Professional Documents
Culture Documents
DOCTORA L T H E S I S
Maria Johansson
Maria Johansson
Department of Mathematics
Luleå University of Technology
971 87 Luleå, Sweden
November 2007
Supervisor
Lars-Erik Persson,
Luleå University of Technology, Sweden
Published 2007
Printed in Sweden by University Printing Office, Luleå
To Hans, Emma and Nils
iv
Abstract
v
vi Abstract
vii
viii Preface
See also [9], [32], [48], [49], [50], [51] and [52].
Acknowledgements
ix
x Acknowledgements
Conventions and notations
The notations R, Z, N stand for the field of real numbers, the group of integers
and the semigroup of natural numbers, respectively. If n ∈ N \{0} we denote
Rn+ := {x = (x1 , ..., xn ) : xi ≥ 0, i = 1, 2, ...n} and R+ := R1+ .
Constants are always positive and denoted by C and may be different
at different places. Throughout this thesis all functions are assumed to be
measurable. Moreover, n ∈ Z+ , 1 ≤ p < ∞, p = p−1 p
(p = ∞ if p = 1),
v (x) and u (x) are weights (positive and measurable functions on Rn ) and
xi
xii Conventions and notations
Contents
Abstract v
Preface vii
Acknowledgements ix
1 Introduction 1
1.1 On Hardy’s original inequalities and its pre-history . . . . . . 1
1.2 Some further developments of the Hardy inequality . . . . . . 4
1.3 On the limiting inequalities of the discrete Hardy inequality . 5
1.4 On the limiting inequalities of the continuous Hardy inequality 6
1.5 On Hardy and Pólya-Knopp type inequalities on the cone of
decreasing functions . . . . . . . . . . . . . . . . . . . . . . . 8
1.6 On the connection between the one condition case and the
two conditions case . . . . . . . . . . . . . . . . . . . . . . . . 10
1.7 On multidimensional Hardy-type inequalities . . . . . . . . . 12
1.8 On the Hardy inequality with measures . . . . . . . . . . . . 13
xiii
xiv Contents
Introduction
and the continuous Hardy inequality: suppose that p > 1 and f is a non-
negative p-integrable function on (0, ∞). Then f is integrable over the in-
terval (0, x) for all x > 0 and
⎛ x ⎞p
∞ p ∞
⎝ 1 f (t)dt⎠ dx ≤ p
f p (x)dx. (1.2)
x p−1
0 0 0
1
2 Carleman type inequalities and Hardy type inequalities
for monotone functions
1. The inequalities (1.1) and (1.2) are the standard forms of Hardy’s
inequalities, which can be found in many textbooks in analysis and have been
highlighted first in the famous book [44] by G.H. Hardy, J.E. Littlewood and
G. Pólya.
2. By restricting (1.2) to the class of step functions it can easily be proved
that (1.2) in fact implies (1.1). This important fact seems to have been first
mentioned by E. Landau (cf. [43, p. 154] and [61, Section 8], see also section
1.3). p
p
3. The constant p−1 in both inequalities (1.1) and (1.2) is sharp in
the sense that it can not be replaced by any smaller number.
4. The inequalities (1.1) and (1.2) imply the following information:
∞
∞
∞ ∞
If f p (x)dx < ∞, then (Hf (x))p dx < ∞, (1.4)
0 0
1
x
where f (x) ≥ 0 and Hf (x) := x f (x) dx is the continuous Hardy operator.
0
Note that (1.3) and (1.4) are sometimes (see e.g. comments at the end
of Hardy’s paper [41]) called the weak forms of (1.1) and (1.2), respectively.
5. The inequalities (1.1) and (1.2) together with remark 3 above im-
ply the important information that the Hardy operators H and h map the
spaces Lp into Lp and lp into lp , respectively (p > 1), and their norms are
equal to p := p−1
p
. Here, as usual, the spaces Lp and lp are the Lebesgue
spaces consisting of all classes of measurable functions f = f (x) on (0, ∞) ,
respective sequences a = {an } of real numbers, such that
⎛∞ ⎞1/p ∞ 1/p
6. The inequalities (1.1) and (1.2) have been generalized and applied in
analysis and in the theory of differential equations. Some of these develop-
ments, generalizations and applications have been described and discussed
Introduction 3
in several papers and books e.g. [44], [62], [63] and [80], devoted only to this
subject.
Let us now present shortly the inequality that in fact was Hardy’s main
source and motivation when he initiated the research that finally led him
to the inequality (1.1). This was another famous inequality, which was dis-
covered by D. Hilbert in the early1900s. Here we present it in the most
basic form: If ∞ 2
m=1 am < ∞ and
∞ 2
n=1 bn < ∞ (am ≥ 0, bn ≥ 0) , then the
∞ ∞
double series n=1 m=1 am+n m bn
converges (the weak form). More precisely,
the inequality
∞
∞
∞ 1/2 ∞ 1/2
am bn
2
≤π am b2n (1.5)
m+n
n=1 m=1 m=1 n=1
holds with π as the sharp constant. In the literature also the following more
general form of (1.5) is referred to as Hilbert’s inequality:
∞
∞
∞ 1/p ∞ 1/p
am bn π
≤ p
am bnp
, (1.6)
m+n sin π
n=1 m=1 p m=1 n=1
the inequalities (1.2) and (1.1), e.g. E. Landau, G. Pólya, M. Riesz and I.
Schur. For example if the results of these persons had been published in
another way it could have changed the situation and it should not be unfair
to call the discrete inequality (1.1) as the Riesz, or the Landau-Riesz or the
Hardy-Landau-Riesz inequality.
• Mapping properties of the Hardy operator with variable limits and its
limiting (geometric mean) operator.
This inequality was first presented in 1922 in [14] by the Swedish mathemati-
cian Torsten Carleman (1892-1942) and it is called Carleman’s inequality.
6 Carleman type inequalities and Hardy type inequalities
for monotone functions
∞
∞ 1/p
√
holds for some finite and positive constant C and for all sequences {ak }∞
k=1
of positive numbers.
where f (t) > 0 and it is sometimes called Knopp’s inequality with reference
to [54] (cf. Remark 2.12). However, it seems that it was G. Pólya who first
discovered this inequality (see Remark 2.3). Therefore we prefer to call it
Pólya-Knopp’s inequality.
In Chapter 2 we prove that (1.9) implies (1.7). We also present some
proofs of (1.9) and thereby some more proofs of (1.7).
Next we mention that in [84] L.-E. Persson and V.D. Stepanov proved a
weight characterization for the inequality (1.14) by first characterizing the
Introduction 7
Hardy inequality
⎛ ∞⎛ x ⎞q ⎞1 ⎛∞ ⎞1
q p
⎝ ⎝ 1
f (t)dt⎠ u(x)dx⎠ ≤ C ⎝ f (x)v(x)dx⎠
p
(1.10)
x
0 0 0
Theorem 1.1. Let 1 < p ≤ q < ∞. Then the Hardy inequality (1.10) holds
for f ≥ 0 if and only if
AP S = AP S (p, q, u, v) := (1.11)
⎛ ⎛ x ⎞q ⎞ 1q ⎛ t ⎞− p1
t
sup ⎝ u(x)x−q ⎝ v(y)1−p dy ⎠ dx⎠ ⎝ v(x)1−p dx⎠ < ∞.
t>0
0 0 0
AP S ≤ C ≤ p AP S. .
Theorem 1.2. Let 0 < p ≤ q < ∞. Then the inequality (1.14) holds for all
f ≥ 0 if and only if
⎛ ⎞ 1q
t
− p1 ⎝
DP S = DP S (p, q, w) := supt w(x)dx⎠ < ∞, (1.12)
t>0
0
where ⎛ ⎞− q
x p
1
w(x) = exp ⎝ ln v(y)dy ⎠ u(x). (1.13)
x
0
holds for decreasing functions f ∗ (t). Recall that the rearrangement of the
Hardy-Littlewood maximal function M f is equivalent to the Hardy operator
of the rearrangement of |f |. To be more precise, if
1
(M f ) (x) = sup |f (z)| dz, x ∈ Rn ,
x∈Q |Q| Q
Introduction 9
where Q is a cube in Rn with sides parallel to the coordinate axes and |Q|
is its Lebesgue measure, then
∗ 1 t ∗
(M f ) (x) ≈ f (s) ds, t > 0.
t 0
Some historical remarks and further developments of this equivalence can be
found in paper [6] (see also [62, p. 92]).
Hence, to prove that
⎝ 1
f (t) dt u (x) dx⎠ ≤ C ⎝ f (x) v (x) dx⎠
p
x 0
0 0
for the cases 0 < p, q ≤ ∞. We include proofs also for the cases (i) p = ∞,
0 < q < ∞, (ii) q = ∞, 1 < p < ∞ and (iii) p = q = ∞. Hence the results
are new also for the case u ≡ 1. Only one condition is necessary for each
case which means that our conditions are different and simpler than other
corresponding conditions in the literature. In our proof we use a technique
of discretization and anti-discretization developed by A Gogatishvili and L.
Pick in [35] and [36], where the opposite embedding was considered.
Other aspects of generalized Lorentz spaces and Hardy type inequalities
for monotone functions have been studied by several authors among them
10 Carleman type inequalities and Hardy type inequalities
for monotone functions
we mention L. Maligranda (see e.g. [26], [45], [57], [56], [68], [69], [70] and
[71]) and for more references see also the Ph.D. Thesis [7] by S. Barza and
the new book [62] by A. Kufner, L. Maligranda and L.-E. Persson.
We also mention that in 1990 E. Sawyer [93] established the following
remarkable duality principle for weighted Lp spaces on the cone of decreasing
functions:
Suppose that 1 < p < ∞. Let g, v be positive measurable functions on
(0, ∞) with v locally integrable. Then
∞
f (x) g (x) dx
sup 0 1 ≈
0≤f ↓ ∞ p (x) v (x) dx p
0 f
1/p ∞
∞ x p x −p
g (x) dx
g (t) dt v (t) dt v (x) dx + ∞0 1/p .
0 0 0
0 v (x) dx
A very nice proof of this principle was presented by V.D. Stepanov in
[99].
⎝ 1 ⎠ ⎝ ⎠
f (t) dt u (x) dx ≤C f (x) v (x) dx
p
(1.15)
x 0
0 0
and
A1 = A1 (p, q, u, v) := (1.17)
Introduction 11
⎛∞ ⎞1 ⎛ t ⎞ 1
q p
⎝ −q ⎠ ⎝ p −p ⎠
sup u(x)x dx x V (x)v(x)dx < ∞,
t>0
t 0
where p = p
p−1 and
t
V (t) = v(x)dx. (1.18)
0
C ≈ max(A0 , A1 ).
⎝ f dλ dμ⎠ ≤ C ⎝ f dλ⎠ ,
p
0
0 0
where λ and μ are measures on R. For the special case when dμ = u (x) dx
and dλ = v (x) dx this result reads (see [97, Corollary 3.4]):
Theorem 1.4. Let 1 < p ≤ q < ∞ and let u (x) and v (x) be weight functions
on (0, ∞). Then the inequality
⎛∞ ⎞1⎛∞ ⎞1
x q q p
He proved the remarkable result that it was necessary and sufficient to have
three different integral conditions to characterize the weights u and v so that
a Hardy-type inequality holds in the normal Lp (u) → Lq (v) situation when
1 < p ≤ q < ∞.
Moreover, recently in her Ph.D. thesis [103] A. Wedestig proved that if
the weight v is of product type, then in fact we can get a weight charac-
terization in the Sawyer situation described above by only using one of his
conditions. This Wedestig result in fact also holds in dimension n ∈ Z+
and the corresponding limit (Pólya-Knopp type) inequalities can be derived.
These ideas were also further developed and complemented in the Ph.D.
thesis [101] by E. Ushakova.
3. In the paper [10], S. Barza, L.-E. Persson and J. Soria studied multi-
dimensional Hardy-type inequalities on the cone of functions decreasing (or
increasing) in each variable. In particular, in their characterization these
authors were guided to introduce the new concept of decreasing sets. More
results of this type can be found in the paper [11] by S. Barza, L.-E. Persson
and V.D. Stepanov. These ideas were fully developed in the Ph.D. thesis [7]
by S. Barza.
In Chapter 4 of this Ph.D. thesis we prove a multidimensional version
of the Sawyer duality principle for radially decreasing functions and with
general weights (see Theorem 4.3). We also state the corresponding result
for radially increasing functions (see Theorem 4.5). A number of applications
and related results are also included.
Introduction 13
∼
where v denotes the absolute continuous part of ν. Moreover, if C is the
least constant for which (1.21) holds, then M ≤ C ≤ p1/p (p )1/p M for
1 < p < ∞ and C = M for p = 1. Moreover, V. Kokilashvili [60] (see also
[62]) in 1979 announced the general result (without a proof there) that for
1 ≤ p ≤ q < ∞ the inequality (1.21) holds if and only if
⎛ ⎞ 1
r ∼
1−p p
1 dv
M K = sup (μ [r, ∞)) ⎝ q dx⎠ < ∞.
r>0 dx
0
It was proved in [33] that there are infinite many conditions (even four scales
of conditions) for characterizing (1.21) for 1 < p ≤ q < ∞. We remark that
in (1.21) three measures are involved namely dt, dμ (x) and dν (x) .
We also consider the following general two weights Hardy type inequality
⎛ ⎛ ⎞q ⎛∞⎞1 ⎞1
q
p
⎜ ⎜ ⎟ ⎟
⎝ ⎝ f dλ⎠ dμ (x)⎠ ≤ C ⎝ f dλ⎠ ,
p
(1.22)
R (−∞,x] 0
14 Carleman type inequalities and Hardy type inequalities
for monotone functions
[0,∞) [0,x] 0
where u, v and w are weights, for monotone functions and for all cases 0 <
p, q < ∞.
Chapter 2
15
16 Carleman type inequalities and Hardy type inequalities
for monotone functions
Finally, we include some facts about Torsten Carleman and his work,
which we have found, by studying [1], [58] and the recent article [67] by L.
Maligranda. This description partly complements the information given by
L. Gårding in his book [30].
k
ai = 1.
i=1
k
x1 +x2 +...+xi
G= e− i
i=1
k
H= e−xi = 1.
i=1
This problem can be solved by using the Lagrange multiplier method. Unfor-
tunately this leads to some technical calculations, which, of course Carleman
carried out in an elegant way. We leave out these calculations here, and only
refer to Carleman’s paper [14], where all the details are presented. The result
is that Mk < e for all k ∈ Z+ . Carleman then showed separately that the
inequality (2.1) is strict when the sum on the left hand side converges.
Remark 2.1. In the same paper [14], Carleman proved that the inequality
(2.1) does not hold in general for any constant c < e, i.e., that the constant
e is sharp.
∞
p p
∞
1
k
p
ai < apk , p > 1. (2.3)
k p−1
k=1 i=1 k=1
1/p
Replace ai with ai and note that by using that x = eln x and the definition
of the derivative we find that
k p
1
1/p
k k
1 1/p 0
ai = exp ln ai − ln ai
k p
i=1 i=1 i=1
k
→ exp D(ln axi ) ( when p → ∞)
i=1 x=0
k
k
= exp axi ln ai / axi
i=1 i=1 x=0
k 1/k
1
k
= exp ln ai = ai
k
i=1 i=1
and
we p see that (2.3) leads to the non-strict inequality (2.1) since
p
p−1 → e when p → ∞. Observe that this method does not automati-
cally prove that we have strict inequality in (2.2) and this has to be proved
separately (see for example our later proofs).
Remark 2.2. G.H. Hardy formulated his inequality (2.3) in 1920 in [42]
and proved it 1925 in [43] but it seems that Carleman did not know about the
inequality (2.3) at this time, since he does not refer to the simple connection
that holds according to the proof above. This is somewhat remarkable since
Carleman worked together with Hardy at that time, see for example their
joint paper [15].
Remark 2.3. The above means that (2.1) may be considered as a limit
inequality for the scale (2.3) of Hardy inequalities. This was pointed out by
G.H. Hardy in 1925 in the paper [43, p. 156], but he pronounced that it was
G. Pólya who made him aware of this interesting fact.
We now present two proofs which are based on variations of the arithmetic-
geometric mean inequality (the AG-inequality).
18 Carleman type inequalities and Hardy type inequalities
for monotone functions
k
ai = ci ci ai ≤ ci ci ai. (2.4)
k
i=1 i=1 i=1 i=1 i=1
(1+i)i
We now choose ci = ii−1
, i = 1, 2, ..., k. Then
k 1/k
ci =k+1 (2.5)
1
1
= ci ai
k (k + 1)
i=1 k=i
∞ ∞ ∞
ci ai
1 i
= = ai 1 + ≤e ai .
i i
i=1 i=1 i=1
The strict inequality holds, since we cannot have equality at the same time in
all terms of the inequality. This can only occur if ci ai = c for some constant
i ∞
1
c > 0 i.e. ai = c 1+i i
i , i = 1, 2, ... but this can not hold since ai is
i=1
i
convergent (note that 1+i i
→ e when i → ∞).
Remark 2.4. This idea of proof was presented by G. Pólya (see [86, p. 249])
but here we have more closely followed the presentation which can be found
in L. Hörmander’s book [47], p. 24.
Remark 2.5. We note that Proof 3 shows that for finite sums the inequality
(2.1) holds even for some constant strictly less than e. More precise, for
N = 1, 2, ...we have
N N
k
√ 1
k
a1 a2 ...ak ≤ 1+ ak .
k
k=1 k=1
k
ai ≤ (k!)−1/k iai . (2.6)
k
i=1 i=1
k
√ −1/k
k
a1 a2 ...ak ≤ (k!) iai
k
k=1 k=1 i=1
∞
k
e
≤ iai
k(k + 1)
k=1 i=1
∞ ∞
1
= e iai
k(k + 1)
i=1 k=1
∞
= e ai .
i=1
Remark 2.6. In the paper [42, p. 77], G.H. Hardy presented essentially
this proof but he also pronounced that it was G. Knopp who pointed out this
proof to him.
