Nonreecting Boundary Conditions for Euler

Equations in Generalized Curvilinear Coordinates

January, 2010, University of Toledo
Adrian Sescu

current aliation (2013): Department of Aerospace Engineering, Mississippi State University, MS
39762
Abstract
This work was intended to derive nonreecting boundary conditions in generalized curvilinear coordinates
(three-dimensions), relying on the original analysis of Giles ([1]). The theoretical analysis is almost com-
pletely done (except the rigorous proof of the well-possednes in the general case), but the equations are not
tested whatsoever. The document is made available online with the hope that it will provide guidance to
those who are interested in pursuing such an analysis and validation.
1 3D Linearized Euler Equations in Cartesian Coordinates
The non-linear, non-conservative (primitive) form of the Euler equations in Cartesian coordinates
can be compactly written:
Q
t
+AQ
x
+BQ
y
+CQ
z
= 0 (1)
where Q =
_
u v w p
_
T
is the vector of the primitive variables and A, B and C are 5 5
matrices.
A =
_

_
u 0 0 0
0 u 0 0
1

0 0 u 0 0
0 0 0 u 0
0 p 0 0 u
_

_
; B =
_

_
v 0 0 0
0 v 0 0 0
0 0 v 0
1

0 0 0 v 0
0 0 p 0 v
_

_
; C =
_

_
w 0 0 0
0 w 0 0 0
0 0 w 0 0
0 0 0 w
1

0 0 0 p w
_

_
(2)

sescu@ae.msstate.edu
1
Here, is the density, p is the pressure, and u, v and w are the velocity components. The
linearization is done by decomposing the primitive vector Q into a mean part and a perturbation
part:
Q = Q+Q

(3)
where
Q =
_
u v w p
_
T
and Q

=
_

_
T
(4)
and because the perturbations are assumed to be much smaller than the mean ow, only the terms
of order O(Q

) are kept, while the other terms of order O(Q

2
) are neglected. Thus the linearized
Euler equations, after the variables are non-dimensionalized using the mean density and the speed
of sound, are:
Q
t
+AQ

x
+BQ

y
+CQ

z
= 0 (5)
where the matrices A, B and C are dened as
A =
_

_
u 1 0 0 0
0 u 0 0 1
0 0 u 0 0
0 0 0 u 0
0 1 0 0 u
_

_
; B =
_

_
v 0 1 0 0
0 v 0 0 0
0 0 v 0 1
0 0 0 v 0
0 0 1 0 v
_

_
; C =
_

_
w 0 0 1 0
0 w 0 0 0
0 0 w 0 0
0 0 0 w 1
0 0 0 1 w
_

_
(6)
2 3D Linearized Euler Equations in Curvilinear Coordinates
Consider the transformation between the curvilinear coordinates and the Cartesian coordinates:
= (t)
= (x, y, z, t)
= (x, y, z, t) (7)
= (x, y, z, t)
Assume for now that the grid is not moving: the time dependence in the above relations is
dropped out. Using the chain-rule formulation applied to Eq. (5), the linearized Euler equations are
written in curvilinear coordinates system as:
2
Q

t
+
_

x
A+
y
B+
z
C

+
_

x
A+
y
B+
z
C

(8)
+
_

x
A+
y
B+
z
C

= 0
Now denote

A =
x
A+
y
B+
z
C

B =
x
A+
y
B+
z
C (9)

C =
x
A+
y
B+
z
C
As a result, Eq. (8) can be written as:
Q

t
+

AQ

+

BQ

+

CQ

= 0 (10)
The matrices

A,

B and

C are:

A =
_

_
U
x

y

z
0
0 U 0 0
x
0 0 U 0
y
0 0 0 U
z
0
x

y

z
U
_

_
(11)

B =
_

_
V
x

y

z
0
0 V 0 0
x
0 0 V 0
y
0 0 0 V
z
0
x

y

z
V
_

_
(12)

C =
_

_
W
x

y

z
0
0 W 0 0
x
0 0 W 0
y
0 0 0 W
z
0
x

y

z
W
_

_
(13)
3
The contravariant mean velocities are given by:
U =
x
u +
y
v +
z
w
V =
x
u +
y
v +
z
w (14)
W =
x
u +
y
v +
z
w
3 Fourier Analysis of the 3D Linearized Euler Equations in Curvilinear
Coordinates
A three-dimensional wave-like solution in the form:
Q

=

Qe
i(k+l+mt)
(15)
is considered and introduced into Eq. (10). This will generate a matrix equation for the 3D wave
amplitudes as unknowns. The matrix has the form:
_
I + k

A+ l

B+ m

C
_

Q =
_

_

1

2

3
0
0 0 0
1
0 0 0
2
0 0 0
3
0
1

2

3

_

_
_
_
_
_
_
_
_
_
_
r
1
r
2
r
3
r
4
r
5
_
_
_
_
_
_
_
_
_
= 0 (16)
or
_

l
1

l
2

l
3

l
4

l
5
_
_

_

1

2

3
0
0 0 0
1
0 0 0
2
0 0 0
3
0
1

2

3

_

_
= 0 (17)
where
= Uk + V l + Wm

1
=
x
k +
x
l +
x
m

2
=
y
k +
y
l +
y
m (18)

3
=
z
k +
z
l +
z
m
4
Eq. (16) or (17) is a homogeneous set of equations; in order to admit non-trivial solution the
determinant of the matrix must satisfy:
det(I + k

A+ l

B+ m

C) = 0 (19)
or

3
_

2
1

2
2

2
3
_
= 0 (20)
3.1 Eigenvalues
If k is the unknown of Eq. (20) the rst three roots are identical:
k
1,2,3
=
V l Wm
U
(21)
The other two roots are determined by solving the equation:
_
Uk + V l + Wm
_
2
(
x
k +
x
l +
x
m)
2
(
y
k +
y
l +
y
m)
2
(
z
k +
z
l +
z
m)
2
= 0 (22)
Denote:
|| =
_
(
x
)
2
+ (
y
)
2
+ (
z
)
2
|| =
_
(
x
)
2
+ (
y
)
2
+ (
z
)
2
|| =
_
(
x
)
2
+ (
y
)
2
+ (
z
)
2
|| =
x

x
+
y

y
+
z

z
(23)
|| =
x

x
+
y

y
+
z

z
|| =
x

x
+
y

y
+
z

z
= V l Wm
= l|| + m||
= l
2
||
2
+ m
2
||
2
+ 2lm||
The discriminant is:
= 4
_
+ U
_
2
4
_
||
2
U
2
_
_

2
_
5
So the fourth and the fth roots are:
k
4,5
=
2
_
+ U
_

2
_
||
2
U
2
_
or
k
4,5
=
_
+ U
_
(1 S)
_
||
2
U
2
_ (24)
where
S =

_
1
_
||
2
U
2
_
(
2
)
_
+ U
_
2
(25)
If is complex with Im() > 0 then the right propagating waves are those for which Im(k) > 0.
This is because the amplitude of the waves is proportional to exp[Im()(t /c)] where c is the
aparent velocity of propagation of the amplitude. If and k are real then the standard result is
the analysis of dispersive wave propagation using the group velocity. For real the incoming waves
are those which either have Im(k) > 0, or have real k and
k
> 0 To determine the direction of
propagation of the waves, lets solve Eq. (20) for too. The roots are easily found:

1,2,3
= Uk + V l + Wm (26)

