140 L. Levaggi – A. Tabacco
5. Wavelet function spaces for the half-line
We now have all the tools to build the detail spaces and the wavelets on the half-line. Recalling the abstract construction, we start from level j
= 0 and look for a complement space W
+
such that V
1 +
= V
+
⊕ W
+
note that the sum is not, in general, orthogonal and W
+
⊥ e V
. To this end, let us consider the basis functions of V
1 +
and let us write them as a sum of a function of V
+
and a function which will be an element of W
+
. Since we have based our construction on the existence of a multilevel decomposition on the real line,
we report two equations that will be largely used in the sequel see, e.g., [6]:
ϕ
1k
= 2
1 p
−
1 2
X
˜ n
≤k−2m≤ ˜n
1
˜h
k−2m
ϕ
0m
+ X
1−n
1
≤k−2m≤1−n
˜g
k−2m
ψ
0m
,
58 ψ
0m
= 2
1 2
−
1 p
1+2m− ˜ n
X
l=1+2m− ˜ n
1
g
l−2m
ϕ
1l
. 59
Interior wavelets. Since V
+
contains the subspace V
+
defined in 32, V
1 +
contains the subspace T
1
V
+
= ϕ
1k [0,+∞
: k ≥ k
∗
. Considering equation 59, let us determine the integer m such that all the indices l in the sum are greater or equal to k
∗
. This is equivalent to 2m
≥ k
∗
+ ˜n
1
− 1, so we set m
≥ k
∗
+ ˜n
1
− 1 2
=: m
∗
. 60
Since, see 2.9 in [11], X
n∈
˜h
n
h
n−2k
= δ
0k
, ∀k ∈
✂
, 61
it is easy to see that ˜n
1
− n is always odd; thus, by 39,
m
∗
= ˜n
1
− n − 1
2 +
δ 2
. 62
Let us set W
I
: = span
n ψ
0m [0,+∞
: m ≥ m
∗
o ;
63 we observe that W
I
can be identified with a subspace of W , thus it is orthogonal to e
V and
W
I
⊆ W
+
. The functions ψ
0m
: = ψ
0m [0,+∞
are called interior wavelets. Border wavelets. Let us now call W
B
a generic supplementary space of W
I
in W
+
and set V
B 1
= T
1
V
B
. P
ROPOSITION
8. The dimension of the space W
B
is m
∗
.
Wavelets on the interval 141
Proof. Let K 0 be an integer such that on [K , +∞ all non-vanishing wavelets and scaling
functions are interior ones. Then V
B 1
⊕ span ϕ
1k
: k
∗
≤ k −n + 2K
= h
V
B
⊕ span ϕ
0k
: k
∗
≤ k −n + K
i ⊕
h W
B
⊕ span ψ
0m
: m
∗
≤ m ≤ K − 1 i
, since, by 58, the first interior wavelet used to generate ϕ
1,−n +2K
is ψ
0K
. Then the result easily follows.
To build W
B
we need some functions that, added to W
I
, will generate both V
B 1
and the interior scaling functions that cannot be obtained in 58 using V
+
and W
I
. Thanks to 51, we only have to consider the problem of generating interior scaling functions. Let us now look
for the functions ϕ
1k
generated by ϕ
0m
, for m ≥ k
∗
, and by ψ
0m
, with m ≥ m
∗
. Let us work separately on the two sums of 58:
a we must have m ≥ k − ˜n
1
2. Imposing m ≥ k
∗
and seeking for integer solutions, we get
k − ˜n
1
2 ≥ k
∗
= −n + δ ;
64 b similarly, we obtain m
≥ n − 1 + k2. Again, we want m ≥ m
∗
, so we must have n
− 1 + k 2
≥ m
∗
. Using 62, this means
n − 1 + k
2 ≥
−n + ˜n
1
− 1 2
+ δ
2 .