Furthermore
m´(x) = log(1/ak ), x ∈ (k − 1, k) , (2.9)
and since m (0) = 0 and m is convex.
m(x) m(x) − m(0) m(k) − m(0)
= ≤ (2.10)
x x k
1
k
m(k)
= = log(1/ai ) for all x ≤ k.
k k
i=1
We now make a substitution and use Hölder’s inequality and (2.8). Then we
find, for every A > 0 and r > 1,
A rA
1 −m(x)/x 1
e dx ≤ e−m(x)/x dx
r r
0 0
A
= e−m(rx)/rx dx
0
A
≤ e−m(x)/rx−((r−1)/r)m (x) dx
0
⎛ A ⎞1/r ⎛ A ⎞ (r−1)
r
≤ ⎝ e−m(x)/x dx⎠ ⎝ e−m (x) dx⎠
0 0
so that
A A
r
−m(x)/x
e dx ≤ r r−1 e−m (x) dx.
0 0
r
We let A → ∞, r → 1+ and note that then r r−1 → e so that
∞ ∞
e−m(x)/x dx ≤ e e−m (x) dx. (2.11)
0 0
respectively
∞ ∞ k
e−m(x)/x dx = e−m(x)/x dx
0 k=1k−1
∞
∞
k 1/k
1
k
1/k
≥ exp − log (1/ai ) = ai .
k
k=1 i=1 k=1 i=1
The non-strict inequality (2.1) follows by using these estimates and (2.11).
The strict inequality follows from the fact that we can not have equality in
(2.10) at the same time for all x and k.
Remark 2.7. This is L. Carleson’s proof (see [16]) and in fact he proved
that the even more general inequality (2.11) holds for every piecewise dif-
ferentiable convex function m(x) on [0, ∞] such that m(0) = 0. In fact,
Carleson formulated and proved his inequality in the following slightly more
general form:
∞ ∞
p −m(x)/x
x e dx ≤ e p+1
xp e−m (x) dx, p > −1. (2.12)
0 0
n
nGn + k (bk − 1) Gk ≤ ak bkk (2.13)
k=1 k=1
which holds for all n = 1, 2, ... and all positive sequences {bk } and where
k 1/k
!
Gk = ai .
i=1
In particular, we see that if
a) bk = 1, k = 1, 2, ... , then
n
Gn ≤ ak = An ,
n
k=1
b) bk = 1 + k1 , k = 1, 2, ... ,then
n n
k
1
nGn + Gk ≤ 1+ ak ,
k
k=1 k=1
which implies that the non-strict inequality (2.1) follows when n → ∞. The
strict inequality can also be proved by using the arguments in the following
proof of the inequality (2.13): We use the elementary inequality
1 + a (x − 1) ≤ xa , x > 0, a > 1, (2.14)
(a simple proof of (2.14) can be obtained by putting α = a1 and replacing x
with xa in the following form of the AG-inequality: xα 11−α ≤ αx+(1 − α) 1).
We now use (2.14) with a = k and x = GGk−1 k
bk (k ≥ 2) to obtain that
k
Gk Gk ak k
1+k bk − 1 ≤ bk = b ,
Gk−1 Gk−1 Gk−1 k
which can be written as
Gk−1 + k (Gk bk − Gk−1 ) ≤ ak bkk . (2.15)
We now prove (2.13) by simple induction. First we observe that G1 = a1 so
that
G1 + (b1 − 1) G1 = a1 b1
i.e. (2.13) holds for n = 1. Assume that the inequality (2.13) holds for
n = N ∈ Z+ . Now us this induction assumption and (2.15) with k = N + 1
and we get that
N +1
(N + 1) GN +1 + k (bk − 1) Gk
k=1
N
= (N + 1) bN +1 GN +1 + k (bk − 1) Gk
k=1
N
= N GN + k (bk − 1) Gk + (N + 1) bN +1 GN +1 − N GN
k=1
N +1
+1
≤ ak bkk + aN +1 bN
N +1 = ak bkk
k=1 k=1
i.e. (2.13) holds even for n = N +1 and, by the induction axiom, we conclude
that (2.13) holds for all n ∈ Z+ .
Carleman’s inequality - history and proofs 23
Remark 2.8. The proof above is somewhat more complicated than the other
proofs but it leads to a better result. In fact, this method of proving inequal-
ities uses a well-known principle, which is sometimes referred to as Redhef-
fer’s recursion principle (see [81]). This principle can also be used to improve
several other classical inequalities.
Let a(n) = {a1 , a2 , ..., an } be a positive sequence (n = 1, 2, ...). We define
the powermeans Mr,n of a(n) in the following way
⎧ 1/r
⎪
⎪ 1 r
n
⎨ n⎪ ak , r = 0,
Mr,n = Mr,n a(n) = k=1 1/n
⎪
⎪ !n
⎪
⎩ ak , r = 0.
k=1
Note that An = M1,n , Gn = M0,n and Hn = M−1,n are the usual arithmetic,
geometric and harmonic means, respectively. We also look at the following
sequence of powermeans:
M r,n = (Mr,1 , Mr,2 , ..., Mr,n ) .
In 1996 B. Mond and J. Pečarić proved the following interesting inequality
between iterative powermeans, see [73]:
Ms,n (M r,n ) ≤ Mr,n (M s,n ) , (2.16)
for every r ≤ s. We have equality if and only if a1 = ... = an . The proof in
[73] is fairly complicated so we do not show it here. But we are convinced
that there is a more elementary proof and we leave that as an open question
to the reader. Our next proof is based on the Mond-Pečaric result.
n
Gk ≤ ai . (2.17)
n k
k=1 k=1 i=1
n
ai ≤ ai , k ≤ n,
i=1 i=1
we find that
n
n
n
Gk ≤ √
n
ak . (2.18)
k=1
n! k=1
24 Carleman type inequalities and Hardy type inequalities
for monotone functions
nn nn−1 n
= < en−1 , i.e., √ < e1−1/n .
n! (n − 1)! n
n!
By combining this inequality with (2.17) we find that
k 1/k
n
n
1−1/n
ai <e ak . (2.19)
k=1 i=1 k=1
This non-strict inequality (2.1) follows when we let n → ∞. The fact that
the inequality actually is strict follows from the fact that equality in (2.18)
only can occur when all ai are equal, but this can not be true under our
∞
assumption that ak is convergent.
k=1
Remark 2.9. More information about how (2.16) can be used to prove and
improve inequalities can be found in the papers [25] and [22].
Remark 2.10. We note that if we, in the proof above, combine (2.17) with
the following variant of the AG-inequality
1/n 1/n
n
1
k
1
ai = (a1 (a1 + a2 ) ... (a1 + a2 + ... + an ))1/n
k n!
k=1 i=1
1/n
1 (na1 + (n − 1) a2 + ... + an )
≤
n! n
we get the following strict improvement of (2.19):
k 1/k n
n
k−1
1−1/n
ai <e 1− ak . (2.20)
n
k=1 i=1 k=1
f (x)dx = ak (2.21)
0 k=1
Carleman’s inequality - history and proofs 25
and
⎛ ⎞ ⎛ x ⎞
∞ x ∞
k+1
1 1
exp ⎝ ln f (t)dt⎠ dx = exp ⎝ ln f (t)dt⎠ dx. (2.22)
x x
0 0 k=0 k 0
k+1
ai = exp ln ai dx (2.24)
k+1
i=1 k i=1
k+1
1
k
x−k
≤ exp ln ai + ln ak+1 ) dx
x x
k i=1
⎛ x ⎞
k+1
1
= exp ⎝ ln f (t)dt⎠ dx.
x
k 0
The crucial inequality in (2.24) depends on the fact that the integrand
is a weighted arithmetic mean of the numbers ln ai , i = 1, 2, ..., k + 1, with
weights x1 , ..., x1 (k − 1 weights) and x−(k−1)
k . Here k − 1 ≤ x ≤ k and since
the sequence is decreasing the mean value is smallest for x = k, i.e., when
all weights = k1 . Now (2.1) follows by combining (2.21) - (2.24).
We now present some simple proofs of (2.2) (and thereby some more
proofs of (2.1)).
x
Proof 8 We note that the function m (x) = − ln f ∗ (t) dt fulfils the condi-
0
tions to use Carleson’s inequality (2.12) (here f ∗ (t) is the decreasing rear-
rangement of the function f ). Therefore, according to (2.12), it holds that
⎛ x ⎞
∞ ∞
⎝ 1 ∗ ⎠
x exp
p
ln f (t) dt dx ≤ e p+1
xp f ∗ (x) dx, (2.25)
x
0 0 0
for every p > −1. Carleson’s argument shows that we in fact have strict
inequality in (2.25) and especially by using (2.25) for p = 0 we get Pólya-
Knopp’s inequality (2.2).
26 Carleman type inequalities and Hardy type inequalities
for monotone functions
Remark 2.11. Carleson did not note this fact explicitly in his paper [16]
since he obviously was mainly interested in giving a simple proof of the in-
equality (2.1).
We now present two other proofs of (2.2) and thereby of (2.1) which, like
Carleson’s proof, only are based on convexity arguments.
We integrate, use (2.26) - (2.28) and change the order of integration and find
that
⎛ x ⎞ ⎛ x ⎞
∞ ∞
1 1
exp ⎝ ln f (t)dt⎠ dx ≤ e− ln x+1 ⎝ tf (t)dt⎠ dx
x x
0 0 0 0
⎛ x ⎞
∞ ∞ ∞ ∞
1 ⎝ 1
=e tf (t)dt⎠ dx = e tf (t) dx = e f (t)dt.
x2 x2
0 0 0 t 0
The strict inequality follows since equality in Jensen’s inequality requires that
∞
tf (t) is constant a.e. but this can not occur since f (x)dx is convergent.
0
Carleman’s inequality - history and proofs 27
Remark 2.12. The proof above is partly related to Knopp’s original idea
(see [59] p. 211). However Knopp worked with the interval [1, x] instead of
[0, x] and hence Jensen’s inequality can not be used. Moreover, Knopp never
wrote out the inequality (2.2) explicitly even if it is sometimes referred in the
literature as this is the case, see for example [44, p. 250] and [72, p. 143].
Remark 2.13. By modifying the proof above we can easily prove even some
weighted versions of (2.2) for instance the following
⎛ x ⎞
∞ ∞
⎝ 1 ⎠ e
exp ln f (t)dt x dx <
p
f (x)xp dx
x 1−p
0 0 0
for every p < 1 which is more general than (2.2), cf. also (2.25).
Proof 10 We first note that if we replace f (t) by f (t)/t in (2.2), then by
(2.27) the left hand side in (2.2) equals
⎛ ⎛ x ⎞⎞
∞ x
1
exp ⎝ ⎝ ln f (t)dt − ln tdt⎠⎠ dx
x
0 0 0
⎛ ⎞
∞ x
1 dx
= e exp ⎝ ln f (t)dt⎠ .
x x
0 0
Thus, (2.2) can equivalently be written in the maybe more natural form
⎛ x ⎞
∞ ∞
1 dx dx
exp ⎝ ln f (t)dt⎠ < 1 f (x) . (2.29)
x x x
0 0 0
In order to prove (2.29) we use the fact that the function f (u) = eu is convex
and Jensen’s inequality :
⎛ x ⎞ ⎛ x ⎞
∞ ∞
1 dx 1
exp ⎝ ln f (t)dt⎠ ≤ ⎝ f (t)dt⎠ dx
x x x2
0 0 0 0
⎛∞ ⎞
∞
1
= f (t) ⎝ dx⎠ dt
x2
0 t
∞
dt
= f (t) .
t
0
Remark 2.14. There are some great similarities between Proof 3 of Carle-
man’s inequality and Proof 10 of Pólya -Knopp’s inequality. After changing
order of summation and integration, respectively, in the concluding calcula-
tions we get
∞
∞
1 1 1 1
= and dx = ,
k (k + 1) i x2 t
k=i t
respectively, and this is crucial for the proofs. We also note that (2.29) does
not follow from Carleson’s inequality (2.12) since p = −1.
1924. He died 1949. Carleman was a remarkable person and there are many
remarks concerning him (see e.g. Professor B. Kjellberg’s interesting and
very personal description in [58, p. 93]).
30 Carleman type inequalities and Hardy type inequalities
for monotone functions
Chapter 3
holds for some finite and positive constant C and for all sequences {ak }∞
k=1
of positive numbers. Moreover, we give upper and lower estimates of the
best constant C in the inequality (the corresponding operator norm).
We have already given some examples of sharpenings of Carleman’s in-
equality for finite sums (see for example remark 2.5 and (2.13), (2.19) and
(2.20)). In this chapter we give some additional sharpenings and generaliza-
tions of the Carleman and Pólya-Knopp inequalities. We start by noting the
following:
Remark 3.1. Proof 9 is of course similar to Proof 10 but it contains the
important information that (2.1) in fact can be equivalently rewritten in the
31
32 Carleman type inequalities and Hardy type inequalities
for monotone functions
form (2.29) with the best constant 1. By using this observation, and by
modifying the proof, we find that, in fact, the following more general theorem
holds (cf. [53], Theorem 4.1):
Theorem 3.1. Let 0 < b ≤ ∞. Let φ be a positive and strictly monotone
convex function on (a, c), −∞ ≤ a < c ≤ ∞. Then
⎛ x ⎞
b b
1 dx x dx
φ⎝ f (t)dt⎠ < φ (f (x)) 1 − (3.1)
x x b x
0 0 0
for every realvalued function on (0, b) such that a < f (x) < c.
Proof. We use Jensen’s inequality, change the order of integration and get
⎛ x ⎞ ⎛ ⎞
b b x
1 dx ⎝ f (t)dt⎠ 1 dx
φ⎝ φ−1 f (t)dt⎠ ≤
x x x2
0 0 0 0
⎛ b ⎞
b
1
= f (t) ⎝ dx⎠ dt
x2
0 t
b
1 1
= f (t) − dt
t b
0
b
t dt
= f (t) 1 − .
b t
0
Remark 3.2. For b = ∞ the inequality (3.1) is strict under the condition
that the integral on the right hand side converges and Theorem 3.1 and the
proof essentially coincide with Theorem 4.1 in [53]. The case described in
Theorem 3.1 has been treated even more generally in [23].
Remark 3.3. By choosing φ (u) = eu and replacing f (x) with ln f (x) we see
that (3.1) becomes (2.29) and thereby the equivalent inequality (2.1) and by
choosing φ (u) = up we find that (3.1) with b = ∞ implies Hardy’s inequality
in the form
⎛ x ⎞p
∞ ∞
⎝ 1 ⎠ dx dx
f (t)dt < f p (x) , p ≥ 1, (3.2)
x x x
0 0 0
Carleman’s inequality - some new sharpenings and gener- 33
alizations
which for the case p > 1 (after some substitutions) can be written in the
usual form (see (1.2))
⎛ x ⎞p
∞ p ∞
⎝ 1 p
g(t)dt⎠ dx < gp (x)dx, p > 1, (3.3)
x p−1
0 0 0
where g (x) = f x1−1/p x−1/p .
Note especially that Hardy’s inequality writ-
ten in the form (3.2) holds even when p = 1 but that the inequality (3.3) then
has no meaning.
Remark 3.4. In particular, Theorem 3.1 implies the following improvements
of Pólya-Knopp’s and Hardy’s inequalities for finite intervals (0, b) , b < ∞:
⎛ x ⎞
b b
1 x
exp ⎝ f (t)dt⎠ dx ≤ e f (x) 1 − dx,
x b
0 0 0
and
⎛ ⎞p
b0 x p b0 (p−1)/p
⎝1 f (t)dt⎠ dx ≤
p
1−
x
dx,
x p−1 b0
0 0 0
respectively, where b0 = bp/(p−1) and g (x) = f x(p−1)/p x−1/p as before.
These inequalities has by the way been proved in the paper [25] (see also
[22]) with a different method which is built on the inequalities between mixed
means (cf. our Proof 7).The idea in this remark is further developed and
applied in [23].
Another interesting question which has been studied is to find general
weighted versions of the inequality (2.2). Partly guided by the development
concerning Hardy type inequalities (see for example the books [63] and [80])
one has asked:
Let 0 < p, q < ∞. Find necessary and sufficient conditions for the weights
(i.e. the positive and measurable functions) u (x) and v (x) so that
⎛ ∞⎛ ⎛ x ⎞⎞q ⎞1/q ⎛∞ ⎞1/p
⎝ ⎝exp ⎝ 1 ln f (t)dt⎠⎠ u (x) dx⎠ ≤ C ⎝ f p (x)u (x) dx⎠
x
0 0 0
(3.4)
holds with a stable estimate of the operator norm (= the smallest constant
so that (3.4) holds).
The following results were recently proved by L.E. Persson and V.D.
Stepanov [84]:
34 Carleman type inequalities and Hardy type inequalities
for monotone functions
Theorem 3.2. Let 0 < p ≤ q < ∞. Then the inequality (3.4) holds if and
only if
⎛ x ⎞1/q
D := supx−1/p ⎝ w(s)ds⎠ < ∞,
x>0
0
where ⎛ ⎛ ⎞⎞q/p
s
1 1
w (s) = ⎝exp ⎝ ln dt⎠⎠ u (s)
s v (t)
0
and
D ≤ C ≤ e1/p D.
Theorem 3.3. Let 0 < q < p < ∞. Then the inequality (3.4) holds if and
only if
⎛ ⎛ ⎞r/p ⎞1/r
∞ x
⎜ 1 ⎟
B := ⎝ ⎝ w(s)ds⎠ w(x)dx⎠ < ∞,
x
0 0
Remark 3.5. These results were proved in [84]. We also refer the reader to
some earlier results of this type which can be found in the papers [46], [73]
and [82]. A further development of Theorem 3.2 and 3.3 can be found in [76]
and the Ph.D thesis by M. Nassyrova [75].
