Corner Compatibility and Well-Posedness

At the corner where a spacelike initial hypersurface meets a timelike boundary, the initial and boundary prescriptions describe the same solution and cannot be chosen independently. Evaluating the boundary condition at the initial time gives the zeroth compatibility condition. Differentiating it in time and using the evolution equations to replace time derivatives gives a hierarchy. Smooth solutions require compatibility to every order; a finite Sobolev or classical regularity claim requires only the corresponding finite number of conditions.

Required background. The Gibbons–Hawking–York Boundary Term supplies initial, boundary, and corner geometry, while PDEs and Green Functions for Gravitation supplies trace regularity, hyperbolic energy estimates, and differentiated boundary operators.

Helpful background. The Einstein Equation supplies the evolution and constraint equations; Hypersurface Normals and Tangential Projectors supplies induced corner data; Formulations of General Relativity orients the volume; and Causality and Global Structure supplies the local domain of dependence.

Compatibility hierarchy for a model wave

Consider

$$u_{tt}=u_{xx}+F(u,\partial u)$$

on $x\le0$, with initial data

$$u(0,x)=f(x), \qquad u_t(0,x)=h(x),$$

and Dirichlet boundary data $u(t,0)=q(t)$. At the corner $(0,0)$,

$$q(0)=f(0)$$

is order-zero compatibility. One time derivative gives

$$q'(0)=h(0).$$

A second derivative and the evolution equation give

$$q''(0)=f''(0)+F(f,h,f')\big|_{x=0}.$$

Higher derivatives follow recursively. Each time derivative of the boundary operator is evaluated at $t=0$, and every time derivative of the unknown is replaced using the evolution equation and previously determined derivatives of the initial data. For Neumann, Sommerfeld, or coupled characteristic conditions, the same procedure applies but involves the corresponding normal and tangential traces.

Regularity-dependent theorem statement

For a specified strongly or symmetrically hyperbolic IBVP with a smooth noncharacteristic or theorem-admissible boundary, assume that the boundary conditions yield the required energy estimate. If the initial and boundary data have the stated Sobolev regularity and satisfy the compatibility conditions through the trace order required by that estimate, the local solution has the claimed regularity. If compatibility fails at order $k$, one cannot in general obtain a solution whose derivatives through the corresponding order extend continuously or possess the asserted Sobolev trace at the corner.

This is not a universal formula “$H^m$ requires exactly $m$ conditions.” The number shifts with the differential order of the system, whether a first- or second-order reduction is used, the order of the boundary operator, and the trace theorem. The operative existence theorem must state the count. Smooth $C^\infty$ data require the recursion at all orders because every differentiated boundary equation must match.

Friedrich and Nagy formulate the smooth Einstein consistency condition by matching the formal time expansions of the reduced fields and the prescribed boundary functions at every order in Friedrich and Nagy 1999, §7.4, pp. 650–651. Their local existence result assumes precisely these smooth compatibility conditions: Friedrich and Nagy 1999, Theorem 8.1, pp. 651–652.

Einstein corner data

Let $S_0$ meet $\mathcal T$ in the two-surface $C$. Compatibility includes more than equality of metric components:

• the induced two-metric computed from the initial metric must match the one implied by the boundary geometry;
• the boost angle between the future normal to $S_0$ and the outward normal to $\mathcal T$ must match the boundary threading;
• initial extrinsic curvature and gauge sources determine first time derivatives of the metric or frame at $C$;
• constraints and subsidiary boundary conditions determine incoming constraint derivatives;
• physical boundary data must match the curvature constructed from the initial geometry at the corner;
• repeated reduced equations determine all higher normal and time derivatives needed for the target regularity.

The hierarchy is formulation dependent even though the need for compatibility is geometric. A change of gauge or reduction changes which derivatives are listed as primitive data, but it cannot permit two unequal traces of the same smooth spacetime field.

Kreiss and collaborators assume smooth initial data compatible with their Sommerfeld-type conditions and obtain boundary and bulk derivative estimates for the harmonic system in Kreiss et al. 2007, Theorem 3 and Appendix 1, pp. 8–10 and 14–17, PDF. Their estimate is part of the proof; formal compatibility without a stable boundary operator would not prove well-posedness.

A controlled failure at second order

Take the homogeneous wave equation with

$$f(x)=0, \qquad h(x)=0, \qquad q(t)=t^2.$$

Then $q(0)=f(0)$ and $q'(0)=h(0)$, so compatibility holds through first order. But

$$q''(0)=2 \ne f''(0)=0.$$

If a $C^2$ solution existed at the corner, the boundary trace would give $u_{tt}(0,0)=2$, while the PDE and initial trace would give $u_{tt}(0,0)=u_{xx}(0,0)=0$, a contradiction. Weaker solutions can exist with reduced corner regularity; failure does not necessarily imply that every weak formulation has no solution. Indeed an explicit solution on $t\ge0$, $x\le0$ is

$$u(t,x)=(t+x)_+^2, \qquad y_+=\max(y,0).$$

It has the required zero initial data and $u(t,0)=t^2$. It is $C^1$ across the incoming characteristic $t+x=0$, but its second derivatives jump there. In the distributional sense $u_{tt}=u_{xx}=2\,\mathbf 1_{t+x\gt0}$, so the wave equation still holds. The lost corner regularity travels into the domain along a characteristic.

Compatibility, constraints, and well-posedness

These are three separate requirements:

1. Constraints restrict the initial and some boundary fields so that the reduced solution can satisfy Einstein’s equation.
2. Corner compatibility makes the initial and boundary traces belong to one field with the desired regularity.
3. Energy estimates give continuous dependence and uniqueness for the reduced IBVP.

Passing the first two does not establish the third. Passing the energy estimate with incompatible data gives a solution only in a weaker class, possibly with a corner singularity. Passing compatibility at a finite order does not justify a smooth theorem.

Practical construction

One reliable workflow is to choose constraint-satisfying initial data and gauge sources first, use the reduced equations to compute the boundary jet at $C$, and then construct boundary functions with that prescribed finite or infinite jet. Borel extension supplies smooth functions realizing a formal jet, but it does not solve the evolution or guarantee global boundary behavior. In numerical work, a ramp can repair compatibility if its initial derivatives are deliberately chosen to match this jet. A generic ramp chosen only to reduce the amplitude or abruptness does not establish the required conditions.

Common pitfalls

Not every theorem needs infinitely many conditions. Infinite compatibility belongs to smooth solutions; finite regularity has a finite trace requirement.

Formal compatibility is not nonlinear well-posedness. A boundary estimate and subsidiary constraint argument remain necessary.

Matching field values is only order zero. Extrinsic curvature, gauge, and radiative data enter the first and higher derivatives.

Exercise: Sommerfeld corner conditions

For $u_{tt}=u_{xx}$ on $x\le0$ with

$$(u_t+u_x)(t,0)=q(t),$$

derive the zeroth and first differentiated compatibility conditions in terms of $f=u|_{t=0}$ and $h=u_t|_{t=0}$.

Solution

At $t=0$ the boundary condition gives

$$h(0)+f'(0)=q(0).$$

Differentiate once in time:

$$u_{tt}(0,0)+u_{xt}(0,0)=q'(0).$$

Using the wave equation and the initial traces gives

$$f''(0)+h'(0)=q'(0).$$

Further conditions alternate spatial derivatives of $f$ and $h$ according to the same recursion. The signs would change if the computational domain or the chosen outgoing normal were reversed.

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.
• 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, no. 23 (2007): 5973–5984. DOI; Open PDF.