betaFeedback

WorksMathematics

Fefferman-Stein Type Distribution Function Inequalities on Product Spaces

Abstract

We establish on Rn×RmR^n\times R^m a Fefferman–Stein type distribution function inequality in which the nontangential maximal function of the product Poisson extension is controlled by the Lusin area integral function: ∣{N>λ}∣≤C(∣{A>λ}∣+λ−2∫{A≤λ}A2 dx). |\{N>λ\}|\le C\left(|\{A>λ\}|+λ^{-2}\int_{\{A\leλ\}}A^2\,dx\right). The central idea is to establish a local L2L^2 estimate for the radial maximal function. This yields a new proof of the corresponding one-parameter inequality, avoiding Green's formula and the treatment of sawtooth regions, and extends to the multiparameter setting.

∣{N>λ}∣≤C(∣{A>λ}∣+λ−2∫{A≤λ}A2 dx).|\{N>\lambda\}| \le C\left(|\{A>\lambda\}|+\lambda^{-2}\int_{\{A\le\lambda\}} A^{2}\,dx\right).

The central idea is to establish a local L2L^{2} estimate for the radial maximal function. This yields a new proof of the corresponding one-parameter inequality, avoiding Green’s formula and the treatment of sawtooth regions, and extends to the multiparameter setting.

Introduction and main results

Maximal functions and square functions are basic tools in the study of Hardy space theory. Comparing these functions is a fundamental problem in this theory, particularly in connection with the following two Fefferman–Stein type distribution function inequalities for the area integral function AA and the nontangential maximal function NN:

∣{A>λ}∣≤C∣{N>λ}∣+Cλ2∫{N≤λ}N2 dx.(1)|\{A>\lambda\}| \le C|\{N>\lambda\}|+\frac{C}{\lambda^{2}}\int_{\{N\le\lambda\}}N^{2}\,dx. \tag*{(1)}
∣{N>λ}∣≤C∣{A>λ}∣+Cλ2∫{A≤λ}A2 dx.(2)|\{N>\lambda\}| \le C|\{A>\lambda\}|+\frac{C}{\lambda^{2}}\int_{\{A\le\lambda\}}A^{2}\,dx. \tag*{(2)}

Both inequalities have been studied systematically in the one-parameter setting. In the multiparameter setting, however, (1.1) has been extensively investigated, whereas progress on (1.2) has remained elusive. In this paper we study the latter inequality on product spaces and establish (1.2).

In the one-parameter theory, distribution function comparisons in both directions between the area integral function and the nontangential maximal function have long been classical. Burkholder and Gundy [3] studied distribution function inequalities for the area integral using probabilistic and martingale methods, connecting localized distribution control with global integral estimates. Fefferman and Stein [8] established the corresponding real-variable framework in their theory of HpH^{p} spaces of several variables. By using Green’s formula and a subtle treatment of sawtooth regions, they obtained distribution estimates in both directions, namely the one-parameter versions of (1.1) and (1.2). Subsequent developments included weighted comparisons, geometric extensions to Lipschitz domains, and sharper ratio estimates. The weighted theory of Gundy and Wheeden [11], Dahlberg’s results on Lipschitz domains [7], and the ratio estimates of Fefferman, Gundy, Silverstein, and Stein [9] illustrate the role of distribution comparisons in weighted theory, complex boundary geometry, and stronger integral control, respectively. Murai and Uchiyama [15] studied good-λ\lambda estimates, while the work of Bañuelos and Moore [1, 2] further addressed sharp comparisons and the density of the area integral.

The difficulty in the multiparameter setting arises because the two groups of spatial variables have independent time variables, while the area integral function involves mixed derivatives in the two groups of space-time variables. The HpH^p theory on the bidisc developed by Malliavin and Malliavin [13] and by Gundy and Stein [10] was pioneering. In [10], Gundy and Stein first established (1.1) and a probabilistic version of (1.2) on the bidisc. Subsequently, Merryfield [14], Sato [16], and Chen [4] (also see Chen [6] for its proof strategy in the one-parameter case) independently obtained distribution estimates of type (1.1), controlling the area integral function by the nontangential maximal function on various product domains;. Chen and Wang [5] extended (1.1) to products of complete Riemannian manifolds with nonnegative Ricci curvature, while Li [12] extended it to product Shilov boundaries. For the reverse distribution inequality (1.2), however, there has been no further breakthrough.

Let d1=nd_1=n, d2=md_2=m, where n,m≥1n,m\ge1, and set X=Rd1×Rd2X=\mathbb{R}^{d_1}\times\mathbb{R}^{d_2}. For i=1,2i=1,2, let Δi\Delta_i denote the Laplacian in the iith group of variables, and define

Pt(i)=e−t−Δi,u(y1,t1;y2,t2)=Pt1(1)Pt2(2)f(y1,y2).P_t^{(i)}=e^{-t\sqrt{-\Delta_i}},\qquad u(y_1,t_1;y_2,t_2)=P_{t_1}^{(1)}P_{t_2}^{(2)}f(y_1,y_2).

Write zi=(yi,ti)z_i=(y_i,t_i), ∇i⊥=(∇yi,∂ti)\nabla_i^\perp=(\nabla_{y_i},\partial_{t_i}), vdi=∣BRdi(0,1)∣v_{d_i}=|B_{\mathbb{R}^{d_i}}(0,1)|, and

Γα(x)=Γ1,α1(x1)×Γ2,α2(x2),Γi,a(xi)={(yi,ti):∣yi−xi∣<ati}.(3)\Gamma_\alpha(x)=\Gamma_{1,\alpha_1}(x_1)\times\Gamma_{2,\alpha_2}(x_2),\qquad\Gamma_{i,a}(x_i)=\{(y_i,t_i):|y_i-x_i|<at_i\}. \tag*{(3)}

Vectors and tensors between the two groups of variables are equipped with their natural Hilbert norms (which are simply Euclidean norms). For α=(α1,α2)\alpha=(\alpha_1,\alpha_2) and β=(β1,β2)∈(0,∞)2\beta=(\beta_1,\beta_2)\in(0,\infty)^2, define

N+f(x)=sup⁡t1,t2>0∣u(x1,t1;x2,t2)∣.(4)N^+f(x)=\sup_{t_1,t_2>0}|u(x_1,t_1;x_2,t_2)|. \tag*{(4)}
Nαf(x)=sup⁡(z1,z2)∈Γα(x)∣u(z1,z2)∣.(5)N_\alpha f(x)=\sup_{(z_1,z_2)\in\Gamma_\alpha(x)}|u(z_1,z_2)|. \tag*{(5)}
Aβf(x)=(∬Γβ(x)∣∇1⊥∇2⊥u(y,t)∣2∏i=12ti dyi dtivdi(βiti)di)1/2.(6)A_\beta f(x)=\left(\iint_{\Gamma_\beta(x)}|\nabla_1^\perp\nabla_2^\perp u(y,t)|^2\prod_{i=1}^{2}\frac{t_i\,dy_i\,dt_i}{v_{d_i}(\beta_i t_i)^{d_i}}\right)^{1/2}. \tag*{(6)}

For a nonnegative function FF, write dF(λ)=∣{x:F(x)>λ}∣d_F(\lambda)=|\{x:F(x)>\lambda\}|, and set

Θλ(s)=min⁡{s2,λ2},Φλ(F)=∫XΘλ(F(x)) dx.(7)\Theta_\lambda(s)=\min\{s^2,\lambda^2\},\qquad\Phi_\lambda(F)=\int_X\Theta_\lambda(F(x))\,dx. \tag*{(7)}

We have the following result.

Theorem 1.1. For every f∈Cc∞(Rn×Rm)f\in C_c^\infty(\mathbb{R}^n\times\mathbb{R}^m) and λ>0\lambda>0, we have

Φλ(Nαf)≤CΦλ(Aβf),(8)\Phi_\lambda(N_\alpha f)\le C\Phi_\lambda(A_\beta f), \tag*{(8)}

where CC depends only on the dimensions and on α,β\alpha,\beta, and Φλ\Phi_\lambda is defined by (1.7). The corresponding one-parameter result holds for Hilbert-valued functions, with a constant independent of the target Hilbert space.