65 Since 64 and 65 have to be both satisfied, we obtain the following condition
k ≥ −2n
+ ˜n
1
+ 2δ − 1 = 2k
∗
+ ˜n
1
− 1 . 66
Indeed, this can be seen considering all possible situations. For instance, if n and δ are even,
then ˜n
1
is odd and l
δ 2
m =
δ 2
. If k satisfies both 64 and 65, so does k − 1; thus we can look for
the least k as an even integer. In this case l
k− ˜ n
1
2
m =
k− ˜ n
1
2
+
1 2
and l
n −1+k
2
m =
n −1+k
2
+
1 2
, and 66 easily follows. The other cases are dealt with similarly. Let us set
k = 2k
∗
+ ˜n
1
− 1 , 67
so that span
n ϕ
1k [0,+∞
: k ≥ k
o ⊆ V
+
⊕ W
I
. We are left with the problem of generating some functions of V
1 +
, precisely ϕ
1k
with k
∗
≤ k k. Observe that we have to generate k
− k
∗
functions using a space of dimension m
∗
=
k−k
∗
2
. In fact one can show that one out of two ϕ
1k
, for k = k
∗
, . . . , k
− 1, depends on the previous ones through elements of level zero.
142 L. Levaggi – A. Tabacco
P
ROPOSITION
9. Let e = 0 if δ is even, e = 1 if δ is odd. For all 1 ≤ m ≤ m
∗
− e, one has ϕ
1,k−2m [0,+∞
∈ S
m
⊕ V
I
⊕ W
I
, with
S
m
: = span
n ϕ
1,k−2l+1 [0,+∞
: 1 ≤ l ≤ m
o .
Proof. Let us set ˜n
1
− n = 2r + 1 with r 0 recall that ˜n
1
− n is odd. Indeed, it is not
difficult to see that for r = 0 there is nothing to prove. Observe that, by 58, for any l ∈
✂
, we have
ϕ
1,k−2l
=2
1 p
−
1 2
X
n≥k
∗
−l
˜h
k−2l−2n
ϕ
0n
+ X
n≥m
∗
+ j
δ 2
k −l
˜g
k−2l−2n
ψ
0n
=2
1 p
−
1 2
˜h
˜ n
1
−1
ϕ
0,k
∗
−l
+ ˜h
˜ n
1
−3
ϕ
0,k
∗
−l+1
+ . . . + ˜g
−n
ψ
0,m
∗
+ j
δ 2
k −l
+ ˜g
−n −2
ψ
0,m
∗
+ j
δ 2
k −l+1
+ . . . 68
and ϕ
1,k−2l+1
=2
1 p
−
1 2
X
n≥k
∗
−l
˜h
k−2l+1−2n
ϕ
0n
+ X
n≥m
∗
+ j
δ 2
k −l
˜g
k−2l+1−2n
ψ
0n
=2
1 p
−
1 2
˜h
˜ n
1
ϕ
0,k
∗
−l
+ ˜h
˜ n
1
−2
ϕ
0,k
∗
−l+1
+ . . . + ˜g
−n +1
ψ
0,m
∗
+ j
δ 2
k −l
+ ˜g
−n −1
ψ
0,m
∗
+ j
δ 2
k −l+1
+ . . . .
69
Let us prove the stated result by induction on m. For m = 1 we consider the linear combination
h
n
ϕ
1,k−2
+ h
n +1
ϕ
1,k−1
= 2
1 p
−
1 2
h
n
˜h
˜ n
1
−1
+ h
n +1
˜h
˜ n
1
ϕ
0,k
∗
−1
+ X
n≥k
∗
c
1n
ϕ
0n
+ h
n
˜g
−n
+ h
n +1
˜g
−n +1
ψ
0,m
∗
+ j
δ 2
k −1
+ X
n≥m
∗
+ j
δ 2
k
d
1n
ψ
0n
, for some coefficients c
1n
and d
1n
. Writing 61 with k =
n − ˜
n
1
+1 2
6= 0 and X
n∈
˜g
n
h
n−2k
= 0 see 3.29 in [11] with k
= −n , we have
h
n
˜h
˜ n
1
−1
+ h
n +1
˜h
˜ n
1
= 0 , h
n
˜g
−n
+ h
n +1
˜g
−n +1
= 0 .
