Skip to content
TheChatArchiveThe Chat Archive
gpt-4o-2024-08-06· OpenAIAi2 WildChat-4.8M (AI2)

You are a helpful assistant generating synthetic data that captures System 1 and System…

  1. User

    You are a helpful assistant generating synthetic data that captures System 1 and System 2 thinking, creativity, and metacognitive reflection. Follow these steps in sequence, using tags [sys1] and [end sys1] for System 1 sections and [sys2] and [end sys2] for System 2 sections.

    1. Identify System 1 and System 2 Thinking Requirements:

      • Carefully read the text.
      • Identify parts of the text that require quick, straightforward responses (System 1). Mark these sections with [sys1] and [end sys1].
      • Identify parts that require in-depth, reflective thinking (System 2), marked with [sys2] and [end sys2].
    2. Apply Step-by-Step Problem Solving with Creativity and Metacognitive Reflection for System 2 Sections:

      2.1 Understand the Problem:

      • Objective: Fully comprehend the issue, constraints, and relevant context.
      • Reflection: "What do I understand about this issue? What might I be overlooking?"
      • Creative Perspective: Seek hidden patterns or possibilities that could reveal deeper insights or innovative connections.

      2.2 Analyze the Information:

      • Objective: Break down the problem logically.
      • Reflection: "Am I considering all factors? Are there any assumptions that need challenging?"
      • Creative Perspective: Explore unique patterns or overlooked relationships in the data that could add depth to the analysis.

      2.3 Generate Hypotheses:

      • Objective: Propose at least 10 hypotheses, each with a Confidence Score (0.0 to 1.0) and Creative Score (0.0 to 1.0), reflecting originality, surprise, and utility.
      • Reflection: "Have I explored all possible explanations or approaches, both conventional and unconventional?"
      • Creative Perspective: Consider novel angles that might provide unexpected insights.

      2.4 Anticipate Future Steps and Obstacles:

      • Objective: Make predictions, accounting for potential outcomes and obstacles.
      • Reflection: "What challenges might I face? Is my plan flexible for different scenarios?"
      • Creative Perspective: Visualize unforeseen outcomes and adapt plans to make use of them effectively.

      2.5 Evaluate Hypotheses:

      • Objective: Assess hypotheses based on feasibility, risk, and potential impact.
      • Evaluation: Refine Confidence and Creative Scores as needed.
      • Reflection: "Am I unbiased in my assessment? Which options fit best with the overall objectives?"
      • Creative Perspective: Identify hidden opportunities or overlooked details in each hypothesis.

      2.6 Select the Best Hypothesis:

      • Objective: Choose the most promising, strategic hypothesis.
      • Reflection: "Why does this hypothesis stand out? How does it uniquely address the issue?"
      • Creative Perspective: Consider any underutilized potential in the selected approach.

      2.7 Implement the Hypothesis:

      • Objective: Outline actionable steps for testing the hypothesis.
      • Reflection: "Is this plan practical? What resources or preparation are required?"
      • Creative Perspective: Refine steps to maximize effectiveness and yield unexpected benefits.

      2.8 Monitor and Review Progress:

      • Objective: Review progress, noting areas for improvement.
      • Reflection: "What’s working well? What could be improved?"
      • Creative Perspective: Look for emerging patterns that could refine future approaches.

      2.9 Reflect and Capture Insights:

      • Objective: Summarize lessons learned and insights gained for future reference.
      • Reflection: "What new understanding has emerged from this process?"
      • Creative Perspective: Identify innovative insights or patterns that could be applied to similar challenges.
    3. Generate Text Output with Interleaved System 1 and System 2 Responses:

      • Use the tags [sys1] and [sys2] throughout.
      • Aim for a lengthier, detailed response. Combine both direct, straightforward System 1 insights and reflective, deeply analytical System 2 segments to capture a blend of quick observations and thoughtful analysis.

    Example Input Text:

    You are tasked with analyzing a sudden shift in customer preferences in a tech market. There has been a noticeable decline in demand for physical devices, with customers increasingly interested in digital-only options. The challenge is to understand this shift, generate hypotheses for why it might be occurring, and explore potential strategies for adapting to this new trend.


    Processed Output with System 1 and System 2 Thinking:

    The tech market is experiencing a shift in customer demand, moving away from physical devices toward digital-only options.

    [sys1] Customer demand has shifted from physical devices to digital-only options. We need to understand why this change is happening and find strategies to address it. Key factors may include cost, convenience, and technological trends. [end sys1]

    [sys2] 2.1 Understand the Problem: The goal is to understand the underlying reasons for the shift in demand from physical devices to digital-only options, then develop a strategy to adapt.

    Reflection: "I understand that preferences are changing, but what might be driving this? Is it primarily cost, or are there other factors such as convenience or sustainability?"

    Creative Perspective: "Could there be a larger trend in digital minimalism or a preference for eco-friendly solutions that we’re missing? What unseen motivations might explain this shift?"

    2.2 Analyze the Information: There are multiple possible factors driving this shift, from economic influences to cultural shifts. It’s essential to isolate each factor and understand its impact.

    Reflection: "Am I fully considering the various economic and social influences? Could there be a technological factor, like better internet speeds, that makes digital-only products more accessible?"

    Creative Perspective: "Are there patterns or trends in other markets that could shed light on this shift? Could this be part of a larger trend toward virtual experiences?"

    2.3 Generate Hypotheses:

    1. Customers prefer digital options due to lower costs. (Confidence: 0.8, Creative: 0.4)
    2. There’s a growing trend toward minimalism and reduced physical clutter. (Confidence: 0.7, Creative: 0.7)
    3. Digital products offer greater flexibility and ease of use. (Confidence: 0.6, Creative: 0.6)
    4. Environmental concerns are pushing consumers away from physical goods. (Confidence: 0.6, Creative: 0.8)
    5. Advances in tech make digital-only options more functional. (Confidence: 0.8, Creative: 0.5)
    6. Pandemic-era remote work increased demand for digital solutions. (Confidence: 0.7, Creative: 0.6)
    7. Media coverage of the environmental impact of physical devices affects preferences. (Confidence: 0.5, Creative: 0.7)
    8. There’s an increase in global digital literacy, expanding market access. (Confidence: 0.6, Creative: 0.6)
    9. Customers view digital as more convenient and scalable for future needs. (Confidence: 0.7, Creative: 0.5)
    10. Younger consumers prefer the aesthetics and convenience of digital products. (Confidence: 0.6, Creative: 0.6)

    Reflection: "Have I considered all possible influences? Are there any surprising factors that could explain this shift?"

    Creative Perspective: "Could specific social trends, like the rise of influencer culture or digital-first lifestyles, be influencing customer choices?"

    2.4 Anticipate Future Steps and Obstacles: Objective: Anticipate possible challenges, such as resistance from segments still preferring physical products.

    Reflection: "What market obstacles might we face if we shift our focus to digital-only? Are there sub-segments that still prioritize physical products?"

    Creative Perspective: "Could expanding digital options help us reach a more global audience? Are there emerging trends that we could leverage in our strategy?"

    [end sys2]

    [sys1] To address this shift, consider a strategy that incorporates both digital-only offerings and educational campaigns about the benefits of digital solutions. Use insights from customer feedback and current trends to guide product development. Focus on flexibility and adaptation to cater to different customer segments. [end sys1]


    abstract: 'We study multivariable spectral multipliers F(L1,L2)F(L_1,L_2) acting on Cartesian product of ambient spaces of two self-adjoint operators L1L_1 and L2L_2. We prove that if FF satisfies Hörmander type differentiability condition then the operator F(L1,L2)F(L_1,L_2) is of Calderón-Zygmund type. We apply obtained results to the analysis of quasielliptic operators acting on product of some fractal spaces. The existence and surprising properties of quasielliptic operators have been recently observed in works of Bockelman, Drenning and Strichartz. This paper demonstrates that Riesz type operators corresponding to quasielliptic operators are continuous on LpL^p spaces. This solves the problem posed in [@BS (1.3) p. 1363].' address: 'Adam Sikora, Mathematical Sciences Institute, Bldg 27, Australian National University, ACT 0200 Australia ' author:

    • Adam Sikora date: '28/07/2008' title: Multivariable spectral multipliers and analysis of quasielliptic operators on fractals

    I dedicate this paper to the memory of my teachers Andrzej Hulanicki and Tadeusz Pytlik.

    Introduction

    Suppose that LL is a self-adjoint operator acting on L2(X,μ)L^2(X,\mu), where XX is a measure space with measure μ\mu. Such an operator admits a spectral resolution EL(λ)E_L(\lambda) and for any bounded Borel function F ⁣:R→CF\colon {\mathbf{R}}\to {\mathbf{C}}, we define the operator F(L)F(L) by the formula F(L)=∫RF(λ)dEL(λ).F(L)=\int_{\mathbf{R}}F(\lambda) d E_L(\lambda). By the spectral theorem the operator F(L)F(L) is continuous on L2(X,μ)L^2(X,\mu). Spectral multiplier theorems investigate sufficient conditions on function FF which ensure that the operator F(L)F(L) extends to a bounded operator on LqL^q for some qq in the range 1≤q≤∞1\le q \le \infty. The theory of spectral multipliers constitutes an important field of Harmonic analysis and there exists a vast literature devoted to the topic, see for example [@Ale1; @Ale2; @ChS; @CoS; @DM; @DOS; @He1; @Her1; @Mi; @MM; @MRS] and references within. The main aim of this paper is to develop a theory of multivariable spectral multipliers of two self-adjoint independent operators acting on the Cartesian product of their ambient spaces. One could also consider here three or more operators but, for simplicity, we limit the discussion to the two dimensional version.

