Unimodular Fourier Multipliers

• Unimodular Fourier multipliers are operators defined via a real phase function that produces modulus-one symbols, ensuring L2 norm preservation.
• They leverage stationary phase and time-frequency decomposition methods to analyze Lp norm growth and sharp boundedness thresholds.
• Their theoretical framework extends to noncommutative and quantum settings, linking classical dispersive PDE analysis with operator algebra techniques.

A unimodular Fourier multiplier is a Fourier multiplier operator of the form $f \mapsto \mathcal{F}^{-1}(e^{i\Phi}\widehat{f})$, where $\Phi$ is a real-valued phase function, so that the multiplier symbol $e^{i\Phi(\xi)}$ has modulus one for all $\xi$. Such operators are canonical in harmonic analysis due to their connections to dispersive equations, transference principles, representation theory of locally compact groups, and sharp multiplier theory in various function spaces. The theory of unimodular Fourier multipliers encompasses phenomena ranging from norm growth, boundedness and optimality in $L^p$ and noncommutative settings, to deeper connections between operator algebras, geometric group theory, and harmonic analysis.

1. Fundamental Definition and Model Examples

A unimodular Fourier multiplier associated with the symbol $e^{i\Phi}$ acts on Schwartz functions $f$ by: $$\widehat{T_\Phi f}(\xi) = e^{i\Phi(\xi)}\,\widehat{f}(\xi)$$ or, equivalently,

$$T_\Phi f(x) = \int_{\mathbb{R}^d} e^{i(\Phi(\xi)+x\cdot\xi)}\,\widehat{f}(\xi)\,d\xi.$$

For parameterized families, e.g., $T^t_\Phi$ with symbol $e^{it\Phi(\xi)}$, the parameter $t\in\mathbb{R}$ scales the oscillatory nature.

• If $\Phi(\xi)=|\xi|^2$, $T^t_\Phi$ is the free Schrödinger evolution.
• For $\Phi(\xi)=|\xi|$, $T^t_\Phi$ realizes the wave propagator.
• General $\Phi$ may be homogeneous, smooth, or possess weaker regularity depending on context (Bulj, 2023, Nicola et al., 2018).

Unimodular multipliers are $L^2$-unitary: they preserve $L^2$ norms due to the modulus one property of the multiplier.

2. $L^p$ Boundedness, Norm Growth, and Sharp Thresholds

(a) Generic Growth of Norms

Given a 0-homogeneous smooth phase $\Phi$ ($\Phi(\lambda\xi)=\Phi(\xi)$ for $\lambda>0$), and $d\ge2$, for generic $\Phi$ (specifically, those for which the restriction $\Phi|_{S^{d-1}}$ is a Morse function), the $L^p$ operator norm of $T_\Phi^t$ satisfies: $$\bigl\|T_\Phi^t\bigr\|_{L^p\rightarrow L^p} \gtrsim_{d,p, \Phi} \langle t\rangle^{d|1/p-1/2|} \quad \text{as } |t|\to\infty, \quad \langle t\rangle=(1+t^2)^{1/2}$$ (Bulj, 2023). This rate is optimal; previous constructions by Kovač–Bulj and Stolyarov exhibited phases for which such growth is realized, demonstrating that maximal norm growth is a generic phenomenon and that earlier conjectures positing slower possible growth (notably Maz'ya's conjecture) are invalid in generic settings.

(b) Sharp Sufficiency & Necessity Conditions in Wiener Amalgam and Modulation Spaces

On Wiener amalgam spaces $W^{p,q}_s$ in $\mathbb{R}^n$, for homogeneous phases $\mu(\xi)=|\xi|^\beta$ with $\beta\in(0,2]$ and nondegeneracy, the sharp threshold for boundedness $e^{i\mu(D)}:W^{p,q}_\sigma\to W^{p,q}$ is: $$\sigma\geq n(\beta-1)|1/p-1/q|, \text{ with strict inequality if } p\neq q$$ (Guo et al., 2018). For standard modulation spaces $M^{p,q}_s$, for highly oscillatory phases $\mu$ with second derivatives in a suitable Wiener amalgam class, one has boundedness, provided appropriate weight-loss proportional to $d(\alpha-2)|1/p-1/2|$ for phases of homogeneity $\alpha>2$ (Nicola et al., 2018). Similar optimality holds for general homogeneous $\mu$ on quasi-Banach Orlicz modulation spaces with precise dependence on the index $r$ governing quasi-norm type and the homogeneity $\alpha$ (Petersson, 18 May 2025). For $\alpha$-modulation spaces, extremely weak time-frequency assumptions on the second derivatives suffice to guarantee boundedness, extending earlier uniform covering results to the more general $\alpha$-covering regime (Zhao et al., 2019).

(c) Main Techniques

• Stationary Phase: For generic phases, lower bounds on $L^p$ norm growth are extracted using stationary phase arguments exploiting nondegenerate critical points on the sphere (Bulj, 2023).
• Time-Frequency and Dyadic Decomposition: For sharp threshold results in function spaces, a combination of local Taylor expansion of the phase and careful decomposition in phase-space uncovers the precise loss of regularity or weight necessary for boundedness (Guo et al., 2018, Nicola et al., 2018, Zhao et al., 2019).

