighth International Workshop on Water

Waves and Floating Bodies, St. John’s, Newfoundland 1993.

Korsmeyer, F. T., Lee, C.-H., & Newman, J. N. ‘The computation of ship interaction

forces in restricted water,’ Journal of Ship Research 1993.

Newman, J. N. ‘Wave eﬀects on deformable bodies,’ Applied Ocean Research, 16, 1, 47-59

1994.

Fathi, D., ‘Linear Hydrodynamic Loads due to Elastic Deformations of Large Volume

34

Structures’, thesis prepared at M.I.T. for siv. ing. degree from NTH, Trondheim 1993.

Fathi, D., Lee, C.-H., & Newman, J. N. ‘Computation of wave induced motions on a

ﬂexible container,’ Symposium on Hydroelasticity, Trondheim, Norway 1994.

Lee, C.-H. & Newman, J. N. ‘Second-order wave eﬀects on oﬀshore structures’ BOSS ‘94,

MIT 1994.

Ferreira, M. D. & Lee, C.-H. ‘Computation of mean forces and moments in multibody

interaction,’ BOSS ‘94, MIT 1994.

Zhu, X. & Lee C.-H. ‘Removing the irregular frequencies from wave-body interactions,’

9-th International Workshop on Water Waves and Floating Bodies, Kuju, Japan. 1994

Newman, J. N. ‘To second order and beyond,’ TLP Technology Symposium, Texas Section

of the Society of Naval Architects and Marine Engineers, Houston 1995.

Newman, J. N. & Lee, C.-H. ‘Runup on a vertical cylinder in long waves,’ Tenth Interna-

tional Workshop on Water Waves and Floating Bodies, Oxford, UK 1995.

35

APPENDICES

NOTES ON BODIES OR PANELS IN THE FREE SURFACE

JNN - September 1993

1. INTRODUCTION

These notes document recent work intended to (1) extend WAMIT to bodies where part or all of the

submerged surface is in the plane of the free surface, with zero draft; and (2) removal of irregular

frequencies by adding panels in the free surface interior to the body with an imposed homogeneous

Neumann (or Dirichlet) condition. The latter development was intended to follow the theory in

the report by Kleinman, denoted here by [K], but diﬀerences and questions have arisen which are

the principal motivation for the notes.

The notation of the WAMIT User Manual is followed. The most important diﬀerences from [K]

are the deﬁnition of the unit normal vector, here taken to be into the body and out of the ﬂuid

domain, and the Green function G, deﬁned here by equation (2.4)

G(x; ξξ ξ) =

1

r

+

1

r

+

2K

π

∞

0

dk

e

k(z+ζ)

k −K

J

0

(kR) (1.1)

r

2

= (x −ξ)

2

+ (y −η)

2

+ (z −ζ)

2

(1.2)

r

2

= (x −ξ)

2

+ (y −η)

2

+ (z + ζ)

2

, (1.3)

where J

0

(x) is the Bessel function of zero order. The correponding function γ in [K] is related to

G via the equation G = −2πγ. For simplicity the depth is assumed inﬁnite.

The velocity potential φ is governed by Laplace’s equation in the ﬂuid domain, subject to a pre-

scribed Neumann condition φ

n

= V on the body surface S

b

, the linear free surface condition

φ

z

−Kφ = 0, on z = 0 (1.4)

and a radiation condition at inﬁnity. The wavenumber is K = ω

2

/g, and g is the acceleration of

gravity.

The Green function G satisﬁes the free surface condition (1.4) with respect to both coordinate

systems. More explicitly,

G

z

−KG = 0, on z = 0 (1.5)

G

ζ

−KG = 0, on ζ = 0 (1.6)

It is not obvious that these relations can be applied in the limit when r = r

= 0, i.e. when the

source and ﬁeld points coincide on the free surface. In the vicinity of this singular point it is known

that G is of the form

G(x; ξξ ξ) =

1

r

+

1

r

−2Ke

K(z+ζ)

log(r

+|z + ζ|) + (γ −log 2) + r

+ O(r

2

log r

)

(1.7)

1

If this expansion is substituted in (1.5) or (1.6), it can be conﬁrmed that the same relations apply

even at the singular point. Thus there is no ‘delta function’ in the free-surface conditions (1.5) and

(1.6).

2 CONVENTIONAL INTEGRAL EQUATION FOR THE VELOCITY POTENTIAL

In the ‘normal’ case of a ﬂoating body, the submerged surface S

b

is entirely below the plane z = 0,

except for a line of intersection of the two surfaces. The ﬂuid domain D

+

is exterior to the body,

below the free surface, and extends to inﬁnity horizontally and vertically. The complementary

domain D

−

is inside the body, below the plane z = 0.

With the above deﬁnitions, it follows from Green’s theorem that

2πα(x)φ(x) +

S

b

φ(ξξ ξ)

∂G

∂n

ξ

dξξ ξ =

S

b

V (ξξ ξ)Gdξξ ξ. (2.1)

Here, as in [K], α is equal to 2 for points in D

+

, 1 for points on S

b

, and 0 for points in D

−

.

The deﬁnitions of α in (2.1) are consistent with the ‘jump conditions’ corresponding to the singu-

larity 1/r in G. Thus, in the limits where the point x approaches S

b

from D

±

,

I

±

(φ) = lim

x→S

±

S

b

φ(ξξ ξ)

∂

∂n

ξ

1

r

dξξ ξ = ∓2πφ +

−−−

S

b

φ(ξξ ξ)

∂

∂n

ξ

1

r

dξξ ξ (2.2)

Here the symbol

− denotes, as in the simpler Cauchy principal-value integral, that a surface of

vanishingly small area including the singular point r = 0 is excluded from the integral. This is the

appropriate interpretation of the integral on the left side of (2.1) for the case where x is on S

b

.

Thus (2.1) is rewritten in the more explicit form

2πφ(x) +

−−−

S

b

φ(ξξ ξ)

∂G

∂n

ξ

dξξ ξ =

S

b

V (ξξ ξ)Gdξξ ξ (x ∈ S

b

) (2.3)

3 INTEGRAL EQUATIONS FOR S

b

IN THE FREE SURFACE

The objective here is to extend (2.1) to the case where part or all of the body surface coincides

with the free surface. Examples of such bodies are a ﬂoating caisson where the interior ‘roof’ is in

the plane z = 0, and a ‘cicular dock’ consisting of a disk of zero draft.

In these cases, for those portions of S

b

lying in the free surface, the two Rankine singularities in the

Green function coalesce, and the eﬀect is to double the jump in (2.2). However caution is required

to correctly account for the jump when the image source 1/r

**is included. For completeness, in
**

relation to the corresponding equations [K, 51b-52b], we write the following derivatives and limits:

∂

∂z

1

r

=

−(z −ζ)

r

3

(3.1a)

∂

∂z

1

r

=

−(z + ζ)

r

3

(3.1b)

2

∂

∂ζ

1

r

=

(z −ζ)

r

3

(3.1c)

∂

∂ζ

1

r

=

−(z + ζ)

r

3

(3.1d)

Thus

∂

∂z

1

r

+

1

r

=

−2z

r

3

on ζ = 0 (3.1e)

∂

∂ζ

1

r

+

1

r

= 0 on ζ = 0 (3.1f)

In the limit z → 0− (3.1e) tends to a delta function with area 4π, but (3.1f) is identically zero

and does not contribute a jump. The former is consistent with [K] equation (51b), but the latter

contradicts the jump in (52b).

I will proceed assuming (3.1f) is correct. Two alternative derivations are given for the case where

S

b

coincides with the plane z = 0.

First start with (2.1), with the ﬁeld point in the ﬂuid domain, below the body surface, and approach

the surface at a point where the surface coincides with the plane z = 0. Prior to the limit, the free

term in (2.1) has the value 4πφ. In the limit the same value applies, since the singular contribution

to the kernel from the two Rankine singularties is

∂

∂n

ξ

1

r

+

1

r

S

b

=

∂

∂ζ

1

r

+

1

r

ζ=0

= 0 (3.1)

Thus, (2.1) is applicable to such a body surface with the understanding that α = 2 when the point

x is on the body surface, in the plane z = 0. Note in this case that the unit normal is positive

upwards, pointing from the ﬂuid domain toward the complementary domain above the free surface.

In the case of a body which is entirely in the plane z = 0, the resulting integral equation is

4πφ(x) +

−−−

S

b

φ(ξξ ξ)

∂G

∂ζ

dξξ ξ =

S

b

V (ξξ ξ)Gdξξ ξ. (x ∈ S

b

) (3.2)

As an alternative approach, consider a body with small ﬁnite draft T, e.g. with a ﬂat bottom and

vertical sides. In this case (2.3) is applicable on the exact body. Neglecting the contribution from

the vertical sides, the contribution from the image source to the left-hand side of (2.3) is

S

b

φ(ξξ ξ)

∂

∂ζ

1

r

dξξ ξ =

S

b

φ(ξξ ξ)

−(z + ζ)

r

3

dξξ ξ on z = ζ = −T (3.2a)

In the limit T → 0 the kernel tends to a delta function with area 2π, augmenting the free term in

(2.3) and thus in agreement with (3.2).

Using the free-surface boundary conditions (1.5 - 1.6), two equivalent integral equations may be

written in the forms

4πφ(x) +

−−−

S

b

φ(ξξ ξ)

∂G

∂z

dξξ ξ =

S

b

V (ξξ ξ)Gdξξ ξ. (x ∈ S

b

) (3.3)

3

and

4πφ(x) + K

S

b

φ(ξξ ξ)Gdξξ ξ =

S

b

V (ξξ ξ)Gdξξ ξ. (x ∈ S

b

) (3.4)

Equation (3.4) is most eﬀective for numerical computations since the logarithmic singularity in

(1.7) is not ampliﬁed by diﬀerentiation. WAMIT has been modiﬁed in accordance with (3.4), and

used to solve the circular dock problem. The results appear to be consistent. It is essential in this

case to use the ‘ILOG=1’ option to integrate the logarithmic singularity in (1.7) analytically over

each panel in the free surface.

Presumably the integral equations (3.2-4) have no homogeneous solutions, or irregular frequencies.

This can be inferred from the fact that the interior volume between S

b

and the free surface is zero.

4 EXISTENCE OF IRREGULAR FREQUENCIES

Hereafter the body is assumed to have nonzero submerged volume, to intersect the free surface

normally, and no portion of its surface lies in the plane z = 0 except for the normal intersection

along a waterline contour. Deﬁne S

i

as the portion of the plane z = 0 interior to S

b

. The union

of S

b

and S

i

is the closed surface S, with the unit normal deﬁned in a consistent manner to point

into the interior volume D

−

. Deﬁne an interior potential φ

which is harmonic in D

−

, and satisﬁes

the free surface condition (1.4) on S

i

. Applying Green’s theorem in the usual manner,

2π

α(x) −2

φ

(x) +

S

b

φ

(ξξ ξ)

∂G

∂n

ξ

dξξ ξ =

S

b

V

(ξξ ξ)Gdξξ ξ. (4.1)

Here V

= φ

n

is the corresponding normal velocity on S

b

.

Now assume that the interior potential φ

satisﬁes the homogeneous Dirichlet condition φ

= 0 on

S

b

. It is known that nontrivial solutions exist for certain body shapes at discrete eigenfrequencies,

and it is assumed that the same is true for general bodies. Physically, these are analogous to

‘sloshing modes’ corresponding to standing waves inside the body, except that the homogeneous

body boundary condition is Dirichlet instead of Neumann. For such solutions the normal velocity

V

is generally nonzero on S

b

, but from (4.1) and the boundary condition φ

= 0 on S

b

it follows

that

S

b

V

(ξξ ξ)Gdξξ ξ = 0 (x ∈ S

b

∪ D

+

) (4.2)

This proves that there is a discrete set of eigenfrequencies (i.e. the irregular frequencies) where the

corresponding normal velocity V

**is orthogonal to G on the body.
**

Next we deﬁne an external ‘radiation’ potential φ which satisﬁes the usual conditions outside the

body, as in §1, with the speciﬁed normal derivative φ

n

= V = V

on S

b

. This is essentially a

radiation problem with a special distribution of normal velocity on the body. The solution is

assumed to exist and to be nontrivial and unique, just as for any more conventional radiation (or

scattering) problem. However for this particular external potential the right-hand side of (2.3) will

vanish when x is on S

b

. It follows that φ is a homogeneous solution of (2.3). This proves the

existence of irregular frequencies in the context of the exterior potential problem.

4

5 EXTENDED-BOUNDARY INTEGRAL EQUATIONS

In [K] it is shown that while (2.3) may have homogeneous solutions at the irregular frequencies,

there is (at most) one solution of the integral equation (2.1) in the extended domain S = S

b

∪ S

i

.

Thus we augment (2.3) by the additional equation

S

b

φ(ξξ ξ)

∂G

∂n

ξ

dξξ ξ =

S

b

V (ξξ ξ)Gdξξ ξ (x ∈ S

i

) (5.1)

A more useful second-kind equation [K, 86] is derived which, in the present notation, takes the

form

2πΦ(x) +

−−−

S

b

Φ(ξξ ξ)

∂G

∂n

ξ

dξξ ξ +

1

2

−−−

S

i

Φ(ξξ ξ)

∂G

∂n

ξ

dξξ ξ =

S

b

V (ξξ ξ)Gdξξ ξ (x ∈ S

b

∪ S

i

) (5.2)

In [K] it is shown (using the questionable jump relation (52b)) that (5.2) has (at most) only one

solution, and that this solution is equal to φ on S

b

.

At a more pragmatic level, our numerical experience indicates that the appropriate replacement

for (5.2) is

2πΦ(x) +

−−−

S

b

Φ(ξξ ξ)

∂G

∂n

ξ

dξξ ξ −

1

2

−−−

S

i

Φ(ξξ ξ)

∂G

∂n

ξ

dξξ ξ =

S

b

V (ξξ ξ)Gdξξ ξ (x ∈ S

b

∪ S

i

) (5.3)

The only diﬀerence is the sign preceding the integral over S

i

.

In fact, the numerical tests are performed with a modiﬁed version of WAMIT where the inﬂuence

functions correspond to the slightly diﬀerent integral equations

2πΦ(x) +

−−−

S

b

Φ(ξξ ξ)

∂G

∂n

ξ

dξξ ξ +

S

i

Φ(ξξ ξ)

∂G

∂n

ξ

dξξ ξ =

S

b

V (ξξ ξ)Gdξξ ξ (x ∈ S

b

) (5.4a)

−4πΦ(x) +

S

b

Φ(ξξ ξ)

∂G

∂n

ξ

dξξ ξ −K

S

i

Φ(ξξ ξ)Gdξξ ξ =

S

b

V (ξξ ξ)Gdξξ ξ (x ∈ S

i

) (5.4b)

Equations (5.3) and (5.4) are equivalent after multiplying the unknown Φ by a factor -2 in the

domain of S

i

, and using (1.6) to replace the normal derivative of G on S

i

in (5.4b). This should be

borne in mind in judging the relevance of the statement that the numerical tests support (5.3a-b),

but not (5.2).

Note that the ﬁrst and third terms in (5.4b) are precisely the same as (minus) the left side of (3.2),

suggesting a possible connection with the absence of irregular frequencies in the dock problem.

Finally we want to show that there are no homogeneous solutions of (5.4), that is no nontrivial

solutions Φ

0

of the pair of integral equations

2πΦ

0

(x) +

−−−

S

b

Φ

0

(ξξ ξ)

∂G

∂n

ξ

dξξ ξ +

S

i

Φ

0

(ξξ ξ)

∂G

∂n

ξ

dξξ ξ = 0 (x ∈ S

b

) (5.5a)

5

−4πΦ

0

(x) +

S

b

Φ

0

(ξξ ξ)

∂G

∂n

ξ

dξξ ξ +

S

i

Φ

0

(ξξ ξ)

∂G

∂n

ξ

dξξ ξ = 0 (x ∈ S

i

) (5.5b)

For this purpose, asume that Φ

0

is a nontrivial solution of (5.5) and deﬁne the following two

potentials (which together are used in place of [K 88]):

φ

b

=

S

b

Φ

0

(ξξ ξ)

∂G

∂n

ξ

dξξ ξ (x ∈ D

−

) (5.6a)

φ

i

=

S

i

Φ

0

(ξξ ξ)

∂G

∂n

ξ

dξξ ξ = −K

S

i

Φ

0

Gdξξ ξ (x ∈ D

−

∪ S

b

∪ D

+

) (5.6b)

The following conditions apply on S

b

:

φ

b

= 2πΦ

0

(x) +

−−−

S

b

Φ

0

(ξξ ξ)

∂G

∂n

ξ

dξξ ξ (x ∈ S

b

) (5.7a)

φ

i

=

S

i

Φ

0

(ξξ ξ)

∂G

∂n

ξ

dξξ ξ (x ∈ S

b

) (5.7b)

Thus, from (5.5a),

φ

b

+ φ

i

= 0 (x ∈ S

b

) (5.8)

The following conditions apply as z → 0− on S

i

:

φ

b

=

S

b

Φ

0

(ξξ ξ)

∂G

∂n

ξ

dξξ ξ (x ∈ S

i

) (5.9a)

φ

i

=

−−−

S

i

Φ

0

(ξξ ξ)

∂G

∂n

ξ

dξξ ξ (x ∈ S

i

) (5.9b)

The free-surface condition (1.5) can be invoked to show that

φ

bz

= Kφ

b

= K

S

b

Φ

0

(ξξ ξ)

∂G

∂n

ξ

dξξ ξ (x ∈ S

i

) (5.10a)

where (5.9a) is used in the latter relation. For the analogous derivative of φ

i

, the jump associated

with (3.1e) gives an extra contribution and it follows from the last form of (5.6b) that

φ

iz

= Kφ

i

−4πKΦ

0

= K

−−−

S

i

Φ

0

(ξξ ξ)

∂G

∂n

ξ

dξξ ξ −4πKΦ

0

(x ∈ S

i

) (5.10b)

Combining (5.10a) and (5.10b), and invoking (5.5b),

φ

bz

+ φ

iz

= 0 (x ∈ S

i

) (5.11)

¿From (5.8) and (5.11), and the uniqueness proof for solutions of Laplace’s equation with combined

Dirichlet and Neumann boundary conditions, it follows that

φ

b

+ φ

i

= 0 (x ∈ D

−

) (5.12)

The remainder of the proof follows as in [K 89-95].

6

WAMIT V3.1S (Second-Order Module) – Theoretical Background

by J. N. Newman

Department of Ocean Engineering, MIT

Cambridge, MA 02139, USA

28 March 1991

1. INTRODUCTION

The WAMIT Second-Order Module, currently Version 3.1S, is designed to solve the second-

order boundary-value problem for the interaction of monchromatic or bichromatic incident

waves with a prescribed body. As in the ﬁrst-order program (Version 3.1ID), the body is

deﬁned by an ensemble of panels. In the second-order solution the boundary conditions

on both the body and free surface must be re-considered, with particular care required

in the case of the free surface. This document is intended to describe the corresponding

analysis in suﬃcient detail to explain the logic of the program. As the second-order module

is extended and reﬁned this document will be modiﬁed in parallel. The date and version

number above should be referenced in this context.

In the notation to be followed the subscripts i, j, k are used to denote the frequencies of

diﬀerent linear solutions, and , m, n are used to denote the Fourier components of the

same solutions, respectively. κ will be used for the wavenumber, to distinguish from the

integer subscript k. First-order components such as the velocity potentials φ

j

in (1.2) are

distinguished from second-order components φ

ij

by the number of subscripts.

The velocity potential is expanded in the form

Φ(x, t) = Φ

(1)

(x, t) +

2

Φ

(2)

(x, t) +· · · (1.1)

where x is a ﬁxed Cartesian coordinate system, and t denotes time. Assuming a discrete

spectrum with frequency components ω

j

> 0,

Φ

(1)

(x, t) = Re

j

φ

j

(x)e

iω

j

t

(1.2)

Φ

(2)

(x, t) = Re

i

j

_

φ

+

ij

(x)e

i(ω

i

+ω

j

)t

+ φ

−

ij

(x)e

i(ω

i

−ω

j

)t

_

(1.3)

1

The second-order potentials φ

±

ij

can be deﬁned to satisfy the symmetry relations

φ

+

ij

= φ

+

ji

and φ

−

ij

= φ

−∗

ji

(1.4)

since anti-symmetric components will not contribute to (1.3).

The free-surface boundary conditions satisﬁed by these potentials are

∂

2

Φ

(1)

∂t

2

+g

∂Φ

(1)

∂z

= 0 (1.5)

∂

2

Φ

(2)

∂t

2

+g

∂Φ

(2)

∂z

= Q(x, y; t) (1.6)

on z = 0. Here the inhomogeneous right-hand-side of the second-order free-surface condi-

tion (1.6) deﬁnes the quadratic forcing function

Q =

1

g

∂Φ

(1)

∂t

∂

∂z

(

∂

2

Φ

(1)

∂t

2

+g

∂Φ

(1)

∂z

) −

∂

∂t

(∇Φ

(1)

· ∇Φ

(1)

) (1.7)

where the right-hand-side is to be evaluated on z = 0. In the following evaluations of

second-order products of ﬁrst-order oscillatory quantities use is made of the relation

Re(Ae

iω

i

t

)Re(Be

iω

j

t

) =

1

2

Re(Ae

iω

i

t

)(Be

iω

j

t

+ B

∗

e

−iω

j

t

)

where (

∗

) denotes the complex conjugate. Adopting a form for Q analogous to (1.3),

Q(x, t) = Re

i

j

_

Q

+

ij

(x)e

i(ω

i

+ω

j

)t

+ Q

−

ij

(x)e

i(ω

i

−ω

j

)t

_

(1.8)

As in (1.4) it is appropriate to symmetrize the functions Q

±

ij

such that

Q

+

ij

= Q

+

ji

and Q

−

ij

= Q

−∗

ji

(1.9)

Combining (1.2) and (1.7) gives the expressions

Q

+

ij

=

i

4g

ω

i

φ

i

(−ω

2

j

∂φ

j

∂z

+ g

∂

2

φ

j

∂z

2

) +

i

4g

ω

j

φ

j

(−ω

2

i

∂φ

i

∂z

+g

∂

2

φ

i

∂z

2

)

−

1

2

i(ω

i

+ω

j

)∇φ

i

· ∇φ

j

(1.10)

and

2

Q

−

ij

=

i

4g

ω

i

φ

i

(−ω

2

j

∂φ

∗

j

∂z

+ g

∂

2

φ

∗

j

∂z

2

) −

i

4g

ω

j

φ

∗

j

(−ω

2

i

∂φ

i

∂z

+ g

∂

2

φ

i

∂z

2

)

−

1

2

i(ω

i

−ω

j

)∇φ

i

· ∇φ

∗

j

(1.11)

where the right-hand-sides of (1.10-11) are evaluated on z = 0. With these deﬁnitions the

free-surface boundary condition for the second-order potential is given by

−(ω

i

±ω

j

)

2

φ

±

ij

+ g

∂φ

±

ij

∂z

= Q

±

ij

(1.12)

on z = 0.

3

2. INCIDENT WAVES

The incident wave potential Φ

I

is deﬁned to be the total ﬁrst- and second-order velocity

potential that would exist in the absence of the body. Each ﬁrst-order component is deﬁned

by the amplitude A

j

, frequency ω

j

, and vector wavenumber K

j

with Cartesian components

(κ

j

cos β

j

, κ

j

sin β

j

, 0). Here β

j

is the angle of incidence relative to the x−axis. The total

ﬁrst-order velocity potential is

Φ

(1)

I

= Re ig

j

A

j

ω

j

Z(κ

j

z) exp i(ω

j

t −K

j

· x) (2.1)

Here, for a ﬂuid of depth h,

Z(κ

j

z) =

cosh(κ

j

(z +h))

cosh(κ

j

h)

(2.2)

In accordance with the ﬁrst-order free-surface condition,

κ

j

tanh(κ

j

h) = ω

2

j

/g ≡ ν

j

(2.3)

It is helpful to anticipate the relations

_

∂Z(κ

j

z)

∂z

_

z=0

= ν

j

and

_

∂

2

Z(κ

j

z)

∂z

2

_

z=0

= κ

2

j

(2.4)

Combining (2.1) and (1.10-11) gives

Q

+

ij

=−

1

2

ig

2

A

i

A

j

exp(−i(K

i

+K

j

) · x)

__

κ

2

j

−ν

2

j

2ω

j

_

+

_

κ

2

i

−ν

2

i

2ω

i

_

+

(ω

i

+ω

j

)

ω

i

ω

j

(K

i

· K

j

−ν

i

ν

j

)

_

(2.5)

and

Q

−

ij

=

1

2

ig

2

A

i

A

∗

j

exp(−i(K

i

−K

j

) · x)

__

κ

2

j

−ν

2

j

2ω

j

_

−

_

κ

2

i

−ν

2

i

2ω

i

_

−

(ω

i

−ω

j

)

ω

i

ω

j

_

K

i

· K

j

+ ν

i

ν

j

__

(2.6)

Solutions for the second-order components of the incident-wave potential follow from (1.12)

in the form

φ

±

ij

=

Q

±

ij

(x, y)Z(κ

±

ij

z)

−(ω

i

±ω

j

)

2

+gκ

±

ij

tanhκ

±

ij

h

(2.7)

4

where

κ

±

ij

= |K

i

±K

j

| (2.8)

Subsequently in the analysis of the free-surface integral of the second-order problem, a

Fourier-Bessel expansion of the ﬁrst-order incident-wave potential will be required. In

terms of the polar coordinates x = ρ cos θ, y = ρ sinθ the appropriate expansion is

φ

Ij

=

igA

j

ω

j

Z(κ

j

z)

∞

n=0

n

(−i)

n

J

n

(κ

j

ρ) cos n(θ −β

j

) (2.9)

where

0

= 1,

n

= 2 for (n ≥ 1), and J

n

is the ﬁrst-kind Bessel function of order n. It will

be convenient to redeﬁne the coeﬃcients in (2.9), writing this equation in the form

φ

Ij

= Z(κ

j

z)

∞

n=0

J

n

(κ

j

ρ)[A

c

jn

cos nθ + A

s

jn

sin nθ] (2.10)

where the coeﬃcients are

_

A

c

jn

A

s

jn

_

=

igA

j

ω

j

n

(−i)

n

_

cos nβ

j

sin nβ

j

_

(2.11)

3. VELOCITY POTENTIAL DUE TO THE BODY

In this section general properties will be listed for the solution of the velocity potential

Φ

B

due to the presence of the body. This includes both the scattered potential Φ

S

which

accounts for the eﬀects of the incident waves on the ﬁxed body, and for the radiation

potential Φ

R

due to the motions of the body about its ﬁxed mean position. In all cases

the decompositions and expansions (1.1-1.3) are applicable.

