betaFeedback

WorksMathematics

Sharp weighted bounds for reverse fractional integral operators

Abstract

In this paper, we prove the one-weight reverse fractional integral operators ∥Isf∥Lq(wq)≥Cw∥f∥Lp(wp) \|I_s f\|_{L^q(w^q)}\geq C_w\|f\|_{L^p(w^p)} holds for all nonnegative measurable functions ff, with ns<p<1\frac ns<p<1 and 1q=1p−sn\frac1q=\frac1p-\frac sn, if and only if wp/(1−p)∈Ap(s−n)n(1−p)w^{p/(1-p)}\in A_{\frac{p(s-n)}{n(1-p)}}. Furthermore, we show the lower bound Cw≳n,s,p[w]p,qmin⁡{p(n−s)n(1−p),p(n−s)sp−n}, C_w\gtrsim_{n,s,p}[w]_{p,q}^{\min\left\{\frac{p(n-s)}{n(1-p)},\frac{p(n-s)}{sp-n}\right\}}, and the exponent is sharp. We also investigate when the reverse weak-type estimate ∥Isf∥Lnn−s,∞(wnn−s)≥ns vn(n−s)/n[wnn−s]A1(n−s)/n∥f∥L1(w) \|I_s f\|_{L^{\frac{n}{n-s},\infty}(w^{\frac{n}{n-s}})}\geq \frac{n}{s}\,v_n^{(n-s)/n}[w^{\frac{n}{n-s}}]^{(n-s)/n}_{A_1}\|f\|_{L^1(w)} holds if and only if wnn−s∈A1w^{\frac{n}{n-s}}\in A_1.

Introduction

The classical fractional integral operators (or Riesz potentials) IαI_\alpha are defined by

Iα(f)(x)=γα∫Rnf(x−y)∣y∣n−α dy,I_\alpha(f)(x)=\gamma_\alpha\int_{\mathbb{R}^n}\frac{f(x-y)}{|y|^{n-\alpha}}\,dy,

where 0<α<n0<\alpha<n and γα=2−απ−n/2Γ(n−α2)Γ(α2)\gamma_\alpha=2^{-\alpha}\pi^{-n/2}\frac{\Gamma\left(\frac{n-\alpha}{2}\right)}{\Gamma\left(\frac{\alpha}{2}\right)}. This family of fractional potential operators was first systematically investigated by Riesz [32]. The (Lp,Lq)(L^p,L^q)-boundedness of Riesz potentials was proved by Hardy and Littlewood [9] in the one-dimensional case, and by Sobolev [35] in higher dimensions. The endpoint weak-type (L1,Ln/(n−α),∞)(L^1,L^{n/(n-\alpha),\infty}) boundedness was subsequently obtained by Zygmund [39]. The classical result is the following.

Theorem A. Let 0<α<n0<\alpha<n and let p,qp,q satisfy 1≤p<q<∞1\le p<q<\infty and 1p−1q=αn\frac{1}{p}-\frac{1}{q}=\frac{\alpha}{n}. Then for p>1p>1,

∥Iα(f)∥Lq(Rn)≤C(n,p,α)∥f∥Lp(Rn).\|I_\alpha(f)\|_{L^q(\mathbb{R}^n)}\le C(n,p,\alpha)\|f\|_{L^p(\mathbb{R}^n)}.

When p=1p=1,

∥Iα(f)∥Ln/(n−α),∞(Rn)=sup⁡λ>0λ∣{x∈Rn:∣Iαf∣>λ}∣(n−α)/n≤C(n,α)∥f∥L1(Rn).\|I_\alpha(f)\|_{L^{n/(n-\alpha),\infty}(\mathbb{R}^n)} =\sup_{\lambda>0}\lambda\left|\{x\in\mathbb{R}^n:|I_\alpha f|>\lambda\}\right|^{(n-\alpha)/n} \le C(n,\alpha)\|f\|_{L^1(\mathbb{R}^n)}.

Lieb [20] calculated the sharp constant in the conformally symmetric case p=2nn+αp = \frac{2n}{n+\alpha}, q=2nn−αq = \frac{2n}{n-\alpha} (see also [8]). More recently, Tang and Wang [37] determined sharp weak-type constants for classical fractional integral operators.

The study of reverse fractional integral operators was initiated by Dou and Zhu [6], who proved the existence of extremal functions and computed the best constant for p=q=2nn+sp = q = \frac{2n}{n+s}. Chen, Liu, Lu and Tao [3] established the reverse weighted Hardy–Littlewood–Sobolev inequality, namely the reverse Stein–Weiss inequality,

∫Rn∫Rn∣x∣α∣x−y∣s−nf(x)g(y)∣y∣β dx dy≥Cn,α,β,p,q′∥f∥Lq′∥g∥Lp\int_{\mathbb{R}^{n}}\int_{\mathbb{R}^{n}} |x|^{\alpha}|x-y|^{s-n}f(x)g(y)|y|^{\beta}\,dx\,dy \ge C_{n,\alpha,\beta,p,q'}\|f\|_{L^{q'}}\|g\|_{L^{p}}

for any nonnegative functions f∈Lq′(Rn)f \in L^{q'}(\mathbb{R}^{n}), g∈Lp(Rn)g \in L^{p}(\mathbb{R}^{n}), and p,q′∈(0,1)p,q' \in(0,1), α,β,s−n>0\alpha,\beta,s-n>0 such that

1p+1q′−α+β+s−nn=2.\frac{1}{p}+\frac{1}{q'}-\frac{\alpha+\beta+s-n}{n}=2.

Kassymov, Ruzhansky and Suragan [16] subsequently extended the reverse weighted and unweighted Hardy–Littlewood–Sobolev inequality to homogeneous groups. For the endpoint case, the present authors jointly with Zhang [11] established the sharp reverse weak estimates and the extremal functions for the reverse fractional integral operators IsI_s. For related reverse Hardy–Littlewood–Sobolev inequality, we refer the reader to [2, 23].

In the weighted setting, Stein and Weiss [36] established the following weighted Hardy–Littlewood–Sobolev inequality (also called the Stein–Weiss inequality):

∫Rn∫Rn∣x∣−α∣x−y∣−λf(x)g(y)∣y∣−β dx dy≤Cn,α,β,p,q∥f∥Lp(Rn)∥g∥Lq(Rn),(1)\int_{\mathbb{R}^{n}}\int_{\mathbb{R}^{n}} |x|^{-\alpha}|x-y|^{-\lambda}f(x)g(y)|y|^{-\beta}\,dx\,dy \le C_{n,\alpha,\beta,p,q}\|f\|_{L^{p}(\mathbb{R}^{n})}\|g\|_{L^{q}(\mathbb{R}^{n})}, \tag*{(1)}

where 1<p,q<∞1<p,q<\infty, α,β,λ\alpha,\beta,\lambda satisfy

1p+1q+α+β+λn=2,1p+1q≥1,\frac{1}{p}+\frac{1}{q}+\frac{\alpha+\beta+\lambda}{n}=2,\qquad\frac{1}{p}+\frac{1}{q}\ge1,
α+β≥0,α<np′,β<nq′,0<λ<n.\alpha+\beta\ge0,\qquad\alpha<\frac{n}{p'},\qquad\beta<\frac{n}{q'},\qquad0<\lambda<n.

The ApA_p weight theory began with Muckenhoupt’s theorem [21]: the Hardy–Littlewood maximal operator MM is bounded on Lp(w)L^p(w) for 1<p<∞1<p<\infty if and only if the weight ww satisfies the ApA_p condition

[w]Ap:=sup⁡Q(1∣Q∣∫Qw)(1∣Q∣∫Qw−1/(p−1))p−1<∞.(2)[w]_{A_p}:=\sup_Q\left(\frac{1}{|Q|}\int_Q w\right)\left(\frac{1}{|Q|}\int_Q w^{-1/(p-1)}\right)^{p-1}<\infty. \tag*{(2)}

Hunt, Muckenhoupt, and Wheeden [12] proved that the ApA_p condition also characterizes the boundedness of the Hilbert transform

Hf(x)=p.v.⁡1π∫f(y)x−y dyHf(x)=\operatorname{p.v.}\frac{1}{\pi}\int\frac{f(y)}{x-y}\,dy

on Lp(w)L^p(w). Coifman and Fefferman [4] developed the general theory for Calderón–Zygmund operators and introduced the class A∞=⋃p>1ApA_\infty=\bigcup_{p>1}A_p. Muckenhoupt and Wheeden [22] characterized the boundedness of the fractional integral operators as follows:

(∫Rn∣Iαf(x)w(x)∣q dx)1/q≤C(∫Rn∣f(x)w(x)∣p dx)1/pfor all f≥0.(3)\left(\int_{\mathbb{R}^{n}}|I_\alpha f(x)w(x)|^q\,dx\right)^{1/q}\le C\left(\int_{\mathbb{R}^{n}}|f(x)w(x)|^p\,dx\right)^{1/p}\quad\text{for all }f\ge0. \tag*{(3)}

where 1/q=1/p−α/n1/q = 1/p - \alpha/n, holds if and only if [w]p,q<∞[w]_{p,q} < \infty, where the weight characteristic is defined as

[w]p,q=sup⁡Q(1∣Q∣∫Qwq dx)(1∣Q∣∫Qw−p′ dx)q/p′<∞.[w]_{p,q} = \sup_Q \left(\frac{1}{|Q|}\int_Q w^q\,dx\right)\left(\frac{1}{|Q|}\int_Q w^{-p'}\,dx\right)^{q/p'} < \infty.