4,5
= Uk + V l + Wm
_

2
1
+
2
2
+
2
3
(27)
The analysis of the rst three waves is straigthforward. If Im() > 0 then it is clear that
Im(k) > 0. If Im() = 0 the group velocities for the rst three roots are all equal to U which means
that, for U > 0 the corresponding waves are incoming waves at x = 0 and outgoing waves at x = 1.
For the fourth and the fth roots the group velocities are:
c
g4,5
= U
2
1

x
+ 2
2

y
+ 2
3

z
2
_

2
1
+
2
2
+
2
3
(28)
If the l and m components of the wavenumber are small (hypothesis used to derive the approx-
imate boundary conditions) the second term of the right-hand-side of Eq. (28) tends to ||. The
6
fourth wave is incoming wave at x = 0 and outgoing wave at x = 1. For 0 < U < || (subsonic
case) the fth wave is an outgoing wave at x = 0 and an incoming wave at x = 1. For supersonic
case (U > ||) the fth wave behaves as the fourth wave (incoming wave at x = 0 and outgoing
wave at x = 1). It can be proved that if and/or S are complex then one of the two roots for
k has imaginary part, while the other has a negative imaginary part: k
4
is dened to be the root
with positive imaginary part (right-running wave), and k
5
is a root with negative imaginary part
(left-running wave).
Remark: U is the contravariant velocity: the relation 0 < U < || is translated into Cartesian
coordinates as 0 < u < 1, taking into account that
x
= 1,
y
= 0 and
z
= 0 (in Cartesian
coordinates).
Next the right and the left eigenvectors corresponding to matrix equations (16) and (17), respec-
tively, will be determined.
3.2 Eigenvectors
3.2.1 Root 1: entropy wave
k
1
=
V l Wm
U
, = Uk
1
+ V l + Wm (29)
The right eigenvectors are determined by solving the matrix equation (16). First, denote

1
=
l

,
2
=
m

(30)
The matrix equation (16) becomes:
_

_
0
1

2

3
0
0 0 0 0
1
0 0 0 0
2
0 0 0 0
3
0
1

2

3
0
_

_
_
_
_
_
_
_
_
_
_
r
1
r
2
r
3
r
4
r
5
_
_
_
_
_
_
_
_
_
= 0 (31)
The general solution (depending on 3 independent constants C
1
, C
2
and C
3
) to this set of equa-
tions is given by:
7
_
_
_
_
_
_
_
_
_
r
1
r
2
r
3
r
4
r
5
_
_
_
_
_
_
_
_
_
=
_
_
_
_
_
_
_
_
_
C
1

2
C
2
+
3
C
3

1
C
2

1
C
3
0
_
_
_
_
_
_
_
_
_
(32)
The choice of these constants will give the rst three right or left eigenvectors. The rst right
eigenvector (keeping the notation of Giles [1]) is chosen by setting C
1
= 1/|| and C
2
= C
3
= 0.
u
R
1
=
_
_
_
_
_
_
_
_
_
1/||
0
0
0
0
_
_
_
_
_
_
_
_
_
(33)
The left eigenvectors are determined by solving the matrix equation (17) which becomes:
_

l
1

l
2

l
3

l
4

l
5
_
_

_
0
1

2

3
0
0 0 0 0
1
0 0 0 0
2
0 0 0 0
3
0
1

2

3
0
_

_
= 0 (34)
The general solution (depending on 3 independent constants D
1
, D
2
and D
3
) to this set of
equations is given by:
_

l
1

l
2

l
3

l
4

l
5
_
=
_
D
1

2
D
2
+
3
D
3

1
D
2

1
D
3
D
1
_
(35)
The rst left eigenvector is chosen by setting D
1
= ||, D
2
= 0 and D
3
= 0:
u
L
1
=
_
|| 0 0 0 ||
_
(36)
The vector v
L
1
is obtained:
lim

1
0,
2
0
(k

1
) =
1
U
8
v
L
1
=
1
U
u
L
1

A =
_
|| 0 0 0 ||
_
(37)
This choice of the eigenvector corresponds to the entropy wave traveling downstream at a speed
U. This can be veried by noting that the only non-zero term in the right eigenvector is the density,
so that the wave has varying entropy, no vorticity and constant pressure. The left eigenvector
measures entropy in the sense that u
L
1
U is equal to the linearized entropy,
x
(p

). The right
eigenvector u
R
1
is chosen such that the left eigenvector is normal to it.
3.2.2 Root 2: rst vorticity wave
k
2
=
V l Wm
U
, = Uk
2
+ V l + Wm (38)
The second right eigenvector corresponding to the triple-root is chosen by setting C
1
= C
3
= 0
and C
2
=
U

in Eq. (32). Taking into account that

2
=
y
k
2
+
y
l +
y
m and
1
=
x
k
2
+
x
l +
x
m,
and replacing k
2
using Eq. (38), the right eigenvector is obtained as
u
R
2
= C
R
2
_
_
_
_
_
_
_
_
_
0

y
_
1 V
1
W
2
_
+ (
y

1
+
y

2
) U

x
_
1 V
1
W
2
_
+ (
x

1
+
x

2
) U

0
0
_
_
_
_
_
_
_
_
_
(39)
and still keeping a multiplicative constant, C
R
2
, for the requirement that the left eigenvector be nor-
mal to the right eigenvector. This choice is not unique. Any linear combination of the eigenvectors is
itself an eigenvector, and the only constraint is the required orthogonality condition. The motivation
of this choice is a knowledge of the distict behavior of the entropy and vorticity variables in uid
dynamics; this choice will also simplies the algebra at later stages of this analysis.
The second left eigenvector corresponding to the triple-root is chosen by setting D
1
= D
3
= 0
and D
2
=
U

in Eq. (35). Taking into account that

2
=
y
k
2
+
y
l +
y
m and
1
=
x
k
2
+
x
l +
x
m,
and replacing k
2
using Eq. (38), the left eigenvector is obtained as
u
L
2
=
_
0

l
2,2

l
2,3
0 0
_
(40)
where

l
2,2
=
_

y
_
1 V
1
W
2
_
+ (
y

1
+
y

2
) U

l
2,3
=
_

x
_
1 V
1
W
2
_
+ (
x

1
+
x

2
) U

(41)
9
Now the constant C
R
2
must be determined such that
_
u
L
2
u
R
2

1
=0,
2
=0
= 1. Only the right
eigenvector u
R
2
is modied, and its new form is:
u
R
2
=
1

2
2
_
_
_
_
_
_
_
_
_
0

y
_
1 V
1
W
2
_
+ (
y

1
+
y

2
) U

x
_
1 V
1
W
2
_
+ (
x

1
+
x

2
) U

0
0
_
_
_
_
_
_
_
_
_
(42)
where

2
=
_

2
x
+
2
y
(43)
The vector v
L
2
is obtained:
lim

1
0,
2
0
(k

2
) =
1
U
v
L
2
=
1
U
u
L
2

A =
_
0

l
2,2

l
2,3
0

l
2,5
_
(44)
where

l
2,5
=
y
(
x

1
+
x

2
)
x
(
y

1
+
y

2
)
This root correponds to a vorticity wave rotating around -axis, and traveling downstream at a
speed U.
3.2.3 Root 3: second vorticity wave
k
3
=
V l Wm
U
, = Uk
3
+ V l + Wm (45)
The third right eigenvector corresponding to the triple-root is chosen by setting C
1
= C
2
= 0 and
C
3
=
U

in Eq. (32). Taking into account that

3
=
z
k
3
+
z
l +
z
m and
1
=
x
k
3
+
x
l +
x
m,
and replacing k
3
using Eq. (45), the right eigenvector is obtained as
10
u
R
3
= C
R
3
_
_
_
_
_
_
_
_
_
0

z
_
1 V
1
W
2
_
+ (
z

1
+
z

2
) U

0
_

x
_
1 V
1
W
2
_
+ (
x

1
+
x

2
) U

0
_
_
_
_
_
_
_
_
_
(46)
and still keeping a multiplicative constant, C
R
3
, for the requirement that the left eigenvector be
normal to the right eigenvector.
The third left eigenvector corresponding to the triple-root is chosen by setting D
1
= D
2
= 0 and
D
3
=
U

in Eq. (35). Taking into account that

3
=
z
k
3
+
z
l +
z
m and
1
=
x
k
3
+
x
l +
x
m,
and replacing k
3
using Eq. (45), the left eigenvector is obtained as
u
L
3
=
_
0

l
3,2
0

l
3,4
0
_
(47)
where

l
3,2
=
_

z
_
1 V
1
W
2
_
+ (
z

1
+
z

2
) U

l
3,4
=
_

x
_
1 V
1
W
2
_
+ (
x

1
+
x

2
) U

(48)
The constant C
R
3
must be determined such that
_
u
L
3
u
R
3

1
=0,
2
=0
= 1. Only the right eigenvector
u
R
3
is modied, and its new form is:
u
R
3
=
1