Theorem 1.1 implies the reverse distribution estimate (2) and the corresponding one-sided strong and weak LpL^p estimates for 0<p<20 < p < 2. The central idea of its proof is to establish the following local L2L^2 estimate for the radial maximal function, thereby reducing the study of the nontangential maximal function to that of the radial maximal function.

Let MiM_i be the centered Hardy–Littlewood maximal operator on the iith spatial factor Rdi\mathbb{R}^{d_i}, and let M1M2M_1M_2 denote their composition in the indicated order.

Theorem 1.2. There exists δ∈(0,1)\delta\in(0,1), depending only on n,m,βn,m,\beta, such that for every measurable E⊂XE \subset X with ∣Ec∣<∞|E^c| < \infty, the set

GE={x:M1M21Ec(x)≤δ}G_E = \{x : M_1M_2\mathbf{1}_{E^c}(x) \le\delta\}

satisfies

∣GEc∣≤C∣Ec∣,∫GE(N+f)2 dx≤C∫E(Aβf)2 dx.(9)|G_E^c| \le C|E^c|,\qquad\int_{G_E}(N^+f)^2\,dx \le C\int_E(A_\beta f)^2\,dx. \tag*{(9)}

This holds for every f∈L2(X)f \in L^2(X). The corresponding one-parameter Hilbert-valued result also holds.

One-parameter preliminaries

This section presents the one-parameter results needed below.

Lemma 2.1. Suppose c1,…,ck≥0c_1,\ldots,c_k \ge0. For all s1,…,sk≥0s_1,\ldots,s_k \ge0, we have

Θλ(∑j=1kcjsj)≤k∑j=1kmax⁡{1,cj2}Θλ(sj).(10)\Theta_\lambda\left(\sum_{j=1}^{k}c_js_j\right) \le k\sum_{j=1}^{k}\max\{1,c_j^2\}\Theta_\lambda(s_j). \tag*{(10)}

If nonnegative functions Fj→FF_j \to F in L2(X)L^2(X), then Φλ(Fj)→Φλ(F)\Phi_\lambda(F_j) \to\Phi_\lambda(F) for every λ>0\lambda> 0.

Proof. By the Cauchy–Schwarz inequality, (∑jcjsj)2≤k∑j(cjsj)2\left(\sum_j c_js_j\right)^2 \le k\sum_j(c_js_j)^2. Considering separately the case where some cjsj≥λc_js_j \ge\lambda and the case where all cjsj<λc_js_j < \lambda, we obtain

min⁡{k∑j(cjsj)2,λ2}≤k∑jmin⁡{(cjsj)2,λ2}.\min\left\{k\sum_j(c_js_j)^2,\lambda^2\right\} \le k\sum_j\min\{(c_js_j)^2,\lambda^2\}.

The first assertion follows from Θλ(cs)≤max⁡{1,c2}Θλ(s)\Theta_\lambda(cs) \le\max\{1,c^2\}\Theta_\lambda(s). Also,

∣Θλ(a)−Θλ(b)∣≤∣a−b∣(a+b),a,b≥0.(11)|\Theta_\lambda(a)-\Theta_\lambda(b)| \le|a-b|(a+b),\qquad a,b \ge0. \tag*{(11)}

Integrating and applying Cauchy–Schwarz inequality gives the final assertion.

For every dimension d≥1d \ge1, fix a radial function φd∈Cc∞(B(0,1/4))\varphi_d \in C^\infty_c(B(0,1/4)) with φd≥0\varphi_d \ge0, ∫φd=1\int\varphi_d=1, and φd>0\varphi_d>0 on B(0,1/8)B(0,1/8). Set

ψd(x)=φd(x)−2−dφd(x/2).(12)\psi_d(x)=\varphi_d(x)-2^{-d}\varphi_d(x/2). \tag*{(12)}

Then ∫ψd=0\int\psi_d=0 and supp⁡ψd⊂B(0,1/2)\operatorname{supp}\psi_d \subset B(0,1/2). Choose also a nonnegative χd∈Cc∞(B(0,1))\chi_d \in C^\infty_c(B(0,1)) with ∫χd=1\int\chi_d=1. Write ψd,t(x)=t−dψd(x/t)\psi_{d,t}(x)=t^{-d}\psi_d(x/t) and χd,t(x)=t−dχd(x/t)\chi_{d,t}(x)=t^{-d}\chi_d(x/t).

Let H\mathcal{H} be a Hilbert space (it may be a Euclidean space, whose dimension depends on the gradient and differs between the one- and two-parameter settings, or an L2L^2 space. For simplicity, we use the same notation throughout). Define

Ed(H)=L2(Rd×(0,∞),t dx dt;H),\mathcal{E}_d(\mathcal{H})=L^2(\mathbb{R}^d\times(0,\infty),t\,dx\,dt;\mathcal{H}),
∥H∥Ed(H)=(∫0∞∫Rdt∥H(x,t)∥H2 dx dt)1/2.(13)\|H\|_{\mathcal{E}_d(\mathcal{H})}=\left(\int_0^\infty\int_{\mathbb{R}^d}t\|H(x,t)\|_{\mathcal{H}}^2\,dx\,dt\right)^{1/2}. \tag*{(13)}

Write ψt=ψd,t\psi_t=\psi_{d,t} and χt=χd,t\chi_t=\chi_{d,t}. Define

TsH=∫s∞ψt−s∗Ht dt,EsH=TsH−χs∗SH,SH=∫0∞ψt∗Ht dt.(14)T_sH=\int_s^\infty\psi_{t-s}*H_t\,dt,\qquad E_sH=T_sH-\chi_s*SH,\qquad SH=\int_0^\infty\psi_t*H_t\,dt. \tag*{(14)}

where Ht(x)=H(x,t)H_t(x)=H(x,t). The following maximal lemma for TsT_s is crucial.

Lemma 2.2. There exists a nonnegative k∈L1(Rd)k\in L^1(\mathbb{R}^d) such that, for H∈Ed(H)H\in\mathcal{E}_d(\mathcal{H}),

∥SH∥22≤C∥H∥Ed(H)2,(15)\|SH\|_2^2\leq C\|H\|_{\mathcal{E}_d(\mathcal{H})}^2, \tag*{(15)}
sup⁡s>0∥EsH(x)∥H2≤C∫0∞t(Kt∗∥Ht∥H2)(x) dt,(16)\sup_{s>0}\|E_sH(x)\|_{\mathcal{H}}^2\leq C\int_0^\infty t\bigl(K_t*\|H_t\|_{\mathcal{H}}^2\bigr)(x)\,dt, \tag*{(16)}

where Kt(x)=t−dk(x/t)K_t(x)=t^{-d}k(x/t). Consequently,

∥sup⁡s>0∥TsH(⋅)∥H∥L2(Rd)2≤C∥H∥Ed(H)2.(17)\left\|\sup_{s>0}\|T_sH(\cdot)\|_{\mathcal{H}}\right\|_{L^2(\mathbb{R}^d)}^2\leq C\|H\|_{\mathcal{E}_d(\mathcal{H})}^2. \tag*{(17)}

The constants are independent of the target Hilbert space H\mathcal{H} and of HH.

Proof. By a limiting argument, we may assume that supp⁡H\operatorname{supp}H is compact in Rd×(0,∞)\mathbb{R}^d\times(0,\infty). Since ψ\psi is radial, compactly supported, and has mean zero,

∣ψ^(ξ)∣≤Cmin⁡{∣ξ∣,∣ξ∣−1}.|\widehat{\psi}(\xi)|\leq C\min\{|\xi|,|\xi|^{-1}\}.

For f∈L2(Rd;H)f\in L^2(\mathbb{R}^d;\mathcal{H}), Plancherel’s theorem gives

∫0∞∥ψt∗f∥22dtt=∫Rd∥f^(ξ)∥H2[∫0∞∣ψ^(tξ)∣2dtt]dξ≤C∥f∥22.\int_0^\infty\|\psi_t*f\|_2^2\frac{dt}{t} =\int_{\mathbb{R}^d}\|\widehat{f}(\xi)\|_{\mathcal{H}}^2\left[\int_0^\infty|\widehat{\psi}(t\xi)|^2\frac{dt}{t}\right]d\xi \leq C\|f\|_2^2.