The boundary condition on the mean position S

B

of the body can be written in the generic

form

∂φ

∂n

= q

B

on S

B

(3.1)

Similarly, on the free surface,

−ω

2

φ + g

∂φ

∂z

= q

F

on z = 0 (3.2)

where ω is the frequency corresponding to the component φ.

5

The solution for φ can be derived from Green’s theorem in terms of the integral equation

2πφ(x) +

_ _

S

B

φ(ξξ ξ)

∂G(x; ξξ ξ)

∂n

ξ

dS =

_ _

S

B

q

B

(ξξ ξ)G(x; ξξ ξ)dS

+

_ _

S

F

q

F

(ξξ ξ)G(x; ξξ ξ)dS

(3.3)

In this equation φ represents any ﬁrst- or second-order velocity potential, subject to (3.1)

and (3.2) and a suitable radiation condition at inﬁnity. G is the free-surface Green function

corresponding to the same frequency ω and wavenumber as φ. Since (3.3) can be applied

to an arbitrary closed surface, this equation holds on the mean position of S

B

and on the

plane z = 0 corresponding to the mean position of S

F

, and it is not necessary to apply

(3.3) to the exact boundaries of the ﬂuid domain.

For the ﬁrst-order components of the potential q

F

= 0, and (3.3) reduces to the conven-

tional form of the integral equation for φ on the body. In the ﬂuid domain the solution for

each ﬁrst-order component of the potential can be expressed in the form

φ(x) =

_ _

S

B

σ(ξξ ξ)G(x; ξξ ξ)dS (3.4)

For the formulation based on Green’s theorem, σ is the linear operator

σ =

1

4π

_

q

B

−φ

∂

∂n

_

(3.5)

Alternatively, in the source formulation, the parameter σ is deﬁned as the ﬁrst-order source

strength, and can be found directly as the solution of the integral equation

2πσ(x) +

_ _

S

B

σ(ξξ ξ)

∂G(x; ξξ ξ)

∂n

x

dS = q

B

(x) (3.6)

In the subsequent analysis of the free-surface integral of the second-order problem, a Fourier

expansion of the ﬁrst-order body potential will be required to complement (2.9) in the far-

ﬁeld domain outside of a partition circle of radius b which is suﬃciently large to completely

surround the body. Consider a generic wavenumber κ which may be equal to either κ

i

, κ

j

,

or the wavenumber κ

k

of the second-order potential. With the additional restriction that

κb is suﬃciently large, evanascent (local) nonradiating components of the Green function

can be excluded from consideration, and

G 2πicZ(κz)Z(κζ)H

(2)

0

(κR) (3.7)

Here the function Z is deﬁned by (2.2), R is the horizontal distance between the source

and ﬁeld points (x, ξξ ξ), and H

(2)

0

(κR) = J

0

(κR) − iY

0

(κR) is the Hankel function of the

6

second kind. (In the following analysis the superscript (2) will be deleted.) The constant

c is deﬁned by

c =

ν

2

−κ

2

κ

2

h −ν

2

h + ν

cosh

2

(κh) =

−κ

2

κ

2

h −ν

2

h +ν

(3.8)

The error in the far-ﬁeld approximation (3.7) is of order exp(−CR/h) where the constant C

is greater than π/2. In the inﬁnite-depth limit the corresponding error is of order (κR)

−2

,

or of order (κR)

−3

in the special case where z = 0.

Graf’s addition theorem (Abramowitz & Stegun, eq. 9.1.79) may be used to expand (3.7)

in a Fourier-Bessel series analogous to (2.9):

G(x; ξξ ξ) 2πicZ(κz)Z(κζ)

∞

n=0

n

H

n

(κρ)J

n

(κρ

) cos n(θ −θ

) (3.9)

where ξ = ρ

cos θ

, η = ρ

sin θ

, and (3.9) is valid in the domain (ρ ≥ ρ

). Substituting

this result into (3.4) gives the corresponding far-ﬁeld representation of the body potential

φ(x) Z(κz)

∞

n=0

H

n

(κρ)[B

c

n

cos nθ +B

s

n

sin nθ] (3.10)

where the coeﬃcients in this expansion are

_

B

c

n

B

s

n

_

= 2πic

n

_ _

S

B

σ(ξξ ξ)Z(κζ)J

n

(κρ

)

_

cos nθ

sin nθ

_

dS (3.11)

4. THE FREE-SURFACE INTEGRAL

The free-surface integral which is displayed as the last term in (3.3) will be considered

here, with the ‘quadratic forcing function’ q

F

deﬁned by the relations (1.6-12). Since

we are considering the evaluation of the second-order velocity potential Φ

B

due to the

presence of the body, the incident-wave potential Φ

I

should be subtracted from the left

side of the free-surface conditions (1.6) and (1.12), and also from the evaluation of the

quadratic forcing function. Thus we seek to evaluate the diﬀerence between (1.10-1.11)

and (2.5-2.6), which includes terms on the right side of (1.10-1.11) due to (a) quadratic

interactions of the body potential with itself, and (b) cross-terms between Φ

B

and Φ

I

.

For purposes of numerical evaluation, the free-surface integral is composed of two parts,

separated by the partition circle of radius ρ = b. Following the methodology of Kim and

Yue, the integration in the inner domain ρ < b is carried out numerically. Thus the inner

free surface is discretized with quadrilateral panels, in an analogous manner to the body

surface. To avoid the evaluation of second-order derivatives of the ﬁrst-order potentials in

(1.10-11), the surface integral is transformed using Stokes’ theorem to a form involving only

7

ﬁrst-order derivatives and line integrals around the inner and outer boundaries. Details

are given in the paper by Kim.

4.1. THE FREE-SURFACE INTEGRAL IN THE FAR FIELD

In the remainder of this section the ‘far-ﬁeld integral’ in the domain (b ≤ ρ < ∞) is

considered. For this purpose the incident- and body-potentials are deﬁned by the Fourier-

Bessel expansions (2.10) and (3.10), respectively.

The ﬁrst task is to evaluate (1.10) and (1.11). We begin with the sum-frequency compo-

nent Q

+

ij

, considering ﬁrst the ‘BB’ component due to quadratic interactions of the body

potential (3.10). Associating the Fourier indices (, m) with the frequency indices (i, j)

respectively, we obtain

Q

+

BiBj

=

i

4

∞

=0

∞

m=0

_

_

ω

i

(κ

2

j

−ν

2

j

) + ω

j

(κ

2

i

−ν

2

i

) −2(ω

i

+ω

j

)ν

i

ν

j

¸

H

(κ

i

ρ)H

m

(κ

j

ρ)

(B

c

i

cos θ + B

s

i

sin θ)(B

c

jm

cos mθ + B

s

jm

sin mθ)

−2(ω

i

+ω

j

)κ

i

κ

j

H

(κ

i

ρ)H

m

(κ

j

ρ)

(B

c

i

cos θ + B

s

i

sin θ)(B

c

jm

cos mθ + B

s

jm

sin mθ)

−2(ω

i

+ω

j

)

m

ρ

2

H

(κ

i

ρ)H

m

(κ

j

ρ)

(B

s

i

cos θ −B

c

i

sin θ)(B

s

jm

cos mθ −B

c

jm

sin mθ)

_

(4.1)

where H

**denotes the ﬁrst derivative of the Hankel function with respect to its argument.
**

4.2. THE FOURIER INTEGRALS IN THE ANGULAR COORDINATE

The integration with respect to the angular coordinate θ can be evaluated after multiplica-

tion by the angular components of the Green function (3.9). Thus we consider the Fourier

integrals

_

2π

0

Q

+

ij

_

cos nθ

sin nθ

_

dθ

The only nonzero components involve the integrals

_

2π

0

cos θ cos mθ cos nθdθ =

π

n

[δ

n,|−m|

+δ

n,+m

] ≡ λ

+

mn

(4.2a)

_

2π

0

sin θ sin mθ cos nθdθ =

π

n

[δ

n,|−m|

−δ

n,+m

] ≡ λ

−

mn

(4.2b)

8

_

2π

0

cos θ sin mθ sin nθdθ =

π

[δ

,|n−m|

−δ

,n+m

] = λ

−

mn

(4.2c)

_

2π

0

sin θ cos mθ sin nθdθ =

π

m

[δ

m,|n−|

−δ

m,n+

] = λ

−

nm

(4.2d)

where δ

mn

is the Kronecker delta function, equal to one if the subscripts are equal and

otherwise equal to zero. The factors λ deﬁned above are nonzero if and only if = m+n,

or m = n + , or n = +m.

The Fourier integrals of the quadratic forcing function (4.1) are then given by

_

2π

0

Q

+

BiBj

_

cos nθ

sin nθ

_

dθ =

i

4

∞

=0

∞

m=0

_

_

ω

i

(κ

2

j

−ν

2

j

) + ω

j

(κ

2

i

−ν

2

i

) −2(ω

i

+ω

j

)ν

i

ν

j

¸

H

(κ

i

ρ)H

m

(κ

j

ρ)

_

B

c

i

B

c

jm

λ

+

mn

+ B

s

i

B

s

jm

λ

−

mn

B

s

i

B

c

jm

λ

−

nm

+ B

c

i

B

s

jm

λ

−

mn

_

−2(ω

i

+ ω

j

)

_

κ

i

κ

j

H

(κ

i

ρ)H

m

(κ

j

ρ)

_

B

c

i

B

c

jm

λ

+

mn

+ B

s

i

B

s

jm

λ

−

mn

B

s

i

B

c

jm

λ

−

nm

+ B

c

i

B

s

jm

λ

−

mn

_

+

m

ρ

2

H

(κ

i

ρ)H

m

(κ

j

ρ)

_

B

c

i

B

c

jm

λ

−

mn

+B

s

i

B

s

jm

λ

+

mn

−B

c

i

B

s

jm

λ

−

nm

−B

s

i

B

c

jm

λ

−

mn

_

__

(4.3)

The second term, which involves derivatives of the Hankel functions, can be reduced using

the relation H

ν

=

1

2

(H

ν−1

−H

ν+1

). Thus

H

(κ

i

ρ)H

m

(κ

j

ρ) =

1

4

_

H

−1

H

m−1

−H

−1

H

m+1

+ H

+1

H

m+1

−H

+1

H

m−1

_

(4.4)

Similarly, the third term resulting from the angular derivatives can be reduced using the

recurrence relation 2νH

ν

/z = H

ν−1

+ H

ν+1

to remove the factor ρ

−2

. Three alternative

relations result; the one which is most useful is obtained by setting ν equal to and then

to m:

H

(κ

i

ρ)H

m

(κ

j

ρ)ρ

−2

=

κ

i

κ

j

4m

_

H

−1

H

m−1

+H

−1

H

m+1

+H

+1

H

m+1

+H

+1

H

m−1

_

(4.5)

Substituting (4.4-5) in (4.3) gives

9

_

2π

0

Q

+

BiBj

_

cos nθ

sin nθ

_

dθ =

i

4

∞

=0

∞

m=0

_

_

ω

i

(κ

2

j

−ν

2

j

) + ω

j

(κ

2

i

−ν

2

i

) −2(ω

i

+ω

j

)ν

i

ν

j

¸

H

(κ

i

ρ)H

m

(κ

j

ρ)

_

B

c

i

B

c

jm

λ

+

mn

+ B

s

i

B

s

jm

λ

−

mn

B

s

i

B

c

jm

λ

−

nm

+ B

c

i

B

s

jm

λ

−

mn

_

−

1

2

(ω

i

+ ω

j

)κ

i

κ

j

_

H

−1

(κ

i

ρ)H

m−1

(κ

j

ρ) + H

+1

(κ

i

ρ)H

m+1

(κ

j

ρ)

_

_

(B

c

i

B

c

jm

+ B

s

i

B

s

jm

)(λ

+

mn

+ λ

−

mn

)

(B

s

i

B

c

jm

−B

c

i

B

s

jm

)(λ

−

nm

−λ

−

mn

)

_

+

1

2

(ω

i

+ ω

j

)κ

i

κ

j

_

H

−1

(κ

i

ρ)H

m+1

(κ

j

ρ) +H

+1

(κ

i

ρ)H

m−1

(κ

j

ρ)

_

_

(B

c

i

B

c

jm

−B

s

i

B

s

jm

)(λ

+

mn

−λ

−

mn

)

(B

s

i

B

c

jm

+ B

c

i

B

s

jm

)(λ

−

nm

+ λ

−

mn

)

_

_

(4.6)

The corresponding integrals for the diﬀerence-frequency function are

_

2π

0

Q

−

BiBj

_

cos nθ

sin nθ

_

dθ =

i

4

∞

=0

∞

m=0

_

_

ω

i

(κ

2

j

−ν

2

j

) −ω

j

(κ

2

i

−ν

2

i

) −2(ω

i

−ω

j

)ν

i

ν

j

¸

H

(κ

i

ρ)H

∗

m

(κ

j

ρ)

_

B

c

i

B

c∗

jm

λ

+

mn

+ B

s

i

B

s∗

jm

λ

−

mn

B

s

i

B

c∗

jm

λ

−

nm

+ B

c

i

B

s∗

jm

λ

−

mn

_

−

1

2

(ω

i

−ω

j

)κ

i

κ

j

_

H

−1

(κ

i

ρ)H

∗

m−1

(κ

j

ρ) + H

+1

(κ

i

ρ)H

∗

m+1

(κ

j

ρ)

_

_

(B

c

i

B

c∗

jm

+ B

s

i

B

s∗

jm

)(λ

+

mn

+ λ

−

mn

)

(B

s

i

B

c∗

jm

−B

c

i

B

s∗

jm

)(λ

−

nm

−λ

−

mn

)

_

+

1

2

(ω

i

−ω

j

)κ

i

κ

j

_

H

−1

(κ

i

ρ)H

∗

m+1

(κ

j

ρ) +H

+1

(κ

i

ρ)H

∗

m−1

(κ

j

ρ)

_

_

(B

c

i

B

c∗

jm

−B

s

i

B

s∗

jm

)(λ

+

mn

−λ

−

mn

)

(B

s

i

B

c∗

jm

+ B

c

i

B

s∗

jm

)(λ

−

nm

+ λ

−

mn

)

_

_

(4.7)

It is helpful to deﬁne the factors

Ω

±

ij

= ω

i

(κ

2

j

−ν

2

j

) ±ω

j

(κ

2

i

−ν

2

i

) −2(ω

i

±ω

j

)ν

i

ν

j

(4.8)

10

and

Λ

±

ij

= (ω

i

±ω

j

)κ

i

κ

j

(4.9)

(These deﬁnitions diﬀer by multiplicative factors from the corresponding parameters used

by Kim.)

Equations (4.6) and (4.7) can then be written in a more compact form, after noting that

the ﬁrst factor contained within large parentheses in each equation is equivalent to half

the sum of the second and third factors. Thus

_

2π

0

Q

+

BiBj

_

cos nθ

sinnθ

_

dθ =

i

8

∞

=0

∞

m=0

__

Ω

+

ij

H

(κ

i

ρ)H

m

(κ

j

ρ)

−Λ

+

ij

_

H

−1

(κ

i

ρ)H

m−1

(κ

j

ρ) + H

+1

(κ

i

ρ)H

m+1

(κ

j

ρ)

__

_

(B

c

i

B

c

jm

+ B

s

i

B

s

jm

)(λ

+

mn

+ λ

−

mn

)

(B

s

i

B

c

jm

−B

c

i

B

s

jm

)(λ

−

nm

−λ

−

mn

)

_

+

_

Ω

+

ij

H

(κ

i

ρ)H

m

(κ

j

ρ) + Λ

+

ij

_

H

−1

(κ

i

ρ)H

m+1

(κ

j

ρ) + H

+1

(κ

i

ρ)H

m−1

(κ

j

ρ)

__

_

(B

c

i

B

c

jm

−B

s

i

B

s

jm

)(λ

+

mn

−λ

−

mn

)

(B

s

i

B

c

jm

+ B

c

i

B

s

jm

)(λ

−

nm

+ λ

−

mn

)

_

_

(4.10)

Similarly for the diﬀerence-frequency integral

_

2π

0

Q

−

BiBj

_

cos nθ

sinnθ

_

dθ =

i

8

∞

=0

∞

m=0

__

Ω

−

ij

H

(κ

i

ρ)H

∗

m

(κ

j

ρ)

−Λ

−

ij

_

H

−1

(κ

i

ρ)H

∗

m−1

(κ

j

ρ) + H

+1

(κ

i

ρ)H

∗

m+1

(κ

j

ρ)

__

_

(B

c

i

B

c∗

jm

+ B

s

i

B

s∗

jm

)(λ

+

mn

+ λ

−

mn

)

(B

s

i

B

c∗

jm

−B

c

i

B

s∗

jm

)(λ

−

nm

−λ

−

mn

)

_

+

_

Ω

−

ij

H

(κ

i

ρ)H

∗

m

(κ

j

ρ) + Λ

−

ij

_

H

−1

(κ

i

ρ)H

∗

m+1

(κ

j

ρ) + H

+1

(κ

i

ρ)H

∗

m−1

(κ

j

ρ)

__

_

(B

c

i

B

c∗

jm

−B

s

i

B

s∗

jm

)(λ

+

mn

−λ

−

mn

)

(B

s

i

B

c∗

jm

+ B

c

i

B

s∗

jm

)(λ

−

nm

+ λ

−

mn

)

_

_

(4.11)

4.3. THE INTEGRALS IN THE RADIAL COORDINATE

In the context of the free-surface integral represented by the last term in (3.3) it is necessary

to multiply the preceding results by the Hankel functions in (3.9), and integrate in the radial

domain b < ρ < ∞. Since the diﬀerential element of surface area is ρdρdθ, the integrals

which must be evaluated in (4.10) and (4.11) are of the basic form

11

_

∞

b

H

(κ

i

ρ)H

m

(κ

j

ρ)H

n

(κ

k

ρ)ρdρ = b

2

_

∞

1

H

(αx)H

m

(βx)H

n

(γx)xdx ≡ b

2

F

mn

(α, β, γ)

(4.12)

Additional integrals will also be required to analyze the interaction between the body and

incident-wave potentials, where one of the ﬁrst pair of Hankel functions in (4.12) is replaced

by the corresponding Bessel function J

ν

, of the same order and with the same argument.

Similarly, in (4.11), the second Hankel function is replaced by its complex conjugate or the

Hankel function of the ﬁrst kind.

To make the above results more explicit, we deﬁne the following nondimensional integrals:

F

(1,2)

mn

=

_

∞

1

H

(2)

(αx)H

(1,2)

m

(βx)H

(2)

n

(γx)xdx (4.13a)

G

(1,2)

mn

=

_

∞

1

J

(αx)H

(1,2)

m

(βx)H

(2)

n

(γx)xdx (4.13b)

H

(1,2)

mn

=

_

∞

1

H

(2)

(αx)J

m

(βx)H

(2)

n

(γx)xdx (4.13c)

(The superscript for the third function is superﬂuous, but it is retained for uniformity in

the notation.) With these deﬁnitions applied, and the contributions included from the

incident-wave potential, the appropriate radial integrals are

12

_

∞

b

H

(2)

n

(κ

k

ρ)ρdρ

_

2π

0

Q

+

ij

(ρ, θ)

_

cos nθ

sin nθ

_

dθ =

i

8

b

2

∞

=0

∞

m=0

__

Ω

+

ij

F

(2)

,m,n

(κ

i

b, κ

j

b, κ

k

b) −Λ

+

ij

_

F

(2)

−1,m−1,n

(κ

i

b, κ

j

b, κ

k

b)

+F

(2)

+1,m+1,n

(κ

i

b, κ

j

b, κ

k

b)

_

_

_

(B

c

i

B

c

jm

+ B

s

i

B

s

jm

)(λ

+

mn

+ λ

−

mn

)

(B

s

i

B

c

jm

−B

c

i

B

s

jm

)(λ

−

nm

−λ

−

mn

)

_

+

_

Ω

+

ij

F

(2)

,m,n

(κ

i

b, κ

j

b, κ

k

b) + Λ

+

ij

_

F

(2)

−1,m+1,n

(κ

i

b, κ

j

b, κ

k

b)

+F

(2)

+1,m−1,n

(κ

i

b, κ

j

b, κ

k

b)

_

_

_

(B

c

i

B

c

jm

−B

s

i

B

s

jm

)(λ

+

mn

−λ

−

mn

)

(B

s

i

B

c

jm

+B

c

i

B

s

jm

)(λ

−

nm

+λ

−

mn

)

_

+

_

Ω

+

ij

G

(2)

,m,n

(κ

i

b, κ

j

b, κ

k

b) −Λ

+

ij

_

G

(2)

−1,m−1,n

(κ

i

b, κ

j

b, κ

k

b)

+G

(2)

+1,m+1,n

(κ

i

b, κ

j

b, κ

k

b)

_

_

_

(A

c

i

B

c

jm

+ A

s

i

B

s

jm

)(λ

+

mn

+ λ

−

mn

)

(A

s

i

B

c

jm

−A

c

i

B

s

jm

)(λ

−

nm

−λ

−

mn

)

_

+

_

Ω

+

ij

G

(2)

,m,n

(κ

i

b, κ

j

b, κ

k

b) + Λ

+

ij

_

G

(2)

−1,m+1,n

(κ

i

b, κ

j

b, κ

k

b)

+G

(2)

+1,m−1,n

(κ

i

b, κ

j

b, κ

k

b)

_

_

_

(A

c

i

B

c

jm

−A

s

i

B

s

jm

)(λ

+

mn

−λ

−

mn

)

(A

s

i

B

c

jm

+ A

c

i

B

s

jm

)(λ

−

nm

+ λ

−

mn

)

_

+

_

Ω

+

ij

H

(2)

,m,n

(κ

i

b, κ

j

b, κ

k

b) −Λ

+

ij

_

H

(2)

−1,m−1,n

(κ

i

b, κ

j

b, κ

k

b)

+H

(2)

+1,m+1,n

(κ

i

b, κ

j

b, κ

k

b)

_

_

_

(B

c

i

A

c

jm

+ B

s

i

A

s

jm

)(λ

+

mn

+ λ

−

mn

)

(B

s

i

A

c

jm

−B

c

i

A

s

jm

)(λ

−

nm

−λ

−

mn

)

_

+

_

Ω

+

ij

H

(2)

,m,n

(κ

i

b, κ

j

b, κ

k

b) + Λ

+

ij

_

H

(2)

−1,m+1,n

(κ

i

b, κ

j

b, κ

k

b)

+H

(2)

+1,m−1,n

(κ

i

b, κ

j

b, κ

k

b)

_

_

_

(B

c

i

A

c

jm

−B

s

i

A

s

jm

)(λ

+

mn

−λ

−

mn

)

(B

s

i

A

c

jm

+ B

c

i

A

s

jm

)(λ

−

nm

+ λ

−

mn

)

_

_

(4.14)

13

_

∞

b

H

(2)

n

(κ

k

ρ)ρdρ

_

2π

0

Q

−

ij

(ρ, θ)

_

cos nθ

sin nθ

_

dθ =

i

8

b

2

∞

=0

∞

m=0

__

Ω

−

ij

F

(1)

,m,n

(κ

i

b, κ

j

b, κ

k

b) −Λ

−

ij

_

F

(1)

−1,m−1,n

(κ

i

b, κ

j

b, κ

k

b)

+F

(1)

+1,m+1,n

(κ

i

b, κ

j

b, κ

k

b)

_

_

_

(B

c

i

B

∗c

jm

+ B

s

i

B

∗s

jm

)(λ

+

mn

+ λ

−

mn

)

(B

s

i

B

∗c

jm

−B

c

i

B

∗s

jm

)(λ

−

nm

−λ

−

mn

)

_

+

_

Ω

−

ij

F

(1)

,m,n

(κ

i

b, κ

j

b, κ

k

b) + Λ

−

ij

_

F

(1)

−1,m+1,n

(κ

i

b, κ

j

b, κ

k

b)

+F

(1)

+1,m−1,n

(κ

i

b, κ

j

b, κ

k

b)

_

_

_

(B

c

i

B

∗c

jm

−B

s

i

B

∗s

jm

)(λ

+

mn

−λ

−

mn

)

(B

s

i

B

∗c

jm

+B

c

i

B

∗s

jm

)(λ

−

nm

+λ

−

mn

)

_

+

_

Ω

−

ij

G

(1)

,m,n

(κ

i

b, κ

j

b, κ

k

b) −Λ

−

ij

_

G

(1)

−1,m−1,n

(κ

i

b, κ

j

b, κ

k

b)

+G

(1)

+1,m+1,n

(κ

i

b, κ

j

b, κ

k

b)

_

_

_

(A

c

i

B

∗c

jm

+ A

s

i

B

∗s

jm

)(λ

+

mn

+ λ

−

mn

)

(A

s

i

B

∗c

jm

−A

c

i

B

∗s

jm

)(λ

−

nm

−λ

−

mn

)

_

+

_

Ω

−

ij

G

(1)

,m,n

(κ

i

b, κ

j

b, κ

k

b) + Λ

−

ij

_

G

(1)

−1,m+1,n

(κ

i

b, κ

j

b, κ

k

b)

+G

(1)

+1,m−1,n

(κ

i

b, κ

j

b, κ

k

b)

_

_

_

(A

c

i

B

∗c

jm

−A

s

i

B

∗s

jm

)(λ

+

mn

−λ

−

mn

)

(A

s

i

B

∗c

jm

+ A

c

i

B

∗s

jm

)(λ

−

nm

+ λ

−

mn

)