2
3
_
_
_
_
_
_
_
_
_
0

z
_
1 V
1
W
2
_
+ (
z

1
+
z

2
) U

0
_

x
_
1 V
1
W
2
_
+ (
x

1
+
x

2
) U

0
_
_
_
_
_
_
_
_
_
(49)
where

3
=
_

2
x
+
2
z
(50)
The vector v
L
3
is obtained:
lim

1
0,
2
0
(k

1
) =
1
U
11
v
L
3
=
1
U
u
L
3

A =
_
0

l
3,2
0

l
3,4

l
3,5
_
(51)
where

l
3,5
=
z
(
x

1
+
x

2
)
x
(
z

1
+
z

2
) (52)
This root correponds to a vorticity wave rotating around -axis, and traveling downstream at a
speed U.
3.2.4 Root 4: rst acoustic wave
Using the notations in Eqs. (18) and (23) and also the Eq. (25):
k
4
=
_
+ U
_
(1 + S)
_
||
2
U
2
_ , = Uk
4
+ V l + Wm +
_

2
1
+
2
2
+
2
3
(53)
The matrix equation used to determined the right eigenvectors for this eigenvalue is:
_

_

1

2

3
0
0 0 0
1
0 0 0
2
0 0 0
3
0
1

2

3

_

_
_
_
_
_
_
_
_
_
_
r
1
r
2
r
3
r
4
r
5
_
_
_
_
_
_
_
_
_
= 0 (54)
where
= Uk
4
+ V l + Wm =
_

2
1
+
2
2
+
2
3
(55)
The general solution (depending on a constant C
4
) is:
_
_
_
_
_
_
_
_
_
r
1
r
2
r
3
r
4
r
5
_
_
_
_
_
_
_
_
_
=
_
_
_
_
_
_
_
_
_
C
4

1
C
4

2
C
4

2
C
4
C
4
_
_
_
_
_
_
_
_
_
(56)
12
The rst right eigenvector is obtained by setting C
4
=
1

:
u
R
4
= C
R
4
_
_
_
_
_
_
_
_
_
1 Uk

4
V
1
W
2

x
k

4
+
x

1
+
x

y
k

4
+
y

1
+
y

z
k

4
+
z

1
+
z

2
1 Uk

4
V
1
W
2
_
_
_
_
_
_
_
_
_
(57)
where C
R
4
is an arbitrary multiplicative constant and
k

4
=
_

U
_
(1 + S

)
_
||
2
U
2
_ (58)
with the new denitions in terms of
1
and
2
:

= 1 V
1
W
2

=
1
|| +
2
||

=
2
1
||
2
+
2
2
||
2
+ 2
1

2
|| (59)
S

_
1
_
||
2
U
2
_
(

2
)
_

U
_
2
The left eigenvector is obtain by solving the matrix equation:
_

l
1

l
2

l
3

l
4

l
5
_
_

_

1

2

3
0
0 0 0
1
0 0 0
2
0 0 0
3
0
1

2

3

_

_
= 0
which result in the general solution (depending on a constant D
4
):
_

l
1

l
2

l
3

l
4

l
5
_
=
_
0
1
D
4

2
D
4

3
D
4
D
4
_
The left eigenvector is obtained by setting D
4
=
1

:
13
u
L
4
= C
L
4
_
0

l

4,2

4,3

4,4

4,5
_
(60)
where C
L
4
is an arbitrary multiplicative constant and

4,2
= (
x
k

4
+
x

1
+
x

2
)

4,3
= (
y
k

4
+
y

1
+
y

2
)

4,4
= (
z
k

4
+
z

1
+
z

2
) (61)

4,5
= 1 Uk

4
V
1
W
2
The constants C
R
4
and C
L
4
must be determined such that
_
u
L
4
u
R
4

1
=0,
2
=0
= 1. The new forms
of the eigenvectors u
R
4
and u
L
4
are:
u
R
4
=
_
U +||
_
2||
2
_
_
_
_
_
_
_
_
_
1 Uk

4
V
1
W
2

x
k

4
+
x

1
+
x

y
k

4
+
y

1
+
y

z
k

4
+
z

1
+
z

2
1 Uk

4
V
1
W
2
_
_
_
_
_
_
_
_
_
(62)
and
u
L
4
=
_
U +||
_
_
0

l

4,2

4,3

4,4

4,5
_
(63)
Taking into account that:
lim

1
0,
2
0
(k

4
) =
1
U +||
(64)
the vector v
L
4
is obtained:
v
L
4
=
1
U +||
u
L
4

A =
_
0

l
4,2

l
4,3

l
4,4

l
4,5
_
(65)
where

l
4,2
=
x
+
1
_
U
x
V
x
_
+
2
_
U
x
W
x
_

l
4,3
=
y
+
1
_
U
y
V
y
_
+
2
_
U
y
W
y
_

l
4,4
=
z
+
1
_
U
z
V
z
_
+
2
_
U
z
W
z
_
(66)

l
4,5
= U + k

4
_
||
2
U
2
_
+
1
_

UV
_
+
2
_

UW
_
(67)
This wave corresponds to an isentropic, irrotational acoustic wave, traveling downstream.
14
3.2.5 Root 5: second acoustic wave
Using the notations in Eqs. (18) and (23) and also the Eq. (25):
k
5
=
_
+ U
_
(1 S)
_
||
2
U
2
_ , = Uk
5
+ V l + Wm
_

2
1
+
2
2
+
2
3
(68)
The matrix equation for this eigenvalue is:
_

_

1

2

3
0
0 0 0
1
0 0 0
2
0 0 0
3
0
1

2

3

_

_
_
_
_
_
_
_
_
_
_
r
1
r
2
r
3
r
4
r
5
_
_
_
_
_
_
_
_
_
= 0 (69)
where
= Uk
5
+ V l + Wm =
_

2
1
+
2
2
+
2
3
(70)
The general solution (depending on a constant C
5
) is:
_
_
_
_
_
_
_
_
_
r
1
r
2
r
3
r
4
r
5
_
_
_
_
_
_
_
_
_
=
_
_
_
_
_
_
_
_
_
C
5

1
C
5

2
C
5

2
C
5
C
5
_
_
_
_
_
_
_
_
_
The rst right eigenvector is obtained by setting C
5
=
1

:
u
R
5
= C
R
5
_
_
_
_
_
_
_
_
_
Uk

5
+ V
1
+ W
2
1
(
x
k

5
+
x

1
+
x

2
)
(
y
k

5
+
y

1
+
y

2
)
(
z
k

5
+
z

1
+
z

2
)
Uk

5
+ V
1
+ W
2
1
_
_
_
_
_
_
_
_
_
(71)
where C
R
5
is an arbitrary multiplicative constant and
15
k

5
=
_

U
_
(1 S

)
_
||
2
U
2
_ (72)
with the new denitions given by Eq. (59)
Following the same routine, the left eigenvector is obtain as:
u
L
5
= C
L
5
_
0

l

5,2

5,3

5,4

5,5
_
(73)
where where C
L
5
is an arbitrary constant and

5,2
= (
x
k

5
+
x

1
+
x

2
)

5,3
= (
y
k

5
+
y

1
+
y

2
)

5,4
= (
z
k

5
+
z

1
+
z

2
) (74)

5,5
= Uk

5
+ V
1
+ W
2
1
The constants C
R
5
and C
L
5
must be determined such that
_
u
L
5
u
R
5

1
=0,
2
=0
= 1. The new forms
of the eigenvectors u
R
5
and u
L
5
are:
u
R
5
=
_
U ||
_
2||
2
_
_
_
_
_
_
_
_
_
Uk

