--- title: "(r,d)-Stability: A Cross-Disciplinary Overview" url: https://www.emergentmind.com/topics/r-d-stability type: topic --- # (r,d)-Stability: A Cross-Disciplinary Overview Across the supplied literature, “\((r,d)\)-stability” is not a standardized term with a single accepted definition. The exact expressions that do appear belong to several distinct theories: \(r\)-stability for closed spacelike hypersurfaces in Lorentzian geometry, \(d\)-stability and \(d\)-\(\sigma\)-stability for subsets of random metric spaces, \(\sigma\)-stability for subsets of random normed modules, \(B\)-stability in Boolean-valued settings, and \((\mathfrak D,\mathcal G,\circ)\)-stability in matrix theory. A separate line of work on exponential Riesz bases in \(\mathbb R^d\) studies explicit stability bounds and generalizations of Kadec’s theorem, but the supplied record does not contain the theorem statements or constants of that paper [1003.0338] [1904.07405] [2401.08421] [1805.05558] [1403.2341]. ## 1. Terminological status The main source of ambiguity is that the symbols \(r\), \(d\), and \(D\) index different objects in different literatures. In the Lorentzian hypersurface literature, \(r\) indexes the mean-curvature order and the associated Jacobi-type operator. In random metric spaces, \(d\) is the random metric itself, so \(d\)-\(\sigma\)-stability means stability defined through metric gluing. In matrix theory, \(D\)-stability refers to structured spectral localization under diagonal or related transformations, and the generalized notation is \((\mathfrak D,\mathcal G,\circ)\)-stability rather than \((r,d)\)-stability. In the exponential-basis literature, \(d\) is the ambient Euclidean dimension, while the natural second parameter is a perturbation radius or admissible bound rather than an established symbol \(r\) [1003.0338] [1904.07405] [2401.08421] [1805.05558] [1403.2341]. | Literature | Exact notion used | Relation to “\((r,d)\)-stability” | |---|---|---| | Lorentzian hypersurfaces | \(r\)-stability | Exact \(d\) parameter absent [1003.0338] | | Random metric spaces | \(d\)-stability, \(d\)-\(\sigma\)-stability | Closest exact use of a metric-indexed \(d\) [1904.07405] | | Random normed modules and Boolean sets | \(\sigma\)-stability, \(B\)-stability, \(d\)-decomposability | Comparison and unification results, but no exact “\((r,d)\)” label [2401.08421] | | Matrix theory | \((\mathfrak D,\mathcal G,\circ)\)-stability | General parametric stability template, not the queried notation [1805.05558] | | Exponential bases | stability bounds in dimension \(d\) | A plausible radius–dimension reading, but the supplied text lacks exact results [1403.2341] | A common misconception is therefore to treat “\((r,d)\)-stability” as a single established technical notion. The supplied arXiv literature supports a narrower conclusion: it is best understood as a terminological crossroads connecting several non-equivalent stability formalisms. ## 2. \(r\)-stability in Lorentzian geometry In the geometric literature represented here, \(r\)-stability is a variational notion for closed spacelike hypersurfaces in a time-oriented Lorentz manifold of constant sectional curvature \(c\). The ambient manifold is assumed to be conformally stationary, and the main theorem additionally assumes the existence of a closed conformal vector field \(V\) and a Killing vector field \(W\). The hypersurface \(x:M^n\to \overline M_c^{\,n+1}\) is closed and spacelike, and the stability problem is formulated for hypersurfaces with constant \((r+1)\)-th mean curvature [1003.0338]. The definition is exact: \(x\) is \(r\)-stable if \[ \mathcal A_r''(0)\le 0 \] for all volume-preserving variations. Equivalently, if \[ \mathcal G=\left\{f\in C^\infty(M): \int_M f\,dM=0\right\}, \] then \(x\) is \(r\)-stable iff \[ \mathcal J_r''(0)\le 0 \text{ for all } f\in\mathcal G. \] The second variation is governed by the operator \[ L_r(f)=\operatorname{tr}(P_r\,\operatorname{Hess}f), \] where \(P_r\) is the \(r\)-th Newton transformation. In ambient constant curvature, \[ L_r(f)=\operatorname{div}(P_r\nabla f). \] The central characterization theorem states that, under the hypotheses of the paper, if \[ \lambda = c(n-r)\binom{n}{r}H_r - nH_1\binom{n}{r+1}H_{r+1} + (r+2)\binom{n}{r+2}H_{r+2} \] is