betaFeedback

WorksMathematics

Hyperbolic systems with non-diagonalisable principal part and variable multiplicities, IV: time-dependent characteristics, representation of solutions and regularity

Abstract

We consider first-order linear hyperbolic systems with time-dependent pseudo-differential principal symbols and characteristic roots of variable multiplicity. Building on the time-independent work of Kamotski and the first author, we introduce a finite-type condition along the lifted bicharacteristic flow that accounts for the contact of time-dependent characteristic roots. Under this condition, we construct a representation of solutions using time-dependent Fourier integral operators and a Picard series whose terms become increasingly smooth. This leads to an asymptotic expansion in smoothness and local LpL^p regularity estimates with the expected loss of derivatives. We also apply these results to higher-order scalar weakly hyperbolic equations with involutive characteristics and time-dependent principal symbols. Finally, we consider upper triangular systems with possibly non-diagonalisable principal parts and derive a representation formula involving integrated Fourier integral operators. This extends the earlier results of Garetto, Jäh and the first author to time-dependent principal parts.

Introduction

The present paper has two objectives. First, we extend the representation formulas and regularity results of Kamotski and Ruzhansky [12] to hyperbolic systems with time-dependent principal symbols and characteristic roots of variable multiplicity. Second, we continue the study of hyperbolic systems with non-diagonalisable principal parts and variable multiplicities initiated in [4] and developed further in [5, 6]. In this fourth part of the series, we establish representation formulas and regularity estimates for such systems with time-dependent principal symbols.

Problem setting

We consider the Cauchy problem for first-order hyperbolic systems

{Dtu=P(t,x,Dx)u,(t,x)∈[0,T]×Rn,u(0,x)=u0(x),(1)\begin{cases} D_tu = P(t,x,D_x)u, \quad(t,x) \in[0,T] \times\mathbb{R}^n,\\ u(0,x) = u_0(x), \tag*{(1)} \end{cases}

where Dt=−i∂tD_t=-i\partial_t and PP is an m×mm \times m matrix of classical pseudo-differential operators of order one, depending smoothly on tt. The principal symbol of PP has smooth real eigenvalues a1,…,ama_1,\ldots,a_m, which depend on (t,x,ξ)(t,x,\xi) and may coincide with variable multiplicities. We study two classes of such systems:

  1. systems whose principal symbol is microlocally smoothly diagonalisable (Sections 3–4);

  1. upper triangular systems, which need not be diagonalisable (Section 5).

2020 Mathematics Subject Classification. Primary 35L45; Secondary 35S30, 35B65, 35L40, 35L30, 35A27.

Key words and phrases. Hyperbolic systems, variable multiplicity, time-dependent characteristics, Fourier integral operators, Picard series, LpL^p regularity, non-diagonalisable systems.

For the first class, we extend the approach of [12] by introducing a finite-type condition adapted to the time-dependent characteristic roots. For the second, we extend the representation theory of [5] to time-dependent principal symbols.

Background and motivation

Strictly hyperbolic systems and constant multiplicity

For strictly hyperbolic systems, the principal symbol can be diagonalised, and the solution of (1) can be represented as a sum of Fourier integral operators acting on the initial data [10, 3]. The calculus of Fourier integral operators then provides a well-established framework for studying regularity and the propagation of singularities [2], while the spectral asymptotics of PP are described by the Weyl formula [9, 1]. When the characteristic roots have constant multiplicity, the principal symbol can be reduced to block-diagonal form, allowing much of the strictly hyperbolic theory to be carried over to this setting. We refer to the works of Taylor [21], Kumano-go [13], and Kumano-go, Taniguchi and Tozaki [14]. Further accounts can be found in Ivrii [11] and, for analytic coefficients, in Nishitani [17].

Variable multiplicity and finite-type condition

When eigenvalues of a system have variable multiplicity, the block-diagonal structure breaks down, making direct diagonalisation impossible. Kucherenko [15] addressed this difficulty by constructing formal asymptotic solutions in this setting. The first step toward a rigorous theory came from Rozenblum [18], who built the fundamental solution using the Picard series, a method of successive approximation. He proved that this solution admits a well-controlled asymptotic expansion with respect to smoothness. His approach relied on a transversality condition: wherever two eigenvalues aja_j and aka_k coincide at a point (x,ξ)(x,\xi), their Poisson bracket must satisfy

{aj,ak}(x,ξ)≠0.\{a_j,a_k\}(x,\xi) \ne0.

Kamotski and the first author [12] later extended this theory by relaxing that requirement. Instead of full transversality, they imposed the weaker finite-type condition: for some integer ν≥1\nu\ge1, the iterated Poisson brackets satisfy

{aj,{aj,…,{aj,ak}…}}(ν)(x,ξ)≠0.\{a_j,\{a_j,\ldots,\{a_j,a_k\}\ldots\}\}^{(\nu)}(x,\xi) \ne0.

This condition permits so-called involutive characteristics, where the Poisson bracket itself vanishes. Under these more general geometric conditions on the characteristics, they showed that solutions are represented as a Picard series whose terms are iterated Fourier integral operators. This series serves as an asymptotic expansion with respect to smoothness, and it yields a precise description of propagation of singularities together with sharp regularity properties of solutions. These results were applied in two directions: to establish regularity estimates for scalar weakly hyperbolic equations with involutive characteristics, and to derive the first and second terms of the spectral asymptotics for the corresponding self-adjoint elliptic systems.

Building on this framework, Garetto, Jäh, and the first author proved well-posedness of the Cauchy problem in anisotropic Sobolev spaces for non-diagonalisable hyperbolic systems with variable multiplicities [4], and established a representation formula for the solution in terms of Fourier integral operators [5], which in turn yields results on regularity and propagation of singularities; see [6] for the case of singular coefficients and [7] for the higher order hyperbolic equations with multiplicities. These results extend those of [12] to the non-diagonalisable case, although the representation formula of [5] still assumes that the principal symbol is independent of time.

The time-independent principal symbol is a restrictive assumption. For example, in many physical systems such as in elasticity, in electromagnetism in inhomogeneous media, and in geophysical wave propagation, the coefficients depend on time. Without this assumption, we can extend the theory to a wider class of problems. Therefore, this paper focuses on the extension of the framework of [12] to the principal symbol depending on time and spatial variables. This is not a direct generalisation: when the time-dependent principal symbol changes the geometry of the characteristics and the structure of the Hamiltonian flow, and the Picard series must be constructed differently. With these modifications, we obtain the representation formulas, regularity estimates and propagation results in the general setting. Moreover, the same approach works also for non-diagonalisable upper triangular systems.

The geometric condition

In [12], using iterated Poisson brackets, they measure the order of contact between characteristic surfaces as follows

Hajνak={aj,{aj,…,{aj,ak}…}⏟ν times}.H_{a_j}^{\nu}a_k=\{a_j,\underbrace{\{a_j,\ldots,\{a_j,a_k\}\ldots\}}_{\nu\ \text{times}}\}.

When aja_j varies with time, the bicharacteristic flow gains an explicit time component, so the spatial Poisson bracket alone fails to represent the contact geometry. We suggest the operator as a natural substitute

Djf:=∂tf+Haj(f),j=1,…,m,\mathcal{D}_j f:=\partial_t f+H_{a_j}(f),\qquad j=1,\ldots,m,

where

Hg(f)={g,f}=∑k=1n(∂g∂ξk∂f∂xk−∂g∂xk∂f∂ξk)H_g(f)=\{g,f\}=\sum_{k=1}^{n}\left(\frac{\partial g}{\partial\xi_k}\frac{\partial f}{\partial x_k}-\frac{\partial g}{\partial x_k}\frac{\partial f}{\partial\xi_k}\right)

denotes the Poisson bracket in the spatial variables. The integral curves of Dj\mathcal{D}_j are the lifted bicharacteristics of aja_j in the extended phase space [0,T]×T∗Rn[0,T]\times T^*\mathbb{R}^n:

t˙=1,x˙=∂ξaj(t,x,ξ),ξ˙=−∂xaj(t,x,ξ).\dot{t}=1,\qquad\dot{x}=\partial_{\xi}a_j(t,x,\xi),\qquad\dot{\xi}=-\partial_xa_j(t,x,\xi).

Differentiation along these curves, rather than along the purely spatial Hamiltonian flow, is what the time-dependent setting demands.

We assume that, at points where two non-identical characteristic roots coincide, aj=aka_j=a_k, the lifted bicharacteristics generated by aja_j meet the coincidence set {aj=ak}\{a_j=a_k\} with finite order of contact. Equivalently, there exists a uniform integer M≥1M\geq1 such that, at every such point, Djν(aj−ak)≠0\mathcal{D}_j^\nu(a_j-a_k)\ne0 for some 1≤ν≤M1\leq\nu\leq M. This leads to the following finite-type contact condition along the lifted bicharacteristic flow.

Condition C. The eigenvalues a1,…,ama_1,\ldots,a_m of the principal symbol of PP are smooth and real, and there exists an integer M≥1M\geq1 such that for every (t,x,ξ)∈[0,T]×T∗Rn(t,x,\xi)\in[0,T]\times T^*\mathbb{R}^n with ξ≠0\xi\ne0 and every pair j≠kj\ne k, if aj(t,x,ξ)=ak(t,x,ξ)a_j(t,x,\xi)=a_k(t,x,\xi) and aj≢aka_j\not\equiv a_k near (t,x,ξ)(t,x,\xi), then there exists an integer 1≤ν≤M1\leq\nu\leq M for which

Djν(aj−ak)(t,x,ξ)=(∂t+Haj)ν−1(∂t(aj−ak)+Haj(aj−ak))(t,x,ξ)≠0.(2)\mathcal{D}_j^\nu(a_j-a_k)(t,x,\xi)=\left(\partial_t+H_{a_j}\right)^{\nu-1}\left(\partial_t(a_j-a_k)+H_{a_j}(a_j-a_k)\right)(t,x,\xi)\ne0. \tag*{(2)}

For the time-independent eigenvalues, the operator Dj\mathcal{D}_j reduces to HajH_{a_j}, and Condition C recovers the autonomous condition of [12]:

Djν(aj−ak)(x,ξ)=Hajν(aj−ak)(x,ξ)=−{aj,{aj,…,{aj,ak}…}}⏟ν times(x,ξ)≠0,1≤ν≤M.\mathcal{D}_j^\nu(a_j-a_k)(x,\xi)=H_{a_j}^\nu(a_j-a_k)(x,\xi)=-\underbrace{\{a_j,\{a_j,\ldots,\{a_j,a_k\}\ldots\}\}}_{\nu\ \text{times}}(x,\xi)\ne0,\qquad1\leq\nu\leq M.

In this context, Condition C is the natural time-dependent generalisation of the finite-type condition introduced in [12], serving as the geometric foundation for the subsequent analysis.

Main results

Before stating the precise theorems, we describe our results informally. For diagonalisable systems, the solution of (1.1) can be written as u=EVu0u=EVu_0, where EE is a propagator built from time-dependent Fourier integral operators associated with the lifted bicharacteristic flows of the eigenvalues, and VV is a Picard series. Under Condition C, the terms of this series become progressively smoother, so the Picard series is an asymptotic expansion in smoothness. This smoothing property yields LpL^p regularity for solutions of (1.1), as well as LpL^p estimates for higher-order scalar weakly hyperbolic equations with involutive characteristics, for all 1<p<∞1<p<\infty. Finally, for upper triangular systems, which need not be diagonalisable, we obtain a representation formula for the solution in terms of integrated Fourier integral operators.

For diagonalisable systems, we decompose P=A+BP=A+B, where AA is a smooth time-dependent family of pseudo-differential operators of order one whose principal symbol is microlocally smoothly diagonalisable, and BB collects the lower-order terms as a smooth time-dependent family of pseudo-differential operators of order zero. The solution of (1.1) is then written as u(t)=E(t)V(t)u0u(t)=E(t)V(t)u_0, where E(t)E(t) is the propagator of the diagonal principal part and V(t)=I+∑ℓ≥1Qℓ(t)V(t)=I+\sum_{\ell\geq1}Q_\ell(t) is the Picard series correcting for the off-diagonal contributions. Both EE and VV are constructed precisely in Section 2, with the Picard series given in (2.6)–(2.7).

Our first main result shows that the iterated Picard terms QℓQ_\ell gain smoothness as ℓ\ell increases. Consequently, the series is an asymptotic expansion in smoothness, and the regularity results that follow are derived from this property.

Theorem 1.1 (Smoothing of the Picard series). Assume that A(t,x,ξ)A(t,x,\xi) is microlocally smoothly diagonalisable and that its eigenvalues aj(t,x,ξ)a_j(t,x,\xi) satisfy Condition C. Then there exists a sequence 0≤N(ℓ)→∞0\leq N(\ell)\to\infty such that each operator Qℓ(t)Q_\ell(t) defined in (2.7) extends to a bounded map

Qℓ(t):Lcomp2(Rn)⟶HlocN(ℓ)(Rn),t∈[0,T].Q_\ell(t):L^2_{\mathrm{comp}}(\mathbb{R}^n)\longrightarrow H^{N(\ell)}_{\mathrm{loc}}(\mathbb{R}^n),\qquad t\in[0,T].

In particular, the Picard series (2.6) is an asymptotic expansion with respect to smoothness.

The smoothing property of Theorem 1.1, combined with the abstract regularity principle of Theorem 4.1, transfers regularity from the data to the solution. This yields the following LpL^p estimates for solutions of (1.1).

References

  1. [1]J. J. Duistermaat and V. Guillemin, The spectrum of positive elliptic operators and periodic bicharacteristics, Invent. Math. 29 (1975), 39–79.
  2. [2]J. J. Duistermaat and L. Hörmander, Fourier integral operators. II, Acta Math. 128 (1972), 183–269.
  3. [3]J. J. Duistermaat, Fourier Integral Operators, Birkhäuser, Boston, 1996.
  4. [4]C. Garetto, C. Jäh, and M. Ruzhansky, Hyperbolic systems with non-diagonalisable principal part and variable multiplicities, I: well-posedness, Math. Ann. 372 (2018), no. 3-4, 1597–1629.DOI
  5. [5]C. Garetto, C. Jäh, and M. Ruzhansky, Hyperbolic systems with non-diagonalisable principal part and variable multiplicities, II: microlocal analysis, J. Differential Equations 269 (2020), no. 10, 7881–7905.arxiv.org/abs/2001.04709
  6. [6]C. Garetto, B. Sabitbek, Hyperbolic systems with non-diagonalisable principal part and variable multiplicities, III: singular coefficients, Math. Ann. 390 (2024), no. 1, 1583–1613.DOI
  7. [7]C. Garetto, B. Sabitbek, C∞ well-posedness of higher order hyperbolic pseudo-differential equations with multiplicities, Acta Applicandae Mathematicae. 196, 4 (2025)DOI
  8. [9]L. Hörmander, The spectral function of an elliptic operator, Acta Math. 121 (1968), 193–218.
  9. [10]L. Hörmander, Fourier integral operators. I, Acta Math. 127 (1971), 79–183.
  10. [11]V. Ya. Ivrii, Linear hyperbolic equations, in: Yu. V. Egorov and M. A. Shubin (Eds.), Partial Differential Equations IV, Encyclopaedia of Mathematical Sciences 33, Springer-Verlag, Berlin, 1993, pp. 149–235.DOI
  11. [12]I. Kamotski and M. Ruzhansky, Regularity properties, representation of solutions, and spectral asymptotics of systems with multiplicities, Comm. Partial Differential Equations 32 (2007), no. 1, 1–35.
  12. [13]H. Kumano-go, pseudo-differential Operators, MIT Press, Cambridge, MA, 1982.
  13. [14]H. Kumano-go, K. Taniguchi, and Y. Tozaki, Multi-products of phase functions for Fourier integral operators with an application, Comm. Partial Differential Equations 3 (1978), no. 4, 349–380.
  14. [15]V. V. Kucherenko, Asymptotic behavior of the solution of the system A(x, −iℏ∂/∂x)u = 0 as h → 0 in the case of characteristics with variable multiplicity, Izv. Akad. Nauk SSSR Ser. Mat. 38 (1974), 625–662.
  15. [16]Y. Morimoto, Fundamental solution for hyperbolic equation with involutive characteristics of variable multiplicity, Comm. Partial Differential Equations 4 (1979), no. 6, 609–643.DOI
  16. [17]T. Nishitani, Hyperbolic Systems with Analytic Coefficients: Well-posedness of the Cauchy Problem, Lecture Notes in Mathematics, vol. 2097, Springer, Cham, 2014.
  17. [18]G. V. Rozenblum, Spectral asymptotic behavior of elliptic systems, Zap. Nauchn. Sem. Leningrad. Otdel. Mat. Inst. Steklov. 96 (1980), 255–271.
  18. [19]A. Seeger, C. D. Sogge, and E. M. Stein, Regularity properties of Fourier integral operators, Ann. of Math. (2) 134 (1991), no. 2, 231–251.DOI
  19. [21]M. E. Taylor, pseudo-differential Operators, Princeton University Press, Princeton, NJ, 1981.

Paper details

vlnc:W_zjs7zbu4
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Claude (Anthropic)

View statement
Looking for related papers…
Contents