    Multivariable spectral multiplier

    We begin our discussion by explaining the definition of multivariable spectral multipliers. We consider two self-adjoint operators LjL_j, j=1,2j=1,2 acting on spaces L2(Xj)L^2(X_j). The tensor product operators L1⊗1L_{1} \otimes 1 and 1⊗L2 1 \otimes L_{2} act on L2(X1×X2)L^2({X}_1 \times {X}_2), where X1×X2{X}_1 \times {X}_2 is the Cartesian product of X1X_1 and X2X_2 with the product measure μ=μ1×μ2\mu=\mu_1 \times \mu_2. To simplify notation we will write L1L_1 and L2L_2 instead of L1⊗1L_{1} \otimes 1 and 1⊗L2 1 \otimes L_{2}. Note that there is a unique spectral decomposition EE such that for all Borel subsets A⊂R2A\subset {\mathbf{R}}^2, E(A)E(A) is a projection on L2(X1×X2)L^2({X}_1 \times{X}_2) and such that for any Borel subsets Aj⊂RA_j \subset {\mathbf{R}}, j=1,2j=1,2 one has E(A1×A2)=EL1(A1)⊗EL2(A2).E(A_1\times{A}_2)=E_{L_1}(A_1)\otimes E_{L_2}(A_2). Hence for any function F ⁣:R2→CF\colon {\mathbf{R}}^2 \to {\mathbf{C}} one can define the operator F(L1,L2)F(L_1,L_2) acting as operators on space L2(X1×X2,)L^2({X}_1 \times {X}_2,) by the formula \labelequwdF(L1,L2)=∫R2F(λ1,λ2)dE(λ1,λ2).\label{equwd} F(L_1,L_2)=\int_{{\mathbf{R}}^2}F(\lambda_1,\lambda_2) d E(\lambda_1,\lambda_2). A straightforward variation of classical spectral theory arguments shows that for any bounded Borel function F ⁣:R2→CF\colon {\mathbf{R}}^2 \to {\mathbf{C}} the operator F(L1,L2)F(L_1,L_2) is continuous on L2(X1×X2)L^2({X}_1 \times {X}_2) and its norm is bounded by ∥F∥L∞\|F\|_{L^\infty}. In this paper we are looking for necessary smoothness conditions on function FF so that the operator F(L1,L2)F(L_1,L_2) is bounded on a range of other Lp(X1×X2)L^p({X}_1 \times {X}_2) spaces. The condition on function FF which we use is a variant of the differentiability condition in Hörmdander-Mikhlin Fourier multiplier result, see [@Her1; @Mi] and [@Her2 Theorem 7.9.6].

    One motivation for multivariable spectral multiplier results comes from Riesz transform like operators, for example aL1+bL2cL1+dL2\mboxorL1L2(cL1+dL2)2,\frac{aL_1+bL_2}{cL_1+dL_2} \quad \mbox{or} \quad \frac{L_1L_2}{(cL_1+dL_2)^2}, where a,b∈Ra,b\in {\mathbf{R}} are arbitrary real numbers and c,d>0c,d>0. Here we assume that the operators LiL_i are positive for i=1,2i=1,2. It is easy to note that then such operators are bounded on L2(X1×X2)L^2({X}_1 \times {X}_2). We prove that under standard assumptions such operators are of Calderón-Zygmund type, that is, they are bounded on all Lp(X1×X2)L^p({X}_1 \times {X}_2) spaces for 1<p<∞1<p<\infty and of weak type (1,1)(1,1). Of course one can consider a much larger family of operators of this type. For the Laplace operators acting on some fractal spaces and for some real numbers c,dc,d the above operators are still bounded on L2(X1×X2)L^2({X}_1 \times {X}_2) (and as we prove it in this paper on other Lp(X1×X2)L^p({X}_1 \times {X}_2) spaces 1<p<∞1<p< \infty) [even though cd<0cd <0]{}. We describe this intriguing phenomenon in the following section.

    Product of fractals and quasielliptic operators

    In [@BS] and [@DS] it is shown that for the Laplace operators defined on some fractal spaces, the set of ratios of eigenvalues have gaps. This means that there are intervals (α,β)(\alpha,\beta), 0<α<β0<\alpha<\beta, such that for any two eigenvalues λi\lambda_i, λj\lambda_j of the Laplacian acting on the same fractal spaces one has λjλi∉(α,β)\frac{\lambda_j}{\lambda_i}\notin (\alpha,\beta). For example if one considers the Dirichlet or Neumann Laplacian on Sierpiński Gasket SGSG then λjλi∉(α,β)\frac{\lambda_j}{\lambda_i}\notin (\alpha,\beta), where \labelgapα=lim⁡n→∞ψn(5)ψn(3)≈2.0611\mboxandβ=lim⁡n→∞ψn(3)ψn+1(3)≈2.4288\label{gap} \alpha=\lim_{n\to \infty} \frac{\psi_n(5)}{\psi_n(3)}\approx 2.0611 \quad \mbox{and} \quad \beta=\lim_{n\to \infty} \frac{\psi_n(3)}{\psi_{n+1}(3)}\approx 2.4288 and ψ(x)=(5−25−4x)/2\psi (x) = (5 -\sqrt{25 - 4x})/2, see [@BS]. The existence of gaps in the set of ratios of eigenvalues has a surprising consequence. Namely one can consider the product of two copies of such fractal spaces and operators L1L_1 and L2L_2 which are copies of the same Laplace operator acting on first and second variable respectively. Now it is easy to notice that if (α,β)(\alpha,\beta) is a gap in the set of ratios of eigenvalues α<γ<β\alpha<\gamma < \beta, a,b∈Ra,b\in {\mathbf{R}}, c,d>0c,d>0 and dc=γ\frac{d}{c}=\gamma then the operators \labelqellaL1+bL2cL1−dL2\mboxorL1L2(cL1−dL2)2,\label{qell} \frac{aL_1+bL_2}{cL_1-dL_2} \quad \mbox{or} \quad \frac{L_1L_2}{(cL_1-dL_2)^2}, are bounded on L2L^2. Following [@BS] we call operators of the form cL1−dL2{cL_1-dL_2} [quasielliptic]{}. It was asked in [@BS] whether the above operators are bounded on other LpL^p spaces. This question is the initial motivation of this paper. We prove that these operators are indeed bounded on all LpL^p spaces for 1<p<∞1<p<\infty and of weak type (1,1)(1,1).

    The idea of multivariable spectral multipliers is of independent interest. It seems to be possible to obtain more general versions of multivariable spectral multipliers and some Marcinkiewicz type variations of these results. However, here we concentrate on obtaining a possibly simple proof of weak type (1,1)(1,1) for Riesz transform type operators corresponding to quasielliptic operators.

    Doubling condition

    Before we state our main result we have to describe our basic assumptions. As it is usually the case in theory of spectral multipliers we require the doubling condition and some version of Gaussian estimates for semigroups generated by operators LjL_j, j=1,2j=1,2, see [@Ale2; @DOS]. We assume that the considered ambient spaces XjX_j j=1,2j=1,2 are equipped with a Borel measure μj\mu_j and distances ρj\rho_j. Let B(x,r)={y∈X ⁣: ρ(x,y)<r}B(x,r)= \{ y \in X \colon \, {\rho}(x,y) < r \} be the open ball centred at xx and radius rr. We suppose throughout that X1,X2{X_1},X_2 satisfy the doubling condition, that is there exists a constant CC such that \labeldoubling1.1μj(B(xj,2r))≤Cμj(B(xj,r))∀xj∈Xj,∀r>0,j=1,2.\label{doubling1.1} \mu_j(B(x_j, 2r)) \le C \mu_j(B(x_j, r)) \quad \forall x_j \in X_j, \forall r > 0, j=1,2. Note that ([doubling1.1]) implies that there exist positive constants CC and d1,d2d_1,d_2 such that \labeldμj(B(xj,tr))≤C(1+t)djμj(B(xj,r))∀t>0,xj∈Xj,r>0.\label{d} \mu_j(B(x_j, t r)) \le C (1+t)^{d_j} \mu_j(B(x_j, r)) \quad \forall t > 0, x_j \in X_j, r > 0. In the sequel we always assume that ([d]) holds. Note that all fractal spaces which we discuss here satisfy condition ([d]). In fact for these spaces μ(B(x,r))∼rd\mu(B(x, r)) \sim r^d for all r≤1r\le 1, see [@Ki; @Str2].

    Kernels of operators

    Suppose that TT is a bounded operator on L2(X).L^2(X). We say that a measurable function KT ⁣:X2→CK_T\colon X^2 \to {\mathbf{C}} is the (singular) kernel of TT if \labelabcde⟨Tf1,f2⟩=∫XTf1f2‾dμ=∫XKT(x,y)f1(y)f2(x)‾dμ(x)dμ(y).\label{abcde} \langle T f_1,f_2\rangle = \int_X T f_1\overline{f_2} d\mu = \int_X K_{T}(x,y) f_1(y)\overline{f_2(x)} d\mu(x) d\mu(y). for all f1,f2∈Cc(X)f_1,f_2\in C_c(X) (for all f1,f2∈Cc(X)f_1,f_2\in C_c(X) such that supp⁡f1∩supp⁡f2=∅{\operatorname{supp}}f_1 \cap {\operatorname{supp}}f_2 = \emptyset respectively). It is well known that if TT is bounded from L1(X)L^1(X) to Lq(X)L^q(X), where 1<q1<q, then TT is a kernel operator, and ∥T∥L1→Lq=sup⁡y∈X∥KT(⋅,y)∥Lq .\|T\|_{L^1 \to L^q} = \sup_{y\in X} \| K_{T}(\cdot , y)\|_{L^{q}}\, . In addition, if SS is continuous on Lq(X)L^q(X) then for almost all y∈Xy\in X \labelker1KST,y=SKT,y\label{ker1} K_{ST,y}=SK_{T,y} where KT,y(x)=KT(x,y)K_{T,y}(x)=K_T(x,y). Next we denote the weak type (1,1)(1,1) norm of an operator TT on a measure space (X,μ)(X,\mu) by ∥T∥L1→L1,∞=sup⁡λ μ({x∈X: ∣Tf(x)∣>λ})\|T\|_{L^1 \to L^{1,\infty}} =\sup \lambda \ \mu ( \{x\in X : \ |Tf(x)| > \lambda \}), where the supremum is taken over λ>0\lambda > 0 and functions ff with L1L^1 norm less than one; this is often called the “operator norm”, though in fact it is not a norm.