By duality,

∣⟨SH,f⟩∣≤[∫0∞t∥Ht∥22 dt]1/2[∫0∞∥ψ~t∗f∥22dtt]1/2≲∥H∥Ed(H)∥f∥2,|\langle SH,f\rangle| \leq\left[\int_0^\infty t\|H_t\|_2^2\,dt\right]^{1/2} \left[\int_0^\infty\|\widetilde{\psi}_t*f\|_2^2\frac{dt}{t}\right]^{1/2} \lesssim\|H\|_{\mathcal{E}_d(\mathcal{H})}\|f\|_2,

where ψ~(x)=ψ(−x)\widetilde{\psi}(x)=\psi(-x). This proves (2.6). We next prove (2.7). We can write

EsH(x)=∫0∞(es,t∗Ht)(x) dt,(18)E_sH(x)=\int_0^\infty(e_{s,t}*H_t)(x)\,dt, \tag*{(18)}

where es,t=1{t>s}ψt−s−χs∗ψte_{s,t}=\mathbf{1}_{\{t>s\}}\psi_{t-s}-\chi_s*\psi_t. Choose a sufficiently large constant C0C_0 and put ar(x)=r−d1{∣x∣<C0r}(x)a_r(x)=r^{-d}\mathbf{1}_{\{|x|<C_0r\}}(x). Taylor’s formula and the compact supports of ψ,χ\psi,\chi yield

∣es,t(x)∣≤C(t/s)as(x),0<t<s/2,|e_{s,t}(x)|\leq C(t/s)a_s(x),\qquad0<t<s/2,
∣es,t(x)∣≤C(s/t)at(x),t>2s,|e_{s,t}(x)|\leq C(s/t)a_t(x),\qquad t>2s,
∣χs∗ψt(x)∣≤Cas(x),s/2≤t≤2s,|\chi_s*\psi_t(x)|\leq Ca_s(x),\qquad s/2\leq t\leq2s,
∣ψt−s(x)∣≤Cat−s(x),s<t<2s.|\psi_{t-s}(x)|\leq Ca_{t-s}(x),\qquad s<t<2s.

Accordingly, decompose EsH(x)E_sH(x) into four terms:

EsH(x)=∫0s/2(es,t∗Ht)(x) dt+∫2s∞(es,t∗Ht)(x) dt−∫s/22s(χs∗ψt∗Ht)(x) dt+∫s2s(ψt−s∗Ht)(x) dt=I1(x)+I2(x)+I3(x)+I4(x).\begin{aligned} E_sH(x) &=\int_0^{s/2}(e_{s,t}*H_t)(x)\,dt+\int_{2s}^{\infty}(e_{s,t}*H_t)(x)\,dt\\ &\quad-\int_{s/2}^{2s}(\chi_s*\psi_t*H_t)(x)\,dt+\int_s^{2s}(\psi_{t-s}*H_t)(x)\,dt\\ &=I_1(x)+I_2(x)+I_3(x)+I_4(x). \end{aligned}

Set Ft(x)=∥Ht(x)∥H2F_t(x)=\|H_t(x)\|_{\mathcal{H}}^2. Cauchy–Schwarz inequality gives

∥I1(x)∥H2≤C∫0s/2ts(as∗Ft)(x) dt,∥I2(x)∥H2≤C∫2s∞ts(at∗Ft)(x) dt,∥I3(x)∥H2≤C∫s/22st(as∗Ft)(x) dt.\begin{aligned} \|I_1(x)\|_{\mathcal{H}}^2 &\le C\int_0^{s/2}\frac{t}{s}(a_s*F_t)(x)\,\mathrm{d}t,\\ \|I_2(x)\|_{\mathcal{H}}^2 &\le C\int_{2s}^{\infty}\frac{t}{s}(a_t*F_t)(x)\,\mathrm{d}t,\\ \|I_3(x)\|_{\mathcal{H}}^2 &\le C\int_{s/2}^{2s}t(a_s*F_t)(x)\,\mathrm{d}t. \end{aligned}

For I4I_4, we have

∥I4(x)∥H2≤[∫s2st(t−st)α(at−s∗Ft)(x) dt][∫s2s(t−st)−αdtt]≤Cα∫s2st(t−st)α(at−s∗Ft)(x) dt,\begin{aligned} \|I_4(x)\|_{\mathcal{H}}^2\le\left[\int_s^{2s}t\left(\frac{t-s}{t}\right)^\alpha(a_{t-s}*F_t)(x)\,\mathrm{d}t\right]\left[\int_s^{2s}\left(\frac{t-s}{t}\right)^{-\alpha}\frac{\mathrm{d}t}{t}\right] \\ &\le C_\alpha\int_s^{2s}t\left(\frac{t-s}{t}\right)^\alpha(a_{t-s}*F_t)(x)\,\mathrm{d}t, \end{aligned}

since

∫s2s(t−st)−αdtt=∫01u−α(1+u)α−1 du=Cα<∞\int_s^{2s}\left(\frac{t-s}{t}\right)^{-\alpha}\frac{\mathrm{d}t}{t} =\int_0^1u^{-\alpha}(1+u)^{\alpha-1}\,\mathrm{d}u=C_\alpha<\infty

whenever α∈(0,1)\alpha\in(0,1). Take

k(z)=C((1+∣z∣)−d−1+∣z∣α−d1{∣z∣<C1})∈L1(Rd).k(z)=C\left((1+|z|)^{-d-1}+|z|^{\alpha-d}\mathbf{1}_{\{|z|<C_1\}}\right)\in L^1(\mathbb{R}^d).

Estimate (2.7) follows from

sup⁡s>2ttsas(z)≤Ctsup⁡s>max⁡(2t,∣z∣/C0)s−d−1≤Ct−d(1+∣z∣t)−d−1,\sup_{s>2t}\frac{t}{s}a_s(z)\le Ct\sup_{s>\max(2t,|z|/C_0)}s^{-d-1}\le Ct^{-d}\left(1+\frac{|z|}{t}\right)^{-d-1},
sup⁡0<s<t/2stat(z)≤Cat(z)≤Ct−d1{∣z∣<C0t},\sup_{0<s<t/2}\frac{s}{t}a_t(z)\le Ca_t(z)\le Ct^{-d}\mathbf{1}_{\{|z|<C_0t\}},
sup⁡t/2≤s≤2tas(z)≤Ct−d1{∣z∣<C1t},\sup_{t/2\le s\le2t}a_s(z)\le Ct^{-d}\mathbf{1}_{\{|z|<C_1t\}},
sup⁡0<r<t/2(rt)αar(z)≤Ct−α∣z∣α−d1{∣z∣<C1t}.\sup_{0<r<t/2}\left(\frac{r}{t}\right)^\alpha a_r(z)\le Ct^{-\alpha}|z|^{\alpha-d}\mathbf{1}_{\{|z|<C_1t\}}.

Finally, to prove (2.8), note that

∥χs∗SH(x)∥H≤Cs−d∫∣x−y∣<s∥SH(y)∥H dy≤CM(∥SH∥H)(x),\|\chi_s*SH(x)\|_{\mathcal{H}}\le Cs^{-d}\int_{|x-y|<s}\|SH(y)\|_{\mathcal{H}}\,\mathrm{d}y\le CM(\|SH\|_{\mathcal{H}})(x),

where MM is the Hardy–Littlewood maximal operator on Rd\mathbb{R}^d. Combining this with (2.6)–(2.7) proves (2.8).

For every θ∈Sd−1\theta\in S^{d-1}, radiality of ψ\psi and Fubini’s theorem give

cd:=∫0∞e−rψ^d(rθ) dr=∫Rdϕd(y)(11+(y⋅θ)2−11+4(y⋅θ)2) dy>0.(19)\begin{aligned} c_d&:=\int_0^\infty e^{-r}\widehat{\psi}_d(r\theta)\,\mathrm{d}r\\ &=\int_{\mathbb{R}^d}\phi_d(y)\left(\frac{1}{1+(y\cdot\theta)^2}-\frac{1}{1+4(y\cdot\theta)^2}\right)\,\mathrm{d}y>0. \tag*{(19)} \end{aligned}

This value is independent of θ\theta. We have the following reconstruction formula.

