A Refinement in Čech Cohomology of Coron’s Necessary Condition Bryce Christopherson, Farhad Jafari Abstract Coron established a homological obstruction to continuous feedback stabilization of nonlinear control systems x ˙ = f ⁡ ( x , u ) \dot{x}=f(x,u) with f ∈ C ⁡ ( Ω , ℝ n ) f\in C(\Omega,\mathbb{R}^{n}) and f ⁡ ( 0 , 0 ) = 0 f(0,0)=0 , showing that local asymptotic stabilizability implies the induced homomorphism f ∗ f_{*} satisfies f ∗ ​ ( H n − 1 ​ ( Σ ϵ ) ) = H n − 1 ​ ( S n − 1 ) f_{*}\big(H_{n-1}(\Sigma_{\epsilon})\big)=H_{n-1}(S^{n-1}) , where Σ ϵ = { ( x , u ) ∈ Ω : f ( x , u ) ≠ 0 , ∥ x ∥ < ϵ , ∥ u ∥ < ϵ } \Sigma_{\epsilon}=\{(x,u)\in\Omega:f(x,u)\neq 0,\enskip\|x\|<\epsilon,\enskip\|u\|<\epsilon\} . In this paper, we refine Coron’s necessary condition using Čech cohomology and the Vietoris–Begle mapping theorem. Specifically, we prove that the closed version of Σ ϵ \Sigma_{\epsilon} , i.e. Σ ¯ ϵ = { ( x , u ) ∈ Ω : f ( x , u ) ≠ 0 , ∥ x ∥ ≤ ϵ , ∥ u ∥ ≤ ϵ } \overline{\Sigma}_{\epsilon}=\{(x,u)\in\Omega:f(x,u)\neq 0,\enskip\|x\|\leq\epsilon,\enskip\|u\|\leq\epsilon\} , must be a Čech cohomology ( n − 1 ) (n-1) -sphere and that the restriction f | Σ ¯ ϵ : Σ ¯ ϵ → f ⁡ ( Σ ¯ ϵ ) f|_{\overline{\Sigma}_{\epsilon}}:\overline{\Sigma}_{\epsilon}\rightarrow f(\overline{\Sigma}_{\epsilon}) induces an isomorphism on the Čech cohomology groups of Σ ¯ ϵ \overline{\Sigma}_{\epsilon} in all degrees. This strengthens Coron’s condition from a constraint on the top class to a full cohomological rigidity statement. MSC Classification]Primary 93D15, 93C10; Secondary 34D20, 54C60 keywords nonlinear stabilization, feedback laws, gain constraints, Brockett’s condition, continuous openness pacs [ † † equal-contributors: These authors contributed equally to this work. 1 Introduction In this paper, we consider autonomous control systems of the form \dot{x}=f(x,u) with f\in C(\Omega,\mathbb{R}^{n}) and f(0,0)=0 and \Omega\subseteq\mathbb{R}^{n}\times\mathbb{R}^{m} a neighborhood of the origin. We are interested in stabilizing these systems. That is (as in, e.g., ( 11 , Definition 10.11) ), we are looking at the following property. Definition 1 (Local Asymptotic Stabilizability - Feedback Laws) . Given a control system \dot{x}=f(x,u) with f\in C(\Omega,\mathbb{R}^{n}) and f(0,0)=0 , we say the system is locally asymptotically stabilizable by means of feedback laws if there exists a neighborhood of the origin \mathcal{O}\subseteq\textrm{proj}_{1}(\Omega) and a feedback u\colon\mathcal{O}\rightarrow\textrm{proj}_{2}(\Omega) satisfying u(0)=0 that renders the origin a locally asymptotically stable equilibrium ( 6 , Definition 2.1) of the closed-loop system \dot{x}=f\big(x,u(x)\big) . When we reference local asymptotic stability throughout the paper as defined in Definition 1 , we will always assume that the feedback u(x) satisfies the following additional criterion: Criterion 1 (Uniqueness Criterion - Feedback Stabilization) . The state-feedback controller u(x) is such that f\big(x,u(x)\big) is continuous and \dot{x}=f\big(x,u(x)\big) has a unique solution x(t) for all t at all x_{0} in a neighborhood of the origin. There is some disagreement as to whether 1 ought to be included as part