    In the sequel we will always require the following Gaussian estimates for the heat kernel corresponding to the operators L1L_1 and L2L_2.

    [1b] Let LjL_j, j=1,2j=1,2 be self-adjoint positive definite operators. We assume that the semigroups generated by LjL_j on L2(Xj)L^2(X_j) have the kernels pt[j](xj,yj)=Kexp⁡(−tLj)(xj,yj)p^{[j]}_{t}(x_j,y_j)=K_{\exp(-tL_j)}(x_j,y_j) defined by [([abcde])]{} which for some constants Cj,bj>0C_j,b_j>0 and m>1{m}>1 satisfy the following Gaussian upper bounds \label1.3kernel∣pt[j](xj,yj)∣≤Cjμ(B(yj,t1/m))−1exp⁡(−bjρj(xj,yj)m/(m−1)t1/(m−1))\label{1.3kernel} \vert p^{[j]}_{t}(x_j,y_j) \vert \le C_j\mu(B(y_j, t^{1/m}))^{-1} \exp \Big(-b_j\frac{{\rho}_j(x_j,y_j)^{m/(m-1)}}{t^{1/(m-1)}} \Big) for all t>0t>0 and j∈{1,2}j\in \{1,2\}.

    We will call pt[j](x,y)p^{[j]}_{t}(x,y) the heat kernels associated with LjL_j. Such estimates are typical for elliptic or sub-elliptic differential operators of order mm (see e.g. [@Da2]). But such estimates hold also for most of the Laplace type operators acting on fractals, see [@Ki; @Str2].

    Note that Assumption [1b] implies that the operator Kexp⁡(sL1+tL2)K_{\exp(sL_1+tL_2)} has a L∞L^\infty kernel given by the following formula Kexp⁡(sL1+tL2)((x1,x2),(y1,y2))=ps[1](x1,y1)pt[2](x2,y2)K_{\exp(sL_1+tL_2)}((x_1,x_2),(y_1,y_2))=p^{[1]}_{s}(x_1,y_1)p^{[2]}_t(x_2,y_2) for all s,t>0s,t>0.

    Main result

    We define a family of dilations {δt}t>0\{\delta_t\}_{t>0} acting on functions F ⁣:R2→CF\colon {\mathbf{R}}^2 \to C by the formula δtF(λ1,λ2)=F(tλ1,tλ2).\delta_{t}F(\lambda_1,\lambda_2) = F(t\lambda_1,t\lambda_2). Next let us recall that the norms in Sobolev spaces Wsp(Rn)W_{s}^{p}({\mathbf{R}}^n) are defined by the formula ∥F∥Wsp(Rn)=∥(Δ+I)s/2F∥Lp(Rn),\|F\|_{W_{s}^{p}({\mathbf{R}}^n)}=\|(\Delta+I)^{s/2}F\|_{L^p({\mathbf{R}}^n)}, where Δ\Delta is the standard Laplace operator. Now we can formulate our main spectral multiplier result.

    [main] Suppose that spaces XjX_j satisfy doubling condition [([d])]{} with constants d1d_1 and d2d_2. Further assume that operators LjL_j satisfy Assumption [[1b]]{} with the same order mm. Let F ⁣:R2→CF\colon {\mathbf{R}}^2 \to {\mathbf{C}} be a continuous function and let η∈Cc∞(0,∞)\eta \in C_c^\infty(0,\infty) be an auxiliary nonzero cut-off function. Suppose that for some s>d1+d22s>\frac{d_1+d_2}{2} \labelhcsup⁡t>0∥η~ δtF∥Ws∞<∞,\label{hc} \sup_{t> 0} \Vert \widetilde{\eta} \, \delta_{t }F \Vert_{W^\infty_s}<\infty, where η~(x,y)=η(x+y)\widetilde{\eta}(x,y)=\eta(x+y). Then the operator F(L1,L2)F(L_1,L_2) is of weak type (1,1)(1,1) and is bounded on Lq(X)L^q(X), X=X1×X2X=X_1\times X_2 for all 1<q<∞1<q<\infty. In addition ∥F(L1,L2)∥L1→L1,∞≤Cssup⁡t>0∥η~ δtF∥Ws∞(R2).\|F({L_1,L_2})\|_{L^1 \to L^{1,\infty}} \le C_s \sup_{t > 0} \Vert \widetilde{\eta} \, \delta_t F \Vert_{W^\infty_s({\mathbf{R}}^2)}.

    [Remarks]{} 1. We assume that L1L_1 and L2L_2 are positive so F(L1,L2)F(L_1,L_2) depends only on the restriction of FF to [0,∞)2[0,\infty)^2. However, it is easier to state Theorem [main] if one considers functions F ⁣:R2→CF\colon {\mathbf{R}}^2 \to {\mathbf{C}}.

    2. The condition on function FF in Theorem [main] is similar to the condition in Hörmdander-Mikhlin Fourier multiplier result, see [@Her1; @Mi] and [@Her2 Theorem 7.9.6]. The only difference is that we use the space Ws∞W^\infty_s instead of Ws2W^2_s. The significance of this difference is discussed in [@DOS].

    3. It is not difficult to check that if for all ∣I∣≤[d1+d22]+1|I| \le \left[\frac{d_1+d_2}{2}\right]+1, where [a][a] is an integer part of aa, sup⁡λ∈R2∣λ∣∣I∣∣∂IF(λ)∣<∞,\sup_{\lambda\in {\mathbf{R}}^2}|\lambda|^{|I|}|\partial^I F(\lambda)| < \infty, then FF satisfies condition ([hc]), see [@Ale1; @Ale2; @LM]. In a sense condition ([hc]) is a fractional exponent version of the above condition in which ∣I∣|I| must be an integer. The above condition is an illuminating illustration of condition ([hc]). In fact condition ([hc]) can be stated equivalently using interpolation between integer cases of the above definition.

    Note that η\eta is only an auxiliary function and that condition ([hc]) does not depend of η\eta.

    4. It would be interesting to obtain a version of Theorem [main] with condition ([hc]) replaced by sup⁡λ∈R2∣λI∂IF(λ)∣<∞,\sup_{\lambda\in {\mathbf{R}}^2}|\lambda^I\partial^I F(\lambda)| < \infty, for all ∣I∣≤l|I|\le l for some sufficiently large ll, see [@MRS Theorem 1.3].

    5. It could be also interesting to try to obtain some multivariable spectral multipliers results similar to Theorem [main] using the techniques developed in [@DR; @LLM].

    Notation

    In subsequent sections we use the following notation. We put XP=X1×X2X_P=X_1\times X_2, xP=(x1,x2)∈XPx_P=(x_1,x_2)\in X_P and yP=(y1,y2)y_P=(y_1,y_2). Next μP=μ1×μ2\mu_P=\mu_1\times \mu_2 and we set ρP(xP,yP)=max⁡{(ρ1(x1,y1),ρ2(x2,y2)}\rho_P(x_P,y_P) =\max\{(\rho_1(x_1,y_1),\rho_2(x_2,y_2)\}. If the discussed results hold separately for both X1X_1 and X2X_2 spaces we just skip index i,ji,j and use X,x,y,μX,x,y,\mu etc.

    Proof of Theorem [main]

    We split the proof of Theorem [main] into a few lemmas. First we show the following straightforward consequences of Assumption [1b]

    [ltu] Suppose that [([1.3kernel])]{} and [([d])]{} hold. Then for all r,t>0r,t>0 ∫X−B(y,r)∣pt(x,y)∣2dμ(x)≤Cμ(B(y,t1/m))−1exp⁡(−brm/(m−1)t1/(m−1)).\int_{X-B(y,r)}|p_{t}(x,y)|^2 d\mu(x) \le C \mu(B(y,t^{1/m}))^{-1}\exp \left(-b\frac{r^{{m}/{(m-1)}}}{ {t^{1/(m-1)}}}\right). In particular ∥pt(x, ⋅ )∥L2(X)2=∥pt( ⋅ ,x)∥L2(X)2≤Cμj(B(x,t1/m))−1.\|p_{t}(x,{\, \cdot \,})\|^2_{L^2(X)}= \|p_{t}({\, \cdot \,},x)\|^2_{L^2(X)} \le C \mu_j(B(x,t^{1/m}))^{-1}.

    By ([1.3kernel]) and ([d]) (see also [@CD Lemma 2.1]) ∫X−B(y,r) ⁣ ⁣∣pt(x,y)∣2dμ(x)≤Cμ(B(y,t1/m))2 ⁣ ⁣∫X−B(y,r) ⁣ ⁣exp⁡(−2bρ(x,y)m/tm−1)dμ(x)≤Cexp⁡(−brm/tm−1)μ(B(y,t1/m))2∫Xexp⁡(−bρ(x,y)m/tm−1)dμ(x)≤Cμ(B(y,t1/m))exp⁡(−brm/tm−1).\begin{aligned} \int\limits_{X-B(y,r)}\!\!|p_{t}(x,y)|^2d\mu(x) \le \frac{C}{\mu(B(y,t^{1/m}))^2} \!\!\int\limits_{X-B(y,r)}\!\!\exp\left(-2b\sqrt[m-1]{\rho(x,y)^m/t}\right) d\mu(x)\\ \le \frac{C\exp(-b\sqrt[m-1]{r^m/t})}{\mu(B(y,t^{1/m}))^2} \int_{X}\exp\big(-b\sqrt[m-1]{\rho(x,y)^{m}/t}\big) d\mu(x)\\ \le \frac{C\mu(B(y,t^{1/m}))}{\exp(-b\sqrt[m-1]{r^{m}/t})}.\end{aligned}

    Second we prove the following lemma

    [P] Suppose that ∥pt[j](xj, ⋅ )∥L2(Xj,μj)2≤Cμj(B(xj,t1/m))−1\|p^{[j]}_{t}(x_j, {\, \cdot \,})\|^2_{L^2(X_j,\mu_j)} \le C \mu_j(B(x_j,t^{{1}/{m}}))^{-1}. Then ∥KF(L1,L2)(xP, ⋅ )∥L2(XP)2=∥KF‾(L1,L2)( ⋅ ,xP)∥L2(XP)2≤C∥F∥L∞2∏j=1,2μj(B(xj,R−1))−1\begin{aligned} \Vert K_{{F}(L_1,L_2)} (x_P,{\, \cdot \,}) \Vert^2_{L^2(X_P)}= \Vert K_{\overline{F}(L_1,L_2)} ({\, \cdot \,},x_P) \Vert^2_{L^2(X_P)} \nonumber \\ \le C \Vert F \Vert_{L^\infty}^2 \prod_{j=1,2}\mu_j(B(x_j,R^{-1}))^{-1} \end{aligned} for any Borel function F ⁣:R2→CF \colon {\mathbf{R}}^2 \to {\mathbf{C}} such that supp⁡F⊂[0,Rm]×[0,Rm]{\operatorname{supp}}F\subset [0,{R}^{{m}}]\times [0,{R}^{{m}}].