_

+

_

Ω

−

ij

H

(1)

,m,n

(κ

i

b, κ

j

b, κ

k

b) −Λ

−

ij

_

H

(1)

−1,m−1,n

(κ

i

b, κ

j

b, κ

k

b)

+H

(1)

+1,m+1,n

(κ

i

b, κ

j

b, κ

k

b)

_

_

_

(B

c

i

A

∗c

jm

+ B

s

i

A

∗s

jm

)(λ

+

mn

+ λ

−

mn

)

(B

s

i

A

∗c

jm

−B

c

i

A

∗s

jm

)(λ

−

nm

−λ

−

mn

)

_

+

_

Ω

−

ij

H

(1)

,m,n

(κ

i

b, κ

j

b, κ

k

b) + Λ

−

ij

_

H

(1)

−1,m+1,n

(κ

i

b, κ

j

b, κ

k

b)

+H

(1)

+1,m−1,n

(κ

i

b, κ

j

b, κ

k

b)

_

_

_

(B

c

i

A

∗c

jm

−B

s

i

A

∗s

jm

)(λ

+

mn

−λ

−

mn

)

(B

s

i

A

∗c

jm

+ B

c

i

A

∗s

jm

)(λ

−

nm

+ λ

−

mn

)

_

_

(4.15)

14

The four factors which involve sums and diﬀerences of the parameters λ will now be

considered, with reference to the deﬁnitions (4.2). Excluding the special case = m = n =

0, where

λ

+

000

= 2π and λ

−

000

= 0

the four factors can be simpliﬁed in the forms

(λ

+

mn

+ λ

−

mn

) = πδ

,m+n

+πδ

m,+n

(4.16a)

(λ

−

nm

−λ

−

mn

) = πδ

,m+n

−πδ

m,+n

(4.16b)

(λ

+

mn

−λ

−

mn

) = πδ

n,+m

(4.16c)

(λ

−

nm

+ λ

−

mn

) = πδ

n,+m

(4.16d)

In the three-dimensional space , m, n, the ﬁrst two factors are nonzero only on the two

planes = m + n and m = + n, whereas the last two factors are nonzero only on

the complementary plane n = + m. Collectively these three planes form a cone with

triangular sections, and with its apex at the origin. The four expressions (4.16) apply also

on the intersections of each pair of adjacent planes. Thus the double sums in (4.14-15)

can be expressed in terms of the separate contributions from each of the three surfaces. A

compact form can be derived if we ﬁrst deﬁne the three functionals

U

(ij)±

m

(F) = Ω

±

ij

F

,m,−m

(κ

i

b, κ

j

b, κ

k

b)

−Λ

±

ij

_

F

−1,m−1,−m

(κ

i

b, κ

j

b, κ

k

b) +F

+1,m+1,−m

(κ

i

b, κ

j

b, κ

k

b)

¸

(4.17a)

V

(ij)±

m

(F) = Ω

±

ij

F

,m,m−

(κ

i

b, κ

j

b, κ

k

b)

−Λ

±

ij

_

F

−1,m−1,m−

(κ

i

b, κ

j

b, κ

k

b) +F

+1,m+1,m−

(κ

i

b, κ

j

b, κ

k

b)

¸

(4.17b)

W

(ij)±

m

(F) = Ω

±

ij

F

,m,+m

(κ

i

b, κ

j

b, κ

k

b)

+ Λ

±

ij

_

F

−1,m+1,+m

(κ

i

b, κ

j

b, κ

k

b) +F

+1,m−1,+m

(κ

i

b, κ

j

b, κ

k

b)

¸

(4.17c)

If the series are truncated at the upper limit = M, m = M, (4.14-15) are equivalent to

15

_

∞

b

H

(2)

n

(κ

k

ρ)ρdρ

_

2π

0

Q

+

ij

(ρ, θ)

_

cos nθ

sin nθ

_

dθ =

πi

8

b

2

_

M

=n

(1 + δ

n

)

_

U

(ij)+

m

(F

(2)

)

_

(B

c

i

B

c

jm

+B

s

i

B

s

jm

)

(B

s

i

B

c

jm

−B

c

i

B

s

jm

)

_

+ U

(ij)+

m

(G

(2)

)

_

(A

c

i

B

c

jm

+ A

s

i

B

s

jm

)

(A

s

i

B

c

jm

−A

c

i

B

s

jm

)

_

+ U

(ij)+

m

(H

(2)

)

_

(B

c

i

A

c

jm

+ B

s

i

A

s

jm

)

(B

s

i

A

c

jm

−B

c

i

A

s

jm

)

_

_

m=−n

±

M−n

=0

(1 +δ

0

)

_

V

(ij)+

m

(F

(2)

)

_

(B

c

i

B

c

jm

+ B

s

i

B

s

jm

)

(B

s

i

B

c

jm

−B

c

i

B

s

jm

)

_

+ V

(ij)+

m

(G

(2)

)

_

(A

c

i

B

c

jm

+A

s

i

B

s

jm

)

(A

s

i

B

c

jm

−A

c

i

B

s

jm

)

_

+ V

(ij)+

m

(H

(2)

)

_

(B

c

i

A

c

jm

+B

s

i

A

s

jm

)

(B

s

i

A

c

jm

−B

c

i

A

s

jm

)

_

_

m=+n

+

n−1

=1

_

W

(ij)+

m

(F

(2)

)

_

(B

c

i

B

c

jm

−B

s

i

B

s

jm

)

(B

s

i

B

c

jm

+ B

c

i

B

s

jm

)

_

+ W

(ij)+

m

(G

(2)

)

_

(A

c

i

B

c

jm

−A

s

i

B

s

jm

)

(A

s

i

B

c

jm

+A

c

i

B

s

jm

)

_

+ W

(ij)+

m

(H

(2)

)

_

(B

c

i

A

c

jm

−B

s

i

A

s

jm

)

(B

s

i

A

c

jm

+ B

c

i

A

s

jm

)

_

_

m=n−

_

(4.18)

16

_

∞

b

H

(2)

n

(κ

k

ρ)ρdρ

_

2π

0

Q

−

ij

(ρ, θ)

_

cos nθ

sin nθ

_

dθ =

πi

8

b

2

_

M

=n

(1 + δ

n

)

_

U

(ij)−

m

(F

(1)

)

_

(B

c

i

B

∗c

jm

+ B

s

i

B

∗s

jm

)

(B

s

i

B

∗c

jm

−B

c

i

B

∗s

jm

)

_

+ U

(ij)−

m

(G

(1)

)

_

(A

c

i

B

∗c

jm

+ A

s

i

B

∗s

jm

)

(A

s

i

B

∗c

jm

−A

c

i

B

∗s

jm

)

_

+ U

(ij)−

m

(H

(1)

)

_

(B

c

i

A

∗c

jm

+ B

s

i

A

∗s

jm

)

(B

s

i

A

∗c

jm

−B

c

i

A

∗s

jm

)

_

_

m=−n

±

M−n

=0

(1 + δ

0

)

_

V

(ij)−

m

(F

(1)

)

_

(B

c

i

B

∗c

jm

+ B

s

i

B

∗s

jm

)

(B

s

i

B

∗c

jm

−B

c

i

B

∗s

jm

)

_

+ V

(ij)−

m

(G

(1)

)

_

(A

c

i

B

∗c

jm

+ A

s

i

B

∗s

jm

)

(A

s

i

B

∗c

jm

−A

c

i

B

∗s

jm

)

_

+ V

(ij)−

m

(H

(1)

)

_

(B

c

i

A

∗c

jm

+ B

s

i

A

∗s

jm

)

(B

s

i

A

∗c

jm

−B

c

i

A

∗s

jm

)

_

_

m=+n

+

n−1

=1

_

W

(ij)−

m

(F

(1)

)

_

(B

c

i

B

∗c

jm

−B

s

i

B

∗s

jm

)

(B

s

i

B

∗c

jm

+ B

c

i

B

∗s

jm

)

_

+ W

(ij)−

m

(G

(1)

)

_

(A

c

i

B

∗c

jm

−A

s

i

B

∗s

jm

)

(A

s

i

B

∗c

jm

+ A

c

i

B

∗s

jm

)

_

+ W

(ij)−

m

(H

(1)

)

_

(B

c

i

A

∗c

jm

−B

s

i

A

∗s

jm

)

(B

s

i

A

∗c

jm

+ B

c

i

A

∗s

jm

)

_

_

m=n−

_

(4.19)

Special attention is required when the integers , m are equal to zero. Since the Bessel and

Hankel functions are odd functions of the (integer) order it follows that F

(1,2)

−1,m,n

= −F

(1,2)

1,m,n

,

and similarly for the case m = ±1 and for the integrals (4.13b,c). Thus it follows from

(4.17) that U

0

= W

0

and V

0m

= W

0m

. It can then be shown, with regard for the limits

of each sum in (4.18-19), and for the factors of two applied to the ﬁrst term in each of the

ﬁrst two sums, that the contributions from these equations are equivalent to those from

(4.14-15) when one or more of the integers , m, n is equal to zero.

Note also that when n = 0 the contributions from the ﬁrst two sums in (4.18-19) are

identical, giving a net contribution of twice that of the ﬁrst sum in the case of the cosine

integrals (upper factors) and zero for the sine integrals (lower factors).

4.4. EVALUATION OF THE RADIAL INTEGRALS

Suitable algorithms are required to evaluate the three families of integrals deﬁned by (4.13).

From the asymptotic approximations of the Bessel and Hankel functions it is apparent

that, for suﬃciently large values of x, the integrands are proportional to x

−1/2

exp(iσx),

17

where the factor σ involves sums and diﬀerences of the three parameters α, β, γ. These

integrals are convergent provided σ = 0. Hereafter it is assumed that the three parameters

α, β, γ > 0, and σ = 0. The former assumption is appropriate since these parameters

correspond physically to the products of three positive wavenumbers with the partition

radius b. The latter assumption will be considered in future work.

Numerical integration along the semi-inﬁnite real x-axis is not practical, due to the very

slow convergence and oscillatory feature of the integrand. These diﬃculties can be over-

come simply by adopting appropriate contours of integration in the complex z-plane, de-

pending on the sign of σ. To determine the appropriate contours for each of the integrals

(4.13), consideration must be given to the basic deﬁnitions H

(1,2)

ν

= J

ν

± iY

ν

, where the

Bessel functions J, Y are real on the real axis, and the corresponding asymptotic approxi-

mations

H

(1,2)

ν

(z)

_

2

πz

exp(±i(z −

1

2

νπ −

π

4

)) (4.20)

which are valid for |z| >> 1 and |z| >> ν. When the order is large it also is necessary to

consider the complementary approximations

J

ν

(z) ∼ z

ν

, Y

ν

∼ z

−ν

(4.21)

¿From the basic relations between the Hankel and Bessel functions the integrals (4.13c)

can be evaluated as one-half times the sum of the two integrals deﬁned by (4.13a), and a

similar approach can be followed with (4.13b). Such a decomposition is essential in some

cases for the contour integrals, since for large values of the argument J

ν

(z) includes two

components proportional respectively to exp(±iz) which may require diﬀerent choices of

the contour of integration. However when >> αx or m >> βx it is clear from (4.21)

that severe cancellation errors will result from this approach. This diﬃculty is avoided by

integrating (4.13) along a ﬁnite segment of the positive real axis 1 ≤ x ≤ x

max

, before

deforming the contour of integration into the complex plane. This procedure is robust

provided the parameter x

max

is suﬃciently large to satisfy the inequalities ≤ αx

max

and

m ≤ βx

max

.

The remaining integrals are of the general form

_

∞

x

max

f(x)e

iσx

dx (4.22)

where f(x) denotes a slowly-varying function of order x

−1/2

, for large x, and in all cases

σ = (±α ± β − γ). If σ > 0 the contour of integration in (4.22) may be deformed in the

ﬁrst quadrant of the complex plane z = x +iy, and a convenient choice is the straight line

upward parallel to the imaginary axis from z = x

max

to z = x

max

+ i∞. Thus (4.22) is

replaced in this case by the semi-inﬁnite integral

18

_

∞

x

max

f(x)e

iσx

dx = i exp(iσx

max

)

_

∞

0

f(x

max

+ iy)e

−σy

dy, (σ > 0) (4.23)

Conversely, if σ < 0, the contour of integration may be deformed in the fourth quadrant

along a similar contour parallel to the negative imaginary axis.

For programming purposes it is convenient to modify these integrals so that in all cases

the contour of integration is in the ﬁrst quadrant. Thus, when σ < 0, (4.22) is replaced by

_

∞

x

max

f(x)e

iσx

dx =

_

_

∞

x

max

f

∗

(x)e

−iσx

dx

_

∗

=

−i

_

∞

0

_

f

∗

(x

max

+ iy)e

−iσx+σy

_

∗

dy (σ < 0)

(4.24)

Note that the function f(z) is deﬁned, apart from a multiplicative exponential factor, by

the integrands of (4.13), and the conjugate function f

∗

(z) therefore involves the same

integrands except that the Hankel functions of ﬁrst- and second-kind are reversed.

In both the ﬁnite integral along the real axis and the semi-inﬁnite integrals (4.23-4) nu-

merical integration is practical using adaptive Romberg quadratures to ensure a speci-

ﬁed tolerance of accuracy. It is computationally eﬃcient to evaluate all three integrals

(4.13a,b,c) together, for all required combinations of the orders (, m, n), with the same

integration algorithm and diﬀerent integrands. The domain of integration is subdivided,

based on the maximum value of |σ|, so that large values of this parameter do not retard

convergence. The parameter x

max

is set equal to the largest value of the ratios /α or

m/β, except that if the largest ratio is less than 1.1, x

max

is set equal to one and the

integral along the real axis is skipped. For the semi-inﬁnite integrals in y, the length of the

ﬁrst segment is set equal to four divided by the maximum value of the exponential factor

σ. The length of each subsequent segment is increased in a linear manner, proportional to

the number of the segment, to take advantage of the exponentially diminishing magnitude

of the integrand in that segment. Two convergence tests are required, ﬁrst within each

ﬁnite segment to determine that the order of the Romberg quadratures is suﬃcient, and

secondly after the (converged) integral in each segment is added to the total integral to

determine when the range of integration can be truncated. Typically a tolerance of 10

−8

is

used, with absolute accuracy speciﬁed if the integral is smaller than one in absolute value,

and relative accuracy in the converse case. In the program this procedure is followed for

the complete set of required integrals, of all combinations of the orders , m, n required in

(4.18-19) and for all three of the integrals deﬁned in (4.13). Convergence tests are applied

to all members of this set simultaneously, and the integration of the entire set continues

until the convergence test is satisﬁed for the entire set. This precludes the possibility

of false convergence for one particular integrand, and permits the simultaneous recursive

evaluation of the required Bessel and Hankel functions of all orders.

Eﬀective subroutines for the Bessel and Hankel functions are required, for both the integral

along the real axis where the arguments are all real, and also in the integrals (4.23-4) where

the arguments are in the ﬁrst quadrant. These will be described separately below.

19

For the functions with real argument the functions J

ν

(x) and Y

ν

(x) are evaluated following

the general methodology described in Numerical Recipes, §6.4, and the Hankel functions

are evaluated from the relations J ± iY . For J

ν

(x) the backward recursion algorithm of

Miller (cf. Abramowitz & Stegun, pp. 385-6) is used with the starting value ν = M

chosen to be the smallest even integer M which satisﬁes the inequalities M > 1.36|x| +24

and M > N + 12, where N is the maximum order required. Normalization is based on

the Neumann series with unit value (Abramowitz & Stegun, equation 9.1.46). Numerical

experiments indicate that the accuracy of the results is at least 14 decimals, with relative

accuracy preserved for x << ν. Chebyshev expansions are used to evaluate the functions

Y

0

, Y

1

and the functions Y

ν

are evaluated for all required higher orders by forward recursion.

A similar backward recursion algorithm is used to evaluate the functions J

ν

(z) with com-

plex argument z in the ﬁrst quadrant, using the same starting value M deﬁned above

except that |x| is replaced by |z|. In this case it is appropriate to use as the normalization

relations the Neumann series for cos, sinz (Abramowitz & Stegun, equations 9.1.47-8), se-

lecting whichever of these functions has the largest absolute value to avoid instability near

the zeros on the real axis. To avoid the possibility of overﬂow, the subroutine evaluates

the normalized functions exp(iz)J

ν

(z) and applies an appropriate exponential factor to

the ﬁnal integrand.

For the Hankel functions of the ﬁrst kind, with complex argument z in the ﬁrst quadrant,

forward recursion is used based on the starting values of order zero and one. The evaluation

of these is based on two complementary algorithms, suitable for small and large values of

|z|. For small |z| ascending series are summed to evaluate J

0,1

and Y

0,1

, (Abramowitz

& Stegun equations 9.1.10-11), and these are then combined to form the corresponding

Hankel functions. These ascending series are used in the quadrilateral domain of the plane

z = x + iy such that x ≥ 0, 0 ≤ y ≤ 3 and x + y ≤ 8. The series are truncated with a

total of 22 terms. For larger values of z outside the quadrilateral domain, rational fraction

approximations of the form described by Luke (Table 66) are eﬀective. To obtain suﬃcient

accuracy in the domain where |z| is large, the coeﬃcients in Luke’s Tables 66.1 and 66.6

have been extended to n = 10, giving rational-fraction approximations as the ratios of

tenth-degree polynomials. Comparison of the two complementary algorithms along the

common boundary of the partition indicates that the minimum relative accuracy of this

subroutine is 14 decimals. The rational-fraction approximations are not eﬀective for the

Hankel functions of the second kind, but the latter can be evaluated directly from the

relation H

(2)

ν

= 2J

ν

− H

(1)

ν

, using the values of J

ν

obtained from backward recursion.

The exponential normalization factors exp(∓iz) are applied to H

(1,2)

ν

, respectively, and

corrected in the ﬁnal evaluation of the integrand.

For checking purposes it is useful to derive recursion formulae for the integrals (4.13). For

this purpose we consider the more general integrals

F

(µ)

mn

=

_

∞

1

H

(αx)H

m

(βx)H

n

(γx)x

µ

dx (4.25)

where H

ν

denotes any of the three functions J

ν

, H

(1)

ν

, or H

(2)

ν

. These integrals are deﬁned

20

in the ordinary sense for µ < 3/2 (assuming the sum of the wavenumbers is nonzero), and

will be applied ultimately with µ = 1. Integrating the derivative of the integrand and

employing the formula H

ν

= H

ν−1

−(ν/z)H

ν

to diﬀerentiate the integrand in accordance

with the chain rule, it follows that

−H

(α)H

m

(β)H

n

(γ) = αF

(µ)

−1,m,n

+βF

(µ)

,m−1,n

+γF

(µ)

,m,n−1

−(+m+n−µ)F

(µ−1)

,m,n

(4.26)

Using the recursion 2νH

ν

/z = H

ν−1

+ H

ν+1

to evaluate the last term in (4.13), with the

index ν replaced successively by , m, n, three relations are obtained as follows:

F

(µ−1)

,m,n

=

α

2

[F

(µ)

−1,m,n

+F

(µ)

+1,m,n

] =

β

2m

[F

(µ)

,m−1,n

+F

(µ)

,m+1,n

] =

γ

2n

[F

(µ)

,m,n−1

+F

(µ)

,m,n+1

]

(4.27)

Combining each of these alternative relations with (4.26) and setting µ = 1 gives the

desired relations

F

+1,m,n

=

2

α( + m+ n −1)

_

αF

−1,m,n

+ βF

,m−1,n

+ γF

,m,n−1

+H

(α)H

m

(β)H

n

(γ)

_

−F

−1,m,n

(4.28a)

F

,m+1,n

=

2m

β( +m + n −1)

_

αF

−1,m,n

+ βF

,m−1,n

+γF

,m,n−1

+ H

(α)H

m

(β)H

n

(γ)

_

−F

,m−1,n

(4.28b)

F

,m,n+1

=

2n

γ( + m+ n −1)

_

αF

−1,m,n

+βF

,m−1,n

+ γF

,m,n−1

+H

(α)H

m

(β)H

n

(γ)

_

−F

,m,n−1

(4.28c)

The relations used here for diﬀerentiation and recursion of the Hankel functions apply to

any linear combination of Bessel functions of the ﬁrst and second kind, including the Hankel

functions of the ﬁrst and second kind. Thus the same generality applies to the relations

(4.28). In particular, one of the Hankel functions may be replaced by the Bessel function

J

ν

, as is required in (4.13b-c). The only requirements here are that the starting values for

the recursion are evaluated for the appropriate combination of Bessel and Hankel functions,

and that the triple product of Hankel functions H

(α)H

m

(β)H

n

(γ) which appears in (4.28)

21

is likewise replaced by the corresponding product of the appropriate Bessel and Hankel

functions, i.e. the integrand of (4.13a,b,c) evaluated at the lower limit of integration.

Collectively the three recursion relations (4.28) can be used to evaluate the integrals F, G, H

for any combination of the three orders , m, n, provided suitable starting values are avail-

able. To proceed in this manner throughout a three-dimensional cube 0 ≤ (, m, n) ≤ M it

is suﬃcient to evaluate the eleven starting values where the three integers have the values

000, 100, 010, 001, 011, 101, 110, 111, 200, 020, 002. However this procedure has two

important disadvantages. First, to evaluate only the required ‘two-dimensional’ combina-

tions of , m, n in (4.18) and (4.19) it is not eﬃcient to evaluate the ‘three-dimensional’

combinations as is required in the recursion relations. More importantly, for large values

of α, β, γ the recursion relations are unstable with substantial cancellation error occur-

ring over several successive values of the integers. (Since forward recursion is stable for

the Hankel functions it was originally thought that this problem would not be serious.

However the inhomogeneous terms in (4.28) appear to be important in this context, and

numerical tests indicate that there is a substantial loss of accuracy in the use of (4.28) for

, m, n > O(10) unless the parameters α, β, γ = O(1).)

The recursion relations are nevertheless useful to check the subroutines used to evaluate

(4.13). Extensive tests have been made in this manner, throughout the two-dimensional

space where the three integrals (4.13) are required in (4.18-19) up to the maximum order

M = 32; with a speciﬁed Romberg tolerance (typically 10

−8

) the diﬀerences between the

left- and right-hand sides of the recursion relations (4.28) are consistently smaller than

this tolerance.

5. REFERENCES

Abramowitz, M., & Stegun, I. A., 1964 Handbook of Mathematical Functions with For-

mulas, Graphs, and Mathematical Tables, Government Printing Oﬃce, Washington and

Dover, New York.

Kim, M.-H. 1990 ‘Second-order sum-frequency wave loads on large-volume structures,’

App. Ocean Research, to appear.

Kim, M.-H. & Yue, D. K. P. 1989 ‘The complete second-order diﬀraction solution for an

axisymmetric body Part 1. Monochromatic incident waves,’ J. Fluid Mech. 200, 235–264.

Kim, M.-H. & Yue, D. K. P. 1990 ‘The complete second-order diﬀraction solution for

an axisymmetric body Part 2. Bichromatic incident waves and body motions,’ J. Fluid

Mech., 211, 557–593.

Lee, C.-H. & Newman, J. N. 1989 ‘First- and second-order wave eﬀects on a submerged

spheroid,’ J. Ship Research, to appear.

Ogilvie, T. F. 1983 ‘Second-order hydrodynamic eﬀects on ocean platforms,’ International

Workshop on Ship and Platform Motions, Berkeley.

22

CONTENTS 1. INTRODUCTION 1.1 Description of the Problem 2. THE FIRST-ORDER PROBLEM 2.1 Boundary Value Problem 2.1.1 Multiple bodies 2.1.2 Body deformations 2.1.3 Bodies near walls 2.2 Integral Equations 2.2.1 Green function 2.2.2 Linear system 2.3 Removal of the Irregular Frequencies 3. THE SECOND-ORDER PROBLEM 3.1 Boundary Value Problem 3.2 Second-order Incident Waves 3.3 Second-order Scattering Waves 4. THE FIRST- AND SECOND-ORDER FORCE 4.1 Force and Moment 4.1.1 Coordinate system 4.1.2 Coordinate transform 4.1.3 Pressure integration 4.1.4 Hydrostatic force and moment of O(1) 4.1.5 Linear force and moment 4.1.5 Second-order force and moment 4.2 Haskind Force and the Second-order Indirect Force 4.3 Equations of Motion REFERENCES A REFERENCES B APPENDICIES

Chapter 1 INTRODUCTION

This Theory Manual describes the theoretical background of the computer program WAMIT (WaveAnalysisMIT). It provides an overview of the theories and the computational methodologies in a concise manner. For detailed information on the selected subjects, some of the published and unpublished articles are included in the Appendix. The use of the program WAMIT and its capabilities are described in a separate User Manual. WAMIT is a panel program designed to solve the boundary-value problem for the interaction of water-waves with prescribed bodies in ﬁnite- and inﬁnite- water depth. Viscous eﬀects of the ﬂuid are not considered throughout and thus the ﬂow ﬁeld is potential without circulation. A perturbation series solution of the nonlinear boundary value problem is postulated with the assumption that the wave amplitude is small compared to the wave length. It is also assumed that the body stays at its mean position and, if it is not ﬁxed, the oscillatory amplitude of the body motion is of the same order as the wave amplitude. The time harmonic solutions corresponding to the ﬁrst- and second-terms of the series expansion are solved for a given steady-state incident wave ﬁeld. The incident wave ﬁeld is assumed to be represented by a superposition of the fundamental ﬁrst-order solutions of particular frequency components in the absence of the body. The boundary value problem is recast into integral equations using the wave source potential as a Green function. The integral equation is then solved by a ‘panel’ method for the unknown velocity potential and/or the source strength on the body surface. Using the latter, the ﬂuid velocity on the body surface is evaluated. For incident waves of speciﬁc frequencies and wave headings, the linear problem is solved ﬁrst. Chapter 2 describes the linear boundary value problem and the solution procedure. The quadratic interaction of two linear solutions deﬁnes the boundary condition for the second-order problem. Chapter 3 is devoted to the discussion of the second-order problem. In Chapter 4, the rigid body motion is discussed in the context of the perturbation series and the expressions for the hydrodynamic forces are derived. Chapter 5 lists the references quoted in this manual and Chapter 6 lists the references illustrating the computational results obtained from WAMIT. The Appendix includes the articles on the subjects which are not explained in detail in this manual. Since the following will focus primarily on the hydrodynamic interaction, it is appropri1

since it has only one large wetted surface. If only ﬁrst-order quantities are required.3) where ρ is the density of the ﬂuid and g is the gravitational acceleration. submerged or surface piercing. t) = −ρ( ∂Φ 1 + ∇Φ · ∇Φ + gz) ∂t 2 (1. An extremely thin submerged body may be diﬃcult to analyze when the small thickness between the two facing surfaces compared to the other dimensions of the body may render the integral equation ill-conditioned1 .1) Here t denotes the time and x=(x.2) The pressure follows from Bernoulli’s equation: p(x.ate to mention here restrictions on the bodies which may be analyzed using WAMIT. z) denotes the Cartesian coordinates of a point in space. The ﬂuid velocity is given by the gradient of the velocity potential V(x. a thin body ﬂoating on the free surface (in other words the draft is very small) is not included in this category. the bodies need not be rigid. y. The bodies may be ﬁxed. Flexible bodies can be analyzed if the oscillatory displacements of the body surface are described by speciﬁed mode shapes. However. The undisturbed free surface is the z = 0 plane with the ﬂuid domain z < 0. free of separation or lifting eﬀects. x may be expressed as the sum of the component vectors as x = xi + yj + zk. The velocity potential satisﬁes the nonlinear free-surface condition: ∂Φ 1 ∂Φ ∂Φ + 2∇Φ · ∇ + ∇Φ · ∇(∇Φ · ∇Φ) = 0 +g 2 ∂t ∂z ∂t 2 applied on the exact free surface 1 ∂Φ 1 + ∇Φ · ∇Φ)z=ζ ζ(x. Multibody interactions between any combinations of these bodies can be analyzed. a body touching the free surface may be analyzed. 1.1 Description of the Problem The ﬂow is assumed to be potential. In the extreme case of the latter.5) See Martin and Rizzo (1993) 2 . constrained or neutrally buoyant and they may be bottom mounted. t) = ∇Φ = ∂Φ ∂Φ ∂Φ i+ j+ k ∂x ∂y ∂z (1. and it is governed by the velocity potential Φ(x.4) (1. t) which satisﬁes Laplace’s equation in the ﬂuid domain: ∇2 Φ = 0 (1. y) = − ( g ∂t 2 1 (1.