The [w]p,q[w]_{p,q} condition was first introduced in [22]. Sawyer [29] first gave the two-weight characterization for the maximal operator, and then set up the two-weight bounds for fractional and Poisson integrals [30]. Sawyer and Wheeden [31] extended the two-weight characterization of fractional integrals to Euclidean and homogeneous spaces. Pérez established two-weight norm inequalities for Riesz potentials [27], and subsequently for potential operators and fractional maximal operators [28].

Once the weighted boundedness of the operator is obtained, it is natural to consider how the operator bound depends on the weight. Buckley [1] proved that for the maximal operator,

∥M∥Lp(w)→Lp(w)≤c [w]Ap1/(p−1).\|M\|_{L^p(w)\to L^p(w)} \leq c\,[w]_{A_p}^{1/(p-1)}.

and that the exponent 1/(p−1)1/(p-1) is sharp. For singular integral operators, the central open problem was so-called A2A_2 conjecture, which asserts that for every Calderón–Zygmund operator TT,

∥T∥L2(w)→L2(w)≤CT[w]A2.\|T\|_{L^2(w)\to L^2(w)} \leq C_T [w]_{A_2}.

Petermichl [25, 26] showed that

∥T∥Lp(w)→Lp(w)≤c [w]Apmax⁡{1,1/(p−1)},\|T\|_{L^p(w)\to L^p(w)} \leq c\,[w]_{A_p}^{\max\{1,1/(p-1)\}},

where TT is either the Hilbert transform or one of the Riesz transforms

Rjf(x)=cnp.v.⁡∫Rnxj−yj∣x−y∣n+1f(y) dy.R_j f(x) = c_n \operatorname{p.v.}\int_{\mathbb{R}^n} \frac{x_j-y_j}{|x-y|^{n+1}} f(y)\,dy.

The A2A_2 conjecture was resolved for general Calderón–Zygmund operators by Hytönen [13]. This landmark proof was subsequently simplified by Lerner [19], who gave an elegant argument via sparse pointwise domination. The sharp extrapolation theorem of Dragičević, Grafakos, Pereyra, and Petermichl [7] extended the sharp p=2p=2 result to all 1<p<∞1 < p < \infty, yielding

∥T∥Lp(w)→Lp(w)≲[w]Apmax⁡(1,1/(p−1)).\|T\|_{L^p(w)\to L^p(w)} \lesssim[w]_{A_p}^{\max(1,1/(p-1))}.

Furthermore, the sharp quantitative reverse Hölder theorem of Hytönen, Pérez, and Rela [15] and the sharp Fujii–Wilson A∞A_\infty bounds of Hytönen and Pérez [14] have also been established. Recently, Lacey, Moen, Pérez, and Torres [17] obtained the corresponding sharp bounds for weighted fractional integral operators IαI_\alpha:

Theorem B. Let 1<p<n/α1 < p < n/\alpha and qq be defined by the equation 1/q=1/p−α/n1/q = 1/p - \alpha/n, and [w]p,q<∞[w]_{p,q} < \infty. Then,

∥Iα∥Lp(wp)→Lq(wq)≤c [w]p,q(1−α/n)max⁡{1,p′/q}.\|I_\alpha\|_{L^p(w^p)\to L^q(w^q)} \leq c\,[w]_{p,q}^{(1-\alpha/n)\max\{1,p'/q\}}.

Furthermore this estimate is sharp.

Further one-weight and two-weight results for fractional integral operators were obtained by Cruz-Uribe and Moen [5], who extend Theorem B to

∥Iα∥Lp(wp)→Lq(wq)≲[w]Ap1/p([w]A∞1/p′+[w1−p′]A∞1/p).\|I_{\alpha}\|_{L^p(w^p)\to L^q(w^q)} \lesssim[w]_{A_p}^{1/p} \left([w]_{A_\infty}^{1/p'}+[w^{1-p'}]_{A_\infty}^{1/p}\right).

These existing results largely focus on power-weight settings, leaving open the general Muckenhoupt type one-weight characterization and sharp quantitative bounds for the best constant in reverse fractional integral operators. In the present paper we address two main questions for the reverse fractional integral operators IsI_s with s>ns>n:

  1. What weight condition characterizes the boundedness of the reverse fractional integral operator, and what is the sharp dependence of the best constant on that weight?

  1. What is the corresponding weight condition for endpoint boundedness?

The reverse fractional integral operators IsI_s are defined by

Is(f)(x)=∫Rn∣x−y∣n−sf(y) dy.I_s(f)(x)=\int_{\mathbb{R}^n}|x-y|^{n-s}f(y)\,dy.

where s>ns>n. Throughout the strong-type discussion, set p′=p/(p−1)<0p'=p/(p-1)<0 and q′=q/(q−1)∈(0,1)q'=q/(q-1)\in(0,1). Define the scale-invariant weight characteristic

[w]p,q:=sup⁡B(1∣B∣∫Bw−p′ dx)−1/p′(1∣B∣∫Bwq dx)−1/q.[w]_{p,q}:=\sup_B\left(\frac{1}{|B|}\int_B w^{-p'}\,dx\right)^{-1/p'}\left(\frac{1}{|B|}\int_B w^q\,dx\right)^{-1/q}.

Here the supremum runs over all Euclidean balls B⊂RnB\subset\mathbb{R}^n. By direct computation,

[w−p′]A1+p′/q=[w]p,q−p′,(4)[w^{-p'}]_{A_{1+p'/q}}=[w]_{p,q}^{-p'}, \tag*{(4)}

so that [w]p,q<∞[w]_{p,q}<\infty is equivalent to w−p′∈A1+p′/qw^{-p'}\in A_{1+p'/q}. Our first main theorem provides the qualitative one-weight characterization.

Theorem 1.1. Let n≥1n\ge1, s>ns>n, n/s<p<1n/s<p<1, and let q<0q<0 satisfy 1/q=1/p−s/n1/q=1/p-s/n. Let w>0w>0 almost everywhere. Then there exists a constant C>0C>0 such that

∥Isf∥Lq(wq)≥C∥f∥Lp(wp)(5)\|I_sf\|_{L^q(w^q)}\ge C\|f\|_{L^p(w^p)} \tag*{(5)}

for every nonnegative measurable ff if and only if [w]p,q<∞[w]_{p,q}<\infty.

Remark 1. Let w(x)=∣x∣γw(x)=|x|^\gamma. Then (1.4) holds if and only if

n−np<γ<s−np.(6)n-\frac{n}{p}<\gamma<s-\frac{n}{p}. \tag*{(6)}

In fact, by Theorem 1.1, the required condition is ∣x∣−p′γ∈A1+p′/q|x|^{-p'\gamma}\in A_{1+p'/q}, then we have

−n<−p′γ<n(1+p′/q−1)=np′q.-n<-p'\gamma<n(1+p'/q-1)=\frac{np'}{q}.

Thus n−np=n/p′<γ<−n/q=s−npn-\frac{n}{p}=n/p'<\gamma<-n/q=s-\frac{n}{p}. The range (1.5) agrees with the one-weight specialization of the reverse Stein-Weiss parameter range in [3]. This gives an independent consistency check on the Muckenhoupt characterization.

Next, we obtain the sharp weighted bounds for the reverse fractional integral operators:

Theorem 1.2. Let n≥1n \ge1, s>ns > n, n/s<p<1n/s < p < 1, and let q<0q < 0 satisfy 1/q=1/p−s/n1/q = 1/p - s/n and [w]p,q<∞[w]_{p,q} < \infty. Then,

∥Is∥Lp(wp)→Lq(wq)≥cn,s,p[w]p,qmin⁡{p(n−s)n(1−p),p(n−s)sp−n}.\lVert I_s \rVert_{L^p(w^p)\to L^q(w^q)} \ge c_{n,s,p} [w]_{p,q}^{\min\left\{\frac{p(n-s)}{n(1-p)},\frac{p(n-s)}{sp-n}\right\}}.

Furthermore this estimate is sharp.