    Set G1=FG2,\mboxwhereG2(λ1,λ2)=exp⁡(−R−m(λ1+λ2)).G_1=\frac{F}{G_2}, \quad \mbox{\rm where} \quad G_2(\lambda_1,\lambda_2)=\exp(-{R}^{-{m}}(\lambda_1+\lambda_2)). Then ∥G1(L1,L2)∥L2(XP)→L2(XP)≤∥G1∥L∞≤e∥F∥L∞.\Vert G_1(L_1,L_2) \Vert_{L^2(X_P) \to L^2(X_P)} \le \|G_1\|_{L^\infty}\le e\|F\|_{L^\infty}. By ([ker1]) the operator F(L1,L2)F(L_1,L_2) has the kernel given by the formula KF(L1,L2)(xP,yP)=[G1(L1,L2)KG2(L1,L2)( ⋅ ,yP)](xP).K_{F(L_1,L_2)}(x_P,y_P) = \left[{G}_1(L_1,L_2)K_{G_2(L_1,L_2)}({\, \cdot \,},y_P)\right](x_P). Now ∫XP∣KF(L1,L2)(xP,yP)∣2dμP(yP)\labelGF≤∥G1(L1,L2)∥L2→L22∏j=1,2∥pR−m[j]( ⋅ ,xj)∥L2(Xj)2≤C∥G1∥L∞2∏j=1,2μj(B(xj,R−1))−1≤C∥F∥L∞2∏j=1,2μj(B(xj,R−1))−1.\begin{aligned} &&\int_{X_P} \vert K_{F(L_1,L_2)}(x_P,y_P) \vert^{2} d \mu_P(y_P)\label{GF} \\ &&\le \Vert {G}_1(L_1,L_2) \Vert_{L^2 \to L^2}^2 \prod_{j=1,2}\Vert p^{[j]}_{{R}^{-m}}({\, \cdot \,}, x_j ) \Vert_{L^2(X_j)}^2\nonumber \\ &&\le C \Vert G_1\Vert^2_{L^\infty} \prod_{j=1,2}\mu_j(B(x_j,{R}^{-1}))^{-1} \le C \Vert F \Vert^2_{L^\infty}\prod_{j=1,2}\mu_j(B(x_j,{R}^{-1}))^{-1} .\nonumber\end{aligned}

    Third we show that (compare [@DOS; @Ou])

    [jeden] For any s1,s2≥0s_1,s_2 \ge 0 there exists a constant CC such that \labelsss∫X1×X2∏j=1,2(∣p(1+iτj)R−m[j](xj,yj)∣2ρj(xj,yj)sj)dμ1×μ2(x1,x2)≤C∏j=1,2(μj(B(yj,R−1))−1R−sj(1+∣τj∣)sj)\begin{aligned} \label{sss} \int_{X_1\times X_2} \prod_{j=1,2}\Big(|p^{[j]}_{ (1+i\tau_j){R}^{-m}}(x_j,y_j)|^2\rho_j(x_j,y_j)^{s_j}\Big) d \mu_1\times\mu_2(x_1,x_2) \nonumber \\ \le C \prod_{j=1,2}\Big(\mu_j(B(y_j,{R}^{-1}))^{-1} {R}^{-s_j}(1+|\tau_j|)^{s_j}\Big) \end{aligned} for all τj∈R\tau_j\in {\mathbf{R}} and R>0{R}>0.

    Note that the integral ([sss]) is a product of integrals and it is enough to show that the above estimates hold for j=1j=1 and j=2j=2 separately, that is ∫Xj∣p(1+iτj)R−m[j](xj,yj)∣2ρj(xj,yj)sjdμj(xj)≤Cμj(B(yj,R−1))−1R−sj(1+∣τj∣)sj\begin{aligned} \int_{X_j}| p^{[j]}_{ (1+i\tau_j){R}^{-{m}}}(x_j,y_j)|^2\rho_j(x_j,y_j)^{s_j} d \mu_j(x_j) \\ \le C \mu_j(B(y_j,{R}^{-1}))^{-1} {R}^{-s_j}(1+|\tau_j|)^{s_j}\end{aligned} holds for j=1j=1 and j=2j=2. The proof for j=1j=1 and j=2j=2 is the same so to simplify notation we skip the index jj. The rest of the proof follows closely the proof in [@DOS] and we describe it here for the sake of completeness. First, we assume that ∥f∥L2(X)=1\|f\|_{L^2(X)}=1 and that supp⁡f⊂X−B(y,r){\operatorname{supp}}f \subset X-B(y,r). Next, we define the holomorphic function Fy ⁣:{z∈C ⁣: ℜe z>0}→CF_y\colon \{z\in {\mathbf{C}}\colon\ \Re e\,z > 0\} \to {\mathbf{C}} by the formula Fy(z)=e−zRmμ(B(y,1/R))(∫Xpz(x,y)f(x)dμ(x))2.F_y(z)=e^{-zR^m}\mu(B(y,1/R)) \bigg(\int_X p_z(x,y)f(x) d\mu(x)\bigg)^2. By the same argument as in ([GF]) if we put z=∣z∣eiθz=|z|e^{i\theta}, then ℜe z=∣z∣cos⁡θ\Re e \, z =|z|\cos\theta and ∥pz( ⋅ ,y)∥L22=∥p∣z∣cos⁡θ( ⋅ ,y)∥L22\|p_{z}({\, \cdot \,},y)\|^2_{L^2}=\|p_{|z|\cos\theta}({\, \cdot \,},y)\|^2_{L^2}. Hence by Lemma [ltu] ∣Fy(z)∣≤e−Rm∣z∣cos⁡θμ(B(y,1/R))∥p∣z∣cos⁡θ( ⋅ ,y)∥L22≤Ce−Rm∣z∣cos⁡θμ(B(y,1/R))μ(B(y,∣z∣cos⁡θm))≤Ce−Rm∣z∣cos⁡θ(1+1Rm∣z∣cos⁡θ)d/m≤CR−d(∣z∣cos⁡θ)−d/m.\begin{aligned} |F_y(z)| & \le &e^{-R^m|z|\cos\theta}\mu(B(y,1/R))\|p_{|z|\cos\theta}({\, \cdot \,},y )\|^2_{L^2} \\ & \le & Ce^{-R^m|z|\cos\theta} \frac{\mu(B(y,1/R))}{\mu(B(y,\sqrt[m]{|z|\cos\theta}))} \le Ce^{-R^m|z|\cos\theta}\Big(1+\frac{1}{R^{m}|z|\cos\theta}\Big)^{d/m} \\ & \le & CR^{-d}(|z|\cos\theta)^{-d/m}.\end{aligned} Similarly for θ=0\theta=0 by Lemma [ltu] ∣Fy(∣z∣)∣≤CR−d∣z∣−d/mexp⁡(−brm/(m−1)∣z∣1/(m−1)).|F_{y}(|z|)|\le C R^{-d} |z|^{-d/m} \exp\bigg(-\frac{br^{m/(m-1)}}{|z|^{1/(m-1)}}\bigg). Now let us recall the following version of Phragmen-Lindelöf Theorem

    [dav] Suppose that function FF is analytic on the half-plane C+={z∈C ⁣:ℜe z>0}{\mathbf{C}}_+=\{ z\in {\mathbf{C}}\colon \Re e\, z > 0 \} and that ∣F(∣z∣eiθ)∣≤a1(∣z∣cos⁡θ)−β1\vert F (|z|e^{i\theta}) \vert \le a_1 (|z|\cos\theta)^{-{\beta_1}} ∣F(∣z∣)∣≤a1∣z∣−β1exp⁡(−a2∣z∣−β2)\vert F(|z|) \vert \le a_1 |z|^{-{\beta_1}}\exp(-a_2|z|^{-{\beta_2}}) for some a1,a2>0a_1, a_2 > 0, β1≥0{\beta_1} \ge 0, β2∈(0,1]{\beta_2}\in(0,1], all z∈C+z \in {\mathbf{C}}_+. Then ∣F(∣z∣eiθ)∣≤a12β1(∣z∣cos⁡θ)−β1exp⁡(−a2β22∣z∣−β2cos⁡θ)\vert F(|z|e^{i\theta}) \vert \le a_1 2^{\beta_1} (|z|\cos\theta)^{-{\beta_1}} \exp\Big(-\frac{a_2{\beta_2}}{2}|z|^{-{\beta_2}}\cos\theta\Big) for all z∈C+z \in {\mathbf{C}}_+.