N N
lk k 1 k
Gk + ≤ 1− 1+ ak , (3.5)
k (k + 1) N +1 k
k=1 k=1 k=1
where
[x] " 2
√
#
Gk := k
a1 a2 ...ak and lk := x∗k−i+1 − x∗i .
i=1
Here [x] is the usual integer part of x and {x∗k } is the sequence {xk }
rearranged in decreasing order.
Carleman’s inequality - some new sharpenings and gener- 35
alizations
Remark 3.6. For previous results of this type we also refer to the papers
[2], [3], [4], [83], [104], [105] and the references found there. We note that
k
by using the estimate lk ≥ 0, 1 + k1 < e and letting N → ∞ we get (2.1)
as a special case of (3.5).
Remark 3.7. Refinements of Carleman’s inequality (2.1) with e replaced by
k
1 + k1 has been known since at least 1967 (see [90] and [91] and compare
with our Proof 6). We also note that the factor 1 − Nk+1 in (3.5) means that
the ”usual” sum on the right hand side of the inequality has been replaced
by the equivalent Cesarosum, i.e. we have calculated the arithmetic mean of
partial sums. This mean value is of course strictly less than the ”usual” sum
since the terms are positive.
Remark 3.8. In the paper [104] P. Yan and G. Sun proved that Carleman’s
inequality (2.1) can be improved in the following way:
∞
k 1/k ∞ −1/2
1
ai <e 1+ ak . (3.6)
k + 1/5
k=1 i=1 k=1
k
This result easily follows from (3.5) by estimating the important factor 1 + k1
in the following way:
−1/2
1 k 1
1+ ≤e 1+ (3.7)
k k + c∗
2
where c∗ = 8−e
e2 −4
≈ 0, 1802696 < 1/5. The inequality (3.7) does not hold for
numbers smaller than c∗ . (See [53], Remark 12). This means that by using
(3.5) we see that the Yan-Sun inequality (3.6) actually can be replaced with
the sharper inequality
∞
k 1/k ∞ ∞ −1/2
bk
1
ai + <e 1+ ak .
k (k + 1) k + c∗
k=1 i=1 k=1 k=1
1 k
Remark 3.9. The factor 1 + k has also been of interest in some other
new papers. For example M. Gyllenberg and P. Yan recently proved in the
paper [39] that
∞
1 k an
1+ =e 1−
k n=1
(1 + k)n
where all an are positive and can be calculated recursively. For example
a1 = 12 , a2 = 24
1 1
, a3 = 48 etc. This answers an earlier question raised by
Yang (see [105]).
36 Carleman type inequalities and Hardy type inequalities
for monotone functions
Remark 3.10. We have noted before that Carleson’s inequality (2.12) gives
both (2.1) and (2.2) as special cases. Another inequality with that property
has recently been proved, namely the following see ([53], theorem 3.1):
⎛ ⎞
B x
1
exp ⎝ ln f (t)dM (t)⎠ dM (x)
M (x)
0 0
B
M∗ (x)
+e 1− f (x)dM (x)
M (x)
0
B
M∗ (x)
≤ e 1− f (x)dM (x).
M (B)
0
∞
1 ∞ 1
√ q
p
Remark 3.11. This result was also stated and proved in the recent Ph.D.
thesis [103] of A. Wedestig. However, our proof here is different.
Carleman’s inequality - some new sharpenings and gener- 37
alizations
−1/p
Proof. Assume that (3.9) holds. Let first w1 = 0, and replace ak with ãk dk
in (3.8). Then (3.8) is equivalent to
⎛ k q/k k −q/kp ⎞ 1q ∞ 1
∞
p p
⎝ ãi di bk ⎠ ≤ C ãk
k=1 i=1 i=1 k=1
! −q/kp
k
or, if wk = i=1 di bk ,
⎛ k q/k ⎞ 1q ∞ 1
∞
p
Now if {a∗k }∞
k=1 is decreasing rearrangement of {ãk }∞
k=1 , then obviously
⎛ k q/k ⎞ 1q ⎛ k q/k ⎞ 1q
∞
∞
⎝ ãi wk ⎠ ≤ ⎝ a∗i wk ⎠ . (3.11)
k=1 i=1 k=1 i=1
∞ k
k q
∗1
= exp log ai k wk dx
k=1k−1 i=1
∞ k
q
1
k−1
1
= exp log a∗i + log a∗k wk dx
k k
k=1k−1 i=1
∞ k
q
1
k−1
x − (k − 1)
≤ exp log a∗i + log a∗k wk dx
x x
k=1k−1 i=1
⎡ ⎛ ⎞⎤q
∞ k
x
= ⎣exp ⎝ 1 ln f ∗ (t)dt⎠⎦ w(x)dx
x
k=1k−1 0
⎛⎡ ⎛ x ⎞⎤q ⎞1/q
∞
1
= ⎝ ⎣exp ⎝ ln f ∗ (t)dt⎠⎦ w(x)dx⎠ .
x
0 0
38 Carleman type inequalities and Hardy type inequalities
for monotone functions
C ≤ e1/p D.
= C a∗p
k =C ãpk
k=1 k=1
i.e. (3.10) and thus (3.8) holds whenever (3.13) holds. In fact, for N ∈ Z+ ,
we have
⎛ x ⎞1/q N +1 1/q
Hence
⎛ ⎞1/q
x
supx−1/p ⎝ w(t)dt⎠ (3.15)
x>0
0
⎛ −q/kp ⎞1/q
N +1
k
≤ sup N −1/p ⎝ di bk ⎠ = B1 ,
N >0 k=1 i=1
1/q
≤ max 1, 21/q−1 w1 + C ãpk .
k=2
Therefore, by using (3.9), (3.13), (3.15) and (3.16), we conclude that (3.8)
holds.
On the contrary, assume that (3.8) (and thus (3.10)) holds for all positive
sequences. In particular, let ãk = 1, k = 1, ..., N + 1 and ãk = 0, k > N + 1.
Then the left hand side in (3.10) can be estimated as follows:
⎛ k q/k ⎞1/q ⎛ q/k ⎞1/q N +1 1/q
∞
N +1
k
⎝ ãi wk ⎠ ≥⎝ ãi wk ⎠ = wk .
k=1 i=1 k=1 i=1 k=1
(2N ) wk ≤ C,
k=1
Remark 3.13. In [46] the weighted Carleman’s inequality (3.8) was char-
acterized with another condition than B1 .
40 Carleman type inequalities and Hardy type inequalities
for monotone functions
Chapter 4
(4.1)
where u and v are weights, i.e. positive and locally integrable functions on
Rn . For the case 0 < p ≤ q < ∞ cf. [10] .
If q = 1 < p < ∞, then (4.1) with n = 1 is a variant of the duality
theorem given in [93] (cf. also [99]).
An explicit form of (4.1) for n ∈ Z+ was given in [8], but only when v
is of product type, that is weights of the form v (x) = v (x1 , x2 , ..., xn ) =
v1 (x1 ) v2 (x2 ) ...vn (xn ), vi ≥ 0, i = 1, 2, ..., n.
We say that f : Rn → R is a radially decreasing (increasing) function
if f (x) = f (y) when |x| = |y| and f (x) ≥ f (y) (f (x) ≤ f (y)) when
41
42 Carleman type inequalities and Hardy type inequalities
for monotone functions
|x| < |y| and we write f ↓ r and f ↑ r , respectively (see also [65] for further
explanations and applications of these notions).
In this Chapter we will prove a duality formula of the type (4.1) for
radially decreasing functions and with general weights (see Theorem 4.3).
We also state the corresponding result for radially increasing functions (see
Theorem 4.5). In particular, these results imply that we can describe map-
ping properties of operators defined on cones of such monotone functions.
Moreover, we point out that these results can also be used to describe map-
ping properties of operators between some corresponding general weighted
multidimensional Lebesgue spaces (see Theorem 4.6 and cf. also [99] for the
case n = 1). We illustrate this useful technique for the identity operator (see
Corollary 4.7) and for the Hardy integral operator over the n-dimensional
sphere (see Corollary 4.9).
where B (x) is the ball in Rn centered at the origin and with radius |x|. We
need the following well-known n-dimensional form of Hardy’s inequality (see
[27]):
is satisfied.
Proof. Apply Theorem
4.1 (ii) with W (x) = v(x),
p
U (x) = v 1−p (x) B(x) v(y)dy and q = p. Denote
vn (s) = sn−1 v(sσ)dσ , (4.5)
Σn−1
⎛ −p ⎞ 1
|x| p
1
= sup V p (α) ⎝ vn (s) ds v (x) dx⎠
α>0 |x|≥α 0
−p 1
1
∞ t p
1 1 1
−1 1
= sup 1 V
p (α) V p (α) = 1 < ∞ .
α>0 (p − 1) p (p − 1) p
Therefore, by Theorem 4.1 (ii), (4.4) holds with a constant
1 1
1
c≤ 1 (p ) p p p = p and the proof is complete.
(p −1) p
where −1
I1 = v1p g1
and 1
p
p −p
I2 =
G (t) V (t) v (t) dt
Rn
with V (t) = B(t) v (x) dx and G (t) = B(t) g (x) dx.
A Sawyer duality principle for radially monotone func- 45
tions in Rn
Remark 4.1. Theorem 4.1 for the case n = 1 is just Theorem 1 in [93].
However, our proof below is based on the technique in [99] and our investi-
gation in Section 4.1.
Moreover, we use the test function f (x) = |t|>|x| h(t)dt, where h (t) ≥ 0
(note that f is radially decreasing), Lemma 4.2 and usual duality in Lpv -
space to find that
f (x) g (x) dx
Rn
C (g) = sup 1
Rn f (x) v (x) dx
0≤f ↓ p p
Rn |x|<|t| h (t) dt g (x) dx
≥ sup p 1
h≥0 p
R n |x|<|t| h (t) dt v (x) dx
Rn h (t) g (x) dxdt
|x|<|t|
= sup p 1
h≥0 p
R n |x|<|t| h (t) dt v (x) dx
1 Rn h |x|<|t| g
≥ sup
p h≥0 hp V p v 1−p p1
Rn
hV G
1 n Vv
= sup R v
p h≥0 hV p p1
Rn v v
1
1 G p p
= v = I2 .
p Rn V
) *∞
1 s
= gn (z) dz +
V (s) 0 |t|
∞
1 s
+ 2
s n−1
v (sσ) dσ gn (z) dz ds
|t| V (s) Σn−1 0
∞ ∞ s
1 1
≤ gn (z) dz + 2
vn (s) gn (z) dz ds.
V (∞) 0 |t| V (s) 0
+ ,- . + ,- .
K1 K2
Moreover,
1 p 1
∞ p
p 1
K1p v = gn (z) dz v (x) dx
Rn Rn V (∞) 0
1
∞
1 p
= s n−1
g (sσ) dσds v (x) dx
V (∞) 0 Σn−1 Rn
−1+ p1 − p1
= v1 g1 = v1 g1 = I1 ,
⎛ p ⎞ 1
∞ p
1 s
=⎝ 2
vn (s) gn (z) dz ds v (t) dt⎠
Rn |t| V (s) 0
⎛ p ⎞ p1
∞ s
sn−1 v (sσ)
=⎝ z n−1 g (zδ) dzdδ dσds v⎠
Rn |t| Σn−1 V 2 (s) 0 Σn−1
⎛ p ⎞ 1
p
v (x)
=⎝ g (y) dy v (t) dt⎠
Rn Rn \B(t) V 2 (x) B(x)
1
p
−p
≤ p V (t) g v (t) dt = p I2 .
Rn B(|x|)
The upper bound of C (g) follows by combining the last estimates and the
proof is complete.
Remark 4.2. According to our proof we see that the duality constant C (g)
in (4.6) can in fact be estimated in the following more precise way:
The proof above is self-contained and not depending directly on the one-
dimensional result (only on similar arguments as V.D. Stepanov used when
he proved this case). Here we give another shorter proof where we directly
use the (Sawyer) onedimensional result.
where s, y ∈ (0, ∞) and σ, τ ∈ n−1 . By using that f (sσ) = f (s) since f
is radial, we get:
∞
f (sσ) g (sσ) sn−1 dσds
Rn f g 0 n−1
1 = 1/p
p p ∞
Rn f v 0
f p (sσ) v (sσ) sn−1 dσds
n−1
∞
/ (s) ds
f (s) G
0
= 1/p ,
∞ /
0 f (s) V (s) ds
p
⎛ p ⎞ p1
∞ − 1 ∞ ∞
s /
G (y) dy
p
⎜ 0 ⎟
V/ (s) ds /
G (s) ds +⎝ p V/ (s) ds⎠ := I.
0 0 0 V/ (y) dy
s
0
Moreover,
A Sawyer duality principle for radially monotone func- 49
tions in Rn
− 1
∞ p ∞
I = v (sσ) s n−1
dσ ds g (sσ) s n−1
dσds
0 n−1 0 n−1
⎛ p ⎞ p1
∞
s g (yτ ) y n−1 dτ dy
⎜ 0 n−1 ⎟
+⎝ p v (sσ) sn−1 dσds⎠
0 s v (yτ ) y n−1 dτ dy n−1
0 n−1
− p1
= v (t) dt g (t) dt
Rn Rn
⎛ p ⎞ p1
⎜ B(t) g (x) dx ⎟
+⎝ p v (t) dt⎠ (4.11)
Rn
B(t) v (x) dx
Proof. We now use Theorem 4.1 (i) (instead of (ii) as in the proof of Lemma
4.2) and obtain as in the proof of (4.4):
p
v(x) f (y)dy dx ≤ p f p (x)V1p (x) v 1−p (x)dx.
Rn B(x) Rn
By using this estimate the proof follows similarly as the proof of Theorem
4.3 so we leave out the details.
50 Carleman type inequalities and Hardy type inequalities
for monotone functions
Remark 4.3. In fact, similar to Remark 4.2 and Corollary 4.4, we find that
1 1
q p
q
(T f (x)) u (x) dx ≤c f (x)v (x) dx
p
(4.12)
Rn Rn
⎝ −p
∗
T g (y) dy V (x) v (x) dx⎠ (4.13)
Rn B(x)
1
q
q 1−q
≤ c g (x)u (x) dx
Rn
Proof. Assume first that (4.12) holds for all 0 < f ↓ r. Then, by using
Corollary 4.4, duality and Hölder’s inequality, we find that
p
∗ −p n f (x) T ∗ g (x) dx
T g (y) dy V (x) v (x) dx ≤ sup R 1
Rn f (x) v (x) dx
Rn B(x) f ↓r p p
T f (x) g (x) dx
n
= sup R 1
Rn f (x) v (x) dx
f ↓r p p
A Sawyer duality principle for radially monotone func- 51
tions in Rn
1
q 1 −q q
≤ sup 1
f ↓r f p (x) v (x) dx p
R n
1
q
q 1−q
=c g (x) u (x) dx .
Rn
On the contrary assume that (4.13) holds for all g ≥ 0. Then, by using
Corollary 4.4 again, we have that
1
p
q 1−q
pc (g (x)) (u (x)) dx
Rn
⎛ p ⎞ 1
p
⎝ −p
≥p ∗
T g (t) dt V (x) v (x) dx⎠
Rn B(x)
f (x) T ∗ g (x) dx
Rn Rn T f (x) g (x) dx
≥ 1 = 1
Rn f (x) v (x) dx Rn f (x) v (x) dx
p p p p
where
1
g (x) u− q (x)
h (x) = .
1
q q1
−q
Rn g (x) u (x) dx
Remark 4.4. By modifying the proof above we see that a similar dual-
ity result also holds for positive radially
increasing
functions. In fact, in
this case we just need to replace B(x) by Rn \B(x) and V (x) by V1 (x) =
Rn \B(x) v (x) dx in (4.13).
Corollary
4.7. Let 1 < p ≤ q < ∞ and suppose that u, v are weights on Rn
with Rn v = ∞ and V (x) = B(x) v (y) dy.
a) The following conditions are equivalent:
i) The inequality
1 1
q p
q
f u ≤c p
f v (4.15)
Rn Rn
⎛ p ⎞ 1
p
⎝ −p
g (y) dy V1 (x) v (x) dx⎠
Rn Rn \B(x)
1
q
q 1−q
≤ c g (x)u (x) dx
Rn
Proof. a) The equivalence of i) and ii) is just a special case of Theorem 4.6.
Moreover, the fact that ii) and iii) are equivalent follows from Theorem 4.1
A Sawyer duality principle for radially monotone func- 53
tions in Rn
(i) with f replaced by g, q by p , p by q , W by vV −p and U by u1−q . In
fact, then (4.16) is equivalent to (note that (1 − q) (1 − q ) = 1 )
1 1 − 1 1
p
|x|<α v
p q q
−p
sup vV u = sup 1 u < ∞.
α>0 |x|>α |x|<α α>0 (p − 1) p |x|<α
The proof of b) follows similarly by just using Remark 4.4 and Theorem 4.1
(ii).
Remark 4.5. The equivalence of i) and iii) can also be proved using the
technique from [10]. For the case q < p cf. also [11].
Proposition 4.8. Let 1 < p ≤ q < ∞ and suppose that u and v are weights
on Rn with Rn v = ∞ and V (x) = B(x) v (x) dx. If 0 ≤ f is a radially
decreasing function in Rn , then
q 1 1
q
p
f (y) dy u (x) dx ≤c f (x)v (x) dx
p
(4.17)
Rn B(x) Rn
and
1 1
p q
p −p
sup |B (x)| V (x) v (x) dx u(x)dx < ∞.
α>0 B(α) Rn \B(α)
(4.19)
Here, as usual, |B (x)| denotes the Lebesgue measure of the ball with
center at 0 and with radius |x|.
Remark 4.6. For the case n = 1 this result is due to V. Stepanov (see [99],
Theorem 2).