Remark 2. Through the proof of Theorem 1.2, we have the following better bounds,

C∗(w)≥cn,s,p[w]p,q−1min⁡{[w−p′]A∞1/q,[wq]A∞1/p′}.(7)C_*(w) \ge c_{n,s,p} [w]_{p,q}^{-1} \min\left\{[w^{-p'}]_{A_\infty}^{1/q}, [w^q]_{A_\infty}^{1/p'}\right\}. \tag*{(7)}

Combining (7), (1.3), the inequalities

[w−p′]A∞≲n,p,q[w−p′]A1+p′/qand[wq]A1+q/p′=[w−p′]A1+p′/qq/p′,[w^{-p'}]_{A_\infty} \lesssim_{n,p,q} [w^{-p'}]_{A_1+p'/q} \quad\text{and}\quad [w^q]_{A_1+q/p'} = [w^{-p'}]_{A_1+p'/q}^{q/p'},

we obtain the desired lower bound in Theorem 1.2.

The critical point is pc=2nn+sp_c = \frac{2n}{n+s}, where the two symmetric terms coincide, giving p′=qp' = q. Thus

Γ∗={p(s−n)sp−n,ns<p≤2nn+s,p(s−n)n(1−p),2nn+s≤p<1.\Gamma_* = \begin{cases} \frac{p(s-n)}{sp-n}, & \frac{n}{s} < p \le\frac{2n}{n+s}, \\ \frac{p(s-n)}{n(1-p)}, & \frac{2n}{n+s} \le p < 1. \end{cases}

At the conformally symmetric reverse Hardy-Littlewood-Sobolev exponent p=2n/(n+s)p = 2n/(n+s), the sharp exponent is

C∗(w)≳[w]p,q−2.C_*(w) \gtrsim[w]_{p,q}^{-2}.

At the endpoint p=1p = 1, we obtain a lower bound estimate of the reverse fractional integral operators.

Theorem 1.3. Let n≥1n \ge1, s>ns > n. The inequality

∥Isf∥Lnn−s,∞(wnn−s)≥nsvnn−sn[wnn−s]A1n−sn∥f∥L1(w)\lVert I_s f \rVert_{L^{\frac{n}{n-s},\infty}(w^{\frac{n}{n-s}})} \ge\frac{n}{s} v_n^{\frac{n-s}{n}} [w^{\frac{n}{n-s}}]_{A_1}^{\frac{n-s}{n}} \lVert f \rVert_{L^1(w)}

holds for all f≥0f \ge0 with a positive constant if and only if wnn−s∈A1w^{\frac{n}{n-s}} \in A_1.

For two quantities XX, YY depending on parameters pp, ss, nn, . . . , the notation X≲YX \lesssim Y (resp. X≳YX \gtrsim Y) means X≤CYX \le CY (resp. X≥cYX \ge cY) for constants cc, C>0C > 0 that depend only on the indicated subscripts; X≈YX \approx Y means both X≲YX \lesssim Y and X≳YX \gtrsim Y.

The paper is organised as follows. Section 2 collects facts on ApA_p and A∞A_\infty weights, the bilinear reduction, and some lemmas. Section 3 proves necessity of the Muckenhoupt condition. In Section 4, a negative-index duality argument reduces the lower bound to a bilinear case, and we prove Theorems 1.1 and 1.2. We use power-weight to show that the exponent is sharp. Section 5 treats the endpoint p=1p = 1 and proves Theorem 1.3 by the layer-cake identity, the A1A_1 condition, and Tonelli’s theorem, while necessity follows by testing on approximate point masses and applying the Lebesgue differentiation.

Preliminaries

In this section, we recall facts on ArA_r and A∞A_\infty weights and establish the bilinear and geometric lemmas used below.

In the main theorems, ww is positive and finite almost everywhere; local integrability of ww itself is not required. We use w(E)=∫Ew(x) dxw(E)=\int_E w(x)\,dx when this integral is defined. For 1<r<∞1<r<\infty, a positive, locally integrable function uu belongs to ArA_r if

[w]Ar:=sup⁡B(1∣B∣∫Bw)(1∣B∣∫Bw−1/(r−1))r−1<∞.[w]_{A_r}:=\sup_B\left(\frac{1}{\lvert B\rvert}\int_B w\right)\left(\frac{1}{\lvert B\rvert}\int_B w^{-1/(r-1)}\right)^{r-1}<\infty.

Ball and cube definitions yield the same ArA_r class, with comparison constants depending on the dimension and the exponent, see [21].

The dual weight is

w∗=w−1/(r−1),w^*=w^{-1/(r-1)},

and

[w∗]Ar′=[w]Ar1/(r−1).(8)[w^*]_{A_{r'}}=[w]_{A_r}^{1/(r-1)}. \tag*{(8)}

Introducing the scale-invariant characteristic

[w]p,q=sup⁡B(1∣B∣∫Bw−p′ dx)−1/p(1∣B∣∫Bwq dx)−1/q.[w]_{p,q}=\sup_B\left(\frac{1}{\lvert B\rvert}\int_B w^{-p'}\,dx\right)^{-1/p}\left(\frac{1}{\lvert B\rvert}\int_B w^q\,dx\right)^{-1/q}.

Equivalently, [w]p,q<∞[w]_{p,q}<\infty means

w−p′=wp′/(1−p)∈A1+p′/q=Ap(s−n)/(n(1−p)).w^{-p'}=w^{p'/(1-p)}\in A_{1+p'/q}=A_{p(s-n)/(n(1-p))}.

It is well known that A∞=⋃r>1ArA_\infty=\bigcup_{r>1}A_r. Equivalently, we use the Fujii-Wilson characteristic

[w]A∞:=sup⁡B1w(B)∫BM(wχB)(x) dx,(9)[w]_{A_\infty}:=\sup_B\frac{1}{w(B)}\int_B M(w\chi_B)(x)\,dx, \tag*{(9)}

where MM is the uncentered Hardy–Littlewood maximal operator. Equivalent A∞A_\infty characteristics are discussed in, for example, [38, 14]. The crucial quantitative fact is the sharp reverse-Hölder theorem of Hytönen–Pérez–Rela [15]. In the Euclidean setting it implies that there are dimensional constants cn,Cn>0c_n,C_n>0 such that, for every w∈A∞w\in A_\infty, every ball BB, and every measurable E⊂BE\subset B,

w(E)w(B)≤Cn(∣E∣∣B∣)δw,δw≥cn[w]A∞.(10)\frac{w(E)}{w(B)}\le C_n\left(\frac{\lvert E\rvert}{\lvert B\rvert}\right)^{\delta_w},\qquad\delta_w\ge\frac{c_n}{[w]_{A_\infty}}. \tag*{(10)}

For q<0q<0 and a positive measurable weight ww, define

∥f∥Lq(w):=(∫Rnf(x)qw(x) dx)1/q,\lVert f\rVert_{L^q(w)}:=\left(\int_{\mathbb{R}^n}f(x)^q w(x)\,dx\right)^{1/q},

with the standard extended-value convention. The LpL^p norm for p<1p<1 is not a norm for a vector space. For q<0q<0, and a weight ww, the reverse weak-type quasi-norm is

∥f∥Lq,∞(w):=inf⁡λ>0λw({x:f(x)<λ})1/q,\lVert f\rVert_{L^{q,\infty}(w)}:=\inf_{\lambda>0}\lambda w(\{x:f(x)<\lambda\})^{1/q},

where the sublevel set uses f<λf<\lambda.

For q<0<p≤1q<0<p\le1, we say an operator TT is of weak-type (p,q)(p,q), if there exists a constant C(p,q,n)>0C(p,q,n)>0 such that ∥f∥Lq,∞≥C(p,q,n)∥f∥Lp\lVert f\rVert_{L^{q,\infty}}\ge C(p,q,n)\lVert f\rVert_{L^p} for all f∈Lp(Rn)f\in L^p(\mathbb{R}^n).

Lemma 2.1 ([6]). For 0<r<10<r<1, put r′=r/(r−1)<0r'=r/(r-1)<0. For nonnegative measurable functions f,gf,g, the reverse Hölder inequality holds whenever the quantities are defined:

∫f(x)g(x) dx≥∥f∥Lr∥g∥Lr′.\int f(x)g(x)\,dx\ge\lVert f\rVert_{L^r}\lVert g\rVert_{L^{r'}}.

We also need the exact dual formula for a negative exponent.

Lemma 2.2. Let q<0q<0 and q′=q/(q−1)∈(0,1)q'=q/(q-1)\in(0,1). For every positive measurable ff,

∥f∥Lq=inf⁡∥g∥Lq′=1, g≥0∫Rnf(x)g(x) dx.\lVert f\rVert_{L^q}=\inf_{\lVert g\rVert_{L^{q'}}=1,\ g\ge0}\int_{\mathbb{R}^n}f(x)g(x)\,dx.