of Definition 1 , as Coron 10 and Zabczyk do 29 . On one hand, a number of results may be stated more cleanly when adopting this additional restriction. On the other, it is also possible to define various more general notions of solutions (e.g. Filippov and Krasovskii) which allow for non-unique trajectories and still exhibit convergence in the desired fashion. All the same, it is necessary for us and we will adopt it here. A variety of conditions describing whether a system is locally asymptotically stabilizable by means of continuous feedback laws have been derived. As a nowhere-near-comprehensive listing, see, e.g., control Lyapunov techniques 1 ; 20 ; 3 ; 13 , topological conditions 5 ; 6 ; 10 ; 11 ; 14 ; 8 ; 17 ; 19 ; 18 , homological conditions 10 , standard Lyapunov-theoretic techniques derived from solutions to a Hamilton–Jacobi–Bellman equation 15 , and many others 11 ; 27 ; 26 ; 22 ; 21 . However, fully characterizing the existence or nonexistence of such a locally asymptotically stabilizing controller satisfying Criterion 1 for an arbitrary system was elusive until recently. To allow for concise wording of this characterization, we first need a definition: Definition 2 (Sections, Local Sections) . Given topological spaces X , Y and a function F\colon X\rightarrow Y , a local section of F is a continuous mapping \sigma\colon\mathcal{O}\rightarrow X for some open set \mathcal{O}\subseteq Y such that F\circ\sigma=\iota , where \iota denotes the inclusion map \iota\colon\mathcal{O}\rightarrow Y . If x_{0}\in X , we say that F has a local section at x_{0} if there exists a local section \sigma:\mathcal{O}\rightarrow X of F such that \mathcal{O} is an open neighborhood of F(x_{0})=y_{0} and \sigma(y_{0})=x_{0} . With this in hand, here is the complete characterization of local asymptotic and exponential stabilizability by feedback obtained by Christopherson, Mordukhovich and Jafari, 9 . Theorem 2 . For k\geq 0 , the system \dot{x}=f(x,u) with f\in C(\Omega,\mathbb{R}^{n}) and f(0,0)=0 is locally asymptotically (resp. exponentially) stabilizable by means of a C^{k} feedback law u if and only if there exists a C^{k} local section \alpha of f at the origin such that \textrm{proj}_{1}\circ\alpha is a homeomorphism (resp. diffeomorphism) and \dot{x}=(\textrm{proj}_{1}\circ\alpha)^{-1}(x) has the origin as a locally asymptotically (resp. exponentially) stable equilibrium. Moreover, every such control u is of the form u=\textrm{proj}_{2}\circ\alpha\circ(\textrm{proj}_{1}\circ\alpha)^{-1} for some \alpha satisfying the above. 2 Coron’s Condition Theorem 2 provides a complete characterization of stabilizability, but is not always practically useful in that it is often difficult to verify that such a local section does or does not exist. Accordingly, ’no-go’ theorems ruling out stabilizability are still quite attractive. The most well known of these is Brockett’s theorem, which requires that the vector field f in a control system \dot{x}=f(x,u) with f\in C(\Omega,\mathbb{R}^{n}) and f(0,0)=0 be such that f(\mathcal{O}) be a neighborhood of the origin in the codomain whenever \mathcal{O}\subseteq\Omega is a neighborhood of the origin in the domain. In 1989, Coron strengthened