2 The Kernel-Weighted Defect
The Volterra–Lévy Itô decomposition, the \(\nu \)-singular/\(\nu \)-regular splitting, the defect vanishing characterization, and \(\nu \)-singularity persistence. The central object is the kernel-weighted Leibniz defect \(\Gamma _V^f = \tilde{\delta }_J(g_f^K)\), which equals the structurally unhedgeable component of risk [1, Theorem 5.2].
2.1 Process, Payoff, and Bridge
CombinedEnergySpace.VolterraProcessData def:CombinedEnergySpace Data for a Volterra process: \(V_T = \delta _W(\text{kernel\_ cont}) + \tilde{\delta }_J(\text{kernel\_ jump})\), decomposing the terminal value into its continuous and jump stochastic integrals against the kernel \(K(T,\cdot )\) [1, Definition 2.1, eq. (2.2)].
CombinedEnergySpace.SmoothPayoff def:CombinedEnergySpace A smooth function \(f \in C_b^3(\mathbb {R})\) applied to \(V_T\). Carries \(f\), \(f'\), the Taylor remainder \(g_f^K(v,z) = f(v+Kz) - f(v) - f'(v)Kz\) (the kernel-weighted defect integrand), with \(g_f^K(v,0) = 0\) and a uniform quadratic bound \(|g_f^K(v,z)| \leq C_{\mathrm{defect}} \cdot z^2\) where \(C_{\mathrm{defect}} = \tfrac {1}{2}\| f''\| _\infty K(T,t)^2\) [1, Theorem 3.1, eq. (3.7)].
CombinedEnergySpace.HilbertJumpBridge def:VolterraProcessData, def:SmoothPayoff Connects the Hilbert Itô formula (from OperatorDerivative.lean [3]) and the Banach defect theory (from BanachLevyComplete.lean [4]) on the combined space \(H_W \times \tilde{H}_J\). The bridge carries exactly one stochastic analysis input: the Lévy–Itô formula (levy_ito, from Applebaum [2], Theorem 4.4.7) applied to the semimartingale \(V_T(t) = \int _0^t K(T,s)\, dL_s\). It also carries the Taylor split (algebraic), the defect coherence (connecting the abstract Banach element to the pointwise formula), and the payoff-process connection. The dependency chain is: process \(\to \) payoff \(\to \) bridge \(\to \) decomposition theorem.
2.2 The Volterra–Lévy Itô Decomposition
CombinedEnergySpace.kernel_weighted_defect def:HilbertJumpBridge \(\Gamma _V^f := \tilde{\delta }_J(g_f^K)\), the compensated Poisson integral of the defect integrand. This is the kernel-weighted Leibniz defect—the precise content invisible to Malliavin calculus, and the structurally unhedgeable component of risk in Volterra–Lévy markets [1, Theorem 5.2, eq. (5.10)].
CombinedEnergySpace.defect_is_taylor_remainder def:HilbertJumpBridge, def:SmoothPayoff \(g_f^K(v,z) = f(v + K \cdot z) - f(v) - f'(v) \cdot K \cdot z\). The defect integrand IS the Taylor remainder of \(f\) composed with the kernel—connecting the abstract \(\tilde{H}_J\) element to the paper’s pointwise formula [1, eq. (3.7)]. ALGEBRAIC: unfolds defect_coherence.
Immediate from the bridge’s defect_coherence field.
CombinedEnergySpace.ito_composition def:HilbertJumpBridge, def:kernel_weighted_defect \(f(V_T) = f(0) + \delta _W(f' \cdot K) + \tilde{\delta }_J(\text{linear}) + \Gamma _V^f + \text{correction}\). This is the five-term identity obtained by composing the Lévy–Itô formula with the Taylor split of the jump integrand. ALGEBRAIC: from levy_ito, taylor_split, and map_add (linearity of \(\tilde{\delta }_J\)).
Rewrite with levy_ito, substitute taylor_split, apply map_add, close with abel.
CombinedEnergySpace.volterra_levy_ito_decomposition lem:ito_composition
This is the paper’s Theorem 3.1 [1]. It is a theorem (derived from ito_composition by algebraic rearrangement), not a structure field. The decomposition composes the Hilbert Itô formula (OperatorDerivative.lean [3]) and the Banach defect theory (BanachLevyComplete.lean [4]) on the combined space \(H_W \times \tilde{H}_J\).
Apply ito_composition, rearrange with abel.
CombinedEnergySpace.two_regime_decomposition def:kernel_weighted_defect, def:HilbertJumpBridge \(\tilde{\delta }_J(h_f^K) = \tilde{\delta }_J(\text{linear}) + \Gamma _V^f\). The compensated Poisson integral splits into a \(\nu \)-singular part (linear in jump size \(z\), with infinite \(\tilde{H}_J\)-norm) and a \(\nu \)-regular part (the defect \(\Gamma _V^f\), with finite \(\tilde{H}_J\)-norm). This is the content of [1, Theorem 5.2(c)]. ALGEBRAIC: from taylor_split and map_add.
