A substitute of harmonic majorization

Liliana de Rosa and Carlos Segovia

Abstract. A substitute of harmonic majorization of powers of harmonic func-

2
tions in R+ is given for one sided kernels and one sided doubling weights. The
substitute of harmonic majorization obtained is applied to the study of a gener-
alization of the Tλ,r functions of Fefferman and Stein.

It is well known, see [SW] and [GR], that if u(x, t) is a harmonic function
defined on the upper half plane (t > 0) , 1 ≤ q < ∞ and δ > 0 , then there exists
a function g(x) ≥ 0 such that

|u(x, t)|δ ≤ Pt ∗ g(x),

1 t 2
where Pt (x) = π t2 +x2 is the Poisson kernel on R+ and
Z Z
g(x)q dx ≤ u∗ (x)qδ dx,

where u∗ (x) = sup|x−y|<t |u(y, t)| . The purpose of this paper is to generalize these
results to kernels different from the Poisson kernel and other maximal functions.
The result we obtain is stated in Theorem (5). We apply Theorem (5) to the study
of a one sided version of the Tλ,r function of Fefferman and Stein [FS] and the
results obtained in this direction are stated in Theorem (18).
The Lebesgue measure of a set E ⊆ IR is denoted by |E| and its characteristic
function by χ E (x) . As usual, a weight w(x) is a Lebesgue measurable and non-
negative function defined on IR . A function f (x) belongs to Lp (w) , 0 < p ≤ ∞ ,
R 1/p
∞ p
if kf kLp (w) = −∞
|f (x)| w(x)dx is finite. If w(x) ≡ 1 , we simply write

1991 Mathematical Subject Classification. Primary 42B30; Secondary 42B25.

Key words and phrases. Maximal functions, harmonic functions, weights.

1
2 l. de rosa and c. segovia

kf kLp or kfRkp . If E ⊆ IR is a Lebesgue measurable set, we denote its w -measure

by w(E) = E w(t)dt . The set function w(E) is a measure on the σ -algebra of all
Lebesgue measurable subset of IR . If w([x, x + 2t]) ≤ cw · w([x, x + t]) holds for
every x ∈ IR and t > 0 , we shall say that w(x) is a right hand doubling weight.
A weight w(x) belongs to the class A− p , 1 < p < ∞ , if there exists a constant
cw,p such that for every a < b < c , we have

Z  Z !p−1
c b
1
− p−1
w(y)dy w(y) dy ≤ cw,p (c − a)p ,
b a

and w(x) belongs to A1− if

Z x+s
1
sup w(y)dy ≤ cw,1 w(x)
s>0 s x

holds for almost every x in IR .

We define the left sided Hardy-Littlewood maximal M − f (x) as
Z
1 x
M − f (x) = sup |f (y)|dy.
s>0 s x−s

E. Sawyer proved in [S] that in the case 1 < p < ∞ , there exists a constant cw,p
such that
kM − f kLp (w) ≤ cw,p kf kLp (w)
if and only if w ∈ A−
p . Moreover, the inequality

sw({x : M − f (x) > s}) ≤ cw,1 kf kL1 (w)

holds for every s > 0 , if and only if w ∈ A−1 .

Let w(x) ≥ 0 defined on IR . Given a Lebesgue measurable function f (x) ,
for every bounded interval I we define mI (f, w) as
Z
1
mI (f, w) = |f (y)|w(y)dy, if 0 < w(I) < ∞
w(I) I
and
mI (f, w) = 0, if w(I) = 0 or w(I) = ∞ .
Then, the maximal function Mw f (x) is defined as

(1) Mw f (x) = sup mI (f, w),

x∈I
 a substitute of harmonic majorization 3

where the supremum is taken over all bounded intervals I such that x ∈ I . Since
we work on the real line it is well known that there exists a constant cq such that
the weighted norm inequality

(2) kMw f kLq (w) ≤ cq kf kLq (w)

holds if 1 < q < ∞ , and

s w({x ∈ IR : Mw f (x) > s}) ≤ ckf kL1 (w)

holds for every s > 0 .