5
+ V
1
+ W
2
1
(
x
k

5
+
x

1
+
x

2
)
(
y
k

5
+
y

1
+
y

2
)
(
z
k

5
+
z

1
+
z

2
)
Uk

5
+ V
1
+ W
2
1
_
_
_
_
_
_
_
_
_
(75)
and
u
L
5
=
_
U ||
_
_
0

l

5,2

5,3

5,4

5,5
_
(76)
Taking into account that:
lim

1
0,
2
0
(k

5
) =
1
U ||
(77)
the vector v
L
5
is obtained:
16
v
L
5
=
1
U ||
u
L
5

A =
_
0

l
5,2

l
5,3

l
5,4

l
5,5
_
(78)
where

l
5,2
=
x

1
_
U
x
V
x
_

2
_
U
x
W
x
_

l
5,3
=
y

1
_
U
y
V
y
_

2
_
U
y
W
y
_

l
5,4
=
z

1
_
U
z
V
z
_

2
_
U
z
W
z
_
(79)

l
5,5
= U k

5
_
||
2
U
2
_

1
_

UV
_

2
_

UW
_
This wave corresponds to an isentropic, irrotational acoustic wave, traveling upstream.
4 First-Order Unsteady Boundary Conditions for 3D Curvilinear Euler
Equations
The one-dimensional, non-reecting boundary conditions are obtained by ignoring all variations in
the y- and z- directions and setting
1
= 0 and
2
= 0. In these conditions:

= 0,

= 0,

= 1, S

=
||
U
(80)
w
R
n
= u
R
n
|

1
=0,
2
=0
(81)
w
L
n
= u
L
n
|

1
=0,
2
=0
= v
L
n
|

1
=0,
2
=0
(82)
The right eigenvectors w
R
n
|
n=1,2,3,4,5
are:
w
R
1
=
_
_
_
_
_
_
_
_
_
1/||
0
0
0
0
_
_
_
_
_
_
_
_
_
, w
R
2
=
1

2
2
_
_
_
_
_
_
_
_
_
0

x
0
0
_
_
_
_
_
_
_
_
_
, w
R
3
=
1

2
3
_
_
_
_
_
_
_
_
_
0

z
0

x
0
_
_
_
_
_
_
_
_
_
17
w
R
4
=
1
2||
_
_
_
_
_
_
_
_
_
_
1

x
1
||

y
1
||

z
1
||
1
_
_
_
_
_
_
_
_
_
_
, w
R
5
=
1
2||
_
_
_
_
_
_
_
_
_
_
1

x
1
||

y
1
||

z
1
||
1
_
_
_
_
_
_
_
_
_
_
(83)
The left eigenvectors w
L
n
|
n=1,2,3,4,5
are:
w
L
1
=
_
|| 0 0 0 ||
_
w
L
2
=
_
0
y

x
0 0
_
w
L
3
=
_
0
z
0
x
0
_
(84)
w
L
4
=
_
0
x

y

z
||
_
w
L
5
=
_
0
x

y

z
||
_
The transformation to and from 1-D characteristics variables is given by the next two matrix
equations:
_
_
_
_
_
_
_
_
_
c
1
c
2
c
3
c
4
c
5
_
_
_
_
_
_
_
_
_
=
_

_
|| 0 0 0 ||
0
y

x
0 0
0
z
0
x
0
0
x

y

z
||
0
x

y

z
||
_

_
_
_
_
_
_
_
_
_
_

_
_
_
_
_
_
_
_
_
(85)
and
_
_
_
_
_
_
_
_
_

_
_
_
_
_
_
_
_
_
=
_

_
1/|| 0 0
1
2||
1
2||
0
1

2
2

y

1

2
3

z
1
2||
2

x

1
2||
2

x
0
1

2
2

x
0
1
2||
2

y

1
2||
2

y
0 0
1

2
3

x
1
2||
2

z

1
2||
2

z
0 0 0
1
2||
1
2||
_

_
_
_
_
_
_
_
_
_
_
c
1
c
2
c
3
c
4
c
5
_
_
_
_
_
_
_
_
_
(86)
where c
1
c
2
c
3
c
4
and c
5
are the amplitudes of the ve characteristics waves. Lets check the consis-
tency with the Cartesian coordinates. If

x
= 1,
y
= 0,
z
= 0

x
= 0,
y
= 1,
z
= 0 (87)

x
= 0,
y
= 0,
z
= 1
18
then the Eqs. (85) and (86) become:
_
_
_
_
_
_
_
_
_
c
1
c
2
c
3
c
4
c
5
_
_
_
_
_
_
_
_
_
=
_

_
1 0 0 0 1
0 0 1 0 0
0 0 0 1 0
0 1 0 0 1
0 1 0 0 1
_

_
_
_
_
_
_
_
_
_
_

_
_
_
_
_
_
_
_
_
(88)
and
_
_
_
_
_
_
_
_
_

_
_
_
_
_
_
_
_
_
=
_

_
1 0 0
1
2
1
2
0 0 0
1
2

1
2
0 1 0 0 0
0 0 1 0 0
0 0 0
1
2
1
2
_

_
_
_
_
_
_
_
_
_
_
c
1
c
2
c
3
c
4
c
5
_
_
_
_
_
_
_
_
_
(89)
being the same with the results from the thesis of Shivaji [3].
At the inow boundary, the correct unsteady, non-reecting boundary conditions for a subsonic
ow are:
_
_
_
_
_
_
_
c
1
c
2
c
3
c
4
_
_
_
_
_
_
_
= 0 (90)
while at the outow boundary:
c
5
= 0 (91)
5 Approximate, Quasi-3D, Unsteady Boundary Conditions for Curvilinear
Euler Equations
The left eigenvectors v
L
n
are all functions of
1
and
2
which are assumed to be small (quasi-3D). It
is prefered to have these functions in polynomial form; the straightforward way to transform them
into polynomial functions is by using Taylor series expansions. The expansions are written as:
19
v
L
n
(
1
,
2
) = v
L
n
(0, 0) +
1
_
(v
L
n
)

1
_

1
=
2
=0
+
2
_
(v
L
n
)

2
_

1
=
2
=0
+ HOT, (92)
n = 1, 2, 3, 4, 5
where HOT stands for higher-order-terms. Equations (92) are exact.
The rst order approximation which was discussed in the previous section is obtained by keeping
only the rst term in the Taylor series expansion. The second order approximation is obtained by
taking into account the rst three terms and neglecting the HOT. Higher order approximation is
possible by considering more terms in the series, but the algebra becomes more complex.
The second order approximation is:
v
L
n
(
1
,
2
) v
L
n
(
1
,
2
) = v
L
n
(0, 0) +
1
_
(v
L
n
)

1
_

1
=
2
=0
+
2
_
(v
L
n
)

2
_

1
=
2
=0
= u
L
n
(0, 0) +
l

_
k
n

(u
L
n
)

A
_

1
=
2
=0
+
m

_
k
n

(u
L
n
)

A
_

1
=
2
=0
(93)
This produces the boundary conditions:
_
u
L
n
|

1
=
2
=0
+ l
_
k
n

(u
L
n
)

A
_

1
=
2
=0
+ m
_
k
n

(u
L
n
)

A
_

1
=
2
=0
_
Q = 0 (94)
Replacing , l and m by i

t
, i

and i

, respectively, Eq. (94) becomes:

_
u
L
n
Q
t

_
k
n

(u
L
n
)

A
_
Q

_
k
n

(u
L
n
)

A
_
Q

1
=
2
=0
= 0 (95)
The next step is to take the derivatives (u
L
n
)

1
and (u
L
n
)

2
, and evaluate u
L
n
,
_
kn

(u
L
n
)

A
_
and
_
kn

(u
L
n
)

A
_
in the hypothesis that
1
= 0 and
2
= 0. This will give ve equations representing
the approximate, quasi-3D boundary conditions: the rst four equations are solved at the inow,
and the fth at the outow.
At the inow, the boundary conditions are:
_