∞ ∞
= t n−1
g (tδ) dδdt dz = gn (t) dt dz
B(x) |z| Σn−1 B(x) |z|
|x| ∞
= sn−1 gn (t) dt dσds
0 Σn−1 s
|x| |x| |x| ∞
= |Σn−1 | sn−1 gn (t) dt ds + gn (t) dtds
0 s 0 |x|
|x| t |x| ∞
= |Σn−1 | s n−1
ds gn (t) dt + |Σn−1 | ds gn (t) dt
0 0 0 |x|
|x| t
= dσ sn−1 ds tn−1 g (tδ) dδ dt
0 Σn−1 0 Σn−1
|x| ∞
+ dσds tn−1 g (tδ) dδdt
0 Σn−1 |x| Σn−1
= |B (y)| g (y) dy + |B (x)| g (y) dy
B(x) Rn \|B(x)|
+ ,- . + ,- .
I1 I2
= I1 (x) + I2 (x) .
obtain that
1
p
p −p
(I1 (x)) V (x) v (x) dx (4.21)
Rn
⎛ p ⎞ 1
p
⎝ −p
= |B (y)| g (y) dy V (x) v (x) dx⎠
Rn B(x)
1
q u (x)1−q q
− 1 1
p
B(α) v (y) dy
q
⎝ −p
= |B (x)| g (y) dy V (x) v (x) dx⎠
Rn \B(α) Rn \B(x)
1
q
q 1−q
≤ c g (x) u (x) dx
Rn
because
1 1
p q
p −p
sup |B (x)| V (x) v (x) dx u (x) dx <∞
α>0 B(α) Rn \B(α)
Now assume that (4.17) holds i.e. that (4.20) holds. Then, in particular,
1 1
p q
p −p q 1−q
(I1 (x)) V (x) v (x) dx ≤c g (x) u (x) dx (4.23)
Rn Rn
and, by using Theorem 4.1 (i) and arguing as above, we find that (4.18)
holds. Moreover, (8.13) holds also with I1 replaced by I2 so that, by using
Theorem 4.1 (ii) and again arguing as in the sufficiently part, we see that
(4.19) holds. The proof is complete.
Remark 4.7. For the case when also the weights are radially decreasing or
increasing some of our results can be written in a more suitable form. Here
we only state the following consequence of Proposition 4.8:
Corollary 4.9. Let 1 < p ≤ q < ∞ and let f (x) and B (x) be positive and
radially decreasing in Rn . Then the Hardy inequality
q 1
q
1
f (y) dy |B (x)|b dx
Rn |B (x)| B(x)
1
p
a
≤ c f (x) |B (x)| dx
p
Rn
Remark 4.8. It is easy to see that Theorem 4.1 is true also if Rn is replaced
by Rn+ or even some more general cone in Rn since the proof essentially
consists of using polar coordinates and carrying over the problem to the (well-
known) one dimensional question. Therefore, by modifying our proofs, we see
that all our results in this chapter indeed holds also when Rn is replaced by
Rn+ or even general cones in Rn as defined in [24].
5.1 Introduction
As mentioned in the introduction E. Sawyer [93] (see Theorem 1.3) proved
the following theorem by performing a general approach for general operators
(In [100] V.D. Stepanov gave a direct proof, which also gave good estimates
of the equivalence constants).
Theorem 5.1. Let 1 < p ≤ q < ∞. Then the Hardy inequality
⎛ ∞⎛ x ⎞q ⎞1 ⎛∞ ⎞1
q p
59
60 Carleman type inequalities and Hardy type inequalities
for monotone functions
and
A1 = A1 (p, q, u, v) := (5.3)
⎛ ⎞1 ⎛ t ⎞ 1
∞ q p
sup ⎝ u(x)x dx⎠ ⎝ x V
−q p −p
(x)v(x)dx⎠ < ∞,
t>0
t 0
where p = p
p−1 and
t
V (t) = v(x)dx. (5.4)
0
C ≈ max(A0 , A1 ).
Theorem 5.2. Let 1 < p ≤ q < ∞. Then the Hardy inequality (5.1) holds
for all f ↓≥ 0 if and only if (5.2) and
A2 = A2 (p, q, u, v) := (5.5)
⎛ y⎛ x ⎞q ⎞1 ⎛ y ⎞− 1
q p
⎝ ⎝ p −p ⎠ −q ⎠ ⎝ p −p ⎠
sup t V (t)v(t)dt u(x)x dx t V (t)v(t)dt <∞
y>0
0 0 0
⎝ f (x)u(x)dx
q ⎠ ≤C ⎝ f (x)v(x)dx
p ⎠ (5.7)
0 0
holds for all f ↓≥ 0 if and only if (5.2) holds. Moreover, the constant C = A0
is the best possible, where A0 is defined by (5.2).
A proof of this Lemma can be found e.g. in [100] (for more references
see e.g. the Ph.D. thesis [7] by S. Barza).
x x ∞ ∞ x
1 1 1
f (t)dt = h(y)dydt = h(y)dy + yh(y)dy. (5.8)
x x x
0 0 t x 0
F := ⎝ ⎝1 f (t)dt⎠ u(x)dx⎠
x
0 0
62 Carleman type inequalities and Hardy type inequalities
for monotone functions
⎛ ∞⎛ ∞ ⎞q ⎞1
x q
⎝ ⎝ 1 ⎠ ⎠
= h(y)dy + yh(y)dy u(x)dx
x
0 x 0
⎛ ⎞1
⎛ ∞⎛ x ⎞q ⎞1
∞ q q
1
≤ ⎝ f (x) u(x)dx
q ⎠ + ⎝ ⎝ yh(y)dy ⎠ u(x)dx ⎠
x
0 0 0
⎛∞ ⎞1
q
:= ⎝ f (x)u(x)dx
q ⎠ + F1 .
0
The first term on the right hand side can be estimated by using Lemma 5.3
and, thus,
⎛∞ ⎞1
p
F ≤ A0 ⎝ f (x)v(x)dx
p ⎠ + F1 . (5.9)
0
y
Moreover, by integrating by parts and setting H(t)dt = G(y), we get that
0
x x x
yH(y) xG(x) G(y) yv(y)
dy = − dy + G(y)dy
V (y) V (x) V (y) V 2 (y)
0 0 0
x
xG(x) yv(y)
≤ + G(y)dy.
V (x) V 2 (y)
0
Moreover,
⎛ ⎞
y x y y
G(y) = h(x) v(t)dtdx = v(t) ⎝ h(x)dx⎠ dt (5.12)
0 0 0 t
⎛ ⎞
y ∞ y
≤ v(t) ⎝ h(x)dx⎠ dt ≤ f (t)v(t)dt.
0 t 0
If we apply Theorem 1.1 to the function f (t) v (t) instead of f (t) and the
q
−q−1
weights V (x) p v (x) xq instead of u (x) and v (x)1−p instead of v (x) we
find that the condition (1.11) becomes
⎛ t ⎛ x ⎞q ⎞
Aq
PS = sup ⎝ V (x)q/p−1−q v(x) ⎝ v (1−p)(1−p ) dy ⎠ dx⎠ ×
t>0
0 0
⎛ t ⎞− pq
⎝ v(x)(1−p)(1−p ) dx⎠
0
⎛ t ⎞⎛ t ⎞− pq
p
= sup ⎝ V (x)q/p−1 v(x))dx⎠ ⎝ v(x)dx⎠ = ,
t>0 q
0 0
64 Carleman type inequalities and Hardy type inequalities
for monotone functions
and, thus,
⎛∞ ⎞q
p
q q
q
F2 ≤ p p A0 ⎝ f (x)v(x)dx
p ⎠ . (5.13)
0
Denoting
yv(y)
G(y) = Φ(y)
V 2 (y)
and
xp V −p (x)v(x) = Ψ(x)1−p (5.14)
and, again applying Theorem 1.1, (5.5) and (5.12), we find that
⎛ ⎞q ⎛∞ ⎞q
∞ x p
q q
F3q = ⎝1 Φ(y)dy ⎠ u(x)dx ≤ p A2 ⎝ Φ (x)Ψ(x)dx⎠
p
x
0 0 0
⎛ ⎞q
∞ p
q
= p Aq2 ⎝ G (x)v(x)V (x)
p −p
dx⎠
0
⎛ ⎛ ⎞p ⎞q
∞ x p
q q
≤ p A2 ⎝ ⎝ f (y)v(y)dy ⎠ −p
v(x)V (x) dx ⎠
0 0
⎛ ⎞q
∞ p
2q
≤ p Aq2 ⎝ f (x) v(x)dx⎠ .
p
(5.15)
0
Here we in the last step apply Theorem 1.1 with q = p, f (x) replaced by
f (x) v (x), v (x) replaced by v 1−p (x) and u (x) replaced by v (x) V −p (x) xp
so that the condition (1.11) AP S = 1. By combining the estimates (5.9),
(5.10), (5.13) and (5.15) we find that (5.1) holds with a constant C satisfying
2
C ≤ 1 + p1/q p A0 + p A2 ),
Applying this function to the right hand side of (5.1) and changing order of
integration we obtain that
⎛∞ ⎞1/p ⎛ y⎛ y ⎞ ⎞1
p
⎝ ⎝ 1
f (s)ds⎠ u(x)dx⎠
x
0 0
⎛ ⎛ ⎛ ⎞ 1 ⎞q ⎞ 1q
∞ x y p
⎜ ⎜1 ⎝ ⎟ ⎟
= ⎝ ⎝ tp V (t)−p −1 v(t)dt⎠ ds⎠ u(x)dx⎠
x
0 0 s
⎛ ⎛ ⎛ ⎞ 1 ⎞q ⎞ 1q
y x y p
⎜ ⎜ ⎟ ⎟
≥ ⎝ ⎝ ⎝ tp V (t)−p −1 v(t)dt⎠ ds⎠ u(x)x−q dx⎠ . (5.17)
0 0 s
⎝ p −p −1
t V (t) v(t)dt⎠ ds
0 s
⎛ ⎞1
x y p
p
⎝ −p −1
≥ s p V (t) v(t)dt⎠ ds
0 s
⎛ ⎛ y ⎞− 1 ⎞
x x p
1 ⎜p
⎝ V (z)−p −1 v(z)dz ⎠ ⎟
≥ sp ⎝ V (t)−p −1 v(t)dt⎠ ds := I
p
0 s t
66 Carleman type inequalities and Hardy type inequalities
for monotone functions
1
x
Hence C ≥ A2
1 . Moreover, since f (x) ≤ x f (t)dt when f is decreasing,
p(p ) p 0
it follows that A0 ≤ C, by choosing ft (x) = χ[0,t] (x) and taking supremum.
Hence also the lower estimate of (5.6) holds and the proof is complete.
Remark 5.1. The crucial part of the proof above is to use the results in
[84], [99] and the technique in [100] combined with finding explicitly the cor-
responding test functions. For the estimates C ≈ max (A0 , A2 ) we have found
that
1 2
1/q
max A0 , A2 ≤ C ≤ 1 + p p A0 + p A2 ) (5.19)
p (p )1/p
but we strongly believe that these estimates can be improved.
Theorem 5.4. Let 0 < p ≤ q < ∞. Let u (x) and v (x) be weight functions
and assume that v (x) is decreasing. Then the inequality
⎛ ∞⎛ ⎛ x ⎞⎞q ⎞1 ⎛∞ ⎞1
q p
⎝ ⎝exp ⎝ 1
ln f (t)dt⎠⎠ u(x)dx⎠ ≤ C ⎝ f (x)v(x)dx⎠ (5.20)
p
x
0 0 0
where p 2 pr
k (r) = 1 + r qr r + r ,
1
Proof. Assume that (5.21) holds and let g(x) = f (x)v(x) p . Then as v(x) is
decreasing and f (x) is decreasing so is g(x). Moreover, if w(x) be defined as
in (1.13), then (5.20) is equivalent to
⎛ ⎛ ⎛ ⎞⎞q ⎞1 ⎛ ⎞1
∞ x q ∞ p
qr
Note that if r > 1 and 1 < p ≤ q < ∞, then we have 1 < r ≤ p < ∞. By
applying Jensen’s inequality we find that if the inequality
⎛ ⎛ ⎞ qr ⎞1 ⎛∞ ⎞1
∞ x p
q
p
⎜ ⎝ qp ⎟
⎝ h(t)dt⎠ w(x)x s dx⎠ ≤ C ⎝ h (x)dx⎠
− r
(5.25)
0 0 0
68 Carleman type inequalities and Hardy type inequalities
for monotone functions
holds for all decreasing functions h (x) , then also (5.24) holds with the same
constant C. Now we replace p by r and q by qr p , u (x) with w(x) and let
v(x) = 1 in the inequality (5.1) and apply Theorem 5.2. This means that
the inequality (5.25) holds if and only if
⎛ t ⎞ 1q
r
qr p
− 1
A0 (r, , w, 1) = DP S (p, q, w) = supt p ⎝ w(x)dx⎠ < ∞,
p t>0
0
and
⎛ t ⎞ 1q
r
qr p 1
A2 (r, , w, 1) = DP S (p, q, w) = supt− p ⎝ w(x)dx⎠ < ∞,
p t>0
0
Remark 5.2. We have obtained the same condition for characterizing the
Pólya-Knopp inequality for decreasing functions as for all positive functions
in [84] but our upper estimate of the best constant C in (5.22) is different.
and
⎛ ⎛ ⎞p ⎞ 1 ⎛∞ ⎞ 1
∞ ∞ p
q
⎜ ⎝ f (t) ⎠ p −p ⎟ ⎝ q 1−q ⎠
⎝ dt x V (x) v(x)dx⎠ ≤ C f (x)u(x) dx ,
t
0 x 0
(5.27)
where V is defined by (5.4). Then, by using the Hardy inequality and the
so called Muckenhoupt condition to characterize the inequalities (5.26) and
(5.27), E. Sawyer received the before mentioned result in Theorem 5.1.
Recently the weighted Hardy inequality has been characterized with some
new conditions (see e.g. [64] and [102]). More generally, nowadays we know
that the Hardy inequality (5.1) for 1 < p ≤ q < ∞ holds for all f ≥ 0 if
and only if just one of the following (infinite many equivalent) conditions is
satisfied (see [33]):
Theorem 5.5. Let 1 < p ≤ q < ∞, 0 < s < ∞ and define
⎛ ⎞1
⎛ t ⎞q 1 −s q ⎛ ⎞s
∞ p x
⎜ ⎟
A1 (s) := sup ⎜ dt⎟
⎝ u (t) ⎝ v 1−p (y)dy ⎠ ⎠
⎝ v 1−p (t)dt⎠ ,
0<x<∞
x 0 0
⎛ ⎞1
⎛ t ⎞q 1
+s q ⎛ ⎞−s
x p x
⎜ ⎟
A2 (s) := sup ⎜
⎝ u (t) ⎝ v 1−p (y)dy ⎠ dt⎟
⎠
⎝ v 1−p (t)dt⎠ ,
0<x<∞
0 0 0
⎛ ⎞ 1
⎛∞ ⎞p 1 −s p ⎛ ⎞s
x q ∞
⎜ ⎟
A3 (s) := sup ⎜ dt⎟
⎝ v 1−p (t) ⎝ u(y)dy ⎠ ⎠
⎝ u(y)dy ⎠ ,
0<x<∞
0 t t
⎛ ⎞ 1
⎛ ⎞p 1
+s p ⎛∞ ⎞−s
∞ ∞ q
⎜ ⎟
A4 (s) := sup ⎜ dt⎟
⎝ v 1−p (t) ⎝ u(y)dy ⎠ ⎠
⎝ u(y)dy ⎠ .
0<x<∞
x t t
(5.28)
Then the Hardy inequality
⎛ ∞⎛ x ⎞q ⎞1 ⎛∞ ⎞1
q p
holds for all measurable functions f ≥ 0 if and only if any of the quantities
Ai (s) is finite. Moreover, for the best constant C in (5.29) we have C ≈
Ai (s), i = 1, 2, 3, 4.
This gives us the possibility to characterize the weighted Hardy inequal-
ity for decreasing functions with some new conditions, by just using the
technique of E. Sawyer described above. We can therefore formulate the
following generalization of our Theorem 5.2:
Theorem 5.6. Let 1 < p ≤ q < ∞. Then the inequality (5.1) holds for all
f ↓≥ 0 if and only if for any s, r, 0 < s, r < ∞, one of the quantities
⎛ ⎞ 1
⎛ t ⎞p 1q −s p ⎛ ⎞s
∞ x
⎜ −p
⎟
B1 (s) := sup ⎜⎝ V (t) v (t)
⎝ u(y)dy ⎠ dt⎟
⎠
⎝ u(t)dt⎠ ,
0<x<∞
x 0 0
⎛ ⎞ 1
⎛ t ⎞p 1q +s p ⎛ ⎞−s
x x
⎜ ⎟
B2 (s) := sup ⎜ −p
⎝ V (t) v (t)
⎝ u(y)dy ⎠ dt⎟
⎠
⎝ u(t)dt⎠ ,
0<x<∞
0 0 0
⎛ ⎞1
⎛∞ ⎞q 1 −s q
x p
⎜ ⎟
B3 (s) := sup ⎜⎝ u(t) ⎝ V (y)−p v (y) dy ⎠ dt⎟
⎠ ×
0<x<∞
0 t
⎛ ⎞s
∞
⎝ −p
V (y) v (y) dy ⎠ ,
t
⎛ ⎞1
⎛∞ ⎞q 1 +s q
∞ p
⎜ ⎟
B4 (s) := sup ⎜⎝ u(t) ⎝ V (y)−p v (y) dy ⎠ dt⎟
⎠ ×
0<x<∞
x t
⎛ ⎞−s
∞
⎝ V (y)−p v (y) dy ⎠
t
and one of the quantities
⎛ ⎞1
⎛ t ⎞q 1 −r q ⎛ ⎞r
∞ p x
⎜ ⎟
E1 (r) := sup ⎜ dt⎟
−q ⎝
Ψ1−p (y)dy ⎠ ⎝ Ψ1−p (t)dt⎠ ,
0<x<∞
⎝ u (t) t ⎠
x 0 0
A new characterization of the Hardy inequality and its 71
limit Pólya-Knopp inequality for decreasing functions
⎛ ⎞1
⎛ t ⎞q 1 +r q
x p
⎜ ⎟
E2 (r) := sup ⎜
⎝ u (t) t−q ⎝ Ψ1−p (y)dy ⎠ dt⎟
⎠ ×
0<x<∞
0 0
⎛ x ⎞−r
⎝ Ψ1−p (t)dt⎠ ,
0
⎛ ⎞ 1
⎛ ⎞p 1
−r p ⎛ ⎞r
x ∞ q ∞
⎜ ⎟
E3 (r) := sup ⎜⎝ Ψ 1−p
(t) ⎝ u(y)y −q dy ⎠ dt⎟
⎠
⎝ u(y)y −q dy ⎠ ,
0<x<∞
0 t t
⎛ ⎞ 1
⎛∞ ⎞p 1 +r p
∞ q
⎜ ⎟
E4 (r) := sup ⎜ dt⎟
⎝ Ψ
1−p
(t) ⎝ u(y)y −q dy ⎠ ⎠ ×
0<x<∞
x t
⎛∞ ⎞−r
⎝ u(y)y −q dy ⎠
t
where Ψ(x) is defined by (5.14) so we can again use Theorem 5.5 now with
s, u (x) and v (x) replaced by r, u (x) x−q and Ψ(x) respectively, and we get
the second bunch of conditions in Theorem 5.6. The proof is complete.