    Now if ∣z∣eiθ=(1+iτ)R−m|z|e^{i\theta}=(1+i\tau)R^{-m}, then ∣z∣=R−m(1+∣τ∣2)1/2|z|=R^{-m}(1+|\tau|^2)^{1/2}, cos⁡θ=(1+∣τ∣2)−1/2\cos\theta=(1+|\tau|^2)^{-1/2} and ∣z∣cos⁡θ=R−m|z|\cos\theta = R^{-m}. Putting a1=CR−da_1=CR^{-d}, a2=brm/(m−1)a_2=br^{m/(m-1)}, β1=d/m{\beta_1}=d/m and β2=1/(m−1){\beta_2}=1/(m-1) in Lemma [dav] we conclude that ∣Fy((1+iτ)R−m)∣≤C′exp⁡(−b′(rR/(1+∣τ∣))m/(m−1)).\begin{aligned} |F_y((1+i\tau)R^{-m})| \le C' \exp{\Big(-b'(rR/(1+|\tau|))^{m/(m-1)}\Big)}.\end{aligned} Hence μ(B(y,1/R))∫X−B(y,r)∣p(1+iτ)R−m(x,y)∣2dμ(x)≤Cexp⁡(−b′(rR/(1+∣τ∣))m/(m−1)).\begin{aligned} \mu(B(y,1/R))\int_{X-B(y,r)} |p_{(1+i\tau)R^{-m}}(x,y)|^2d \mu(x) \\ \le C \exp{\Big(-b'(rR/(1+|\tau|))^{m/(m-1)}\Big)}.\end{aligned} Finally, we have ∫X∣p(1+iτ)R−m(x,y)∣2ρ(x,y)sdμ(x)=∑k≥0∫k(1+∣τ∣)R−1≤ρ(x,y)≤(k+1)(1+∣τ∣)R−1∣p(1+iτ)R−m(x,y)∣2ρ(x,y)sdμ(x)≤(1+∣τ∣)sR−s∑k≥0(k+1)s∫X−B(y,k(1+∣τ∣)R−1)∣p(1+iτ)R−m(x,y)∣2dμ(x)≤Cμ(B(y,1/R))−1R−s(1+∣τ∣)s.\begin{aligned} &&\hspace{-2cm}\int_X |p_{(1+i\tau)R^{-m}}(x,y)|^2{\rho}(x,y)^s d \mu(x)\\ &=& \sum_{k \ge 0} \int_{ k (1+|\tau|)R^{-1} \le {\rho}(x,y) \le (k+1) (1+|\tau|)R^{-1}} |p_{(1+i\tau)R^{-m}}(x,y)|^2{\rho}(x,y)^s d \mu(x)\\ &\le& (1+|\tau|)^s R^{-s}\sum_{k \ge 0} (k+1)^s \int_{ X - B(y,k (1+|\tau|)R^{-1}) } |p_{(1+i\tau)R^{-m}}(x,y)|^2 d \mu(x)\\ &\le& C\mu(B(y,1/R))^{-1}R^{-s}(1+|\tau|)^{s}. \end{aligned}

    Next we show that

    [waga] Suppose that L1L_1 and L2L_2 satisfy Assumption [([1b])]{}, R>0R>0 and s=s1+s2s=s_1+s_2. Then for any ε>0{\varepsilon}> 0 there exists a constant C=C(s,ε)C=C(s,{\varepsilon}) such that ∫X∣KF(L1,L2)((x1,x2),(y1,y2))∣2∏j=1,2(1+Rρj(xj,yj))sjdμ1×μ2(x1,x2)≤C∥δRmF∥Ws2+ε∞2∏j=1,2μj(B(yj,R−1))−1\begin{aligned} \int_{X} \vert K_{F(L_1,L_2)} ((x_1,x_2),(y_1,y_2))\vert^{2} \prod_{j=1,2}(1 + {R}\rho_j(x_j,y_j) )^{s_j} d\mu_1\times \mu_2(x_1,x_2) \nonumber\\ \le C \Vert \delta_{R^m} F \Vert_{W^\infty_{\frac{s}{2} + {\varepsilon}}}^{2} \prod_{j=1,2}\mu_j(B(y_j,R^{-1}))^{-1} \end{aligned} for all Borel functions FF such that supp⁡F⊆[0,Rm]2{\operatorname{supp}}F \subseteq [0, {R}^{{m}}]^2.

    Set G(λ1,λ2)=eλ1+λ2δRmF(λ1,λ2).G(\lambda_1,\lambda_2)= e^{\lambda_1+\lambda_2}\delta_{{R}^{m}} F(\lambda_1,\lambda_2). In virtue of the Fourier inversion formula F(L1,L2)=G(R−mL1,R−mL2)exp⁡(−R−m(L1+L2))=14π2∫R2exp⁡(∑j=1,2(iτj−1)R−mLj)G^(τ1,τ2)dτ1dτ2,\begin{aligned} &&F(L_1,L_2)=G(R^{-m}L_1, R^{-m}L_2)\exp{(-R^{-m}(L_1+L_2)) }\\&&\hspace{0.5cm}=\frac{1}{4\pi^2}\int_{{\mathbf{R}}^2} \exp{\left( \sum_{j=1,2}(i\tau_j-1)R^{-m}L_j \right)} \widehat{G}(\tau_1,\tau_2) d \tau_1 d \tau_2,\end{aligned} where G^\widehat{G} is the Fourier transform of the function GG. Hence KF(L1,L2)((x1,x2),(y1,y2))=14π2∫R2G^(τ1,τ2)∏j=1,2p(1−iτj)R−m[j](xj,yj)dτ1dτ2.\begin{aligned} &&K_{F(L_1,L_2)}((x_1,x_2),(y_1,y_2)) \\&&\hspace{0.5cm}= \frac{1}{4\pi^2} \int_{{\mathbf{R}}^2}\widehat{G}(\tau_1,\tau_2) \prod\limits_{j=1,2} p^{[j]}_{(1-i\tau_j){R}^{-m}}(x_j,y_j) d \tau_1 d \tau_2.\end{aligned} Thus by Lemma [jeden] and Lemma [ltu] (∫X1×X2∣KF(L1,L2)((x1,x2),(y1,y2))∣2×(1+Rρ1(x1,y1))s1(1+Rρ2(x2,y2))s2dμ1×μ2(x1,x2))12≤∫R2 ⁣∣G^(τ1,τ2)∣(∏j=1,2∫Xj ⁣∣p(1−iτj)R−m[j](xj,yj)∣2(1+Rρj(xj,yj))sjdμj(xj))12dτ1dτ2≤C(∏j=1,2μj(B(yj,R−m)−12)∫R2∣G^(τ1,τ2)∣(1+∣τ1∣)s1/2(1+∣τ2∣)s2/2dτ1dτ2≤C(∏j=1,2μj(B(yj,R−m)−12)(∫R2∣G^(τ1,τ2)∣2(1+τ12+τ22)s1+s2+ε+22)12×(∫R2(1+τ12+τ22)−2−ε2)12\labeldwa≤C∥G∥Ws1+s2+2+ε22∏j=1,2μj(B(yj,R−m)−12.\begin{aligned} &&\Bigg( \int_{X_1\times X_2} \vert K_{F(L_1,L_2)}((x_1,x_2),(y_1,y_2)) \vert^{2} \nonumber \\ &&\hspace{2cm} \times(1 + R{\rho}_1(x_1,y_1) )^{s_1}(1 + R{\rho}_2(x_2,y_2) )^{s_2} d\mu_1\times \mu_2(x_1,x_2) \Bigg)^{\frac{1}{2}} \nonumber\\ &&\le\int\limits_{{\mathbf{R}}^2} \! \vert \widehat{G}(\tau_1,\tau_2)\vert \Big(\prod_{j=1,2}\int\limits_{X_j} \! |p^{[j]}_{(1-i \tau_j)R^{-m}}(x_j,y_j)|^2 (1+R{\rho}_j(x_j,y_j))^{s_j} d\mu_j(x_j)\Big)^{\frac{1}{2}} d \tau_1 d \tau_2 \nonumber\\ &&\le C\Big( \prod_{j=1,2}\mu_j(B(y_j,R^{-m})^{-\frac{1}{2}}\Big) \int_{{\mathbf{R}}^2} \vert \widehat{G}(\tau_1,\tau_2) \vert (1+|\tau_1|)^{s_1/2}(1+|\tau_2|)^{s_2/2} d \tau_1 d \tau_2 \nonumber\\ &&\hspace{1cm} \le C \Big(\prod_{j=1,2}\mu_j(B(y_j,R^{-m})^{-\frac{1}{2}}\Big)\left(\int_{{\mathbf{R}}^2} \vert \widehat{G}(\tau_1,\tau_2) \vert^{2} ({1+\tau_1^2+\tau_2^{2}})^{\frac{s_1+s_2+{\varepsilon}+2}{2}}\right)^{\frac{1}{2}} \nonumber \\ &&\hspace{2cm} \times \left(\int_{{\mathbf{R}}^2}({1+\tau_1^2+\tau_2^{2}})^{\frac{-2-{\varepsilon}}{2}}\right)^{\frac{1}{2}} \label{dwa}\\ && \hspace{4cm} \le C \|G\|_{W^2_{\frac{s_1+s_2+2+{\varepsilon}}{2}}}\prod_{j=1,2}\mu_j(B(y_j,R^{-{m}})^{-\frac{1}{2}}. \nonumber\end{aligned} However, supp⁡F⊆[0,Rm]2{\operatorname{supp}}F \subseteq [0, {R}^{{m}}]^2 and supp⁡δRmF⊆[0,1]2{\operatorname{supp}}\delta_{{R}^m}F \subseteq [0,1]^2 so \labeldwa2∥G∥W(s+2+ε)/22≤C∥δRmF∥W(s+2+ε)/22≤C∥δRmF∥W(s+2+ε)/2∞.\label{dwa2} \|G\|_{W^2_{{(s+2+{\varepsilon})}/{2}}} \le C \|\delta_{R^{m}} F\|_{W^2_{(s+2+{\varepsilon})/{2}}} \le C \|\delta_{R^{m}} F\|_{W^\infty_{(s+2+{\varepsilon})/{2}}}. From the last two estimates one can obtain a multiplier result in which the required order of differentiability of the function F F is of 11 greater than that of Lemma [waga]. To get rid of this additional 11 we use Mauceri-Meda interpolation argument, see [@MM] and [@DOS]. First we note that the estimates from Lemma [waga] are equivalent to the following inequality ∫X∣KδR−mF(L1,L2)((x1,x2),(y1,y2))∣2∏j=1,2(1+Rρj(xj,yj))sjdμ1×μ2(x1,x2)≤C∥F∥Ws1+s22+ε∞2∏j=1,2μj(B(yj,R−1))−1\labeldww\begin{aligned} \int_{X} \vert K_{\delta_{R^{-m}}F(L_1,L_2)} ((x_1,x_2),(y_1,y_2))\vert^{2} \prod_{j=1,2}(1 + R\rho_j(x_j,y_j) )^{s_j} d\mu_1\times \mu_2(x_1,x_2) \nonumber \\ \le C \Vert F\Vert_{W^\infty_{ {\frac{s_1+s_2}{2}}+ {\varepsilon}}}^{2} \prod_{j=1,2} \mu_j(B(y_j,R^{-1}))^{-1} \label{dww}\end{aligned} for all bounded Borel functions FF such that supp⁡F⊂[0,1]2{\operatorname{supp}}F\subset [0,1]^2. Now we define the linear operator Ky1,y2,R ⁣:L∞([0,1]2)→L2(X1×X2,μ1×μ2)K_{y_1,y_2,{R}} \colon L^\infty([0,1]^2) \to L^2{(X_1\times X_2,\mu_1\times \mu_2)} by the formula Ky1,y2,R(F)=KδR−mF(L1,L2)( ⋅ ,(y1,y2)).K_{y_1,y_2,{R}}(F)=K_{\delta_{R^{-m}} F(L_1,L_2)}({\, \cdot \,},(y_1,y_2)). By Lemma [P] ∥Ky1,y2,R∥L∞([0,1]2)→L2(X1×X2,μ1×μ2)2≤C∏j=1,2μj(B(yj,R−1))−1.\Big\Vert K_{y_1,y_2,{R}}\Big\Vert_{L^\infty([0,1]^2) \to L^2(X_1 \times X_2 ,\mu_1 \times \mu_2)}^2\le C \prod_{j=1,2} \mu_j(B(y_j,R^{-1}))^{-1}. Next we put Ly1,y2,s1,s2,R2=L2(X1×X2,μy1,y2,s1,s2,R)L^2_{y_1,y_2,s_1,s_2,{R}}=L^2(X_1\times X_2,\mu_{y_1,y_2,s_1,s_2,{R}}), where dμy1,y2,s1,s2,R(x1,x2)=∏j=1,2(1+Rρj(xj,yj))sjdμ1×μ2(x1,x2)d\mu_{y_1,y_2,s_1,s_2,R}(x_1,x_2) =\prod_{j=1,2}(1+R\rho_j(x_j,y_j))^{s_j}d\mu_1\times \mu_2(x_1,x_2) By ([dwa]) and ([dwa2]) ∥Ky,R∥W(s1+s2+2+ε)/2∞([0,1]2)→Ly1,y2,s1,s2,R22≤C∏j=1,2μj(B(yj,R−1))−1.\Big\Vert K_{y,R}\Big\Vert_{W^\infty_{(s_1+s_2+2+{\varepsilon})/2}([0,1]^2) \to L^2_{y_1,y_2,s_1,s_2,R}}^2 \le C \prod_{j=1,2} \mu_j(B(y_j,R^{-1}))^{-1}. By interpolation, for every θ∈(0,1)\theta \in (0,1) there exists a constant CC such that ∥δRF(L1,L2)( ⋅ ,(y1,y2))∥Ly1,y2,θs1,θs2,R22≤C∥F∥[L∞,W(s1+s2+2+ε)/2∞][θ]2∏j=1,2μj(B(yj,R−1))−1.\begin{aligned} && \Vert \delta_{R}F(L_1,L_2)({\, \cdot \,},(y_1,y_2) ) \Vert_{L^{2}_{y_1,y_2,\theta s_1, \theta s_2,R}} ^2\\&& \le {C}\Vert F \Vert_{{[L^\infty, W^\infty_{(s_1+s_2+2+{\varepsilon})/2}]_{[\theta]}}}^2 \prod_{j=1,2} \mu_j(B(y_j,R^{-1}))^{-1}.\end{aligned} In particular, for all s>0s>0, θ∈(0,1) \theta \in (0, 1) and ε′>ε{\varepsilon}' > {\varepsilon} ∥δR−mF(L1,L2)( ⋅ ,(y1,y2))∥Ly1,y2,θs1,θs2,R22≤C∥F∥W(θs1+θs2+2θ+θε′)/2∞2∏j=1,2μj(B(yj,R−1))−1.\begin{aligned} &&\Vert \delta_{R^{-m}}F(L_1,L_2)({\, \cdot \,},(y_1,y_2) ) \Vert_{L^{2}_{y_1,y_2,\theta s_1, \theta s_2,R}}^2 \\&& \le {C}\Vert F \Vert_{{ W^\infty_{(\theta s_1+\theta s_2+2\theta +\theta {\varepsilon}')/2}}}^2 \prod_{j=1,2} \mu_j(B(y_j,R^{-1}))^{-1}.\end{aligned} Hence putting sj′=sj/θs'_j={s_j}/{\theta} in the above inequality and taking θ\theta small enough we obtain ∥δR−mF(L1,L2)( ⋅ ,(y1,y2))∥Ly1,y2,s1′,s2′,R22≤C∥F∥W(s1′+s2′)/2+ε"∞2∏j=1,2μj(B(yj,R−1))−1.\begin{aligned} &&\Vert \delta_{R^{-m}}F(L_1,L_2)({\, \cdot \,},(y_1,y_2) ) \Vert_{L^{2}_{y_1,y_2,s'_1, s'_2,R}}^2 \\ &&\le {C} \Vert F \Vert_{{ W^\infty_{(s'_1+s'_2)/2+{\varepsilon}"}}}^2\prod_{j=1,2} \mu_j(B(y_j,R^{-1}))^{-1}.\end{aligned} This proves ([dww]) and Lemma [P]