Lemma 2.3. Let q∈L2(Rd;H)q\in L^2(\mathbb{R}^d;\mathcal{H}) and v(x,t)=Ptq(x)v(x,t)=P_tq(x), where PtP_t is the Poisson semigroup on Rd\mathbb{R}^d. For every s>0s>0, the following identity holds in L2L^2:

v(x,s)=−cd−1Ts(∂tv(⋅,t))(x).(20)v(x,s)=-c_d^{-1}T_s(\partial_tv(\cdot,t))(x). \tag*{(20)}

Proof. For ξ≠0\xi\ne0, the Fourier multiplier on the right-hand side of (20) is

−cd−1∫0∞ψd^(rξ)(−∣ξ∣)e−(s+r)∣ξ∣ dr=e−s∣ξ∣.-c_d^{-1}\int_{0}^{\infty}\widehat{\psi_d}(r\xi)(-|\xi|)e^{-(s+r)|\xi|}\,dr=e^{-s|\xi|}.

Here we use the substitution r∣ξ∣↦rr|\xi|\mapsto r and the definition of cdc_d. Since ∂tv∈Ed(H)\partial_t v\in\mathcal{E}_d(\mathcal{H}), Lemma 2.2 justifies (20). □

Lemma 2.4. Let FF be a measurable function on R+d+1\mathbb{R}^{d+1}_+ with values in a Hilbert space H\mathcal{H}, and, for c>0c>0, define

ScF(x)=(∫Γc(x)∥F(y,t)∥H2t dy dtvd(ct)d)1/2.S_cF(x)=\left(\int_{\Gamma_c(x)}\|F(y,t)\|_{\mathcal{H}}^2\frac{t\,dy\,dt}{v_d(ct)^d}\right)^{1/2}.

For fixed a,b>0a,b>0, we have

Φλ(SaF)≤CΦλ(SbF).(21)\Phi_\lambda(S_aF)\le C\Phi_\lambda(S_bF). \tag*{(21)}

Proof. The assertion is immediate if the right-hand side is infinite. Otherwise, set E={SbF≤λ}E=\{S_bF\le\lambda\}, B=EcB=E^c, and G={M1B≤ε}G=\{M1_B\le\varepsilon\}, where MM is the centered Hardy–Littlewood maximal operator. Its weak type (1,1)(1,1) bound gives ∣Gc∣≤Cε−1∣B∣|G^c|\le C\varepsilon^{-1}|B|. For any (y,t)(y,t) such that G∩B(y,at)≠∅G\cap B(y,at)\ne\varnothing, choose x∈G∩B(y,at)x\in G\cap B(y,at). Since B(y,bt)⊂B(x,(a+b)t)B(y,bt)\subset B(x,(a+b)t),

1−∣E∩B(y,bt)∣∣B(y,bt)∣≤Ca,bM1B(x)≤Ca,bε.1-\frac{|E\cap B(y,bt)|}{|B(y,bt)|}\le C_{a,b}M1_B(x)\le C_{a,b}\varepsilon.

Fix ε\varepsilon sufficiently small that ∣E∩B(y,bt)∣/∣B(y,bt)∣≥1/2|E\cap B(y,bt)|/|B(y,bt)|\ge1/2. Thus

∣G∩B(y,at)∣∣B(y,at)∣≤1≤2∣E∩B(y,bt)∣∣B(y,bt)∣.\frac{|G\cap B(y,at)|}{|B(y,at)|}\le1\le2\frac{|E\cap B(y,bt)|}{|B(y,bt)|}.

Tonelli’s theorem gives ∫G(SaF)2≤2∫E(SbF)2\int_G(S_aF)^2\le2\int_E(S_bF)^2. Using Φλ(F)≤∫GF2+λ2∣Gc∣\Phi_\lambda(F)\le\int_G F^2+\lambda^2|G^c| and B={SbF>λ}B=\{S_bF>\lambda\} proves (21). □

Lemma 2.5. Let vv be a Hilbert-valued harmonic function on R+d+1\mathbb{R}^{d+1}_+. If y∈B(x,at)y\in B(x,at) and η>a\eta>a, then

∥v(y,t)−v(x,t)∥H≤CAηv(x),(22)\|v(y,t)-v(x,t)\|_{\mathcal{H}}\le CA_\eta v(x), \tag*{(22)}

where AηvA_\eta v is the one-parameter area integral function.

Proof. Choose c>0c>0, depending only on a,ηa,\eta, sufficiently small that every ball of radius ctct centered on the horizontal segment from (x,t)(x,t) to (y,t)(y,t) lies in Γη(x)\Gamma_\eta(x). By subharmonicity of ∥∇v∥H2\|\nabla v\|_{\mathcal{H}}^2 and the mean-value inequality on balls,

∥∇v(z,t)∥H2≤Ct−d−1∫B((z,t),ct)∥∇v(w,s)∥H2 dw ds≤Ct−2(Aηv(x))2.\|\nabla v(z,t)\|_{\mathcal{H}}^2\le Ct^{-d-1}\int_{B((z,t),ct)}\|\nabla v(w,s)\|_{\mathcal{H}}^2\,dw\,ds\le Ct^{-2}(A_\eta v(x))^2.

Integrating along the horizontal segment proves the assertion. □

We now give the one-parameter Hilbert-valued results required in the proofs of Theorems 1.2 and 1.1. For q∈L2(Rd;H)q\in L^2(\mathbb{R}^d;\mathcal{H}), write v(x,t)=Ptq(x)v(x,t)=P_tq(x) and define

N+q(x)=sup⁡t>0∥v(x,t)∥H,N^+q(x)=\sup_{t>0}\|v(x,t)\|_{\mathcal{H}},
Naq(x)=sup⁡t>0,∣y−x∣<at∥v(y,t)∥H,N_aq(x)=\sup_{\substack{t>0,\\ |y-x|<at}}\|v(y,t)\|_{\mathcal{H}},
Abq(x)=(∫Γb(x)∥∇v(y,t)∥H2t dy dtvd(bt)d)1/2.A_bq(x)=\left(\int_{\Gamma_b(x)}\|\nabla v(y,t)\|_{\mathcal{H}}^2\frac{t\,dy\,dt}{v_d(bt)^d}\right)^{1/2}.

Lemma 2.6. For fixed b>0b>0, there exists δ>0\delta>0 such that, for any measurable E⊂RdE\subset\mathbb{R}^{d} with ∣Ec∣<∞|E^{c}|<\infty, the corresponding set GE={x:M1Ec(x)≤δ}G_E=\{x:M1_{E^{c}}(x)\le\delta\} satisfies

∣GEc∣≤C∣Ec∣,∫GE(N+q)2 dx≤C∫E(Abq)2 dx,(23)|G_E^{c}|\le C|E^{c}|,\qquad\int_{G_E}(N^{+}q)^{2}\,dx\le C\int_{E}(A_bq)^{2}\,dx, \tag*{(23)}

with constants independent of the Hilbert space H\mathcal{H} and b,q,Eb,q,E.

Proof. Put Wb(y,t)=B(y,bt)W_b(y,t)=B(y,bt) and ρE(y,t)=∣E∩Wb(y,t)∣/∣Wb(y,t)∣\rho_E(y,t)=|E\cap W_b(y,t)|/|W_b(y,t)|. Tonelli’s theorem yields

∫E(Abq)2 dx=∫Rd∫0∞∥∇v(y,t)∥H2ρE(y,t)t dt dy.\int_E(A_bq)^2\,dx=\int_{\mathbb{R}^{d}}\int_0^\infty\|\nabla v(y,t)\|_{\mathcal{H}}^2\rho_E(y,t)t\,dt\,dy.