Remark 5.4. By using Theorem 5.6 and arguing as in the proof of Theo-
rem 5.4 we can get a number of other characterizations of the Pólya-Knopp
inequality (5.20) then that is stated in Theorem 5.4.
Chapter 6
6.1 Introduction
Let us first recall that in the Sawyer characterization of Hardy’s inequality
for decreasing functions two independent conditions were used (see Theorem
1.3 = Theorem 5.1 and c.f. also our alternative descriptions in Theorem 5.2
and 5.6).
On the contrary, in the next result by G. Sinnamon [97] only one condi-
tion is necessary to characterize the corresponding Hardy type inequality.
Theorem 6.1. Let 1 < p ≤ q < ∞ and let u (x) and v (x) be weight functions
on (0, ∞). Then the inequality
⎛ ∞⎛ x ⎞q ⎞1 ⎛∞ ⎞1
q p
73
74 Carleman type inequalities and Hardy type inequalities
for monotone functions
Remark 6.1. The main argument of Sinnamon was to prove (via so called
level functions) that (6.1) is in fact equivalent to that
⎛ ⎞1 ⎛ ⎞1
∞ q ∞ p
holds for all decreasing f ≥ 0 and use the fact that a characterization for such
continuous embeddings between Lorentz spaces is known. (Here, as usual,
x
V (x) = v (t) dt.)
0
At a first glance the two inequalities (5.1) and (6.1) seems to be similar,
but to characterize (5.1) we need two conditions while (6.1) can be charac-
terized by only one condition. Moreover, in [93] E. Sawyer gives an example
for the fact that neither (5.2) nor (5.3) is sufficient for (5.1) to hold namely
the following:
Example: If u (x) = xq−1 and v (x) = xp−1 , then (5.2) holds but not
(5.3), while if u (x) = xq (x + 1)−q−1 and v (x) = χ(0,1) (x) , then (5.3) holds
but not (5.2).
Let us now look at another example:
Example: If v(x) ≡ 1, then
⎛ ⎞ 1q
t 1
−p
A0 = sup ⎝ u(x)dx⎠ t
0<t<∞
0
and ⎛∞ ⎞1
q
1
p
A1 = sup ⎝ x u(x)dx⎠ t .
−q
0<t<∞
t
We also note that if u (x) is replaced by u (x) x−q in (6.1)-(6.2) and v (x) ≡ 1,
then (5.1) coincides with (6.1) and (5.2) coincides with (6.2). In particular,
this means that in this case the conditions A0 < ∞ and A1 < ∞ must be
equivalent. This can be proved directly but it is also a consequence of the
general equivalence result proved in [33] (see Theorem 5.5 and Remark 6.2):
A Unified approach to Sawyer and Sinnamon characteri- 75
zations of the Hardy inequality for decreasing functions
Remark 6.2. The equivalence of the conditions A0 < ∞ and A1 < ∞ for
v (x) ≡ 1 follows by using Theorem 5.5 with u (x) = u (x) x−q and A1 with
s = p1 and A2 with s = 1p .
In Section 6.2 we state and prove the main result of this Chapter (The-
orem 6.2), which generalize the above mentioned result of Sinnamon (see
Corollary 6.3) and also put some new light on the result of Sawyer for the
case when the weight v (x) is increasing. Section 6.3 is reserved for some
concluding remarks.
Theorem 6.2. Let 1 < p ≤ q < ∞ and let u (x) , v (x) and ϕ (x) be weight
functions on (0, ∞), where ϕ (x) is decreasing. Then the inequality
⎛ ∞⎛ x ⎞q ⎞1 ⎛∞ ⎞1
q p
⎝ ⎝ f (t) ϕ (t) v (t) dt⎠ u (x) dx⎠ ≤ C ⎝ f p (x) v (x) dx⎠ (6.4)
0 0 0
holds for all C < ∞ and decreasing f if and only if one of the following
conditions holds for some s > 0:
⎛ ⎞1
⎛ ⎞q 1
−s q
∞ x p
⎜ ⎟
Dϕ (s) := sup ⎜ dx⎟
⎝ u (x) ⎝ v (y) ϕp (y) dy ⎠ ⎠ × (6.5)
t>0
t 0
⎛ ⎞s
t
⎝ v (x) ϕp (x) dx⎠ < ∞.
0
⎛ ⎞ 1
⎛∞ ⎞p 1
−s p
t q
⎜ ⎟
Dϕ∗ (s) := sup ⎜ dx⎟
⎝ v (x) ϕp (x) ⎝ u (y) dy ⎠ ⎠ × (6.6)
t>0
0 x
⎛ ⎞s
∞
⎝ u (x) dx⎠ < ∞.
t
76 Carleman type inequalities and Hardy type inequalities
for monotone functions
⎛ ⎞1
⎛ x ⎞q 1 +s q
t p
⎜ ⎟
Eϕ (s) := sup ⎜
⎝ u (x)
⎝ v (y) ϕp (y) dy ⎠ dx⎟
⎠ × (6.7)
t>0
0 0
⎛ ⎞−s
t
⎝ p
v (x) ϕ (x) dx⎠ < ∞.
0
⎛ ⎞ 1
⎛∞ ⎞p 1 +s p
∞ q
⎜ ⎟
Eϕ∗ (s) := sup ⎜ dx⎟
⎝ v (x) ϕp (x) ⎝ u (y) dy ⎠ ⎠ × (6.8)
t>0
t x
⎛ ⎞−s
∞
⎝ u (x) dx⎠ < ∞.
t
Proof. First we use Theorem 5.5 with u (x), v (x) = ϕ−p (x) to conclude
that, for any s > 0, the conditions (6.5) - (6.8) are equivalent. In particular,
now we assume that (6.5) holds for some s, 0 < s < p1 . Since f is decreasing
we can write ∞
f p (x) = h (y) dy, where h ≥ 0.
x
Thus, by using the Fubini theorem, we find that (6.4) can equivalently be
written as
⎛ ⎛ ⎛ ⎞1 ⎞q ⎞ 1q
∞ x ∞ p
⎜ ⎜ ⎝ ⎟ ⎟
⎝ ⎝ h (y) dy ⎠ v (t) ϕ (t) dt⎠ u (x) dx⎠ (6.9)
0 0 t
⎛ ⎛ ⎞ ⎞1
∞ y p
Now, we define
x 1
p 1 p
Ṽϕ (t) = v (t) ϕ (t) dt and C0 = .
1 − sp
0
A Unified approach to Sawyer and Sinnamon characteri- 77
zations of the Hardy inequality for decreasing functions
⎛ ∞ q ⎞1
x ∞ 1 q
p 1 1
= ⎝ h (y) dy ϕ (t) Ṽϕ−s (t) Ṽϕs (t) v p (t) v p (t) dt u (x) dx⎠
0 t
0
⎛ q
∞ x ∞ p
≤⎝ −sp
h (y) dy Ṽϕ (t) v (t) ×
0 t
0
q 1
x q
p
−sp p
Ṽϕ (t) ϕ (t) v(t)dt u (x) dx
0
⎛∞ ⎞1
x ∞ q
1
q
p q p
−s
≤ C0 ⎝ h (y) dy Ṽϕsp (t) v (t) dt Ṽϕ (x) u (x) dx⎠
0 t
0
⎛ ⎛∞ ⎞ p ⎞ p1
∞ ∞
1
q
⎜ q p −s ⎟
≤ C0 ⎝ h (y) dy Ṽϕ (t) v (t) ⎝ Ṽϕ
sp
(x) u (x) dx⎠ dt⎠
t
0 t
⎛ ⎛ ⎛∞ ⎞ p ⎞ ⎞ p1
∞ q
y 1
⎜ ⎜ q p −s ⎟ ⎟
≤ C0 ⎝ h (y) ⎝ Ṽϕ (t) v (t) ⎝ Ṽϕ
sp
(x) u (x) dx⎠ dt⎠ dy ⎠
0
0 t
⎛ ⎞1 ⎛ ∞ ⎞1
∞
1
q y p
q p
−s
≤ C0 supṼϕs (t) ⎝ Ṽϕ (x) u (x) dx ⎠ ⎝ h (y) v (t) dt dy ⎠
t≥0 0
t 0
⎛ ⎞1
1 ∞ p
1 p y
= D1 (s) ⎝ h (y) v (t) dt dy ⎠
1 − sp 0
0
⎛ ⎞1
1 ∞ p
1 p
= Dϕ (s) ⎝ f (t) v (t) dt⎠ .
p
1 − sp
0
78 Carleman type inequalities and Hardy type inequalities
for monotone functions
where t > 0 is fixed. Then the integral on the right hand side of (6.4) is
equal to
⎛ t
p
⎝ p
= Ṽϕ−(1+ps) (t) ϕp (x) v (x) dx (6.10)
p (1 − s) − 1
0
⎞1
∞ p
+ Ṽϕ−(1+ps) (x) ϕp (t) v (x) dx⎠
t
p 1
p −ps 1 −ps p
= Ṽϕ (t) + Ṽϕ (t)
p (1 − s) − 1 ps
p 1
p 1 p −s
= + Ṽϕ (t) .
p (1 − s) − 1 ps
Moreover, the left hand side of (6.4) is greater than
⎛ ∞⎛ x ⎞q ⎞1
q )
*
⎝ ⎝ f (t) ϕ (t) v (t) dt⎠ u (x) dx⎠ = since p + 1 = p
p
t 0
⎛
⎛
∞ t
1
p − +s
=⎝ ⎝
p
Ṽϕ (t) ϕp (y) v (y) dy +
p (1 − s) − 1
t 0
⎞q ⎞1
x 1
q
− +s
p
(y) ϕ (y) v (y) dy ⎠ u (x) dx⎠
p
Ṽϕ
t
⎛
∞
1
1
p p
−s p p
−s
=⎝ Ṽϕ (t) + Ṽϕ (x) −
p (1 − s) − 1 p (1 − s) − 1
t
A Unified approach to Sawyer and Sinnamon characteri- 79
zations of the Hardy inequality for decreasing functions
1 q 1
q
p −s
Ṽϕ p (t) u (x) dx
p (1 − s) − 1
⎛∞ ⎞1
q
p q p1 −s
= ⎝ Ṽϕ (x) u (x) dx ⎠ .
p (1 − s) − 1
t
By taking supremum over all t > 0 we find that (6.5) holds for 0 < s < 1p .
Hence, according to Theorem 5.5, all conditions (6.5) - (6.8) holds for all
s > 0. The proof is complete.
holds for some C < ∞ and for all decreasing f ≥ 0 if and only if one of the
following conditions holds for some s > 0 :
⎛ ⎞1
⎛ x ⎞q 1 −s q ⎛ ⎞s
∞ p x
⎜ ⎟
D1 (s) = sup ⎜
⎝ u (x)
⎝ v (y) dy ⎠ dx⎟
⎠
⎝ v (t) dt⎠ < ∞.
t>0
t 0 0
80 Carleman type inequalities and Hardy type inequalities
for monotone functions
⎛ ⎞ 1
⎛∞ ⎞p 1 −s p ⎛ ⎞s
t q ∞
⎜ ⎟
D1∗ (s) = sup ⎜
⎝ v (x)
⎝ u (y) dy ⎠ dx⎟
⎠
⎝ u (x) dx⎠ < ∞.
t>0
0 x t
⎛ ⎞1
⎛ x ⎞q 1 +s q ⎛ ⎞−s
t p x
⎜ ⎟
E1 (s) = sup ⎜
⎝ u (x)
⎝ v (y) dy ⎠ dx⎟
⎠
⎝ v (t) dt⎠ < ∞.
t>0
0 0 0
⎛ ⎞ 1
⎛∞ ⎞p 1 +s p ⎛ ⎞−s
∞ q ∞
⎜ ⎟
E1∗ (s) = sup ⎜
⎝ v (x) ⎝ u (y) dy ⎠ dx⎟
⎠
⎝ u (x) dx⎠ < ∞.
t>0
t x t
1
Moreover, by applying Theorem 6.2 with ϕ (t) = v(t) and u (x) replaced
−q
by u (x) x we obtain the following alternative to the Sawyer result for the
case when v (t) is increasing.
Corollary 6.4. Let 1 < p ≤ q < ∞ and let u (x) and v (x) be weight
functions on (0, ∞), where v (x) is increasing. Then the inequality
⎛ ⎛ ⎞q ⎛∞ ⎞1 ⎞1
∞ x q p
holds for some C < ∞ and for all decreasing f ≥ 0 if and only if one of the
following conditions holds for some s > 0 :
⎛ ⎞1
⎛ ⎞q 1
−s q
∞ x p
⎜ ⎟
F1 (s) = sup ⎜ dx⎟
⎝ u (x) x−q ⎝ v 1−p (y) dy ⎠ ⎠ ×
t>0
t 0
⎛ ⎞s
t
⎝ 1−p
v (x) dx⎠ < ∞.
0
A Unified approach to Sawyer and Sinnamon characteri- 81
zations of the Hardy inequality for decreasing functions
⎛ ⎞ 1
⎛∞ ⎞p 1 −s p
t q
⎜ ⎟
F1∗ (s) = sup ⎜
⎝ v
1−p
(x) ⎝ u (y) y −q dy ⎠ dx⎟
⎠ ×
t>0
0 x
⎛∞ ⎞s
⎝ u (x) x−q dx⎠ < ∞.
t
⎛ ⎞1
⎛ x ⎞q 1
+s q
t p
⎜ ⎟
F2 (s) = sup ⎜ dx⎟
⎝ u (x) x−q ⎝ v 1−p (y) dy ⎠ ⎠ ×
t>0
0 0
⎛ ⎞−s
t
⎝ v 1−p (x) dx⎠ < ∞.
0
⎛ ⎞ 1
⎛ ⎞p 1
+s p
∞ ∞ q
⎜ ⎟
F2∗ (s) = sup ⎜
⎝ v 1−p
(x) ⎝ u (y) y −q dy ⎠ dx⎟
⎠ ×
t>0
t x
⎛ ⎞−s
∞
⎝ u (x) x−q dx⎠ < ∞.
t
Remark 6.6. According to our proof of Theorem 6.2 we see that the suf-
ficient part of the Theorem holds in general i.e. without the additional as-
sumption that ϕ (x) is decreasing.
Remark 6.7. Note that condition F1 (s) < ∞, with 0 < s < p1 coincides
with that obtained by Wedestig in [102] to characterize (6.11) for all functions
f ≥ 0. The explanation of this phenomenon is that among all rearrangements
∞
of the function f the integral f p (x) v (x) dx with v (x) non-decreasing, will
0
be the smallest possible when f is rearranged in decreasing order. It follows
that if v (x) is increasing, then (6.11) for all functions is equivalent to (6.11)
restricted to the cone of decreasing functions.
Remark 6.8. In view of Sinnamon’s result that (6.1) and (6.3) are in fact
equivalent our result also gives directly new results concerning embedding be-
tween Lorentz spaces. More exactly, let f ∗ denote the decreasing rearrange-
ment of function f from the measurespace Ω. Then, for 1 < p ≤ q < ∞, the
inequality
⎛∞ ⎞1 ⎛∞ ⎞1
q p
holds for some finite constant C if and only if any of the conditions in Corol-
lary 6.3 holds for any s > 0 and with u (x) replaced by Vu(x)
q (x) . Except for the
case E1 p1 < ∞ used by Sinnamon we think this fact is not pointed out
before in the literature.
7.1 Introduction
Let M+ be the class consisting of all Borel functions f : [0, ∞) → [0, +∞]
and M ↓ (M ↑) be a subclass of M+ , which consists of all decreasing
(increasing) functions f ∈ M+ . Suppose that λ, μ and ν are positive Borel
σ-finite measures on [0, ∞) and u, v, w ∈ M+ are weight functions.
In this chapter we study for 0 < p, q, s < ∞ the problem when the Hardy
inequality of the form
1 1
q p
83
84 Carleman type inequalities and Hardy type inequalities
for monotone functions
where
(Hf ) (x) := (H1 f ) (x) = f udλ (7.4)
[0,x]
we may and shall restrict our studies to the inequality (7.3). All the charac-
terizations of (7.1) can be easily reproduced from the results for (7.3).
The weighted inequality (7.3) for f ∈ M ↓, when λ = μ = ν is the
Lebesgue measure, was essentially characterized in [93] and [100] with the
complement for the case 0 < q < 1 = p in [98] and recent contribution in
[13] for the case 0 < q < p < 1. In fact, [93], [100], [98] and [13] deal with
the case u(x) = 1, but a weight u can be incorporated with no change in
the arguments. A piece of historical remarks and the literature can be found
in ( [62] and [63, Chapter 6]). We summarize these results in the following
theorems:
and
∞ 1 t s p s −p 1
p
q
A1 := sup v u w w(s)ds
t>0 t 0 0 0
and C ≈ A0 + A1 .