    Next we show the following simple consequence of doubling condition ([d]).

    [waga2] Suppose that [([d])]{} holds. Then for any s1>d1s_1 > d_1 and s2>d2s_2 > d_2 there exists ε>0{\varepsilon}>0 such that ∫XP−B(yP,r)∏j=1,2(1+Rρj(xj,yj))−sjdμ(x)≤C(1+rR)−ε∏j=1,2μ(B(yj,R−1))\int_{{X_P-B(y_P,r)}}\prod_{j=1,2}(1 + {{R}}\rho_j(x_j,y_j) )^{-s_j} d\mu(x) \le C {(1+rR)^{-{\varepsilon}}}\prod_{j=1,2} \mu(B(y_j,{R}^{-1})) where B(yP,r)={xP∈XP ⁣: ρP(xP,yP)<r}B(y_P,r)=\{x_P \in X_P \colon \, \rho_P(x_P,y_P) < r\}.

    Choose ε>0{\varepsilon}>0 such that sj′=sj−ε>djs'_j= s_j-{\varepsilon}>d_j and note that inf⁡XP−B(yP,r)∏j=1,2(1+Rρj(xj,yj))−ε≤(1+rR)−ε.\inf_{{X_P-B(y_P,r)}}\prod_{j=1,2}(1 + {R}\rho_j(x_j,y_j) )^{-{\varepsilon}}\le {(1+rR)^{-{\varepsilon}}}. Hence ∫XP−B(yP,r)∏j=1,2(1+Rρj(xj,yj))−sjdμP(xP)≤(1+rR)−ε∫XP∏j=1,2(1+Rρj(xj,yj))−sj′dμP(xP).\begin{aligned} \int_{{X_P-B(y_P,r)}}\prod_{j=1,2}(1 + {R}\rho_j(x_j,y_j) )^{-s_j} d\mu_P(x_P) \\ \le {(1+rR)^{-{\varepsilon}}} \int_{{X_P}} \prod_{j=1,2}(1 + {R}\rho_j(x_j,y_j) )^{-s'_j} d\mu_P(x_P).\end{aligned} Next ∫XP∏j=1,2(1+Rρj(xj,yj))−sj′dμP(xP)=∏j=1,2∫Xj(1+Rρj(xj,yj))−sj′dμj(xj)\int_{{X_P}} \prod_{j=1,2}(1 + {R}\rho_j(x_j,y_j) )^{-s'_j} d\mu_P(x_P) =\prod_{j=1,2} \int_{{X_j}} (1 + {R}\rho_j(x_j,y_j) )^{-s'_j} d\mu_j(x_j) so it is enough to show that for j=1j=1 and j=2j=2 one has ∫Xj(1+Rρj(xj,yj))−sj′dμj(xj)≤μj(B(yj,R−1)).\int_{{X_j}} (1 + {R}\rho_j(x_j,y_j) )^{-s'_j} d\mu_j(x_j) \le \mu_j(B(y_j,{R}^{-1})). As before we skip the index jj in the proof. \labelrr1∫X(1+Rρ(x,y))−sdμ(x)≤μ(B(y,R−1))+∑k≥0∫2k≤Rρ(x,y)≤2k+1(Rρ(x,y))−sdμ(x)≤μ(B(y,R−1))+∑k≥0(2k)−sμ(B(y,2k+1/R))≤C∑k≥0(2k)d−sμ(B(y,R−1))≤μ(B(y,R−1)).\begin{aligned} \label{rr1} \int\limits_{X} (1+ R{\rho}(x,y) )^{-s}d\mu(x) \le \mu(B(y, R^{-1})) + \sum_{k \ge 0} \int\limits_{2^k \le R{\rho}(x,y) \le 2^{k+1}} (R {\rho}(x,y))^{-s}d\mu(x) \\ \le \mu(B(y, R^{-1})) +\sum_{k \ge 0} (2^k)^{-s} \mu(B(y, 2^{k+1}/R)) \le C\sum_{k \ge 0} (2^k)^{d-s} \mu(B(y, R^{-1}))\\ \le \mu(B(y, R^{-1})).\end{aligned}

    To prove that operator is of weak type (1,1) we usually use estimates for the gradient of the kernel. The following theorem replaces the gradient estimates in the proof of Theorem [main].

    [CZ] Suppose that ∥F∥L∞≤C1\| F \|_{L^\infty} \le C_1 and that \labellap1sup⁡r∈R+sup⁡yP∈XP∫XP−B(yP,r)∣KF(1−Φr)(L1,L2)(xP,yP)∣dμP(xP)≤C1,\label{lap1} \sup_{r\in{\mathbf{R}}^+}\sup_{y_P\in X_P}\int_{ X_P-B(y_P,r)} |K_{F(1-\Phi_{r})(L_1,L_2)}(x_P,y_P)| d \mu_P(x_P) \leq C_1 , where Φr(L1,L2)=exp⁡(−rm(L1+L2))\Phi_{r}(L_1,L_2)=\exp(-r^m(L_1+L_2) ). Then ∥F(L1,L2)∥L1→L1,∞≤CC1.\| F(L_1,L_2)\|_{L^1\to L^{1,\infty}} \le C C_1.

