[Paper Review] Maximum principles for boundary-degenerate linear parabolic differential operators

Paul M. N. Feehan|arXiv (Cornell University)|Jun 20, 2013

Advanced Mathematical Modeling in Engineering39 references3 citations

TL;DR

This paper establishes weak and strong maximum principles for boundary-degenerate linear parabolic partial differential operators, where the diffusion coefficient vanishes on a portion of the parabolic boundary. It shows that uniqueness and comparison principles hold using only the non-degenerate boundary portion, enabling sharper a priori estimates and solutions with improved regularity, particularly relevant for stochastic volatility models in mathematical finance.

ABSTRACT

We develop weak and strong maximum principles for boundary-degenerate, linear, parabolic, second-order partial differential operators, $Lu := -u_t- r(aD^2u)-\langle b, Du angle + cu$, with \emph{partial} Dirichlet boundary conditions. The coefficient, $a(t,x)$, is assumed to vanish along a non-empty open subset, $\mydirac_0!\sQ$, called the \emph{degenerate boundary portion}, of the parabolic boundary, $\mydirac!\sQ$, of the domain $\sQ\subset\RR^{d+1}$, while $a(t,x)$ may be non-zero at points in the \emph{non-degenerate boundary portion}, $\mydirac_1!\sQ := \mydirac!\sQ\less\bar{\mydirac_0!\sQ}$. Points in $\mydirac_0!\sQ$ play the same role as those in the interior of the domain, $\sQ$, and only the non-degenerate boundary portion, $\mydirac_1!\sQ$, is required for boundary comparisons. We also develop comparison principles and a priori maximum principle estimates for solutions to boundary value and obstacle problems defined by boundary-degenerate parabolic operators, again where only the non-degenerate boundary portion, $\mydirac_1!\sQ$, is required for boundary comparisons. Our results complement those in our previous articles [arXiv1204.6613, arXiv:1305.5098].

Motivation & Objective

• To develop weak and strong maximum principles for boundary-degenerate linear parabolic operators where the diffusion coefficient vanishes on a subset of the parabolic boundary.
• To show that uniqueness and comparison results for solutions depend only on the non-degenerate boundary portion, not the full boundary.
• To provide a priori estimates and comparison principles for boundary value and obstacle problems under degeneracy.
• To reconcile classical Fichera theory with modern function space frameworks used in stochastic processes and mathematical finance.
• To demonstrate that solutions can be continuous up to the degenerate boundary without requiring additional Dirichlet conditions there, improving on classical results.

Proposed method

• Define the parabolic operator $ Lu = -u_t - r(aD^2u) - ra{b, Du} + cu $, where $ a $ vanishes on a non-empty open subset $ ot{ar{ abla}}_0 cal Q $ of the parabolic boundary.
• Distinguish between the degenerate boundary portion $ ot{ar{ abla}}_0 cal Q $ and the non-degenerate portion $ ot{ar{ abla}}_1 cal Q = ot{ar{ abla}} cal Q \setminus \overline{\not{\bar{\nabla}}_0\ncal Q} $, with only $ \not{\bar{\nabla}}_1\ncal Q $ required for boundary comparisons.
• Apply the weak maximum principle property to derive a priori estimates for solutions in $ C^2_s $ and $ C^2 $ function spaces.
• Use viscosity solution techniques and Fichera-type sign conditions on the Fichera function $ \mathfrak{b} $ to ensure uniqueness without imposing Dirichlet data on the degenerate boundary.
• Establish comparison principles and maximum principle estimates for obstacle problems, again relying only on the non-degenerate boundary portion.
• Demonstrate that the classical Fichera framework imposes stronger regularity assumptions than necessary, and show that modern function space paradigms yield stronger uniqueness results.

Experimental results

Research questions

• RQ1 Can maximum principles be established for boundary-degenerate parabolic operators when Dirichlet data is prescribed only on the non-degenerate boundary portion?
• RQ2 How does the Fichera sign condition on the boundary influence uniqueness and comparison principles in the presence of degeneracy?
• RQ3 What is the role of the degenerate boundary portion in the maximum principle, and can it be treated analogously to the interior of the domain?
• RQ4 How do modern function space frameworks (e.g., $ C^{2+eta}_s $) improve upon classical Fichera theory in the context of degenerate PDEs?
• RQ5 To what extent can a priori estimates and comparison principles be derived for obstacle problems involving boundary-degenerate operators?

Key findings

• The weak maximum principle holds for $ L $-subharmonic functions in $ C^2_s $ on both bounded and unbounded domains, with the maximum attained on the non-degenerate boundary portion $ \not{\bar{\nabla}}_1\ncal Q $.
• A priori estimates for solutions to boundary value problems are derived using only the non-degenerate boundary data, improving upon classical results that require data on the full boundary.
• For the parabolic Heston operator, the first boundary value problem in the Fichera framework differs from modern formulations: Dirichlet data on the degenerate boundary is not required for uniqueness when $ 0 < \beta < 1 $.
• Solutions to obstacle problems can be shown to satisfy maximum principles and comparison estimates using only the non-degenerate boundary portion, even when the operator is degenerate on part of the boundary.
• The classical Fichera maximum principle imposes stronger regularity assumptions (e.g., $ C^2 $ up to the degenerate boundary) than necessary; modern frameworks allow continuity up to the degenerate boundary without requiring smoothness.
• The uniqueness result in this paper is stronger than that of Fichera’s theorem when $ 0 < \beta < 1 $, as it requires $ g = 0 $ only on $ \not{\bar{\nabla}}_1\ncal Q $, not on the entire boundary $ \not{\bar{\nabla}}\ncal Q $.