_
|| 0 0 0 ||
0
y

x
0 0
0
z
0
x
0
0
x

y

z
||
_

_
Q
t
+
_

_
g
11
g
12
g
13
g
14
g
15
g
21
g
22
g
23
g
24
g
25
g
31
g
32
g
33
g
34
g
35
g
41
g
42
g
43
g
44
g
45
_

_
Q

+
_

_
h
11
h
12
h
13
h
14
h
15
h
21
h
22
h
23
h
24
h
25
h
31
h
32
h
33
h
34
h
35
h
41
h
42
h
43
h
44
h
45
_

_
Q

= 0 (96)
20
and at the outow, they are:
_
0
x

y

z
||

Q
t
+
_
g
51
g
52
g
53
g
54
g
55

+
_
h
51
h
52
h
53
h
54
h
55

= 0 (97)
where
g
11
= 0
g
12
= 0
g
13
= 0
g
14
= 0
g
15
= 0
g
21
= 0
g
22
= U
y
V
y
g
23
= U
x
+ V
x
g
24
= 0
g
25
=
x

x
g
31
= 0
g
32
= U
z
V
z
(98)
g
33
= 0
g
34
= U
x
+ V
x
g
35
=
x

x
g
41
= 0
g
42
= U
x
+ V
x
g
43
= U
y
+ V
y
g
44
= U
z
+ V
z
g
45
= U
||
||
+ V ||
g
51
= 0
g
52
= U
x
V
x
g
53
= U
y
V
y
g
54
= U
z
V
z
g
55
= U
||
||
+ V ||
and
21
h
11
= 0
h
12
= 0
h
13
= 0
h
14
= 0
h
15
= 0
h
21
= 0
h
22
= U
y
W
y
h
23
= U
x
+ W
x
h
24
= 0
h
25
=
x

x
h
31
= 0
h
32
= U
z
W
z
(99)
h
33
= 0
h
34
= U
x
+ W
x
h
35
=
x

x
h
41
= 0
h
42
= U
x
+ W
x
h
43
= U
y
+ W
y
h
44
= U
z
+ W
z
h
45
= U
||
||
+ W||
h
51
= 0
h
52
= U
x
W
x
h
53
= U
y
W
y
h
54
= U
z
W
z
h
55
= U
||
||
+ W||
It can be veried that the above equations reduce to the Cartesian 3D Giles boundary conditions.
Thus, according to Eq. (87), the Eqs. (96) and (97) become:
_

_
1 0 0 0 1
0 0 1 0 0
0 0 0 1 0
0 1 0 0 1
_

_
Q
t
+
_

_
0 0 0 0 0
0 u v 0 0
0 0 0 v 0
0 v u 0 v
_

_
Q
y
22
+
_

_
0 0 0 0 0
0 0 w 0 0
0 u 0 w 1
0 w 0 u w
_

_
Q
z
= 0 (100)
and
_
0 1 0 0 1

Q
t
+
_
0 v u 0 v

Q
y
+
_
0 w 0 u w

Q
z
= 0 (101)
being consistent with the results from the thesis of Shivaji [[3]].
6 Analysis of Well-Posedness
The one-dimensional approximation of the boundary conditions is always well-posed (demonstrated
by using the energy method). The second order approximation need to be analyzed and corrected
in case of ill-posedness. This follows next.
6.1 Inow Boundary Conditions from the Second-Order Approximation
The aim is to verify that there is no incoming mode which exactly satises the boundary conditions.
The analysis is done in a frame of reference that is moving with speed V in -direction and W
in -direction. The well-posedness in this frame of reference is both necessary and sucient for
well-posedness in the original frame of reference. This simplies the algebra by setting V = W = 0.
If there are N

incoming waves, then the generalized incoming mode may be written as

U (, , , t) =
_
4

n=1
a
n
u
R
n
e
ikn
_
e
i(l+mt)
(102)
with Im() 0. The wavenumbers (divided by frequency) are given by:
k

1
= k

2
= k

3
=
1
U
(103)
k

4
=
_

+ U
_
(1 + S

)
_
||
2
U
2
_ (104)
where
23
S

_
1
_
||
2
U
2
_
(

1)
_

+ U
_
2
(105)
and the denitions given by Eq. (59). Following the theory given in Giles work [], the critical
matrix

C
[44]
corresponding to the inow boundary has the elements:
c
nj
= v
L
n
u
R
j
(106)
The vectors
v
L
n
= u
L
n
|

1
=
2
=0
+
1
_
k

n
(u
L
n
)

A
_

1
=
2
=0
+
2
_
k

n
(u
L
n
)

A
_

1
=
2
=0
, (107)
n = 1, 2, 3, 4
are needed rst with V = W = 0. They are:
v
L
1
=
_

l
1,1

l
1,2

l
1,3

l
1,4

l
1,5
_
(108)
where

l
1,1
=
x

l
1,2
= 0

l
1,3
= 0 (109)

l
1,4
= 0

l
1,5
=
x
v
L
2
=
_

l
2,1

l
2,2

l
2,3

l
2,4

l
2,5
_
(110)
where

l
2,1
= 0

l
2,2
=
y

1
U
y

2
U
y

l
2,3
=
x
+
1
U
x
+
2
U
x
(111)

l
2,4
= 0

l
2,5
=
1
(
x

x
)
2
(
x

x
)
24
v
L
3
=
_

l
3,1

l
3,2

l
3,3

l
3,4

l
3,5
_
(112)
where

l
3,1
= 0

l
3,2
=
z

1
U
z

2
U
z

l
3,3
= 0 (113)

l
3,4
=
x
+
1
U
x
+
2
U
x

l
3,5
=
1
(
x

x
)
2
(
x

x
)
v
L
4
=
_

l
4,1

l
4,2

l
4,3

l
4,4

l
4,5
_
(114)
where

l
4,1
= 0

l
4,2
=
x
+
1
U
x
+
2
U
x

l
4,3
=
y
+
1
U
y
+
2
U
y
(115)

l
4,4
=
z
+
1
U
z
+
2
U
z

l
4,5
= || +
1
U
||
||
+
2
U
||
||
Thus, the elements of the critical matrix

C are:
c
11
= 1
c
12
= 0
c
13
= 0
c
14
= 0 (116)
c
21
= 0
c
22
=
1

2
2
_
_

y
+
1
U
y
+
2
U
y
_
2
+
_

x
+
1
U
x
+
2
U
x
_
2
_
c
23
=
1

2
3
__

y
+
1
U
y
+
2
U
y
_ _

z
+
1
U
z
+
2
U
z
_
c
24
= 0
c
31
= 0
25
c
32
=
1

2
2
__

y
+
1
U
y
+
2
U
y
_ _

z
+
1
U
z
+
2
U
z
_
c
33
=
1

2
3
_
_

z
+
1
U
z
+
2
U
z
_
2
+
_

x
+
1
U
x
+
2
U
x
_
2
_
c
34
= 0
c
41
= 0
c
42
= 0
c
43
= 0
c
44
=
_
U +||
_
2||
2
_

_
k

+
2
1
U
2

+
2
2
U
2

+
1
|| +
2
|| + 2
1

2
U||
+k

4
U
_

1
|| +
2
||

1
U
||

2
U
||

_
+

+
1
U
||

+
2
U
||

_
The requirement that there is no non-trivial incoming mode satisfying the boundary conditions
is equivalent to the statement that the determinant of the above matrix is non-zero for all real l and
m and complex with Im() 0.
In other words, the procedure would be this:
-assume that the determinant det(

C) = 0;
-if after manipulations we get FALSE, then the problem is well-posed
-if we get a combination of , l and m that make det(

C) = 0, then the problem is ill-posed
Assuming that everything else (grid metrics, mean ow) is constant, the determinant det(

C) is
a function of
1
and
2
. So, det(

C) = 0 will produce a relation between
1
and
2
(unless the result
is FALSE).
To simplify the analysis a little bit, lets assume that the grid is orthogonal at the boundary.
Under this hypothesis:
|| = || = || = 0 (117)
c
44
=
_
U +||
_
2||
2
_
k

