Systematic classification of polynomial families yielding distributional defect identities for second-order operators

Hörmander's 2001 paper 'Asgeirsson's Mean Value Theorem and Related Identities' proved that the Asgeirsson mean value theorem for the ultrahyperbolic equation is witnessed by an explicit compactly supported distribution, and built a calculus in which one-dimensional homogeneous distributions $\chi_\pm^a$ composed with particular polynomials — the quartic $T_R$ attached to a pair of nondegenerate quadratic forms, a cubic polynomial obtained by inverting the characteristic surface $\{T_R=0\}$, and a multi-factor generalization — extend to global distributions and satisfy exact delta-layer defect identities under second-order operators, with prefactors depending on the signatures of the quadratic data. The paper develops these families case by case, under hypotheses that exclude low-dimensional and reducible situations. The problem posed here is to turn these examples into a systematic theory: classify the polynomial families admitting such extension and defect identities (admissible signatures, critical-set structure, precise prefactors, support decomposition), treat the exceptional low-dimensional and reducible cases where the paper's hypotheses fail, and settle the analytic status of the two-valued simple-layer densities produced by the Lorentz-signature instances, including their global single-valued realizations and their role in progressive solutions.

Background

For a linear partial differential equation, a mean value theorem asserts that solutions satisfy an identity between integral averages over geometric figures; the prototypical example is the Gauss mean value property of harmonic functions. Asgeirsson (1937) proved the corresponding statement for the ultrahyperbolic equation $(\Delta_x-\Delta_y)u=0$ on $\mathbb{R}^{2m}$, where $x,y\in\mathbb{R}^m$: if $u$ is continuous near the compact set $K_R=\{(x,y):|x|+|y|\le R\}$, then the integral of $u$ over the sphere $\{(x,0):|x|=R\}$ equals its integral over the sphere $\{(0,y):|y|=R\}$. Fritz John's monograph 'Plane Waves and Spherical Means Applied to Partial Differential Equations' connected this theorem to the geometry of the space of lines: solutions of the four-variable ultrahyperbolic equation arise as the compatibility conditions for a function on line space to be the X-ray transform (the operator integrating a function along lines) of a function on Euclidean 3-space.

A powerful reformulation, which Hörmander takes as his point of departure, uses distribution theory. A distribution is a continuous linear functional on the space $C_0^\infty$ of smooth compactly supported test functions; surface measures and Dirac measures are distributions. Writing $f_R$ for the difference of the two surface measures in Asgeirsson's theorem, the identity $\langle u,f_R\rangle=0$ for all solutions of the homogeneous equation is equivalent to the existence of a compactly supported distribution $\mu_R$ with $(\Delta_x-\Delta_y)\mu_R=f_R$ (the same duality principle underlying fundamental solutions, i.e. distributional solutions with a point source). Hörmander's Theorem 2.1 makes this explicit: with the quartic polynomial $A_R(x,y)=(|x|^2-|y|^2)^2-2R^2(|x|^2+|y|^2)+R^4$, one may take $\mu_R=\tfrac12\chi_+^{(1-m)/2}(\pi A_R/4R^2)$, where $\chi_+^a$ is the one-dimensional homogeneous distribution defined for $\operatorname{Re}a>-1$ by $t_+^a/\Gamma(a+1)$ and continued analytically in the exponent $a$, and where the composition denotes the pullback of $\chi_+^a$ by the scalar function $\pi A_R/4R^2$, regularized near the set where the polynomial is critical. The formula is closely related to the fundamental solution of the wave equation in $\mathbb{R}^{m+1}$; the paper's abstract states that the work was originally motivated by the progressive solutions of the wave equation (solutions concentrated near expanding wave fronts) studied by G. Friedlander and M. Riesz.