As usual, by C0∞ we mean the class of all functions with compact support
and derivatives of all orders, and S stand for the class of functions with rapidly
decreasing derivatives of all orders.
Let γ be a non-negative integer and Φγ the class of all C0∞ -functions ψ(x) ,
such that there exists an interval I = [0, β] satisfying
(i) support (ψ) ⊂ I
(ii) |I|γ+1 kDγ ψk∞ ≤ 1.
Given a bounded function f (x) with compact support, we shall consider the
∗
left sided maximal function f−,γ defined as
∗
(3) f−,γ (x) = sup |f ∗ ψ(x)|

where the supremum is taken over all ψ belonging to Φγ , see [RS].

Let r > 0 and s > 1 + r , we define the function
xr χ (0,∞) (x).
(4) h(x) =
(1 + x)s

For any function ϕ(x) we denote by ϕt (x) the function ϕt (x) = t−1 ϕ(x/t) .
With these definitions and notations we can state the main result of this paper.

(5) Theorem. Let w(x) be a right hand doubling weight, k a non-negative

integer and ϕ(x) ∈ S with support contained in [0, ∞) . Given a bounded function
f (x) with bounded support there exists a non-negative function g(x) such that

(6) |f ∗ ϕt (x)| ≤ c(g ∗ ht (x))k+2

and
Z Z
(7) g(x)q w(x)dx ≤ c ∗
f−,k+1 (x)q/(k+2) w(x)dx
4 l. de rosa and c. segovia

hold with 1 ≤ q < ∞ and a constant c depending on k, ϕ, h and q .

In order to prove Theorem (5) we shall need some results.

(8) Lemma. Let w(x) be a right hand doubling weight. If {xk } is a sequence
of real numbers and {rk } and {λk } are sequences of positive numbers, then given
a > 1 , there exists a constant cq,a (w) such that
X X

λk χ (xk ,xk +ark ) ≤ cq,a (w) λk χ (xk ,xk +rk )
q L (w) q L (w)

holds for 1 ≤ q < ∞ . The constant c depends on w and q .

Proof. The proof of this Lemma for a doubling weight is sketched in Lemma 4,
page 115 of [ST]. For the sake of completeness we give the proof here. First of all
we observe that it is enough to prove this lemma if w((xk , xk + rk )) < ∞ for every
k . Actually, we can assume that 0 < w((xk , xk + rk )) < ∞ , k ≥ 1 , and since
w(x) is a right hand doubling weight it follows that 0 < w((xk , xk + ark )) < ∞
for every k . Let g(x) ≥ 0 and Mw g(x) the maximal function defined in (1). For
xk ≤ z ≤ xk + rk , we have
Z xk +ark
g(x)w(x)dx ≤ w(xk , xk + ark )Mw g(z).
xk

Therefore, since w(x) es a right hand doubling weight, we get

Z xk +ark Z xk +rk
g(x)w(x)dx ≤ c Mw g(z)w(z)dz,
xk xk

for every k . Then

X Z xk +ark X Z xk +rk
λk g(x)w(x)dx ≤ c λx Mw g(z)w(z)dz,
k xk k xk

which is equivalent to
Z !
∞ X
λk χ(xk ,xk +ark ) (x) g(x)w(x)dx
−∞ k
Z !
∞ X
≤c λk χ(xk ,xk +rk ) (z) Mw g(z)w(z)dz.
−∞ k
 a substitute of harmonic majorization 5

Applying Hölder’s inequality and (2), we have that the right hand side above
is bounded by

X χ
C λk (xk ,xk +rk ) kgkLq0 (w) ,

k Lq (w)

and the Lemma follows immediately with cq,a (w) = C .