Since ∥∂tv∥H≤∥∇v∥H\|\partial_t v\|_{\mathcal{H}}\le\|\nabla v\|_{\mathcal{H}}, setting H=1{ρE≥1/2}∂tvH=1_{\{\rho_E\ge1/2\}}\partial_t v gives ∥H∥Ed(H)2≤2∫E(Abq)2 dx\|H\|_{\mathcal{E}_d(\mathcal{H})}^{2}\le2\int_E(A_bq)^2\,dx. For GE={M1Ec≤δ}G_E=\{M1_{E^c}\le\delta\}, the weak type (1,1)(1,1) bound of MM gives ∣GEc∣≤Cδ−1∣Ec∣|G_E^c|\le C\delta^{-1}|E^c|. If ∣y−x∣<t/2|y-x|<t/2, then B(y,bt)⊂B(x,(b+1/2)t)B(y,bt)\subset B(x,(b+1/2)t), so

1−ρE(y,t)≤CbM1Ec(x).1-\rho_E(y,t)\le C_bM1_{E^c}(x).

Taking δ≤(2Cb)−1\delta\le(2C_b)^{-1} ensures ρE(y,t)≥1/2\rho_E(y,t)\ge1/2. The support of the reconstruction kernel ψt−s\psi_{t-s} satisfies ∣y−x∣<(t−s)/2<t/2|y-x|<(t-s)/2<t/2. Hence, for x∈GEx\in G_E, all derivative evaluation points in the reconstruction lie in {ρE≥1/2}\{\rho_E\ge1/2\}. By the one-parameter reconstruction formula (Lemma 2.3) and Lemma 2.2,

∫GE(N+q)2 dx≤C∫GEsup⁡s>0∥TsH(x)∥H2 dx≤C∥H∥Ed(H)2.\int_{G_E}(N^{+}q)^2\,dx\le C\int_{G_E}\sup_{s>0}\|T_sH(x)\|_{\mathcal{H}}^2\,dx\le C\|H\|_{\mathcal{E}_d(\mathcal{H})}^{2}.

This proves the lemma.

In Lemma 2.6, take E={Abq≤λ}E=\{A_bq\le\lambda\}. The definition of Φλ\Phi_\lambda gives

Φλ(N+q)≤C[∫{Abq≤λ}(Abq)2 dx+λ2∣{Abq>λ}∣]=CΦλ(Abq).(24)\Phi_\lambda(N^{+}q)\le C\left[\int_{\{A_bq\le\lambda\}}(A_bq)^2\,dx+\lambda^2|\{A_bq>\lambda\}|\right]=C\Phi_\lambda(A_bq). \tag*{(24)}

Proposition 2.7. For every q∈L2(Rd;H)q\in L^2(\mathbb{R}^{d};\mathcal{H}) and a,b>0a,b>0,

Φλ(Naq)≤CΦλ(Abq),(25)\Phi_\lambda(N_aq)\le C\Phi_\lambda(A_bq), \tag*{(25)}

where CC is independent of H\mathcal{H}, λ\lambda, and qq, and depends only on a,ba,b and the dimension dd.

Proof. Take η>a\eta>a. Lemma 2.5 gives Naq≤N+q+CAηqN_aq\le N^{+}q+CA_\eta q. Applying Lemma 2.1, (24), and Lemma 2.4 in succession yields

Φλ(Naq)≤C{Φλ(N+q)+Φλ(Aηq)}≤CΦλ(Aηq)≤CΦλ(Abq).\Phi_\lambda(N_aq)\le C\{\Phi_\lambda(N^{+}q)+\Phi_\lambda(A_\eta q)\}\le C\Phi_\lambda(A_\eta q)\le C\Phi_\lambda(A_bq).

This proves the proposition.

The preceding proposition yields the one-parameter version of (2). Thus Proposition 2.7 gives a new proof of that inequality, avoiding Green’s formula and the treatment of sawtooth regions. We next extend this approach to the multiparameter setting, treating the two-parameter case as an example.

Proofs of the multiparameter theorems

Building on the one-parameter Hilbert-valued estimates in § 2, we establish a product operator estimate, a product reconstruction formula, and geometric lemmas for the product setting, and then prove the main theorems.

Two-parameter reconstruction and product geometric lemmas

Apply the preceding constructions in each spatial factor. Write ψi,r=ψdi,r\psi_{i,r}=\psi_{d_i,r} and χi,r=χdi,r\chi_{i,r}=\chi_{d_i,r}, with corresponding operators Si,Ti,s,Ei,sS_i,T_{i,s},E_{i,s}, positive kernels Ki,tK_{i,t}, and averaging operators Vi,sh=χi,s∗ihV_{i,s}h=\chi_{i,s}*_{i}h. For a Hilbert space H\mathcal{H}, define

∥H∥En,m(H)2=∫(0,∞)2∫Rn×Rmt1t2∣H(y1,t1;y2,t2)∣2 dy1 dy2 dt1 dt2.(26)\|H\|_{\mathcal{E}_{n,m}(\mathcal{H})}^{2}=\int_{(0,\infty)^{2}}\int_{\mathbb{R}^{n}\times\mathbb{R}^{m}}t_{1}t_{2}|H(y_{1},t_{1};y_{2},t_{2})|^{2}\,dy_{1}\,dy_{2}\,dt_{1}\,dt_{2}. \tag*{(26)}

When no confusion can arise, we abbreviate ∥H∥H\|H\|_{\mathcal{H}} as ∣H∣|H|. Here Q+\mathbb{Q}_{+} denotes the positive rational numbers.

Proposition 3.1. For every Hilbert space H\mathcal{H} and H∈En,m(H)H\in\mathcal{E}_{n,m}(\mathcal{H}),

∥sup⁡s1>0,s2>0∣T1,s1T2,s2H(⋅)∣∥L2(X)2≤C∥H∥En,m(H)2.(27)\left\|\sup_{s_{1}>0,s_{2}>0}|T_{1,s_{1}}T_{2,s_{2}}H(\cdot)|\right\|_{L^{2}(X)}^{2}\leq C\|H\|_{\mathcal{E}_{n,m}(\mathcal{H})}^{2}. \tag*{(27)}

Proof. By density, we may assume that HH is smooth and compactly supported in Rn×Rm×(0,∞)2\mathbb{R}^{n}\times\mathbb{R}^{m}\times(0,\infty)^{2}. Operators acting in different variables commute, and

T1,s1T2,s2=V1,s1V2,s2S1S2+E1,s1V2,s2S2+V1,s1S1E2,s2+E1,s1E2,s2.T_{1,s_{1}}T_{2,s_{2}}=V_{1,s_{1}}V_{2,s_{2}}S_{1}S_{2}+E_{1,s_{1}}V_{2,s_{2}}S_{2} +V_{1,s_{1}}S_{1}E_{2,s_{2}}+E_{1,s_{1}}E_{2,s_{2}}.

Denote the pointwise suprema of the H\mathcal{H}-norms of these four terms by U11,U10,U01,U00U_{11},U_{10},U_{01},U_{00}, respectively. The first term satisfies

U11≤CM1M2∣S1S2H∣,∥U11∥22≤C∥S1S2H∥22≤C∥H∥En,m(H)2.U_{11}\leq CM_{1}M_{2}|S_{1}S_{2}H|,\qquad\|U_{11}\|_{2}^{2}\leq C\|S_{1}S_{2}H\|_{2}^{2}\leq C\|H\|_{\mathcal{E}_{n,m}(\mathcal{H})}^{2}.

In the second term, E1,s1E_{1,s_{1}} and V2,s2V_{2,s_{2}} act in different variables and hence commute, so

U10≤CM2(sup⁡s1>0∣E1,s1S2H∣).U_{10}\leq CM_{2}\left(\sup_{s_{1}>0}|E_{1,s_{1}}S_{2}H|\right).

By (2.7), Fubini’s theorem, and (2.6) applied in the second variable for each fixed (y1,t1)(y_{1},t_{1}),

∥sup⁡s1>0∣E1,s1S2H∣∥22≤C∫Rn∫0∞t1∥(S2H)t1(y1,⋅)∥L2(Rm;H)2 dt1 dy1≤C∥H∥En,m(H)2.\left\|\sup_{s_{1}>0}|E_{1,s_{1}}S_{2}H|\right\|_{2}^{2} \leq C\int_{\mathbb{R}^{n}}\int_{0}^{\infty}t_{1}\|(S_{2}H)_{t_{1}}(y_{1},\cdot)\|_{L^{2}(\mathbb{R}^{m};\mathcal{H})}^{2}\,dt_{1}\,dy_{1} \leq C\|H\|_{\mathcal{E}_{n,m}(\mathcal{H})}^{2}.