Thus the total ﬁrst order potential for the wave-body interaction can be expressed by a sum of components having circular frequency ωj > 0: Φ(1) (x. the velocity potential is expanded in a form Φ(x.9) and (1. t) = Φ(1) (x. The directional spreading of the incident waves is not shown explicitly in (1. from the equations (1. t) + Φ(2) (x.8) (1. In (1.4). These boundary conditions are discussed in later chapters. y) = − 1 ∂Φ (1) g ∂t (1. the free surface elevation also takes the form of the perturbation series.10) 1 ∂Φ (2) 1 1 ∂Φ (1) ∂ 2 Φ(1) ζ (2) (x. ζ(x. In that case α is used for the rotational modes.11). Frequently ξ is used. t) = Re j φj (x)eiωj t (1.10). At the second-order. However if it is necessary. y) = − ( + ∇Φ(1) · ∇Φ(1) − ) g ∂t 2 g ∂t ∂z∂t In the equations (1. to denote only the translational modes.6) and with aid of the Taylor expansion with respect to the mean free surface. y) = ζ (1) (x.6) When the body is not ﬁxed the motion amplitude2 of the body is also expanded in a perturbation series (1. in a restrictive sense. (1.6) and (1. one can easily incorporate it by expressing φj (x) as a sum of the velocity potential components each of which corresponds to a particular wave heading. with the understanding that the real part of the time harmonic solution is physically relevant. y) + ζ (2) (x.5) and (1. Given a wave spectrum. 2 3 . y) + · · · with ζ (1) (x. the free surface boundary conditions for Φ(1) and Φ(2) imposed on z = 0 are derived. indepedent of time. it is customary to assume the spectrum is expressed as a linear superposition of the ﬁrst-order incident waves of diﬀerent frequencies.11) Here we introduce the complex velocity potential φj (x).With the assumption of a perturbation solution in terms of a small wave slope of the incident waves. For a moving body.9) (1.11) φj (x) denotes the ﬁrst-order solution in the presence of the incident wave of frequency ωj and the wave heading βj .6) and (1. From the equations (1. the total velocity potential takes a form Here ξ includes both the translational and the rotational modes of the rigid body.7) and with the Taylor expansion of Φ with respect to the mean body surface. the body boundary condition is derived.8-10). t) + · · · (1. the right-hand sides are evaluated on z = 0.7) ξ = ξ (1) + ξ (2) + · · · From (1.

These are determined by two linear incident wave components: one has frequency ωj and wave heading βj and the other ωi and βi . we discuss the solution procedures for the complex velocity potentials φj and φ± in (1. In the next two chapters. If one consider the directional spreading of the incident waves. φ± (x) are represented by a double summation ij over wave headings (βi . βj ).13) Here we may assume ωi ≥ ωj ≥ 0 without loss of the generality.12).and diﬀerence-frequency velocity potential with frequencies ωi + ωj and ωi − ωj .11) and (1. t) = Re i j φ+ (x)ei(ωi +ωj )t + φ− (x)ei(ωi −ωj )t ij ij (1.Φ(2) (x. φ+ and φ− are ij ij referred to as the sum. ij 4 . respectively.12) with symmetric conditions φ+ = φ+ ij ji and φ− = φ−∗ ij ji (1.

5) . Thus we omit the subscript j indicating the frequency component in this Chapter. we review the ﬁrst-order boundary value problem and its solution procedure. For a ﬂuid of depth h.1) ω representing a plane progressive wave of the circular frequency ω and the wave heading angle β. Linearity of the problem allows it to be distinguished from the disturbance due to the the motion of the body which is represented by the radiation potential φR .4) An eﬃcient root ﬁnding algorithm for the computation of κ from equation (2. β is an angle of incidence to the positive x − axis. The velocity potential of the ﬁrst-order incident wave is deﬁned by igA Z(κz)e−iκ(x cos β+y sin β) (2.12). it is given by g cosh(κ(z + h)) (2. ωj from the discrete spectrum (1. The scattering velocity potential φS represents the disturbance to the incident wave due to the presence of the body in its ﬁxed position. Thus the total velocity potential is given by φ = φI + φS + φR = φD + φR 5 (2.Chapter 2 THE FIRST-ORDER PROBLEM In this Chapter. A denotes the complex wave amplitude.4) is explained in Newman (1990). where κ = (2. Here we consider one particular frequency component.3) Z(κz) = cosh(κh) ω2 κ tanh κh = g where the wavenumber κ is the real root of the dispersion relation (2.2) ω2 ≡ ν. The function Z represents the depth dependence of the ﬂow and is given by φI = Z(κz) = eκz for inﬁnite water depth.

the linear free-surface condition: φz − νφ = 0 on z = 0. pitch and yaw in the increasing order of j. These modes are referred to as surge. roll. Finally. and a condition on the sea bottom: ∇φ → 0 as z → −∞.1 Boundary Value Problem The total ﬁrst-order velocity potential φ. along with each of its components. or φz = 0 on z = −h (2.9a) (2.10) (2. sway. the conditions on the body surface complete the description of the boundary value problem. heave.and ﬁnite.11) (2. 2. They take the form ∂φk = nk ∂n and ∂φI ∂φS =− ∂n ∂n (2. the scattering and the radiation potentials are subject to a radiation condition stating that the wave energy associated with the disturbance due to the body is carried away from the body in all direction in the far ﬁeld.where the diﬀraction potential φD is deﬁned to be the sum of φI and φS .7) 6 . In addition. and φk the corresponding unit-amplitude radiation potential (speciﬁcally.water depth. respectively.6) Here ξk is the complex amplitude of the oscillatory motion in mode k of the six degrees of freedom.8) (2. unitamplitude means the unit-amplitude linear or angular velocity of the rigid body motion). satisﬁes the Laplace equation in the ﬂuid domain: ∇2 φ = 0. The radiation potential itself is a linear combination of the components corresponding to the modes of motion such that 6 φR = iω k=1 ξ k φk (2.9b) for inﬁnite.

1). 1 7 . n5 . Then the corresponding radiation potential φk satisﬁes ∂φk = uk · n (2. Examples are the bending modes of structural deformation and the compressible modes of a ﬂexible body.1 These topics are reviewed brieﬂy next. the total radiation potential consists of 6N components. This is done by deﬁning φk as the velocity potential corresponding to a particular mode of one body while the other bodies are kept stationary. the velocity potential is subject to the condition (2. 2.12) Variations to the canonical problem speciﬁed above include multibody interactions.2 Body deformations Appication can also be made to analyze the more general modes of motion of the structure. the mode shapes are described in terms of the vector displacement of the body boundary. 2. along with the exploitation of the symmetry of the bodies. n2 . The resultant wave ﬁeld Walls are assumed to coincide with the x = 0 or the y = 0 planes. The unit vector n is normal to the body boundary and it is assumed the normal vector points out of the ﬂuid domain. In this way. it follows that ∂φD =0 ∂n (2.3 Bodies near walls ∂φ =0 ∂n When walls are present. generalized modes and the presence of the vertical walls. 2. beyond the rigid body motions. When there are two walls they meet perpendicularly at the origin. From (2.1 Multiple bodies The decomposition of the radiation potential into components.13) ∂n on the body boundary. To analyze these modes. The details of the analysis for generalized modes and its applications are described in Newman (1994). corresponding to the modes of the rigid body motion.where (n1 .1. where N is the number of bodies. can be extended to multi-body interaction.5). n6 ) = x × n.14) on the walls.1. Let uk (x) be the mode shape given as a function of the points on the body boundary. a reﬂected wave (φr ) should be I added to φI to describe the wave ﬁeld in the absence of the body.1. n3 ) = n and (n4 . If we deﬁne the incident wave as in (2. x is the position of a point on the body boundary.

17) (2. (2. x) (2. the symmetric component satisﬁes ∂φ =0 ∂n and the antisymmetric component satisﬁes φ=0 (2. the computational domain for φD .4 Integral Equations The boundary value problem is solved by the integral equation method.19) 2πφD (x) + = 4πφI (x) ∂nξ SB Alternatively. However when these equations are solved using a low order panel method.18) SB SB where SB denotes the body boundary. From a computational point of view. x) = ∂nξ ξ dξ (− ∂φI ξ )G(ξ .15) If the body geometry is symmetric with respect to the x = 0 plane.satisﬁes (2. φS and φR can be reduced to half or one quadrant of the entire domain. the velocity potentials are expressed as a sum of the symmetric and antisymmetric components with respect to the planes of symmetry. This is due to the hypersingular integral arising from the double derivative of the constant strength wave 8 .14) and it is a standing wave with maximum free surface elevation 2|A|.18) to (2. We need to evaluate the ﬂuid velocity on the body surface and in the ﬂuid domain for the computation of the second-order wave forces as discussed in Chapter 4. Thus the velocity potential φk on the body boundary is obtained from the integral equation 2πφk (x) + ξ ξ dξ φk (ξ ) ξ ∂G(ξ .20) SB SB and the diﬀraction potential follows from (2. x) = ∂nξ ξ ξ dξ nk G(ξ .20).5).) The total velocity potential is then expressed in the form φ = φI + φr + φS + φR I (2. x) ξ ξ dξ φD (ξ ) (2.20) due to the simpler form of the right-hand-side. they can not predict the velocity accurately on or close to the body boundary. the scattering potential can be obtained from 2πφS (x) + ξ ξ dξ φS (ξ ) ξ ∂G(ξ .19) is slightly more eﬃcient than (2. On these planes. the y = 0 plane or both. The velocity may be computed from the spacial derivatives of Green’s integral equations. In these cases. x) ∂n (2. (2.16) 2. (The actual free surface elevation depends on the incident wave heading β.1. The corresponding equation for the total diﬀraction velocity potential φD is (Korsmeyer et al (1988)) ξ ∂G(ξ .

1. which is referred to as the wave source potential.5 The Green Function The Green function G(x. the discrete form of the equation (2.26) r 2 = (x − ξ)2 + (y − η)2 + (z − ζ)2 9 .22) are solved by the panel method.21) SB and that of σS corresponding to the scattering potential φS 2πσS (x) + SB ξ ξ dξ σS (ξ ) ξ ∂G(ξ .22) The ﬂuid velocity on the body boundary or in the ﬂuid domain due to φk or φS is then obtained from ∇φ(x) = ∇ SB ξ ξ ξ dξ σ(ξ )G(ξ . The integral equation for the source strength σk corresponding to the radiation potential φk takes the form (Lee and Newman (1991)) 2πσk (x) + ξ ξ dξ σk (ξ ) ξ ∂G(ξ .1). The wetted body surface is represented by an ensemble of quadrilateral panels (a triangular panel is a special type of quadrilateral panel where two vertices coalesce). (2.source potential distributed on a quadrilateral panel.20) takes the form N EQN 2πφS (xk ) + n=1 N EQN φS (xn ) − n=1 Sn ξ dξ ξ ∂G(ξ . x) = nk ∂nx (2. For example. and in inﬁnite water depth it is deﬁned by (Wehausen and Laitone (1960)) G(x.25) (2.24) = ∂φI (xn ) ∂n Sn where NEQN is the total number of panels (unknowns) and xk are the coordinates of the centroid of the k-th panel. ξ ) = 1 1 2ν + + r r π ∞ 0 dk ek(z+ζ) J0 (kR) k−ν (2.23) The ﬂuid velocity due to the incident wave is evaluated directly from (2. x) (2. xk ) (2. ξ ). For this reason. 2. x) ∂φI =− ∂nx ∂n (2.18) to (2. Integral equations. xk ) ∂nξ ξ ξ dξ G(ξ . It satisﬁes the free-surface and radiation conditions. the ﬂuid velocity is computed based on the source formulation. The unknowns are assumed to be constant over each panel and the integral equation is enforced at the centroid of each panel (a collocation method). is the velocity potential at the point x due to a point source of strength −4π located at the point ξ .

The integration of the latter part over a panel is carried out using either one or four point Gauss quadrature. it is deﬁned by 1 1 + +2 r r ∞ 0 (2.6 Linear system The linear system. In both expressions (2. in (2. is O( ). The relative computational eﬃciency between the iterative method and Gauss elimination can be illustrated by the ratio of the number of ﬂoating point operations.27) G(x. In the equation (2.30) may not be evaluated accurately using Gauss quadrature in this situation. The former is attached in the appendix. the Fourier k−integration is indented above the pole on the real axis in order to enforce the radiation condition. The remaining wave part of the Green function is evaluated based on the algorithms described in Newman (1992). the inﬂuence due to the continuous distribution of the Rankine part of the wave source potential on a quadrilateral panel is evaluated based on the algorithms described in Newman (1985). 2.28). the former to the 3KM latter. The details on the iterative method is described in Lee (1988). The logarithmic singularity in (2. Where NEQN is the number of unknowns or the dimension of the NEQN linear system and K is the number of iterations. In ﬁnite depth. the velocity potentials are subject to the conditions (2. The Green function is then modiﬁed to be a sum of (2. generated by implementing the solution procedure described above (for example (2.24) the number of the 10 . The algorithm for evaluating the inﬂuence of the logarithmic singularity distributed over a panel can be found in Newman and Sclavounos (1988).24)). special care is required for the evaluation of Green function as is discussed in Newman (1993) and Zhu (1994).17).28) (2.r 2 = (x − ξ)2 + (y − η)2 + (z + ζ)2 where J0 (x) is the Bessel function of zero order. G takes a form G(x.29) (r )2 = (x − ξ)2 + (y − η)2 + (z + ζ + 2h)2 .14). When there are walls or when we exploit the symmetry/antisymmetry of the ﬂow.24). (2. In this case.30) where γ is the Euler constant. When the panel is on the free surface.1.25) and (2. ξ ) = 1 1 + − 2νeν(z+ζ) (log(r + |z + ζ|) + (γ − log 2) + r + O(r log r )) r r (2. (As an example. This ratio.16) or (2. a block iterative method or Gauss elimination.15) or (2. ξ ) = dk (k + ν) cosh k(z + h) cosh k(ζ + h) −kh e J0 (kR) k sinh kh − ν cosh kh (2. When the ﬁeld and source points are close together and in the vicinity of the free surface. M is the number of the right-hand sides of the linear system sharing a left-hand side. is solved by an iterative method.28) and its image sources. the image source must be placed at the reﬂected points of x with respect to the planes of symmetry.

Si .31a) SB Sf SB SB ξ ξ dξ φk (ξ ) Sf ξ ξ dξ φk (ξ ) SB (2. x) + ∂nξ ξ ξ dξ φk (ξ ) ξ ∂G(ξ . The opposite limit is the case when the dimension of the blocks is one. which is identical to the iterative method. 11 . x) = ∂nξ ξ ∂G(ξ . This accelerates the rate of convergence as the dimension of the block increases.31b) where φk is an artiﬁcial velocity potential deﬁned in the interior domain. Some examples are i) barges with shallow draft.18-20). Gauss elimination is not only slower than the iterative method for large NEQN but also requires suﬃciently large core memory to store the entire matrix elements. The block iterative method is based on the same algorithm as the iterative method. x) = ∂nξ ξ ξ dξ nk G(ξ . Equations (2.18-22) have nonunique solutions at the common irregular frequencies corresponding to the Dirichlet eigen-frequencies for the closed domain deﬁned by the body boundary. ii) bodies with large ﬂare and iii) multiple bodies separated by small gaps. all or part of the matrix may be stored on the auxiliary storage devise (hard disk) and then retrieved at each iteration. Numerical solutions of these equations are erroneous near the irregular frequencies. the back substitution is performed for the diagonal blocks. since only φk on SB is physically relevant. x) ξ ξ dξ nk G(ξ . The eigenmodes correspond to the solution with the homogeneous Dirichlet condition on SB and the linear free-surface condition on Si . Another case where the iterative method may be slowly convergent is in the linear system for the extended boundary integral equation (see Section 2. in the iterative method. The extended boundary integral equations for φk are 2πφk (x) + −4πφk (x) + ξ ξ dξ φk (ξ ) ξ ∂G(ξ . For these problems. These two equations are solved simultaneously for φk on SB and φk on Si .) In most applications the iterative method converges in 10-15 iterations and is the most eﬃcient way to solve the linear system. We discard φk after we solve the equations. x) + ∂nξ ξ ∂G(ξ . There are some problems for which the iterative method is slowly convergent or nonconvergent due to bad conditioning of the linear system. On the other hand.right-hand-side vectors is the same as the number of the diﬀerent wave headings of φI for a ﬁxed left-hand-side matrix. The block iterative method requires core memory to store the elements of one block for LU decomposition.7 Removal of Irregular Frequencies The integral equations (2. since the LU decomposition is performed using the core memory.3). Gauss elimination or the block iterative method should be used. and the interior free surface. The extended boundary integral equation method is applied to remove the irregular frequency eﬀect from the velocity potentials in (2.31a) and (2.1. x) (2.31b) are for x on SB and Si . but Gauss elimination is applied locally for the speciﬁed diagonal blocks. respectively. The limiting case is the same as Gauss elimination. 2. At each stage of the iterations.

Zhu (1994) also makes a comparison with other methods for the removal of the irregular frequency eﬀect and demonstrates the eﬀectiveness of the present method.32b) ∂φI ∂n The scattering potential takes the same form as the radiation potential but − replaces nk on the right-hand side of the equations (2. The proper ∂n condition of the function V (x) is discussed in Lee et al (1995). x) = 4πφI (x) ∂nξ ξ ∂G(ξ . x) + ∂nξ Sf ξ ξ dξ σ (ξ ) ξ ∂G(ξ .33b) ∂φI ) for the radition (scattering) problem. where g(x) = nk (g(x) = − 12 . x) + ∂nξ ξ ξ dξ σ (ξ ) ξ ∂G(ξ . x) = 4πφI (x) ∂nξ 2πφD (x) + SB ξ ξ dξ φD (ξ ) Sf ξ ξ dξ φD (ξ ) (2.31a) and (2.32a) −4πφD (x) + SB ξ ξ dξ φD (ξ ) Sf ξ ξ dξ φD (ξ ) (2.The equations for the diﬀraction potential φD are given by ξ ∂G(ξ .32b). Lee et al (1995) and Zhu (1994).33a) SB Sf −4πσ (x) + SB ξ ξ dξ σ(ξ ) ξ ∂G(ξ . x) + ∂nξ ξ ∂G(ξ . The details of the derivation of these equations are described in Kleinman (1982). x) + ∂nξ ξ ∂G(ξ . The extended boundary integral equation for the source formulation of the radiation or scattering problem takes a form 2πσ(x) + ξ ξ dξ σ(ξ ) ξ ∂G(ξ . x) = −V (x) ∂nξ (2. x) = g(x) ∂nξ (2.

the second-order potential. Section 6. t) (3.11)1 . In this Chapter we restrict our attention to the solution of the second-order boundary value problem.4) When body is not ﬁxed. The forcing function on the body boundary is given by (Ogilvie (1983)) QB = − ξ ∂H ∂ΦI ∂(ξ (1) + α(1) × x) α +n· x + (α(1) × n) · ( − ∇Φ(1) ) ∂n ∂t ∂t ξ −n · ((ξ (1) + α(1) × x) · ∇)∇Φ(1) + 1 (2) ∂ξk nk k=1 ∂t 6 (2) (3.Chapter 3 THE SECOND-ORDER PROBLEM From the quadratic interaction of the two linear wave components of the frequencies ωi and ωj in the discrete spectrum (1. t) is subject to the free surface condition ∂Φ(2) ∂ 2 Φ(2) = QF (x.12). 13 . t) +g ∂t2 ∂z and the body boundary condition (3.4)) QF = ∂ 1 ∂Φ(1) ∂ ∂ 2 Φ(1) ∂Φ(1) ( ) − (∇Φ(1) · ∇Φ(1) ) +g g ∂t ∂z ∂t2 ∂z ∂t (3. Other second-order quantities.4) below. The second-order velocity potential Φ(2) (x. are discussed in Chapter 4. we have the second-order waves of the frequency components ωi + ωj and ωi − ωj in (1.1) ∂Φ(2) = QB (x. obtainable from the ﬁrst-order solution.e. such as the quadratic second-order force . i.2) ∂n where the inhomogeneous right-hand-side of the second-order free-surface condition (3. y.3) QF is to be evaluated on z = 0. the ﬁrst-order motion aﬀects the second-order solution as shown in the equation (3.1) deﬁnes the quadratic forcing function (Newman (1977.

ΦI is referred to as the second-order incident wave potential. With this in mind.4) is the quadratic components of the coordinate transformation matrix due to the rotation of the body and it (2) takes a form (3. t) = Re i j Q+ (x)ei(ωi +ωj )t + Q− (x)ei(ωi −ωj )t ij ij (3.12).3).8) and ∗ ∂ 2 φ∗ i i ∂ 2 φi j 2 ∂φj 2 ∂φi ωi φi (−ωj + g 2 ) − ωj φ∗ (−ωi +g 2 ) j 4g ∂z ∂z 4g ∂z ∂z 1 − i(ωi − ωj )∇φi · ∇φ∗ (3.6) with the symmetry condition (3. Thus we may decompose Q± into the quadratic interactions of the two incident wave potentials F Q− = F 14 . From combining (1. we have the expressions for the complex amplitudes of the free-surface forcing functions. Q is expressed as Q(x. In (3. we will omit the subscript ij hereafter.14) and (3. In accordance with the deﬁnition of the second-order potential (1. consist of the incident wave potential (φI ) and the body disturbances (φB = φS + φR ).and diﬀerency-frequency problems.and diﬀerence. use is made of the relation 1 Re(Aeiωi t )Re(Beiωj t ) = Re[(Aeiωi t )(Beiωj t + B ∗ e−iωj t )] (3. nk is the appropriate component of n for the translational modes and of a component of (x×n) for the rotational modes.7) ± The other second-order quantities such as the motion amplitude ξ (2) = ξij are expressed in a form of (3.15).9) j 2 For future reference.7).8-9).6) With the symmetry condition.4).frequency forcings on z = 0 are given by Q+ = F ∂ 2 φj ∂ 2 φi i i 2 ∂φj 2 ∂φi ωi φi (−ωj + g 2 ) + ωj φj (−ωi +g 2 ) 4g ∂z ∂z 4g ∂z ∂z 1 − i(ωi + ωj )∇φi · ∇φj 2 (3. we note that the ﬁrst-order potential φi and φj in (3. This is the second-order potential in the absence of the body and will be discussed in Section 3. Q+ = Q+ ij ji and Q− = Q−∗ ij ji (3.2. In the following evaluations of second-order products of two ﬁrst-order oscillatory quantities.and it is to be evaluated on the mean body boundary.11) and (3.5) 2 where (∗ ) denotes the complex conjugate. for sum. Sum. H in (3.