1 1
(b) 0 < q < p < ∞, 1 < p < ∞, r := q − 1p , max (B0 , B1 ) < ∞, where
1
∞ t − pr t s q pr t q r
B0 := w u v(s)ds u v (t) dt ,
0 0 0 0 0
and
⎛ ⎞1
∞ ∞ r t s p s −p r
p
r
p
B1 := ⎝ v u w w(s)ds v(t)dt⎠
0 t 0 0 0
and C ≈ B0 + B1 .
Hardy inequality with three measures on monotone func- 85
tions
and C ≈ B0 + B1 .
(d) 0 < p ≤ q < ∞, 0 < p ≤ 1, max (A0 , A1 ) < ∞, where
t ∞ 1 t − p1
q
A1 := sup u v w
t>0 0 t 0
and C ≈ A0 + A1 .
It is important to note, that the weighted case of (7.3) for 1 < p, q < ∞
was solved in [93] by proving the principle of duality which allows to reduce
an inequality with a positive operator on monotone functions to an inequality
with modified operator on non-negative functions. The other cases, when
p, q ∈ (1, ∞) were studied by different methods.
Our aim is twofold. First we study the inequality (7.3) in the case 0 < p ≤
1 proving a complete analog of the parts (c) and (d) of Theorem 7.1 (Section
7.3). In the case 0 < q < p ≤ 1 our method is based on the characterization
of the Hardy inequality on nonnegative functions in the case 0 < q < 1 = p,
which we establish in Section 7.3 (Theorem 7.12). This approach is direct
and different from discretization methods of [13] and [37].
Hardy inequality (7.3) on monotone functions with two different measures
was recently investigated by G. Sinnamon [97]. Namely, for 1 < p < ∞ and
0 < q < ∞ the author established the equivalence of (7.3) with u ≡ v ≡ w ≡
1 and dλ = dν for f ∈ M+ to the same inequality restricted to f ∈ M ↓ .
Moreover, such equivalence takes place also for more general operator than
(7.4), that is for the operator
(Kf ) (x) = k (x, y) f (y) dλ (y)
[0,x]
holds. If γ ∈ (−1, 0) and Λ̄f (0) < +∞, then (7.7) holds.
The following two statements can be obtained from [[96], Lemma 2.1]
(see also [[97], Proposition 1.5]).
Remark 7.1. Two similar lemmas are valid for the approximation from
above.
The following statements are taken from [96] and concern the weighted
Lpλ [0, ∞) − Lqμ [0, ∞) inequality with the operator of the form
(Ku f ) (x) = k (x, y) u (y) f (y) dλ (y) .
[0,x]
holds for all f ∈ M+ if and only if A : = max (A0,1 , A0,2 ) < ∞, where
1 1
q p
p
A0,1 := sup v(x)k (x, t)q dμ (x) u dλ ,
t∈[0,∞) [t,∞) [0,t]
1 1
q p
p p
A0,2 := sup vdμ k (t, y) u (y) dλ (y) .
t∈[0,∞) [t,∞) [0,t]
If 1 < q < p < ∞ and 1r = 1q − 1p , then inequality (7.9) holds for all f ∈ M+
if and only if B : = max (B0,1 , B0,2 ) < ∞, where
B0,1 :=
⎛ r r ⎞1
q q
r
⎝ p p
v(x)k (x, t)q dμ (x) u dλ u (t) dλ (t)⎠ ,
[0,∞) [t,∞) [0,t]
B0,2 :=
⎛ r r ⎞1
p p
r
⎝ p p
vdμ k (t, y) u (y) dλ (y) v (t) dμ (t)⎠ .
[0,∞) [t,∞) [0,t]
88 Carleman type inequalities and Hardy type inequalities
for monotone functions
The next statement is an analog of the previous theorem for the operator
Ku∗ of the dual form
(Ku∗ f ) (x) = k (y, x) u (y) f (y) dλ (y)
[x,∞)
1 1
q p
p p
A∗0,2 := sup vdμ k (y, t) u (y) dλ (y) .
t∈[0,∞) [0,t] [t,∞)
If 1 < q < p < ∞ and 1r = 1q − 1p , then inequality (7.10) holds for all f ∈ M+
if and only if B∗ : = max B∗0,1 , B∗0,2 < ∞, where
B∗0,1 :=
⎛ r r ⎞1
q q
r
⎝ p p
q
v(x)k (t, x) dμ (x) u dλ u (t) dλ (t)⎠ ,
[0,∞) [0,t] [t,∞)
B∗0,2 :=
⎛ r r ⎞1
p p
r
⎝ p p
vdμ k (y, t) u (y) dλ (y) v (t) dμ (t)⎠ .
[0,∞) [0,t] [t,∞)
1 1 1
Theorem 7.9. If 0 < q < p < ∞, and r = q − p then
1 ⎛ r ⎞1
q q r
[0,∞) F vdμ
q
1 ≈ ⎝ dν(y)⎠ .
−1
sup w(y) W vdμ
F ∈M↓ p [0,∞) [y,∞)
[0,∞) F p wdν
(7.13)
Analogous results take place for F ∈ M ↑ .
Theorem 7.10. If 0 < p ≤ q < ∞, then
1 1
q q q
[0,∞) F vdμ [x,∞) vdμ
sup 1 = sup 1 . (7.14)
F ∈M↑ p x∈[0,∞) p
p
[0,∞) F wdν [x,∞) wdν
1 1
Theorem 7.11. If 0 < q < p < ∞ and r = q − 1p , then
1 ⎛ r ⎞1
q q r
[0,∞) F vdμ
q
sup ≈ ⎝ w(y) W̄ −1
vdμ dν(y)⎠ .
1
F ∈M↑ p [0,∞) [0,y]
[0,∞) F p wdν
(7.15)
Note that Theorems 7.9 and 7.11 with q = 1 give analogs of Sawyer’s
principle of duality with general Borel measures.
where
w dνa
w̃ := and w̃(x)↓ := ess inf λ w̃ (t) . (7.17)
u dλ t∈[0,x]
Moreover, C ≈ B.
Proof. Let us start with proving that (7.16) is equivalent to the following
inequality
q 1
q
dνa
f udλ v (x) dμ (x) ≤C fw dλ. (7.18)
[0,∞) [0,x] [0,∞) dλ
q 1
q
For a general w̃↓ we may and shall suppose that w̃↓ (x) < ∞ for all x > 0.
Let N ∈ N and
wN (x) := χ[0,N ] (x) w̃↓ (x) .
92 Carleman type inequalities and Hardy type inequalities
for monotone functions
(0)
Then wN (+∞) = 0 and similar to Lemma 7.5 we find that wN ∈ M ↓ and
hn ∈ M+ (n ∈ N) such that
(0)
(1) wN (x) ≤ wN (x) for all x ∈ [0, ∞).
(0)
(2) wN (x) = wN (x) for λ-a.e. x ∈ [0, ∞).
(0)
(3) wN,k (x) := [x,∞) hk dλ ≥ wN (x) for all x ∈ [0, ∞).
(4) The sequence {wN,k (x)}k≥1 is decreasing in k for all x ∈ [0, ∞) and
(0)
wN (x) = lim wN,k (x) λ-a.e. x ∈ [0, ∞). Then, by the previous part of the
k→∞
proof, for any f ∈ M+ we have that
q 1
q
f dλ v (x) dμ (x)
[0,N ] [0,x]
⎛ q ⎞ 1−q
1−q
q
v (z) dμ (z)
⎝ v (x) dμ (x)⎠ f wN,k dλ.
[0,N ] [0,x] wN,k (z) [0,N ]
1 1
By (3) and (1) we have that wN,k (z) ≤ wN (z) and by (4), (2) and the Mono-
tone Convergence Theorem, it yields that
f f f
lim dλ = (0)
dλ = dλ.
k→∞ [0,x] wN,k [0,x] w w
[0,x] N
N
f dλ v (x) dμ (x)
[0,N ] [0,x]
⎛ ⎞ 1−q
q
1−q
q
v (z) dμ (z)
⎝ v (x) dμ (x)⎠ f wN dλ
[0,N ] [0,x] wN (z) [0,N ]
Hardy inequality with three measures on monotone func- 93
tions
⎛ q ⎞ 1−q
1−q
q
⎝ v (z) dμ (z) ⎠
= v (x) dμ (x) f w̃↓ dλ
[0,N ] [0,x] w̃↓ (z) [0,N ]
≤B f w̃↓ dλ.
[0,∞)
⎝ v (z) dμ (z)
v (y) dμ (y)⎠ C
[0,∞) [0,y] wk (z)
Moreover, C ≈ B ≈ B0 + B1 , where
1
q
1
B0 := vdμ w (0)− p
[0,x]
94 Carleman type inequalities and Hardy type inequalities
for monotone functions
⎛ q ⎞ 1−q
1−q
q
q
⎝ − 1−q ⎠
B1 := w vdμ v (x) dμ (x) .
[0,∞) [x,∞)
Proof. It follows from Theorem 7.12, Lemma 7.3 and [97, Theorem 2.6].
Denote
Λ (t) := Λu (t) = udλ (7.23)
[0,t]
and observe that by the change f p → f in the inequality (7.3) we get the
following equivalent inequality:
p
1 q
q
Theorem 7.14. (a) Let 0 < p ≤ q < ∞ and 0 < p ≤ 1. Then (7.3) holds
for all f ∈ M ↓ if and only if
− 1 1
p q
and
− 1 1
p q
and C ≈ A0 + A1
(b) Let 0 < q < 1 = p. Then (7.3) holds for all f ∈ M ↓ if and only if
⎛ ⎞ 1−q
1
1−q
q
and
⎛ ⎞ 1−q
q
1−q
q
⎝ Λ (s)
B1 := ess supλ v(t)dμ(t) v (t) dμ (t)⎠ <∞
[0,∞) [0,x) s∈[0,t] W (s)
and C ≈ B0 + B1
Hardy inequality with three measures on monotone func- 95
tions
p
(s)
(c) Let 0 < q < p < 1, Vp (t) := ess supλ ΛW (s) . Then (7.3) holds for all
s∈[0,t]
f ∈ M ↓ if
⎛ p ⎞ p−q
p−q
pq
B0 := ⎝ w (y) W −1
Λ vdμ
q
dν (y)⎠ < ∞,
[0,∞) [y,∞)
and
⎛ q ⎞ p−q
p−q
pq
Proof. (a) Since f ∈ M ↓, then (Hu f ) (x) ≥ f (x) Λ (x) and (7.3) implies
1 1
q p
f Λ vdμ
q q
≤C p
f wdν , f ∈M↓.
[0,∞) [0,∞)
1 1
p q
= Λ (t) vdμ ,
[t,∞)
where
J :=
⎛ ⎛ ⎞q ⎞1
p−1 p
q
⎜ ⎝ ⎟
⎝ hudλ h(y)u(y)Λ (y) dλ(y)⎠ v (x) dμ (x)⎠ .
p
[0,∞) [0,x] [y,∞)
For the first term on the right hand side of (7.26), according to Theorem
7.8, we have that
1 1
q p
For the second term on the right hand side of (7.26), by the Minkowski
inequality with pq ≥ 1 and Lemma 7.3, we find that
⎛ p−1 p ⎞1
q
p
p 1
p
≈ A1 hudλ w(s)dν(s)
[0,∞) [s,∞)
1
p
≤ A1 f p wdν
[0,∞)
≤ (A0 + A1 ) f p wdν
[0,∞)
98 Carleman type inequalities and Hardy type inequalities
for monotone functions
1
q
q
(Hf ) vdμ ≤C f wdν, f ∈ M ↓, (7.29)
[0,∞) [0,∞)
that
1
q
q 1
q
q 1
q
q 1
q
W
hdλ v (x) dμ (x) ≤C h udλ.
[0,∞) [0,x] [0,∞) Λ
+ hudλ Λq (x)v(x)dμ(x)
[0,∞) [x,∞)
q 1 1
q q
+B0 f wdν ≤ (B0 + B1 ) f wdν.
[0,∞) [0,∞)
For an arbitrary f ∈ M ↓ we use the arguments from the end of the part
(a).
(c) Sufficiency.
To prove (7.3) we again, suppose first that f ∈ M ↓
, f (x) = [x,∞) hudλ for λ-a.e. x ∈ [0, ∞), where h ∈ M+ and f (x) ≥
[x,∞) hudλ for all x ∈ [0, ∞). Then, arguing as before and applying Minkowskii’s
inequality, we find that
p
1
q q
Hf p v(x)dμ(x)
[0,∞)
⎛ ⎛ 1 ⎞q ⎞ pq
p
+⎝ hudλ Λ (x)v(x)dμ(x)⎠
q
[0,∞) [x,∞)
⎛ q ⎞p p
p
q
q
q
≤⎝ hΛp udλ v(x)dμ(x)⎠ + f Λq vdμ
p
For an arbitrary f ∈ M ↓ we again use the arguments from the end of the
part (a).
Necessity. The inequality B0 ≤ C follows by using similar arguments as
in the proof of A0 ≤ C and B0 ≤ C in the parts (a) and (b).
For the rest it is sufficient to show that (7.24) implies the inequality
C B1 .
Suppose for simplicity, that Vp (0) = 0. Let
$ r
0
p
g (t) := max 2 , m ∈ Z : 2 ≤ Vp (t)
m m
and $ r
0
p
τm := inf y ∈ [0, ∞) : 2 ≤ Vp (y) .
m
and define r
q
[x,∞) vdμ
f (t) := dg (x) .
[t,∞) W (x)
Then f ∈ M ↓ and, by Lemma 7.3, it yields that
r
q
≥ v (x)
m [τm ,τm+1 )
⎛ ⎛ r ⎞ p1 ⎞q ⎞ 1q
q
⎜ ⎜ [s,∞) vdμ ⎟ ⎟ ⎟
×⎜
⎝ ⎝ dg (s)⎠ dΛ (y)⎟ ⎟
⎠ dμ (x)⎠
[0,τm ] [y,τm ] W (s)
⎛ r
p
≥ ⎝ vdμ vdμ
m [τm ,τm+1 ) [τm ,∞)
q 1
q
− p1 1
× W (τm ) (g (τm ) − g (y)) dΛ (y) p
[0,τm ]
⎛ r q ⎞ 1q
m
p
2 p Λ (τm )
⎝ vdμ vdμ 1
⎠
m [τm ,τm+1 ) [τm ,∞) W (τm ) p
102 Carleman type inequalities and Hardy type inequalities
for monotone functions
⎛ r ⎞1
p
q
≥ ⎝ 2m
vdμ v (s) dμ (s)⎠
m [τm ,τm+1 ) [s,∞)
⎛ r ⎞1
r p
q
r
⎝ p
Vp (s) vdμ v (s) dμ (s)⎠ =: B2,1
q
[0,∞) [s,∞)
pr
r B q ⇒C B .
With such f (x) the inequality (7.24) gives that: C p B2,1 2,1 2,1
Now, if we put f = χ{0} in (7.24), then we find that
1 − 1 1 − 1
q q
W (0) p 1 p
C≥ vdμ = vdμ =: B2,0 .
[0,∞) Λp (0) [0,∞) Vp (0)
Theorem 7.15. Let 0 < p ≤ q < ∞ and 0 < p ≤ 1. Then, (7.3) holds for
all f ∈ M ↑ if and only if
1
q
1
Ā1 := sup Λ (x, t) v (x) dμ (x)
q
W̄ − p (x) < ∞,
t∈[0,∞) [t,∞)
where
Λ (x, t) := udλ,
[t,x]
and C ≈ Ā1 .
and let 0 < p < 1. Then, by Minkowskii inequality and Lemma 7.2, we find
Hardy inequality with three measures on monotone func- 103
tions
that
hudλ u (s) dλ (s)
[0,x] [0,s]
⎛ p−1 ⎞1
p
1
q
(Hf )q vdμ
[0,∞)
⎛ ⎛ ⎞q
p−1 p
⎜ ⎝
≤ ⎝ hudλ h (y) u (y) Λ (x, y) dλ (y)⎠
p
[0,∞) [0,x] [0,y]
1
× v (x) dμ (x)) q
⎛ p−1
≤ ⎝ hudλ h (y) u (y)
[0,∞) [0,y]
p ⎞1
p
q
≈ Ā1 f p wdν .
[0,∞)
⎝ w(y) W −1 ∗
(T g)udλ dν(y)⎠ (7.33)
[0,∞) [y,∞)
1
q
q
≤ C q vdμ , g ∈ M+ ,
[0,∞)
holds with (T ∗ g)(z) = [0,∞) k(z, x)g(z)v(z)dμ(z).
(b) The inequality (7.32) for f ∈ M ↑ holds if and only if the following
inequality holds:
⎛ p ⎞ 1
p
Now let us present our result for the case 1 < p, q < ∞ .
Theorem 7.17. Let k(x, y) = [y,x] W −1 udλ and f ∈ M ↓ . The inequality
(7.3) holds for 1 < p ≤ q < ∞ if and only if A = max {A0,1 + A0,2 } < ∞,
where
1 1
p q
p
A0,1 : = sup w(y)k(t, y) dν(y) vdμ ,
t∈[0,∞) [0,t] [t,∞)
Hardy inequality with three measures on monotone func- 105
tions
1 1
p q
⎛ r r ⎞1
q q
r
B0,2 : = ⎝ wdν q
v(x)k(x, t) dμ(x) w(t)dν(t)⎠
[0,∞) [0,t] [t,∞)
1 1
and r = q − 1p . Moreover, if C is the best constant in (7.3), then C = B.