The L2L^{2} boundedness of M2M_{2} therefore gives ∥U10∥22≤C∥H∥En,m(H)2\|U_{10}\|_{2}^{2}\leq C\|H\|_{\mathcal{E}_{n,m}(\mathcal{H})}^{2}. The third term is treated symmetrically.

For the final term, apply (2.7) successively, first in the first variable and then in the second, moving the suprema over the time variables inside the nonnegative integrals. We obtain

U00(x1,x2)2≤C∫0∞∫0∞t1t2∫Rn∫RmK1,t1(x1−y1)K2,t2(x2−y2)×∣H(y1,t1;y2,t2)∣2 dy2 dy1 dt2 dt1.U_{00}(x_{1},x_{2})^{2}\leq C\int_{0}^{\infty}\int_{0}^{\infty}t_{1}t_{2}\int_{\mathbb{R}^{n}}\int_{\mathbb{R}^{m}}K_{1,t_{1}}(x_{1}-y_{1})K_{2,t_{2}}(x_{2}-y_{2}) \times|H(y_{1},t_{1};y_{2},t_{2})|^{2}\,dy_{2}\,dy_{1}\,dt_{2}\,dt_{1}.

Integrating in x1,x2x_{1},x_{2} and using ∥Ki,t∥1=∥ki∥1\|K_{i,t}\|_{1}=\|k_{i}\|_{1} yields the L2L^{2} estimate for the double error term. Combining all four estimates proves (3.2).

We next give the two-parameter reconstruction formula for functions harmonic in each group of variables.

Lemma 3.2. For the product Poisson extension u(y1,t1;y2,t2)=Pt1(1)Pt2(2)f(y1,y2)u(y_1,t_1;y_2,t_2)=P_{t_1}^{(1)}P_{t_2}^{(2)}f(y_1,y_2), we have

u(y1,s1;y2,s2)=(cncm)−1T1,s1T2,s2(∂t1∂t2u(⋅,t1;⋅,t2))(y1,y2).(28)u(y_1,s_1;y_2,s_2)=(c_nc_m)^{-1}T_{1,s_1}T_{2,s_2}(\partial_{t_1}\partial_{t_2}u(\cdot,t_1;\cdot,t_2))(y_1,y_2). \tag*{(28)}

The formula on the right uses only values of uu at points satisfying ti>sit_i>s_i and ∣yi−xi∣<(ti−si)/2<ti/2|y_i-x_i|<(t_i-s_i)/2<t_i/2.

Proof. First apply Lemma 2.3 in the first spatial variable, treating the second spatial variable and its time variable as Hilbert-valued parameters. Apply the same lemma in the second spatial variable. Multiplying the constants from the two one-parameter reconstructions and canceling the two minus signs gives (3.3). Since ψi,ti−si\psi_{i,t_i-s_i} is supported in B(0,(ti−si)/2)B(0,(t_i-s_i)/2), the reconstruction integral uses only values of uu in the stated cones.

Lemma 3.3. For every measurable set E⊂XE\subset X,

∫E(Aβf)2 dx=∫(0,∞)2∫X∣∇1⊥∇2⊥u(y,t)∣2ρE(y,t)t1t2 dy dt.(29)\int_E(A_\beta f)^2\,dx=\int_{(0,\infty)^2}\int_X|\nabla_1^\perp\nabla_2^\perp u(y,t)|^2\rho_E(y,t)t_1t_2\,dy\,dt. \tag*{(29)}

where

ρE(y,t)=ρEβ(y,t)=∣E∩Wβ(y,t)∣∣Wβ(y,t)∣,Wβ(y,t)=B(y1,β1t1)×B(y2,β2t2),\rho_E(y,t)=\rho_E^\beta(y,t)=\frac{|E\cap W_\beta(y,t)|}{|W_\beta(y,t)|},\qquad W_\beta(y,t)=B(y_1,\beta_1t_1)\times B(y_2,\beta_2t_2),

and β∈(0,∞)2\beta\in(0,\infty)^2. The identity remains valid if ∇1⊥∇2⊥u(y,t)\nabla_1^\perp\nabla_2^\perp u(y,t) is replaced by an arbitrary measurable function and AβA_\beta by the square function SβS_\beta in Lemma 3.5 below.

Proof. Substitute the definition of the area integral function and apply Tonelli’s theorem. The details are omitted.

Lemma 3.4. Suppose ai,βi>0a_i,\beta_i>0. There exists δ>0\delta>0, depending only on these numbers and the dimensions, such that for every E⊂XE\subset X with ∣Ec∣<∞|E^c|<\infty, the set GE={x:M1M21Ec(x)≤δ}G_E=\{x:M_1M_2\mathbf{1}_{E^c}(x)\le\delta\} satisfies

∣GEc∣≤C∣Ec∣,(30)|G_E^c|\le C|E^c|, \tag*{(30)}
ρE(y,t)≥12if x∈GE,∣yi−xi∣<aiti (i=1,2).\rho_E(y,t)\ge\frac{1}{2}\quad\text{if }x\in G_E,\quad|y_i-x_i|<a_it_i\ (i=1,2).

Proof. Let B=EcB=E^c. By the L2L^2 boundedness of M1M2M_1M_2,

∣{M1M21B>δ}∣≤δ−2∥M1M21B∥22≤Cδ−2∣B∣.|\{M_1M_2\mathbf{1}_B>\delta\}|\le\delta^{-2}\|M_1M_2\mathbf{1}_B\|_2^2\le C\delta^{-2}|B|.

For the second estimate, put Bi∗=B(xi,(βi+ai)ti)B_i^*=B(x_i,(\beta_i+a_i)t_i). Then B(yi,βiti)⊂Bi∗B(y_i,\beta_it_i)\subset B_i^*, so

1−ρE(y,t)≤Cβ,a∣B1∗∣∣B2∗∣∫B1∗×B2∗1B(w) dw≤Cβ,aM1M21B(x).1-\rho_E(y,t)\le\frac{C_{\beta,a}}{|B_1^*||B_2^*|}\int_{B_1^*\times B_2^*}\mathbf{1}_B(w)\,dw\le C_{\beta,a}M_1M_2\mathbf{1}_B(x).

Taking δ≤(2Cβ,a)−1\delta\le(2C_{\beta,a})^{-1} proves the second estimate.

Lemma 3.5. Let FF be a Hilbert-valued measurable function on the product space. Define

ScF(x)=(∬Γc(x)∥F(y,t)∥H2∏i=12ti dyi dtivdi(citi)di)1/2.S_cF(x)=\left(\iint_{\Gamma_c(x)}\|F(y,t)\|_{\mathcal H}^2\prod_{i=1}^2\frac{t_i\,dy_i\,dt_i}{v_{d_i}(c_it_i)^{d_i}}\right)^{1/2}.

Then

Φλ(SaF)≤CΦλ(SbF),(31)\Phi_\lambda(S_aF)\le C\Phi_\lambda(S_bF), \tag*{(31)}

where CC depends only on a,ba,b and the dimensions.

Proof. The assertion is immediate if the right-hand side is infinite. Otherwise, set E={SbF≤λ}E=\{S_bF\leq\lambda\}, B=EcB=E^c, and G={M1M21B≤ε}G=\{M_1M_2\mathbf{1}_B\leq\varepsilon\}. Then ∣Gc∣≤Cε∣B∣|G^c|\leq C_\varepsilon|B|. If G∩Wa(z)G\cap W_a(z) is nonempty, choose a point in this intersection as a base point. By Lemma 3.4, for a sufficiently small fixed ε\varepsilon we have ρEb(z)≥1/2\rho_E^b(z)\geq1/2, and hence

∣G∩Wa(z)∣∣Wa(z)∣≤2∣E∩Wb(z)∣∣Wb(z)∣.\frac{|G\cap W_a(z)|}{|W_a(z)|}\leq2\frac{|E\cap W_b(z)|}{|W_b(z)|}.

