Professional Documents
Culture Documents
SIAM J. NUMER. ANAL. c 2010 Society for Industrial and Applied Mathematics
Vol. 48, No. 1, pp. 346371
OPTIMALLY BLENDED SPECTRAL-FINITE ELEMENT SCHEME
FOR WAVE PROPAGATION AND NONSTANDARD REDUCED
INTEGRATION
MARK AINSWORTH
Abstract. We study the dispersion and dissipation of the numerical scheme obtained by taking
a weighted averaging of the consistent (nite element) mass matrix and lumped (spectral element)
mass matrix for the small wave number limit. We nd and prove that for the optimum blending the
resulting scheme (a) provides 2p +4 order accuracy for pth order method (two orders more accurate
compared with nite and spectral element schemes); (b) has an absolute accuracy which is O(p
3
)
and O(p
2
) times better than that of the pure nite and spectral element schemes, respectively; (c)
tends to exhibit phase lag. Moreover, we show that the optimally blended scheme can be eciently
implemented merely by replacing the usual Gaussian quadrature rule used to assemble the mass and
stiness matrices by novel nonstandard quadrature rules which are also derived.
Key words. numerical dispersion, numerical dissipation, high order numerical wave propagation
AMS subject classications. 65N15, 65N30, 65N35, 35J05
DOI. 10.1137/090754017
1. Introduction. The development and analysis of numerical methods for wave
propagation frequently centers on the issue of controlling errors due to numerical
dispersion and dissipation [4, 8]. The dispersive properties of standard Galerkin nite
element methods of arbitrary order were considered in [1]. However, spectral element
methods [9, 10, 11, 12] have attracted considerable interest in the computational wave
propagation community thanks in part to the fact that the mass matrix is diagonal,
and their superior phase accuracy compared with standard nite elements [2].
Even as early as 1984, the possibility of employing a weighted average of the
nite element and spectral element schemes has been conjectured as a means by
which to obtain the most promising, cost-eective method for computational wave
propagation [13]. Many authors have even commented on the eectiveness of the
scheme obtained by forming a simple average of the spectral and nite element schemes
in the case of rst order elements, but no systematic treatment or analysis seems to
be available. Seriani and Oliveira [15] consider the possibility of blending the methods
using a criterion whereby the phase error vanishes at a particular, user-specied, value
of the normalized wave number. However, this approach means that the blending
parameter is frequency and mesh dependent and may actually result in an increase
in the phase error at frequencies that were originally resolved by the pure nite and
spectral element approaches.
A more natural approach to the selection of the blending parameter that is more in
the spirit of the design of methods for computational wave propagation is to maximize
the order of accuracy in the phase error. This criterion is adopted in the present work.
Received by the editors March 26, 2009; accepted for publication (in revised form) December 3,
2009; published electronically April 16, 2010.
http://www.siam.org/journals/sinum/48-1/75401.html
(x)
2
u(x) = 0, x R.
2.1. Piecewise linear approximation in one dimension. Suppose we dis-
cretize (2.2) using piecewise linear nite elements on a uniform mesh of size h > 0
and seek an approximation of the form
jZ
u
j
j
(x),
where
j
is the usual piecewise linear hat function associated with node x
j
. We obtain
the following equation for the value u
j
of the approximation at node x
j
= jh, j Z:
(2.3) u
j+1
2u
j
+u
j1
+
2
6
(u
j+1
+ 4u
j
+u
j1
) = 0,
where = h. This equation admits nontrivial solutions of the form u
j
= e
ij
(1)
h
provided that
(1)
h satises
(1)
h = cos
1
_
6 2
2
6 +
2
_
Copyright by SIAM. Unauthorized reproduction of this article is prohibited.
348 MARK AINSWORTH AND HAFIZ ABDUL WAJID
or, writing the above expression as a series in ,
(2.4)
(1)
h =
3
24
+ .
This result shows that the nite element approximation tends to exhibit phase lead
compared with the corresponding solutions of the continuous equation (2.2).
An alternative approach is to approximate the problem using the spectral element
method. This results in the following equation for the coecients u
j
:
(2.5) u
j+1
2 u
j
+ u
j1
+
2
u
j
= 0,
leading to nontrivial solutions of the form u
j
= e
ij
(1)
h
, where
(1)
h = cos
1
_
1
2
2
_
or, again expanding as a series in ,
(2.6)
(1)
h = +
3
24
+ .
This means that the spectral element scheme tends to exhibit phase lag. In search of
a numerical scheme with superior phase accuracy, we follow the suggestion of Marfurt
[13] and form a blended scheme by taking a linear combination of (2.3) and (2.5):
(2.7) u
j+1
2u
j
+u
j1
+
2
6
_
(1 )u
j+1
+ 2(2 +)u
j
+ (1 )u
j1
= 0,
where [0, 1] is a parameter whose value is to be determined.
Proceeding as before, we discover that the scheme admits nontrivial solutions of
the form u
j
= e
ij
(1)
h
, where
(1)
h = cos
1
_
2
(2 +) 6
2
( 1) 6
_
,
or, writing the above expression as a series in ,
(2.9)
(1)
h = +
3
24
(2 1) +
5
1920
(20
2
20 + 9) + .
The above expression reduces to those obtained for the nite and spectral element
schemes in the cases = 0 and = 1, respectively. However, more interestingly, we
observe that by choosing = 1/2, two additional orders of accuracy in the phase are
obtained.
2.2. Implementation via nonstandard quadrature rules. The practical
implementation of the blended scheme may at rst sight appear to entail the assembly
of the mass matrices for both the nite and spectral element schemes, which would
be rather unattractive. We can construct another piecewise linear nite element
approximation in which the entries in the mass and stiness matrices are approximated
using the nonstandard quadrature rule
(2.10)
_
1
1
f(x)dx Q
(1)
(f) = f () +f () ,
Copyright by SIAM. Unauthorized reproduction of this article is prohibited.
OPTIMALLY BLENDED SCHEME FOR WAVES 349
where =
_
1
3
(1 + 2). This rule is exact for linear functions, but not products of
linear functions, meaning that the entries appearing in the mass matrix are under-
integrated. The quadrature rule (2.10) is used to develop a composite quadrature rule
I
(1)
,h
on R given by
_
R
f(x)dx
h
2
jZ
_
f(
+
j
) +f(
j
)
_
= I
(1)
,h
(f),
where
j
= (j +
1
2
)h
h
2
, j Z.
The new piecewise linear nite element approximation is then dened by seeking
a nontrivial function of the form
U
h
(; x) =
jZ
u
j
j
(x), x R,
such that
(2.11) I
(1)
,h
(
x
U
h
r
)
2
I
(1)
,h
(U
h
r
) = 0
for all r Z. Interestingly, the resulting scheme gives precisely the same stencil as
(2.7) for the coecients
_
u
j
_
jZ
, where, as before, = h:
u
j+1
2u
j
+u
j1
+
2
6
_
(1 )u
j+1
+ 2(2 +)u
j
+ (1 )u
j1
= 0.
In other words, the scheme coincides with the blended scheme in the case of linear
elements, meaning that the blended scheme can be realized in practice by replacing
the standard Gaussian quadrature rule by the nonstandard rule (2.10). Similarly,
the optimally blended scheme can be obtained by using the quadrature rule (2.10) in
conjunction with the choice = 1/2.