Proof. Because of Corollary 7.16 (a) the inequality (7.3) is equivalent to
⎛ p ⎞ 1
p
⎝ w(y) W (x)−1
gvdμ u(x)dλ(x) dν(y)⎠
[0,∞) [y,∞) [x,∞)
1
q
q
≤ C q vdμ , g ∈ M+ . (7.34)
[0,∞)
By changing the order of integration in the left hand side of (7.34) we obtain
the Hardy inequality with Oinarov kernel of the form
⎛ p ⎞ 1
p
⎝ w(y) 1/q
f (z)k(z, y)v(z) dμ(z) dν(y) ⎠
[0,∞) [y,∞)
1
q
q
≤ C f dμ .
[0,∞)
106 Carleman type inequalities and Hardy type inequalities
for monotone functions
1 1
p q
Moreover, if C is the best constant in (7.3), then C = Ā. 1In the case 12 < q <
p < ∞ the inequality (7.3) holds if and only if B̄ = max B̄0,1 + B̄0,2 < ∞,
where
⎛ r r ⎞1
p p
r
p
B̄0,1 : = ⎝ w(y)k̄(y, t) dν(y) vdμ v(t)dμ(t)⎠ ,
[0,∞) [t,∞) [0,t]
⎛ r r ⎞1
q q
r
1 1
and r = q − 1p . Moreover, if C is the best constant in (7.3), then C = B ∗ .
Chapter 8
Characterizations of embeddings in
Lorentz spaces
8.1 Introduction
Let (R, μ) be a totally σ-finite measure space with a non-atomic measure
μ, and let M (R, μ) be the set of all extended complex-valued μ-measurable
functions on R. For f ∈ M (R, μ), let
f∗ (t) = μ ({x ∈ R; |f (x)| > t}) , t ∈ (0, ∞) ,
be a distribution function of f . The non-increasing rearrangement f ∗ of f
is defined by
f ∗ (t) = inf {x > 0; f∗ (x) ≤ t} , t ∈ (0, ∞) .
We will assume that μ(R) = ∞. Everywhere in this chapter, we assume
that u, v, and w are weights, that is, locally integrable non-negative functions
on (0, ∞) and we denote
s s
V (s) = v (t) dt and U (s) = u (t) dt.
0 0
We assume that u is such that U (t) > 0 for every t ∈ (0, ∞) and denote
t
1
fu∗∗ (t) = f ∗ (s) u(s)ds.
U (t) 0
107
108 Carleman type inequalities and Hardy type inequalities
for monotone functions
Definition 8.1. Let p ∈ (0, ∞) and let u and v be weight functions. Then
we define the following four spaces (of Lorentz and Marcinkiewicz type):
1
∞ p
Λp (v) = f ∈ M (R, μ) ; f Λp (v) = f ∗p (t) v (t) dt <∞ ;
0
Λ∞ (v) = f ∈ M (R, μ) ; f Λ∞ (v) = ess sup f ∗ (t)v(t) < ∞ ;
t∈(0,∞)
1
∞ p
Γpu (v) = f ∈ M (R, μ) ; f Γpu (v) = fu∗∗p (t) v (t) dt <∞ ;
0
Γ∞
u (v) = f ∈ M (R, μ) ; f Γ∞
u (v)
= ess sup fu∗∗ (t)v (t) <∞ .
t∈(0,∞)
8.2 Preliminaries
Let us now recall some definitions and basic facts concerning discretization
and anti-discretization which can be found in [35] and [36].
respectively.
1 ϕ (t) U (t)
lim ϕ (t) = lim = lim = lim = 0.
t→0+ t→∞ ϕ (t) t→∞ U (t) t→0+ ϕ (t)
ϕ (xk ) ≈ ϕ (t) if k ∈ Z1 ,
ϕ (xk ) ϕ (t)
≈ if k ∈ Z2 .
U (xk ) U (t)
Let us recall [35, Lemma 2.7] that if ϕ ∈ ΩU , then there always exists a
discretizing sequence for ϕ with respect to U .
The proof follows from the following simple observation: If F and G are
non-negative functions on (0, ∞) and F is non-increasing, then
Lemma 8.2 ([36], Lemma 1.5). Let u, v be weights and let ϕ be defined
by
v(τ )
ϕ(t) = ess sup U (s) ess sup , t ∈ (0, ∞). (8.2)
s∈(0,t) τ ∈(s,∞) U (τ )
Γ∞ ∞
u (v) = Γu (ϕ),
Theorem 8.3 ([35], Theorem 2.11). Let p, q, r ∈ (0, ∞). Assume that
U is an admissible function, let ν be a non-negative non-degenerate Borel
measure on [0, ∞) and let ϕ be the fundamental function of ν with respect to
U q . Moreover, let σ ∈ ΩU and let {xk } be a discretizing sequence for ϕ with
respect to U q . Then
r
ϕ(xk ) q
r
ϕ(t) q −1
r dν(t) ≈ r .
[0,∞) σ(t) p k∈Z σ(xk ) p
Lemma 8.4 ([35], Corollary 2.13). Let q ∈ (0, ∞), let U be an admissible
function, let f ∈ ΩU , let ν be a positive non-degenerated Borel measure
on [0, ∞) and let ϕ be the fundamental function of ν with respect to U q .
Moreover, assume that {xk } is a discretizing sequence for ϕ with respect to
U q . Then
q 1 1
f (t)
q
f (xk ) q q
dν (t) ≈ ϕ (xk ) .
[0,∞) U (t) U (xk )
k∈Z
By using Lemma 3.1 and Lemma 3.2 from [35] we get the following result:
112 Carleman type inequalities and Hardy type inequalities
for monotone functions
q q
xk xk
sup h τk ≈ sup h τk ,
k∈Z 0 k∈Z xk−1
∞ q
xk+1 q
−1
h τk ≈ h τk−1 ,
k∈Z xk k∈Z xk
and ∞ q xk+1 q
−1
sup h τk ≈ sup h τk−1 .
k∈Z xk k∈Z xk
Lemma 8.7 ([35], Lemma 3.6). Let q ∈ (0, ∞), let U be an admissible
function, let ν be a positive non-degenerated Borel measure on [0, ∞) and let
ϕ be the fundamental function of ν with respect to U q . Assume that {xk } is
a discretizing sequence for ϕ with respect to U q . Then
∞ ∞ q
∞ q
|f (t)|dt |f (t)|dt
dν(x) ≈ ϕ(xk )
0 0 U (t) + U (x) 0 U (t) + U (xk )
k∈Z
xk xk+1 q
Lemma 8.8 ([35], Lemma 3.7). Let q ∈ (0, ∞), let U be an admissible
function and let ν be a non-degenerate positive Borel measure. Moreover, let
ϕ be the fundamental function of ν with respect to U q and let f be a measur-
able function on [0, ∞). If {xk } is a discretizing sequence for ϕ with respect
to U q , then
q q
|f (y)|
|f (y)|
sup dν(x) ≈ sup ϕ(xk )
[0,∞) y∈(0,∞) U (x) + U (y) k∈Z y∈(0,∞)
U (xk ) + U (y)
q
Lemma 8.10 ([35], Lemma 3.9). Let U be an admissible function and let
ϕ ∈ ΩU . Let {xk } be a discretizing sequence for ϕ with respect to U . Then
|f (y)| |f (y)|
sup ϕ(x) sup ≈ sup ϕ(xk ) sup
x∈(0,∞) 0<y<∞ U (x) + U (y) k∈Z 0<y<∞ U (xk ) + U (y)
≈ sup ϕ(xk )U (xk )−1 sup |f (y)| + sup ϕ(xk ) sup |f (y)|U (y)−1
k∈Z xk−1 <y<xk k∈Z xk <y<xk+1
ak vk ≤C ak wk
k∈Z k∈Z
1
−r r
r
wk p vkq < ∞.
k∈Z
Lemma 8.12 ([80] and [98]). We have the following Hardy type inequali-
ties:
114 Carleman type inequalities and Hardy type inequalities
for monotone functions
p
xk t p
p − pp
p
the best constant C ≈ xk−1 xk−1 u(s)ds V (t) u(t)dt and the
⎝ ⎝ 1 u(t)dt ⎠ ⎠
A8 := w(x)dx < ∞.
0 U (x) 0 ess supv (s)
0<s<t
116 Carleman type inequalities and Hardy type inequalities
for monotone functions
Our next goal is to formulate Theorem 8.13 in a more compact and nice
form. We need the following definition of the associate space (also called the
Köthe dual) Λp (v)u of Λp (v) .
In the classical situation (that is when u = 1) the associate norm Λp (v) was
characterised by E. Sawyer [93] for 1 < p < ∞ and by V. D. Stepanov [100]
for 0 < p ≤ 1. In the case p = ∞ it was considered in [21] (see also [34]).
In fact we can describe Λp (v) in the following way:
⎧
⎪
⎪ Γ ∞ t
, 0<p≤1
⎪
⎪ V (t)
1/p
⎪
⎪
⎨ p
Γp t v(t) p
, 1<p<∞
Λ (v) =
p
V (t) (8.7)
⎪
⎪
⎪
⎪
⎪ 1 1
⎩ Λ ess supv(s) ,
⎪ p = ∞.
0<s<t
Using Γqu (w) and Λp (v)u we can now formulate Theorem 8.13 in the
following compact form:
(i) 0 < p ≤ q ≤ ∞
3 3 3 3
sup 3χ(0,t) 3Γq (w) 3χ(0,t) 3Λp (v) U (t)−1 < ∞;
t u u
pq
(ii) 0 < q < p ≤ ∞, r = p−q
∞3 1/r
3 3 3r
3χ(0,t) 3r−q
q w (t) U (t)−r 3
χ (0,t)
3 p dt < ∞.
Γ (w) u Λ (v) u
0
Remark 8.2. Theorem 8.13 and Theorem 8.14 are in fact equivalent, the
only difference is that Theorem 8.14 is stated in a more compact form. All
the eight conditions A1 − A8 in Theorem 8.13 can be obtained from the two
conditions in Theorem 8.14.
8.4 Proofs
Proof of Theorem 8.13: We start with the upper bounds (sufficiency).
Let 0 < p < ∞. First we use the fact that f is non-increasing if and only
if f p is non-increasing and a standard argument shows that it is enough to
∞
prove (8.1) for f satisfying f ∗ (t)p = t h(s)ds, where h is some positive
measurable function on (0, ∞). That is, when 0 < q < ∞ we only have to
prove that
t ∞ 1 q 1q
∞
1 p
h(z)dz u(s)ds w (t) dt
0 U (t) 0 s
∞ ∞ 1
p
≤C h(s)ds v (t) dt (8.9)
0 t
t 1 q 1q
∞ ∞
1 p
h(z)dz u(s)ds w(t)dt
0 U (t) 0 s
∞ 1
p
≤C h(t)V (t)dt ,
0
and
t ∞ 1 ∞ 1
ϕ(t) p p
sup h(z)dz u(s)ds ≤ C h(t)V (t)dt , (8.11)
t∈(0,∞) (t)
U 0 s 0
respectively.
Assume first that 0 < q < ∞. Define
∞
w (s) ds
ϕ (x) = U q (x) . (8.12)
0 (U (s) + U (x))q
Then ϕ ∈ ΩU q and therefore there exists a discretizing sequence for ϕ with
k)
respect to U q . Let {xk } be one such sequence. Then ϕ (xk ) and Uϕ(x
q (x ) .
k
Furthermore, there is a decomposition Z = Z1 ∪Z2 , Z1 ∩Z2 = ∅ such that for
every k ∈ Z1 and t ∈ [xk , xk+1 ], ϕ (xk ) ≈ ϕ (t) and for every k ∈ Z2 and
k)
t ∈ [xk , xk+1 ] , Uϕ(x ϕ(t)
q (x ) ≈ U q (t) .
k
For the left-hand side of (8.9), using Lemma 8.4 with dυ(t) = w (t) dt
t ∞ 1
and f (t) = 0 s h (z) dz p u(s)ds we get that
t ∞ 1 q 1q
∞
1 p
h (z) dz u(s)ds w (t) dt
0 U (t) 0 s
q 1q
xk ∞ 1
p ϕ (xk )
≈ h (z) dz u(s)ds
0 s U (xk )q
k∈Z
[using Lemma8.6]
1 q 1q
xk ∞ p ϕ (xk )
≈ h (z) dz u(s)ds
xk−1 s U (xk )q
k∈Z
1 q 1q
xk xk ∞ p ϕ (xk )
= h (z) dz + h (z) dz u(s)ds
xk−1 s xk U (xk )q
k∈Z
Characterizations of embeddings in Lorentz spaces 119
1 q 1q
xk xk p ϕ (xk )
≈ h (z) dz u(s)ds
xk−1 s U (xk )q
k∈Z
1 q 1q
xk ∞ p ϕ (xk )
+ h (z) dz u(s)ds
xk−1 xk U (xk )q
k∈Z
1q 1 q
xk
ϕ (xk )
xk p
≈ h (z) dz u(s)ds
xk−1 s U (xk )q
k∈Z
q xk q 1
∞ p ϕ (xk )
q
+ h (z) dz u(s)ds
xk xk−1 U (xk )q
k∈Z
xk
[using the fact that xk−1 u(s)ds = U (xk ) − U (xk−1 ) ≈ U (xk )]
1 q 1q
xk xk p ϕ (xk )
≈ h (z) dz u(s)ds
xk−1 s U (xk )q
k∈Z
1
∞ q
p
q
+ h (z) dz ϕ (xk )
k∈Z xk
1 q 1q
xk xk p ϕ (xk )
≈ h (z) dz u(s)ds
xk−1 s U (xk )q
k∈Z
1
xk+1 q
p
q
Now we shall distinguish several cases. Starting with the case 0 < p ≤ 1.
Using (8.3) for I and (8.5) for II we find that
1 q 1q
xk xk p ϕ (xk )
I =: h (z) dz u(s)ds
xk−1 s U (xk )q
k∈Z
120 Carleman type inequalities and Hardy type inequalities
for monotone functions
q
ϕ (xk ) t
− 1
≤ sup u (s) ds V (t) p ×
U (xk )q xk−1 <t<xk x k−1
k∈Z
q ⎞ 1q
xk p
h (s) V (s) ds ⎠
xk−1
⎛ q ⎞ 1q
ϕ (xk ) − pq
xk p
II = : h (z) dz ϕ (xk )
k∈Z xk
q 1q
− qp
xk+1 p
≤ ϕ (xk ) V (xk ) h (s) V (s) ds . (8.15)
k∈Z xk
q
(i) In the case 0 < p ≤ 1, p ≤ q < ∞ (i.e. p ≥ 1) we have according to
(8.14) that
1 ∞ 1
ϕ (xk ) q −1
p
I ≤ sup sup U (t) V (t) p h (s) V (s) ds ,
k U (xk ) xk−1 <t<xk 0
q
and, similarly, if p ≥ 1 we have according to (8.15) that
∞ 1
1 p
− p1
II ≤ supϕ (xk ) V (xk ) q h (s) V (s) ds
k 0
pq
(ii) For the case 0 < q < p ≤ 1, r = p−q , by applying Hölder’s inequality
for sums to the right-hand side of (8.14) and the right hand side of (8.15)
Characterizations of embeddings in Lorentz spaces 121
p
with exponents q and rq , we find that
r r 1 1
ϕ (xk ) q − p1
r ∞ p
I r sup U (t) V (t) h (s) V (s) ds
U (xk ) xk−1 <t<xk 0
k∈Z
1 1
r
− pr
r ∞ p
II ϕ (xk ) V (xk )
q h (s) V (s) ds .
k∈Z 0
Now let us assume that 1 < p < ∞. Using (8.4) for I and (8.5) for II, we
get that
⎛ q
ϕ (xk ) xk p p
−p
p
(iii) Suppose that 1 < p ≤ q < ∞. Then, according to (8.17) and (8.18),
we obtain that
1 1 1
p ∞
ϕ (xk ) q xk p
−pp p
I sup U (t) V (t)
p u(t)dt h (s) V (s) ds ,
k>0 U (xk ) xk−1 0
∞ 1
1 1 p
−
II supϕ (xk ) q V (xk ) p h (s) V (s) ds .
k 0
122 Carleman type inequalities and Hardy type inequalities
for monotone functions
⎛ r ⎞ 1r
ϕ q (xk )
r
xk p ∞ 1
p p
− pp
I ⎝ U (t) p V (xk ) u(t)dt ⎠ h (s) V (s) ds
U (xk )r xk−1 0
k∈Z
and 1
r ∞ 1
r p
− pr
II ϕ (xk ) V (xk )
q h (s) V (s) ds .
k∈Z 0
I + II
⎛ r ⎞ 1r
ϕ q (xk ) xk
r
p
p p p
−p −p
⎝ V (t) U (t) p u(t)dt + V (x )
k U (x k )p ⎠ ×
U (xk )r xk−1
k∈Z
∞ 1
p
h (s) V (s) ds
0
Characterizations of embeddings in Lorentz spaces 123
⎛ r ⎞ 1r
ϕ q (xk ) xk
r
p
p
−
⎝ r V (t)−p U (t)p v(t)dt + V (xk ) p U (xk )p ⎠ ×
U (xk ) xk−1
k∈Z
∞ 1
p
h (s) V (s) ds
0
⎛ r ⎞ 1r
r ∞ p −p ∞ 1
U (t) V (t) v(t)dt p ⎠ p
⎝ ϕ q (xk ) p
h (s) V (s) ds
k∈Z 0 (U (t) + U (xk )) 0
∞ 1
p
≤ A4 h (s) V (s) ds .
