Limit-Viability Criterion Overview

• Limit-viability criterion is a set of explicit inequalities and solvability obstacles that delineate the boundary between persistence and extinction in systems approaching a critical parameter limit.
• It applies across diverse fields such as population dynamics, portfolio optimization, and network percolation, providing a framework for analyzing phase transitions and viable solution spaces.
• By identifying critical thresholds through eigenvalue sign changes and optimality conditions, the criterion enables precise predictions of system viability under parametric perturbations.

The limit-viability criterion is a family of sharp necessary-and-sufficient conditions, typically in the form of inequalities or solvability obstacles, that delineate the boundary between persistence and extinction, successful solution and obstruction, or global viability and failure in the singular or limiting regime of a parametric system ("as a parameter approaches its limiting value"). This criterion appears in a wide range of mathematical, physical, biological, and algorithmic contexts where feasibility, persistence, or solvability is determined by behavior at or near singular limits.

1. Mathematical Formalism and Prototypical Examples

A limit-viability criterion arises in singularly perturbed systems, large-population limits, or infinite parameter regimes, where the system exhibits a bifurcation between viable and non-viable behavior. The canonical example is the criticality threshold for persistence in parabolic Lotka-Volterra equations with rare mutations, as formulated in "Survival criterion for a population subject to selection and mutations" (Costa et al., 2020). There, one considers the family of PDEs

$$\partial_t n_\varepsilon(t,x) - \varepsilon\, \Delta_x n_\varepsilon = \tfrac{1}{\varepsilon} n_\varepsilon R(x, I_\varepsilon(t)), \quad I_\varepsilon(t) = \int \psi(x)\, n_\varepsilon(t,x)\, dx,$$

with $\varepsilon \to 0$. The criterion is that the asymptotic fate (survival/extinction) is dictated by the sign of the per-capita growth function $R(x,0)$ on the zero-level set of the effective initial condition.

For semimartingale portfolio models, weak and local market viability are characterized by the No-Unbounded-Profit-with-Bounded-Risk (NUPBR) property and the existence of optimal portfolios under a limiting family of measures converging to the real-world measure (Choulli et al., 2012). In stochastic evolutionary models with random fitness, the fixation probability formula for a rare mutant converges to a limit governed by the difference in mean minus half-variance of scaled viability coefficients (Kroumi et al., 2023).

A general pattern emerges: limit-viability criteria provide explicit algebraic, analytical, or variational conditions separating sustainable/persistent solutions from those that collapse, in a precise limiting regime. They serve as "phase boundaries" for persistence versus extinction, solvability versus obstruction, or global viability versus collapse.

2. Derivation in Reaction-Diffusion and Population Models

In trait-structured population dynamics with small mutation rates, the asymptotic analysis of

$$\partial_t n_\varepsilon - \varepsilon \Delta_x n_\varepsilon = \frac{1}{\varepsilon} n_\varepsilon R(x, I_\varepsilon)$$

leads, under the Hopf–Cole transform $n_\varepsilon = \exp(u_\varepsilon/\varepsilon)$, to a constrained Hamilton–Jacobi equation in the limit $\varepsilon\to 0$: $$\partial_t u = |\nabla u|^2 + R(x, I(t)), \quad \max_x u(t,x) = 0,$$ where the support of the emergent population concentrates on the zero-level set $\Gamma_t = \{x : u(t,x)=0\}$. The survival/extinction trichotomy is as follows (Costa et al., 2020):

• Persistence: If the intersection of the support of the initial data with $\{x: R(x,0)>0\}$ is nonempty, the population persists globally in time.
• Extinction on an Interval: If $\Gamma_0 \subset \{x: R(x,0) \le -C\}$ for some $C>0$, then extinction occurs uniformly up to some finite $T_0$.
• Punctual (Instantaneous) Extinction: If $\Gamma_0 \subset \{x: R(x,0) \le 0\}$ but the inclusion is not strict, extinction occurs at isolated times.

A necessary-and-sufficient criterion is thus: Survival if and only if $R(x,0)>0$ somewhere on the initial zero-level set $\Gamma_0$. This generalizes using principal eigenvalues in linearizations: $$\lambda^* < 0 \;\Leftrightarrow\; \text{persistence}, \qquad \lambda^* > 0 \;\Leftrightarrow\; \text{extinction}.$$

3. Limit-Viability in Evolutionary Stochastic Processes

For structured populations in the island model with random per-generation viability coefficients $S_A, S_B$, scaling as $1/D$, the frequency process becomes, in the limit of large number of demes $D$, a diffusion with generator

$$\mathcal{L}f(x) = M(x) f'(x) + \tfrac{1}{2} V(x) f''(x),$$

where the drift $M(x)$ and variance $V(x)$ depend on the scaled means $\mu_A, \mu_B$ and variances $\sigma_A^2, \sigma_B^2$. In the strong-structure, low-dispersal limit,

$$M(x) \to x(1-x)\left[(\mu_A - \tfrac{1}{2}\sigma_A^2) - (\mu_B - \tfrac{1}{2} \sigma_B^2)\right],$$

and type $A$ has larger fixation probability than $B$ if and only if

$$\mu_A - \tfrac{1}{2} \sigma_A^2 > \mu_B - \tfrac{1}{2} \sigma_B^2.$$

This is the stochastic limit-viability criterion for evolutionary success under environmental heterogeneity (Kroumi et al., 2023).

