Inf-Sup Stability for Stokes

• Inf-Sup Stability for the Stokes equations is a key property that ensures well-posedness and numerical robustness of mixed variational formulations in incompressible fluid mechanics.
• Finite element pairs, such as MINI and Hood–Taylor, employ techniques including the use of Fortin operators to maintain a discrete inf-sup condition and guarantee convergence and accurate pressure approximations.
• Stabilization methods like grad-div and pressure-jump terms are critical in mitigating spurious pressure modes and improving error estimates, especially in regimes with small viscosity.

The inf-sup (Ladyzhenskaya–Babuška–Brezzi, LBB) stability condition for the Stokes equations is the central structural property that guarantees well-posedness and numerical robustness of mixed variational formulations in incompressible fluid mechanics. The constancy and behavior of the inf-sup constant, in both continuous and discrete settings, underpin the convergence and error estimates for velocity–pressure approximations, influence practical selection of finite element pairs, and determine the necessity and form of stabilization for diverse discretization strategies.

1. Continuous Formulation and the Inf-Sup Condition

Let $\Omega\subset\mathbb{R}^d$ $(d=2,3)$ be a bounded Lipschitz domain. The canonical stationary Stokes problem is: find $(u,p)\in V\times Q$, with

$$V := H_0^1(\Omega)^d, \qquad Q := L_0^2(\Omega) = \left\{ q\in L^2(\Omega): (q,1)_\Omega=0 \right\},$$

such that for all $(v,q) \in V \times Q$,

$a(u,v) + b(v,p) = (f,v), \qquad b(u,q) = 0,$

where $a(u,v) = \nu (\nabla u, \nabla v)_\Omega$, $b(v,p) = -(\nabla\cdot v, p)_\Omega$, and $\nu>0$ is the viscosity (Frutos et al., 2016).

The continuous inf-sup (LBB) condition asserts the existence of $\beta_0>0$—independent of the data—such that

$(d=2,3)$0

where $(d=2,3)$1, $(d=2,3)$2. This property, together with coercivity of $(d=2,3)$3 on $(d=2,3)$4, yields unique solvability and $(d=2,3)$5-stability of the pressure.

2. Discrete Inf-Sup Stability: Finite Element Frameworks

Let $(d=2,3)$6 be finite element subspaces, typically based on piecewise polynomial discretizations on a family of shape-regular, quasi-uniform meshes $(d=2,3)$7 of $(d=2,3)$8. The discrete inf-sup condition is

$(d=2,3)$9

with $(u,p)\in V\times Q$0 bounded below independently of $(u,p)\in V\times Q$1, and ideally, of other parameters (e.g., viscosity). For optimal numerical performance and error control, pairs such as MINI ($(u,p)\in V\times Q$2/$(u,p)\in V\times Q$3), quadratic-linear Hood–Taylor ($(u,p)\in V\times Q$4), and similar, are established as inf-sup stable on standard meshes (Frutos et al., 2016).

The proof of discrete inf-sup typically proceeds via construction of a Fortin operator $(u,p)\in V\times Q$5 such that $(u,p)\in V\times Q$6 for all $(u,p)\in V\times Q$7 and $(u,p)\in V\times Q$8. This allows transfer of the continuous inf-sup to the discrete setting (Frutos et al., 2016).

3. Impact of Inf-Sup Stability on Error Estimates and Robustness

The discrete inf-sup constant $(u,p)\in V\times Q$9 enters all pressure error/control inequalities. For grad-div stabilized Galerkin methods, error bounds such as

$$V := H_0^1(\Omega)^d, \qquad Q := L_0^2(\Omega) = \left\{ q\in L^2(\Omega): (q,1)_\Omega=0 \right\},$$0

hold with constants independent of the viscosity $$V := H_0^1(\Omega)^d, \qquad Q := L_0^2(\Omega) = \left\{ q\in L^2(\Omega): (q,1)_\Omega=0 \right\},$$1 for inf-sup stable pairs (e.g., $$V := H_0^1(\Omega)^d, \qquad Q := L_0^2(\Omega) = \left\{ q\in L^2(\Omega): (q,1)_\Omega=0 \right\},$$2–$$V := H_0^1(\Omega)^d, \qquad Q := L_0^2(\Omega) = \left\{ q\in L^2(\Omega): (q,1)_\Omega=0 \right\},$$3) (Frutos et al., 2016). The inf-sup constant specifically bounds the pressure error via estimates such as

$$V := H_0^1(\Omega)^d, \qquad Q := L_0^2(\Omega) = \left\{ q\in L^2(\Omega): (q,1)_\Omega=0 \right\},$$4

making small inf-sup constants detrimental to pressure accuracy. The uniformity of $$V := H_0^1(\Omega)^d, \qquad Q := L_0^2(\Omega) = \left\{ q\in L^2(\Omega): (q,1)_\Omega=0 \right\},$$5 is thus critical for robust performance, particularly in regimes of small viscosity.