Proof. Reverse Hölder’s inequality with exponents q′q' and qq gives the inequality for every admissible gg. Put A=∫Rnfq(x) dxA=\int_{\mathbb{R}^n}f^q(x)\,dx. If 0<A<∞0<A<\infty, then g=A−1/q′fq−1g=A^{-1/q'}f^{q-1} has ∥g∥q′=1\lVert g\rVert_{q'}=1 and

∫Rnf(x)g(x) dx=A1/q=∥f∥q.\int_{\mathbb{R}^n}f(x)g(x)\,dx=A^{1/q}=\lVert f\rVert_q.

If A=∞A=\infty, choose increasing measurable sets EjE_j for which 0<∫Ejfq(x) dx<∞0<\int_{E_j}f^q(x)\,dx<\infty and ∫Ejfq(x) dx→∞\int_{E_j}f^q(x)\,dx\to\infty. The same construction with fq−1χEjf^{q-1}\chi_{E_j} gives integrals (∫Ejfq(x) dx)1/q→0=∥f∥q(\int_{E_j}f^q(x)\,dx)^{1/q}\to0=\lVert f\rVert_q. □

The strong (p,q)(p,q) inequality for the weighted reverse fractional integral operators can then be equivalently reformulated as a bilinear form controlled from below by a product of norms. Given F(y)=f(y)w(y)F(y)=f(y)w(y), define

SwF(x)=w(x)Is(F/w)(x).S_wF(x)=w(x)I_s(F/w)(x).

Then

∥Isf∥Lq(wq)=∥SwF∥Lq,∥f∥Lp(wp)=∥F∥Lp.\lVert I_sf\rVert_{L^q(w^q)}=\lVert S_wF\rVert_{L^q},\qquad\lVert f\rVert_{L^p(w^p)}=\lVert F\rVert_{L^p}.

By Lemma 2.2, the best strong constant equals the best constant in

Bw(F,G):=∬Rn×Rn∣x−y∣s−nF(y)G(x)w(x)w(y) dy dx≥C∥F∥p∥G∥q′.(11)B_w(F,G):=\iint_{\mathbb{R}^n\times\mathbb{R}^n}|x-y|^{s-n}F(y)G(x)\frac{w(x)}{w(y)}\,dy\,dx\ge C\lVert F\rVert_p\lVert G\rVert_{q'}. \tag*{(11)}

This formulation exhibits the symmetry

Cp,q′(w)=Cq′,p(w−1).(12)C_{p,q'}(w)=C_{q',p}(w^{-1}). \tag*{(12)}

It will be useful in the sharpness argument.

Next, we introduce the following lemma which play an important roles in the proof of the main theorems.

Lemma 2.3. Let ∥F∥Lp=∥G∥Lq′=1\lVert F\rVert_{L^p}=\lVert G\rVert_{L^{q'}}=1 for q<0q<0. Assume [w]p,q<∞[w]_{p,q}<\infty and let B=B(z,r)B=B(z,r). Fix η∈(0,1)\eta\in(0,1). If

∫BF(y)p dy≥η,∫(2B)cG(x)q′ dx≥η,\int_B F(y)^p\,dy\ge\eta,\qquad\int_{(2B)^c}G(x)^{q'}\,dx\ge\eta,

then

Bw(F,G)≥cn,s,p,η[w]p,q−1[w−p′]A∞1/q.(13)B_w(F,G)\ge c_{n,s,p,\eta}[w]_{p,q}^{-1}[w^{-p'}]_{A_\infty}^{1/q}. \tag*{(13)}

If instead

∫BG(x)q′ dx≥η,∫(2B)cF(y)p dy≥η,\int_B G(x)^{q'}\,dx \ge\eta,\qquad\int_{(2B)^c} F(y)^p\,dy \ge\eta,

then

Bw(F,G)≥cn,s,p,η[w]p,q−1[wq]A∞1/p′.\mathcal{B}_w(F,G) \ge c_{n,s,p,\eta}[w]_{p,q}^{-1}[w^q]_{A_\infty}^{1/p'}.

Proof. We prove the first assertion. Decompose

(2B)c=⋃k≥1Ak,Ak=2k+1B∖2kB,(2B)^c=\bigcup_{k\ge1} A_k,\qquad A_k=2^{k+1}B\setminus2^kB,

and set qk=∫AkG(x)q′ dxq_k=\int_{A_k}G(x)^{q'}\,dx. Then

∑k≥1qk≥η.\sum_{k\ge1}q_k\ge\eta.

If y∈By\in B and x∈Akx\in A_k, then

∣x−y∣≥c2kr.|x-y|\ge c2^kr.

Hence

Bw(F,G)≥c∑k≥1(2kr)s−n(∫BF(y)w(y) dy)(∫AkG(x)w(x) dx).\mathcal{B}_w(F,G)\ge c\sum_{k\ge1}(2^kr)^{s-n}\left(\int_B\frac{F(y)}{w(y)}\,dy\right)\left(\int_{A_k}G(x)w(x)\,dx\right).

Reverse Hölder gives

∫BF(y)w(y) dy≥(∫BF(y)p dy)1/p(∫Bw−p′(y) dy)1/p′≥η1/p(∫Bw−p′(y) dy)1/p′,\int_B\frac{F(y)}{w(y)}\,dy\ge\left(\int_BF(y)^p\,dy\right)^{1/p}\left(\int_Bw^{-p'}(y)\,dy\right)^{1/p'}\ge\eta^{1/p}\left(\int_Bw^{-p'}(y)\,dy\right)^{1/p'},

and

∫AkG(x)w(x) dx≥qk1/q′(∫Akwq(y) dy)1/q≥qk1/q′(∫2k+1Bwq(y) dy)1/q.\int_{A_k}G(x)w(x)\,dx\ge q_k^{1/q'}\left(\int_{A_k}w^q(y)\,dy\right)^{1/q}\ge q_k^{1/q'}\left(\int_{2^{k+1}B}w^q(y)\,dy\right)^{1/q}.

Applying the definition of the weight condition on 2k+1B2^{k+1}B yields

Bw(F,G)≥cη1/p∑k≥11qk1/q′(2kr)s−n(∫Bw−p′(y) dy)1/p′(∫2k+1Bwq(y) dy)1/q≥cn,s,p,η[w]p,q−1∑k≥1(w−p′(2k+1B)w−p′(B))−1/p′1qk1/q′.(14)\begin{aligned} \mathcal{B}_w(F,G)&\ge c\eta^{1/p}\sum_{k\ge1}\frac{1}{q_k^{1/q'}}(2^kr)^{s-n}\left(\int_Bw^{-p'}(y)\,dy\right)^{1/p'}\left(\int_{2^{k+1}B}w^q(y)\,dy\right)^{1/q}\\ &\ge c_{n,s,p,\eta}[w]_{p,q}^{-1}\sum_{k\ge1}\left(\frac{w^{-p'}(2^{k+1}B)}{w^{-p'}(B)}\right)^{-1/p'}\frac{1}{q_k^{1/q'}}. \tag*{(14)} \end{aligned}

By the A∞A_\infty estimate (10), we have

w−p′(B)w−p′(2k+1B)≤Cn2−knδu,δu≥cn[w−p′]A∞.\frac{w^{-p'}(B)}{w^{-p'}(2^{k+1}B)}\le C_n2^{-kn\delta_u},\qquad\delta_u\ge\frac{c_n}{[w^{-p'}]_{A_\infty}}.

Therefore

(w−p′(2k+1B)w−p′(B))−1/p′≥cn2−nδup′k,where 2nδup′>1.\left(\frac{w^{-p'}(2^{k+1}B)}{w^{-p'}(B)}\right)^{-1/p'}\ge c_n2^{-\frac{n\delta_u}{p'}k},\qquad\text{where }2^{\frac{n\delta_u}{p'}}>1.

Set ak=2−nδuk/p′a_k=2^{-n\delta_u k/p'}. Since 1/q+1/q′=11/q+1/q'=1 and q<0q<0, discrete reverse Hölder gives

∑k≥1(w−p′(2k+1B)w−p′(B))−1/p′qk1/q′≳n∑k≥1akqk1/q′≥(∑k≥1akq)1/q(∑k≥1qk)1/q′≥η1/q′2−nδu/p′(1−2−nδuq/p′)−1/q.\begin{aligned} \sum_{k\geq1}\left(\frac{w^{-p'}(2^{k+1}B)}{w^{-p'}(B)}\right)^{-1/p'}q_k^{1/q'}\gtrsim_n\sum_{k\geq1}a_kq_k^{1/q'} \\ &\geq\left(\sum_{k\geq1}a_k^q\right)^{1/q}\left(\sum_{k\geq1}q_k\right)^{1/q'} \\ &\geq\eta^{1/q'}2^{-n\delta_u/p'}\left(1-2^{-n\delta_uq/p'}\right)^{-1/q}. \end{aligned}

We may choose 0<δu≤10<\delta_u\leq1 with δu≳n[w−p′]A∞−1\delta_u\gtrsim_n [w^{-p'}]_{A_\infty}^{-1}. Since q/p′>0q/p'>0, the last expression is bounded below by cn,s,p,ηδu−1/q≳n,s,p,η[w−p′]A∞1/qc_{n,s,p,\eta}\delta_u^{-1/q}\gtrsim_{n,s,p,\eta}[w^{-p'}]_{A_\infty}^{1/q}. Together with (2.7), this proves (2.6). The second assertion follows by exchanging (F,p,w)(F,p,w) with (G,q′,w−1)(G,q',w^{-1}); the weight characteristic is unchanged under this exchange.