(9) Lemma. Let −∞ < α < β < ∞ and I = (α, β) . Let k be a non-negative
integer and a(x) a function satisfying
(i) kak∞ < ∞ ,
(ii) support of a(x) contained in I and
R
(iii) if ` is a non-negative integer, ` ≤ k , then x` a(x)dx = 0 .
Then, for ϕ(x) ∈ S with support contained in [0, ∞) we have
a) if x ≤ α then ϕt ∗ a(x) = 0 ,
b) if x ≥ α then for any integer n ≥ 1
 n+1
kak∞ kDn ϕk∞ x − α
(10) |ϕt ∗ a(x)| ≤ ,
(n + 1)! t

and

(11) |ϕt ∗ a(x)| ≤ kak∞ kϕk1 ,

c) if x ≥ β + (β − α) then for any non-negative integers u and v

 k+2 x−α u


β−α t
|ϕt ∗ a(x)| ≤ ck,u,v,ϕ kak∞
v .
t 1 + x−α
t

Proof. By a change of variables, it is enough to prove the Lemma for the case
α = 0 . Let us show part a). It is evident since ϕt ∗ a(x) is the convolution of
functions supported on [0, ∞) . Part b) (10): Since x = 0 is a zero of ϕ of infinite
order then ϕ(z) = z n Dn ϕ(θz)/n! with 0 < θ < 1 , for any integer n ≥ 1 .
Thus, |ϕ(z)| ≤ kDn ϕk∞ |z|n /n! . Then, we have
Z β Z x/t
|ϕt ∗ a(x)| ≤ kak∞ |ϕt (x − y)|dy ≤ kak∞ |ϕ(z)|dz.
0 0

Therefore,
kak∞ kDn ϕk∞ n+1
|ϕt ∗ a(x)| ≤ (x/t) .
(n + 1)!
6 l. de rosa and c. segovia

Part b) (11) is obvious. As for part c), taking into account (iii) we have
Z "   X k  ` #
−1 x−y −y 1 ` x
(12) ϕt ∗ a(x) = t ϕ − D ϕ a(y)dy.
t t `! t
`=0

Since ϕ(x) ∈ S and the support of ϕ(x) is contained in [0, ∞) , given any non-
negative integers u and v we have
 k+1 
D ϕ(z) ≤ ck,u,v,ϕ z u /(1 + z)v ,
for z ≥ 0 . Then, by Taylor’s formula the integrand in (12) is bounded by
 k+1
1 β (x/t)u
ck,u,v,ϕ kak∞ ,
(k + 1)! t (1 + x/t)v
thus, integrating we get c).

(13) Lemma. Let h(x) be the function defined in (4). Let I = [α, β] and Ie =
[α, β + β − α] . Then,
a) if x ≤ α then ht ∗ χIe (x) = 0 ,
b) if α ≤ x ≤ β + β − α and x − α ≤ 2t then
 r+1
3−s x−α
ht ∗ χIe (x) ≥ ,
r+1 t

c) if α ≤ x ≤ β + β − α and x − α ≥ 2t then
ht ∗ χIe (x) ≥ 3−s ,

and
d) if β + β − α ≤ x then


x−α r
χ −r β − α t
ht ∗ Ie (x) ≥ 2
s .
t 1 + x−α
t

Proof. Without loss of generality we can assume that α = 0 . Part a) is obvious.

If 0 ≤ x ≤ 2β and x ≤ 2t , we have
Z 2β Z x/t
ht ∗ χIe (x) = ht (x − y)dy = h(z)dz
0 0
Z x/t
zr 3−s  x r+1
= dz ≥ ,
0 (1 + z)s r+1 t
 a substitute of harmonic majorization 7

which proves b). Let 0 ≤ x ≤ 2β and x ≥ 2t . We have

Z x/t Z 2
zr
ht ∗ χIe (x) = h(z)dz ≥ s
dz ≥ 3−s ,
0 1 (1 + z)
which proves c). Finally, if x ≥ 2β ,
Z 2β Z β
χ
ht ∗ Ie (x) = ht (x − y)dy ≥ ht (x − y)dy
0
 0 r
Z x/t x−β
β t −r β (x/t)r
= h(z)dz ≥
s ≥ 2 ,
x−β
t
t 1 + xt t (1 + x/t)s

which proves part d).