3. Unimodular Multipliers as a Universal Test Family

A central structural result establishes that the theory of general bounded Fourier multipliers $T_m$ is inextricably linked to that for unimodular symbols. Specifically, for $1\leq p<\infty$: $$m \in \mathcal{M}_p(\mathbb{R}^n) \iff \exists c>0, s\in(0,1),\,\, \bigl\|e^{i t m}\bigr\|_{\mathcal{M}_p}\leq e^{c|t|^s}\ \forall t$$ where $\mathcal{M}_p$ is the $L^p$ multiplier algebra (Carro et al., 22 Jan 2026). This equivalence enables the reduction of $L^p$ multiplier theory to uniform subexponential norm growth estimates for exponential families $e^{itm}$, and underlies new theorems for rough, singular, and oscillatory integral operators.

4. Unimodular Multipliers in Noncommutative and Quantum Settings

Unimodular Fourier multipliers on group von Neumann algebras and quantum groups are analyzed through the structure of tracial von Neumann algebras and the associated noncommutative $L^p$ and Lorentz spaces. For second-countable unimodular groups $G$, the Plancherel trace provides an appropriate noncommutative integration, allowing joint treatment of $L^p$–spaces for $1 < p < \infty$ (Akylzhanov et al., 2016, Zhang, 2022, Ruzhansky et al., 2024, Akylzhanov et al., 2015). The norm of a (possibly non-commutative) Fourier multiplier is controlled by singular values of its symbol in $L^{r,\infty}(VN_R(G))$ with $1/r=1/p-1/q$.

In this framework, the $L^p\rightarrow L^q$ norm of a left-invariant operator $T_A$ (with symbol $A\in VN_R(G)$) satisfies: $$\|T_A\|_{L^p(G)\rightarrow L^q(G)} \lesssim \|A\|_{L^{r,\infty}(VN_R(G))}, \qquad r=(1/p-1/q)^{-1}$$ (Akylzhanov et al., 2016, Ruzhansky et al., 2024, Zhang, 2022, Akylzhanov et al., 2015).

Noncommutative Multiplier Structures

• Separating Multipliers: In noncommutative $L^p$-spaces $L^p(VN(G))$, unimodular multipliers $M_\phi$ associated with $\phi(s)=c\psi(s)$, for unitary characters $\psi$ and $|c|=1$, are precisely the isometric and separating maps (Arhancet et al., 2023).
• Absolute Dilations: Any selfadjoint, unital, completely positive Fourier multiplier $M_\varphi$ ($\varphi$ positive-definite, $\varphi(e)=1$) admits an invertible $L^p$-space dilation as a shift and compression on a larger von Neumann algebra (constructed via fermionic Fock spaces and crossed products) (Duquet, 2022).

5. Multilinear, Amenability, and Operator Algebraic Aspects

Multilinear unimodular multipliers generalize to $n$-linear operator-valued maps with multiplicatively bounded norm, connecting Fourier multipliers in group von Neumann algebras with Schur multipliers and matrix-valued operator theory. For unimodular, amenable groups, the norm equality between multilinear Fourier and Schur multiplier mappings is established, with transference principles, whereas in the nonamenable setting, the symmetry is broken (Caspers et al., 2022).

Operator-algebraic structure is further highlighted in the decomposable/CB Fourier multipliers $M^{\infty,\textrm{dec}}(G)$, characterized via Herz–Schur multipliers and projections from the space of completely bounded weak* operators, and tied to group amenability and inner amenability (Arhancet et al., 2022). In discrete $G$, every decomposable multiplier is unimodular.

6. Connections to Riesz Transforms and Sobolev Multiplier Theory

Unimodular multipliers bridge to the theory of noncommutative Riesz transforms and spectral multipliers. For functions $m(g)=\tilde{m}(b_\psi(g))$ (where $b_\psi$ is the cocycle for a conditionally negative length function $\psi$), H\"ormander–Mihlin-type theorems hold in terms of Sobolev–Besov norms of $\tilde{m}$ and Littlewood–Paley averages of noncommutative Riesz transforms (Junge et al., 2014). Constants are independent of dimension and admit Besov-type refinements.

7. Implications, Open Directions, and Optimality

The theory of unimodular Fourier multipliers reveals deep structural insights:

• Generic unimodular oscillatory multipliers attain maximal $L^p$ norm growth, invalidating earlier conjectured bounds (Bulj, 2023).
• Boundedness criteria in time–frequency spaces uncover phase-space decomposability with minimal regularity assumptions (Nicola et al., 2018, Zhao et al., 2019, Petersson, 18 May 2025).
• The equivalence between general and unimodular multiplier theory suggests that boundedness and optimality for rough singular/integral operators can be reduced to subexponential norm control for unimodular families (Carro et al., 22 Jan 2026).
• Operator-algebraic frameworks enable generalization to noncommutative and quantum settings, tying unimodularity, amenability, and positivity to structural results (Arhancet et al., 2022, Duquet, 2022).

Open directions include the extension of sharp threshold theory to rougher phases or weights, full characterization of multiplier classes on nonunimodular groups, and the infinite-dimensional and quantum group regimes (Akylzhanov et al., 2015, Zhang, 2022, Petersson, 18 May 2025).