4. Model Examples and Practical Recommendations

Well-established inf-sup stable FE pairs on quasi-uniform and shape-regular meshes include:

Pair	Velocity Space	Pressure Space	Stability Proof
MINI	$$V := H_0^1(\Omega)^d, \qquad Q := L_0^2(\Omega) = \left\{ q\in L^2(\Omega): (q,1)_\Omega=0 \right\},$$6	$$V := H_0^1(\Omega)^d, \qquad Q := L_0^2(\Omega) = \left\{ q\in L^2(\Omega): (q,1)_\Omega=0 \right\},$$7	Fortin operator
Hood–Taylor	$$V := H_0^1(\Omega)^d, \qquad Q := L_0^2(\Omega) = \left\{ q\in L^2(\Omega): (q,1)_\Omega=0 \right\},$$8	$$V := H_0^1(\Omega)^d, \qquad Q := L_0^2(\Omega) = \left\{ q\in L^2(\Omega): (q,1)_\Omega=0 \right\},$$9	H(div)-lifting, Fortin
Crouzeix–Raviart (odd $(v,q) \in V \times Q$0)	$(v,q) \in V \times Q$1	$(v,q) \in V \times Q$2	Macro-element technique
Trace-FEM P2–P1	$(v,q) \in V \times Q$3	$(v,q) \in V \times Q$4	Fortin function, stabilization (Olshanskii et al., 2019)

For equal-order pairs or unfitted/cut meshes, stabilization (grad-div, ghost-penalties, pressure-jump terms) is mandatory to recover inf-sup stability (Frutos et al., 2016, Johansson et al., 2015, Olshanskii et al., 2019). For grad-div, choose $(v,q) \in V \times Q$5 for inf-sup-stable (e.g., $(v,q) \in V \times Q$6–$(v,q) \in V \times Q$7) and $(v,q) \in V \times Q$8 for equal-order (e.g., MINI), to balance error terms (Frutos et al., 2016).

5. CutFEM, Trace-FEM, and Extended Scenarios

High-order cut finite element (CutFEM) methods for the Stokes problem on composite or unfitted meshes rely critically on local (interior patch) inf-sup stability, Nitsche terms for interface enforcement, and elementwise stabilization to maintain a positive inf-sup constant $(v,q) \in V \times Q$9 even under arbitrary interface geometry (Johansson et al., 2015). The analysis decomposes the pressure space into constant and nonconstant modes across subdomains and leverages patchwise inf-sup for those with interface stabilization (ghost-penalties, least-squares) to eliminate small cut-cell instabilities.

For surface PDEs, such as the Stokes system posed on smooth surfaces, trace FE methods enable discretization using bulk element traces. Inf-sup stability is achieved via normal derivative stabilization in a suitably defined mesh-dependent norm, ensuring a uniform lower bound on the discrete inf-sup constant with respect to both mesh size $a(u,v) + b(v,p) = (f,v), \qquad b(u,q) = 0,$0 and cut geometry (Olshanskii et al., 2019).

6. Role in Stabilization and Advanced Discretizations

Grad-div stabilization, pressure-jump stabilization, and bubble-enriched schemes represent mechanisms to enforce or enhance inf-sup stability for pairs which otherwise may be unstable (e.g., equal-order $a(u,v) + b(v,p) = (f,v), \qquad b(u,q) = 0,$1–$a(u,v) + b(v,p) = (f,v), \qquad b(u,q) = 0,$2). For example, grad-div stabilization augments the variational problem with the term $a(u,v) + b(v,p) = (f,v), \qquad b(u,q) = 0,$3, improving mass-conservation and guaranteeing error bounds that do not degenerate as $a(u,v) + b(v,p) = (f,v), \qquad b(u,q) = 0,$4. However, stabilization cannot substitute discrete inf-sup failure: stabilization parameters and terms must be tuned to interact with the underlying pairing (Frutos et al., 2016, Massing et al., 2012).

For high-order and nonconforming elements, macro-element and patchwise analysis (e.g., for Crouzeix–Raviart pairs of arbitrary odd degree, or Taylor–Hood on anisotropic meshes) is essential to guarantee $a(u,v) + b(v,p) = (f,v), \qquad b(u,q) = 0,$5 uniformity, exploiting local divergence-invertibility through explicit basis functions or Vandermonde arguments (Carstensen et al., 2021, Barrenechea et al., 2017). For generalized meshes (polytopic, hybridized, or barycentric-refined), specialized patchwise inf-sup proofs are employed, often relying on decompositions of the pressure space and explicit lifting operators (Guzman et al., 2017, Antonietti et al., 2020).

7. Theoretical and Practical Consequences

The inf-sup constant $a(u,v) + b(v,p) = (f,v), \qquad b(u,q) = 0,$6 governs three essential properties:

• Stability: Well-posedness of the discretized mixed problem.
• Error bounds: Appears reciprocally in pressure-error estimates, pressure-robustness, and in the uniformity of velocity error bounds especially as parameters (e.g., viscosity) vary.
• Robustness and flexibility: Uniform inf-sup stability enables the use of highly anisotropic/refined meshes, unfitted/cut methods, higher order, or nonstandard finite element spaces, with guaranteed numerical reliability.

A breakdown in inf-sup stability leads to spurious pressure modes, numerical locking, and/or loss of convergence. Consequently, satisfying (and quantifying) the discrete inf-sup condition is fundamental to both theoretical analysis and the engineering practice of Stokes and Navier–Stokes finite element simulation.

---

• (Frutos et al., 2016) Analysis of the grad-div stabilization for the time-dependent Navier-Stokes equations with inf-sup stable finite elements
• (Johansson et al., 2015) High Order Cut Finite Element Methods for the Stokes Problem
• (Carstensen et al., 2021) Crouzeix-Raviart triangular elements are inf-sup stable
• (Barrenechea et al., 2017) The inf-sup stability of the lowest order Taylor-Hood pair on Anisotropic Meshes
• (Olshanskii et al., 2019) Inf-sup stability of the trace P2-P1 Taylor-Hood elements for surface PDEs