From Lemmas (9) and (13) we get the following proposition:

(14) Proposition. Let I , a(x) and ϕ(x) as in Lemma (9) and, h(x) and Ie,
as in Lemma (13). If u = r(k + 2) , v = s(k + 2) and n = (r + 1)(k + 2) − 1 , we
have
h ik+2
(15) |ϕt ∗ a(x)| ≤ cr,s,k,ϕ kak∞ ht ∗ χIe (x) .

Proof. It is obvious that (15) holds if x ≤ α . Let α ≤ x ≤ β + β − α and

x − α ≤ 2t . By (10) and part b) of Lemma (13), we have

(r + 1)k+2 s(k+2) n+1 h ik+2

|ϕt ∗ a(x)| ≤ 3 kD ϕk∞ kak∞ ht ∗ χIe (x) ,
(n + 1)!
and therefore (15) holds in this case. If α ≤ x ≤ β + β − α and x − α ≥ 2t , by
(11) and part c) of Lemma (13) it follows that
h ik+2
|ϕt ∗ a(x)| ≤ 3s(k+2) kϕk1 kak∞ ht ∗ χIe (x) ,

proving (15) also in this case. Finally if x ≥ β + β − α by part c) of Lemma (9)

and part d) of Lemma (13), we get
 k+2 x−α u


β−α t
|ϕt ∗ a(x)| ≤ ck,u,v,ϕ kak∞
v
t 1 + x−α
t
h ik+2
≤ ck,r,s,ϕ 2r(k+2) kak∞ ht ∗ χIe (x) ,
8 l. de rosa and c. segovia

ending the proof.

∗
Let f (x) be a bounded function with bounded support and let f−,k+1 (x) be
the maximal function defined in (3). This function is finite everywhere on IR . The
sets
∗
Ωi = {x : f−,k+1 (x) > 2i }, i ∈ ZZ
are open and bounded. By Theorem (2.2) of [RS], if Iij stand for the connected
components of Ωi , there exist functions aij (x) such that

(i) kaij k∞ ≤ c,
(ii) the support of aij is contained in Iij ,
(iii) for every integer `, 0 ≤ ` ≤ k,
Z
(16)
x` aij (x)dx = 0, and
∞
X X
(iv) f (x) = 2i aij (x).
−∞ j

It is easy to see that for any ϕ(x) ∈ S

Z ∞
X XZ
i
(v) f (x)ϕ(x)dx = 2 aij (x)ϕ(x)dx .
−∞ j
Now, we can prove Theorem (5).
Proof. Let ϕ(x) belong to S and support of ϕ(x) contained in [0, ∞) . By
Proposition (14) the functions aij (x) described in (16) satisfy
 k+2
|aij ∗ ϕt (x)| ≤ c χ Ie ∗ ht (x) .
ij

Then, by (v), we get

 k+2
X X
|f ∗ ϕt (x)| ≤ c  2i/(k+2) χ ∗ ht (x) .
Ieij
i j

If we define X X
g(x) = 2i/(k+2) χ (x),
Ieij
i j

we have
k+2
|f ∗ ϕt (x)| ≤ c (g ∗ ht (x)) ,
 a substitute of harmonic majorization 9

which proves (6). On the other hand, by Lemma (8), it follows that

X
X
kgkLq (w) ≤ c
2i/(k+2) χ Iij
.
i j
Lq (w)

Then, taking into account that

X 21/k+2
2i/k+2 χ Ωi (x) ≤ f∗ (x)1/k+2 ,
i
21/k+2 − 1 −,k+1

we obtain the inequality (7).