The engine for the generalizations is Hörmander's Theorem 3.3. Let $F$ be a smooth real function on a manifold whose critical set on its zero level, $S=\{F=0,\,F'=0\}$, is a submanifold of codimension $v>2$ on which the Hessian $F''$ has rank $v$ (a nondegenerate transverse quadratic structure), and let $L$ be a second-order differential operator that annihilates $\chi_\pm^{(2-v)/2}(F)$ off $S$. Then these compositions, a priori defined only outside $S$, admit canonical distributional extensions across $S$, and the extensions satisfy an exact defect identity: $L$ applied to them equals an explicit constant times $\gamma\,\delta_{S,F}$, where $\gamma=LF/(2v)$ on $S$, the constants involve $\sin(\pi v_\pm/2)$ with $(v_+,v_-)$ the signature (the pair of positive and negative dimensions) of the transverse quadratic form $F''/2$, and $\delta_{S,F}$ is the intrinsically normalized Dirac density supported on $S$ (Proposition 3.2). Applied to the quartic $T_R=\mathscr{Q}(x,y)^2-2R^2(\mathscr{Q}_1(x)+\mathscr{Q}_2(y))+R^4$ attached to a $\mathscr{Q}$-orthogonal splitting $V=V_1\oplus V_2$ with $\mathscr{Q}=\mathscr{Q}_1-\mathscr{Q}_2$ nondegenerate and $\dim V_1=\dim V_2$, this yields distributions $U_\pm$ that solve the constant-coefficient equation $\mathscr{Q}^*(\partial)U_\pm=0$ off the two singular quadrics $\{(x,0):\mathscr{Q}_1(x)=R^2\}$ and $\{(0,y):\mathscr{Q}_2(y)=R^2\}$, and whose defect under $\mathscr{Q}^*(\partial)$ is an explicit pair of delta layers (distributions of the form smooth density times a Dirac measure concentrated on a surface) on those quadrics, with signature-dependent sine prefactors; the construction is invariant under the Kelvin transform (the inversion $X\mapsto R^2X/\mathscr{Q}(X)$, familiar from the classical inversion symmetry of Newtonian potentials). Inverting the characteristic surface $\{T_R=0\}$ at a regular point produces surfaces governed by a cubic polynomial $P(t,x,y)=2(t_1^2-a^2)t_2+(t_1+a)\mathscr{Q}_1(x)+(t_1-a)\mathscr{Q}_2(y)$, with two additional hyperbolic coordinates $t_1,t_2$ appended, for which analogous extension and defect identities hold; the construction extends to a multi-factor polynomial built from distinct real parameters $a_1<\dots<a_\mu$, whose defect is a finite sum of delta layers supported on singular paraboloids. In Lorentzian signature the defect formulas involve smooth densities on the characteristic surfaces that take two values ('two-valued' simple-layer densities) and solve the wave equation.

All of this is developed case by case: the quartic, the cubic obtained by inversion, and the multi-factor generalization, each under hypotheses (codimension $v>2$, nondegenerate transverse Hessian, factor spaces of dimension at least two) that exclude degenerate and low-dimensional situations. The paper does not attempt to delimit the class of polynomials for which such constructions work, nor to analyze the exceptional cases or the two-valued densities that its Lorentzian examples produce.

Problem Statement

Hörmander's paper proves its extension and defect identities for three specific polynomial families — the quartic $T_R$ attached to a pair of nondegenerate quadratic forms, a cubic polynomial produced by inverting the characteristic surface $\{T_R=0\}$, and a multi-factor generalization with distinct real parameters — while its extension engine (Theorem 3.3) requires the critical set to be a submanifold of codimension $v>2$ with nondegenerate transverse Hessian, and the factor-space setup avoids the situations in which a factor space has dimension $1$. The directions the paper leaves unpursued constitute one coherent objective: replace the case-by-case examples by a systematic theory of the polynomial-composition defect calculus. Specifically:

(1) Classification. Determine and characterize the polynomial families $P$, and the second-order operators $L$ acting on them, for which the homogeneous compositions $\chi_\pm^a(P)$, at the exponent $a=(2-v)/2$ fixed by the codimension $v$ of the critical set $\{P=0,\,P'=0\}$, admit canonical distributional extensions and satisfy explicit delta-layer defect identities: which signatures are admissible, what structure the critical set $\{P=0,\,P'=0\}$ must have, what the precise prefactors are, and how the supports of the extensions decompose among the components of the zero set. A polynomial $P$ admits a canonical distributional extension when the local-limit construction of Theorem 3.3 (analytic continuation of the homogeneous pullback, regularized near $S$) yields a uniquely determined distribution, and the classification is sought over the class of nonconstant polynomials obtained from finitely many nondegenerate quadratic forms by finite sums and products and by the inversion construction that produces the cubic (the polynomial governing the inverted zero set $\{T_R=0\}$), together with parameterization over distinct real parameters $R>0$ and $a_1<\dots<a_\mu$; this list delimits the class rather than enumerating separate mandatory sub-deliverables, and it excludes restricting to polynomials that compose a single nondegenerate quadratic form. The known families ($T_R$, the cubic, the multi-factor polynomial) must be recovered as instances of the classification, and the classification must delimit its own boundary within this class, identifying polynomials for which no such identity holds together with explicit obstructions.

(2) Exceptional low-dimensional and reducible cases. Analyze the cases the hypotheses exclude: in particular the 'trivial/reducible' situations in which one factor space has dimension $1$, so that the analytic-continuation exponents of the one-dimensional homogeneous distributions can reach integer values and the extension mechanism changes character. Determine the precise behavior there: whether alternative factorizations or modified distributional constructions still yield extension and defect identities, and what replaces the sine-prefactor formulas.

(3) Two-valued Lorentzian densities. In the Lorentz-signature instances of the construction, the defect formulas exhibit two-valued smooth densities on characteristic surfaces — smooth densities carrying two branches over the surface, equivalently smooth densities on a double cover of the characteristic surface — that solve the wave equation. Establish their analytic and geometric status: whether they admit global single-valued realizations, what their interpretation is, and what role they play in constructing progressive solutions in the sense of Friedlander and Riesz.