U
_
+ U +

(118)
k

4
=
U

U
2
(1 + S

) (119)
S

=
1
U
_
_
U
2

_
+
2

(120)
where
26
=
2
1

+
2
2

(121)
One of the result of det(

C) = 0 is c
44
= 0. Replacing k

4
will give the equation:
S

=
_
U +

U
(122)
Replacing S

, and solving for , gives:

=
2
1

+
2
2

U
2
(123)
The conclusion is:
if = +iU

l
2

+ m
2

(124)
satisfying the condition that Im() 0, then:

2
1

+
2
2

U
2
and S

U
2
(125)
with the correct branch of the square root being taken to ensure that Im(k
4
) 0.
To evaluate the critical matrix under the conditions given by Eqs. (117) and (125), denote rst:

1
=
x
+
1
U
x
+
2
U
x

2
=
y
+
1
U
y
+
2
U
y
(126)

3
=
z
+
1
U
z
+
2
U
z
The critical matrix is:

C =
_

_
1 0 0 0
0
1

2
2
(
2
2
+
2
1
)
1

2
3
(
2

3
) 0
0
1

2
2
(
2

3
)
1

2
3
(
2
3
+
2
1
) 0
0 0 0 0
_

_
(127)
Next, the second row is multiplied by
2
, and the third row by
3
, then add the second row to
the third row. The result is:
27

C =
_

_
1 0 0 0
0
1

2
2
(
2
2
+
2
1
)
1

2
3
(
2

3
) 0
0
1

2
2

2
(
2
1
+
2
2
+
2
3
)
1

2
3

3
(
2
1
+
2
2
+
2
3
) 0
0 0 0 0
_

_
(128)
It is ease to show that
2
1
+
2
2
+
2
3
= 0 under the condition that the grid is orthogonal at the
boundary (Eq. (117)), and taking into account the Eq. (125). Thus, the critical matrix is:

C =
_

_
1 0 0 0
0 c
22
c
23
0
0 0 0 0
0 0 0 0
_

_
(129)
with c
22
and c
23
given by Eq. (116).
So, there is clearly a non-trivial incoming mode coresponding to the third right eigenvector,
and the inow boundary conditions are ill-posed. Looking at the critical matrix, there is a second
incoming mode (corresponding to the fourth eigenvector), but this is actually a multiple of the rst,
because when = +iU

_
l
2

+m
2

, then k
3
= k
4
=
1
U
. Hence, the initial-boundary-value problem
is ill-posed with two ill-posed modes.
6.2 Ouow Boundary Conditions from the Second-Order Approximation
At the outow, the generalized incoming mode may be written as
U (, , , t) = a
5
u
R
5
e
ik
5

e
i(l+mt)
(130)
with Im() 0. The wavenumber is given by:
k

5
=
_
+ U
_
(1 S

)
_
||
2
U
2
_ (131)
where S

is given by Eq. (105). The correct square root must be taken in the denition of S

to ensure that if and S

are both real then S

is positive, and if and S

are complex then

Im(k
5
) < 0. The critical matrix is actually a scalar, and has the form:

C =
_
U ||
_
2||
2
_

_
k

+
2
1
U
2

+
2
2
U
2

+
1
|| +
2
|| + 2
1

2
U||
+k

5
U
_

1
|| +
2
||

1
U
||

2
U
||

_
+

+
1
U
||

+
2
U
||

_
28
Under the condition that the grid is orthogonal at the boundary (|| = || = || = 0),

C = 0
leads to the same results as before:
if = +iU

l
2

+ m
2

(132)
satisfying the condition that Im() 0, then:
=
2
1

+
2
2

U
2
and S

U
2
(133)
but is this case:
S

=
_
U +

_
+

U
(134)
If we substitute Eqs. (133) into (134), then the result is:

U
2
=

U
2
(135)
which is FALSE. This contradicts the assumption that there is an incoming mode satisfying the
boundary conditions. The conclusion is that the outow boundary condition is well-posed.
7 Modied Boundary Conditions
The inow boundary conditions must be modied to assure the well-posedness. Since the rst three
inow boundary conditions require that a
1
= a
2
= a
3
= 0, the only nedeed condition of orthogonality
is that between v
L
4
and u
R
5
. A new denition is proposed for v
L
4
:
v
L
4
= v
L
4,old
+
1
m
1
_
0
y

x
0 0
_
+
2
m
2
_
0
z
0
x
0
_
(136)
The variables m
1
and m
2
are chosen to minimize v
L
4
u
R
5
, which controls the magnitude of the
reection coecient, and at the same time will produce a well posed boundary condition. The
motivation of this approach is that the second approximation to the scalar wave equation is well-
posed and produces fourth-order reection.
The new form of v
L
4
is:
v
L
4
=
_

l
4,1

l
4,2

l
4,3

l
4,4

l
4,5
_
(137)
29
where

l
4,1
= 0

l
4,2
=
x
+
1
U
x
+
2
U
x

1
m
1

2
m
2

l
4,3
=
y
+
1
U
y
+
2
U
y
+
1
m
1

x
(138)

l
4,4
=
z
+
1
U
z
+
2
U
z
+
2
m
2

l
4,5
= || +
1
U
||
||
+
2
U
||
||
Using Eq. (120), the binomial expansion of S

is:
S

=
||
U
_
1 +
1
2
_
U
2
||
2
||
2
_
+ O
_

2
_
_

||
U
_
1 +
1
2
_
U
2
||
2
||
2
_

_
(139)
Thus, the product v
L
4
u
R
5
will be:
v
L
4
u
R
5
=
_
U ||
_
2||
2
_
A
1

2
1
+ A
2

2
+ A
3

2
2
_
(140)
where
A
1
=
1
2
_
U +||
_
||
2
+ m
1
(
y

y
) (141)
A
2
= m
1
(
y

y
) + m
2
(
z

z
) (142)
A
3
=
1
2
_
U +||
_
||
2
+ m
2
(
z

z
) (143)
The reection coecient is fourth order if m
1
and m
2
are chosen such that v
L
4
u
R
5
0. v
L
4
u
R
5
is
a homogeneous polynomial function of
1
and
2
; it is zero when its coecients are all zero. The
problem is that there are 3 coecients (3 equations) for 2 unknowns, m
1
and m
2
.
Lets check the consistency with the Cartesian coordinates. Using Eq. (87):
A
1
=
1
2
(u + 1) m
1
A
2
= 0
A
3
=
1
2
(u + 1) m
2
30
v
L
4
u
R
5
0 if A
1
= A
2
= 0 which means:
m
1
= m
2
=
1
2
(u + 1)
which is clearly consistent with the Cartesian results. So, the problem is brought by the term A
2
being the coecient of the product between
1
and
2
.
******** need to determine m
1
and m
2
for general case ********
Under the assumption that A
2
0 (?), the values of m
1
and m
2
can be determined:
m
1
=
1
2
_
U +||
_
||
2

y
m
2
=
1
2
_
U +||
_
||
2

z
(144)
The new fourth-order inow boundary conditions will be:
_

_
|| 0 0 0 ||
0
y

x
0 0
0
z
0
x
0
0
x

y

z
||
_

_
Q
t
+
_

_
g
11
g
12
g
13
g
14
g
15
g
21
g
22
g
23
g
24
g
25
g
31
g
32
g
33
g
34
g
35
g
41
g
42
g
43
g
44
g
45
_

_
Q

+
_

_
h
11
h
12
h
13
h
14
h
15
h
21
h
22
h
23
h
24
h
25
h
31
h
32
h
33
h
34
h
35
h
41
h
42
h
43
h
44
h
45
_