The following lemma guarantees that such separated functions in Lemma 2.3 always hold.

Lemma 2.4. Let 0<p<10<p<1, 0<q′<∞0<q'<\infty, and let F,G≥0F,G\geq0 be measurable functions on Rn\mathbb{R}^n with

∫RnF(y)p dy=1,∫RnG(x)q′ dx=1.\int_{\mathbb{R}^n}F(y)^p\,dy=1,\qquad\int_{\mathbb{R}^n}G(x)^{q'}\,dx=1.

And there is a dimensional constant ηn>0\eta_n>0 such that at least one of the following alternatives holds for some ball BB:

∫BFp(y) dy≥ηn,∫(2B)cGq′(x) dx≥12;(15)\int_B F^p(y)\,dy\geq\eta_n,\qquad\int_{(2B)^c}G^{q'}(x)\,dx\geq\frac{1}{2}; \tag*{(15)}
∫BGq′(x) dx≥ηn,∫(2B)cFp(y) dy≥12.(16)\int_B G^{q'}(x)\,dx\geq\eta_n,\qquad\int_{(2B)^c}F^p(y)\,dy\geq\frac{1}{2}. \tag*{(16)}

Proof. For a nonnegative integrable function HH with ∫RnH(y) dy=1\int_{\mathbb{R}^n}H(y)\,dy=1, define

cH(r)=sup⁡z∈Rn∫B(z,r)H(y) dy,rH=inf⁡{r>0:cH(r)≥1/2}.c_H(r)=\sup_{z\in\mathbb{R}^n}\int_{B(z,r)}H(y)\,dy,\qquad r_H=\inf\{r>0:c_H(r)\geq1/2\}.

Since H∈L1H\in L^1, absolute continuity of the integral gives cH(r)→0c_H(r)\to0 as r↓0r\downarrow0, uniformly in the center. Since ∫RnH(y) dy=1\int_{\mathbb{R}^n}H(y)\,dy=1, there holds cH(r)→1c_H(r)\to1 as r→∞r\to\infty. Hence 0<rH<∞0<r_H<\infty, and by definition every ball BB of radius strictly smaller than rHr_H satisfies ∫BH(y) dy<1/2\int_B H(y)\,dy<1/2.

We claim there is a ball BHB_H of radius 2rH2r_H with ∫BHH≥1/3\int_{B_H}H\geq1/3. Indeed, choose rj↓rHr_j\downarrow r_H with rj<2rHr_j<2r_H and cH(rj)≥1/2c_H(r_j)\geq1/2. For each jj there is a center zjz_j with ∫B(zj,rj)H>1/3\int_{B(z_j,r_j)}H>1/3, and B(zj,2rH)⊃B(zj,rj)B(z_j,2r_H)\supset B(z_j,r_j).

Apply this to H=FpH=F^p and H=Gq′H=G^{q'}, obtaining balls BF,BGB_F,B_G of radii 2rF,2rG2r_F,2r_G respectively, each carrying mass ≥1/3\geq1/3. Choose L=16L=16.

Case 1: rF≤rG/Lr_F\leq r_G/L. Then 2BF2B_F has radius 4rF<rG4r_F<r_G, so ∫2BFGq′<1/2\int_{2B_F}G^{q'}<1/2 while ∫BFFp≥1/3\int_{B_F}F^p\geq1/3. Thus (2.8) holds with B=BFB=B_F and ηn=1/3\eta_n=1/3.

Case 2: rG≤rF/Lr_G\leq r_F/L. Symmetrically, (2.9) holds with B=BGB=B_G. Case 3: L−1rG<rF<LrGL^{-1}r_G<r_F<Lr_G. Cover BFB_F, whose radius is 2rF2r_F, by at most NnN_n balls of radius rG/4r_G/4, where NnN_n depends only on nn and LL. Such a bound follows by taking a maximal rG/4r_G/4-separated set of centers in BFB_F: the radius-rG/8r_G/8 balls are disjoint and lie in a ball of controlled radius, so their number is bounded by a volume ratio. Since ∫BFFp(y) dy≥1/3\int_{B_F}F^p(y)\,dy\ge1/3, one of these covering balls, say B∗B^*, satisfies

∫B∗Fp≥13Nn.\int_{B^*}F^p\ge\frac{1}{3N_n}.

The radius of 2B∗2B^* is rG/2<rGr_G/2<r_G, hence ∫2B∗Gq′<1/2\int_{2B^*}G^{q'}<1/2. This proves (2.8) with ηn=(3Nn)−1\eta_n=(3N_n)^{-1}. Interchanging FpF^p and Gq′G^{q'} proves (2.9) as well. □

Necessity of the Muckenhoupt condition

Proposition 1. If (1.4) holds with constant C>0C>0, then

[w]p,q≤Cn,sC−1.(17)[w]_{p,q}\le C_{n,s}C^{-1}. \tag*{(17)}

In particular, w−p′∈A1+p′qw^{-p'}\in A_{1+\frac{p'}{q}}.

Proof. Fix a ball B=B(x0,R)B=B(x_0,R). We first show wq(B)<∞w^q(B)<\infty. For sufficiently large NN, the set DN=B∩{N−1≤w≤N}D_N=B\cap\{N^{-1}\le w\le N\} has positive measure. For f=χDNf=\chi_{D_N}, both sides of (1.4) are positive and finite on BB. If wq(B)=∞w^q(B)=\infty, the negative qq-norm of IsfI_sf is zero, contradicting (1.4). Now let f≥0f\ge0 be supported in BB. For x,y∈Bx,y\in B,

∣x−y∣s−n≤(2R)s−n,|x-y|^{s-n}\le(2R)^{s-n},

so

Isf(x)≤(2R)s−n∫Bf(y) dy.I_sf(x)\le(2R)^{s-n}\int_B f(y)\,dy.

Restricting the integral to BB gives an upper bound for the negative norm. Hence

∥Isf∥Lq(wq)≤(2R)s−n(∫Bf(y) dy)(∫Bw(y)q dy)1/q.\|I_sf\|_{L^q(w^q)}\le(2R)^{s-n}\left(\int_B f(y)\,dy\right)\left(\int_B w(y)^q\,dy\right)^{1/q}.

Combining this with the assumed lower bound yields

C≤(2R)s−n∫Bf(y) dy(∫Bf(y)pw(y)p dy)1/p(∫Bw(y)q dy)1/q.C\le(2R)^{s-n}\frac{\int_B f(y)\,dy}{\left(\int_B f(y)^p w(y)^p\,dy\right)^{1/p}}\left(\int_B w(y)^q\,dy\right)^{1/q}.

Note that the cut-off function fN=w−p′1B∩{N−1≤w≤N}f_N=w^{-p'}1_{B\cap\{N^{-1}\le w\le N\}}, whose integral is always finite. If ∫Bw−p′(y) dy=∞\int_B w^{-p'}(y)\,dy=\infty, the resulting ratio tends to zero, contradicting the assumed lower bound. Hence w−p′w^{-p'} is locally integrable, and using reverse Hölder’s inequality, we have

inf⁡f∫Bf(y) dy(∫Bf(y)pw(y)p dy)1/p=(∫Bw(y)−p′ dy)1/p′.\inf_f\frac{\int_B f(y)\,dy}{\left(\int_B f(y)^p w(y)^p\,dy\right)^{1/p}}=\left(\int_B w(y)^{-p'}\,dy\right)^{1/p'}.

Thus

(∫Bw−p′)−1/p′(∫Bwq)−1/q≤C−1(2R)s−n.\left(\int_B w^{-p'}\right)^{-1/p'}\left(\int_B w^q\right)^{-1/q}\le C^{-1}(2R)^{s-n}.