Wavelets on the interval 143
Thus we have h
n
ϕ
1,k−2 [0,+∞
+ h
n +1
ϕ
1,k−1 [0,+∞
∈ V
I
⊕ W
I
, and the result follows because h
n
6= 0. Set now 1 m
≤ m
∗
− e. As before, we choose a certain linear combination of the scaling functions ϕ
1,k−2l
and ϕ
1,k−2l+1
with 1 ≤ l ≤ m. Then we use 68, 69 to represent them
through functions of level 0. More precisely h
n
ϕ
1,k−2m
+ h
n +1
ϕ
1,k−2m+1
+ . . . + h
n +2m−2
ϕ
1,k−2
+ h
n +2m−1
ϕ
1,k−1
= h
n
˜h
˜ n
1
−1
+ h
n +1
˜h
˜ n
1
ϕ
0,k
∗
−m
+ h
n
˜h
˜ n
1
−3
+ h
n +1
˜h
˜ n
1
−2
+ h
n +2
˜h
˜ n
1
−1
+ h
n +3
˜h
˜ n
1
ϕ
0,k
∗
−m+1
+ . . . +
h
n
˜h
˜ n
1
−2m+1
+ h
n +1
˜h
˜ n
1
−2m+2
+ . . . + h
n +2m−1
˜h
˜ n
1
ϕ
0,k
∗
−1
+ X
n≥k
∗
c
mn
ϕ
0n
+ h
n
˜g
−n
+ h
n +1
˜g
−n +1
ψ
0,m
∗
+
δ 2
−m
+ h
n
˜g
−n −2
+ h
n +1
˜g
−n −1
+ h
n +2
˜g
−n
+ h
n +3
˜g
−n +1
ψ
0,m
∗
+
δ 2
−m+1
+ . . . + h
n
˜g
−n −2m+2
+ h
n +1
˜g
−n −2m+1
+ . . . + h
n +2m−2
˜g
−n +1
ψ
0,m
∗
+
δ 2
−1
+ X
n≥m
∗
+
δ 2
d
mn
ψ
0n
, 70
for some c
mn
and d
mn
. The coefficients of the functions ϕ
0,k
∗
−m
, . . . , ϕ
0,k
∗
−1
can be written as
X
n∈
h
n
˜h
n−2k
= δ
0k
, 71
with k =
n − ˜
n
1
+1 2
, . . . ,
n − ˜
n
1
+1 2
+ m − 1 = −r, · · · , −r + m − 1, respectively. Similarly, the coefficients of the functions ψ
0,m
∗
+
δ 2
−m
, . . . , ψ
0,m
∗
+
δ 2
−1
can be written as X
n∈
h
n−2k
˜g
n
= 0 , with k
= −n ,
· · · , −n − m + 1, respectively. Observe now that m ≤ m
∗
− e = r +
δ 2
. If m
≤ r, all indices k in 71 are negative, so h
n
ϕ
1,k−2m
+ h
n +1
ϕ
1,k−2m+1 [0,+∞
∈ −
h
n +2
ϕ
1,k−2m+2
+ h
n +3
ϕ
1,k−2m+3
+ . . . + h
n +2m−2
ϕ
1,k−2
+ h
n +2m−1
ϕ
1,k−1 [0,+∞
+ V
I
⊕ W
I
144 L. Levaggi – A. Tabacco
and the result is proven by induction since h
n
6= 0. If m r i.e., δ ≥ 2, we get h
n
ϕ
1,k−2m
+ h
n +1
ϕ
1,k−2m+1 [0,+∞
∈ −
h
n +2
ϕ
1,k−2m+2
+ h
n +3
ϕ
1,k−2m+3
+ . . . + h
n +2m−2
ϕ
1,k−2
+ h
n +2m−1
ϕ
1,k−1 [0,+∞
+2
1 p
−
1 2
ϕ
0,k
∗
−m+r [0,+∞
+ V
I
⊕ W
I
. Therefore, by the induction hypothesis, we only have to prove that
ϕ
0,k
∗
−m+r [0,+∞
∈ S
m
⊕ V
I
⊕ W
I
. We immediately get m
− r ≤ m
∗
− e − r =
δ 2
, so we show that ϕ
0,k
∗
−l [0,+∞
∈ S
l
, 1
≤ l ≤ δ
2 ,
72 by induction on l. If l
= 1, from 69 we have ϕ
0,k
∗
−1 [0,+∞
= 2
1 2
−
1 p
˜h
˜ n
1
ϕ
1,k−1 [0,+∞
+ X
n≥k
∗
c