In summary, the nonstandard quadrature rule leads to a scheme which admits a
nontrivial solution given by
(2.12) U
h
(; x) =
jZ
e
ij
(1)
j
(x),
where
(1)
K
f(x, y, z)dxdydz f(, , ) +f(, , ) +f(, , ) +f(, , )
+f(, , ) +f(, , ) +f(, , ) +f(, , ), (2.14)
Copyright by SIAM. Unauthorized reproduction of this article is prohibited.
350 MARK AINSWORTH AND HAFIZ ABDUL WAJID
where =
_
1
3
(1 + 2). If we choose = 0, then the scheme reduces to the standard
nite element approximation while the choice = 1 gives the spectral element scheme.
Consequently, the scheme with a general choice of may be considered as a blended
approximation. We wish to analyze the dispersive properties of the resulting scheme.
Based on our experience in the one dimensional case, we seek a nontrivial solution in
the three dimensional case in the form
u(x, y, z) = U
h
(k
x
; x)U
h
(k
y
; y)U
h
(k
z
; z),
where k
x
, k
y
, k
z
R are constants to be determined and U
h
is dened in (2.12).
Observe that, thanks to (2.12), we could also write
u(x, y, z) =
,m,n
e
ih[
(1)
(k
x
)+m
(1)
(k
y
)+n
(1)
(k
z
)]
(x)
m
(y)
n
(z),
where
(1)
(k
x
; x)
x
q
) I
(1)
,h
(U
h
(k
y
; y)
r
) I
(1)
,h
(U
h
(k
z
; z)
s
)
+I
(1)
,h
(U
h
(k
x
; x)
q
) I
(1)
,h
(
y
U
h
(k
y
; y)
y
r
) I
(1)
,h
(U
h
(k
z
; z)
s
)
+I
(1)
,h
(U
h
(k
x
; x)
q
) I
(1)
,h
(U
h
(k
y
; y)
r
) I
(1)
,h
(
z
U
h
(k
z
; z)
z
s
)
=
2
I
(1)
,h
(U
h
(k
x
; x)
q
) I
(1)
,h
(U
h
(k
y
; y)
r
) I
(1)
,h
(U
h
(k
z
; z)
s
) (2.15)
for all q, r, s Z. Recalling that U
h
(k
x
; x)
q
)I
(1)
,h
(U
h
(k
y
; y)
r
)I
(1)
,h
(U
h
(k
z
; z)
s
) = 0
and as a consequence, we deduce that the new scheme admits a nontrivial solution
provided that
k
2
x
+k
2
y
+k
2
z
=
2
.
The wave number of the discrete solution is given by k
h
, where
k
2
h
=
(1)
(k
x
)
2
+
(1)
(k
y
)
2
+
(1)
(k
z
)
2
,
and then, thanks to (2.9), we deduce that
k
2
h
=
2
+
h
2
12
(2 1)[(k
x
)
4
+ (k
y
)
4
+ (k
z
)
4
] +O(h
4
6
).
We again see that there exists an optimal choice of blending parameter, and moreover,
it coincides with the optimal parameter for the one dimensional case. The arguments
used above extend to any number of dimensions meaning that the optimal blending
parameter is independent of the number of spatial dimensions.
Copyright by SIAM. Unauthorized reproduction of this article is prohibited.
OPTIMALLY BLENDED SCHEME FOR WAVES 351
2.4. Numerical example. In practice, by making use of the nonstandard
quadrature rule, the cost of using the optimally blended scheme is virtually the same
as that of using the pure nite or spectral element scheme, but can result in markedly
superior numerical results. In order to illustrate the potential of such an approach in
multidimensions, we consider the problem
(2.16)
2
u(x, y)
2
u(x, y) = 0 in (0, 1)
2
subject to Dirichlet boundary conditions:
u = e
ik
1
x
on
1
= {x (0, 1), y = 0},
u = e
ik
2
y
on
4
= {y (0, 1), x = 0},
where k
1
and k
2
are user-specied constants satisfying k
2
1
+k
2
2
=
2
, and nonreecting
boundary conditions:
u
x
Gu = 0 on
2
= {y (0, 1), x = 1},
u
y
Gu = 0 on
3
= {x (0, 1), y = 1},
where Gis the usual Dirichlet to Neumann map [8]. The Dirichlet boundary conditions
are chosen so that the exact solution to the boundary value problem (2.16) is the plane
wave solution u(x, y) = e
i(k
1
x+k
2
y)
, with the coecients chosen to be k
1
= 20 and
k
2
= 1. The accuracy of the real components of the spectral, nite, and optimal
scheme solutions obtained with 20 linear elements are compared in Figure 2.1, where
the cuts of the two dimensional solution along the lines y = x and y = 2x relative
to the edge
1
are shown. The phase lead and lag are evident and correspond to
numerical approximations obtained using the nite and spectral element schemes,
respectively. Moreover, the phase accuracy of the numerical approximation obtained
using the optimal scheme is noticeably better than that of nite and spectral element
schemes. In Figure 2.2, we show the eect of increasing the number of elements in
each direction along the same lines as used in Figure 2.1. It is clear that with 30
linear elements the numerical approximations obtained using the nite and spectral
element schemes converge to the exact solution but phase lead and phase lag are still
prominent, while the numerical approximation corresponding to the optimal scheme
is virtually completely resolved.
2.5. Extension to quadratic elements. In the case of quadratic elements, we
obtain two discrete equations corresponding to the nodal and midpoint degrees of
freedom {u
j
} and {u
j+1/2
+u
j+1
_
+
_
140 2
2
(4 +)
_
u
j
+
_
2
2
( 1) 80
(u
j1/2
+u
j+1/2
) = 0 (2.17)
and
_
2
( 1) 40
_
[u
j1
+u
j+1
] +
_
80 2
2
( + 4)
_
u
j+1/2
= 0
with = h. We use the latter relation to express the midpoint degree of freedom
u
j+1/2
and u
j+1
as follows:
u
j+1/2
=
2
( 1) 40
2(
2
( + 4) 40)
_
u
j
+u
j+1
_
, j Z.
Copyright by SIAM. Unauthorized reproduction of this article is prohibited.
352 MARK AINSWORTH AND HAFIZ ABDUL WAJID
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
1.5
1
0.5
0
0.5
1
1.5
2
x
U
Numerical real wave obtained using the spectral element scheme
Numerical real wave obtained using the finite element scheme
Numerical real wave obtained using the optimal scheme
Exact real wave
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
1.5
1
0.5
0
0.5
1
1.5
2
x
U
Numerical real wave obtained using the spectral element scheme
Numerical real wave obtained using the finite element scheme
Numerical real wave obtained using the optimal scheme
Exact real wave
Fig. 2.1. Variation of the numerical approximations of the solution with linear spectral, nite,
and optimal schemes to (2.16) using h = 1 along the lines (a) y = x and (b) y = 2x.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
1.5
1
0.5
0
0.5
1
1.5
2
x
U
Numerical real wave obtained using the spectral element scheme
Numerical real wave obtained using the finite element scheme
Numerical real wave obtained using the optimal scheme
Exact real wave
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
1.5
1
0.5
0
0.5
1
1.5
2
x
U
Numerical real wave obtained using the spectral element scheme
Numerical real wave obtained using the finite element scheme
Numerical real wave obtained using the optimal scheme
Exact real wave
Fig. 2.2. Variation of the numerical approximations of the solution with linear spectral, nite,
and optimal schemes to (2.16) using h = 0.67 along the lines (a) y = x and (b) y = 2x.