_
Q

= 0 (145)
where
g
11
= 0
g
12
= 0
g
13
= 0
g
14
= 0
g
15
= 0
g
21
= 0
g
22
= U
y
V
y
31
g
23
= U
x
+ V
x
g
24
= 0
g
25
=
x

x
g
31
= 0
g
32
= U
z
V
z
(146)
g
33
= 0
g
34
= U
x
+ V
x
g
35
=
x

x
g
41
= 0
g
42
= U
x
+ V
x
+ m
1

y
g
43
= U
y
+ V
y
m
1

x
g
44
= U
z
+ V
z
g
45
= U
||
||
+ V ||
and
h
11
= 0
h
12
= 0
h
13
= 0
h
14
= 0
h
15
= 0
h
21
= 0
h
22
= U
y
W
y
h
23
= U
x
+ W
x
h
24
= 0
h
25
=
x

x
h
31
= 0
h
32
= U
z
W
z
(147)
h
33
= 0
h
34
= U
x
+ W
x
h
35
=
x

x
h
41
= 0
h
42
= U
x
+ W
x
+ m
2

z
h
43
= U
y
+ W
y
h
44
= U
z
+ W
z
m
2

x
32
h
45
= U
||
||
+ W||
8 Summary of Dimensional Curvilinear 3D BC
The inow and outow boundaries are aligned along direction.
Usefull notations:
|| =
_
(
x
)
2
+ (
y
)
2
+ (
z
)
2
|| =
_
(
x
)
2
+ (
y
)
2
+ (
z
)
2
|| =
_
(
x
)
2
+ (
y
)
2
+ (
z
)
2
|| =
x

x
+
y

y
+
z

z
(148)
|| =
x

x
+
y

y
+
z

z
|| =
x

x
+
y

y
+
z

2
=
_
(
x
)
2
+ (
y
)
2

3
=
_
(
x
)
2
+ (
z
)
2
8.1 First-Order Unsteady Boundary Conditions
The transformation to and from 1-D characteristics variables is given by the next two matrix equa-
tions:
_
_
_
_
_
_
_
_
_
c
1
c
2
c
3
c
4
c
5
_
_
_
_
_
_
_
_
_
=
_

_
c
2
|| 0 0 0 ||
0 c
y
c
x
0 0
0 c
z
0 c
x
0
0 c
x
c
y
c
z
||
0 c
x
c
y
c
z
||
_

_
_
_
_
_
_
_
_
_
_

_
_
_
_
_
_
_
_
_
(149)
and
_
_
_
_
_
_
_
_
_

_
_
_
_
_
_
_
_
_
=
_

1
c
2
||
0 0
1
2 c
2
||
1
2 c
2
||
0
y
c
2
2

z
c
2
3
x
2 c||
2

x
2||
2
0
x
c
2
2
0
y
2 c||
2

y
2 c||
2
0 0
x
c
2
3
z
2 c||
2

z
2 c||
2
0 0 0
1
2||
1
2||
_

_
_
_
_
_
_
_
_
_
_
c
1
c
2
c
3
c
4
c
5
_
_
_
_
_
_
_
_
_
(150)
33
where c
1
c
2
c
3
c
4
and c
5
are the amplitudes of the ve characteristics waves.
At the inow boundary, the correct unsteady, non-reecting boundary conditions for a subsonic
ow are:
_
_
_
_
_
_
_
c
1
c
2
c
3
c
4
_
_
_
_
_
_
_
= 0 (151)
while at the outow boundary:
c
5
= 0 (152)
8.2 Approximate, Quasi-3D, Unsteady Boundary Conditions
At the inow, the boundary conditions in terms of ow perturbations are:
_

_
c
2
|| 0 0 0 ||
0 c
y
c
x
0 0
0 c
z
0 c
x
0
0 c
x
c
y
c
z
||
_

_
Q
t
+
_

_
g
11
g
12
g
13
g
14
g
15
g
21
g
22
g
23
g
24
g
25
g
31
g
32
g
33
g
34
g
35
g
41
g
42
g
43
g
44
g
45
_

_
Q

+
_

_
h
11
h
12
h
13
h
14
h
15
h
21
h
22
h
23
h
24
h
25
h
31
h
32
h
33
h
34
h
35
h
41
h
42
h
43
h
44
h
45
_

_
Q

= 0 (153)
where
Q =
_

_
T
(154)
and at the outow, they are:
_
0 c
x
c
y
c
z
||

Q
t
+
_
g
51
g
52
g
53
g
54
g
55

+
_
h
51
h
52
h
53
h
54
h
55

= 0 (155)
where
34
g
11
= 0
g
12
= 0
g
13
= 0
g
14
= 0
g
15
= 0
g
21
= 0
g
22
= U
y
V
y
g
23
= U
x
+ V
x
g
24
= 0
g
25
=
x

x
g
31
= 0
g
32
= U
z
V
z
(156)
g
33
= 0
g
34
= U
x
+ V
x
g
35
=
x

x
g
41
= 0
g
42
= U
x
+ V
x
g
43
= U
y
+ V
y
g
44
= U
z
+ V
z
g
45
= U
||
||
+ V ||
g
51
= 0
g
52
= U
x
V
x
g
53
= U
y
V
y
g
54
= U
z
V
z
g
55
= U
||
||
+ V ||
and
h
11
= 0
h
12
= 0
h
13
= 0
h
14
= 0
h
15
= 0
35
h
21
= 0
h
22
= U
y
W
y
h
23
= U
x
+ W
x
h
24
= 0
h
25
=
x

x
h
31
= 0
h
32
= U
z
W
z
(157)
h
33
= 0
h
34
= U
x
+ W
x
h
35
=
x

x
h
41
= 0
h
42
= U
x
+ W
x
h
43
= U
y
+ W
y
h
44
= U
z
+ W
z
h
45
= U
||
||
+ W||
h
51
= 0
h
52
= U
x
W
x
h
53
= U
y
W
y
h
54
= U
z
W
z
h
55
= U
||
||
+ W||
At the inow, the boundary conditions in terms of characteristic variables are:

t
_
_
_
_
_
_
c
1
c
2
c
3
c
4
_
_
_
_
_
_
+
_

_
g
11
g
12
g
13
g
14
g
15
g
21
g
22
g
23
g
24
g
25
g
31
g
32
g
33
g
34
g
35
g
41
g
42
g
43
g
44
g
45
_

_
_
_
_
_
_
_
_
_
c
1
c
2
c
3
c
4
c
5
_
_
_
_
_
_
_
_
_
+
_

_
h
11
h
12
h
13
h
14
h
15
h
21
h
22
h
23
h
24
h
25
h
31
h
32
h
33
h
34
h
35
h
41
h
42
h
43
h
44
h
45
_

_
_
_
_
_
_
_
_
_
c
1
c
2
c
3
c
4
c
5
_
_
_
_
_
_
_
_
_
= 0 (158)
and at the outow, they are:
36
(c
5
)
t
+
_
g
51
g
52
g
53
g
54
g
55

_
_
_
_
_
_
_
_
_
c
1
c
2
c
3
c
4
c
5
_
_
_
_
_
_
_
_
_
+
_
h
51
h
52
h
53
h
54
h
55

_
_
_
_
_
_
_
_
_
c
1
c
2
c
3
c
4
c
5
_
_
_
_
_
_
_
_
_
= 0 (159)
where
g
11
= 0
g
12
= 0
g
13
= 0
g
14
= 0
g
15
= 0
g
21
= 0
g
22
=
U

2
2
(
y

y
+
x

x
) + V
g
23
=
1

2
3

z
_
U
y
V
y
_
g
24
=
_
c|| + U
2||
2
_
(
x

x
)
g
25
=
_
c|| U
2||
2
_
(
x

x
)
g
31
= 0
g
32
=
1

2
2

y
_
U
z
V
z
_
(160)
g
33
=
U

2
3
(
z

z
+
x

x
) + V
g
34
=
_
c|| + U
2||
2
_
(
x

x
)
37
g
35
=
_
c|| U
2||
2
_
(
x

x
)
g
41
= 0
g
42
=
U

2
2
(
y

y
)
g
43
=
U

2
3
(
z

z
)
g
44
= U
||
||
2
+ V
g
45
= 0
g
51
= 0
g
52
= g
42
=
U

2
2
(
y

y
)
g
53
= g
43
=
U

2
3
(
z

z
)
g
54
= 0
g
55
= g
44
= U
||
||
2
+ V
and
h
11
= 0
h
12
= 0
h
13
= 0
h
14
= 0
h
15
= 0
h
21
= 0
h
22
=
U