constant and \(\operatorname{Div}V\neq 0\) on \(M\), then \[ x \text{ is \(r\)-stable } \iff \lambda \text{ is the first eigenvalue of } L_r. \] Thus \(r\)-stability is a spectral property of the Jacobi operator naturally attached to the \(r\)-th mean curvature, not a metric concatenation property and not a two-parameter \((r,d)\)-theory in the formal sense [1003.0338]. No separate symbol \(d\) occurs in this formalism. This suggests that, if the notation “\((r,d)\)-stability” is imported into this literature, the only natural meaning would be “\(r\)-stability in dimension \(d\),” where \(d\) is interpreted as either the hypersurface dimension \(n\) or the ambient dimension \(n+1\). That interpretation is plausible, but it is not the paper’s own terminology. ## 3. \(d\)-\(\sigma\)-stability in random metric spaces In random metric geometry, the exact relevant term is \(d\)-\(\sigma\)-stability. A random metric space \((E,d)\) over \((\Omega,\mathcal F,P)\) has \[ d:E\times E\to L^0_+(\mathcal F), \] so the metric takes values in nonnegative random variables rather than in \(\mathbb R_+\). For a nonempty subset \(G\subset E\), \(d\)-stability and \(d\)-\(\sigma\)-stability are measurable pasting properties defined through the random distance itself [1904.07405]. The binary notion is \(d\)-stability: for any \(x_1,x_2\in G\) and any \(A\in\mathcal F\), there exists \(x\in G\) such that \[ I_A\, d(x,x_1)=0 \quad\text{and}\quad I_{A^c}\, d(x,x_2)=0. \] The countable notion is \(d\)-\(\sigma\)-stability: for each sequence \(\{x_n:n\in\mathbb N\}\subset G\) and each countable partition \(\{A_n:n\in\mathbb N\}\) of \(\Omega\), there exists \(x\in G\) such that \[ I_{A_n}\, d(x,x_n)=0,\qquad \forall n\in\mathbb N. \] The pasted point is unique, so one writes \[ x=\sum_{n\ge 1} I_{A_n}x_n, \] now meaning metric agreement on each \(A_n\). This is the metric analogue of the older algebraic \(\sigma\)-stability from \(L^0\)-module theory. In a compatible \(L^0(\mathcal F,K)\)-module, if \[ d(I_Ax,I_Ay)=I_A\, d(x,y), \] then \[ G \text{ is \(d\)-\(\sigma\)-stable } \iff G \text{ is \(\sigma\)-stable}. \] In particular, in an RN module with \[ d(x,y)=\|x-y\|, \] metric patching and algebraic patching coincide. This is the basic reason \(d\)-\(\sigma\)-stability functions as the correct extension of countable concatenation from random normed modules to arbitrary random metric spaces [1904.07405]. A central identity is that concatenation is respected by the metric: \[ d\!\left(\sum I_{A_n}x_n,\sum I_{A_n}y_n\right)=\sum_{n\ge 1} I_{A_n}d(x_n,y_n). \] That identity makes the theory operational in variational principles, fixed point theorems, and measurable selection arguments. ## 4. Comparison theorems and structural equivalences The comparison paper on stability notions shows that \(d\)-\(\sigma\)-stability is not an isolated construction but part of a larger web of equivalences. Any random metric space \((S,d)\) can be isometrically embedded into a \(\mathcal T_{\varepsilon,\lambda}\)-complete RN module \(\mathcal U_b(S,L^0(\mathcal F))\) by fixing \(p_0\in S\) and defining \[ T(p)=d(p,\cdot)-d(p_0,\cdot). \] This embedding satisfies \[ \|T(p)-T(q)\|=d(p,q),\qquad \forall p,q\in S. \] For any nonempty \(G\subset S\), \[ G \text{ is } d\text{-}\sigma\text{-stable } (d\text{-stable}) \iff T(G)\text{ is }\sigma\text{-stable }(\text{stable}). \] Accordingly, \(d\)-\(\sigma\)-stability in random metric spaces is explicitly proved to be a special case of module-theoretic \(\sigma\)-stability [2401.08421]. The same paper develops the completeness side of the theory. If \(G\) is a \(d\)-stable subset of a random metric space, then \[ G \text{ is } (\varepsilon,\lambda)\text{-complete} \iff G \text{ is both } d\text{-}\sigma\text{-stable and } L^0\text{-complete}. \] For a \(d\)-\(\sigma\)-stable random metric space \((S,d)\), the following are equivalent: 1. \((S,d)\) is stably compact. 2. \((S,d)\) is stably sequentially compact. 3. \((S,d)\) is random totally bounded and \(L^0\)-complete. 