Now substituting these values into (2.17), we obtain a single equation corresponding
to nodal degrees of freedom:
4
( 1) + 2
2
(3 8) 240
2(
2
( + 4) 40)
_
u
j1
+u
j+1
_
+
4
(2 + 3) 2
2
(3 + 52) + 240
2
( + 4) 40
u
j
= 0, j Z.
This relation is of the same form as (2.3) and (2.5), and we may therefore proceed
as before by inserting a nontrivial solution of the form u
j
= e
ij
(2)
h
into the above
equation. Proceeding as before, we arrive at the expression
(2.18) cos(
(2)
h) =
4
(2 + 3) 2
2
(3 + 52) + 240
4
(1 ) 2
2
(3 8) + 240
,
and hence, for 1, there exists a solution of the form
(2.19)
(2)
h = +
3 2
2880
5
+
63
2
126 + 88
2419200
7
+ .
Copyright by SIAM. Unauthorized reproduction of this article is prohibited.
OPTIMALLY BLENDED SCHEME FOR WAVES 353
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
1.5
1
0.5
0
0.5
1
1.5
2
x
U
Numerical real wave obtained using the spectral element scheme
Numerical real wave obtained using the finite element scheme
Numerical real wave obtained using the optimal scheme
Exact real wave
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
1.5
1
0.5
0
0.5
1
1.5
2
x
U
Numerical real wave obtained using the spectral element scheme
Numerical real wave obtained using the finite element scheme
Numerical real wave obtained using the optimal scheme
Exact real wave
(a) y = x (b) y = 2x
Fig. 2.3. Variation of the numerical approximations of the solution with quadratic spectral,
nite, and optimal schemes to (2.16) using h = 2.5 along the lines (a) y = x and (b) y = 2x.
For = 0 and = 1, the above expression reduces to the ones obtained for nite
element [1] and spectral element [2] schemes. The choice = 2/3 means the rst term
of the above expression vanishes and gives two additional orders of accuracy in the
phase compared with the standard schemes. Furthermore, the absolute value of the
coecient of the leading term with the optimum value of is decreased by factors of
50 and 25 compared to the leading coecient obtained with the nite and spectral
element schemes, respectively.
We can extend the scheme to higher numbers of spatial dimensions in precisely
the same way as we described earlier for linear elements, provided that a suitable
nonstandard quadrature rule can be identied. One obtains the optimally blended
scheme in the case of quadratic elements if the following quadrature rule,
(2.20)
_
1
1
f(x)dx
1
3(3 + 2)
_
5f
_
_
1
5
(3 + 2)
_
+4(2+3)f(0)+5f
_
_
1
5
(3 + 2)
__
,
is used to approximate the entries in the mass and stiness matrices. Moreover, for the
optimum value of = 2/3, (2.20) reduces to the quadrature rule given in [3]. In Figure
2.3, we show the eect of using piecewise quadratic elements instead of piecewise
linear elements along the lines y = x and y = 2x relative to the bottom edge
1
.
As expected, the numerical approximations corresponding to the nite and spectral
element schemes are, respectively, leading and lagging even with quadratic elements,
but again the optimal scheme performs much better even when h is relatively large.
The results obtained by reducing the size of the elements are given in Figure 2.4.
2.6. Extension to cubic elements. Turning to the case of cubic elements, we
have the following expression for the discrete wave number for the blended scheme:
(3)
h = +
4 3
604800
7
+
4
2
15 + 11
63504000
9
+ ,
where the rst term vanishes corresponding to the optimum value of the blending
parameter = 3/4, and we again observe that two additional orders of accuracy are
achieved.
Copyright by SIAM. Unauthorized reproduction of this article is prohibited.
354 MARK AINSWORTH AND HAFIZ ABDUL WAJID
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
1.5
1
0.5
0
0.5
1
1.5
2
x
U
Numerical real wave obtained using the spectral element scheme
Numerical real wave obtained using the finite element scheme
Numerical real wave obtained using the optimal scheme
Exact real wave
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
1.5
1
0.5
0
0.5
1
1.5
2
x
U
Numerical real wave obtained using the spectral element scheme
Numerical real wave obtained using the finite element scheme
Numerical real wave obtained using the optimal scheme
Exact real wave
(a) y = x (b) y = 2x
Fig. 2.4. Variation of the numerical approximations of the solution with quadratic spectral,
nite, and optimal schemes to (2.16) using h = 2 along the lines (a) y = x and (b) y = 2x.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
1.5
1
0.5
0
0.5
1
1.5
2
x
U
Numerical real wave obtained using the spectral element scheme
Numerical real wave obtained using the finite element scheme
Numerical real wave obtained using the optimal scheme
Exact real wave
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
1.5
1
0.5
0
0.5
1
1.5
2
x
U
Numerical real wave obtained using the spectral element scheme
Numerical real wave obtained using the finite element scheme
Numerical real wave obtained using the optimal scheme
Exact real wave
(a) y = x (b) y = 2x
Fig. 2.5. Variation of the numerical approximations of the solution with cubic spectral, nite,
and optimal schemes to (2.16) using h = 5 along the lines (a) y = x and (b) y = 2x.
Once again, we can extend the scheme to higher numbers of spatial dimensions
provided a suitable quadrature rule is available. For cubic (and higher order) elements,
no such rule seems to be known in the literature. However, we may use the following
new quadrature rule (which is a special case of Theorem 2.1):
(2.21)
_
1
1
f(x)dx
7840
681
_
(f(
+
) +f(
+
))
(39 +
681)(
681 3)
+
(f(
) +f(
))
(39
681)(3 +
681)
_
,
where
=
_
2730 70
(p)
h = +
_
_
1 +
1
p
_
1
_
F
(p)
() + C
(p)
F
(p+1)
() +O()
2p+5
,
where F
(p)
() and C
(p)
(p)
h = +C
(p)
F
(p+1)
() +O()
2p+5
,
showing that in general we obtain two additional orders of accuracy with the optimal
choice of blending parameter . Moreover, in Corollary 3.3, we show that the absolute
value of the leading coecient in the error
(p)
(u, v) = (1 )B(u, v) +
B(u, v)
for piecewise polynomials u and v. The dierence between the bilinear forms for the
nite and spectral element schemes lies in the fact that the spectral element scheme
Copyright by SIAM. Unauthorized reproduction of this article is prohibited.
356 MARK AINSWORTH AND HAFIZ ABDUL WAJID
uses the GaussLobatto quadrature rule which we denote by Q
(p)
to evaluate the
integrals, while the nite element scheme evaluates (via GaussLegendre quadrature)
the integrals exactly. Consequently, the bilinear form for the blended scheme (2.22)
should be based on a quadrature rule Q
(p)
for which
Q
(p)
(f) = (1 )
_
1
1
f(x)dx +Q
(p)
(f) f P
2p+1
,
where P
2p+1
denotes the space of polynomials of degree 2p + 1. The following result
constitutes the extension of the nonstandard quadrature rules to elements of arbitrary
order.