this condition to a homological one, revealing a necessary relation between the existence of stabilizing feedback laws and a condition regarding a singular homology group associated to the system and the homomorphism on homology induced by f . Namely, for \epsilon\in(0,\infty] , let \Sigma_{\epsilon}:=\{(x,u)\in\Omega:f(x,u)\neq 0,\enskip\|x\|<\epsilon,\enskip\|u\|<\epsilon\}, that is, \Sigma_{\epsilon}=\big(\left(\mathbb{B}_{\epsilon}^{\mathbb{R}^{n}}(0)\times\mathbb{B}_{\epsilon}^{\mathbb{R}^{m}}(0)\right)\cap\Omega\big)\setminus f^{-1}(0) . Let H_{q}(M) denote the q -th (unreduced) singular homology group with coefficients in \mathbb{Z} for a given topological space M , and let f_{*} be the induced homomorphism from H_{n-1}(\Sigma_{\epsilon}) into H_{n-1}\big(\mathbb{R}^{n}\setminus\left\{0\right\}\big) . Then, Coron’s necessary condition may be stated as follows: Theorem 3 (Coron’s Condition) . If a control system \dot{x}=f(x,u) with f\in C(\Omega,\mathbb{R}^{n}) and f(0,0)=0 is locally asymptotically stabilizable by means of continuous stationary feedback laws, then for all \epsilon>0 it is necessary that f_{*}\left(H_{n-1}(\Sigma_{\epsilon})\right)\cong H_{n-1}\big(\mathbb{R}^{n}\setminus\left\{0\right\}\big). Since \mathbb{R}^{n}\setminus\{0\} deformation retracts to the n-1 sphere S^{n-1} , and since H_{k}(S^{n-1})=\left\{\begin{matrix}0,&k\neq 0,n-1\\ \mathbb{Z},&k=0,n-1\end{matrix}\right. Coron’s condition simply says that it is necessary to have f_{*}\left(H_{n-1}(\Sigma_{\epsilon})\right)=\mathbb{Z} . Coron demonstrated that the condition of Theorem 3 is necessary but not sufficient with the following example: Example 1 (Insufficiency of Theorem 3 10 ) . Consider the system on \mathbb{R}^{3} given by \dot{x}=f(x,u)=\begin{bmatrix}x_{3}^{2}(x_{1}-x_{2})&x_{3}^{2}(x_{2}-x_{3})&u\end{bmatrix}^{T} This system satisfies Coron’s necessary condition ( Theorem 3 ) but is not stabilizable via a continuous feedback. This raises a natural question: is Coron’s top-degree homology condition the whole story? The example above suggests either that stabilization cannot be characterized in a purely (co)homological way (which we view as a likely possibility) or, if it can, that a stronger (co)homological condition is necessary. After some preliminaries in the next section, we will prove a refinement of this condition which is stronger than Theorem 3 . 3 Reframing in Čech Cohomology Throughout this section and the remainder of the paper, all Čech cohomology groups are taken with coefficients in {\mathbb{Z}} , unless explicitly stated otherwise. Early in the development of homological theories for compact metric spaces, Vietoris proved a landmark result known as the Vietoris theorem or Vietoris-Begle mapping theorem 24 ; 23 . This result, concerning the so-called Vietoris homology groups of such a space, showed that continuous mappings whose fibers have the homology of points induce isomorphisms in homology. Specifically, letting \widetilde{VH}_{k} denote the k^{th} reduced Vietoris homology group of a compact metric space, the theorem is as follows: Theorem 4 (Vietoris-Begle Mapping 28 ) . Let X and Y be compact metric spaces and let f:X\rightarrow Y be a continuous