    Theorem [CZ] follows from [@DM Theorem 2]. Indeed one can check that the family of operators exp⁡(−tLP)\exp(-tL_P), where LP=L1+L2L_P=L_1+L_2 satisfies all assumptions from [@DM]. Hence any operator TT which is bounded on L2L^2 and whose kernel KT(1−Φr)(L1,L2)K_{T(1-\Phi_{r})}(L_1,L_2) satisfies condition ([lap1]) is of weak type (1,1)(1,1). See also [@CD; @He1] and [@CoS] for similar results.

    [Remark]{}. In this note we define Calderón-Zygmund type operators in the sense of Theorem [CZ] and [@DM Theorem 2]. Note that this does not mean that the singular kernel of the operator TT is continuous outside the diagonal as it is the case for classical Calderón-Zygmund operators. For quasielliptic operators which we discuss below continuity of the kernel is a question of independent interest.

    Choose a function η\eta in Cc∞(0,1)C_c^\infty(0,1) such that ∑n∈Zη(2nmλ)=1∀λ∈R+.\sum_{n\in {\mathbf{Z}}} \eta(2^{nm}\lambda) = 1 \qquad\forall \lambda \in {\mathbf{R}}_+. Then F(1−Φr)(L1,L2)=∑n∈Zη~nF(1−Φr)(L1,L2),F(1-\Phi_{r})(L_1,L_2) =\sum_{n\in {\mathbf{Z}}} {\widetilde{\eta}}_n F(1-\Phi_{r})(L_1,L_2), where η~n=δ2−nmη~{\widetilde{\eta}}_n=\delta_{2^{-nm}}{\widetilde{\eta}}. Recall that η~(λ1,λ2)=η(λ1+λ2){\widetilde{\eta}}(\lambda_1,\lambda_2)=\eta(\lambda_1+\lambda_2). By Lemma [waga] and Lemma [waga2] for any sj>djs_j>{d_j}, j=1,2j=1,2 there exists ε,ε′>0{\varepsilon}, {\varepsilon}' >0 such that ∫XP−B(yP,r)∣Kη~nF(1−Φr)(L1,L2)(xP,yP)∣dμP(xP)≤(∫XP∣Kη~nF(1−Φr)(L1,L2)(xP,yP)∣2∏j=1,2(1+2nρ(xj,yj))sjdμP(xP))1/2×(∫XP−B(yP,r)∏j=1,2(1+2nρ(xj,yj))−sjdμP(xP))1/2≤C(1+2nr)−ε/2(∏j=1,2μj(B(yj,2−n))1/2)×(∏j=1,2μj(B(yj,2−n))−1/2)∥δ2nm[η~nF(1−Φr)]∥Ws1+s2+ε′2∞≤C(1+2nr)−ε/2∥δ2nm[η~nF(1−Φr)]∥Ws1+s2+ε′2∞.\begin{aligned} && \int_{X_P-B(y_P,r)} | K_{{\widetilde{\eta}}_nF(1-\Phi_{r})(L_1,L_2)}(x_P,y_P)|{d} \mu_P(x_P)\\ &\le& \Big(\int_{X_P} \vert K_{{\widetilde{\eta}}_n F(1-\Phi_{r})(L_1,L_2)} (x_P,y_P)\vert^{2}\prod_{j=1,2} (1 + 2^n{\rho}(x_j,y_j) )^{s_j} d\mu_P(x_P)\Big)^{1/2}\nonumber \\ &&\times\Big(\int_{{X_P-B(y_P,r)}}\prod_{j=1,2} (1 + 2^n{\rho}(x_j,y_j) )^{-s_j} d\mu_P(x_P)\Big)^{1/2} \nonumber \\ &\leq& C (1+2^{n}r)^{-{\varepsilon}/2} \Big( \prod_{j=1,2}\mu_j(B(y_j,{2}^{-n}))^{1/2}\Big) \nonumber \\ &&\times \Big( \prod_{j=1,2}\mu_j(B(y_j,{2}^{-n}))^{-1/2} \Big) \|\delta_{2^{nm}} [{\widetilde{\eta}}_n F(1-\Phi_{r})]\|_{W^\infty_{\frac{s_1+s_2+{\varepsilon}'}{2}}}\nonumber \\ &\leq& C (1+2^{n}r)^{-{\varepsilon}/2}\|\delta_{2^{nm}}[{\widetilde{\eta}}_n F(1-\Phi_{r})]\|_{W^\infty_{\frac{s_1+s_2+{\varepsilon}'}{2}}}.\nonumber\end{aligned} Now for any Sobolev space Wsp(R)W_s^{p}({\mathbf{R}}) and any integer k>sk>s ∥δ2nm[η~nF(1−Φr)]∥Ws∞≤C∥δ2nm[η~nF]∥Wsp∥δ2nm[1−Φr]∥Ck([1/4,1])≤C2nmrm1+2nmrm∥δ2nm[η~nF]∥Ws∞=C(r2n)m1+(r2n)m∥η~δ2nmF∥Ws∞.\begin{aligned} \|\delta_{2^{nm}}[{\widetilde{\eta}}_n F(1-\Phi_{r})]\|_{W^{\infty}_{s}} \leq C \|\delta_{2^{nm}}[{\widetilde{\eta}}_n F]\|_{W_s^{p}} \|\delta_{2^{nm}}[1-\Phi_{r}]\|_{C_k([1/4,1])}\\ \le \frac{C 2^{nm}r^m}{1+2^{nm}r^m} \|\delta_{2^{nm}} [{\widetilde{\eta}}_n F]\|_{W_{s}^{\infty}} =\frac{C (r2^n)^{m}}{1+(r2^n)^{m}} \|{\widetilde{\eta}}\delta_{2^{nm}} F\|_{W_s^{\infty}}.\end{aligned} Hence for s=s1+s2+ε′2s=\frac{s_1+s_2+{\varepsilon}'}{2} \labelkoniecsup⁡yP∈XP∫XP−B(yP,r)∣KF(1−Φr)(L1,L2)(xP,yP)∣dμP(xP)≤C∑n∈Z(r2n)m1+(r2n)m (1+2nr)−ε/2∥η~δ2nmF∥Wsp≤Csup⁡t>0∥η~δtF∥Wsp\begin{aligned} \label{koniec} &&\hspace{-1cm}\sup_{y_P\in X_P} \int_{X_P-B(y_P,r)} |K_{F(1-\Phi_{r})(L_1,L_2)}(x_P,y_P)| d \mu_P(x_P) \\ &\leq& C \sum_{n\in {\mathbf{Z}}} \frac{(r2^n)^{m}}{1+(r2^n)^{m}} \, (1+2^{n}r)^{-{\varepsilon}/2} \|{\widetilde{\eta}}\delta_{2^{nm}} F\|_{W_s^{p}} \leq C\sup_{t>0} \|{\widetilde{\eta}}\delta_{t} F\|_{W_s^{p}} \nonumber\end{aligned} as required to prove Theorem [main].

    Analysis on fractals and quasielliptic operators {#chfra}

    Many interesting examples of spaces and operators satisfying Assumption [1b] are described in the theory of Brownian Motion and Laplace like operators on fractals, see for example [@Ki; @Str2]. A compelling instance of such an operator is the Laplace operator on the Sierpiński gasket SGSG (Neumann or Dirichlet). Assumption [1b] and condition ([d]) hold with d=log⁡3/(log⁡5−log⁡3)=d =\log 3/ (\log 5 -\log 3)= and m=d+1=log⁡5/(log⁡5−log⁡3)m=d+1=\log5/ (\log5-\log3) , see [@BP; @Str1; @Str2]. We are especially interested in this example because the set of ratios of eigenvalues for the Laplacian on SGSG has gaps so it can be used to construct quasielliptic operators. Indeed the main results of [@BS Theorem 2.1] says that

    For any two eigenvalues λi,λj\lambda_i, \lambda_j of the Laplacian on SG λiλj∉(α,β)\frac{\lambda_i}{\lambda_j}\notin (\alpha,\beta) where α\alpha and β\beta are defined by [([gap])]{}.

    There are more examples of Laplacian on fractals such that the sets of ratios of their eigenvalues have gaps. For instance this is the case for hierarchical fractals introduced by Humbly HH(b)HH(b). For definitions and a precise formulation of results, see [@DS Theorems 5.2 and 5.3]. It also can be noticed that the discussed result that holds on SGSG also holds in the setting of [@Te]. It is easy to note that if (α,β)(\alpha,\beta) is a gap in the set of ratios of eigenvalues α<γ<β\alpha<\gamma < \beta, a,b∈Ra,b\in {\mathbf{R}}, c,d>0c,d>0 and dc=γ\frac{d}{c}=\gamma then the operators defined by ([qell]) are bounded on L2L^2. It seems likely that this short description of results concerning quasielliptic operators and gaps in the sets of ratios of eigenvalues will be soon outdated. Therefore, we state our result in an abstract way which could be applied to all quasielliptic operators.

    [one] Suppose that LL is a positive self-adjoint operator acting on L2(X)L^2(X). We say that the interval (α,β)(\alpha, \beta) is a gap in the set of ratios of eigenvalues of LL if λiλj∉(α,β)\frac{\lambda_i}{\lambda_j}\notin (\alpha,\beta) for all λi,λj\lambda_i,\lambda_j in the L2L^2 spectrum of the operator LL.

    Next we consider the Cartesian product of two copies of metric measure space XX and operators L1=L⊗1L_{1}=L \otimes 1 and L2=1⊗L L_2=1 \otimes L acting on L2(X2)L^2(X^2).