Theorem 2.1. Let [0, 1) be xed, and denote the zeros of L
p+1
L
p1
by
{
i
}
p
i=0
. Then {
i
}
p
i=0
are distinct and contained in (1, 1). Let {w
i
}
p
i=0
denote the
weights dened by the rule
(2.23) w
i
=
2[p(1 +) +]
p(p + 1)L
p
(
i
)[L
p+1
(
i
) L
p1
(
i
)]
i = 0, 1, . . . , p.
Then, the weights {w
i
}
p
i=0
are well-dened and positive. Moreover, if we use the
weights and nodes to dene a (p + 1)-point quadrature rule Q
(p)
, then Q
(p)
satises
the following identity:
(2.24) Q
(p)
(f) = (1 )
_
1
1
f(x)dx +Q
(p)
(f) f P
2p+1
,
where Q
(p)
is the (p +1)-point GaussLegendreLobatto quadrature rule dened in [2,
eq. (2.7)]. Consequently, Q
(p)
p+1
(
i
) L
p1
(
i
)
appearing in the denominator for the quadrature weights (2.23) is nonzero. Moreover,
the factor L
p
(
i
) is also nonzero. This can be seen by rst noting that, thanks to the
recurrence relation for Legendre polynomials,
2p + 1
p + 1
i
L
p
(
i
) = L
p+1
(
i
) +
p
p + 1
L
p1
(
i
)
=
_
+
p
p + 1
_
L
p1
(
i
).
If
i
is nonzero, then the right-hand side is nonzero (otherwise we have L
p1
(
i
) =
L
p
(
i
) = L
p+1
(
i
) = 0 which would imply all subsequent Legendre polynomials have
a common zero), and hence L
p
(
i
) is nonzero. Equally well, if
i
does vanish, then
L
p1
(
i
) also vanishes meaning that p is even, and hence L
p
(0) is nonzero.
The proof of this result is given in section 4. Observe that for = 0 and = 1,
Q
(p)
are given for the optimum value of the blending parameter = p/(p + 1). It
is a simple matter to compute the higher order rules using the expressions given in
Theorem 2.1.
The quadrature rule Q
(p)
(k
x
)
2
+
(p)
(k
y
)
2
+
(p)
(k
z
)
2
,
where k
2
x
+k
2
y
+k
2
z
=
2
. Thanks to (3.2) given in Theorem 3.1, we obtain
k
2
hp
=
2
+
_
p!
(2p)!
_
2
_
_
1 +
1
p
_
1
_
h
2p
2p + 1
[k
2p+2
x
+k
2p+2
y
+k
2p+2
z
]
+O(h
2p+2
2p+4
),
which is valid for general and for all p 2. For the optimal choice of = p/(p +1),
using Corollary 3.3, we have
k
2
hp
=
2
+
8
(2p 1)
_
(p + 1)!
(2p + 2)!
_
2
h
2p+2
2p + 3
[k
2p+4
x
+k
2p+4
y
+k
2p+4
z
] +O(h
2p+4
2p+6
).
3. Analysis of dispersion for elements of arbitrary order. Our rst result
gives the discrete dispersion relation for the blending of spectral-nite element ap-
proximation for elements of arbitrary order p N, with blending parameter [0, 1],
and generalizes the particular cases given in section 2. The proof of the following
theorem follows exactly the same arguments used in [2] for the proof of Theorem 4.1
and is therefore omitted.
Theorem 3.1. Let > 0 and consider the sequences {a
p
}
p=1
and {b
p
}
p=1
dened
by the recursion relations
(3.1)
a
p+1
=
2p + 1
b
p
+a
p1
b
p+1
=
2p + 1
a
p
+b
p1
for p N with a
0
= 1, a
1
= 1, b
0
= 0, and b
1
= 1/. Then, the discrete dispersion
relation for the optimal scheme of order p N is given by
(3.2) cos
(p)
h = R
(p)
(2) = (1)
p
(p)
1
() +
(p)
2
()
(p)
1
()
(p)
2
()
,
Copyright by SIAM. Unauthorized reproduction of this article is prohibited.
358 MARK AINSWORTH AND HAFIZ ABDUL WAJID
where = h, and
(p)
1
() = a
p
(b
p1
((p + 1) +p) +p(2p + 1)a
p
)
and
(p)
2
() = b
p
(a
p1
((p + 1) +p) p(2p + 1)b
p
) .
The sequences {a
p
}
p=1
and {b
p
}
p=1
originally appeared in Lemma 5.2 of [2] in the
analysis of the pure spectral element scheme. In particular, we note that the denom-
inator appearing in (3.2) is nonvanishing. This can be seen by rst noting that
(p)
1
()
(p)
2
() = p(2p + 1)(a
2
p
+b
2
p
),
for 1, and then using the following identity,
(3.3)
_
sin( n/2) cos( n/2)
cos( n/2) sin( n/2)
_ _
a
n
b
n
_
=
_
2
_
J
n+1/2
()
Y
n+1/2
()
_
,
proved in [2] to see that
a
2
p
+b
2
p
=
2
_
J
p+1/2
()
2
+Y
p+1/2
()
2
,
where J and Y are cylindrical Bessel functions of the rst and second kinds, respec-
tively [5]. Finally make use of the identity in the rst equation of (8.479) of [5] to
deduce that
a
2
p
+b
2
p
> 1
for all and p N.
In the case when = 1, the expression (3.2) reduces to the discrete dispersion
relation (4.6) obtained in [2] for spectral element methods, while in the case = 0
expression (3.2) gives an alternative form of the discrete dispersion relation (3.2)
obtained in [1] for nite element methods. As pointed out in [1], R
(p)
(2) is a rational
function of degree [2p/2p] in for all p N which, in the case of the pure nite element
method ( = 0), corresponds to certain types of Pade approximants.
The following theorem proved in section 4 gives the leading term for the error in
the discrete dispersion relation for the blended scheme with parameter [0, 1].
Theorem 3.2. Let p 2 and [0, 1]. Then, the error in the discrete dispersion
relation (3.2) is given by
cos
(p)
h cos =
_
1
_
1 +
1
p
__
F
(p)
() C
(p)
F
(p+1)
() +O()
2p+6
or, if is suciently small,
(3.4)
(p)
h =
_
_
1 +
1
p
_
1
_
F
(p)
() +C
(p)
F
(p+1)
() +O()
2p+5
,
where
C
(p)
=
2
(2p + 3)
(2p 1)
_
1 +
1
p
_
2
(2p + 3)
_
1 +
1
p
_
+ 2
2p
2
+p + 1
2p 1
Copyright by SIAM. Unauthorized reproduction of this article is prohibited.
OPTIMALLY BLENDED SCHEME FOR WAVES 359
and
F
(p)
(s) =
1
2
_
p!
(2p)!
_
2
s
2p+1
2p + 1
.
As expected, when = 0 or = 1, the above result reduces to spectral and nite
element schemes, respectively. More interestingly, (3.4) indicates that the blending
term in the error can be eliminated by choosing = p/(p+1), resulting in an additional
two orders of accuracy in the discrete dispersion relation.
Corollary 3.3. Let p 2. For = p/(p+1), the error in the discrete dispersion
relation (3.2) is given by
(3.5)
(p)
h =
4
(2p 1)
_
(p + 1)!
(2p + 2)!
_
2
2p+3
2p + 3
+O()
2p+5
.
Proof. Substitute = p/(p +1) in (3.4), and applying straightforward manipula-
tions, we obtain (3.5) as required.