Since ∣B∣−1/p′−1/q=∣B∣(s−n)/n≃nRs−n|B|^{-1/p'-1/q}=|B|^{(s-n)/n}\simeq_n R^{s-n}, division by ∣B∣(s−n)/n|B|^{(s-n)/n} proves (3.1). □

Strong-type bounds and sharpness

Proof of Theorem 1.2. By homogeneity, normalize ∥F∥p=∥G∥q′=1\|F\|_p=\|G\|_{q'}=1 and apply Lemma 2.4. If (2.8) holds, then Lemma 2.3 gives

Bw(F,G)≥cn,s,p[w]p,q−1[w−p′]A∞1/q.\mathcal{B}_w(F,G)\ge c_{n,s,p}[w]_{p,q}^{-1}[w^{-p'}]_{A_\infty}^{1/q}.

If (2.9) holds, then

Bw(F,G)≥cn,s,p[w]p,q−1[wq]A∞1/p′.\mathcal{B}_w(F,G)\ge c_{n,s,p}[w]_{p,q}^{-1}[w^q]_{A_\infty}^{1/p'}.

Thus, we obtain

Bw(F,G)≥cn,s,p[w]p,q−1min⁡{[w−p′]A∞1/q,[wq]A∞1/p′}.\mathcal{B}_w(F,G)\ge c_{n,s,p}[w]_{p,q}^{-1}\min\{[w^{-p'}]_{A_\infty}^{1/q},[w^q]_{A_\infty}^{1/p'}\}.

Taking the infimum over GG with ∥G∥q′=1\|G\|_{q'}=1 and using Lemma 2.2 proves (1.6).

Proof of Theorem 1.1. The implication (i)⇒\Rightarrow(ii) is Proposition 1. If (ii) holds, then (1.3) gives w−p′∈A1+p′qw^{-p'}\in A_{1+\frac{p'}{q}} and its dual weight wq∈A1+qp′w^q\in A_{1+\frac{q}{p'}}, so both are in A∞A_\infty. The mixed lower bound of Theorem 1.2 is therefore strictly positive and yields (i).

In the following subsections, we first prove the lower bound, then show that the exponent cannot be improved.

Proof. Since

[w−p′]A∞≲n,p,q[w−p′]A1+p′q=[w]p,qp′,[w^{-p'}]_{A_\infty}\lesssim_{n,p,q}[w^{-p'}]_{A_{1+\frac{p'}{q}}}=[w]_{p,q}^{p'},

and, using (2.1),

[wq]A∞≲n,p,q[wq]A1+qp′=[w]p,q−q,[w^q]_{A_\infty}\lesssim_{n,p,q}[w^q]_{A_{1+\frac{q}{p'}}}=[w]_{p,q}^{-q},

Theorem 1.2 gives the lower bound estimate

C∗(w)≳[w]p,q−1min⁡{[w]p,q−p′/q,[w]p,q−q/p′}=[w]p,q−(1+max⁡{p′/q,q/p′})=[w]p,q−Γ∗.C_*(w)\gtrsim[w]_{p,q}^{-1}\min\{[w]_{p,q}^{-p'/q},[w]_{p,q}^{-q/p'}\} =[w]_{p,q}^{-(1+\max\{p'/q,q/p'\})}=[w]_{p,q}^{-\Gamma_*}.

Next, we show the sharpness of the exponent. Note that

∣x∣δ∈Ar⟺−n<δ<n(r−1).(18)|x|^\delta\in A_r\Longleftrightarrow-n<\delta<n(r-1). \tag*{(18)}

and, as δ→−n\delta\to-n and δ→n(r−1)\delta\to n(r-1),

[∣x∣−n+ε]Ar≈n,rε−1,(19)[|x|^{-n+\varepsilon}]_{A_r}\approx_{n,r}\varepsilon^{-1}, \tag*{(19)}
[∣x∣n(r−1)−ε]Ar≈n,rε−(r−1).(20)[|x|^{n(r-1)-\varepsilon}]_{A_r}\approx_{n,r}\varepsilon^{-(r-1)}. \tag*{(20)}

Let

wε−(x)=∣x∣(n−ε)/p′,0<ε<1.w_\varepsilon^-(x)=|x|^{(n-\varepsilon)/p'},\qquad0<\varepsilon<1.

By (4.2),

[wε−]p,q≈ε1/p′.(21)[w_\varepsilon^-]_{p,q}\approx\varepsilon^{1/p'}. \tag*{(21)}

Choose

fε(x)=∣x∣−n+εχB(0,1)(x).f_\varepsilon(x)=|x|^{-n+\varepsilon}\chi_{B(0,1)}(x).

Since −p/p′=1−p-p/p'=1-p,

∥fε∥Lp((wε−)p)p=∫B(0,1)∣x∣−n+ε dx=∣Sn−1∣ε.(22)\|f_\varepsilon\|_{L^p((w_\varepsilon^-)^p)}^p=\int_{B(0,1)}|x|^{-n+\varepsilon}\,dx=\frac{|S^{n-1}|}{\varepsilon}. \tag*{(22)}

For ∣x∣≥2|x|\ge2 and ∣y∣<1|y|<1, ∣x−y∣≤32∣x∣|x-y|\le\frac{3}{2}|x|, hence

Isfε(x)≤(32)s−n∣x∣s−n∣Sn−1∣ε.I_s f_\varepsilon(x)\le\left(\frac{3}{2}\right)^{s-n}|x|^{s-n}\frac{|S^{n-1}|}{\varepsilon}.

And

∫Rn(Isfε(x))q(wε−(x))q dx≥cn,sε−q∫∣x∣>2∣x∣(s−n)q+(n−ε)q/p′ dx=cn,sε−q∣Sn−1∣∫2∞r−1−εq/p′ dr≥cn,s,pε−q−1.\begin{aligned} \int_{\mathbb{R}^n}(I_s f_\varepsilon(x))^q(w_\varepsilon^-(x))^q\,dx\ge c_{n,s}\varepsilon^{-q}\int_{|x|>2}|x|^{(s-n)q+(n-\varepsilon)q/p'}\,dx \\ &=c_{n,s}\varepsilon^{-q}|S^{n-1}|\int_2^\infty r^{-1-\varepsilon q/p'}\,dr \\ &\ge c_{n,s,p}\varepsilon^{-q-1}. \end{aligned}

Therefore

∥Isfε∥Lq((wε−)q)≤Cn,s,pε−1/q−1.(23)\|I_s f_\varepsilon\|_{L^q((w_\varepsilon^-)^q)}\le C_{n,s,p}\varepsilon^{-1/q-1}. \tag*{(23)}

Combining (4.5) and (4.6),

C∗(wε−)≤Cn,s,pε−1/p′−1/q=Cn,s,pε(s−n)/n.(24)C_*(w_\varepsilon^-)\le C_{n,s,p}\varepsilon^{-1/p'-1/q}=C_{n,s,p}\varepsilon^{(s-n)/n}. \tag*{(24)}

(4.4) implies

C∗(wε−)≲[wε−]p,q−(1+p′/q).(25)C_*(w_\varepsilon^-)\lesssim[w_\varepsilon^-]_{p,q}^{-(1+p'/q)}. \tag*{(25)}

Hence, we obtain C∗(w)≳[w]p,q−ΓC_*(w)\gtrsim[w]_{p,q}^{-\Gamma} with Γ≥1+p′/q\Gamma\ge1+p'/q.

On the other hand, let

wε+(x)=∣x∣−n/q+ε/p′,0<ε<1.w_\varepsilon^+(x)=|x|^{-n/q+\varepsilon/p'},\qquad0<\varepsilon<1.

By (4.3),

[wε+]p,q≈ε1/q.(26)[w_\varepsilon^+]_{p,q}\approx\varepsilon^{1/q}. \tag*{(26)}

The bilinear symmetry (2.5) exchanges

(p,−p′,w−p′)⟷(q′,−q,wq),w⟷w−1.(p,-p',w^{-p'})\longleftrightarrow(q',-q,w^q),\qquad w\longleftrightarrow w^{-1}.

For w=wε+w=w_\varepsilon^+,

σ=(wε+)q=∣x∣−n+εq/p′,\sigma=(w_\varepsilon^+)^q=|x|^{-n+\varepsilon q/p'},

which is just the lower-boundary family with the parameters exchanged. Applying (4.7) with −p′-p' and −q-q exchanged gives

C∗(wε+)≲ε−1/p′−1/q=ε(s−n)/n.C_*(w_\varepsilon^+)\lesssim\varepsilon^{-1/p'-1/q}=\varepsilon^{(s-n)/n}.

(4.9) gives

C∗(wε+)≲[wε+]p,q−(1+q/p′).C_*(w_\varepsilon^+)\lesssim[w_\varepsilon^+]_{p,q}^{-(1+q/p')}.

Hence Γ≥1+q/p′\Gamma\ge1+q/p'. Together with (4.8), this proves optimality of Γ∗=max⁡{1+p′/q,1+q/p′}\Gamma_*=\max\{1+p'/q,1+q/p'\} and completes the proof of sharp exponent of Theorem 1.2.