15) As in the ﬁrst-order problem. it is convenient to decompose the total second-order potential into three components: the second order incident wave potential (φ± ). These are proportional to the second-order motion and can be treated separately from the rest of body forcing as is discussed below. φ± describes the disturbance due the second-order motion of the body and R 15 . We S R deﬁne φ± as the potential that satisﬁes the second-order free surface condition in absence I of the body. and two body disturbance waves II IB (Q± ) such that BB (3. the secondI order scattering wave potential (φ± ) and the second-order radiation potential (φ± ). the incident and the body disturbance waves (Q± ).13) are omitted from (3. since they are not a quadratic function of the ﬁrst-order solution.4). The matrices H ± which account for the rotational motion of the body are given by −(α2i α2j + α3i α3j ) 0 1 + −α1i α1j + α3i α3j H = α1i α2j + α1j α2i 2 α2i α3j + α2j α3i α1i α3j + α1j α3i ∗ ∗ −(α2i α2j + α3i α3j ) 0 1 − ∗ ∗ ∗ ∗ H = α1i α2j + α1j α2i −α1i α1j + α3i α3j 2 ∗ ∗ ∗ ∗ α2i α3j + α2j α3i α1i α3j + α1j α3i 0 0 −α1i α1j + α2i α2j 0 0 ∗ ∗ −α1i α1j + α2i α2j (3.and diﬀerence-frequency components of the last term of (3.(Q± ). They are given by Q+ = − B ∂φ+ i(ωi + ωj ) I + n · H +x ∂n 2 i ξ α ξ α + [(αi × n) · (ωj (ξ j + αj × x) − ∇φj ) + (αj × n) · (ωi (ξ i + αi × x) − ∇φi )] 4 1 ξ ξ − (3.10) Q± = Q± + Q± + Q± F II IB BB Next we consider the sum. Q± = i(ωi ± ωj ) B k ξ ± nk (3.12) j j 4 The sum.and diﬀerence-frequency forcing on the body boundary.11-12).11) n · [((ξ i + α i × x) · ∇)∇φj + ((ξ j + α j × x) · ∇)∇φi ] 4 and Q− = − B ∂φ− i(ωi − ωj ) I + n · H −x ∂n 2 i α ξj αj ξ [(αi × n) · (ωj (ξ ∗ + α∗ × x) − ∇φ∗ ) + (α∗ × n) · (ωi (ξ i + αi × x) − ∇φi )] + j j 4 1 ξ ξj n · [((ξ i + αi × x) · ∇)∇φ∗ + ((ξ ∗ + α ∗ × x) · ∇)∇φi ] − (3.14) (3.

is linearly proportional to the motion amplitude. ω j )2 φ ± S ∂φ± + g S = Q± + Q± IB BB ∂z on z = 0 (3.21) B ∂n At the far ﬁeld φ± satisﬁes the same radiation condition as that for φR of the ﬁrst-order.16) where the unit-amplitude radiation velocity potential φk is deﬁned in the same way as φk in Chapter 2. The rest of the second-order potential is deﬁned to be φ± .28). The boundary condition on the free surface.1 Boundary Value Problem The total second-order potential φ± satisﬁes Laplace’s equation in z < 0 and the condition on the sea bottom (equations (2. R For φ± .9a-9b)). Thus based S on the asymptotic behaviours of G and Q± (in other words.7) and (2.18) ∂φk = nk ∂n (3. the scattering potential is subject to −(ωi ± and ∂φ± S = Q± (3. we apply the ‘weak radiation condition’ suggested by Molin (1979).20) 16 . on the body boundary and at the far ﬁeld are speciﬁed below for each component potential. the surface integral at the far ﬁeld can be shown to vanish with the result shown in (3. The incident wave potential is subject to −(ωi ± ωj )2 φ± + g I ∂φ± I = Q± II ∂z on z = 0 (3.17) The unit amplitude radiation potential is subject to −(ωi ± ωj )2 φk + g and ∂φk = 0 on z = 0 ∂z (3. from the behaviours of φi and F φj ). This decomposition is in accordance with the convention for the ﬁrst-order problem where the wave exciting force is the pressure force due to the sum of the incident and scattering waves. The detailed analysis is in Molin (1979). 3. It can be decomposed into the mode dependent components φ± = i(ωi ± ωj ) R 6 ξ k φk k=1 (3. whether the S body is ﬁxed or moving. Finally.19) on the mean body boundary SB .

y)Z(κ± z) ij II = 2 + gκ± tanh κ± h −(ωi ± ωj ) ij ij (3.26) Note that in inﬁnite water depth. φ± . is obtained from Green’s integral S equation 17 .23) Combining (3.25) Solutions for the second-order incident-wave potential follow from (3. Here β is the angle of incidence relative to the x−axis.1) can be written in the form igA Z(κz)e−iK·x ω On z = 0. κ sin β. I 3.27) Q± (x. The ﬁrst-order incident wave potential (2.17) in the form φ± I where κ± = |Ki ± Kj | ij (3.22) =ν and ∂ 2 Z(κz) ∂z 2 z=0 = κ2 (3.and the second-derivatives of Z in vertical direction are φI = ∂Z(κz) ∂z z=0 (3. and vector wavenumber K with Cartesian components (κ cos β.2 The Second-order Incident Waves Each of the two ﬁrst-order incident waves is deﬁned by the amplitude A.24) (3. Thus the sum frequency incident wave potential φ+ vanishes in unidirectional regular and irregular waves. frequency ω.8-9) gives 1 Q+ = − ig 2 Ai Aj exp(−i(Ki + Kj ) · x) II 2 2 κ2 − νj κ2 − νi2 (ωi + ωj ) j + i + (Ki · Kj − νi νj ) 2ωj 2ωi ωi ωj and Q− = II 1 2 ig Ai A∗ exp(−i(Ki − Kj ) · x) j 2 2 κ2 − νj κ2 − νi2 (ωi − ωj ) j − i − Ki · Kj + νi νj 2ωj 2ωi ωi ωj (3. 0). the ﬁrst.22) with (3. QII = 0 if βi = βj .3 The Second-order Scattering Waves The solution for the second-order scattering potential.3.

When the body is I not ﬁxed. the integral over SB is represented by a sum of the B integral over each panel assuming Q± (x) is constant on each panel.25) and (2. The waterline is approximated by line segments (consisting of the sides of the panels adjacent to the free surface) and Q± (x) is evaluated on the B midpoints of the segments. the free-surface integral is evaluated separately in two domain divided by a partition circle of radius ρ = b. We consider the integral over SB ﬁrst.30) ξ dl · G[∇φ × (ξ + α × x)] After substituting (3. S The evaluation of the integrals on the right hand side is described below.28). we ﬁrst modify the terms involving the double spacial derivative of the ﬁrstorder velocity potential by applying Stokes’s theorem to avoid the inaccurate numerical evaluation of the second-order derivatives based on the lower order panel method (Lee and Zhu (1993)). ξ )dS B ξ 1 g SF + Q± (ξ )G(x.29) The integral equation (3.24) (but with G corresponding to the same frequency ω and wavenumber as φ± ). After the solution of (3.30) into Q± . φ± on the free surface and in the ﬂuid domain can be S obtained from 4πφ± (x) + S SB φ± (ξ ) S ξ ∂G(x. since it contains B only the normal velocity due to φ± which are derived in Section 3. ξ )dS F ξ (3.28).28) will be considered next in which the ‘quadratic forcing function’ Q± is given by (3. The free-surface integral which is displayed as the last term in (3.8-9). Following the methodology of 18 .28) is identical to the integral equation for the ﬁrst-order potential and thus the discrete form of it is the same as that of (2.2. ξ dSG{n · [(ξ + α × x) · ∇]∇φ} = ξ dSG{n · [(∇φ · ∇)(ξ + α × x)]} − ξ dS[n · (ξ + α × x)](∇φ · ∇G) SB SB SB + + SB WL dS ∂φ ξ [(ξ + α × x) · ∇G] ∂n (3.2πφ± (x) + S SB φ± (ξ ) S ξ ∂G(x. ξ ) dS = ∂nξ SB Q± (ξ )G(x. b issuﬃciently large to neglect the eﬀect of √ the evanescent waves outside the circle. Q± (x) is evaluated B B on the centroids of the panels. For a ﬁxed body.28) where G is the wave source potential deﬁned in (2. ξ )dS B ξ 1 g SF + Q± (ξ )G(x. ξ ) dS = ∂nξ SB Q± (ξ )G(x. ξ )dS F ξ (3. Here ρ = x2 + y 2 . For the purposes F of numerical evaluation. The left-hand side of (3.28) is solved by the panel method. the evaluation Q± (x) is simple.

The integral on this domain is carried out by Gauss-Chebyshev quadrature in the azimuthal direction and Gauss-Legendre quadrature in the radial direction as discussed in Lee and Zhu (1993). The divergence ∇ and the normal vector n must be interpreted in the two dimensional sense on the z = 0 plane. To achieve a better computational eﬃciency for the remaining surface integral. ξ ) renders Gauss-Chebyshev quadrature ineﬃcient. The domain outside the circle of F radius a to the partition circle forms an annulus. When the ﬁeld point x on the body and the source point ξ on the annulus are close. Inside this circle. the surface integral is transformed. Discussion on the selection of the optimum a is provided in Chapter 11 of the WAMIT User Manual (1995). the free surface is discretized with quadrilateral panels.Kim and Yue(1989). using Gauss theorem. In the inner domain. In the outer domain (b ≤ ρ) both the Green function and the asymptotics of the ﬁrstorder potentials are expanded in Fourier-Bessel series. The method of integration on the inner domain is described below and that on the outer domain follows in the next Section. Thus the integration on this region is the sum of the integral over each panels assuming the constant Q± on each panel evaluated at the centroid.30). the integration in the inner domain ρ < b is carried out numerically. the free-surface integrals are reduced to the sum of the line integrals with respect to the radial coordinate ρ. to a form involving only the ﬁrst-order derivatives and line integrals around the waterline (W L) and the partition circle (P C) as is shown below. a is suﬃciently large to enclose the body. The analysis on the integral over the far-ﬁeld free surface is described in detail in Newman (1991) which is included in the Appendix 19 . in an analogous manner to the body surface. After integrating the trigonometric functions with respect to the angular coordinate. The line integrals over W L and P C are carried out in a similar way to the line integral in (3.8-9).31) SF W L+P C SF [(∇φi · ∇φj )G + φi (∇φj · ∇G)]dS where SF denotes only the inner domain of the free surface. the Rankine singularity of G(x. φi ∂ 2 φj GdS = − ∂z 2 + φi (∇φj · n)Gdl (3. we divide the inner domain further into two parts separated by a circle of radius ‘a with a < b. in order to avoid the evaluation of second-order derivatives of the ﬁrst-order potentials in (3.

X = (X. Y. part of the forces can be obtained without solving the scattering potential.and second-order forces are derived from direct integration of the ﬂuid pressure over a body boundary in Section 4. 4. y. The equations of motion for the ﬁrst. z ) are inertial reference frames but x = (x. x = (x. Z) and x ˆ ˆ ˆ ˆ x = (ˆ.1. y. By making use of Green’s theorem. Z) is a global coordinate system and with Z = 0 the undisturbed free-surface. z) is not. The ﬁrst-order Haskind exciting force and the second-order force via indirect approach are discussed in Section 4.1 Hydro-static and -dynamic Force and Moment Coordinate system We consider three coordinate systems.1) 20 .2. This coordinate system is denoted by x = (ˆ.1.2 Coordinate transform ˆ The position vectors in the x− and x− coordinate systems are related to each other by linear transformation ˆ x = ξ + T tx (4.3. The origin of x may be x ˆ ˆ ˆ displaced from the free-surface and Zo denotes the Z-coordinate of the origin of x.1 4.1. 4. Y. Note that X = (X. z) is a body-ﬁxed coordinate system and z also is positive upward when the body is at rest.AND SECOND-ORDER FORCES The expressions for the ﬁrst.Chapter 4 THE FIRST. y . We introduce a third coordinate system which is ﬁxed in space and coincides with “x at ˆ rest”. y .and second-order problem are derived in Section 4. z ). The positive Z axis points upward.

ξ3 ) and α = (α1 .2) In (4.5) P (ˆ )(ˆ × n)dS x x ˆ (4.and the normal vector by ˆ n = T tn (4.7) . T t is the transpose of T = T3 T2 T1 and T1 . the velocity potential and the body motion amplitudes are functions of time. α2 . They also represent the motion amplitudes of the body in the order of surge-sway-heave and roll-pitch-yaw. though the time t does not appear explicitly. 4. ξ2 . F= and M= ˆ SB ˆ SB P (ˆ )ˆ dS xn (4.3) ξ = (ξ1 .1.6) ˆ where SB denotes the instantaneous wetted body boundary. respectively.4) P(ˆ ) = −ρ[Φt (ˆ ) + ∇Φ(ˆ ) · ∇Φ(ˆ ) + g(ˆ + Zo )] x 2 The hydro-static and -dynamic force and moment are obtained from the integration of the pressure over the instantaneous wetted surface. Further discussion on T can be found in Ogilvie (1983).. ˆ The total pressure at x is given by Bernoulli’s equation 1 x x z x (4. ˆ The pressure on the exact body surface (ˆ ∈ SB ) may be approximated by Taylor x expansion with respect to the mean body surface (x ∈ SB ) ξ P(ˆ ) = P (x) + [ξ + (T t − I)x] · ∇P (x) + . α3 ) are the translational and rotational displacements ˆ of x(-coordinate system) with respect to x. T2 and T3 take forms 1 0 T1 = 0 cos α1 0 − sin α1 cos α2 T2 = 0 sin α2 cos α3 T3 = − sin α3 0 0 sin α1 cos α1 0 − sin α2 1 0 0 cos α2 sin α3 cos α3 0 0 0 1 (4..3 Pressure integration In the following derivation.1-2). it is understood that the pressure. x 21 (4.

Substituting (4.7) into (4.1-2), we have ˆ x = x + ξ (1) + α(1) × x + Hx + ξ (2) + α(2) × x + O(A3 ) ˆ n = n + α(1) × n + Hn + α(2) × n + O(A3 ) The cross product of (4.8) and (4.9) then takes a form α ˆ ˆ x × n = x × n + ξ (1) × n + α(1) × (x × n) + ξ (1) × (α(1) × n) + H(x × n) + ξ (2) × n + α(2) × (x × n) + O(A3 ) (4.10) From (4-4), (4-7) and (4.11), the pressure is expressed by the values on the mean body position P(ˆ ) = −ρ{g(z + Zo ) + [Φt (x) + g(ξ3 + α1 y − α2 x)] x 1 (1) ξ + [ ∇Φ(1) (x) · ∇Φ(1) (x) + (ξ (1) + α (1) × x) · ∇Φt (x) + gHx · ∇z] 2 (2) (2) (2) (2) (4.11) + [Φt (x) + g(ξ3 + α1 y − α2 x)]} + O(A3 ) H in (4.7-11) is the second-order component of T t and it takes a form − 1 ((α2 )2 + (α3 )2 ) 0 0 2 (1) (1) (1) 2 (1) 2 1 H = − 2 ((α1 ) + (α3 ) ) 0 α1 α2 (1) (1) (1) (1) (1) 2 (1) 2 1 α2 α3 − 2 ((α1 ) + (α2 ) ) α1 α3

(1) (1) (1) (1) (1) (1)

(4.8) (4.9)

(4.12)

Substituting the expansions (4.7-11) into (4.5-6), we have the series expansion of the integrals (4.5-6).

4.1.4

**Hydrostatic force and moment of O(1)
**

F = −ρg M = −ρg (z + Zo )ndS (4.13)

SB

These are the buoyancy force and moment when the body is at rest and are expressed by

SB

(z + Zo )(x × n)dS = −ρg

SB

[(z + Zo )x] × ndS

**The following relations are invoked frequently in this Chapter. − −
**

S

S

ψndS =

V V

∇ψdV ∇ × ψdV (4.14)

n × ψdS =

∇ × (z + Zo )x = k × x 22

where S is a closed surface consisting of SB and the waterplane area Awp . V denotes the volume of the body. Using the relations (4.14), the force and moment are expressed in familiar forms

F = ρgV k (4.15) M = ρgV (yb i − xb j) where xb and yb are the x and y coordinates of the center of buoyancy. ˆ In (4.15), i, j, and k are the unit vectors in x coordinate system.

4.1.5

Linear force and moment

**The linear force and moment are obtained from F(1) = − ρ − ρg − ρ M(1) = − ρ − ρg − ρg − ρg
**

(1)

SB

nΦt dS α (α(1) × n)(z + Zo )dS

(1) (1) (1)

SB SB

ng(ξ3 + α1 y − α2 x)dS (4.16)

(1)

SB

(x × n)Φt dS (x × n)(ξ3 + α1 y − α2 x)dS ξ (ξ (1) × n)(z + Zo )dS α [α(1) × (x × n)](z + Zo )dS

(1) (1) (1)

SB SB SB

In (4.16), the ﬁrst terms are the hydrodynamic force and moment and the rest are the hydrostatic. Following the decomposition (2.5), we consider component potentials such (1) (1) (1) (1) (1) that Φ(1) = ΦI + ΦS + ΦR = ΦD + ΦR . The hydrodynamic force and moment are (1) divided into two components: the “wave exciting force” due to ΦD and the force due to (1) ΦR expressed in terms of the added mass and damping coeﬃcients. The integrals of the hydrostatic pressure can be simpliﬁed by applying (4.14) and their variations with the results F(1) = − ρ Φt ndS − ρgAwp(ξ3 + α1 yf − α2 xf )k 23

(1) (1) (1) (1)

SB

(4.17) M(1) = − ρ

SB

(x ×

(1) (1)

(1) n)Φt dS (1) (1) (1) (1)

− ρg[−V ξ2 + Awp yf ξ3 + (V zb + L22 )α1 − L12 α2 − V xb α3 ]i − ρg[ V ξ1 − Awp xf ξ3 − L12 α1 + (V zb + L11 )α2 − V ybα3 ]j where Lij is the second moment over the waterplane area. For example, L12 =

Awp (1) (1) (1) (1)

xydS.

xf and yf are the coordinates of the center of ﬂoatation. Following (1.10), the force can be represented by a discrete spectrum (the moment takes an identical form and is omitted here) F(1) = Re

j (1) Fj eiωj t

(4.18)

4.1.6

**Second-order force and moment
**

(z + Zo )HndS − ρ

(1) (1) (1) (1) α (α(1) × n)[Φt + g(ξ3 + α1 y − α2 x)]dS

**The second-order force and moment are obtained from F(2) = − ρg − ρ +
**

SB SB

1 (1) ξ [ ∇Φ(1) · ∇Φ(1) + (ξ (1) + α(1) × x) · ∇Φt ]ndS − ρg SB 2

SB

(Hx · k)ndS

1 (1) (1) (1) ρg [η (1) − (ξ3 + α1 y − α2 x)]2 1 − n2 dl z 2 WL (2) (2) (2) − ρgAwp (ξ3 + α1 yf − α2 xf )k − ρ M(2) =

1 2 SB

Φt ndS (4.19)

(1) (1) (1)

(2)

ρg

wl

[η (1) − (ξ3 + α1 y − α2 x)]2 1 − n2 (x × n)dl z

− ρ − ρ − ρg − ρ − ρg

1 (1) ξ [ ∇Φ(1) · ∇Φ(1) + (ξ (1) + α(1) × x) · ∇Φt ](x × n)dS SB 2

SB (1) (1) (1) (1) ξ (ξ (1) × n)[Φt + g(ξ3 + α1 y − α2 x)]dS

SB SB

α ξ (1) × (α(1) × n)(z + Zo )dS

(1) (1) (1) (1) α(1) × (x × n)[Φt + g(ξ3 + α1 y − α2 x)]dS

SB

(z + Zo )H(x × n)dS − ρg

(2) (2) (2) (2) (2)

SB

(Hx · k)(x × n)dS

(2) (2) (2) (2) (2)

− ρgi[−V ξ2 + Awp yf ξ3 + (V zb + L22 )α1 − L12 α2 − V xb α3 ] − ρgj[ V ξ1 − Awp xf ξ3 − L12 α1 + (V zb + L11 )α2 − V yb α3 ] 24

19) take forms F(2) = Fq + Fp − ρgAwp (ξ3 + α1 yf − α2 xf )k − ρ M (2) SB (2) (2) (2) n ∂ΦR dS ∂t (4.14) and their variations to the hydrostatic pressure integrals of (4. The quadratic force and moment are deﬁned by 25 . −ρ SB (1) (1) (1) (1) α (α(1) × n)[Φt + g(ξ3 + α1 y − α2 x)]dS α = α × F (1) + ρgα × = α × F (1) + ρgV −ρ SB (1) (1) [−α1 α3 i α (α(1) × n)(z + Zo )dS − α2 α3 j + ((α1 )2 + (α2 )2 )k] (4.21) (2) = Mq + Mp − ρgi[−V ξ2 + Awp yf ξ3 + (V zb + L22 )α1 − L12 α2 − V xb α3 ] − ρgj[V ξ1 − Awp xf ξ3 − L12 α1 + (V zb + L11 )α2 − V yb α3 ] − ρ ∂Φ (x × n) R dS ∂t SB (2) (2) (2) (2) (2) (2) (2) (2) (2) (2) (2) where the subscript q denotes the force and moment due to the quadratic interaction of the ﬁrst-order solution and p due to the second-order potential.− ρ SB (x × n)Φt dS (2) We apply (4.20) (1) (1) (1) (1) SB (1) (1) (1) (1) ξ (ξ (1) × n)[Φt + g(ξ3 + α1 y − α2 x)]dS = ξ (1) × F (1) + ρg −ρ α(1) × (x × n) SB (1) (1) (1) SB (1) α ξ (1) × (α(1) × n)(z + Zo )dS (1) SB [Φt + g(ξ3 + α1 y − α2 x)]dS ξ (ξ (1) × n)(z + Zo )dS + SB α = α (1) × M (1) + ρgα(1) × { ξ = α(1) × M (1) − ρgV α(1) × (ξ α [α(1) × (x × n)](z + Zo )dS} α × k) − ρgV α(1) × [α(1) × (yb i − xb j)] (2) The second-order force or moment due to Φ(2) is decomposed into a part due to ΦI + (2) (2) ΦS and the other part due to ΦR as in the ﬁrst-order. Then the force and moment (4.19).19) and make use of the following relations to simplify (4.

7 Haskind Exciting Force and Indirect Second-order Force The linear wave exciting force and the second-order potential force Fp may be evaluated not from the scattering solution but from the appropriate component of the radiation solution.1.12) the second-order force F(2) can be expressed as (the moment takes an identical form).24) 4.22) 1 ρg = 2 [η (1) − (ξ3 + α1 y − α2 x)]2 1 − n2 (x × n)dl z (1) (1) (1) Mq wl − ρ + + + + 1 (1) ξ [ ∇Φ(1) · ∇Φ(1) + (ξ (1) + α(1) × x) · ∇Φt ](x × n)dS SB 2 ξ (1) × F (1) + α(1) × M (1) 1 (1) (1) (1) (1) (1) (1) (1) (1) ρg(−V ξ1 α3 + V α1 α2 xb − V α2 α3 zb − V ((a1 )2 − (a3 )2 )yb 2 1 (1) (1) (1) (1) (1) (1) −α1 α3 L12 − α2 α3 L22 − ((α1 )2 + (α2 )2 )Zo Awp yf )i 2 1 (1) (1) (1) (1) (1) (1) ρg(−V ξ2 α3 + V α1 α3 zb + V ((a2 )2 − (a3 )2 )xb 2 1 (1) (1) (1) (1) (1) (1) +α1 α3 L11 + α2 α3 L12 + ((α1 )2 + (α2 )2 )Zo Awp xf )j 2 (1) (1) (1) (1) (1) (1) (1) (1) ρg(V ξ1 α1 + V ξ2 α2 + V α2 α3 xb − V α1 α3 yb )k (2) (2) The second-order potential force and moment are given by Fp = −ρ ∂(ΦI + ΦS ) ndS ∂t SB (4.23) Mp = −ρ SB (x × n) (2) ∂(ΦI + ∂t (2) ΦS ) dS Adopting a form analogous to (1. F(2) = Re i j F+ ei(ωi +ωj )t + F− ei(ωi −ωj )t ij ij (4.Fq = 1 ρg 2 wl [η (1) − (ξ3 + α1 y − α2 x)]2 1 − n2 ndl z (1) (1) (1) − ρ 1 (1) ξ [ ∇Φ(1) · ∇Φ(1) + (ξ (1) + α(1) × x) · ∇Φt ]ndS SB 2 1 (1) (1) (1) (1) (1) (1) + α(1) × F (1) − ρgAwp [α1 α3 xf + α2 α3 yf + ((α1 )2 + (α2 )2 )Zo ]k 2 (4. The ﬁrst-order force evaluated in this way is referred to as the Haskind exciting force and the second-order force as the indirect force. 26 .

26) into (4.18-21) into (4.25) with the conditions (2.29). 27 . on the bottom and at the far ﬁeld vanish due to the boundary conditions for φS and φj on these surfaces.1. applied to φS and a component of the radiation potential φj .10) and (2. (Newman (1977. we have SB nj φ± dS = S SB Q± φ± dS + B j 1 g SF (Q± + Q± )φ± dS IB BB j (4. The force for the mode j is given by ± Fpj = −i(ωi ± ωj )ρ SB nj (φ± + φ± )dS I S (4.18)) Thus they do not appear in (4.4. Upon substituting (4. applied to φ± and φ± . Section 6.26).8 Haskind exciting force We consider a complex amplitude of a spectral component of the ﬁrst-order wave exciting force. we have ∂φj ∂φS 0= φS − φj )dS ( (4.28) From Green’s theorem.29) SF It can be shown that the integrals on the sea bottom and at the far ﬁeld vanish due to the boundary conditions.25) From Green’s theorem. we have j S 0= ∂φ± ± ∂φ± ± j φ − S φj )dS + ∂n S ∂n SB ( ( ∂φ± ± ∂φ± ± j φ − S φj )dS ∂z S ∂z (4.30) Thus the force can be obtained from ± Fpj = −i(ωi ± ωj )ρ[ SB (nj φ± + Q± φ± )dS + I B j 1 g SF (Q± + Q± )φ± dS] IB BB j (4. Upon substituting (3. The force for the mode j is given by Fj = −iωρ SB nj (φI + φS )dS (4.9 Indirect second-order force We consider a complex amplitude of a spectral component of the second-order potential force Fp .31) The radiation potential φ± with the frequency ωi ±ωj is often referred to as an assisting j potential.11).26) ∂n SB ∂n It can be shown that integrals over the free surface. the Haskind relation follows in the form Fj = −iωρ (nj φI − ∂φI φj )dS ∂n (4.1.27) SB 4.