1n
ϕ
0n [0,+∞
+ X
n≥m
∗
+
δ 2
−1
d
1n
ψ
0n [0,+∞
∈ S
1
⊕ V
I
⊕ W
I
for some c
1n
and d
1n
. If l 1, using induction, we similarly get ϕ
0,k
∗
−l [0,+∞
= 2
1 2
−
1 p
˜h
˜ n
1
ϕ
1,k−2l+1 [0,+∞
+ X
n≥k
∗
−l+1
c
l,n
ϕ
0n [0,+∞
+ X
n≥m
∗
+
δ 2
−l
d
l,n
ψ
0n [0,+∞
∈ S
l
⊕ V
I
⊕ W
I
again c
ln
and d
ln
are fixed coefficients. Thus we have proven 72, and this completes the proof.
Using this result, setting ψ
0,m
∗
−l
: = ϕ
1,k−2l+1 [0,+∞
− P ϕ
1,k−2l+1 [0,+∞
, l
= 1, . . . , m
∗
, 73
and W
B
= {ψ
0m
| m = 0, . . . , m
∗
− 1} , 74
we get W
+
= W
B
⊕ W
I
, 75
Wavelets on the interval 145
with W
I
as defined in 63. With the same process we can build e
W
+
= e W
B
⊕ e W
I
. As for the scaling functions, we must find a couple of biorthogonal bases for the spaces W
+
and e W
+
. To fix notations, we suppose that e
m
∗
≤ m
∗
otherwise we only have to exchange e
m
∗
and m
∗
in what follows. From the biorthogonality properties on the real line we have hψ
0m
, e ψ
0n
i = δ
mn
, ∀m, n ≥ m
∗
. Note, however, that the modified wavelets we have defined are no longer orthogonal to the inte-
rior ones. In fact, from definition 73, it follows that hψ
0m
, e ψ
0n
i = ϕ
1,k−2m
∗
−m+1
, e ψ
0n
, m
= 0, . . . , m
∗
− 1 , n ≥ m
∗
. Using the refinement equation for wavelets on the real line, we easily show that
hψ
0m
, e ψ
0n
i = 0, n
= 0, . . . , e m
∗
− 1 if m ≥ ˜k + ˜n
1
− 1 2
=: m
∗
hψ
0m
, e ψ
0n
i = 0, m
= 0, . . . , m
∗
− 1 ifn ≥ k
+ n
1
− 1 2
=: e m
∗
. 76
Observing that m
∗
≥ e m
∗
, it is sufficient to find two m
∗
×m
∗
matrices E = e
mr
and e E
= ˜e
ns
such that
m
∗
−1
X
r=0
e
mr
ψ
0r
,
m
∗
−1
X
s=0
˜e
ns
e ψ
0s
+ = δ
mn
, ∀m, n = 0, . . . , m
∗
− 1 . Calling Y the m
∗
× m
∗
matrix of components Y
mn
= hψ
0m
, e ψ
0n
i, this condition is equivalent to E Y e
E
T
= I . Again, it is enough to prove that the matrix Y is non-singular. In fact this follows from the
assumed invertibility of the matrix X defined in 45; since det Y
6= 0 iff W
+
∩ e W
+ ⊥
= {0} , we immediately get the result observing that
W
+
∩ e W
+ ⊥
⊂ V
1 +
∩ e V
⊥ 1
= {0} . Moreover
W
+
⊂ e V
⊥
; indeed, for any v
∈ L
p +
, we have hv − P
v, e
ϕ
0k
i = ˘v
0k
− X
l≥0
˘v
0l
hϕ
0l
, e
ϕ
0k
i = 0 . Finally, for any j
∈
✁
