Diagonal Wavefront Strategy in Microlocal Analysis

• Diagonal wavefront strategy is an analytic method in microlocal analysis that uses diagonal dilation groups and anisotropic wavelets to localize singularities.
• It partitions frequency space into anisotropically oriented wedges, allowing precise detection of singularity locations through rapid decay of wavelet coefficients.
• The approach leverages microlocal admissibility and weak cone-approximation properties to establish a multiple-wavelet criterion that characterizes the wavefront set in any dimension.

The diagonal wavefront strategy is an analytic methodology for microlocal analysis that characterizes the wavefront set $\operatorname{WF}(u)$ of tempered distributions $u\in\mathcal{S}'(\mathbb{R}^d)$ by means of anisotropic continuous wavelet transforms parameterized by diagonal dilation groups. By employing wavelets with Fourier support concentrated in shrinking, anisotropically oriented frequency wedges and utilizing group-theoretic properties of the diagonal group $H_{\mathrm{diag}}$, this approach enables an explicit, multi-directional and multi-scale localization of singularities in phase space. The diagonal wavefront strategy yields a precise, multiple-wavelet criterion for wavefront set membership in arbitrary dimension $d$ based on rapid decay of wavelet coefficients in appropriate geometric configurations (Fell et al., 2014).

1. Diagonal Dilation Groups and Dual Action

The core of the diagonal wavefront strategy is based on the diagonal dilation group

$$H_{\mathrm{diag}} = \left\{\,\mathrm{diag}(a_1, \dots, a_d)\ :\ a_i\in\mathbb{R}\setminus\{0\} \right\}\subset\mathrm{GL}(\mathbb{R}^d).$$

A generic group element is $h = \mathrm{diag}(a_1,\dots,a_d)$ with inverse $h^{-1} = \mathrm{diag}(a_1^{-1},\dots,a_d^{-1})$. The group’s dual action on frequency space is the right action

$$(\xi, h) \mapsto h^{T}\xi = (a_1\xi_1, a_2\xi_2, \dots, a_d\xi_d)^{T},$$

which serves to anisotropically scale different coordinate frequencies.

2. Frequency Localization: Cones and Patches

Microlocal information is accessible via frequency-localized structures. For a compactly supported wavelet $\psi$ with Fourier support within a set $V \Subset \mathbb{R}^d \setminus \{0\}$, and given a spherical cap $W\subset S^{d-1}$ as well as $R>0$, one defines the frequency cone

$$C(W,R) = \{\xi\in\mathbb{R}^d : |\xi|>R,\ \xi/|\xi|\in W\}.$$

Associated to this geometric partition, two cone-affiliated subsets of $H_{\mathrm{diag}}$ are defined: $$\begin{aligned} K_i(W, V, R) &= \left\{h\in H_{\mathrm{diag}}: h^{-T}V\subset C(W,R)\right\}, \ K_o(W, V, R) &= \left\{h\in H_{\mathrm{diag}}: h^{-T}V\cap C(W,R)\neq \emptyset\right\}. \end{aligned}$$ This constructs the analytic foundation for resolving phase-space localization by group action.

3. Microlocal Admissibility and Cone Approximation

For a geometric and analytic identification of microlocal singularities, two properties are critical:

Microlocal admissibility requires a systematic correlation between dilation scales and operator norms. For the diagonal group, operator norms satisfy

$$\|h\| = \max_i |a_i|,\quad \|h^{-1}\| = \max_i |a_i|^{-1} = \|h\|^{-1}.$$

Exactly, microlocal admissibility in direction $\xi$ consists of

$\|h^{-1}\| \leq C\|h\|^{-\alpha_1}$

on $K_o(W_0, V, R_0)$, for suitable neighborhoods $W_0\ni\xi$, some $\alpha_1=1$, and bounded measure growth, which all hold for $H_{\mathrm{diag}}$.