2 Equations of Motion The translational and rotational motions of the body are governed by mˆ gtt = FT x and Lt = MT (4. we express xg . which may diﬀer from the center of mass.36) (4.34) with (4.and the second-order equations of motion. xgtt is the acceleration of the center of mass of the body and FT is the total force due to the ﬂuid pressure. FT .35-6). the body mass and the external force. ˆ xg = xg + ξ + α × xg + Hxg + O(A3) F = F (1) + FB + FE + F (2) + FB + FE + O(A3 ) (1) (1) (2) (2) (4.34).38-39).33) in the body-ﬁxed coordinate system.3). the subscript B denotes the gravitational force on the body mass. describe the motion of the rigid body. If we are interested in the motion about the origin of the body coordinate system.32) ˆ In (4. T MT and ω in perturbation series. The force without 28 . Ig ω t + ω × Ig ω = T MT (4.33) (4.32).37) (4. x.4. MT is the ˆ total moment about the xg and Lt is the time rate of change of the angular momentum in ˆ the x coordinate system.35) where I and Mo are the moment of inertia and moment about o. E external force which is assumed to be expandable in a perturbation series. ˆ To derive the ﬁrst. In (4.33). ω t (ω ) the angular acceleration (velocity) and T the tranform matrix (4. Following Ogilvie(1983). the following relations are to be substitued into (4. (4. ω ω Ig ω = Iω − mxg × (ω × xg ) and ˆ MT = Mo − xg × FT (4.38) T MT = M (1) − xg × F (1) + M (2) − xg × F (2) − ξ (1) × F (1) − α (1) × M (1) α + xg × (α(1) × F (1) ) + Γ(1) + Γ(2) + O(A3 ) (4. It is convenient to express (4.39) In (4. xo .32) and (4. in which the moment of inertia is time invariant.34) ω where Ig is the moment of inertia about the center of mass.

the equations of motion at each order follows.40) (4.45) α1t + α2t α3 (2) (1) (1) = α2t − α1t α3 (2) (1) (1) α3t + α1t α2 (2) (1) (1) (4.34). These are given by ξ (1) ξ (1) F (1) ξ (1) E tt [(M + M + A){ (1) } + (B + B ){ t(1) } + (C + C E ){ (1) }] = { exc }(4.43) (4. leads to the conditions F (0) + FB + FE = 0 and M (0) + MB + ME = 0 (0) (0) (0) (0) (1) (2) (1) (2) (1) (2) (4. Γ(1) and Γ(2) are the ﬁrst.32) and (4.46) Upon substitution of (4.41) The moment ME consists of a pure torque Mτ and the moment due to the external force ˆ ME = Mτ + xe × FE (4. α1t 0 0 t t t ˆ ω = 0 + T1 α2t + T1 T2 0 α3t 0 0 and angular velocity in the x coordinate system. ω .47) (1) Mexc α αtt αt E and ξ (2) ξ (2) F (2) ξ (2) E tt [(M + M + A){ (2) } + (B + B ){ t(2) } + (C + C E ){ (2) }] = { exc }(4.42) ˆ The angular velocity in the x coordinate system can be obtained by linear superposition of the contributions from time derivatives of Euler-angles.35-46) into (4.1 and 4.and second-order components of ˆ Γ = T [MB + MB + ME + ME − xg × (FE + FE ) + O(A3 )] Equilibrium. is obtained from ˆ ω = T tω Thus we have at the ﬁrst-order (1) (4. when the body is at rest.44) ω and at the second-order ω (2) (1) = (1) α2t (1) α3t α1t (4.48) (2) Mexc α αtt αt E 29 .2.subscript is the hydro-static and -dynamic force derived in Sections 4.

49) 0 0 0 0 0 ρAwp 0 ρAwp yf 0 −ρAwp xf 0 0 ρAwp yf ρ(V zb + L22 ) − mzg −ρL12 0 0 0 −ρAwp xf −ρL12 ρ(V zb + L11 ) − mzg 0 0 0 0 −ρV xb + mxg −ρV yb + myg 0 (4. They may or may not be diﬀerent for the ﬁrst and the second-order motions.where A and B are the added mass and damping coeﬃcients and they depend on the frequency. velocity and the motion amplitude.52) Mq + Mp − ξ + ρg[V (1) (1) ξ1 α3 i (1) (2) = ×F (1) −α (1) ×M (1) (1) (1) (1) +V (1) (1) ξ2 α3 j − (V ξ1 α1 − V ξ2 α2 )k] (1) 1 (1) (1) (1) (1) + mg[( ((α1 )2 − (α3 )2 )yg − α1 α2 xg + α2 α3 zg )i 2 1 (1) (1) − ( ((α2 )2 − (α3 )2 )xg + α1 α3 zg )j + (α1 α3 yg − α2 α3 xg )k] 2 α + xg × (α(1) × F (1) ) − mxg × Htt xg − Ig αq tt (1) (1) (0) (0) α − αt × Ig αt + Σ[Hxe × FE − α(1) × ((α(1) × xe ) × FE )] (2) (1) (2) (1) α + {Σ[Mτ + xe × FE − α(1) × Mτ − (xe − xg ) × (α(1) × FE )]}ex 30 . B E and C E are the external force components which are proportional to the unknown body acceleration.47) and (4. respectively.48) are (2) Fexc = (2) Mexc Fq + Fp − mHtt xg + (ΣFE )ex (4. M and C are the same in (4.47) are (1) Fexc = −ρ D(1) (1) SB Φt ndS + (ΣFE )ex (4.48) and they are given by m 0 0 M = 0 mzg −myg 0 0 0 C = g 0 0 0 0 m 0 −mzg 0 mxg 0 0 0 0 m myg mxg 0 0 −mzg myg I11 I21 I31 mzg 0 −mxg I12 I22 I32 −myg mxg 0 I13 I23 I33 (4.51) D(1) (1) dS + (ΣMτ + Σxe × FE )ex (1) (1) Mexc = −ρ SB (x × n)Φt and those on the right-hand side of (4.50) Finally the force and moment on the right-hand side of (4. M E .

In (4.52). αq is a vector and its components are the time derivatives of the second terms of the vector tt elements of (4.46).The subscript ex denotes the part of the force and moment which is not linearly proportional to the motion amplitude. 31 . the velocity or the acceleration at each order.

an update. R. Dept. 1977 ‘Marine Hydrodynamics’. C. & Sclavounos. D.’ J.1S (Second-Order Module) . P. N.D. MIT Press. Lee. Martin & G. & Zhu.-H.-H. Molin. C. Lee. J. Martin & Rizzo 1993 ‘Boundary integral equations for bodies of small. J. R. Kleinman. Wickham.’ J. Newfoundland. T. E. editors. 1991 ‘First.-H. John’s. C. & Yue. 1989 ‘The complete second-order diﬀraction solution for an axisymmetric body Part 1. J. St.-H. Korsmeyer. 1982 ‘On the mathematical theory of the motion of ﬂoating bodies . thickness’ 8-th International Workshop on Water Waves and Floating Bodies. & Newman. A. Newman. 200. J. Newman. 183-190. 197-202. X. 235–264. Lee. 1991 ‘WAMIT V3.-H. N. N. Newman. D. F. Lee. Newman. 32 . Newman J. pp.’ P. but ﬁnite. K.Theoretical Background. C. 12. Cambridge University Press. 1992 ‘The approximation of free-surface Green functions. 14-18. J.N. & Zhu X. N. 1990 “Numerical solutions of the water-wave dispersion relation.. 1979 ‘Second order diﬀraction loads upon three dimensional bodies.-H. John’s.’ Applied Ocean Research 1..’ in ‘Wave Asymptotics. of Ocean Engineering. J. Newfoundland. Journal of Engineering Mathematics. Fluid Mech. 1988 ‘Numerical methods for the solution of three-dimensional integral equations in wave-body interactions.’ 8-th International Workshop on Water Waves and Floating Bodies. P. Lee.” Applied Ocean Research.’ unpublished report included in April 1991 Steering Committee Report. St. B. Monochromatic incident waves.’ DTNSRDC Rep 82/074. Newman. M. Houston. Submitted for publication .. C. N. MIT. No 3. 1988 ‘The analysis of wave interactions with tension leg platforms. 57-67. 19. J. 1993 ‘Second-order diﬀraction and radiation solutions on ﬂoating bodies.and second-order wave eﬀects on a submerged spheroid. 1985 ‘Algorithms for the Free-Surface Green Function’. Thesis.. N. Ship Research 35...’ Ph. Newman.’ OMAE Conference. Vol.REFERENCES A Kim. 1995 “An extended boundary integral equation for the removal of the irregular frequency eﬀect”. N.

1993 ‘Note on bodies and panels in the free surface. J.’ 5th International Conference on the Behaviour of Oﬀshore Structures (BOSS ‘88) Trondheim. WAMIT User Manual 1999 WAMIT Inc. Zhu. N. Dept. 9. Berlin. & Sclavounos. and Laitone. V. D.’ Applied Ocean Research. 1960 ‘Surface waves. 1994 ‘Wave eﬀects on deformable bodies. X. F. 1. 1994 ‘Irregular Frequency Removal from the Boundary Integral Equation for the Wave-body Problem’.. of Ocean Eng. MIT. 1983 ‘Second-order hydrodynamic eﬀects on ocean platforms.’ unpublished report. Berkeley. Springer.Newman. Newman. 1988 ‘The computation of wave loads on large oﬀshore structures.. P.’ International Workshop on Ship and Platform Motions. V. Newman. 33 .. J.’ Handbuch der Physik. Master Thesis. 16. 47-59. Wehausen. J. Ogilvie. T. J. N. 446-778. E. N.

Newman. J. Lee C. Woods Hole. London 1992.’ OMAE Conference. ‘Wave eﬀects on deformable bodies. Lee. T. N. Newman. T. ‘Sensitivity of wave loads to the discretization of bodies’ BOSS ‘92. C. C. ‘The computation of wave loads on large oﬀshore structures.. 16. J. ‘Panel methods in marine hydrodynamics. & Newman. ‘Second-order diﬀraction and radiation solutions on ﬂoating bodies. & Zhu. Newfoundland 1993. & Sclavounos. 1989. D. John’s. fact or ﬁction?’ (Eleventh Georg Weinblum Memorial Lecture). N..-H.-H. J. & Yue.. J. C. Fathi. D.. 1991. N. 1988.. Trondheim. K. J. 4..-H. 36. N. Emmerhoﬀ.. D. Lee C.. P.’ OMAE Conference. ‘Deformable ﬂoating bodies.’ 5th International Conference on the Behaviour of Oﬀshore Structures BOSS ‘88 Trondheim.. ‘Linear Hydrodynamic Loads due to Elastic Deformations of Large Volume 34 . St. 155-168. F. & Newman.’ Eighth International Workshop on Water Waves and Floating Bodies. & Sclavounos.’ Keynote Lecture. ‘The numerical towing tank. Second-order wave loads on a stationary body. Lee. Kim. Hobart. F. Norway. N. ing.’ 8-th International Workshop on Water Waves and Floating Bodies.’ 6th International Workshop on Water Waves and Floating Bodies.. Newfoundland 1993. ‘The analysis of wave interactions with tension leg platforms. Korsmeyer. J. P. J. for siv. ‘The computation of second-order wave loads. X. D. O. 47-59 1994.-H.. St. & Newman.-H.’ Applied Ocean Research. N. John’s. 1. Ship Research. Lee.. C. ‘The computation of ship interaction forces in restricted water. N. 1988. 1989. Newman.’ thesis prepared at M.-H. ‘First. J. J. 1991. N.-H. Newman. ‘Analysis and computation of wave eﬀects on large volume structures. N..I. Lee. Newman. J. M.and second-order wave eﬀects on a submerged spheroid. Lee. H.’ J. Eleventh Australasian Fluid Mechanics Conference. N.’ Journal of Ship Research 1993. C. P. Newman. 1991. Australia 1992.. J. Stavanger. Houston.T. degree from NTH.REFERENCES B Korsmeyer. Newman. Ship Technology Research/Schiﬀstechnik.

D.I. Fathi. X.. ‘Second-order wave eﬀects on oﬀshore structures’ BOSS ‘94.-H. J. Texas Section of the Society of Naval Architects and Marine Engineers. Houston 1995. Japan. ‘Computation of wave induced motions on a ﬂexible container. 1994 Newman. J. & Lee. ‘Computation of mean forces and moments in multibody interaction.Structures’.-H. J.’ Symposium on Hydroelasticity.’ BOSS ‘94. thesis prepared at M. & Newman. Zhu. N.-H. C.T. Norway 1994. D. Kuju. M.’ 9-th International Workshop on Water Waves and Floating Bodies. ‘Removing the irregular frequencies from wave-body interactions. N. UK 1995. Lee. N. for siv.. ing. Trondheim. ‘Runup on a vertical cylinder in long waves. Oxford.’ Tenth International Workshop on Water Waves and Floating Bodies.-H. Lee. 35 . Ferreira. C. N.-H. degree from NTH. & Lee. Newman. C. MIT 1994.’ TLP Technology Symposium. ‘To second order and beyond. MIT 1994. J. Trondheim 1993. C. & Lee C. & Newman.

APPENDICES .

The most important diﬀerences from [K] are the deﬁnition of the unit normal vector.1) r 2 = (x − ξ)2 + (y − η)2 + (z − ζ)2 r = (x − ξ) + (y − η) + (z + ζ) . 2 2 2 2 (1. and (2) removal of irregular frequencies by adding panels in the free surface interior to the body with an imposed homogeneous Neumann (or Dirichlet) condition. and the Green function G. For simplicity the depth is assumed inﬁnite. The wavenumber is K = ω 2 /g. and g is the acceleration of gravity.4) with respect to both coordinate systems. with zero draft. More explicitly. the linear free surface condition φz − Kφ = 0. The notation of the WAMIT User Manual is followed. ξ ) = dk 0 ek(z+ζ) J0 (kR) k−K (1. here taken to be into the body and out of the ﬂuid domain.3) where J0 (x) is the Bessel function of zero order.September 1993 1. denoted here by [K].6) It is not obvious that these relations can be applied in the limit when r = r = 0. Gζ − KG = 0. The Green function G satisﬁes the free surface condition (1.7) .4) 1 2K 1 + + r r π ∞ G(x. on z = 0 (1. when the source and ﬁeld points coincide on the free surface. INTRODUCTION These notes document recent work intended to (1) extend WAMIT to bodies where part or all of the submerged surface is in the plane of the free surface. The correponding function γ in [K] is related to G via the equation G = −2πγ.4) and a radiation condition at inﬁnity. i.NOTES ON BODIES OR PANELS IN THE FREE SURFACE JNN .e. deﬁned here by equation (2. In the vicinity of this singular point it is known that G is of the form G(x.2) (1.5) Gz − KG = 0. but diﬀerences and questions have arisen which are the principal motivation for the notes. on z = 0 (1. subject to a prescribed Neumann condition φn = V on the body surface Sb . The velocity potential φ is governed by Laplace’s equation in the ﬂuid domain. ξ ) = 1 1 + − 2KeK(z+ζ) log(r + |z + ζ|) + (γ − log 2) + r + O(r 2 log r ) r r 1 (1. on ζ = 0 (1. The latter development was intended to follow the theory in the report by Kleinman.

1) to the case where part or all of the body surface coincides with the free surface. 2 CONVENTIONAL INTEGRAL EQUATION FOR THE VELOCITY POTENTIAL In the ‘normal’ case of a ﬂoating body. However caution is required to correctly account for the jump when the image source 1/r is included. and the eﬀect is to double the jump in (2. Examples of such bodies are a ﬂoating caisson where the interior ‘roof’ is in the plane z = 0. The complementary domain D− is inside the body.1b) . it follows from Green’s theorem that ∂G ξ dξ = ∂nξ 2πα(x)φ(x) + Sb ξ φ(ξ ) ξ ξ V (ξ )Gdξ . Sb (2. 51b-52b]. and extends to inﬁnity horizontally and vertically. Thus there is no ‘delta function’ in the free-surface conditions (1. This is the appropriate interpretation of the integral on the left side of (2. it can be conﬁrmed that the same relations apply even at the singular point.If this expansion is substituted in (1. and 0 for points in D− .5) and (1.1) for the case where x is on Sb . in the limits where the point x approaches Sb from D± . With the above deﬁnitions. Thus.1) are consistent with the ‘jump conditions’ corresponding to the singularity 1/r in G. as in the simpler Cauchy principal-value integral. that a surface of vanishingly small area including the singular point r = 0 is excluded from the integral. as in [K]. The ﬂuid domain D+ is exterior to the body.1) is rewritten in the more explicit form ξ 2πφ(x) + −− φ(ξ ) − Sb ∂G ξ dξ = ∂nξ ξ ξ V (ξ )Gdξ Sb (x ∈ Sb ) (2. 1 for points on Sb . α is equal to 2 for points in D+ . The deﬁnitions of α in (2. we write the following derivatives and limits: −(z − ζ) ∂ 1 = ∂z r r3 −(z + ζ) ∂ 1 = ∂z r r3 2 (3. Thus (2.1) Here. I± (φ) = lim x→S± ξ φ(ξ ) Sb ∂ 1 ∂ 1 ξ ξ ξ dξ = ∓2πφ + −− φ(ξ ) − dξ ∂nξ r ∂nξ r Sb (2. except for a line of intersection of the two surfaces. and a ‘cicular dock’ consisting of a disk of zero draft.5) or (1.2) Here the symbol − denotes. the submerged surface Sb is entirely below the plane z = 0. In these cases.6).6). For completeness.1a) (3. below the plane z = 0.2). in relation to the corresponding equations [K.3) 3 INTEGRAL EQUATIONS FOR Sb IN THE FREE SURFACE The objective here is to extend (2. for those portions of Sb lying in the free surface. the two Rankine singularities in the Green function coalesce. below the free surface.

1d) on ζ = 0 on ζ = 0 (3. in the plane z = 0. and approach the surface at a point where the surface coincides with the plane z = 0. The former is consistent with [K] equation (51b). the contribution from the image source to the left-hand side of (2. Note in this case that the unit normal is positive upwards.3) and thus in agreement with (3. In the limit the same value applies.1e) tends to a delta function with area 4π.2).g. with a ﬂat bottom and vertical sides. In the case of a body which is entirely in the plane z = 0.1) has the value 4πφ.1f) is identically zero and does not contribute a jump. since the singular contribution to the kernel from the two Rankine singularties is 1 ∂ 1 + ∂nξ r r Sb = 1 ∂ 1 + ∂ζ r r ζ=0 =0 (3.1) Thus.1e) (3.6). with the ﬁeld point in the ﬂuid domain. pointing from the ﬂuid domain toward the complementary domain above the free surface. but (3.1. In this case (2.1f) is correct. consider a body with small ﬁnite draft T .3) is ξ φ(ξ ) Sb ∂ 1 ξ dξ = ∂ζ r ξ φ(ξ ) Sb −(z + ζ) ξ dξ r3 on z = ζ = −T (3.1) is applicable to such a body surface with the understanding that α = 2 when the point x is on the body surface. I will proceed assuming (3.(z − ζ) ∂ 1 = ∂ζ r r3 −(z + ζ) ∂ 1 = ∂ζ r r3 Thus 1 −2z ∂ 1 + = 3 ∂z r r r 1 ∂ 1 + =0 ∂ζ r r (3. First start with (2.5 . (2. two equivalent integral equations may be written in the forms ξ 4πφ(x) + −− φ(ξ ) − Sb ∂G ξ dξ = ∂z 3 ξ ξ V (ξ )Gdξ . Sb (x ∈ Sb ) (3.2) As an alternative approach. e. Two alternative derivations are given for the case where Sb coincides with the plane z = 0.1). but the latter contradicts the jump in (52b). augmenting the free term in (2.1f ) In the limit z → 0− (3.1c) (3.3) . the free term in (2.3) is applicable on the exact body. Prior to the limit. Using the free-surface boundary conditions (1.2a) In the limit T → 0 the kernel tends to a delta function with area 2π. Sb (x ∈ Sb ) (3. below the body surface. Neglecting the contribution from the vertical sides. the resulting integral equation is ξ 4πφ(x) + −− φ(ξ ) − Sb ∂G ξ dξ = ∂ζ ξ ξ V (ξ )Gdξ .

WAMIT has been modiﬁed in accordance with (3.3) will vanish when x is on Sb . It is known that nontrivial solutions exist for certain body shapes at discrete eigenfrequencies. Next we deﬁne an external ‘radiation’ potential φ which satisﬁes the usual conditions outside the body. (x ∈ Sb ) (3. and no portion of its surface lies in the plane z = 0 except for the normal intersection along a waterline contour.4). However for this particular external potential the right-hand side of (2. This is essentially a radiation problem with a special distribution of normal velocity on the body. and used to solve the circular dock problem. just as for any more conventional radiation (or scattering) problem.2-4) have no homogeneous solutions. Now assume that the interior potential φ satisﬁes the homogeneous Dirichlet condition φ = 0 on Sb . 4 EXISTENCE OF IRREGULAR FREQUENCIES Hereafter the body is assumed to have nonzero submerged volume. 4 .7) is not ampliﬁed by diﬀerentiation. except that the homogeneous body boundary condition is Dirichlet instead of Neumann. to intersect the free surface normally.3). Physically.2) Sb This proves that there is a discrete set of eigenfrequencies (i.and 4πφ(x) + K Sb ξ ξ φ(ξ )Gdξ = Sb ξ ξ V (ξ )Gdξ .1) Here V = φn is the corresponding normal velocity on Sb . the irregular frequencies) where the corresponding normal velocity V is orthogonal to G on the body. Deﬁne Si as the portion of the plane z = 0 interior to Sb . The union of Sb and Si is the closed surface S. 2π α(x) − 2 φ (x) + ξ φ (ξ ) Sb ∂G ξ dξ = ∂nξ ξ ξ V (ξ )Gdξ .7) analytically over each panel in the free surface. Applying Green’s theorem in the usual manner. Sb (4. as in §1.4) on Si . It follows that φ is a homogeneous solution of (2. but from (4. It is essential in this case to use the ‘ILOG=1’ option to integrate the logarithmic singularity in (1.4) is most eﬀective for numerical computations since the logarithmic singularity in (1. The solution is assumed to exist and to be nontrivial and unique.e. The results appear to be consistent. and satisﬁes the free surface condition (1. with the unit normal deﬁned in a consistent manner to point into the interior volume D− . For such solutions the normal velocity V is generally nonzero on Sb .1) and the boundary condition φ = 0 on Sb it follows that ξ ξ V (ξ )Gdξ = 0 (x ∈ Sb ∪ D+ ) (4. Deﬁne an interior potential φ which is harmonic in D− . or irregular frequencies. these are analogous to ‘sloshing modes’ corresponding to standing waves inside the body. This can be inferred from the fact that the interior volume between Sb and the free surface is zero. with the speciﬁed normal derivative φn = V = V on Sb . and it is assumed that the same is true for general bodies. This proves the existence of irregular frequencies in the context of the exterior potential problem. Presumably the integral equations (3.4) Equation (3.

3) The only diﬀerence is the sign preceding the integral over Si . 86] is derived which.2). This should be borne in mind in judging the relevance of the statement that the numerical tests support (5. in the present notation.2).4b).4b) are precisely the same as (minus) the left side of (3.3) by the additional equation ξ φ(ξ ) Sb ∂G ξ dξ = ∂nξ ξ ξ V (ξ )Gdξ Sb (x ∈ Si ) (5.2) is ξ 2πΦ(x) + −− Φ(ξ ) − Sb ∂G 1 ∂G ξ ξ ξ − dξ − −− Φ(ξ ) dξ = ∂nξ 2 ∂nξ Si ξ ξ V (ξ )Gdξ Sb (x ∈ Sb ∪ Si ) (5. takes the form ξ 2πΦ(x) + −− Φ(ξ ) − Sb ∂G 1 ∂G ξ ξ ξ − dξ + −− Φ(ξ ) dξ = ∂nξ 2 ∂nξ Si ξ ξ V (ξ )Gdξ Sb (x ∈ Sb ∪ Si ) (5. Note that the ﬁrst and third terms in (5. At a more pragmatic level. and using (1.2) In [K] it is shown (using the questionable jump relation (52b)) that (5. In fact. Thus we augment (2.4) are equivalent after multiplying the unknown Φ by a factor -2 in the domain of Si .4).4b) −4πΦ(x) + ξ Φ(ξ ) Sb ξ ξ Φ(ξ )Gdξ = Si Sb ξ ξ V (ξ )Gdξ Equations (5. suggesting a possible connection with the absence of irregular frequencies in the dock problem.5a) .3) and (5. that is no nontrivial solutions Φ0 of the pair of integral equations ξ − 2πΦ0 (x) + −− Φ0 (ξ ) Sb ∂G ξ dξ + ∂nξ 5 Si ξ Φ0 (ξ ) ∂G ξ dξ = 0 ∂nξ (x ∈ Sb ) (5.1) in the extended domain S = Sb ∪ Si . there is (at most) one solution of the integral equation (2.3a-b). Finally we want to show that there are no homogeneous solutions of (5. the numerical tests are performed with a modiﬁed version of WAMIT where the inﬂuence functions correspond to the slightly diﬀerent integral equations ∂G ξ dξ + ∂nξ ∂G ξ dξ − K ∂nξ ∂G ξ dξ = ∂nξ ξ 2πΦ(x) + −− Φ(ξ ) − Sb ξ Φ(ξ ) Si ξ ξ V (ξ )Gdξ Sb (x ∈ Sb ) (x ∈ Si ) (5.3) may have homogeneous solutions at the irregular frequencies.2) has (at most) only one solution. but not (5. our numerical experience indicates that the appropriate replacement for (5.5 EXTENDED-BOUNDARY INTEGRAL EQUATIONS In [K] it is shown that while (2.1) A more useful second-kind equation [K.6) to replace the normal derivative of G on Si in (5. and that this solution is equal to φ on Sb .4a) (5.