Endpoint theorem and explicit characteristic bounds

We now consider the endpoint case p=1p=1. For a positive measure σ\sigma, we use

∥h∥Lnn−s,∞(σ)=inf⁡λ>0λσ({x:h(x)<λ})n−sn.\lVert h\rVert_{L^{\frac{n}{n-s},\infty}(\sigma)}=\inf_{\lambda>0}\lambda\sigma(\{x:h(x)<\lambda\})^{\frac{n-s}{n}}.

The A1A_1 characteristic is

[σ]A1=sup⁡B1∣B∣∫Bσess inf⁡Bσ.[\sigma]_{A_1}=\sup_B\frac{\frac{1}{|B|}\int_B\sigma}{\operatorname*{ess\,inf}_B\sigma}.

Proof. We first prove the sufficient. Assume wnn−s∈A1w^{\frac{n}{n-s}}\in A_1 and put A=[wnn−s]A1A=[w^{\frac{n}{n-s}}]_{A_1}. Fix f≥0f\ge0 and λ>0\lambda>0, and set

Eλ={x:Isf(x)<λ}.E_\lambda=\{x:I_s f(x)<\lambda\}.

If ff is nonzero and IsfI_s f is finite at some point, then each superlevel set EλE_\lambda is bounded. Indeed, if ∫Bf(y) dy>0\int_B f(y)\,dy>0 for some ball BB, then Isf(x)I_s f(x) grows at least like ∣x∣s−n|x|^{s-n} as ∣x∣→∞|x|\to\infty. Thus it suffices to consider 0<wnn−s(Eλ)<∞0<w^{\frac{n}{n-s}}(E_\lambda)<\infty. For almost every yy and every r>0r>0, the A1A_1 condition gives

wnn−s(Eλ∩B(y,r))≤wnn−s(B(y,r))≤Awnn−s(y)vnrn.w^{\frac{n}{n-s}}(E_\lambda\cap B(y,r))\le w^{\frac{n}{n-s}}(B(y,r))\le A w^{\frac{n}{n-s}}(y)v_n r^n.

Set Ry=(wnn−s(Eλ)Avnwnn−s(y))1/nR_y=\left(\frac{w^{\frac{n}{n-s}}(E_\lambda)}{A v_n w^{\frac{n}{n-s}}(y)}\right)^{1/n}. The layer-cake identity and Tonelli’s theorem give

∫Eλ∣x−y∣s−nwnn−s(x) dx=(s−n)∫0∞rs−n−1wnn−s(Eλ∖B(y,r)) dr≥(s−n)∫0Ryrs−n−1(wnn−s(Eλ)−Awnn−s(y)vnrn) dr=ns(Avn)n−snwnn−s(Eλ)s/nwnn−s(y)1/qe=ns(Avn)n−snwnn−s(Eλ)s/nw(y).\begin{aligned} \int_{E_\lambda}|x-y|^{s-n}w^{\frac{n}{n-s}}(x)\,dx &=(s-n)\int_0^\infty r^{s-n-1}w^{\frac{n}{n-s}}(E_\lambda\setminus B(y,r))\,dr\\ &\ge(s-n)\int_0^{R_y}r^{s-n-1}\left(w^{\frac{n}{n-s}}(E_\lambda)-A w^{\frac{n}{n-s}}(y)v_n r^n\right)\,dr\\ &=\frac{n}{s}(A v_n)^{\frac{n-s}{n}}w^{\frac{n}{n-s}}(E_\lambda)^{s/n}w^{\frac{n}{n-s}}(y)^{1/q_e}\\ &=\frac{n}{s}(A v_n)^{\frac{n-s}{n}}w^{\frac{n}{n-s}}(E_\lambda)^{s/n}w(y). \end{aligned}

Consequently, Tonelli’s theorem yields

λwnn−s(Eλ)≥∫EλIsf(x)wnn−s(x) dx≥ns(Avn)n−snwnn−s(Eλ)s/n∫Rnf(y)w(y) dy.\lambda w^{\frac{n}{n-s}}(E_\lambda)\ge\int_{E_\lambda}I_s f(x)w^{\frac{n}{n-s}}(x)\,dx\ge\frac{n}{s}(A v_n)^{\frac{n-s}{n}}w^{\frac{n}{n-s}}(E_\lambda)^{s/n}\int_{\mathbb{R}^n}f(y)w(y)\,dy.

Division by wnn−s(Eλ)s/nw^{\frac{n}{n-s}}(E_\lambda)^{s/n} and then taking the infimum over λ\lambda, we obtain

∥Isf∥Lnn−s,∞(wnn−s)≥nsvnn−sn[wnn−s]A1n−sn∥f∥L1(w).\lVert I_s f\rVert_{L^{\frac{n}{n-s},\infty}(w^{\frac{n}{n-s}})}\ge\frac{n}{s}v_n^{\frac{n-s}{n}}[w^{\frac{n}{n-s}}]_{A_1}^{\frac{n-s}{n}}\lVert f\rVert_{L^1(w)}.

We show necessity. First, wnn−sw^{\frac{n}{n-s}} is locally integrable. Indeed, for any ball BB, choose NN so that B∩DNB\cap D_N has positive measure, where DN={N−1≤w≤N}D_N=\{N^{-1}\le w\le N\}. With f=χB∩DNf=\chi_{B\cap D_N}, ∥f∥L1(w)>0\lVert f\rVert_{L^1(w)}>0 and IsfI_s f is bounded on BB. If wnn−s(B)=∞w^{\frac{n}{n-s}}(B)=\infty, which makes the reverse weak norm zero, contradicting

∥Isf∥Lnn−s,∞(wnn−s)≥nsvnn−sn[wnn−s]A1n−sn∥f∥L1(w).\lVert I_s f\rVert_{L^{\frac{n}{n-s},\infty}(w^{\frac{n}{n-s}})}\ge\frac{n}{s}v_n^{\frac{n-s}{n}}[w^{\frac{n}{n-s}}]_{A_1}^{\frac{n-s}{n}}\lVert f\rVert_{L^1(w)}.

Fix NN and choose a point y∈DNy \in D_N that is also a Lebesgue point of wχDNw\chi_{D_N} and wnn−sw^{\frac{n}{n-s}}. Define

fε(z)=χB(y,ε)∩DN(z)∫B(y,ε)∩DNw(x) dx,∫Rnfε(z)w(z) dz=1.f_\varepsilon(z)=\frac{\chi_{B(y,\varepsilon)\cap D_N}(z)}{\int_{B(y,\varepsilon)\cap D_N}w(x)\,dx}, \qquad\int_{\mathbb{R}^n}f_\varepsilon(z)w(z)\,dz=1.

By Lebesgue differentiation, we obtain ∫fε(z) dz→w(y)−1\int f_\varepsilon(z)\,dz\to w(y)^{-1}. For x∈B(y,R)x\in B(y,R),

Isfε(x)≤(R+ε)s−n∫Rnfε(z) dz.I_sf_\varepsilon(x)\le(R+\varepsilon)^{s-n}\int_{\mathbb{R}^n}f_\varepsilon(z)\,dz.

Since nn−s<0\frac{n}{n-s}<0, let ε→0\varepsilon\to0, this yields

1∣B∣∫B(y,R)wnn−s(z) dz≤vn−1(nsvnn−sn[wnn−s]A1n−snwnn−s(y))nn−s(27)\frac{1}{|B|}\int_{B(y,R)}w^{\frac{n}{n-s}}(z)\,dz\le v_n^{-1}\left(\frac{n}{s}v_n^{\frac{n-s}{n}}[w^{\frac{n}{n-s}}]_{A_1}^{\frac{n-s}{n}}w^{\frac{n}{n-s}}(y)\right)^{\frac{n}{n-s}} \tag*{(27)}

for almost every yy and every R>0R>0. For an arbitrary ball B(c,R)B(c,R) and almost every y∈B(c,R)y\in B(c,R),

1∣B∣∫B(c,R)wnn−s(z) dz≤2n1∣B∣∫B(y,2R)wnn−s(z) dz\frac{1}{|B|}\int_{B(c,R)}w^{\frac{n}{n-s}}(z)\,dz\le2^n\frac{1}{|B|}\int_{B(y,2R)}w^{\frac{n}{n-s}}(z)\,dz
≤2nvn−1(nsvnn−sn[wnn−s]A1n−snwnn−s(y))nn−s.\le2^nv_n^{-1}\left(\frac{n}{s}v_n^{\frac{n-s}{n}}[w^{\frac{n}{n-s}}]_{A_1}^{\frac{n-s}{n}}w^{\frac{n}{n-s}}(y)\right)^{\frac{n}{n-s}}.