2
2
(
y

y
+
x

x
) + W
h
23
=
1

2
3

z
_
U
y
W
y
_
h
24
=
_
c|| + U
2||
2
_
(
x

x
)
h
25
=
_
c|| U
2||
2
_
(
x

x
)
h
31
= 0
h
32
=
1

2
2

y
_
U
z
W
z
_
(161)
38
h
33
=
U

2
3
(
z

z
+
x

x
) + W
h
34
=
_
c|| + U
2||
2
_
(
x

x
)
h
35
=
_
c|| U
2||
2
_
(
x

x
)
h
41
= 0
h
42
=
U

2
2
(
y

y
)
h
43
=
U

2
3
(
z

z
)
h
44
= U
||
||
2
+ W
h
45
= 0
h
51
= 0
h
52
= h
42
=
U

2
2
(
y

y
)
h
53
= h
43
=
U

2
3
(
z

z
)
h
54
= 0
h
55
= h
44
= U
||
||
2
+ W
8.3 Modied Boundary Conditions
Under the assumption that A
2
0 (?; this assumption need to be revised), the values of m
1
and m
2
could be determined:
m
1
=
1
2
_
U +||
_
||
2

y
m
2
=
1
2
_
U +||
_
||
2

z
(162)
The new fourth-order (?) inow boundary conditions in terms of ow perturbations will be:
_

_
c
2
|| 0 0 0 ||
0 c
y
c
x
0 0
0 c
z
0 c
x
0
0 c
x
c
y
c
z
||
_

_
Q
t
+
_

_
g
11
g
12
g
13
g
14
g
15
g
21
g
22
g
23
g
24
g
25
g
31
g
32
g
33
g
34
g
35
g
41
g
42
g
43
g
44
g
45
_

_
Q

39
+
_

_
h
11
h
12
h
13
h
14
h
15
h
21
h
22
h
23
h
24
h
25
h
31
h
32
h
33
h
34
h
35
h
41
h
42
h
43
h
44
h
45
_

_
Q

= 0 (163)
where
g
11
= 0
g
12
= 0
g
13
= 0
g
14
= 0
g
15
= 0
g
21
= 0
g
22
= U
y
V
y
g
23
= U
x
+ V
x
g
24
= 0
g
25
=
x

x
g
31
= 0
g
32
= U
z
V
z
(164)
g
33
= 0
g
34
= U
x
+ V
x
g
35
=
x

x
g
41
= 0
g
42
= U
x
+ V
x
+ m
1

y
g
43
= U
y
+ V
y
m
1

x
g
44
= U
z
+ V
z
g
45
= U
||
||
+ V ||
and
h
11
= 0
h
12
= 0
h
13
= 0
h
14
= 0
h
15
= 0
40
h
21
= 0
h
22
= U
y
W
y
h
23
= U
x
+ W
x
h
24
= 0
h
25
=
x

x
h
31
= 0
h
32
= U
z
W
z
(165)
h
33
= 0
h
34
= U
x
+ W
x
h
35
=
x

x
h
41
= 0
h
42
= U
x
+ W
x
+ m
2

z
h
43
= U
y
+ W
y
h
44
= U
z
+ W
z
m
2

x
h
45
= U
||
||
+ W||
At the inow, the new boundary conditions in terms of characteristic variables are:

t
_
_
_
_
_
_
c
1
c
2
c
3
c
4
_
_
_
_
_
_
+
_

_
g
11
g
12
g
13
g
14
g
15
g
21
g
22
g
23
g
24
g
25
g
31
g
32
g
33
g
34
g
35
g
41
g
42
g
43
g
44
g
45
_

_
_
_
_
_
_
_
_
_
c
1
c
2
c
3
c
4
c
5
_
_
_
_
_
_
_
_
_
+
_

_
h
11
h
12
h
13
h
14
h
15
h
21
h
22
h
23
h
24
h
25
h
31
h
32
h
33
h
34
h
35
h
41
h
42
h
43
h
44
h
45
_

_
_
_
_
_
_
_
_
_
c
1
c
2
c
3
c
4
c
5
_
_
_
_
_
_
_
_
_
= 0 (166)
and at the outow, they are:
41
(c
5
)
t
+
_
g
51
g
52
g
53
g
54
g
55

_
_
_
_
_
_
_
_
_
c
1
c
2
c
3
c
4
c
5
_
_
_
_
_
_
_
_
_
+
_
h
51
h
52
h
53
h
54
h
55

_
_
_
_
_
_
_
_
_
c
1
c
2
c
3
c
4
c
5
_
_
_
_
_
_
_
_
_
= 0 (167)
where
g
11
= 0
g
12
= 0
g
13
= 0
g
14
= 0
g
15
= 0
g
21
= 0
g
22
=
U

2
2
(
y

y
+
x

x
) + V
g
23
=
1

2
3

z
_
U
y
V
y
_
g
24
=
_
c|| + U
2||
2
_
(
x

x
)
g
25
=
_
c|| U
2||
2
_
(
x

x
)
g
31
= 0
g
32
=
1

2
2

y
_
U
z
V
z
_
(168)
g
33
=
U

2
3
(
z

z
+
x

x
) + V
g
34
=
_
c|| + U
2||
2
_
(
x

x
)
42
g
35
=
_
c|| U
2||
2
_
(
x

x
)
g
41
= 0
g
42
=
U

2
2
(
y

y
) m
1
g
43
=
U

2
3
(
z

z
) m
1
U

2
3

z
g
44
= U
||
||
2
+ V
g
45
= 0
g
51
= 0
g
52
=
U

2
2
(
y

y
)
g
53
=
U

2
3
(
z

z
)
g
54
= 0
g
55
= U
||
||
2
+ V
and
h
11
= 0
h
12
= 0
h
13
= 0
h
14
= 0
h
15
= 0
h
21
= 0
h
22
=
U

2
2
(
y

y
+
x

x
) + W
h
23
=
1

2
3

z
_
U
y
W
y
_
h
24
=
_
c|| + U
2||
2
_
(
x

x
)
h
25
=
_
c|| U
2||
2
_
(
x

x
)
h
31
= 0
h
32
=
1

2
2

y
_
U
z
W
z
_
(169)
43
h
33
=
U

2
3
(
z

z
+
x

x
) + W
h
34
=
_
c|| + U
2||
2
_
(
x

x
)
h
35
=
_
c|| U
2||
2
_
(
x

x
)
h
41
= 0
h
42
=
U

2
2
(
y

y
) m
2
U

2
2

z
h
43
=
U

2
3
(
z

z
) m
2
h
44
= U
||
||
2
+ W
h
45
= 0
h
51
= 0
h
52
=
U

2
2
(
y

y
)
h
53
=
U

2
3
(
z

z
)
h
54
= 0
h
55
= U
||
||
2
+ W
References
[1] Giles, M. B., Non-Reecting Boundary Conditions for Euler Equations Calculation, AIAA
Journal, Vol. 28, No. 12, pp. 2050-2058, 1990.
[2] Saxer, A. P., Giles, M. B., Quasi-Three-Dimensional Nonreecting Boundary Conditions for
Euler Equations Calculation, J. Prop. Power, Vol. 9, No. 2, pp. 263-271, 1993.
[3] Shivaji, M., Curvilinear Extension to the Giles Non-Reecting Boundary Conditions for Wall-
Bounded Flows, Master Thesis, University of Toledo, 2006.
[4] Moinier, P., Giles, M. B., Eigenmode Analysis for Turbomachinery Applications, Report no.
04/11, Oxford University Computing Laboratory, 2004.
44