In Table 3.1 we give closed form expressions for the rational function R
(p)
()
obtained from Theorem 3.1 along with the leading terms in the error for 1
obtained from Theorem 3.2 for orders p from 1 up to 3 and [0, 1]. Moreover, the
error in the leading term for the optimum value of , i.e., = p/(p+1) obtained from
Corollary 3.3 is given for a small wave number limit.
Table 3.1
The discrete dispersion relation R
(p)
() = cos
(p)
()
1
2
( + 2) 6
2
( 1) 6
2
4
(2 + 3) 2
2
(3 + 52) + 240
4
(1 ) 2
2
(3 8) + 240
3
6
(3 + 4) 4
4
(26 + 135) + 240
2
( + 48) 25200
6
( 1) + 2
4
(8 15) + 120
2
(2 9) 25200
Order p cos
1
R
(p)
() cos
1
R
(p)
() , = p/(p + 1)
1
3
(2 1)
24
+O(
5
)
5
480
2
5
(3 2)
2880
+O(
7
)
7
75600
3
7
(4 3)
604800
+O(
9
)
9
31752000
We make the following observations regarding the optimally blended scheme:
1. The leading term is two orders more accurate compared with the standard
spectral and nite element schemes; see [16, 1, 2, 8], where the leading term in the
Copyright by SIAM. Unauthorized reproduction of this article is prohibited.
360 MARK AINSWORTH AND HAFIZ ABDUL WAJID
10
2
10
1
10
0
10
1
10
15
10
10
10
5
10
0
10
5
|
(
R
(
p
)
(
)
)
2
(
c
o
s
(
)
)
2
|
,
=
0
,
1
,
p
/
(
p
+
1
)
Finite element scheme
Spectral element scheme
Optimal scheme
4
1
1
6
10
2
10
1
10
0
10
1
10
15
10
10
10
5
10
0
10
5
|
(
R
(
p
)
(
)
)
2
(
c
o
s
(
)
)
2
|
,
=
0
,
1
,
p
/
(
p
+
1
)
Finite element scheme
Spectral element scheme
Optimal scheme
6
1
1
8
(a) p = 1 (b) p = 2
10
2
10
1
10
0
10
1
10
15
10
10
10
5
10
0
10
5
|
(
R
(
p
)
(
)
)
2
(
c
o
s
(
)
)
2
|
,
=
0
,
1
,
p
/
(
p
+
1
)
Finite element scheme
Spectral element scheme
Optimal Scheme
1
1
10
8
10
2
10
1
10
0
10
1
10
15
10
10
10
5
10
0
10
5
|
(
R
(
p
)
(
)
)
2
(
c
o
s
(
)
)
2
|
,
=
0
,
1
,
p
/
(
p
+
1
)
Finite element scheme
Spectral element scheme
Optimal scheme
1
12
1
10
(c) p = 3 (d) p = 4
Fig. 3.1. Error in discrete dispersion relations of orders p = 1 to 4 versus wave number for
nite, spectral, and optimal schemes. For pth order nite and spectral element schemes the slope of
the lines is 2p + 2, whereas for optimal scheme the slope of the line is 2p + 4.
expressions was accurate to order O()
2p
. This is illustrated in Figure 3.1, where
it is observed that for a pth order scheme the slope of the lines with spectral and
nite element schemes is 2p + 2, whereas with the optimal scheme the slope is
2p + 4.
2. The coecient of the leading term in the error obtained with the blended
scheme for the optimum value of is 2/(4p
2
1)(2p + 3) and 2p/(4p
2
1)(2p + 3)
times better compared with the leading terms in the error obtained with nite [1] and
spectral [2] element schemes, respectively. This is also illustrated in Figure 3.1, where
it is observed that the absolute value of the error for the optimized scheme is superior
to that of the standard schemes even for modest values of .
3. Figure 3.2 shows the frequency spectra of the nite element, spectral element,
and optimally blended scheme for elements of order p = 1, . . . , 4. In particular, one
can see the so-called cuto frequencies of the schemes corresponding to the values of
at which the magnitude of the rational function R
(p)
(1)
h)
N
o
r
m
a
l
i
z
e
d
F
r
e
q
u
e
n
c
e
y
Finite element scheme
Spectral element scheme
Optimal scheme
Exact continuum
= 3.46
= 2.00
= 2.45
Cutoff frequencies
0 1 2 3 4 5 6 7
0
1
2
3
4
5
6
7
8
9
10
Normalized Wavenumber real(
(2)
h)
N
o
r
m
a
l
i
z
e
d
F
r
e
q
u
e
n
c
e
y
Finite element scheme
Spectral element scheme
Optimal scheme
Exact continuum
(a) p=1 (b) p=2
0 1 2 3 4 5 6 7 8 9 10
0
2
4
6
8
10
12
14
Normalized Wavenumber real(
(3)
h)
N
o
r
m
a
l
i
z
e
d
F
r
e
q
u
e
n
c
e
y
Finite element scheme
Spectral element scheme
Optimal scheme
Exact continuum
0 2 4 6 8 10 12 14
0
2
4
6
8
10
12
14
16
18
20
Normalized Wavenumber real(
(4)
h)
N
o
r
m
a
l
i
z
e
d
F
r
e
q
u
e
n
c
e
y
Finite element scheme
Spectral element scheme
Optimal scheme
Exact continuum
(c) p=3 (d) p=4
Fig. 3.2. Frequency spectra of the one dimensional spectral, nite, and optimally blended
scheme for elements of order p = 1, . . . , 4.
for greater than
(3.6) =
_
12
1 + 2
.
For the nite element scheme ( = 0) and spectral element scheme ( = 1) we obtain
cuto frequencies of 2
6. These
frequencies are indicated in Figure 3.2(a). All of the schemes corresponding to p = 1
elements fail to admit propagating waves for higher frequencies and thus have a single
stopping band extending to innity. In general, the schemes based on elements of
order p have p stopping bands, as shown in Figure 3.2(b)(d).
The so-called spatial resolution limit is dened [16] to be the number of elements
per wavelength corresponding to the cuto frequency. The spatial resolution limit for
p = 1 elements is obtained by inserting (3.6) into (2.8) to obtain
cos
(1)
h = R
(1)
_
_
12
1 + 2
_
= 1
Copyright by SIAM. Unauthorized reproduction of this article is prohibited.
362 MARK AINSWORTH AND HAFIZ ABDUL WAJID
or
(1)
h = ,
giving a spatial resolution limit for the p = 1 elements of
h
=
2
(1)
h
=
2
= 2.
Examination of Figure 3.2(b)(d) reveals that for a pth order scheme, the spatial reso-
lution limit is given by 2/p elements per wavelength in agreement with the observation
of [16].
4. Proofs of the results. This section provides the proofs of general results for
the error in the discrete dispersion relation for the blended scheme.
4.1. Basic polynomials. Let p N be given and [0, 1] be a parameter to
be determined. Dene the bilinear form
(4.1)
B
(v, w) = (1 )
_
1
1
(v
2
vw)dx +Q
(p)
(v
2
vw),
where Q
(p)
is the (p+1) point GaussLobatto quadrature rule and > 0 is a constant.
If v, w P
p
, then v
P
2p2
and the quadrature rule Q
(p)
integrates this product
exactly. Hence, if v, w P
p
, then
(4.2)
B
(v, w) =
_
1
1
v
dx
2
_
(1 )
_
1
1
vwdx +Q
(p)
(vw)
_
.