1e) gives an extra contribution and it follows from the last form of (5.8) ξ Φ0 (ξ ) ∂G ξ dξ ∂nξ ∂G ξ dξ ∂nξ (x ∈ Si ) (5. asume that Φ0 is a nontrivial solution of (5.10b). 6 (x ∈ D− ) (5.5b).9a) ξ − φi = −− Φ0 (ξ ) Si (x ∈ Si ) (5.7b) φb + φi = 0 The following conditions apply as z → 0− on Si : φb = Sb (x ∈ Sb ) (5.11) ¿From (5. Si ξ Φ0 (ξ ) ∂G ξ dξ ∂nξ (x ∈ Sb ) (5.5) and deﬁne the following two potentials (which together are used in place of [K 88]): φb = φi = ξ Φ0 (ξ ) Sb ξ Φ0 (ξ ) ∂G ξ dξ ∂nξ Si (x ∈ D− ) (x ∈ D− ∪ Sb ∪ D+ ) (5.5b) For this purpose. it follows that φb + φi = 0 The remainder of the proof follows as in [K 89-95]. φbz + φiz = 0 (5.5a).7a) φi = Thus.6b) Si ∂G ξ dξ = −K ∂nξ ξ Φ0 Gdξ The following conditions apply on Sb : ξ − φb = 2πΦ0 (x) + −− Φ0 (ξ ) Sb ∂G ξ dξ ∂nξ (x ∈ Sb ) (5.5) can be invoked to show that φbz = Kφb = K Sb ξ Φ0 (ξ ) ∂G ξ dξ ∂nξ (x ∈ Si ) (5.9a) is used in the latter relation. and invoking (5.8) and (5.10b) Combining (5.10a) where (5.12) .6a) (5.9b) The free-surface condition (1.11). For the analogous derivative of φi .−4πΦ0 (x) + Sb ξ Φ0 (ξ ) ∂G ξ dξ + ∂nξ Si ξ Φ0 (ξ ) ∂G ξ dξ = 0 ∂nξ (x ∈ Si ) (5.10a) and (5. the jump associated with (3. from (5. and the uniqueness proof for solutions of Laplace’s equation with combined Dirichlet and Neumann boundary conditions.6b) that ξ − φiz = Kφi − 4πKΦ0 = K−− Φ0 (ξ ) Si ∂G ξ dξ − 4πKΦ0 ∂nξ (x ∈ Si ) (x ∈ Si ) (5.

1S (Second-Order Module) – Theoretical Background by J. m. t) + · · · (1. n are used to denote the Fourier components of the same solutions.2) are distinguished from second-order components φij by the number of subscripts. N. MIT Cambridge.1) where x is a ﬁxed Cartesian coordinate system. This document is intended to describe the corresponding analysis in suﬃcient detail to explain the logic of the program. κ will be used for the wavenumber. t) = Re j φj (x)eiωj t (1. INTRODUCTION The WAMIT Second-Order Module. currently Version 3.1S. j. In the notation to be followed the subscripts i. Assuming a discrete spectrum with frequency components ωj > 0. Newman Department of Ocean Engineering. t) = Re i j φ+ (x)ei(ωi +ωj )t + φ− (x)ei(ωi −ωj )t ij ij 1 (1. the body is deﬁned by an ensemble of panels. As in the ﬁrst-order program (Version 3. The date and version number above should be referenced in this context. respectively.WAMIT V3. and t denotes time. t) + 2 Φ(2) (x. t) = Φ(1) (x. with particular care required in the case of the free surface. USA 28 March 1991 1. In the second-order solution the boundary conditions on both the body and free surface must be re-considered. The velocity potential is expanded in the form Φ(x. First-order components such as the velocity potentials φj in (1. is designed to solve the secondorder boundary-value problem for the interaction of monchromatic or bichromatic incident waves with a prescribed body. MA 02139.2) Φ(2) (x.1ID).3) . k are used to denote the frequencies of diﬀerent linear solutions. As the second-order module is extended and reﬁned this document will be modiﬁed in parallel. and . Φ(1) (x. to distinguish from the integer subscript k.

6) deﬁnes the quadratic forcing function ∂ ∂Φ(1) 1 ∂Φ(1) ∂ ∂ 2 Φ(1) ( ) − (∇Φ(1) · ∇Φ(1) ) +g 2 g ∂t ∂z ∂t ∂z ∂t Q= (1.3).3). t) = Re i j Q+ (x)ei(ωi +ωj )t + Q− (x)ei(ωi −ωj )t ij ij (1. t) ∂t2 ∂z (1.7) where the right-hand-side is to be evaluated on z = 0.5) (1.10) and 2 .2) and (1. Adopting a form for Q analogous to (1. In the following evaluations of second-order products of ﬁrst-order oscillatory quantities use is made of the relation Re(Aeiωi t )Re(Beiωj t ) = 1 Re(Aeiωi t )(Beiωj t + B ∗ e−iωj t ) 2 where (∗ ) denotes the complex conjugate.The second-order potentials φ± can be deﬁned to satisfy the symmetry relations ij φ+ = φ+ ij ji and φ− = φ−∗ ij ji (1.7) gives the expressions Q+ ij ∂ 2 φj ∂ 2 φi i i 2 ∂φj 2 ∂φi +g +g 2 ) = ωi φi (−ωj ) + ωj φj (−ωi 4g ∂z ∂z 2 4g ∂z ∂z 1 − i(ωi + ωj )∇φi · ∇φj 2 (1. The free-surface boundary conditions satisﬁed by these potentials are ∂ 2 Φ(1) ∂Φ(1) =0 +g ∂t2 ∂z ∂Φ(2) ∂ 2 Φ(2) +g = Q(x.8) As in (1. y.6) on z = 0.4) since anti-symmetric components will not contribute to (1. Here the inhomogeneous right-hand-side of the second-order free-surface condition (1. Q(x.4) it is appropriate to symmetrize the functions Q± such that ij Q+ = Q+ ij ji and Q− = Q−∗ ij ji (1.9) Combining (1.

Q− ij ∂φ∗ ∂ 2 φ∗ i i ∂ 2 φi j j 2 ∗ 2 ∂φi = ωi φi (−ωj ) − ωj φj (−ωi +g +g 2 ) 4g ∂z ∂z 2 4g ∂z ∂z 1 − i(ωi − ωj )∇φi · ∇φ∗ j 2 (1.12) 3 .11) where the right-hand-sides of (1. With these deﬁnitions the free-surface boundary condition for the second-order potential is given by −(ωi ± on z = 0. ωj )2 φ± ij ∂φ± ij = Q± +g ij ∂z (1.10-11) are evaluated on z = 0.

3) It is helpful to anticipate the relations ∂Z(κj z) ∂z = νj and ∂ 2 Z(κj z) ∂z 2 = κ2 j (2.7) . frequency ωj . INCIDENT WAVES The incident wave potential ΦI is deﬁned to be the total ﬁrst.1) In accordance with the ﬁrst-order free-surface condition.and second-order velocity potential that would exist in the absence of the body.6) Solutions for the second-order components of the incident-wave potential follow from (1.2.12) in the form φ± = ij Q± (x.1) and (1. κj sin βj . y)Z(κ± z) ij ij −(ωi ± ωj )2 + gκ± tanh κ± h ij ij 4 (2. for a ﬂuid of depth h. Z(κj z) = cosh(κj (z + h)) cosh(κj h) (2. and vector wavenumber Kj with Cartesian components (κj cos βj . Each ﬁrst-order component is deﬁned by the amplitude Aj .2) (1) j Aj Z(κj z) exp i(ωj t − Kj · x) ωj (2. 0).4) z=0 z=0 Combining (2.10-11) gives 1 Q+ = − ig 2 Ai Aj exp(−i(Ki + Kj ) · x) ij 2 2 2 κ2 − νj κ2 − νi (ωi + ωj ) j i (Ki · Kj − νi νj ) + + 2ωj 2ωi ωi ωj and 1 Q− = ig 2 Ai A∗ exp(−i(Ki − Kj ) · x) j ij 2 2 2 κ2 − νj κ2 − νi (ωi − ωj ) j i − − Ki · Kj + νi νj 2ωj 2ωi ωi ωj (2. The total ﬁrst-order velocity potential is ΦI = Re ig Here. 2 κj tanh(κj h) = ωj /g ≡ νj (2. Here βj is the angle of incidence relative to the x−axis.5) (2.

on the free surface. and Jn is the ﬁrst-kind Bessel function of order n. −ω 2 φ + g ∂φ = qF ∂z on z = 0 (3. and for the radiation potential ΦR due to the motions of the body about its ﬁxed mean position. The boundary condition on the mean position S B of the body can be written in the generic form ∂φ = qB ∂n Similarly.9) where 0 = 1.1-1. y = ρ sin θ the appropriate expansion is igAj = Z(κj z) ωj n=0 ∞ n φIj n (−i) Jn (κj ρ) cos n(θ − βj ) (2. It will be convenient to redeﬁne the coeﬃcients in (2.11) 3.2) on S B (3.1) where ω is the frequency corresponding to the component φ. VELOCITY POTENTIAL DUE TO THE BODY In this section general properties will be listed for the solution of the velocity potential ΦB due to the presence of the body. writing this equation in the form ∞ φIj = Z(κj z) n=0 Jn (κj ρ)[Ac cos nθ + As sin nθ] jn jn (2.9). a Fourier-Bessel expansion of the ﬁrst-order incident-wave potential will be required.10) where the coeﬃcients are Ac jn As jn = igAj ωj n (−i) n cos nβj sin nβj (2. In terms of the polar coordinates x = ρ cos θ.3) are applicable.8) Subsequently in the analysis of the free-surface integral of the second-order problem. 5 . This includes both the scattered potential ΦS which accounts for the eﬀects of the incident waves on the ﬁxed body.where κ± = |Ki ± Kj | ij (2. n = 2 for (n ≥ 1). In all cases the decompositions and expansions (1.

κj .3) can be applied to an arbitrary closed surface. With the additional restriction that κb is suﬃciently large. and G 2πicZ(κz)Z(κζ)H0 (κR) (2) (3. and can be found directly as the solution of the integral equation 2πσ(x) + SB ξ σ(ξ ) ∂G(x. R is the horizontal distance between the source (2) and ﬁeld points (x.7) Here the function Z is deﬁned by (2. In the ﬂuid domain the solution for each ﬁrst-order component of the potential can be expressed in the form φ(x) = SB ξ σ(ξ )G(x. ξ )dS In this equation φ represents any ﬁrst. this equation holds on the mean position of SB and on the plane z = 0 corresponding to the mean position of SF .The solution for φ can be derived from Green’s theorem in terms of the integral equation 2πφ(x) + SB ξ φ(ξ ) ∂G(x. and (3.3) reduces to the conventional form of the integral equation for φ on the body. ξ )dS (3. G is the free-surface Green function corresponding to the same frequency ω and wavenumber as φ. ξ ). in the source formulation. Consider a generic wavenumber κ which may be equal to either κi . subject to (3.6) In the subsequent analysis of the free-surface integral of the second-order problem. evanascent (local) nonradiating components of the Green function can be excluded from consideration. the parameter σ is deﬁned as the ﬁrst-order source strength. ξ ) dS = qB (x) ∂nx (3.4) For the formulation based on Green’s theorem. Since (3.5) Alternatively. ξ )dS (3. ξ ) dS = ∂nξ SB ξ qB (ξ )G(x.3) to the exact boundaries of the ﬂuid domain.3) SF + ξ qF (ξ )G(x. σ is the linear operator σ= ∂ 1 qB − φ 4π ∂n (3.2) and a suitable radiation condition at inﬁnity.9) in the farﬁeld domain outside of a partition circle of radius b which is suﬃciently large to completely surround the body.1) and (3. and it is not necessary to apply (3. For the ﬁrst-order components of the potential qF = 0.2). and H0 (κR) = J0 (κR) − iY0 (κR) is the Hankel function of the 6 . or the wavenumber κk of the second-order potential. a Fourier expansion of the ﬁrst-order body potential will be required to complement (2.or second-order velocity potential.

with the ‘quadratic forcing function’ qF deﬁned by the relations (1. For purposes of numerical evaluation.4) gives the corresponding far-ﬁeld representation of the body potential ∞ c s Hn (κρ)[Bn cos nθ + Bn sin nθ] φ(x) Z(κz) n=0 (3.9) where ξ = ρ cos θ . η = ρ sin θ . the incident-wave potential ΦI should be subtracted from the left side of the free-surface conditions (1. To avoid the evaluation of second-order derivatives of the ﬁrst-order potentials in (1. in an analogous manner to the body surface.79) may be used to expand (3.11) due to (a) quadratic interactions of the body potential with itself.3) will be considered here. the integration in the inner domain ρ < b is carried out numerically. THE FREE-SURFACE INTEGRAL The free-surface integral which is displayed as the last term in (3.9): ∞ G(x.10-11). ξ ) 2πicZ(κz)Z(κζ) n=0 n Hn (κρ)Jn (κρ ) cos n(θ − θ ) (3. 9. Since we are considering the evaluation of the second-order velocity potential ΦB due to the presence of the body. In the inﬁnite-depth limit the corresponding error is of order (κR)−2 . eq.11) and (2.6) and (1. and also from the evaluation of the quadratic forcing function.10-1.6-12).10-1.7) in a Fourier-Bessel series analogous to (2. Following the methodology of Kim and Yue. and (3.1. and (b) cross-terms between ΦB and ΦI . Thus the inner free surface is discretized with quadrilateral panels. (In the following analysis the superscript (2) will be deleted. or of order (κR)−3 in the special case where z = 0.11) 4. the free-surface integral is composed of two parts. separated by the partition circle of radius ρ = b. Graf’s addition theorem (Abramowitz & Stegun.6).second kind.) The constant c is deﬁned by −κ2 ν 2 − κ2 cosh2 (κh) = 2 κ2 h − ν 2 h + ν κ h − ν2h + ν c= (3.7) is of order exp(−CR/h) where the constant C is greater than π/2.10) where the coeﬃcients in this expansion are c Bn s Bn = 2πic n SB ξ σ(ξ )Z(κζ)Jn (κρ ) cos nθ sin nθ dS (3. Substituting this result into (3.12).5-2. which includes terms on the right side of (1.9) is valid in the domain (ρ ≥ ρ ). the surface integral is transformed using Stokes’ theorem to a form involving only 7 . Thus we seek to evaluate the diﬀerence between (1.8) The error in the far-ﬁeld approximation (3.

1. considering ﬁrst the ‘BB’ component due to quadratic interactions of the body ij potential (3. THE FOURIER INTEGRALS IN THE ANGULAR COORDINATE The integration with respect to the angular coordinate θ can be evaluated after multiplication by the angular components of the Green function (3. THE FREE-SURFACE INTEGRAL IN THE FAR FIELD In the remainder of this section the ‘far-ﬁeld integral’ in the domain (b ≤ ρ < ∞) is considered.and body-potentials are deﬁned by the FourierBessel expansions (2.9). 4.10) and (1. For this purpose the incident.| −m| + δn. +m ] ≡ λ+ mn ≡ λ− mn (4.| [δn. Thus we consider the Fourier integrals 2π 0 Q+ ij cos nθ dθ sin nθ The only nonzero components involve the integrals 2π cos θ cos mθ cos nθdθ = 0 2π π n [δn.2a) (4. We begin with the sum-frequency component Q+ . The ﬁrst task is to evaluate (1.2b) sin θ sin mθ cos nθdθ = 0 π n −m| +m ] 8 . we obtain Q+ BiBj i = 4 ∞ ∞ 2 2 ωi (κ2 − νj ) + ωj (κ2 − νi ) − 2(ωi + ωj )νi νj j i =0 m=0 H (κi ρ)Hm (κj ρ) c s c s (Bi cos θ + Bi sin θ)(Bjm cos mθ + Bjm sin mθ) − 2(ωi + ωj )κi κj H (κi ρ)Hm (κj ρ) c s c s (Bi cos θ + Bi sin θ)(Bjm cos mθ + Bjm sin mθ) m − 2(ωi + ωj ) 2 H (κi ρ)Hm (κj ρ) ρ s c s c (Bi cos θ − Bi sin θ)(Bjm cos mθ − Bjm sin mθ) (4.ﬁrst-order derivatives and line integrals around the inner and outer boundaries.11).10) and (3. respectively.1) where H denotes the ﬁrst derivative of the Hankel function with respect to its argument.10). Associating the Fourier indices ( . Details are given in the paper by Kim. 4. j) respectively.2. m) with the frequency indices (i.10). − δn.

or n = + m.|n−m| −δ .5) Substituting (4.1) are then given by 2π 0 Q+ BiBj cos nθ i dθ = sin nθ 4 ∞ ∞ 2 2 ωi (κ2 − νj ) + ωj (κ2 − νi ) − 2(ωi + ωj )νi νj j i c c s s Bi Bjm λ+ + Bi Bjm λ− mn mn s c c s Bi Bjm λ− m + Bi Bjm λ− n mn =0 m=0 H (κi ρ)Hm (κj ρ) − 2(ωi + ωj ) κi κj H (κi ρ)Hm (κj ρ) c c s s Bi Bjm λ+ + Bi Bjm λ− mn mn s c c s Bi Bjm λ− m + Bi Bjm λ− n mn (4.|n− | − δm. or m = n + .3) + c c s s Bi Bjm λ− + Bi Bjm λ+ m mn mn H (κi ρ)Hm (κj ρ) 2 c B s λ− s B c λ− ρ −Bi jm n m − Bi jm mn The second term.3) gives 9 . The Fourier integrals of the quadratic forcing function (4.n+m ] = λ− mn (4.4-5) in (4. which involves derivatives of the Hankel functions.2π cos θ sin mθ sin nθdθ = 0 2π π π [δ . can be reduced using the relation Hν = 1 (Hν−1 − Hν+1 ).2d) sin θ cos mθ sin nθdθ = 0 m [δm. the third term resulting from the angular derivatives can be reduced using the recurrence relation 2νHν /z = Hν−1 + Hν+1 to remove the factor ρ−2 . Thus 2 1 H 4 H (κi ρ)Hm (κj ρ) = −1 Hm−1 −H −1 Hm+1 +H +1 Hm+1 −H +1 Hm−1 (4. equal to one if the subscripts are equal and otherwise equal to zero. the one which is most useful is obtained by setting ν equal to and then to m: H (κi ρ)Hm (κj ρ)ρ−2 = κi κj H 4 m −1 Hm−1 + H −1 Hm+1 + H +1 Hm+1 + H +1 Hm−1 (4.4) Similarly.n+ ] = λ− m n where δmn is the Kronecker delta function.2c) (4. Three alternative relations result. The factors λ deﬁned above are nonzero if and only if = m + n.

8) 10 .6) 1 + (ωi + ωj )κi κj H −1 (κi ρ)Hm+1 (κj ρ) + H +1 (κi ρ)Hm−1 (κj ρ) 2 c c s s (Bi Bjm − Bi Bjm )(λ+ − λ− ) mn mn s c c s (Bi Bjm + Bi Bjm )(λ− m + λ− ) n mn The corresponding integrals for the diﬀerence-frequency function are ∞ ∞ 2π 0 Q− BiBj cos nθ i dθ = sin nθ 4 H 2 2 ωi (κ2 − νj ) − ωj (κ2 − νi ) − 2(ωi − ωj )νi νj j i c c∗ s s∗ Bi Bjm λ+ + Bi Bjm λ− mn mn s c∗ c s∗ Bi Bjm λ− m + Bi Bjm λ− n mn =0 m=0 ∗ (κi ρ)Hm (κj ρ) 1 ∗ ∗ − (ωi − ωj )κi κj H −1 (κi ρ)Hm−1 (κj ρ) + H +1 (κi ρ)Hm+1 (κj ρ) 2 c c∗ s s∗ (Bi Bjm + Bi Bjm )(λ+ + λ− ) mn mn s c∗ c s∗ (Bi Bjm − Bi Bjm )(λ− m − λ− ) n mn (4.7) 1 ∗ ∗ + (ωi − ωj )κi κj H −1 (κi ρ)Hm+1 (κj ρ) + H +1 (κi ρ)Hm−1 (κj ρ) 2 c c∗ s s∗ (Bi Bjm − Bi Bjm )(λ+ − λ− ) mn mn s c∗ c s∗ (Bi Bjm + Bi Bjm )(λ− m + λ− ) n mn It is helpful to deﬁne the factors 2 2 Ω± = ωi (κ2 − νj ) ± ωj (κ2 − νi ) − 2(ωi ± ωj )νi νj j i ij (4.2π 0 Q+ BiBj cos nθ i dθ = sin nθ 4 ∞ ∞ 2 2 ωi (κ2 − νj ) + ωj (κ2 − νi ) − 2(ωi + ωj )νi νj j i c c s s Bi Bjm λ+ + Bi Bjm λ− mn mn s c c s Bi Bjm λ− m + Bi Bjm λ− n mn =0 m=0 H (κi ρ)Hm (κj ρ) 1 − (ωi + ωj )κi κj H −1 (κi ρ)Hm−1 (κj ρ) + H +1 (κi ρ)Hm+1 (κj ρ) 2 c c s s (Bi Bjm + Bi Bjm )(λ+ + λ− ) mn mn s c c s (Bi Bjm − Bi Bjm )(λ− m − λ− ) n mn (4.

Thus 2π 0 Q+ BiBj cos nθ i dθ = sin nθ 8 − Λ+ H ij ∞ ∞ =0 m=0 Ω+ H (κi ρ)Hm (κj ρ) ij +H +1 (κi ρ)Hm+1 (κj ρ) −1 (κi ρ)Hm−1 (κj ρ) c c s s (Bi Bjm + Bi Bjm )(λ+ + λ− ) mn mn s c c s (Bi Bjm − Bi Bjm )(λ− m − λ− ) n mn −1 (κi ρ)Hm+1 (κj ρ) + Ω+ H (κi ρ)Hm (κj ρ) + Λ+ H ij ij +H +1 (κi ρ)Hm−1 (κj ρ) c c s s (Bi Bjm − Bi Bjm )(λ+ − λ− ) mn mn s c c s (Bi Bjm + Bi Bjm )(λ− m + λ− ) n mn (4. after noting that the ﬁrst factor contained within large parentheses in each equation is equivalent to half the sum of the second and third factors. and integrate in the radial domain b < ρ < ∞.11) are of the basic form 11 . the integrals which must be evaluated in (4.3.10) Similarly for the diﬀerence-frequency integral 2π 0 Q− BiBj cos nθ i dθ = sin nθ 8 − Λ− H ij ∞ ∞ =0 m=0 ∗ Ω− H (κi ρ)Hm (κj ρ) ij ∗ −1 (κi ρ)Hm−1 (κj ρ) +H ∗ +1 (κi ρ)Hm+1 (κj ρ) c c∗ s s∗ (Bi Bjm + Bi Bjm )(λ+ + λ− ) mn mn s c∗ c s∗ (Bi Bjm − Bi Bjm )(λ− m − λ− ) n mn ∗ −1 (κi ρ)Hm+1 (κj ρ) ∗ + Ω− H (κi ρ)Hm (κj ρ) + Λ− H ij ij +H ∗ +1 (κi ρ)Hm−1 (κj ρ) c c∗ s s∗ (Bi Bjm − Bi Bjm )(λ+ − λ− ) mn mn s c∗ c s∗ (Bi Bjm + Bi Bjm )(λ− m + λ− ) n mn (4.3) it is necessary to multiply the preceding results by the Hankel functions in (3.9) (These deﬁnitions diﬀer by multiplicative factors from the corresponding parameters used by Kim. THE INTEGRALS IN THE RADIAL COORDINATE In the context of the free-surface integral represented by the last term in (3.) Equations (4.11) 4. Since the diﬀerential element of surface area is ρdρdθ.9).6) and (4.10) and (4.and Λ± = (ωi ± ωj )κi κj ij (4.7) can then be written in a more compact form.

in (4.2) (2) J (αx)Hm (βx)Hn (γx)xdx H (1.13b) G (1. where one of the ﬁrst pair of Hankel functions in (4. β. the second Hankel function is replaced by its complex conjugate or the Hankel function of the ﬁrst kind.∞ b H (κi ρ)Hm (κj ρ)Hn (κk ρ)ρdρ = b 2 1 ∞ H (αx)Hm (βx)Hn (γx)xdx ≡ b2 F mn (α. and the contributions included from the incident-wave potential.12) Additional integrals will also be required to analyze the interaction between the body and incident-wave potentials.11).2) mn ∞ = 1 H (2) (2) (αx)Jm (βx)Hn (γx)xdx (4.13a) (4.2) (2) (αx)Hm (βx)Hn (γx)xdx (4.2) mn ∞ = 1 H ∞ (2) (1. Similarly. To make the above results more explicit. but it is retained for uniformity in the notation. of the same order and with the same argument. γ) (4.2) mn = 1 (1.12) is replaced by the corresponding Bessel function Jν .) With these deﬁnitions applied.13c) (The superscript for the third function is superﬂuous. we deﬁne the following nondimensional integrals: F (1. the appropriate radial integrals are 12 .

κj b.14) ) (2) .n (κi b.m−1.m. κj b. κj b.m−1.m+1.n (κi b.m. κj b.n (κi b. κk b) (As i − Ac i s Bjm )(λ− m n − λ− mn (4. κj b. κk b) − Λ+ G ij +G + Ω+ G ij + (2) +1. κj b.n (κi b. κj b.n (κi b. κk b) c s (Ac Bjm + As Bjm )(λ+ + λ− ) i i mn mn c Bjm (2) . κk b) + Ω+ H ij + − Λ+ H ij (2) H +1. κj b.n (κi b.n (κi b. κj b.m+1. κj b.n (κi b.n (κi b.m−1. κk b) c s (Bi Ac + Bi As )(λ+ + λ− ) jm jm mn mn + Ω+ H ij + + Λ+ H ij (2) H +1.n (κi b. κk b) s c c s (Bi Bjm − Bi Bjm )(λ− m − λ− ) n mn (2) −1.∞ b (2) Hn (κk ρ)ρdρ 2π 0 Q+ (ρ. κj b. κk b) s c (Bi Ac − Bi As )(λ− m − λ− ) jm jm n mn (2) −1. κj b.m.m+1. κk b) s c c s (Bi Bjm + Bi Bjm )(λ− m + λ− ) n mn (2) −1. κk b) (2) . κj b. κk b) i 2 b 8 ∞ ∞ =0 m=0 Ω+ F ij (2) . κk b) c c s s (Bi Bjm − Bi Bjm )(λ+ − λ− ) mn mn c c s s (Bi Bjm + Bi Bjm )(λ+ + λ− ) mn mn + Ω+ F ij + Λ+ F ij +F + Ω+ G ij (2) +1. κj b.n (κi b. κk b) (2) . κj b.m+1. κj b.n (κi b.m.n (κi b.n (κi b.m. κk b) (2) G +1.m+1. κk b) + Λ+ G ij (2) −1.m−1. κk b) + (2) F +1. κk b) s c (Bi Ac + Bi As )(λ− m + λ− ) jm jm n mn c s (Bi Ac − Bi As )(λ+ − λ− ) jm jm mn mn 13 . κj b. κk b) c s (Ac Bjm − As Bjm )(λ+ − λ− ) i i mn mn c s (As Bjm + Ac Bjm )(λ− m + λ− ) n mn i i (2) −1.m+1.m−1. θ) ij cos nθ dθ = sin nθ − Λ+ F ij (2) −1.n (κi b. κk b) (2) .m−1.n (κi b.n (κi b.n (κi b.m. κj b.