, we set W
j +
= T
j
W
+
and e
W
j +
= e T
j
e W
+
; 77
146 L. Levaggi – A. Tabacco
setting ψ
j m
= T
j
ψ
0m
, e ψ
j m
= e T
j
e ψ
0m
for every j, m ∈
✁
, it is easy to check that the biorthog- onality relations
hψ
j m
, e ψ
j
′
n
i = δ
j j
′
δ
mn
, ∀ j, j
′
, m, n
≥ 0 , hold. Moreover, with a proof similar to the one of Proposition 3, one has
P
ROPOSITION
10. The bases 9
j
= {ψ
j m
: m ∈
✁
} of W
j +
, are uniformly p-stable bases and the bases e
9
j
= e
ψ
j m
: m ∈
✁
of e W
j +
, are uniformly p
′
-stable bases for all j
≥ 0. Moreover, let us state a characterization theorem for Besov spaces based on the biorthogonal
multilevel decomposition V
j +
, e V
j j ≥0
as described in Section 3. With the notation of the Introduction, for 1 p, q
+∞, let us set X
s pq
: =
B
s pq
+
if s
≥ 0 , B
−s p
′
q
′
+ ′
if s 0 ,
and denote by
+
the space of distributions. T
HEOREM
3. Let ϕ ∈ B
s pq
+
for some s 0, 1 p, q
+∞. For all s ∈ S := − mins
, L, mins
, L
\ {0}, the following characterization holds:
X
s pq
=
n v
∈ L
p +
: P
j ≥0
2
sq j
P
k≥0
|v
j k
|
p q p
+∞ o
if s 0 ,
n v
∈
+
: v ∈ X
s pq
, for some s
∈ S and P
j ≥0
2
sq j
P
k≥0
|v
j k
|
p q p
+∞ o
if s 0 ,
where v
j k
= ˆv
j k
= hv, e ψ
j k
i if
s 0 , ˆev
j k
= hv, ψ
j k
i if s 0 .
In addition, for all s ∈ S and v ∈ X
s pq
, we have |v|
X
s pq
≍
X
j ≥0
2
sq j
X
k≥0
|v
j k
|
p
q p
1q
. 78
Finally, if p = q = 2, the characterization and the norm equivalence hold for all index s ∈
− mins ,
L, mins ,
L. Proof. It is sufficient to apply Theorems 1 and 2 since the Bernstein- and Jackson- type in-
equalities have been proven in Proposition 6 and 7, respectively and remember the interpolation result 1.
R
EMARK
4. It is possible to obtain a characterization result using the same representation of a function v
∈ X
s pq
for both positive and negative s. Indeed, if ϕ ∈ B
s pq
, e
ϕ ∈ B
˜s pq
, given any s
∈ − min˜s , e
L, mins ,
L \ {0} and v =
P
j,k≥0
ˆv
j k
ψ
j k
∈ X
s pq
, we have the same norm equivalence as in 78 with
ˆv
j k
instead of v
j k
see, e.g., [14]. We also observe that, in general,
˜s s
and thus Theorem 3 gives characterization for a larger interval.
Wavelets on the interval 147
6. Boundary values of scaling functions and wavelets