We introduce basic polynomials
p
,
p
P
p
for p N dened in [2] by
(4.3)
p
(1) = 1,
p
(1) = (1)
p+1
:
B
(
p
, w) = 0 w P
p
H
1
0
(1, 1)
and
(4.4)
p
(1) = 1,
p
(1) = (1)
p
:
B
(
p
, w) = 0 w P
p
H
1
0
(1, 1).
Observe that if is suciently small, then (4.3) and (4.4) admit a unique solution.
From (4.3) and considering the parity of
p
, we deduce that
p
P
p1
for all p N.
Moreover, w
p
and w
p
P
2p1
, for w P
p
H
1
0
(1, 1), and it follows that the
quadrature rule Q
(p)
is exact in (4.1) for w =
p
, so that
(
p
, w) =
_
1
1
_
p
w
p
w
_
dx = 0 w P
p
H
1
0
(1, 1).
Hence, for
p
the bilinear form (4.1) coincides with the bilinear form
B(v, w) = Q
(p)
(v
)
2
Q
(p)
(vw)
considered in [2], and we may, therefore, for
p
quote results for
p
directly from [2].
In particular, from Theorem 5.1 of [2], we have
(
p
,
p
) =
B(
p
,
p
) = 2
a
p
b
p
,
Copyright by SIAM. Unauthorized reproduction of this article is prohibited.
OPTIMALLY BLENDED SCHEME FOR WAVES 363
where a
p
and b
p
are dened in Theorem 3.1. We shall require the corresponding
expression for
B
(
p
,
p
).
Theorem 4.1. Let p = 2, 3, 4, . . . . Then
(4.5)
B
(
p
,
p
) = 2
pa
p+1
+(p + 1)a
p1
pb
p+1
+(p + 1)b
p1
,
where {a
p
} and {b
p
} are dened in Theorem 3.1 and 2 = .
Proof. For the duration of this proof the superscript on
p
will be omitted since no
confusion is likely to arise. Suppose w P
p1
H
1
0
(1, 1), then w P
2p1
. Using
the fact that the quadrature rule in the bilinear form (4.1) is exact for functions
belonging to P
2p1
, and using denition (4.4), we obtain
(4.6)
_
1
1
(
+
2
)wdx = 0 w P
p1
H
1
0
(1, 1).
Now, we can write F(x) =
(x) +
2
(x) P
p
in the form
(4.7)
(x) +
2
(x) =
p+1
j=1
j
L
j
(x),
where
j
are the scalars and L
j
is the Legendre polynomial of degree j. Now inserting
w(x) = (1 x
2
)L
1
=
2
= =
p2
= 0. Also, parity considerations imply that
p
= 0. Hence,
(4.8) F(x) =
(x) +
2
(x) =
p+1
L
p+1
(x) +
p1
L
p1
(x).
Now, choosing w(x) = (1 x
2
)L
p1
(x) P
p
H
1
0
(1, 1) in denition (4.4), we get
(4.9)
_
1
1
F(x)w(x)dx
2
__
1
1
(x)w(x)dx Q
(p)
(w)
_
= 0.
Also, using (4.8) together with the rst term of the last expression gives
_
1
1
F(x)w(x)dx =
_
1
1
(1 x
2
)
_
p+1
L
p+1
(x)L
p1
(x) +
p1
[L
p1
(x)]
2
dx,
and then exploiting the orthogonality property of the Legendre polynomials, we obtain
(4.10)
_
1
1
F(x)w(x)dx =
2p(p 1)
2p 1
p1
.
The error in the GaussLobatto quadrature rule is given by
(4.11) E =
_
1
1
(x)w(x)dx Q
(p)
(w) =
(p + 1)p
3
2
2p+1
[(p 1)!]
4
(2p + 1)[(2p)!]
3
D
2p
(w),
using (4.10-27) from [14]. Now using the Leibniz rule, we get
(4.12) D
2p
(w) =
(2p)!
p!
2
(p)
(0)w
(p)
(0),
Copyright by SIAM. Unauthorized reproduction of this article is prohibited.
364 MARK AINSWORTH AND HAFIZ ABDUL WAJID
where
(4.13)
(p)
(0) =
p+1
2
L
(p+1)
p+1
(0)
obtained by dierentiating (4.8) p times with respect to x. Moreover,
w
(p)
(x) = D
p
_
(1 x
2
)L
p1
(x)
.
Now using the identity D
p
_
(1 x
2
)L
p1
(x)
= p(p 1)L
(p1)
p1
(x), we get
(4.14) w
(p)
(0) = p(p 1)L
(p1)
p1
(0).
Hence, using L
(p)
p
(0) =
(2p)!
p!
1
2
p
, together with (4.13) and (4.14), (4.12) simplies to
give
D
2p
(w) =
p+1
2
(2p + 1)(p 1)[(2p)!]
3
p
3
2
2p
(2p 1)[(p 1)!]
4
.
Now substituting this value into (4.11) and performing ordinary manipulations gives
(4.15) E =
_
1
1
(x)w(x)dx Q
(p)
(w) =
2
2
(p
2
1)
2p 1
p+1
.
Inserting the values from (4.10) and (4.15) into (4.9), we obtain
p1
=
(p+1)
p
p+1
.
Consequently, (4.8) can be rewritten in the form
(4.16) F(x) =
(x) +
2
(x) =
p+1
_
_
1 +
1
p
_
L
p1
(x) +L
p+1
(x)
_
.
Observe that we may write
(4.17) (x) =
p
(x) +
p2
(x)
for suitable constants and where is given in [2, 5.10] and dened as
(4.18)
p
(x) =
p/2
j=0
_
2
_
j+1
L
(2j+1)
p+1
(x)
and satises the property
(4.19)
p
(x) +
2
p
(x) = L
p+1
(x).
Consequently, we have
F(x) =
_
p
(x) +
2
p
(x)
_
+
_
p2
(x) +
2
p2
(x)
_
;
using the property (4.19), we get
F(x) = L
p+1
(x) L
p1
(x).
Copyright by SIAM. Unauthorized reproduction of this article is prohibited.
OPTIMALLY BLENDED SCHEME FOR WAVES 365
Comparing the last equation with (4.16), we are led to the choices =
p+1
and
=
p+1
(p + 1)/p, and with these values, (4.17) becomes
(4.20) (x) =
p+1
_
_
1 +
1
p
_
p2
(x) +
p
(x)
_
.
Applying the boundary condition (1) = 1, we obtain
p+1
= p/(1), provided
that (1) is nonzero, with (1) = (p +1)
p2
(1) +p
p
(1). Consequently, may be
written in the form
(4.21) (x) =
(x)
(1)
.
We want to obtain a closed form expression for (4.1), and for this we dene
I
(x) =
(x + 1)/2 + (1)
p
(1 x)/2, so
(, ) =
B
(,
I
) +
B
(,
I
)
= [
I
]
1
1
_
1
1
(
+
2
)
I
dx,
and applying integration by parts together with (4.16), we get
(, ) = 2
(1) 2
p+1
_
1 +
_
1 +
1
p
__
=
2
(1)
[
((p + 1)b
p1
+pb
p+1
)
and
(
p
,
p
) = 2
_
(p + 1)a
p1
+pa
p+1
(p + 1)b
p1
+pb
p+1
_
which completes the proof.