The inequality also holds when its left-hand side is zero. Tonelli’s theorem gives ∫G(SaF)2≤2∫E(SbF)2\int_G(S_aF)^2\leq2\int_E(S_bF)^2. Splitting the integral defining Φλ\Phi_\lambda into the parts over GG and GcG^c proves (31).

Proof of Theorem 1.2

Proof of Theorem 1.2. By a limiting argument, we may assume that f∈Cc∞(Rn×Rm)f\in C^\infty_c(\mathbb{R}^n\times\mathbb{R}^m) and u=Pt1(1)Pt2(2)fu=P_{t_1}^{(1)}P_{t_2}^{(2)}f. In Lemma 3.4, take a1=a2=1/2a_1=a_2=1/2 and choose GEG_E accordingly. Then

ρE(y,t)≥12whenever x∈GE,∣yi−xi∣<ti/2.\rho_E(y,t)\geq\frac{1}{2}\qquad\text{whenever }x\in G_E,\quad|y_i-x_i|<t_i/2.

Set

ΩE={(y,t):ρE(y,t)≥1/2},H=1ΩE∂t1∂t2u.\Omega_E=\{(y,t):\rho_E(y,t)\geq1/2\},\qquad H=\mathbf{1}_{\Omega_E}\partial_{t_1}\partial_{t_2}u.

Since ∂t1∂t2\partial_{t_1}\partial_{t_2} is one component of ∇1⊥∇2⊥\nabla_1^\perp\nabla_2^\perp, Lemma 3.3 gives

∥H∥En,m2≤2∫E(Aβf)2 dx.(32)\|H\|_{\mathcal{E}_{n,m}}^2\leq2\int_E(A_\beta f)^2\,dx. \tag*{(32)}

For x∈GEx\in G_E, Lemma 3.2 and the preceding geometric property show that every point used in the reconstruction belongs to ΩE\Omega_E. Thus ∂t1∂t2u\partial_{t_1}\partial_{t_2}u can be replaced by HH in the reconstruction integral. Proposition 3.1 gives

∫GE(N+f)2 dx≤C∫GEsup⁡s1>0,s2>0∣T1,s1T2,s2H∣2 dx≤C∥H∥En,m2≤C∫E(Aβf)2 dx.\int_{G_E}(N^+f)^2\,dx\leq C\int_{G_E}\sup_{s_1>0,s_2>0}|T_{1,s_1}T_{2,s_2}H|^2\,dx \leq C\|H\|_{\mathcal{E}_{n,m}}^2\leq C\int_E(A_\beta f)^2\,dx.

This proves Theorem 1.2.

Lemma 2.6 gives the corresponding one-parameter local estimate for Hilbert-valued functions.

Corollary 3.6. Under the hypotheses of Theorem 1.2, for every f∈L2(Rn×Rm)f\in L^2(\mathbb{R}^n\times\mathbb{R}^m) and λ>0\lambda>0,

Φλ(N+f)≤CΦλ(Aβf).(33)\Phi_\lambda(N^+f)\leq C\Phi_\lambda(A_\beta f). \tag*{(33)}

Proof. Take E={Aβf≤λ}E=\{A_\beta f\leq\lambda\}, so that ∣Ec∣<∞|E^c|<\infty. Split the integral defining Φλ(N+f)\Phi_\lambda(N^+f) over GEG_E and GEcG_E^c, and apply Theorem 1.2.

Proof of Theorem 1.1

Lemma 3.7. For ηi>αi\eta_i>\alpha_i, f∈Cc∞(Rn×Rm)f\in C^\infty_c(\mathbb{R}^n\times\mathbb{R}^m), u=Pt1(1)Pt2(2)fu=P_{t_1}^{(1)}P_{t_2}^{(2)}f, and zi=(yi,ti)z_i=(y_i,t_i), define

Q1(x)=sup⁡z1∈Γ1,α1(x1)(∫Γ2,η2(x2)∣∇2⊥u(z1,z2)∣2t2 dy2 dt2vm(η2t2)m)1/2,Q_1(x)=\sup_{z_1\in\Gamma_{1,\alpha_1}(x_1)}\left(\int_{\Gamma_{2,\eta_2}(x_2)}|\nabla_2^\perp u(z_1,z_2)|^2\frac{t_2\,dy_2\,dt_2}{v_m(\eta_2t_2)^m}\right)^{1/2},
Q2(x)=sup⁡z2∈Γ2,α2(x2)(∫Γ1,η1(x1)∣∇1⊥u(z1,z2)∣2t1 dy1 dt1vn(η1t1)n)1/2.Q_2(x)=\sup_{z_2\in\Gamma_{2,\alpha_2}(x_2)}\left(\int_{\Gamma_{1,\eta_1}(x_1)}|\nabla_1^\perp u(z_1,z_2)|^2\frac{t_1\,dy_1\,dt_1}{v_n(\eta_1t_1)^n}\right)^{1/2}.

Then

Φλ(Q1)+Φλ(Q2)≤CΦλ(Aβf).(34)\Phi_\lambda(Q_1)+\Phi_\lambda(Q_2)\leq C\Phi_\lambda(A_\beta f). \tag*{(34)}

Proof. We first estimate Q1Q_1. Fix η2\eta_2, and let

dμ2,η2(y2,t2)=t2 dy2 dt2vm(η2t2)m,H2=L2(Rm×(0,∞),dμ2,η2;Cm+1).d\mu_{2,\eta_2}(y_2,t_2)=\frac{t_2\,dy_2\,dt_2}{v_m(\eta_2t_2)^m},\qquad\mathcal{H}_2=L^2\left(\mathbb{R}^m\times(0,\infty),d\mu_{2,\eta_2};\mathbb{C}^{m+1}\right).

This is a fixed Hilbert space independent of x2x_2. For each x2x_2, define the H2\mathcal{H}_2-valued function

qx2(y1)(y2,t2)=1{∣y2−x2∣<η2t2}∇2⊥Pt2(2)f(y1,y2).q_{x_2}(y_1)(y_2,t_2)=\mathbf{1}_{\{|y_2-x_2|<\eta_2t_2\}}\nabla_2^\perp P_{t_2}^{(2)}f(y_1,y_2).

Fubini’s theorem and the Fourier transform give

∫Rm∥qx2∥L2(Rn;H2)2 dx2=∫Rn∫Rm×(0,∞)t2∣∇2⊥Pt2(2)f(y1,y2)∣2 dy2 dt2 dy1=12∥f∥22.(35)\int_{\mathbb{R}^m}\|q_{x_2}\|_{L^2(\mathbb{R}^n;\mathcal{H}_2)}^2\,dx_2 =\int_{\mathbb{R}^n}\int_{\mathbb{R}^m\times(0,\infty)}t_2\left|\nabla_2^\perp P_{t_2}^{(2)}f(y_1,y_2)\right|^2\,dy_2\,dt_2\,dy_1 =\frac{1}{2}\|f\|_2^2. \tag*{(35)}

Thus qx2∈L2(Rn;H2)q_{x_2}\in L^2(\mathbb{R}^n;\mathcal{H}_2) for almost every x2x_2. The Hilbert-valued nontangential maximal function of its Poisson extension in the first variable is Q1(⋅,x2)Q_1(\cdot,x_2), and the corresponding area integral function is A(β1,η2)f(⋅,x2)A_{(\beta_1,\eta_2)}f(\cdot,x_2). Applying Proposition 2.7 for each x2x_2, integrating, and then using Lemma 3.5, we obtain

Φλ(Q1)≤CΦλ(A(β1,η2)f)≤CΦλ(Aβf).\Phi_\lambda(Q_1)\leq C\Phi_\lambda(A_{(\beta_1,\eta_2)}f)\leq C\Phi_\lambda(A_\beta f).

The estimate for Q2Q_2 is symmetric: fix x1x_1 and use the one-parameter argument with values in

H1=L2(Rn×(0,∞),t1 dy1 dt1vn(η1t1)n;Cn+1).\mathcal{H}_1=L^2\left(\mathbb{R}^n\times(0,\infty),\frac{t_1\,dy_1\,dt_1}{v_n(\eta_1t_1)^n};\mathbb{C}^{n+1}\right).