0
Let now q = ∞. In the same way, by using Lemma 8.2, Lemma 8.5 and
Lemma 8.6 we obtain for the left-hand side of (8.11) that
t ∞ 1
ϕ(t) p
sup h(z)dz u(s)ds
t∈(0,∞) U (t) 0 s
∞ 1
ϕ(xk ) xk p
≈ sup h(z)dz u(s)ds
k∈Z U (xk ) 0 s
xk 1
ϕ(xk ) xk p
≈ sup h(z)dz u(s)ds
k∈Z U (xk ) xk−1 s
xk+1 1
p
+ sup ϕ(xk ) h(z)dz := III + IV, (8.19)
k∈Z xk
1
ϕ (xk ) − p1
xk+1 p
III sup sup U (t) V (t) h (s) V (s) ds ,
k U (xk ) xk−1 <t<xk xk
xk+1 1
p
− pq
IV supϕ (xk ) V (xk ) h (s) V (s) ds
k xk
124 Carleman type inequalities and Hardy type inequalities
for monotone functions
(vi) Let now 1 < p < ∞, q = ∞. By using (8.4) for III and (8.5) for IV ,
we obtain that
1 1
p p
ϕ (xk ) xk p
− p xk
III sup U (t) p V (t) p u(t)dt h (s) V (s) ds
k∈Z U (xk ) xk−1 xk−1
and 1
− p1
xk+1 p
IV ϕ (xk ) V (xk ) h (s) V (s) ds .
k∈Z xk
1
p
ϕ (xk ) xk
−p p − pp
p
sup V (t) U (t) v(t)dt + V (xk ) U (xk ) ×
k>0 U (xk ) xk−1
∞ 1
p
h (s) V (s) ds
0
1
∞
U (t)p V (t)−p v(t)dt p
supϕ (xk ) ×
k>0 0 U (t)p + U (xk )p
∞ 1
p
h (s) V (s) ds
0
∞ 1
p
= A6 h (s) V (s) ds .
0
Characterizations of embeddings in Lorentz spaces 125
(vii) For the case p = q = ∞, we use Lemma 8.1 and get that
w(x) x ∗
f Γ∞
u (w)
= ess sup f (t)u(t)dt
x∈(0,∞) U (x) 0
w(x) x u(t)dt
≤ ess sup ess sup f ∗ (x)ess supv (s)
x∈(0,∞) U (x) 0 ess supv (s) x∈(0,∞) 0<s<x
0<s<t
≤ A7 f Λ∞ (v) .
⎝ 1
x
u(t)dt ⎠
f Γqu (w) ≤⎝ w(x)dx⎠ ess sup f ∗ (x)ess supv (s)
0 U (x) 0 ess supv (s) x∈(0,∞) 0<s<x
0<s<t
≤ A8 f Λ∞ (v) .
Next we prove the lower bounds (Necessity). Let 0 < q < ∞ and let {xk }
be a discretizing sequence for ϕ from (8.12). Then by (8.13) we have that
1 q 1q
xk xk p ϕ (xk )
h (z) dz u(s)ds
xk−1 s U (xk )q
k∈Z
1
xk+1 q
p
q
+ h (z) dz ϕ (xk )
k∈Z xk
1
xk p
Let 0 < p ≤ 1. For k ∈ Z, let hk be functions that satisfy the Hardy in-
equality (8.3) and the Hölder inequality (8.5), that is, functions hk satisfying
supp hk ⊂ [xk−1 , xk ] ,
xk
hk (s) V (s)ds = 1,
xk−1
1 p
xk xk p
hk (t) dt u(s)ds sup (U (t) − U (xk−1 ))p V (t)−1 ,
xk−1 s xk−1 <t<xk
xk
hk (s) ds V (xk−1 )−1 .
xk−1
126 Carleman type inequalities and Hardy type inequalities
for monotone functions
where {ak } is a sequence of positive real numbers. Then, by using this test
function in (8.20), we get that
1
pq ϕ (xk ) q − pq
q
Now, by using Proposition 8.11 for the case 0 < p ≤ 1, p ≤ q, we find that
1
− p1
supϕ(xk ) q V (xk ) +
k∈Z
1
−1/p ϕ (xk ) q
sup sup (U (t) − U (xk−1 )) V (t) < ∞.
k∈Z xk−1 <t<xk U (xk )
Moreover, according to Lemma 8.10, we see that
1
A1 sup sup V (t)−1/p ϕ (t) q
k∈Z xk−1 <t<xk
1
1
− p1 −1/p ϕ (xk ) q
supϕ(xk ) V (xk ) q + sup sup U (t) V (t)
k∈Z k∈Z xk−1 <t<xk U (xk )
1 1
supϕ(xk ) q V (xk )− p
k∈Z
1
−1/p ϕ (xk ) q
+ sup sup (U (t) − U (xk−1 )) V (t) < ∞.
k∈Z xk−1 <t<xk U (xk )
Let 0 < q < p ≤ 1. We obtain from (8.22) and Proposition 8.11 that
1
r
− r
r
ϕ(xk ) q V (xk ) p
k∈Z
r 1
ϕ (xk ) q r
Therefore, in view of Theorem 8.3, Lemma 8.6 and Lemma 8.8, we have that
r
1
ϕ (x ) q
r
−r/p k
A2 sup U (t) V (t)
r
xk−1 <t<xk U (xk )r
k∈Z
1
r
r
Let 1 < p < ∞. For k ∈ Z, let hk be functions that satisfy the Hardy in-
equality (8.4) and the Hölder inequality (8.5), that is, functions hk satisfying
supp hk ⊂ [xk−1 , xk ] ,
xk
hk (s) V (s)ds = 1,
xk−1
1 p
xk xk p
hk (t) dt u(s)ds
xk−1 s
p
xk p
p
− pp
(U (t) − U (xk−1 )) p V (t) u(t)dt
xk−1
xk
hk (s) ds V (xk−1 )−1 .
xk−1
where {ak } is a sequence of positive real numbers. Then, by using this test
function in (8.20) we get that
⎛ q ⎞ 1q
q
ϕ (xk ) xk p
− pp
p
⎝ p
ak (U (t) − U (xk−1 )) p V (t) u(t)dt ⎠
U (xk )q xk−1
k∈Z
1 1
q
− pq
q
p
p
+ ak ϕ (xk−1 ) V (xk−1 ) ak . (8.24)
k∈Z k∈Z
128 Carleman type inequalities and Hardy type inequalities
for monotone functions
Now, by using Proposition 8.11 for the case 1 < p < q < ∞, we find that
1/p 1
xk
p /p
−p /p ϕ (xk ) q
sup (U (t) − U (xk−1 )) V (t) u(t)dt
k∈Z xk−1 U (xk )
1 1
+ supV (xk )− p ϕ (xk ) q < ∞.
k∈Z
Hence, by using Lemma 8.5 and Lemma 8.9 and performing integration by
part, we have that
1 1/p
ϕ (xk ) q xk
1 1
A3 sup V (t)−p U (t)p v(t)dt + supϕ (xk ) q V (xk )− p
k∈Z U (xk ) xk−1 k∈Z
1 1/p
ϕ (xk ) q xk
−p p
sup V (t) (U (t) − U (xk−1 )) v(t)dt
k∈Z U (xk ) xk−1
1 1
+supϕ (xk ) q V (xk )− p
k∈Z
1 1/p
ϕ (xk ) q xk
− pp p
sup V (t) (U (t) − U (xk−1 )) p u(t)dt
k∈Z U (xk ) xk−1
1 1
+supϕ (xk ) q V (xk )− p < ∞.
k∈Z
Let now 1 < p < ∞, 0 < q < p. Then, by using (8.24) and Proposition 8.11,
we obtain that
⎛ r ⎞ 1r
ϕ (xk )
r/q x k
− pp
p p
⎝ V (t) (U (t) − U (x k−1 )) p u(t)dt ⎠
U (xk )r xk−1
k∈Z
1
r
− pr
r
+ ϕ (xk ) V xk
q < ∞.
k∈Z
Therefore, by using Theorem 8.3 and Lemma 8.7 and integration by part,
we get that
⎛ r ⎞ 1r
r
ϕ (xk ) q xk
p
+ ϕ (xk ) V (xk )
q
k∈Z
Characterizations of embeddings in Lorentz spaces 129
⎛ r ⎞ 1r
r
ϕ (xk ) q xk
p
+ ϕ (xk ) V (xk )
q
k∈Z
⎛ r ⎞ 1r
r
ϕ (xk ) q xk p
−p p
p
Let 0 < p ≤ 1. If we use in (8.25) the test function defined by (8.21), then
we obtain that
ϕ (xk )
1
−1
sup akp sup (U (t) − U (xk−1 )) V (t) p
k∈Z U (xk ) xk−1 <t<xk
1
1 1
p
Now let 1 < p < ∞. Using in (8.25) the test function defined by (8.23) we
obtain that
1
p
ϕ (xk )
1
p
xk p
− pp
sup ak (U (t) − U (xk−1 )) p V (t) u(t)dt
k∈Z U (xk ) xk−1
1
1
p − p1
p
1/p
xk
p /p −p /p ϕ (xk )
sup (U (t) − U (xk−1 )) V (t) u(t)dt
k∈Z xk−1 U (xk )
1
+ supV (xk )− p ϕ (xk ) < ∞.
k∈Z
Hence, by using Lemma 8.5, Lemma 8.9 and performing integration by part,
Characterizations of embeddings in Lorentz spaces 131
we find that
1/p
ϕ (xk ) xk
−p p
A6 sup V (t) U (t) v(t)dt
k∈Z U (xk ) xk−1
1
+ supϕ (xk ) V (xk )− p
k∈Z
1/p
ϕ (xk ) xk
−p p
sup V (t) (U (t) − U (xk−1 )) v(t)dt
k∈Z U (xk ) xk−1
1
+ supϕ (xk ) V (xk )− p
k∈Z
1/p
ϕ (xk ) xk
− pp
p
sup V (t) (U (t) − U (xk−1 )) p u(t)dt
k∈Z U (xk ) xk−1
1
+ supϕ (xk ) V (xk )− p < ∞.
k∈Z
The corresponding estimates for A7 and A8 are trivial, by using Lemma 8.1
and taking the test function f (t) = (ess sup0<s<t v(s))−1 . The proof is com-
plete.
Proof of Theorem 8.14:
It is easy to see that
∗∗ U (t)
χ(0,t) (s) ≈ . (8.26)
u U (s) + U (t)
⎛ ⎞1
∞ q
3 3 U (t) q
3 χ(0,t) 3 ≈⎝ w (s) ds⎠ (8.27)
Γqu (w) (U (t) + U (s))q
0
and
3 3 U (t) w(s)
3 χ(0,t) 3 ≈ ess sup , (8.28)
Γ∞
u (w) U (t) + U (s)
0<s<∞
we find that
3 3 3 3
3 χ(0,t) 3 ≈ 3 χ(0,t) 3
Λp (v)u U (t)p v(t)
Γpu
V (t)p
1
∞ p p
U (s) v (s)
≈ U (t) ds . (8.29)
0 U (t) + U (s) V (s)p
3 3 3 3
sup 3χ(0,t) 3Γq (w) 3χ(0,t) 3Λp (v) U (t)−1
u u
t>0
⎛∞ ⎞1 ⎛ ∞ ⎞ 1
q q p p
2. For the case 1 < p < ∞, q = ∞, by using (8.8) and (8.28), we have
that
3 3 3 3
sup 3χ(0,t) 3Γ∞ (w) 3χ(0,t) 3Λp (v) U (t)−1
t>0 u u
⎛ ∞ p
⎞ 1
p
U (t) w (s) ⎝ U (s) v (s) ⎠
≈ sup ess sup ds
t>0 s∈(0,∞) U (t) + U (s) U (t) + U (s) V (s)p
0
= A6 .
For the case 0 < p ≤ 1, by using (8.8) and (8.26), we obtain that
3 3 3 3
3χ(0,t) 3 ≈ 3χ(0,t) 3
Λp (v)u Γ∞
U (t)
u
V (t)1/p
U (s)
≈ U (t) ess sup V (s)−1/p . (8.30)
t∈(0,∞) U (s) + U (t)
3. For the case, 0 < p ≤ 1, p < q < ∞, by using Lemma 8.2 and (8.30), we
Characterizations of embeddings in Lorentz spaces 133
get that
3 3 3 3
sup 3χ(0,t) 3Γ∞ (w) 3χ(0,t) 3Λp (v) U (t)−1
t>0 u u
3 3 3 3
3 3 3
≈ sup χ(0,t) Γq (w) χ(0,t) 3 U (t)−1
U (t)
t>0 u Γ∞
u
V (t)1/p
⎛ ⎞1
∞ q q
U (t) U (s)
≈ sup ⎝ w (s) ds⎠ ess sup V (s)−1/p
t>0 U (t) + U (s) s∈(0,∞) U (s) + U (t)
0
⎛ ⎞1
∞ q q
U (t)
≈ sup ⎝ w (s) ds⎠ V (t)−1/p
t>0 U (t) + U (s)
0
= A1 .
4. For the case 0 < p ≤ 1, q = ∞, by using (8.28), (8.30) and Lemma 8.2,
we find that
3 3 3 3
sup 3χ(0,t) 3Γ∞ (w) 3χ(0,t) 3Λp (v) U (t)−1
t>0 u u
3 3 3 3
≈ sup 3χ(0,t) 3Γ∞ (w) 3χ(0,t) 3 U (t) U (t)−1
t>0 u Γ∞
u
V (t)1/p
pq
5. Now consider the case 1 < p < ∞, 0 < q < p, r = p−q . By using (8.8)
(8.27), and (8.29), we have that
⎛ ⎞1/r
∞
3 3 3 3
⎝ 3 χ(0,t) 3r−q
q w (t) U (t)−r 3χ(0,t) 3Λp (v) dt⎠
r
Γu (w) u
0
⎛∞ ⎞1/r
3 3r−q 3 3
≈ ⎝ 3 χ(0,t) 3Γq (w) w (t) U (t)−r 3χ(0,t) 3 p U (t)p v(t) dt⎠
r
u Γu
V (t)p
0
134 Carleman type inequalities and Hardy type inequalities
for monotone functions
⎛ ⎛ ⎞ r−q
∞ ∞ q q
⎜ ⎝ U (t) ⎠
≈ ⎝ w (s) ds w (t) ×
(U (t) + U (s))q
0 0
⎛∞ ⎞ r ⎞1/r
p p
⎝ U (t) v (s) ⎟
ds⎠ dt⎠
U (s) + U (s) V (s)p
0
= A4 .
6. For the case 0 < q < p ≤ 1, by using (8.8), (8.27) and (8.30), we obtain
that
⎛ ⎞1/r
∞
3 3 3 3r
⎝ 3 χ(0,t) 3r−q w(t)U (t)−r 3χ(0,t) 3 ∞ dt⎠
Γq (w)u Γu
U (t)
V (t)1/p
0
⎛ ⎛ ⎞ r−q
∞ ∞ q
⎜ ⎝ U (t)q ⎠
≈⎝ q w(s)ds w(t)U (t)−r ×
(U (t) + U (s))
0 0
1r
−1 r
U (t)U (s)V (s) p
ess sup dt
s∈(0,∞) U (t) + U (s)
⎛ ⎛ ⎞ r−q ⎞ 1r
∞ ∞ q r
⎜ U (t)q −1 ⎟
≈⎝ ⎝ q w(s)ds
⎠ w(t)U (t)−r sup U (s)V (s) p dt⎠
(U (t) + U (s)) s∈(0,t)
0 0
= A2 .
3 3 3 3
≈ sup 3χ(0,t) 3Γ∞ (w) 3 χ(0,t) 3
u(τ )
U (t)−1
t>0 u Λ1 ess sups∈(0,τ ) v(s)
t
U (t) w (s) 1 u (τ )
≈ sup ess sup dτ
t>0 s∈(0,∞) U (t) + U (s) U (t) ess sups∈(0,τ ) v(s)
0
Characterizations of embeddings in Lorentz spaces 135
t
w (s) u (τ )
≈ ess sup dτ
s∈(0,∞) U (s) ess sups∈(0,τ ) v(s)
0
= A7 .
8. For the case 0 < q < p = ∞, we have that r = q and by using (8.8),
we find that
⎛∞ ⎞1/r
3 3 3 3r
⎝ 3 χ(0,t) 3r−q w (t) U (t)−r 3χ(0,t) 3 ∞ U (t) dt⎠
Γq (w) u Γu
V (t)1/p
0
⎛ ⎞1/q
∞
3 3q
≈ ⎝ w (t) U (t)−q 3 χ(0,t) 3
u(t)
dt⎠
Λ1 ess sups∈(0,t) v(s)
0
⎛ ⎛ ⎞ ⎞1/q
∞ t
1 u (τ )
= ⎝ ⎝ dτ ⎠ w (t) dt⎠
U (t) ess sups∈(0,τ ) v(s)
0 0
= A8 .
[8] S. Barza, H. Heinig, and L.-E. Persson. Duality theorem over the cone
of monotone functions and sequences in higher dimensions. J. Inequal.
Appl., 7(1):79–108, 2002.
137
138 Carleman type inequalities and Hardy type inequalities
for monotone functions
[17] M. Carro, A. García del Amo, and J. Soria. Weak-type weights and
normable Lorentz spaces. Proc. Amer. Math. Soc., 124:849–857, 1996.
[19] M. Carro and J. Soria. Weighted Lorentz spaces and the Hardy oper-
ator. J. Funct. Anal., 112:480–494, 1993.
[21] M. Carro, J. Soria, Pick L., and V.D. Stepanov. On embeddings be-
tween classical Lorentz spaces. Math. Ineq. Appl., 4:397–428, 2001.
[36] A. Gogatishvili and L. Pick. Embeddings and duality theory for weak
classical Lorentz spaces. Canad. Math. Bull., 49:82–95, 2006.
[81] B. Opic and L. Pick. On the geometric mean operator. J. Math. Anal.
Appl., 183(3):652–662, 1994.
[86] G. Pólya. Proof of an inequality. Proc. London Math. Soc, 24(2), 1926.
144 Carleman type inequalities and Hardy type inequalities
for monotone functions
[94] G. Sinnamon. Spaces defined by level functions and their duals. Studia
Math., 111:19–52, 1994.
[102] A. Wedestig. Some new Hardy type inequalities and their limiting
inequalities. J. Inequal. Pure Appl. Math., 4(3), 2003.