4. \((S,d)\) is random sequentially compact. In the allied paper devoted to random metric spaces, one also has \[ E \text{ is } (\varepsilon,\lambda)\text{-complete} \iff E \text{ is } L^0\text{-complete} \] whenever \((E,d)\) is \(d\)-\(\sigma\)-stable, and on an \((\varepsilon,\lambda)\)-complete RM space, \[ E \text{ is \(d\)-stable } \iff E \text{ is \(d\)-\(\sigma\)-stable}. \] This yields the recurring principle stated there as \[ \text{binary pasting} + \text{completeness} \Rightarrow \text{countable pasting} \] [2401.08421] [1904.07405]. Further unifications are module-theoretic and Boolean-valued. In an order complete RN space, \[ \text{\(d\)-decomposable} \iff \text{\(d\)-stable}, \] and \[ \|\cdot\| \text{ is } d\text{-decomposable} \iff S \text{ is } d\text{-}\sigma\text{-stable}. \] For \(L^p(\mathcal F)\)-normed \(L^\infty\)-modules, the gluing property is derived from the \(\sigma\)-stability of a generating complete RN module. At the Boolean level, a nonempty subset of a Boolean set is universally complete iff it is \(B\)-stable. These results make stability a general theory of measurable or Boolean gluing rather than a single property tied to one category of spaces [2401.08421]. ## 5. Structured spectral stability and generalized \(D\)-stability A different use of parametric stability arises in matrix theory. The relevant general concept is \((\mathfrak D,\mathcal G,\circ)\)-stability: if \(\mathfrak D\subset\mathbb C\) is a stability region, \(\mathcal G\subset \mathcal M^{n\times n}\) is a matrix class, and \(\circ\) is a binary operation, then an \(n\times n\) matrix \(A\) is left \((\mathfrak D,\mathcal G,\circ)\)-stable if \[ \sigma(\mathbf G \circ \mathbf A)\subset \mathfrak D \quad\text{for every } \mathbf G\in \mathcal G. \] The right version requires \[ \sigma(\mathbf A \circ \mathbf G)\subset \mathfrak D \quad\text{for every } \mathbf G\in \mathcal G. \] This framework unifies multiplicative and additive \(D\)-stability, Schur \(D\)-stability, \(H\)-stability, \(D\)-hyperbolicity, \(D\)-positive, block \(D(\alpha)\)-stability, \(H(\alpha)\)-stability, and Hadamard-product variants [1805.05558]. Several classical examples are obtained by choosing \(\mathfrak D\), \(\mathcal G\), and \(\circ\). Multiplicative \(D\)-stability corresponds to \[ \mathfrak D = \{ z\in\mathbb C : \Re z > 0\},\qquad \mathcal G=\mathcal D^+,\qquad \circ=\text{matrix multiplication}, \] so \[ \sigma(DA)\subset \{z\in\mathbb C:\Re z>0\}\quad \forall D\in\mathcal D^+. \] Additive \(D\)-stability replaces multiplication by addition, while Schur \(D\)-stability uses the unit disk \[ \mathfrak D = \{ z\in\mathbb C : |z|<1\}. \] The paper also gives general inclusion, transpose, inversion, scalar-multiplication, and similarity results, together with Lyapunov-type sufficient conditions such as the implication from diagonal stability to multiplicative \(D\)-stability [1805.05558]. This matrix-theoretic theory does not explicitly define “\((r,d)\)-stability.” A plausible implication is that any genuine matrix-theoretic notion with that label would have to be encoded by specifying three pieces of data: a stability region, a structured acting class, and an operation. In that sense, \((\mathfrak D,\mathcal G,\circ)\)-stability is a template into which an external \((r,d)\)-indexed notion could be placed, but it is not itself the queried term. ## 6. Dimension-dependent perturbation stability for exponential bases The remaining nearby usage comes from harmonic analysis. The arXiv record for “Stability of exponential bases on \(d\)-dimensional domains” states only that the paper finds explicit stability bounds for exponential Riesz bases on domains of \(\mathbb R^d\) and that the results generalize Kadec theorem and other stability theorems in the literature. The supplied details also state that the actual theorem text, constants, and proofs are unavailable, so no paper-specific statements about hypotheses, bounds, or sharpness can be attributed beyond that record [1403.2341]. The supplied context nevertheless identifies the standard framework in which such results are usually formulated. One studies \[ \mathcal E(\Lambda)=\{e^{2\pi i \langle \lambda_n,x\rangle}\}_{n\in I} \] in \(L^2(\Omega)\), where \(\Omega\subset \mathbb R^d\) is a bounded measurable domain and \(\Lambda=\{\lambda_n\}\subset \mathbb R^d\) is a perturbation of \(\mathbb Z^d\) or of a lattice. A typical hypothesis is \[ \sup_{n\in \mathbb Z^d}\|\lambda_n-n\|