For
p
we rewrite expressions (5.20) and (5.22) given in [2] in terms of F
(p)
(2)
with higher order terms as the quadrature rule is exact for
p
in the bilinear form
(4.1). Moreover, the estimates (5.20) and (5.22) are the same as (4.16) and (4.15)
in [1], for p = 2N and p = 2N + 1, respectively. Therefore, when p is even and
= (m+ 1/2), m Z, then
E
(p)
(2) = 2
F
(p)
(2)
2
(2p + 1)
2
2p 1
F
(p+1)
(2)
+O(2
2p+4
),
Copyright by SIAM. Unauthorized reproduction of this article is prohibited.
366 MARK AINSWORTH AND HAFIZ ABDUL WAJID
and when p is odd and = m, m Z, then
E
(p)
(2) = 2F
(p)
(2)
(2p + 1)
2
2p 1
2F
(p+1)
(2) +O(2
2p+6
).
Theorem 4.2. Let p N satisfy p 2. Then
1. if p is even, and = (m + 1/2), m Z, then
(4.24) E
(p)
(2) = 2
_
1 +
1
p
_
F
(p)
(2) 2F
(p+1)
(2)
C
(p)
+O(2
2p+6
);
2. if p is odd, and = m, m Z, then
E
(p)
(2) = 2
_
1 +
1
p
_
F
(p)
(2)
2
F
(p+1)
(2)
C
(p)
+O(2
2p+4
), (4.25)
where
C
(p)
=
2
_
1 +
1
p
_
2
2p + 3
2p 1
_
1 +
1
p
_
(2p + 3) 1
and
F
(p)
(2) =
1
2
_
p!
(2p)!
_
2
(2)
2p+1
2p + 1
p 2.
Proof. First, consider the case when p is even. Writing the series {a
p
} and {b
p
}
in (4.23) in terms of Bessel functions using the identity (3.3), we get
(
p
,
p
) =
2
(p + 1)(J
p1/2
() cot +Y
p1/2
()) p(J
p+3/2
() cot +Y
p+3/2
())
(p + 1)(J
p1/2
() Y
p1/2
() cot ) p(J
p+3/2
() Y
p+3/2
() cot )
, (4.26)
where J and Y are cylindrical Bessel functions of the rst and second kinds, respec-
tively, and = (m + 1/2) for all m Z. Adding 2tan to (4.26) and applying
straightforward manipulations give
(4.27)
B
(
p
,
p
) + 2tan =
2
cos
2
Q
p+3/2
()
_
1 Q
p+3/2
() tan
_
1
,
where
(4.28) Q
p+3/2
() =
(p + 1) J
p1/2
() pJ
p+3/2
()
(p + 1) Y
p1/2
() pY
p+3/2
()
,
and for small , i.e., when 1 is given by
(4.29) Q
p+3/2
() =
_
1 +
1
p
_
F
(p)
(2) +
C
(p)
F
(p+1)
(2) + .
With the aid of this estimate, we obtain that
E
(p)
(2) = 2
_
1 +
1
p
_
F
(p)
(2) 2F
(p+1)
(2)
C
(p)
+ .
Copyright by SIAM. Unauthorized reproduction of this article is prohibited.
OPTIMALLY BLENDED SCHEME FOR WAVES 367
The assertions concerning polynomials of odd order are proved in a similar fashion.
Again, using the identity (3.3), we write the series {a
p
} and {b
p
} in (4.23) in terms
of Bessel functions and get
(
p
,
p
) =
2
(p + 1)(Y
p1/2
() cot J
p1/2
()) p(Y
p+3/2
() cot J
p+3/2
())
(p + 1)(J
p1/2
() +Y
p1/2
() cot ) p(J
p+3/2
() +Y
p+3/2
() cot )
, (4.30)
where = m for all m Z. Subtracting 2cot from (4.30) and after simplication,
we obtain
(4.31)
B
(
p
,
p
) 2cot =
2
sin
2
Q
p+3/2
()
_
1 +Q
p+3/2
() cot
_
1
.
Now, using (4.29) in the above expression and simplifying gives
E
(p)
(2) = 2
_
1 +
1
p
_
F
(p)
(2)
2
F
(p+1)
(2)
C
(p)
+
as required.
4.2. Proof of Theorem 3.2. We now prove Theorem 3.2 by using expressions
derived in [2] for cos
(p)
h cos 2 =
2
E
(p)
+ E
(p)
+
obtained in [2] for small ; i.e., 1. Now,
2
E
(p)
= 2F
(p)
(2) 2F
(p+1)
(2)
(2p + 1)
2
2p 1
+
and
E
(p)
(2) = 2
_
1 +
1
p
_
F
(p)
(2) 2F
(p+1)
(2)
C
(p)
+ .
Inserting these values into (4.32) and simplifying gives
(4.33) cos
(p)
h cos 2 = 2
_
1
_
1 +
1
p
__
F
(p)
(2) 2C
(p)
F
(p+1)
(2) + ,
where C
(p)
=
2
(2p+3)
(2p1)
(1 +
1
p
)
2
(2p + 3)(1 +
1
p
) + 2
2p
2
+p+1
2p1
.
For the case when p is odd, we reconsider
(4.34) cos
(p)
h cos 2 =
2
E
(p)
+ E
(p)
+
obtained in [2], where
2
E
(p)
= 2
_
1 +
1
p
_
F
(p)
(2) 2F
(p+1)
(2)
C
(p)
+
Copyright by SIAM. Unauthorized reproduction of this article is prohibited.
368 MARK AINSWORTH AND HAFIZ ABDUL WAJID
and
E
(p)
(2) = 2F
(p)
(2) 2F
(p+1)
(2)
(2p + 1)
2
2p 1
+ .
Now, substituting these values into (4.34) and simplifying results in (4.33), which is
what was required. Finally, for small
(p)
h, the approximation
cos
(p)
h cos 2 = 2(
(p)
h 2) +
gives us the required estimate (3.4).
4.3. Proof of Theorem 2.1.
Proof. Let f P
p
be written in the form
f(x) =
p
j=0
j
(x)f(
j
),
where
j
P
p
satises
j
(
k
) =
jk
j = 0, 1, 2, . . . , p. Applying the quadrature rule
Q
(p)
gives
Q
(p)
(f) =
p
j=0
w
j
f(
j
),
where {
j
}
p
j=0
and {w
j
}
p
j=0
are the nodes and weights of Q
(p)
, respectively. Later,
we show that the quadrature weights dened in (2.23) satisfy
(4.35) w
j
=
_
1
1
j
(x)dx.
Hence, for f P
p
Q
(p)
(f) =
p
j=0
f(
j
)
_
1
1
j
(x)dx =
_
1
1
p
j=0
f(
j
)
j
(x)dx =
_
1
1
f(x)dx,
and so Q
(p)
p
f(x)dx.
Moreover, since vanishes at the nodes of Q
(p)
, we can write
_
1
1
p
f(x)dx = Q
(p)
(
p
f) = Q
(p)
(
p
f +q) = Q
(p)
(f),
Copyright by SIAM. Unauthorized reproduction of this article is prohibited.
OPTIMALLY BLENDED SCHEME FOR WAVES 369
and it follows that Q
(p)
(f) = (1 )
_
1
1
f(x)dx +Q
(p)
(f) f P
2p1
,
and hence identity (2.24) holds for f P
2p1
.