Weak cone-approximation property: Given any cone $C(W,R)$, there exist a refined cone $C(W',R')$ and frequency patch $V_n\subset V$ such that

$K_o(W', V_n, R') \subset K_i(W, V_n, R).$

For $H_{\mathrm{diag}}$, define

$$V_n = \left\{\xi:\ \frac{n}{n+1}<|\xi|<\frac{n+1}{n},\ \left|\frac{\xi}{|\xi|} - e_1 \right| < \frac{1}{n} \right\}$$

and the cone-approximation can be verified via estimates on the action of diagonal entries $a_i$. The diagonal group does not satisfy the strong (single-patch) cone-approximation property due to its structure, admitting only the weak version; thus, multiple wavelets are necessary.

4. Continuous Wavelet Transforms and Decay Criterion

For the $\mathcal{S}(\mathbb{R}^d)$ wavelet $\psi$ and tempered distribution $u$, the representation

$$[\pi(x,h)\psi](y) = |\det h|^{-1/2}\,\psi(h^{-1}(y-x))$$

induces the continuous wavelet transform,

$W_{\psi}u(x,h) = \langle u, \pi(x,h)\psi \rangle.$

By Lemma 3.1 (Fell et al., 2014),

$$\widehat{W_{\psi}(\varphi u)(\cdot,h)}(\xi) = \widehat{\varphi u}(\xi)\,|\det h|^{1/2}\,\overline{\widehat{\psi}(h^T\xi)}$$

for any compactly supported $\varphi u$, so decay of $W_{\psi}u(y,h)$ in $\|h\|$ on cone sets is tightly linked to the smoothness or singularity of $u$ in corresponding phase-space regions.

5. Wavefront Set Characterization: The Diagonal Strategy

The multiple-wavelet criterion for the wavefront set is as follows: For each direction $\xi$ and point $x$, choose wedges $V_n$ as above and corresponding admissible wavelets $\psi_n$ with frequency support in $V_n$.

For any $u\in\mathcal{S}'(\mathbb{R}^d)$, $(x,\xi)\notin\operatorname{WF}(u)$ if and only if there exist neighborhood $U\ni x$, spherical cap $W\ni\xi$, and integer $n_0$ such that for all $n\geq n_0$, all $y\in U$, and all $h\in K_o(W, V_n, R)$,

$$\forall N\;\exists C_N: \quad |W_{\psi_n}u(y,h)| \leq C_N \|h\|^{-N}.$$

This super-polynomial decay in scale for appropriately directed, frequency-localized wavelets directly characterizes wavefront regularity [(Fell et al., 2014), Thm. 4.1 and 6.3]. The proof hinges on the reconstructive and localization capacity of the patches, together with admissibility and approximation properties.

6. Explicit Analysis in Dimension Two

In $d=2$, the group $$H_{\mathrm{diag}} = \{\mathrm{diag}(a,b) : a,b\neq 0\}$$ acts on $\mathbb{R}^2$; wedges $V_n$ can be chosen: $$V_n = \{(r\cos\theta, r\sin\theta) : \frac{n}{n+1}<r<\frac{n+1}{n},\, |\theta|<1/n\}$$ centered near $(1,0)$. Here, $\|h\| = \max\{|a|,|b|\}$, and all microlocal admissibility and weak cone-approximation properties are verified via straightforward coordinate calculations. By permuting coordinates, all directions are accessed, enabling directional singularities throughout $\mathbb{R}^2$ to be analyzed by this approach.

7. Summary and Theoretical Implications

The diagonal wavefront strategy proceeds by:

• Using $H_{\mathrm{diag}}$ for anisotropic continuous wavelet transforms,
• Decomposing frequency space into shrinking wedges $V_n$ around each direction,
• Choosing admissible wavelets $\psi_n$ supported in each $V_n$,
• Verifying microlocal admissibility and weak cone-approximation properties,
• Establishing that $(x,\xi)\notin\operatorname{WF}(u)$ if and only if $W_{\psi_n}u(y,h)$ decays super-polynomially in scale for all large $n$, $y$ near $x$, and $h$ mapping $V_n$ into a fixed cone near $\xi$.

This yields an explicit, multiple-wavelet class of wavefront set characterizations for arbitrary dimension $d$, underpinned by group-theoretic and microlocal structures, and is specialized for diagonal dilations distinct from similitude and shearlet approaches (Fell et al., 2014).