These three components are linked parts of the single objective: component (2) is the completion of the classification at the boundary where its hypotheses fail, and component (3) is the Lorentzian chapter of the same defect calculus.

References

1. Lars Hörmander, 'Asgeirsson's Mean Value Theorem and Related Identities', Journal of Functional Analysis 184(2), 377-401 (2001). DOI: 10.1006/jfan.2001.3743, URL: https://doi.org/10.1006/jfan.2001.3743
2. Leifur Ásgeirsson, 'Über eine Mittelwertseigenschaft von Lösungen homogener linearer partieller Differentialgleichungen 2. Ordnung mit konstanten Koeffizienten', Mathematische Annalen 113, 321-346 (1937). DOI: 10.1007/BF01571637, URL: https://doi.org/10.1007/bf01571637 (the original Asgeirsson mean value theorem)
3. Fritz John, 'Plane Waves and Spherical Means Applied to Partial Differential Equations', Dover Publications, Mineola, NY, 2004 (reprint of the 1955 Interscience edition). ISBN: 0-486-43804-X, URL: https://zbmath.org/?q=an%3A1073.35003
4. F. G. Friedlander, 'Sound Pulses', Cambridge Monographs on Mechanics and Applied Mathematics, Cambridge University Press, 1958. URL: https://zbmath.org/?q=an%3A0079.41001 (progressive-wave solutions of the wave equation)
5. Marcel Riesz, 'L'intégrale de Riemann-Liouville et le problème de Cauchy', Acta Mathematica 81, 1-222 (1949). DOI: 10.1007/BF02395016, URL: https://doi.org/10.1007/bf02395016 (Riesz kernels and the Cauchy problem for the wave equation)
6. Norbert Ortner and Peter Wagner, 'Deduction of L. Hörmander's extension of Ásgeirsson's mean value theorem', Bulletin des Sciences Mathématiques 127(10), 835-843 (2003). DOI: 10.1016/S0007-4497(03)00045-9, URL: https://doi.org/10.1016/S0007-4497(03)00045-9
7. Toshiyuki Kobayashi and Bent Ørsted, 'Analysis on the minimal representation of O(p,q) — III. Ultrahyperbolic equations on $\mathbb{R}^{p-1,q-1}$', Advances in Mathematics 180(2), 551-595 (2003). DOI: 10.1016/S0001-8708(03)00014-8, URL: https://arxiv.org/abs/math/0111086
8. Fikret Gölgeleyen and Masahiro Yamamoto, 'Uniqueness of solution of an inverse source problem for ultrahyperbolic equations', Inverse Problems 36(3), 035008 (2020). DOI: 10.1088/1361-6420/ab63a2, URL: https://arxiv.org/abs/1909.01399
9. Fikret Gölgeleyen and Masahiro Yamamoto, 'Stability of inverse problems for ultrahyperbolic equations', Chinese Annals of Mathematics, Series B 35(4), 527-556 (2014). DOI: 10.1007/s11401-014-0848-6, URL: https://doi.org/10.1007/s11401-014-0848-6
10. Yuxin Wang, Yueyang Shen, Daxuan Deng and Ivo D. Dinov, 'Determinism, well-posedness, and applications of the ultrahyperbolic wave equation in spacekime', Partial Differential Equations in Applied Mathematics 5, 100280 (2022). DOI: 10.1016/j.padiff.2022.100280, URL: https://doi.org/10.1016/j.padiff.2022.100280
11. Viktor Korzyuk and Jan Rudzko, 'Curvilinear Parallelogram Identity and Mean-Value Property for a Semilinear Hyperbolic Equation of the Second Order', Eurasian Mathematical Journal 15(2), 61-74 (2024). DOI: 10.32523/2077-9879-2024-15-2-61-74, URL: https://emj.enu.kz/index.php/main/article/view/343
12. G. G. Baghaturia and M. Z. Menteshashvili, 'A General Integral of a Quasilinear Equation and Application to a Nonlinear Characteristic Problem', Siberian Mathematical Journal 60(6), 940-951 (2019). DOI: 10.1134/S0037446619060028, URL: https://doi.org/10.1134/S0037446619060028
13. Guillem Cobos and Brendan Guilfoyle, 'An Extension of Asgeirsson's Mean Value Theorem for Solutions of the ultra-hyperbolic Equation in Dimension Four' (2020). arXiv: 2011.11330, URL: https://arxiv.org/abs/2011.11330
14. Guillem Cobos and Brendan Guilfoyle, 'A Conformal Mean Value Theorem for Solutions of the Ultrahyperbolic Equation' (2022). arXiv: 2210.08155, URL: https://arxiv.org/abs/2210.08155
15. I. P. Polovinkin, 'Mean Value Theorems for Linear Partial Differential Equations', Journal of Mathematical Sciences 197(3), 399-403 (2014). DOI: 10.1007/s10958-014-1721-6, URL: https://doi.org/10.1007/s10958-014-1721-6