Rewrite with taylor_split, apply map_add.
2.3 Moment Structure and Singularity
NuMomentStructure Captures the integrability dichotomy for stable-type Lévy measures with index \(\gamma \in (1,2)\) [1, Assumption L1]: \(\int |z|^p\, \nu (dz) = \infty \) for \(p {\lt} \gamma \) (\(\nu \)-singular regime) and \(\int |z|^{2p}\, \nu (dz) {\lt} \infty \) for \(2p {\gt} \gamma \) (\(\nu \)-regular regime). The working exponent \(p \in (1,\gamma )\) and the small-jump constant \(c_\gamma \) from the stable density bound \(\nu (dz) \sim c_\gamma |z|^{-(1+\gamma )}\, dz\).
NuMomentStructure.small_jump_2p_moment_pos def:NuMomentStructure \(0 {\lt} 2c_\gamma / (2p - \gamma )\). The small-jump moment is positive and finite in the \(\nu \)-regular regime. CONCRETE: from div_pos.
Rewrite with small_jump_2p_moment_eq, apply div_pos.
NuMomentStructure.p_moment_divergence_exponent def:NuMomentStructure \(p - \gamma {\lt} 0\). The divergence exponent is negative, encoding \(\int |z|^p\, \nu (dz) = \infty \). Matches stable_small_jump_moment_diverges in BanachLevyComplete.lean [4]. CONCRETE.
linarith [hp_upper].
NuMomentStructure.two_p_gt_γ def:NuMomentStructure \(2p {\gt} \gamma \). The convergence exponent exceeds the stable index, encoding \(\int |z|^{2p}\, \nu (dz) {\lt} \infty \). Matches stable_small_jump_moment_finite in BanachLevyComplete.lean [4]. CONCRETE: from \(p {\gt} 1\) and \(\gamma {\lt} 2\).
linarith [hp_lower, hγ_upper].
CombinedEnergySpace.defect_norm_from_uniform_bound def:SmoothPayoff, def:NuMomentStructure \(0 {\lt} C_{\mathrm{defect}}^p \cdot \int _{|z| \leq 1} |z|^{2p}\, \nu (dz)\). Connects the uniform Taylor bound to the scaling law: the product of the defect constant and the small-jump moment is positive, establishing the leading term in the scaling estimate [1, Theorem 6.1, eq. (6.2)]. CONCRETE.
mul_pos with rpow_pos_of_pos and small_jump_2p_moment_pos.
NuSingularityData The kernel norm \(\int _0^T |K(T,s)|^p\, ds {\gt} 0\) (finite by the kernel regularity assumption [1, Assumption K1]) and truncated \(z\)-moments \(\int _{|z| \leq R} |z|^p\, \nu (dz)\) which are unbounded as \(R \to \infty \) (since \(p {\lt} \gamma \) and \(\nu \) is \(\gamma \)-stable near zero).
CombinedEnergySpace.nu_singularity_persists def:NuMomentStructure, def:NuSingularityData \(\forall M,\; \exists R,\; \| K\| _p \cdot \int _{|z| \leq R} |z|^p\, \nu (dz) {\gt} M\). The \(\tilde{H}_J\)-norm of the linear integrand \(K(T,\cdot )z\) is infinite: the product of a positive kernel norm and an unbounded \(z\)-moment exceeds any bound. Hence \(V_T \notin \mathrm{im}(\tilde{\delta }_J)\): the kernel cannot regularize the jump-size singularity, because \(K\) acts on time while the singularity is in \(z\) [1, Proposition 6.1]. CONCRETE: from mul_lt_mul_of_pos_left and field_simp.
Divide \(M\) by kernel_p_norm, use truncated_z_moment_unbounded, multiply back.
2.4 Defect Vanishing
CombinedEnergySpace.JumpPresence def:HilbertJumpBridge Whether the Lévy measure is nontrivial. When \(\nu = 0\) (no jumps), all jump sizes are \(z = 0\), so by defect_at_zero (the Taylor remainder at zero displacement is zero), the defect integrand vanishes in \(\tilde{H}_J\). This connects to the fundamental dichotomy: \(D\) satisfies the Leibniz rule if and only if the driver has no jumps [4, Theorem B].
CombinedEnergySpace.defect_vanishes_without_jumps_volterra def:JumpPresence, def:kernel_weighted_defect When \(\nu = 0\): \(\Gamma _V^f = 0\). The kernel-weighted defect vanishes, recovering perfect hedging in pure rough volatility models (El Euch–Rosenbaum 2019). The defect IS the non-Gaussian content: it vanishes precisely when the driver is Gaussian [1, Proposition 7.1]. ALGEBRAIC: from no_jumps_zero_defect and map_zero.
Rewrite with zero defect integrand, apply map_zero.