4. Limit-Viability in Network Percolation

In multiplex networks with link overlap, mutual percolation (the existence of a giant mutually connected component) is the $\rho \to 0$ limit of a more general node-viability problem. The limit-viability criterion is a discontinuity (hybrid transition) in the fraction $M$ of the network in the GMC, given by the simultaneous solution of

$$\begin{aligned} M^* &= R_\infty + \sum_{m=1}^\infty R(m) \left[1 - e^{-m z M^*}\right]^2, \ 1 &= \sum_{m=1}^\infty 2 R(m) (m z) e^{-m z M^*} \left[1 - e^{-m z M^*}\right], \end{aligned}$$

where $R(m)$, $R_\infty$ encode the overlap-cluster statistics (Min et al., 2014). Overlap lowers the mutual percolation threshold and expands the hysteresis region, exemplifying how structural features modify the locus of viability thresholds.

5. Limit-Viability in Optimization and Evolutionary Search

In Memetic Viability Evolution (mViE), "viability boundaries" $b_j(t)$ for each constraint $g_j(x)\leq 0$ and $b_0(t)$ for the objective are adaptively tightened throughout the search: $$b_j^{(t+1)} = \max\left(0, \min\{ b_j^{(t)},\, g_j(y^{(t)}) + \tfrac{1}{2}(b_j^{(t)} - g_j(y^{(t)})) \} \right),$$ where $y^{(t)}$ is a successful offspring at iteration $t$ (Maesani et al., 2018). Nonviable solutions are discarded at each step, and the search space "contracts" toward the true feasibility region as $t\to\infty$—the limit where all $b_j(t)\to 0$ defines the feasible set's viability boundary.

6. Limit-Viability in Constrained PDEs and Geometry

In the conformal method for the Einstein constraints on asymptotically cylindrical manifolds, the existence of a solution to the coupled Lichnerowicz–Choquet‐Bruhat–York system is equivalent to the nonexistence of a nontrivial solution to a singular ("limit") elliptic equation: $$\operatorname{div}_g L W = \alpha_0 \sqrt{\tfrac{n-1}{n}} |L W| \tfrac{d \tau}{\tau}$$ for some $\alpha_0\in(0,1]$ (Dilts et al., 2014). If the limit equation admits no solution with prescribed decay, the original constraint system is solvable; otherwise, it is obstructed.

7. Limit-Viability in Cosmology, QFT, and Other Domains

• In low-energy bigravity, the region in parameter space where both Vainshtein screening and viable cosmological evolution to early times are possible is determined by inequalities encoding the compatibility of the required non-linear scales: the breakdown of these inequalities signals the failure of cosmological viability—the model cannot simultaneously fit late- and early-time behavior (Kenna-Allison et al., 2018).
• In neural-network quantum field theory surrogates, the validity of the perturbative expansion in $1/N$ for a width-$N$ network is only ensured if both coupling–UV length and network size satisfy

$$\lambda\,\Lambda^{d-2}\xi^{d-2} \ll 1, \quad N \gg \max\{[\Lambda^{d-2} \xi^{d-2}]^n,\, \Lambda^d\xi^d\}$$

for target order $n$ in coupling $\lambda$ (Sen et al., 5 Aug 2025). Any violation prevents accurate extraction of continuum field-theoretic results at finite $N$.

Table: Representative Limit-Viability Criteria

Domain	Criterion / Condition	Reference
Trait-structured population PDEs	$R(x,0)>0$ on support $\Rightarrow$ persistence; $R(x,0)\leq 0$ $\Rightarrow$ extinction	(Costa et al., 2020)
Evolutionary island model	$$\mu_A - \frac12 \sigma_A^2 > \mu_B - \frac12 \sigma_B^2$$ for fixation advantage of $A$	(Kroumi et al., 2023)
Multiplex network percolation	Simultaneous solution of equations (*) for discontinuous onset of GMC	(Min et al., 2014)
Viability evolution in optimization	Sequential contraction $b_j(t)\to 0$ for all $j$ (feasibility as $t\to\infty$)	(Maesani et al., 2018)
Portfolio optimization/finance	Existence of measures $Q(\varepsilon)\to P$ with optimizers under NUPBR	(Choulli et al., 2012)
Infinite-width neural network QFT	$$\lambda\,\Lambda^{d-2}\xi^{d-2}\ll 1;\; N\gg [\Lambda^{d-2}\xi^{d-2}]^n$$	(Sen et al., 5 Aug 2025)
Asymptotically cylindrical GR constraints	Solution/non-solution of the DGH limit equation (elliptic obstacle)	(Dilts et al., 2014)
Bigravity cosmology	Parameters satisfying all inequalities in Section 4 (screening + attractor + early branch)	(Kenna-Allison et al., 2018)

8. Interpretive Remarks and Implications

Limit-viability criteria serve as precise analytic boundaries for persistence, robustness, and feasibility in the limiting regimes of complex models. They are essential in

• predicting viability under parametric perturbations,
• quantifying phase transitions in probabilistic and dynamical systems,
• diagnosing obstructions to solvability in geometric PDEs and optimization,
• and engineering robust algorithms that adapt viability limits dynamically.

Their form—explicit inequalities, saddle-node points, or eigenvalue sign changes—allows for direct computation of critical parameter domains. However, these criteria are often model-specific and require careful derivation in each context; generalizing across systems requires explicit mapping of the underlying singular limits and concentration phenomena. Theoretical advances in this area have clarified the mechanism of criticality in networked systems, evolutionary dynamics, mathematical finance, and numerical simulation of fields, providing foundational tools for both analysis and design in applied mathematics, physics, and computational optimization.