Application. Let λ > 1 and r > 0 . We define a right hand sided version of
the Tλ,r function of Fefferman and Stein, see [FS], as

Z Z !1/r
δ x
−
Tλ,r,ϕ f (x) = sup δ −λ tλ−2 |f ∗ ϕt (y)|r dydt ,
δ>0 0 x−δ

where |ϕ(x)| ≤ c(1 + |x|)1+ε , ε > 0 , and the support of ϕ(x) is contained in
[0, ∞) .
Proceeding as in [MW] we obtain that if λ < r
 
−
(17) Tλ,r,ϕ f (x) ≤ cM − |f |r/λ (x)λ/r

holds for every x ∈ IR . If λ ≥ r , let k be a non-negative integer such that

λ < r(k + 2) . By Theorem (5) we get

Z Z !1/r
δ x
− −λ λ−2 r(k+2)
Tλ,r,ϕ f (x) ≤ c sup δ t [g ∗ ht (y)] dydt
δ>0 0 x−δ
h ik+2 h ik+2
− −
= c Tλ,r(k+2),h g(x) = c Tλ,s,h g(x) ,

with s = r(k + 2) . Since λ < s , by (17) with s instead of r , we have

h   iλ/r
−
Tλ,r,ϕ f (x) ≤ c M − g s/λ (x) .
 10 l. de rosa and c. segovia

Thus Z h Z h
ip   i λp
r
−
Tλ,r,ϕ f (x) w(x)dx ≤ c M − g s/λ (x) w(x)dx
Z
≤ c g(x)(k+2)p w(x)dx,

λp
if r > 1 and w ∈ A−
λp/r . Taking into account (7) we obtain the inequality

Z Z
p ∗
[Tλ,r,ϕ f (x)] w(x)dx ≤ c f−,k+1 (x)p w(x)dx.

Thus, we have proved the following theorem which generalizes a result of B. Muck-
enhoupt and R.L.Wheeden [MW].

λ −
(18) Theorem. Let p > λr , k + 2 > r and w ∈ Aλp/r . If ϕ(x) ∈ S with
support contained in [0, ∞) , then
Z h ip Z
− ∗
Tλ,r,ϕ f (x) w(x)dx ≤ c f−,k+1 (x)p w(x)dx

holds with a finite constant c depending on λ, r, ϕ, w, k and p .

References
p
[FS] C.Fefferman and E.M.Stein, H spaces of several variables, Acta Math. 129 (1972), 137-193.
[GR] J.Garcı́a-Cuerva and J.L.Rubio de Francia, Weighted Norm Inequalities and Related Topics,
North-Holland, Amsterdam, 1985.
[MW] B.Muckenhoupt and R.L.Wheeden, Norm inequalities for the Littlewood-Paley function gλ∗ ,
Trans. Amer. Math. Soc. 191 (1974), 95-111.
[RS] L.de Rosa and C.Segovia, Weighted H p spaces for one sided maximal functions, Contem-
porary Mathematics (AMS series) 189 (1995), 161-183.
[S] E.Sawyer, Weighted inequalities for the one sided Hardy-Littlewood maximal functions, Trans.
Amer. Math. Soc. 297 (1986), 53-61.
[ST] J.O.Strömberg and A.Torchinsky, Weighted Hardy Spaces, Lecture Notes in Math. 1381,
Springer-Verlag, 1989.
[SW] E.M.Stein and G.Weiss, On the theory of harmonic functions of several variables, I. The
theory of H p spaces. Acta Math. 103 (1960), 25-62.

Department of Mathematic, Univeristy of Buenos Aires, (1428)-Buenos Aires, Ar-

gentina
Current address: Instituto Argentino de Matemática, Viamonte 1636, (1055)-Buenos Aires,
Argentina.
E-mail addresses: segovia@iamba.edu.ar - lderosa@mate.dm.uba.ar