    [frac] Suppose that metric measure space (X,μ,ρ)(X,\mu, \rho) satisfies condition [([d])]{} with doubling constant dd and that operator LL satisfy Assumption [[1b]]{}. Next assume that (α,β)(\alpha, \beta) is a gap in the set of ratios of eigenvalues of LL and that \mboxsupp ω⊂{(λ1,λ2) ⁣: ∣λ1λ2−γ∣≤σ}\mbox{\rm supp}\, \omega \subset \{(\lambda_1,\lambda_2)\colon \, \left|\frac{\lambda_1}{\lambda_2}-\gamma \right| \le \sigma\} for some γ,σ\gamma, \sigma such that α<γ−σ<γ+σ<β\alpha < \gamma-\sigma<\gamma+\sigma < \beta. Assume in addition that for some s>ds>{d} sup⁡t>0∥η~ δt[(1−ω)F]∥Ws∞<∞,\sup_{t> 0} \Vert \widetilde{\eta} \, \delta_{t }\left[(1-\omega) F\right] \Vert_{W^\infty_s}<\infty, where η~ \widetilde{\eta} is defined in the same way as in Theorem [[main]]{}. Then the operator F(L1,L2)F(L_1,L_2) is of weak type (1,1)(1,1) and is bounded on Lq(X2)L^q(X^2) for all 1<q<∞1<q<\infty.

    It is not difficult to note that if (α,β)(\alpha, \beta) is a gap in the set of ratios of eigenvalues of LL then ωF(L1,L2)=0\omega F(L_1,L_2)=0 so F(L1,L2)=(1−ω)F(L1,L2)F(L_1,L_2)=(1-\omega) F(L_1,L_2) and Theorem [frac] follows from Theorem [main].

    Below we describe a straightforward consequence of Theorem [frac].

    [ko] Suppose that for some c,d>0c,d>0, cL1−dL2cL_1-dL_2 is a quasielliptic operator, that is α<dc<β\alpha< \frac{d}{c}< \beta, where (α,β)(\alpha, \beta) is a gap in the set of ratios of eigenvalues of LL described in Definition [[one]]{}. Then for all constants a,b∈Ca,b \in {\mathbf{C}} the operators aL1+bL2cL1−dL2\mboxandL1L2(cL1−dL2)2\frac{aL_1+bL_2}{cL_1-dL_2} \quad \mbox{and} \quad \frac{L_1L_2}{(cL_1-dL_2)^2} are bounded on all LpL^p spaces for 1<p<∞1<p<\infty and of a weak type (1,1)(1,1).

    Corollary [ko] follows from Theorem [frac] if one chooses function ω\omega such that ω(λ1,λ2)=1\omega(\lambda_1,\lambda_2)=1 for all λ1,λ2\lambda_1,\lambda_2 such that ∣λ1λ2−dc∣≤ε\left|\frac{\lambda_1}{\lambda_2}-\frac{d}{c}\right| \le {\varepsilon} for some small ε>0{\varepsilon}>0.

    If α\alpha and β\beta are defined by [([gap])]{} then the Neumann and Dirichlet Laplacian acting on Sierpiński gasket GSGS satisfy all assumption of Theorem [[frac]]{} and Corollary [[ko]]{}.

    [Acknowledgement: ]{} The author is indebted to Robert Strichartz for alerting him to the main research problem studied in the paper and hosting him at Cornell University.

    [10]{}

    G. Alexopoulos. Spectral multipliers on [L]{}ie groups of polynomial growth. , 120(3):973–979, 1994.

    G. Alexopoulos. Spectral multipliers for Markov chains. , 56(3):833–852, 2004.

    M. T. Barlow and E. A. Perkins. Brownian motion on the [S]{}ierpiński gasket. , 79(4):543–623, 1988.

    . , [56]{}(3):1361–1375, [2007.]{}

    Th. Coulhon and X. T. Duong. Riesz transforms for 1≤p≤21\leq p\leq 2. , 351(3):1151–1169, 1999.

    M. Christ and C. D. Sogge. The weak type L\sp1L\sp 1 convergence of eigenfunction expansions for pseudodifferential operators. [Invent. Math.]{} [94]{} (1988), no. 2, 421-453.

    M. Cowling and A. Sikora. A spectral multiplier theorem on SU(2){\rm SU}(2). 238 (2001), no. 1, 1–36.

    E. B. Davies. Uniformly elliptic operators with measurable coefficients. , 132(1):141–169, 1995.

    S. Drenning, R. S. Strichartz. Spectral Decimation on Hambly’s Homogeneous Hierarchical Gaskets. Preprint, 2007.

    X. T. Duong and A. McIntosh. Singular integral operators with non-smooth kernels on irregular domains. , 15(2):233–265, 1999.

    X. T. Duong, El-Maati Ouhabaz and A. Sikora. Plancherel-type estimates and sharp spectral multipliers. 196(2):443–485, 2002.

    X. T. Duong and D.W Robinson. Semigroup kernels, Poisson bounds, and holomorphic functional calculus. 142(1):443–485, 1996.

    W. Hebisch. A multiplier theorem for [S]{}chrödinger operators. , 60/61(2):659–664, 1990.

    L. H[ö]{}rmander. Estimates for translation invariant operators in L\spp{L}\sp{p}spaces. , 104:93–140, 1960.

    L. H[ö]{}rmander. . Springer-Verlag, Berlin, 1983.

    J. Kigami. Cambridge University Press, 2001.

    F. Lancien, G. Lancien and C. Le Merdy. A joint functional calculus for sectorial operators with commuting resolvents. , 77(2):387–414, 1998.

    N. Lohoué and M. Marias. Invariants géométriques des espaces localement symétriques et théorème de multiplicateurs. Preprint, 2007.

    G. Mauceri and S. Meda. Vector-valued multipliers on stratified groups. , 6(3-4):141–154, 1990.

    D. Muller, F. Ricci, Fulvio and E.M. Stein. , 2: 99–233, 1995.

    S. G. Mikhlin. Multidimensional singular integrals and integral equations. Translated from the Russian by W. J. A. Whyte. Pergamon Press, Oxford-New York-Paris 1965.

    El-Maati Ouhabaz. Gaussian estimates and holomorphy of semigroups. , 123(5):1465–1474, 1995.

    R. S. Strichartz. Function spaces on fractals. , 198(1): 43–83, 2003.

    R. S. Strichartz. Princeton University Press, Princeton, 2006

    A. Teplyaev. Spectral analysis on infinite Sierpiński gaskets. , 159(2):537–567, 1998.

  2. Assistant

    The paper discusses the analysis and properties of multivariable spectral multipliers applied to Cartesian products of self-adjoint operators, a valuable concept in harmonic analysis and operator theory. Let's break it down into sections that warrant either quick, straightforward responses or require more in-depth analysis.

    Initial Analysis and Recognition:

    [sys1] The paper addresses the problem of applying multivariable spectral multipliers to products of self-adjoint operators, focusing on conditions for these operators to maintain bounded properties across various spaces. The main result is extending the continuity of these operators on L2(X1×X2)L^2(X_1 \times X_2). [end sys1]

    In-Depth Analysis of Problem and Approach:

    [sys2]

    2.1 Understand the Problem:

    • Objective: Comprehend the issue of spectral multipliers in the context of two independent operators and product spaces.
    • Reflection: "I understand we aim to develop a theory where the operators L1L_1 and L2L_2 interact on a Cartesian product space. Am I overlooking needs for broader function class adaptability or edge cases like discrepancies in operator properties?"
    • Creative Perspective: "Could mapping fractal characteristics onto more conventional spaces reveal new uses or interpretations of these spectral multipliers?"

    2.2 Analyze the Information:

    • Objective: Break down the technical components, such as assumptions on Gaussian estimates and doubling conditions.
    • Reflection: "Are there conditions or factors in the eigenvalues of the operators that we’re assuming will hold? How uniform are these assumptions across different operator types?"
    • Creative Perspective: "What insights can the fractal nature or the quasielliptic properties of the operators provide that aren’t immediately obvious?"

    2.3 Generate Hypotheses:

    1. Fractal space operations maintain operator boundedness under certain spectral gaps. (Confidence: 0.8, Creative: 0.7)
    2. Doubling conditions ensure uniform operator behavior across LpL^p spaces. (Confidence: 0.7, Creative: 0.5)
    3. The uniqueness of spectral decomposition extends well to multivariable operators. (Confidence: 0.6, Creative: 0.6)
    4. Operator adaptability is driven by effective analysis of Gaussian estimates. (Confidence: 0.7, Creative: 0.6)
    5. Current exploration may reveal new facets of fractal dimensions impacting operator work. (Confidence: 0.6, Creative: 0.8)
    6. The presence of eigenvalue gaps can facilitate specific types of operator constructions such as quasielliptic ones. (Confidence: 0.8, Creative: 0.8)
    7. Spectral multiplier theorems can adapt to broader contexts beyond current harmonic analysis applications. (Confidence: 0.5, Creative: 0.7)
    8. Spectral theory applications present untapped potential in non-commutative analysis. (Confidence: 0.4, Creative: 0.9)
    9. Diverse operator designs could lead to novel computational models. (Confidence: 0.5, Creative: 0.6)
    10. Employing heat kernel estimates might solve broader operator theory problems beyond the current scope. (Confidence: 0.7, Creative: 0.7)

    Reflection: "Have I considered the practical implications of each hypothesis in real-world analysis? Have I explored both mathematical elegance and computational feasibility?" Creative Perspective: "Are there potentially groundbreaking insights in how we understand and employ these spectral multipliers across fields?"

    2.4 Anticipate Future Steps and Obstacles:

    • Objective: Anticipate challenges like adapting theories for broader spectra or handling anomalies in eigenvalue distributions.
    • Reflection: "How might emerging computational techniques influence these mathematical theories?"
    • Creative Perspective: "Could leveraging AI provide new angles for analyzing spectral operator behavior?"

    [end sys2]

    Summarized Findings:

    [sys1] The exploration into multivariable spectral multipliers highlights potential in leveraging eigenvalue gaps in fractals to develop bounded operator strategies. Core strengths lie in harnessing the properties of spectral multipliers and conditions like Gaussian bounds and the doubling condition. [end sys1]