Random Tensor Estimates and Deterministic Diagonal Resolutions for Mixed Paracontrolled Operators

Technical Summary: Random Tensor Estimates and Deterministic Diagonal Resolutions for Mixed Paracontrolled Operators

Problem Statement
The paper addresses the construction of singular stochastic partial differential equations (SPDEs), specifically focusing on the convergence of mixed paracontrolled operators arising in quadratic expansions. In dispersive equations, enhanced random objects often take operator-valued forms where the Duhamel multiplier contributes oscillatory structure, but the deterministic resonant product may fall below its classical regularity threshold.

The specific object of study is the mixed second-level operator attached to an ordered triple $(i; j, k)$ in a coupling graph:
$$T_{i;j,k}(w) = I_i(w \prec \Psi_j) \circ \Psi_k$$
where $I_i$ is a Duhamel channel, $\prec$ denotes a paraproduct (low-high interaction), and $\circ$ denotes a resonant product. These operators are singular in the range where the deterministic resonant-product threshold fails (specifically when $\lambda_i + \alpha_j + \alpha_k \le d$). The challenge is to establish operator-level convergence for these blocks in arbitrary dimensions, accommodating finite color families, random initial data, colored-in-time forcing, and frequency-dependent amplitudes, without relying on model-specific cancellations for every instance.

Methodology
The paper isolates a reusable operator-level mechanism by decomposing the mixed operator at a finite cutoff $\Lambda$ into two distinct components via a finite Wick split:
$$T^\tau_\Lambda = D^\tau_\Lambda + B^\tau_\Lambda$$
where $\tau = (i; j, k)$.

1. Finite Wick Split and Decomposition:

   • Deterministic Diagonal ($D^\tau_\Lambda$): This component arises from the scalar expectation of the product of stochastic legs. Due to Fourier-diagonal covariance, the contraction enforces $\ell + r = 0$, which forces the external incidence $n = q$. Thus, $D^\tau_\Lambda$ is a deterministic Fourier-diagonal Volterra multiplier acting on the low input.
   • Centered Random Component ($B^\tau_\Lambda$): This component retains the full paracontrolled incidence $n = q + \ell + r$. It is a centered second-homogeneous Gaussian chaos.

2. Centered Component Estimation (Probabilistic):
   The paper employs a finite random tensor estimate to control $B^\tau_\Lambda$.

   • Local Normalization: The Gaussian legs are normalized using a local covariance square root on each Fourier involution class, handling rank-defective covariances via quotient spaces.
   • Tensor Inequality: The centered block is treated as a random matrix. The authors apply a finite second-order tensor estimate (generalizing results from Deng, Nahmod, and Yue [4] and [9]) involving eight specific "flattening" norms of the coefficient tensor.
   • Incidence Counting: The proof records the specific Schur-type bookkeeping required for the incidence relation $n = q + \ell + r$, showing that the operator norm is bounded by $N^{d/2 - \Gamma_\tau}(M^{d/2} + Q^{d/2})$, where $\Gamma_\tau = \lambda_i + \alpha_j + \alpha_k$.
   • Time Lift: A Banach-valued time-lift argument (Kolmogorov-type) upgrades fixed-time estimates to uniform estimates over the Volterra time simplex, absorbing small time-loss exponents into strict dyadic margins.

3. Deterministic Component Estimation (Analytic):
   The paper treats $D^\tau_\Lambda$ via a "deterministic diagonal resolution."

   • Interface Separation: Instead of proving cancellations within the abstract theorem, the paper requires the user to supply a resolution where the raw diagonal is decomposed into a retained part $R^\tau_\Lambda$ and a subtracted counterterm $C^\tau_\Lambda$.
   • Admissibility: The retained part must satisfy a scalar dyadic Volterra bound (either absolute or oscillatory with phase gains). This allows model-specific phase differences, conjugation symmetries, and orbit sums to be handled as explicit inputs.
   • Orbit Sums: If a model requires grouping branches into orbits (e.g., via time parity or conjugation), the estimate is imposed on the direct sum of the orbit, not individual branches.

Key Contributions

• Abstract Operator-Level Criterion: The paper provides a general convergence theorem (Theorem 2.8) that separates the universal probabilistic mechanism (random tensor estimates) from model-specific deterministic facts (diagonal cancellations and phase gaps).
• Deterministic Diagonal Resolution Interface: It formalizes a "black box" interface where concrete models only need to verify that their retained diagonal kernels satisfy specific scalar Volterra bounds. This abstracts the "color-phase" analysis found in specific equations (like the two-speed Klein–Gordon system in [13]) into a reusable theorem.
• Frequency-Envelope Variant: Theorem 2.12 extends the results to general Gaussian time profiles and frequency-dependent amplitudes using dyadic envelopes rather than power laws, accommodating random initial data and colored-in-time forcing under a unified framework.
• Explicit Incidence Counting: The paper rigorously derives the eight flattening norms required for the tensor estimate in the context of mixed paracontrolled blocks, clarifying the precise combinatorial dependencies on spatial dimension and frequency shells.

Results
The main result (Theorem 2.8) establishes that for a finite family of mixed triples $\mathcal{M}$, if:

1. The centered admissibility conditions (Definition 2.2) are met (ensuring the random tensor summability window is non-empty), and
2. A dyadically admissible deterministic diagonal resolution is chosen (Definition 2.4),

Then the resolved operators $T^{\tau, P}_\Lambda$ converge in $L^p(\Omega)$ and almost surely in the operator topology $\mathcal{L}(E^{2,\sigma}_{T, \mathcal{M}}, Y^\tau_T)$ to a limit $T^{\tau, P} = R^\tau + B^\tau$.

Specifically, the centered component $B^\tau$ maps $X^\sigma_T$ to $C_T H^{s-\lambda_i} \cap L^1_T B^{\sigma-\lambda_i}_{2,\infty}$, while the retained diagonal $R^\tau$ maps the space with time-derivative control $E^{2,\sigma}_{T, \mathcal{M}}$ to the same target.

In the specific case of the three-dimensional equal-gain wave/Klein–Gordon system, the theorem recovers the known mixed-operator window $s_2 < 4\alpha - 3$ and $3 - 3\alpha < \sigma < 4\alpha - 3$ (with strict losses), demonstrating that the abstract mechanism correctly reproduces the constraints of concrete models.

Significance and Claims
The paper claims to extract the "reusable mixed-operator mechanism" underlying concrete color-phase diagonal analyses. Its significance lies in its modularity:

• It does not attempt to reprove a complete solution theory for every coupling graph.
• It does not automatically classify all deterministic diagonal cancellations.
• Instead, it provides the operator-level criterion that an application can invoke once the finite coupling graph, the retained scalar diagonal orbits, and the corresponding Volterra estimates are supplied.

This separation allows the same mixed operator estimate to be reused in diverse contexts, including random initial data problems, colored-in-time noise models, multi-component systems, and other quadratic paracontrolled expansions, provided the specific diagonal resolution inputs are verified. The paper positions itself as a complement to concrete equation-level theories (like [13]), offering the abstract probabilistic engine while leaving the model-specific scalar classification to the user.