Take the essential infimum in yy, we have wnn−s∈A1w^{\frac{n}{n-s}}\in A_1. This completes the proof of Theorem 1.3. □

Remark 3 (The endpoint coefficient need not be sharp). For n=1n=1, s=2s=2, and w≡1w\equiv1, then

∥Isf∥Lnn−s,∞(wnn−s)≥14∥f∥L1(w).\|I_sf\|_{L^{\frac{n}{n-s},\infty}(w^{\frac{n}{n-s}})}\ge\frac{1}{4}\|f\|_{L^1(w)}.

The actual best constant is 1/21/2. Indeed, if 0<M=∫Rf<∞0<M=\int_{\mathbb{R}}f<\infty and ∫R∣xf(x)∣ dx<∞\int_{\mathbb{R}}|x f(x)|\,dx<\infty, Jensen’s inequality gives

I2f(x)=∫R∣x−y∣f(y) dy≥M∣x−yˉ∣,yˉ=M−1∫Ryf(y) dy.I_2f(x)=\int_{\mathbb{R}}|x-y|f(y)\,dy\ge M|x-\bar y|,\qquad\bar y=M^{-1}\int_{\mathbb{R}}yf(y)\,dy.

*Hence ∣{I2f<λ}∣≤2λ/M|\{I_2f<\lambda\}|\le2\lambda/M, so ∥I2f∥L−1,∞≥M/2\|I_2f\|_{L^{-1,\infty}}\ge M/2; If ∫R∣xf(x)∣ dx=∞\int_{\mathbb{R}}|x f(x)|\,dx=\infty, then I2f≡∞I_2f\equiv\infty and the inequality is immediate. If M=∞M=\infty, then I2fI_2f is infinite except possibly at one point, so the same inequality holds in the extended sense. Conversely, for any nonzero compactly supported ff, the sublevel-set length is asymptotic to 2λ/M2\lambda/M as λ→∞\lambda\to\infty, giving the upper bound 1/21/2.

School of Science, Xi’an University of Posts and Telecommunications, Xi’an, 710121, China, E-mail: [email protected]

Laboratory of Mathematics and Complex Systems (Ministry of Education), School of Mathematical Sciences, Beijing Normal University, Beijing, 100875, China, E-mail: [email protected]

References

  1. [1]S. M. Buckley, Estimates for operator norms on weighted spaces and reverse Jensen inequalities, Trans. Amer. Math. Soc. 340 (1993), no. 1, 253–272.DOI
  2. [2]J. A. Carrillo, M. G. Delgadino, J. Dolbeault, R. L. Frank, and F. Hoffmann, Reverse Hardy–Littlewood–Sobolev inequalities, J. Math. Pures Appl. 132 (2019), 133–165.DOI
  3. [3]L. Chen, Z. Liu, G. Lu, and C. Tao, Reverse Stein–Weiss inequalities and existence of their extremal functions, Trans. Amer. Math. Soc. 370 (2018), no. 12, 8429–8450.DOI
  4. [4]R. R. Coifman and C. Fefferman, Weighted norm inequalities for maximal functions and singular integrals, Studia Math. 51 (1974), 241–250.DOI
  5. [5]D. Cruz-Uribe and K. Moen, One and two weight norm inequalities for Riesz potentials, Illinois J. Math. 57 (2013), no. 1, 295–323.
  6. [6]J. Dou and M. Zhu, Reversed Hardy–Littlewood–Sobolev inequality, Int. Math. Res. Not. IMRN 2015 (2015), no. 19, 9696–9726.
  7. [7]O. Dragičević, L. Grafakos, C. Pereyra and S. Petermichl, Extrapolation and sharp norm estimates for classical operators on weighted Lebesgue spaces, Publ. Math. 49 (2005), no. 1, 73–91.DOI
  8. [8]R. Frank and E. Lieb, A new, rearrangement-free proof of the Sharp Hardy–Littlewood–Sobolev In-equality. Spectral Theory, Function Spaces and Inequalities. Oper. Theory: Adv. Appl. 219 (2012), 55–67.DOI
  9. [9]G. Hardy and J. Littlewood, Some properties of fractional integral I. Math. Zeit. (1928), 565–606.DOI
  10. [11]L. Huang, H. Tang and C. Zhang, Endpoint estimates of fractional integral operators when α > n, Bull. Sci. Math. 199 (2025), 103569.DOI
  11. [12]R. Hunt, B. Muckenhoupt, and R. Wheeden, Weighted norm inequalities for the conjugate function and Hilbert transform, Trans. Amer. Math. Soc. 176 (1973), 227–251.
  12. [13]T. P. Hytönen, The sharp weighted bound for general Calderón–Zygmund operators, Ann. of Math. (2) 175 (2012), no. 3, 1473–1506.arxiv.org/abs/1007.4330
  13. [14]T. Hytönen and C. Pérez, Sharp weighted bounds involving A∞, Anal. PDE 6 (2013), no. 4, 777–818.
  14. [15]T. Hytönen, C. Pérez, and E. Rela, Sharp reverse Hölder property for A∞ weights on spaces of homogeneous type, J. Funct. Anal. 263 (2012), no. 12, 3883–3899.
  15. [16]A. Kassymov, M. Ruzhansky, and D. Suragan, Reverse Stein–Weiss, Hardy–Littlewood–Sobolev, Hardy, Sobolev and Caffarelli–Kohn–Nirenberg inequalities on homogeneous groups, Forum Math. 34 (2022), no. 5, 1147–1158.arxiv.org/abs/1912.01460
  16. [17]M. T. Lacey, K. Moen, C. Pérez, and R. H. Torres, Sharp weighted bounds for fractional integral operators, J. Funct. Anal. 259 (2010), no. 5, 1073–1097.arxiv.org/abs/0905.3839
  17. [19]A. K. Lerner, A simple proof of the A₂ conjecture, Int. Math. Res. Not. IMRN 2013 (2013), no. 14, 3159–3170.arxiv.org/abs/1202.2824
  18. [20]E. Lieb, Sharp constants in the Hardy–Littlewood–Sobolev and related inequalities. Ann. of Math. (2) 118 (1983), no. 2, 349–374.
  19. [21]B. Muckenhoupt, Weighted norm inequalities for the Hardy maximal function, Trans. Amer. Math. Soc. 165 (1972), 207–226.DOI
  20. [22]B. Muckenhoupt and R. L. Wheeden, Weighted norm inequalities for fractional integrals, Trans. Amer. Math. Soc. 192 (1974), 261–274.DOI
  21. [23]Q. A. Ngô and V. H. Nguyen, Sharp reversed Hardy–Littlewood–Sobolev inequality on ℝⁿ, Israel J. Math. 220 (2017), 189–223.DOI
  22. [25]S. Petermichl, The sharp bound for the Hilbert transform in weighted Lebesgue spaces in terms of the classical Aₚ characteristic, Amer. J. Math. 129 (2007), no. 5, 1355–1375.
  23. [26]S. Petermichl, The sharp weighted bound for the Riesz transforms, Proc. Amer. Math. Soc. 136 (2008), no. 4, 1237–1249.DOI
  24. [27]C. Pérez, Two weight norm inequalities for Riesz potentials and uniform Lᵖ-weighted Sobolev inequalities, Indiana Univ. Math. J. 39 (1990), 31–44.DOI
  25. [28]C. Pérez, Two weight inequalities for potential and fractional type maximal operators, Indiana Univ. Math. J. 43 (1994), 663–683.DOI
  26. [29]E. T. Sawyer, A characterization of a two-weight norm inequality for maximal operators, Studia Math. 75 (1982), 1–11.DOI
  27. [30]E. T. Sawyer, A characterization of a two-weight norm inequality for fractional and Poisson integrals, Trans. Amer. Math. Soc. 308 (1988), 533–545.DOI
  28. [31]S. Sawyer and R. L. Wheeden, Weighted inequalities for fractional integrals on Euclidean and homogeneous spaces, Amer. J. Math. 114 (1992), 813–874.DOI
  29. [32]M. Riesz, L’intégrale de Riemann–Liouville et le problème de Cauchy. Acta Math. 81 (1949), 1–222.
  30. [35]S. Sobolev, On a theorem in functional analysis (in Russian). Mat. Sob. 46 (1938), 471–497.
  31. [36]E. M. Stein and G. Weiss, Fractional integrals on n-dimensional Euclidean space, J. Math. Mech. 7 (1958), no. 4, 503–514.
  32. [37]H. Tang and Y. Wang, Sharp constants for weak estimates of the Riesz potentials, arXiv:2608.31043v2 [math.CA] (2026).arxiv.org/abs/2608.31043
  33. [38]M. Wilson, Weighted Littlewood–Paley Theory and Exponential-Square Integrability, Lecture Notes in Mathematics, vol. 1924, Springer, Berlin, 2008.DOI
  34. [39]A. Zygmund, On a theorem of Marcinkiewicz concerning interpolation of operators, J. Math. Pures Appl. (9) 35 (1956), 223–248.

Paper details

vlnc:W_p88r4zxj
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents