Characteristics, Constraint Propagation, and Well-Posedness

A relativistic formulation is not well posed merely because its evolution equations resemble waves, and it is not “causal” merely because a few listed speeds equal the speed of light. A defensible result must treat one fully specified gauge-fixed reduction: compute its principal symbol in every spatial direction, exhibit a uniformly controlled characteristic basis or symmetrizer, separate physical, gauge, and constraint fields, derive a closed subsidiary system, and control the boundary flux and initial-boundary compatibility. These ingredients yield an energy estimate in a declared norm. They do not by themselves prove global nonlinear existence, nor do coordinate speeds identify physical signals without a sector analysis.

Required background. Partial differential equations and Green functions supplies symbols, boundary fluxes, and energy estimates. The Einstein equation supplies the constraints and physical solution set.

Helpful background. Hypersurface normals and tangential projectors supplies initial and boundary normals. Formulations of General Relativity explains why the reduction and constraint additions must be named. Differential equations reviews uniqueness and continuous dependence.

The complete symbol of a gauge-fixed reduction

Suppose a first-order formulation has been written as

$$A^0(u,x)\partial_tu+A^i(u,x)\partial_i u=F(u,x),$$

with a specified state vector $u$, reduction constraints, lapse–shift or other gauge equations, and a positive state norm. Freeze a smooth background solution. For every real unit spatial covector $s_i$, form

$$P(s)=(A^0)^{-1}A^is_i.$$

Strong hyperbolicity requires real eigenvalues and a complete eigenvector matrix whose condition number is bounded uniformly over the unit sphere. An equivalent smooth positive symmetrizer may be used. Symmetric hyperbolicity supplies a direction-independent positive matrix that symmetrizes all principal matrices and is especially useful at a boundary. The calculation must include arbitrary directions; axis-only tests can miss a defective oblique block.

Each characteristic field should then be assigned, by its defining combination and its on-constraint behavior, to one of three sectors. Physical fields are represented by gauge-invariant curvature or radiative combinations in the stated background or asymptotic regime. Gauge fields change slicing, coordinates, frame, or gauge source without changing the geometry. Constraint fields vanish on Einstein data or measure a reduction constraint. Degeneracy of speeds does not merge these meanings. A gauge speed outside the metric null cone is not by itself superluminal physical propagation, while an uncontrolled constraint speed can still spoil the mathematical initial-boundary problem.

For second-order-in-space systems, the same questions must be answered using a declared second-order energy or a named first-order or pseudodifferential reduction. Auxiliary derivative variables create additional constraints, and adding them to the equations changes the off-constraint symbol. A proof about one reduction cannot be transferred silently to another.

From characteristic fields to a Cauchy estimate

For constant frozen coefficients, uniform diagonalization gives a Fourier estimate independent of wave number. For variable coefficients, the symmetrizer must have enough regularity for commutator terms, and lower-order coefficients must be bounded in the Sobolev class being used. The desired local estimate has the form

$$\lVert u(t)\rVert_{H^q} \le C_T\left( \lVert u(0)\rVert_{H^q} +\int_0^t\lVert F(\tau)\rVert_{H^q}\,d\tau \right),$$

where $C_T$ is independent of spatial frequency and perturbation wavelength. The regularity index, coefficient assumptions, reduction norm, and time interval are part of the theorem. A pointwise eigenvalue calculation without this uniform estimate establishes neither existence nor continuous dependence.

The harmonic Einstein example makes the distinction concrete. A wave-like principal part is supplemented by harmonic constraints and a particular boundary operator; the resulting frozen half-space problem is strongly well posed only under explicit spectral conditions on that operator. Kreiss, Reula, Sarbach, and Winicour 2007, §§2–3, Theorems 1–3, pp. 3–10, PDF derives both interior and boundary derivative estimates and then explains the local quasilinear iteration. The theorem is evidence for that specified harmonic system, not for every formulation with null characteristic roots.

Subsidiary evolution and boundary flux

Let $C$ collect the Hamiltonian, momentum, gauge, and reduction constraints relevant to the formulation. Derive their evolution from the main equations rather than assuming it:

$$\partial_t C+B^i\partial_iC=MC,$$

or the corresponding second-order system. The right-hand side must be homogeneous in $C$ for the zero-constraint solution to remain invariant. One then checks hyperbolicity of the subsidiary symbol, identifies its incoming fields, and shows that the main boundary conditions impose compatible homogeneous or controlled data on them. Constraint preservation in a boundary-free domain does not exclude constraint influx through a timelike boundary.

For a symmetric hyperbolic main system, first multiply by a symmetrizer so that $A^0$ is positive Hermitian and every $A^i$ is Hermitian; the following $A^\mu$ denote these symmetrized matrices. Choose the outward spatial conormal $\nu_i$ and define $A^\nu=A^i\nu_i$. For $E=\int_{\Sigma_t}u^\dagger A^0u\,dV$, the energy identity contains

$$\frac{dE}{dt} =-\int_{\partial\Sigma_t}u^\dagger A^\nu u\,dS +\text{bulk and forcing terms}.$$

Negative eigenvalues of $(A^0)^{-1}A^\nu$ are incoming relative to the outward normal and require boundary data. A maximally dissipative homogeneous condition chooses a subspace of the correct dimension on which the boundary contribution cannot increase the energy; inhomogeneous data add a controlled boundary norm. Friedrich and Nagy 1999, §3, equations (3.1)–(3.7), pp. 624–628 gives the general boundary-matrix construction and an explicit field example.

The incoming count alone is not sufficient. The boundary conditions must also preserve the subsidiary constraints, obey a uniform boundary-stability condition, and be compatible with the initial data at the corner. For a smooth solution, differentiating the boundary condition repeatedly in time and using the evolution equations must agree with the spatial derivatives determined by the initial data to the required order. For finite Sobolev regularity, the corresponding finite compatibility hierarchy is used.

What a complete proof establishes

A useful proof statement names the formulation and gauge, reduction order, background or neighborhood of states, coefficient regularity, norm, initial data class, boundary normal and incoming convention, boundary operator, and compatibility order. It then establishes a uniform Cauchy or initial-boundary estimate, proves existence and uniqueness in that class, and shows through the subsidiary system that reduced solutions with constraint-compatible data solve Einstein’s equations.

Friedrich and Nagy’s vacuum construction illustrates this architecture in a formulation quite different from generalized harmonic evolution. Their boundary-adapted reduced equations are symmetric hyperbolic and chosen so that a subsidiary argument recovers the full field equations; see Friedrich and Nagy 1999, §§5–6, Theorem 6.1, pp. 634–644. Their data, gauge sources, maximally dissipative curvature conditions, and all-order smooth corner consistency lead to a local existence result in Friedrich and Nagy 1999, §§7–8, Theorem 8.1, pp. 644–652. The theorem’s hypotheses and its stated qualification on general geometric uniqueness must be retained; it is not a formulation-independent assertion about arbitrary timelike boundary data.

The final independent checks are therefore direct. Reconstruct the physical, gauge, and constraint characteristic combinations in one frozen background; verify completeness and symmetrizer bounds as the direction varies; count incoming fields with the declared outward normal; compute the energy flux after imposing the proposed boundary condition; and derive the induced boundary condition for every incoming subsidiary field. Failure of any one check blocks the corresponding well-posedness claim even if the other calculations succeed.

Exercise

Consider the constant-coefficient system on the half-line $x\ge0$,

$$\partial_tu+ \begin{pmatrix}1&0\\0&-1\end{pmatrix} \partial_xu=0.$$

Identify the incoming and outgoing fields using the outward normal at $x=0$. Show that the homogeneous boundary condition $u_1=ru_2$ is maximally dissipative when $|r|\le1$, and explain what fails when $|r|\gt1$.

Solution

The domain is $x\ge0$, so its outward conormal at the finite boundary points toward decreasing $x$: $\nu_x=-1$. Hence

$$A^\nu=- \begin{pmatrix}1&0\\0&-1\end{pmatrix} =\begin{pmatrix}-1&0\\0&1\end{pmatrix}.$$

The field $u_1$ has negative outward-normal speed and enters the domain; $u_2$ leaves it. With

$$E(t)=\frac12\int_0^\infty(u_1^2+u_2^2)\,dx,$$

integration by parts gives

$$\frac{dE}{dt} =\frac12\left(u_1^2-u_2^2\right)_{x=0}.$$

Imposing $u_1=ru_2$ yields

$$\frac{dE}{dt} =\frac12(r^2-1)u_2^2\big|_{x=0}.$$

Thus $|r|\le1$ makes the homogeneous boundary flux non-increasing and prescribes exactly the one incoming field. If $|r|\gt1$, this boundary subspace permits positive flux in the displayed $H=I$ energy and is therefore not maximally dissipative for that symmetrizer. That observation alone is not an ill-posedness theorem; a different estimate or symmetrizer would have to be ruled out separately. Prescribing $u_2$ instead would constrain an outgoing field while leaving the incoming field uncontrolled, so the incoming count would be wrong even though one scalar condition had been imposed.

References

• Friedrich, Helmut, and Gabriel Nagy. “The Initial Boundary Value Problem for Einstein’s Vacuum Field Equation.” Communications in Mathematical Physics 201 (1999): 619–655. doi:10.1007/s002200050571.
• Kreiss, Heinz-Otto, Oscar Reula, Olivier Sarbach, and Jeffrey Winicour. “Well-Posed Initial-Boundary Value Problem for the Harmonic Einstein Equations Using Energy Estimates.” Classical and Quantum Gravity 24 (2007): 5973–5984. arXiv:0707.4188.