Now let f P
2p
be written in the form
f(x) = (x)L
p1
(x) +q(x)
for suitable constant R and q P
2p1
. Since vanishes at the quadrature points
of Q
(p)
, we have
(4.37) Q
(p)
(f) = Q
(p)
(q) = (1 )
_
1
1
q(x)dx +Q
(p)
(q),
where the second step follows from (2.24) applied to q P
2p1
. The orthogonality
property of Legendre polynomials means that
(1 )
_
1
1
(x)L
p1
(x)dx = (1 )
_
1
1
(L
p+1
(x) L
p1
(x))L
p1
(x)dx
= (1 )
_
1
1
L
2
p1
(x)dx.
Furthermore, since L
p+1
and L
p1
coincide at the nodes of the GaussLobatto quadra-
ture rule, we have
Q
(p)
(L
p1
) = Q
(p)
([L
p+1
L
p1
]L
p1
) = (1 )Q
(p)
(L
2
p1
),
and since the GaussLobatto rule has precision 2p 1, we obtain
Q
(p)
(L
p1
) = ( 1)
_
1
1
L
2
p1
(x)dx.
Consequently, we deduce that
(1 )
_
1
1
(x)L
p1
(x)dx +Q
(p)
(L
p1
) = 0,
and then adding times this identity to (4.37) shows that
Q
(p)
(f) = (1 )
_
1
1
f(x)dx +Q
(p)
(f)
for all f P
2p
. It is trivial to see that (2.24) now holds for all f P
2p+1
since both
sides of (2.24) vanish identically when f is the odd function f(x) = x
2p+1
.
The positivity of the weights can be seen by inserting f(x) =
2
j
(x) P
2p
into
(2.24) to obtain for [0, 1)
w
j
= (1 )
_
1
1
2
j
(x)dx +Q
(p)
(
2
j
) > 0.
Copyright by SIAM. Unauthorized reproduction of this article is prohibited.
370 MARK AINSWORTH AND HAFIZ ABDUL WAJID
We now show that the nodes {
i
}
p
i=0
are real, are distinct, and lie within the
interval (1, 1). Suppose this were not the case. Let {
i
}
m
i=0
with m < p be the
points where (x) P
p+1
changes sign in (1, 1), and then the polynomial W(x) =
(x
0
)(x
1
) (x
m
)(x) vanishes at the nodes of Q
(p)
(W) = (1 )
_
1
1
W(x)dx +Q
(p)
(W),
but the right-hand side is nonzero since W does not change sign, and we obtain a
contradiction. Hence m = p.
Above, we have shown that (2.24) holds provided that the weights satisfy (4.35).
We now show that choosing the weights according to (4.35) implies (2.23) holds.
Observe that
J
(x) = (x)/(x
J
)
(
J
) P
p
so that
(4.39) w
J
=
_
1
1
J
(x)dx =
1
(
J
)
_
1
1
(x)
x
J
dx.
We recall the ChristoelDarboux identity [14, 4.7-3]:
p
k=0
L
k
(x)L
k
(y)(2k + 1) =
L
p+1
(x)L
p
(y) L
p
(x)L
p+1
(y)
x y
(p + 1), x = y.
Choose y =
J
and integrate from 1 to 1 with respect to x to get
(4.40)
_
1
1
L
p+1
(x)L
p
(
J
) L
p
(x)L
p+1
(
J
)
x
J
dx =
2
p + 1
, p N.
Now, inserting L
p+1
(x) = (x) +L
p1
(x) and L
p+1
(
J
) = L
p1
(
J
) gives
2
p + 1
= L
p
(
J
)
_
1
1
(x)
x
J
dx +
_
1
1
L
p1
(x)L
p
(
J
) L
p
(x)L
p1
(
J
)
x
J
dx,
and then using (4.40), we obtain
2
p + 1
= L
p
(
J
)
(
J
)w
J
2
p
.
Inserting
(
J
) = L
p+1
(
J
) L
p1
(
J
) in the above equation and performing
straightforward manipulations, we arrive at the conclusion
w
J
=
2[p(1 +) +]
p(p + 1)L
p
(
J
)[L
p+1
(
J
) L
p1
(
J
)]
J = 0, 1, . . . , p
as required.
Copyright by SIAM. Unauthorized reproduction of this article is prohibited.
OPTIMALLY BLENDED SCHEME FOR WAVES 371
REFERENCES
[1] M. Ainsworth, Discrete dispersion relation for hp-version nite element approximation at
high wave number, SIAM J. Numer. Anal., 42 (2004), pp. 553575.
[2] M. Ainsworth and H. A. Wajid, Dispersive and dissipative behaviour of the spectral element
method, SIAM J. Numer. Anal., 47 (2009), pp. 39103937.
[3] S. Challa, High-order Accurate Spectral Elements for Wave Propagation, Masters thesis in
Mechanical Engineering, Clemson University, Clemson, SC, 1998.
[4] G. C. Cohen, Higher-order numerical methods for transient wave equations, Scientic Com-
putation, Springer-Verlag, Berlin, 2002.
[5] I. S. Gradshteyn and I. M. Ryzhik, Table of integrals, series, and products, 4th edition
prepared by Ju. V. Geronimus and M. Ju. Cetlin. Translated from the Russian by Scripta
Technica, Inc. Translation edited by Alan Jerey, Academic Press, New York, 1965.
[6] M. N. Guddati and B. Yue, Modied integration rules for reducing dispersion error in nite
element methods, Comput. Methods Appl. Mech. Engrg., 193 (2004), pp. 275287.
[7] M. N. Guddati and B. Yue, Dispersion-reducing nite elements for transient acoustics, J.
Acoust. Soc. Am., 118 (2005), pp. 21322141.
[8] F. Ihlenburg, Finite element analysis of acoustic scattering, Applied Mathematical Sciences
132, Springer-Verlag, Berlin, 1998.
[9] D. Komatitsch and J. Tromp, Spectral-element simulations of global seismic wave
propagation-I. Validation, Geophys. J. Int., 149 (2002), pp. 390412.
[10] D. Komatitsch and J. Tromp, Spectral-element simulations of global seismic wave
propagation-II. 3-D Models, oceans, rotation, and self-gravitation, Geophys. J. Int., 150
(2002), pp. 303318.
[11] D. Komatitsch and J. P. Vilotte, The spectral-element method: An ecient tool to simulate
the seismic response of 2D and 3D geological structures, Bull. Seismol. Soc. Amer., 88
(1998), pp. 368392.
[12] F. Maggio and A. Quarteroni, Acoustic wave simulation by spectral methods, East-West J.
Numer. Math., 2 (1994), pp. 129150.
[13] K. J. Marfurt, Accuracy of nite-dierence and nite-element modeling of the scalar and
elastic wave equations, Geophysics, 49 (1984), pp. 533549.
[14] A. Ralston, A First Course in Numerical Analysis, McGraw-Hill, New York, 1965.
[15] G. Seriani and S. P. Oliveira, Optimal blended spectral-element operators for acoustic wave
modeling, Geophysics, 72 (2007), pp. SM95SM106.
[16] L. L. Thompson and P. M. Pinsky, Complex wave number Fourier analysis of the p-version
nite element method, Comput. Mech., 13 (1994), pp. 255275.