This gives Φλ(Q2)≤CΦλ(Aβf)\Phi_\lambda(Q_2)\leq C\Phi_\lambda(A_\beta f). □

Proof of Theorem 1.1. We may assume that f∈Cc∞(Rn×Rm)f\in C^\infty_c(\mathbb{R}^n\times\mathbb{R}^m) and uu is its two-parameter Poisson extension. Consider the decomposition

u(z1,z2)−u(x1,t1;x2,t2)=(u(z1,z2)−u(z1;x2,t2))+(u(z1;x2,t2)−u(x1,t1;x2,t2)).u(z_1,z_2)-u(x_1,t_1;x_2,t_2)=\bigl(u(z_1,z_2)-u(z_1;x_2,t_2)\bigr)+\bigl(u(z_1;x_2,t_2)-u(x_1,t_1;x_2,t_2)\bigr).

For the first difference, fix z1z_1, regard u(z1,z2)u(z_1,z_2) as a harmonic function in the second group of variables, and apply Lemma 2.5 with a=α2a=\alpha_2 and η=η2\eta=\eta_2. Its right-hand side is controlled by Q1(x)Q_1(x). For the second difference, fix z2=(x2,t2)z_2=(x_2,t_2) and apply the same lemma in the first group of variables with a=α1a=\alpha_1 and η=η1\eta=\eta_1; its right-hand side is controlled by Q2(x)Q_2(x). Taking suprema over z1,z2z_1,z_2 gives

Nαf≤N+f+C(Q1+Q2).(36)N_\alpha f\leq N^+f+C(Q_1+Q_2). \tag*{(36)}

By Lemma 2.1, (3.8), and Lemma 3.7,

Φλ(Nαf)≤C{Φλ(N+f)+Φλ(Q1)+Φλ(Q2)}≤CΦλ(Aβf).\Phi_\lambda(N_\alpha f)\leq C\{\Phi_\lambda(N^+f)+\Phi_\lambda(Q_1)+\Phi_\lambda(Q_2)\}\leq C\Phi_\lambda(A_\beta f).

This proves Theorem 1.1. □

Corollary 3.8. Under the hypotheses of Theorem 1.1,

∣{Nαf>λ}∣≤C∣{Aβf>λ}∣+Cλ−2∫{Aβf≤λ}(Aβf)2 dx.(37)|\{N_\alpha f>\lambda\}|\leq C|\{A_\beta f>\lambda\}|+C\lambda^{-2}\int_{\{A_\beta f\leq\lambda\}}(A_\beta f)^2\,dx. \tag*{(37)}

Moreover, for every 0<p<20<p<2,

∥Nαf∥Lp≤Cp∥Aβf∥Lp,∥Nαf∥Lp,∞≤Cp∥Aβf∥Lp,∞.(38)\|N_\alpha f\|_{L^p}\leq C_p\|A_\beta f\|_{L^p},\qquad\|N_\alpha f\|_{L^{p,\infty}}\leq C_p\|A_\beta f\|_{L^{p,\infty}}. \tag*{(38)}

Proof. Since λ2∣{Nαf>λ}∣≤Φλ(Nαf)≤CΦλ(Aβf)\lambda^{2}\lvert\{N_{\alpha}f>\lambda\}\rvert\leq\Phi_{\lambda}(N_{\alpha}f)\leq C\Phi_{\lambda}(A_{\beta}f), splitting the right-hand side over {Aβf≤λ}\{A_{\beta}f\leq\lambda\} and {Aβf>λ}\{A_{\beta}f>\lambda\} gives (37). The strong type assertion follows from Tonelli’s theorem and

∫0∞λp−3min⁡{s2,λ2} dλ=2spp(2−p),0<p<2.\int_{0}^{\infty}\lambda^{p-3}\min\{s^{2},\lambda^{2}\}\,d\lambda=\frac{2s^{p}}{p(2-p)},\qquad0<p<2.

For the weak type assertion, writing K=∥F∥Lp,∞K=\lVert F\rVert_{L^{p,\infty}}, we have

Φλ(F)=∫0λ2r∣{F>r}∣ dr≤22−pKpλ2−p.\Phi_{\lambda}(F)=\int_{0}^{\lambda}2r\lvert\{F>r\}\rvert\,dr\leq\frac{2}{2-p}K^{p}\lambda^{2-p}.

Take F=AβfF=A_{\beta}f and apply Theorem 1.1. □\square

Remark 3.9. For any fixed finite number of factors, one can formally expand the tensor product of the operators ViSi+EiV_iS_i+E_i. Each term must be treated separately using the corresponding positive integrable kernel estimates and iterated maximal operator estimates. Thus estimate (2) extends to the multiparameter setting.

Acknowledgments. The author was supported by the National Natural Science Foundation of China (No. 12571109) and the National Key Research and Development Program of the Ministry of Science and Technology of China (No. 2022YFA1005700).

References

  1. [1]R. Bañuelos and C. N. Moore, Sharp estimates for the nontangential maximal function and the Lusin area function in Lipschitz domains, Trans. Amer. Math. Soc. 312 (1989), no. 2, 641–662.
  2. [2]R. Bañuelos and C. N. Moore, Distribution function inequalities for the density of the area integral, Ann. Inst. Fourier 41 (1991), no. 1, 137–171.DOI
  3. [3]D. L. Burkholder and R. F. Gundy, Distribution function inequalities for the area integral, Studia Math. 44 (1972), no. 6, 527–544.DOI
  4. [4]J. Chen, H^p spaces related to multiparameters and their characterizations (in Chinese), M.Sc. thesis, Hangzhou University, 1984.
  5. [5]J. Chen and S. Wang, Comparison between maximal function and square function on positively curved product manifolds, in Chinese Mathematics into the 21st Century (Wu Wen-tsun and Cheng Min-de, eds.), Peking University Press, 1991.
  6. [6]J. Chen, A simple proof of an inequality dominating the area integral by the nontangential maximal function, Pacific J. Math. 190 (1999), no. 1, 41–44.DOI
  7. [7]B. E. J. Dahlberg, Weighted norm inequalities for the Lusin area integral and the nontangential maximal functions for functions harmonic in a Lipschitz domain, Studia Math. 67 (1980), no. 3, 297–314.
  8. [8]C. Fefferman and E. M. Stein, H^p spaces of several variables, Acta Math. 129 (1972), 137–193.
  9. [9]R. Fefferman, R. F. Gundy, M. L. Silverstein and E. M. Stein, Inequalities for ratios of functionals of harmonic functions, Proc. Natl. Acad. Sci. USA 79 (1982), no. 24, 7958–7960.pubmed.ncbi.nlm.nih.gov/16593265
  10. [10]R. F. Gundy and E. M. Stein, Hp theory for the poly-disc, Proc. Natl. Acad. Sci. USA 76 (1979), no. 3, 1026–1029.DOI
  11. [11]R. F. Gundy and R. L. Wheeden, Weighted integral inequalities for the nontangential maximal function, Lusin area integral, and Walsh–Paley series, Studia Math. 49 (1974), no. 2, 107–124.DOI
  12. [12]J. Li, On Fefferman–Stein type inequality for Shilov boundaries and applications, Ann. Sc. Norm. Super. Pisa Cl. Sci. 27 (2026), no. 3, 1329–1354.
  13. [13]M.-P. Malliavin and P. Malliavin, Intégrales de Lusin-Calderón pour les fonctions biharmoniques, Bull. Sci. Math. (2) 101 (1977), 357–384.
  14. [14]K. G. Merryfield, On the area integral, Carleson measures and Hp in the polydisc, Indiana Univ. Math. J. 34 (1985), no. 3, 663–685.
  15. [15]T. Murai and A. Uchiyama, Good λ inequalities for the area integral and the nontangential maximal function, Studia Math. 83 (1986), no. 3, 251–262.DOI
  16. [16]S. Sato, Lusin functions and nontangential maximal functions in the Hp theory on the product of upper half-spaces, Tohoku Math. J. (2) 37 (1985), no. 1, 1–13.DOI

Paper details

vlnc:W_9py384bn
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