κk b) (1) .n (κi b.n (κi b. κk b) (1) . κj b. κj b. κj b.m+1. κk b) s c (Bi A∗c + Bi A∗s )(λ− m + λ− ) jm jm n mn c s (Bi A∗c − Bi A∗s )(λ+ − λ− ) jm jm mn mn 14 .n (κi b.n (κi b. θ) ij cos nθ dθ = sin nθ − Λ− F ij (1) −1. κj b.m.m+1.m+1. κk b) s ∗c c ∗s (Bi Bjm + Bi Bjm )(λ− m + λ− ) n mn (1) −1. κj b. κk b) − Λ− G ij +G + Ω− G ij + (1) +1. κk b) + (1) F +1.n (κi b. κj b.n (κi b.15) ) (1) . κj b.n (κi b. κj b.m.m+1.m.n (κi b.m. κj b.n (κi b.n (κi b.m+1. κj b. κk b) + Ω− H ij + − Λ− H ij (1) H +1.n (κi b.m−1. κj b. κk b) s ∗c c ∗s (Bi Bjm − Bi Bjm )(λ− m − λ− ) n mn (1) −1. κj b.m−1.m+1. κk b) + Λ− G ij (1) −1.n (κi b.n (κi b.n (κi b.m. κj b.m−1.n (κi b. κk b) s c (Bi A∗c − Bi A∗s )(λ− m − λ− ) jm jm n mn (1) −1.m.m−1. κk b) (1) G +1. κk b) c s (Bi A∗c + Bi A∗s )(λ+ + λ− ) jm jm mn mn + Ω− H ij + + Λ− H ij (1) H +1.m−1. κk b) ∗c ∗s (Ac Bjm − As Bjm )(λ+ − λ− ) i i mn mn ∗c ∗s (As Bjm + Ac Bjm )(λ− m + λ− ) n mn i i (1) −1.m−1. κj b. κk b) ∗c ∗s (Ac Bjm + As Bjm )(λ+ + λ− ) i i mn mn ∗c Bjm (1) . κj b. κk b) (1) . κj b.n (κi b.n (κi b. κj b.n (κi b. κk b) (As i − Ac i ∗s Bjm )(λ− m n − λ− mn (4. κj b.∞ b (2) Hn (κk ρ)ρdρ 2π 0 Q− (ρ. κk b) i 2 b 8 ∞ ∞ =0 m=0 Ω− F ij (1) . κk b) c ∗c s ∗s (Bi Bjm − Bi Bjm )(λ+ − λ− ) mn mn c ∗c s ∗s (Bi Bjm + Bi Bjm )(λ+ + λ− ) mn mn + Ω− F ij + Λ− F ij +F + Ω− G ij (1) +1.

n mn In the three-dimensional space . where λ+ = 2π and λ− = 0 000 000 the four factors can be simpliﬁed in the forms (λ+ + λ− ) = πδ mn mn (λ− m − λ− ) = πδ n mn .m+n .14-15) can be expressed in terms of the separate contributions from each of the three surfaces.16d) (λ+ − λ− ) = πδn. whereas the last two factors are nonzero only on the complementary plane n = + m.2). (4. Collectively these three planes form a cone with triangular sections. κj b.16) apply also on the intersections of each pair of adjacent planes. κk b) (4. +m (κi b.14-15) are equivalent to 15 . The four expressions (4.m−1. Excluding the special case = m = n = 0.m.m. κj b. κk b) (4.m−1. −m (κi b. mn mn (λ− m + λ− ) = πδn. − πδm.m.m− ij − Λ± ij (κi b. Thus the double sums in (4. κj b.m+1.17a) V (ij)± (F ) m = Ω± F .m+1. κk b) ij +F +1.17c) If the series are truncated at the upper limit = M . +m (κi b. κk b) − Λ± F ij −1.16a) (4. κk b) (4.m− (κi b. κj b. +m (κi b. −m (κi b. −m (κi b. the ﬁrst two factors are nonzero only on the two planes = m + n and m = + n. and with its apex at the origin. κk b) ij + Λ± F −1. κj b. m. with reference to the deﬁnitions (4. κk b) F −1. +m +m +n +n (4. κj b. κj b.17b) (ij)± W m (F ) = Ω± F . κj b.m+n + πδm. A compact form can be derived if we ﬁrst deﬁne the three functionals U (ij)± m (F ) = Ω± F ij . κj b.16c) (4. κk b) +F +1.m−1.16b) (4. κk b) + F +1. m = M . n.The four factors which involve sums and diﬀerences of the parameters λ will now be considered.m− (κi b.m+1.

18) c s (Bi Ac + Bi As ) jm jm s Ac − B c As ) (Bi jm jm i m= +n + =1 W (ij)+ (2) ) m (F c c s s (Bi Bjm − Bi Bjm ) s c c s (Bi Bjm + Bi Bjm ) c s (Ac Bjm − As Bjm ) i i c s (As Bjm + Ac Bjm ) i i +W +W (ij)+ (2) ) m (G (ij)+ (2) ) m (H c s (Bi Ac − Bi As ) jm jm s Ac + B c As ) (Bi jm jm i m=n− 16 .∞ b (2) Hn (κk ρ)ρdρ 2π 0 Q+ (ρ. θ) ij M cos nθ dθ = sin nθ (ij)+ U m (F (2) ) c c s s (Bi Bjm + Bi Bjm ) s c c s (Bi Bjm − Bi Bjm ) πi 2 b 8 (1 + δ n ) =n +U +U M −n (ij)+ (2) ) m (G (ij)+ (2) ) m (H c s (Ac Bjm + As Bjm ) i i c s (As Bjm − Ac Bjm ) i i c s (Bi Ac + Bi As ) jm jm s c (Bi Ac − Bi As ) jm jm m= −n ± =0 (1 + δ 0 ) V +V +V n−1 (ij)+ (F (2) ) m c c s s (Bi Bjm + Bi Bjm ) s c c s (Bi Bjm − Bi Bjm ) (ij)+ (G (2) ) m (ij)+ (H(2) ) m c s (Ac Bjm + As Bjm ) i i c s (As Bjm − Ac Bjm ) i i (4.

2) (1.13b.19) c s (Bi A∗c + Bi A∗s ) jm jm s A∗c − B c A∗s ) (Bi jm jm i m= +n + =1 W (ij)− (F (1) ) m c ∗c s ∗s (Bi Bjm − Bi Bjm ) s ∗c c ∗s (Bi Bjm + Bi Bjm ) ∗c ∗s (Ac Bjm − As Bjm ) i i ∗c ∗s (As Bjm + Ac Bjm ) i i +W +W (ij)− (G (1) ) m (ij)− (H(1) ) m c s (Bi A∗c − Bi A∗s ) jm jm s A∗c + B c A∗s ) (Bi jm jm i m=n− Special attention is required when the integers . for suﬃciently large values of x. From the asymptotic approximations of the Bessel and Hankel functions it is apparent that. and similarly for the case m = ±1 and for the integrals (4. and for the factors of two applied to the ﬁrst term in each of the ﬁrst two sums.4. the integrands are proportional to x−1/2 exp(iσx). 17 . m are equal to zero.18-19).13).m. that the contributions from these equations are equivalent to those from (4. n is equal to zero.n = −F1. 4. θ) ij M cos nθ dθ = sin nθ (ij)− U m (F (1) ) c ∗c s ∗s (Bi Bjm + Bi Bjm ) s ∗c c ∗s (Bi Bjm − Bi Bjm ) πi 2 b 8 (1 + δ n ) =n +U +U M −n (ij)− (1) ) m (G (ij)− (1) ) m (H ∗c ∗s (Ac Bjm + As Bjm ) i i ∗c ∗s (As Bjm − Ac Bjm ) i i c s (Bi A∗c + Bi A∗s ) jm jm s c (Bi A∗c − Bi A∗s ) jm jm m= −n ± =0 (1 + δ 0 ) V +V +V n−1 (ij)− (F (1) ) m c ∗c s ∗s (Bi Bjm + Bi Bjm ) s ∗c c ∗s (Bi Bjm − Bi Bjm ) (ij)− (G (1) ) m (ij)− (H(1) ) m ∗c ∗s (Ac Bjm + As Bjm ) i i ∗c ∗s (As Bjm − Ac Bjm ) i i (4. Thus it follows from (4. Note also that when n = 0 the contributions from the ﬁrst two sums in (4. giving a net contribution of twice that of the ﬁrst sum in the case of the cosine integrals (upper factors) and zero for the sine integrals (lower factors). Since the Bessel and (1. with regard for the limits of each sum in (4. It can then be shown. m.m.18-19) are identical.c).17) that U 0 = W 0 and V0m = W0m .2) Hankel functions are odd functions of the (integer) order it follows that F−1.∞ b (2) Hn (κk ρ)ρdρ 2π 0 Q− (ρ. EVALUATION OF THE RADIAL INTEGRALS Suitable algorithms are required to evaluate the three families of integrals deﬁned by (4.14-15) when one or more of the integers .n .

where the factor σ involves sums and diﬀerences of the three parameters α.2) Hν (z) 1 π 2 exp(±i(z − νπ − )) πz 2 4 (4.21) ¿From the basic relations between the Hankel and Bessel functions the integrals (4. Hereafter it is assumed that the three parameters α. before deforming the contour of integration into the complex plane. When the order is large it also is necessary to consider the complementary approximations Jν (z) ∼ z ν . due to the very slow convergence and oscillatory feature of the integrand. To determine the appropriate contours for each of the integrals (1. Y are real on the real axis.2) = Jν ± iYν . These integrals are convergent provided σ = 0.22) may be deformed in the ﬁrst quadrant of the complex plane z = x + iy.13). The former assumption is appropriate since these parameters correspond physically to the products of three positive wavenumbers with the partition radius b. for large x. γ. Thus (4.21) that severe cancellation errors will result from this approach. and the corresponding asymptotic approximations (1. These diﬃculties can be overcome simply by adopting appropriate contours of integration in the complex z-plane. Numerical integration along the semi-inﬁnite real x-axis is not practical. Such a decomposition is essential in some cases for the contour integrals.13b).13c) can be evaluated as one-half times the sum of the two integrals deﬁned by (4. γ > 0. and a convenient choice is the straight line upward parallel to the imaginary axis from z = xmax to z = xmax + i∞. Yν ∼ z −ν (4. However when >> αx or m >> βx it is clear from (4. If σ > 0 the contour of integration in (4. and in all cases σ = (±α ± β − γ).22) is replaced in this case by the semi-inﬁnite integral 18 . depending on the sign of σ. since for large values of the argument Jν (z) includes two components proportional respectively to exp(±iz) which may require diﬀerent choices of the contour of integration. This diﬃculty is avoided by integrating (4. The remaining integrals are of the general form ∞ xmax f (x)eiσx dx (4. β. β. and a similar approach can be followed with (4. and σ = 0. The latter assumption will be considered in future work. consideration must be given to the basic deﬁnitions Hν Bessel functions J.13) along a ﬁnite segment of the positive real axis 1 ≤ x ≤ xmax . where the (4.22) where f (x) denotes a slowly-varying function of order x−1/2 .20) which are valid for |z| >> 1 and |z| >> ν. This procedure is robust provided the parameter xmax is suﬃciently large to satisfy the inequalities ≤ αxmax and m ≤ βxmax .13a).

with the same integration algorithm and diﬀerent integrands.23-4) numerical integration is practical using adaptive Romberg quadratures to ensure a speciﬁed tolerance of accuracy. when σ < 0.23) Conversely. and also in the integrals (4.13a. It is computationally eﬃcient to evaluate all three integrals (4. The length of each subsequent segment is increased in a linear manner. This precludes the possibility of false convergence for one particular integrand. and permits the simultaneous recursive evaluation of the required Bessel and Hankel functions of all orders.1.23-4) where the arguments are in the ﬁrst quadrant.22) is replaced by ∞ xmax f (x)e −i iσx ∞ ∞ dx = xmax ∗ f ∗ (x)e−iσx dx −iσx+σy ∗ ∗ = (4.∞ xmax f (x)eiσx dx = i exp(iσxmax ) ∞ 0 f (xmax + iy)e−σy dy. apart from a multiplicative exponential factor. In the program this procedure is followed for the complete set of required integrals. ﬁrst within each ﬁnite segment to determine that the order of the Romberg quadratures is suﬃcient. The parameter xmax is set equal to the largest value of the ratios /α or m/β. the contour of integration may be deformed in the fourth quadrant along a similar contour parallel to the negative imaginary axis.c) together.13). n). In both the ﬁnite integral along the real axis and the semi-inﬁnite integrals (4. for both the integral along the real axis where the arguments are all real. and the conjugate function f ∗ (z) therefore involves the same integrands except that the Hankel functions of ﬁrst. of all combinations of the orders . Typically a tolerance of 10−8 is used.b. For programming purposes it is convenient to modify these integrals so that in all cases the contour of integration is in the ﬁrst quadrant. the length of the ﬁrst segment is set equal to four divided by the maximum value of the exponential factor σ. These will be described separately below.and second-kind are reversed. Convergence tests are applied to all members of this set simultaneously. Two convergence tests are required. xmax is set equal to one and the integral along the real axis is skipped. (4. with absolute accuracy speciﬁed if the integral is smaller than one in absolute value. by the integrands of (4.18-19) and for all three of the integrals deﬁned in (4. m. if σ < 0. except that if the largest ratio is less than 1. based on the maximum value of |σ|. and secondly after the (converged) integral in each segment is added to the total integral to determine when the range of integration can be truncated. Thus.24) (σ < 0) 0 f (xmax + iy)e dy Note that the function f (z) is deﬁned. to take advantage of the exponentially diminishing magnitude of the integrand in that segment. n required in (4. and relative accuracy in the converse case. For the semi-inﬁnite integrals in y. proportional to the number of the segment. so that large values of this parameter do not retard convergence. for all required combinations of the orders ( . (σ > 0) (4.13). and the integration of the entire set continues until the convergence test is satisﬁed for the entire set. The domain of integration is subdivided. 19 . m. Eﬀective subroutines for the Bessel and Hankel functions are required.

pp. equation 9. and these are then combined to form the corresponding Hankel functions. where N is the maximum order required. selecting whichever of these functions has the largest absolute value to avoid instability near the zeros on the real axis. and the Hankel functions are evaluated from the relations J ± iY .1. For the Hankel functions of the ﬁrst kind. These integrals are deﬁned 20 .46). These ascending series are used in the quadrilateral domain of the plane z = x + iy such that x ≥ 0. 0 ≤ y ≤ 3 and x + y ≤ 8. sin z (Abramowitz & Stegun. In this case it is appropriate to use as the normalization relations the Neumann series for cos.1 and 66. (Abramowitz & Stegun equations 9. suitable for small and large values of |z|. with complex argument z in the ﬁrst quadrant.1. but the latter can be evaluated directly from the (2) (1) relation Hν = 2Jν − Hν . Normalization is based on the Neumann series with unit value (Abramowitz & Stegun. §6. Hν . (1.25) where Hν denotes any of the three functions Jν . the coeﬃcients in Luke’s Tables 66.13). giving rational-fraction approximations as the ratios of tenth-degree polynomials. The rational-fraction approximations are not eﬀective for the Hankel functions of the second kind. forward recursion is used based on the starting values of order zero and one. the subroutine evaluates the normalized functions exp(iz)Jν (z) and applies an appropriate exponential factor to the ﬁnal integrand.6 have been extended to n = 10.1 . and corrected in the ﬁnal evaluation of the integrand. For Jν (x) the backward recursion algorithm of Miller (cf.For the functions with real argument the functions Jν (x) and Yν (x) are evaluated following the general methodology described in Numerical Recipes. Abramowitz & Stegun. For checking purposes it is useful to derive recursion formulae for the integrals (4. equations 9.2) The exponential normalization factors exp(∓iz) are applied to Hν .47-8). To avoid the possibility of overﬂow. with relative accuracy preserved for x << ν. For this purpose we consider the more general integrals F (µ) mn ∞ = 1 H (αx)Hm (βx)Hn (γx)xµ dx (1) (2) (4.10-11). using the values of Jν obtained from backward recursion. A similar backward recursion algorithm is used to evaluate the functions Jν (z) with complex argument z in the ﬁrst quadrant. using the same starting value M deﬁned above except that |x| is replaced by |z|. Y1 and the functions Yν are evaluated for all required higher orders by forward recursion. 385-6) is used with the starting value ν = M chosen to be the smallest even integer M which satisﬁes the inequalities M > 1. To obtain suﬃcient accuracy in the domain where |z| is large. respectively. Comparison of the two complementary algorithms along the common boundary of the partition indicates that the minimum relative accuracy of this subroutine is 14 decimals.1 and Y0. rational fraction approximations of the form described by Luke (Table 66) are eﬀective. For small |z| ascending series are summed to evaluate J0. Numerical experiments indicate that the accuracy of the results is at least 14 decimals.4. The evaluation of these is based on two complementary algorithms. The series are truncated with a total of 22 terms.36|x| + 24 and M > N + 12. Chebyshev expansions are used to evaluate the functions Y0 . For larger values of z outside the quadrilateral domain.1. or Hν .

m. In particular.m.m.n+1 ] 2 2m 2n (4.m.m−1. Integrating the derivative of the integrand and employing the formula Hν = Hν−1 − (ν/z)Hν to diﬀerentiate the integrand in accordance with the chain rule.m.n + γF . and that the triple product of Hankel functions H (α)Hm (β)Hn (γ) which appears in (4.n +βF .m.n + γF . it follows that −H (α)Hm (β)Hn (γ) = αF (µ) (µ) (µ) −1.13b-c).m.m.n + F .n ] = [F .n F .m.n+1 = 2n αF γ( + m + n − 1) −1.26) Using the recursion 2νHν /z = Hν−1 + Hν+1 to evaluate the last term in (4.m−1. with the index ν replaced successively by . as is required in (4. three relations are obtained as follows: F (µ−1) .28c) + H (α)Hm (β)Hn (γ) − F .m.m.m.n = α (µ) β γ (µ) (µ) (µ) (µ) (µ) [F −1. The only requirements here are that the starting values for the recursion are evaluated for the appropriate combination of Bessel and Hankel functions. including the Hankel functions of the ﬁrst and second kind.n + F +1.n + βF .m−1.n−1 (4. m.27) Combining each of these alternative relations with (4.m.28a) + H (α)Hm (β)Hn (γ) − F −1.n ] = [F .n−1 (4. n.28) 21 .in the ordinary sense for µ < 3/2 (assuming the sum of the wavenumbers is nonzero).28).m−1.m+1.n +γF .n + βF .n F .n−1 −( +m+n−µ)F (µ−1) .n = 2m αF β( + m + n − 1) −1.13).m.n−1 The relations used here for diﬀerentiation and recursion of the Hankel functions apply to any linear combination of Bessel functions of the ﬁrst and second kind. Thus the same generality applies to the relations (4.m.n + γF .n = 2 αF α( + m + n − 1) −1.28b) + H (α)Hm (β)Hn (γ) − F .m−1. one of the Hankel functions may be replaced by the Bessel function Jν .m.m.n + βF .26) and setting µ = 1 gives the desired relations F +1.n−1 (4.n (4.m.m+1. and will be applied ultimately with µ = 1.n−1 + F .m−1.

Berkeley. However the inhomogeneous terms in (4. I.b.28) for . Kim.-H. More importantly. Lee. Monochromatic incident waves. G. for large values of α.13). 1990 ‘Second-order sum-frequency wave loads on large-volume structures.’ J.. Washington and Dover. Fluid Mech.) The recursion relations are nevertheless useful to check the subroutines used to evaluate (4. Government Printing Oﬃce. First. n in (4.. and Mathematical Tables. 010. New York. 22 .28) can be used to evaluate the integrals F . γ the recursion relations are unstable with substantial cancellation error occurring over several successive values of the integers.-H. γ = O(1). M. M. Kim.18-19) up to the maximum order M = 32. 1983 ‘Second-order hydrodynamic eﬀects on ocean platforms. 1989 ‘First. However this procedure has two important disadvantages. Ship Research.’ J. 110. to appear. Ocean Research. D. and numerical tests indicate that there is a substantial loss of accuracy in the use of (4. to appear. 557–593. & Newman.28) appear to be important in this context. 001. 002.13) are required in (4. & Stegun. Graphs. 020. n) ≤ M it is suﬃcient to evaluate the eleven starting values where the three integers have the values 000.is likewise replaced by the corresponding product of the appropriate Bessel and Hankel functions. F.e.-H.’ J. Fluid Mech. i. M. REFERENCES Abramowitz. m. 011. Bichromatic incident waves and body motions. 235–264.’ International Workshop on Ship and Platform Motions. to evaluate only the required ‘two-dimensional’ combinations of . K. 211. P. Collectively the three recursion relations (4. M. D. provided suitable starting values are available. J. 200.13a. m. m.19) it is not eﬃcient to evaluate the ‘three-dimensional’ combinations as is required in the recursion relations.and second-order wave eﬀects on a submerged spheroid.18) and (4. Extensive tests have been made in this manner.c) evaluated at the lower limit of integration. T. β. H for any combination of the three orders . β. 200.. 111. Ogilvie. n. To proceed in this manner throughout a three-dimensional cube 0 ≤ ( . 101. with a speciﬁed Romberg tolerance (typically 10−8 ) the diﬀerences between the left. C. 1990 ‘The complete second-order diﬀraction solution for an axisymmetric body Part 2. (Since forward recursion is stable for the Hankel functions it was originally thought that this problem would not be serious. n > O(10) unless the parameters α. 1989 ‘The complete second-order diﬀraction solution for an axisymmetric body Part 1. K. A. N. 5. 1964 Handbook of Mathematical Functions with Formulas.-H. the integrand of (4. throughout the two-dimensional space where the three integrals (4. & Yue. P. Kim.and right-hand sides of the recursion relations (4. m.’ App.28) are consistently smaller than this tolerance. & Yue. 100.

- Generalized Thermoelastic Wave Propagation in Circumferential Direction of Transversely Isotropic Cylindrical Curved PlatesUploaded byjperdigon9634
- Dynamic Analysis of Me MsUploaded byHelton Luiz Calado
- Me ScopeUploaded byMelissa Hoppenstedt Hoge
- Tubaldi, Dall'Asta, Dezi 15wcee PRSI BridgesUploaded byEnrico Calcetto
- Cen Et Al (2016) Effect of Mass Ratio on Hydrodynamic Response of a Flexible CylinderUploaded byKarl von Terzaghi
- Blade Group Resonant StressesUploaded bybharathvg8096
- 1-s2.0-S0022460X03003833-mainUploaded byGaurav Verma
- 1-s2.0-S1878522009001246-mainUploaded byAbhileshWarade
- Modes Indicate Cracks in Wind Turbine BladesUploaded bytylerdurdane
- 2005_3.pdfUploaded byNick Bester
- Zheng-Wang Qu et al- The Huggins band of ozone: Unambiguous electronic and vibrational assignmentUploaded byMaxnamew
- 1112012913Dynamics Response Spectrum Analysis.pdfUploaded byHectoreRodriguezl
- 7-3Uploaded bysarasmile2009
- Mdof Frf CurvefitUploaded byAnonymous ep7LE5ZdP5
- GL Ship Vibration 09Uploaded byzeek77
- Mathematics Ncert Ch03 Pair of Linear Equations in Two Variables Ex 3.6 Ans NvbUploaded byarpit
- Cavity Drag and Flow OscillationsUploaded byChris Han
- Acoustic LoggingUploaded byWassef MB
- J. Bredenbeck et al- The vibrational spectrum of deuterated phosphaethyne: A quantum mechanical, classical, and semiclassical analysisUploaded byMaxnamew
- T1073-79.pdfUploaded byJay Patel
- mathematicsUploaded byJain Pradeep
- 10th MathsUploaded byYou Are Not Wasting TIME Here
- Windowing Time DataUploaded byVinay Cam
- Serpentine Mode Converter DesignUploaded byOmRanjan
- Sensor Location Effect on Flexible Robot ControlUploaded byhussalkhafaji
- A Baseline and Vision of Ultrasonic Guided Wave Inspection PotentialUploaded byDemian Pereira
- EMCF-9-F14Uploaded byShitPre Super
- ITE 2702 - CM moratuwa universityUploaded byDilini Dakshika Herath
- AWB130 Dynamics 04 SpectrumUploaded byKishore Mylavarapu
- ANSYS57_tut_modalUploaded bydiclovis

- z-013.pdfUploaded byspsingh72
- P-100(Process).pdfUploaded byspsingh72
- Ships HydrodynamicsUploaded byspsingh72
- Norsok M-501 ed 6 Clarification log 2015 rev 1_finished.pdfUploaded byspsingh72
- FPI Rules E-Feb14Uploaded byspsingh72
- Norsok M-501 FAQ.pdfUploaded byspsingh72
- Fatigue RulesUploaded byspsingh72
- fea-dundesUploaded byspsingh72
- Buckling Strength of Plated StructuresUploaded byspsingh72
- OSS-302_Offshore Riser Systems_Oct 2003Uploaded byspsingh72
- CAP437(Feb 2013)Uploaded byspsingh72
- Rigzone CareerUploaded byspsingh72
- Os e406_2010 04(Free Fall Lifeboats)Uploaded byspsingh72
- 21km+2011Uploaded byspsingh72