surjection such that \widetilde{VH}_{k}\big(f^{-1}(y)\big)=\widetilde{VH}_{k}\big(\{*\})\big)=0 for 0\leq k\leq N-1 and all y\in Y . Then, the induced homomorphism f_{*}:\widetilde{VH}_{k}(X)\rightarrow\widetilde{VH}_{k}(Y) is an isomorphism for k\leq N-1 and a surjection for k=N . The condition that \widetilde{VH}_{k}\big(f^{-1}(y)\big)=\widetilde{VH}_{k}\big(\{*\})\big) for 0\leq k\leq N-1 and all y\in Y is sometimes stated instead as ’ f has (N-1) -acyclic fibers ’ (wherein the (N-1) modifying prefix is omitted when the condition holds for all r\geq 0 ). This is also commonly applied to other homology and cohomology theories as well, so we opt to avoid this terminology here to avoid any confusion regarding which homology or cohomology theory is intended. Later, Vietoris homology would eventually fall out of fashion and give rise to Čech homology and cohomology (Dowker gives a wonderful treatment of the relation between Vietoris and Čech homology/cohomology and the strengths of the latter/deficiencies of the former 12 ). With these more modern tools in hand, Begle would go on to produce the following generalization: Theorem 5 (Begle 16 ; 7 ) . Let X and Y be paracompact spaces, let G be an abelian group, and let f:X\rightarrow Y be a continuous closed surjection. If \check{\mathrm{H}}^{k}(f^{-1}(y);G)=\check{\mathrm{H}}^{k}(\{*\};G) for each 0\leq k0 sufficiently small so that \overline{\mathbb{B}_{2\epsilon}(y)}\subset B . Define A_{\epsilon}(y):=\alpha\big(\overline{\mathbb{B}_{\epsilon}(y)}\big),\qquad W_{\epsilon}(y):=\alpha\big(\mathbb{B}_{2\epsilon}(y)\big)\subset\alpha(\overline{B}). First, we show W_{\epsilon}(y) is open in \alpha(\overline{B}) . Since \tilde{\alpha} is a homeomorphism and \mathbb{B}_{2\epsilon}(y) is open in \overline{B} (with the subspace topology), it follows that W_{\epsilon}(y)=\tilde{\alpha}\big(\mathbb{B}_{2\epsilon}(y)\big) is open in \alpha(\overline{B}) . Now, we show A_{\epsilon}(y) is homeomorphic to a closed n –ball. Because \tilde{\alpha} is a homeomorphism, A_{\epsilon}(y)=\tilde{\alpha}\big(\overline{\mathbb{B}_{\epsilon}(y)}\big) is homeomorphic to \overline{\mathbb{B}_{\epsilon}(y)} , which is a closed n –ball in \mathbb{R}^{n} . Hence A_{\epsilon}(y) is compact and contractible, and therefore is a compact absolute neighborhood retract. Next, we construct a retraction r_{\epsilon}:W_{\epsilon}(y)\rightarrow A_{\epsilon}(y) . Define the radial projection p_{\epsilon}:\mathbb{B}_{2\epsilon}(y)\rightarrow\overline{\mathbb{B}_{\epsilon}(y)} by p_{\epsilon}(z)=\begin{cases}z,&\|z-y\|\leq\epsilon,\\ y+\epsilon\frac{z-y}{\|z-y\|},&\epsilon<\|z-y\|<2\epsilon.\end{cases} This map is continuous and satisfies p_{\epsilon}(z)=z for all z\in\overline{\mathbb{B}_{\epsilon}(y)} . Defining r_{\epsilon}:=\tilde{\alpha}\circ p_{\epsilon}\circ\tilde{\alpha}^{-1}:W_{\epsilon}(y)\rightarrow A_{\epsilon}(y). and noting that \tilde{\alpha} is a homeomorphism, it is straightforward to see that r_{\epsilon} is continuous. Moreover, if w\in A_{\epsilon}(y) , then w=\tilde{\alpha}(z) for some z\in\overline{\mathbb{B}_{\epsilon}(y)} , and hence r_{\epsilon}(w)=\tilde{\alpha}\big(p_{\epsilon}(z)\big)=\tilde{\alpha}(z)=w . Thus r_{\epsilon} is a retraction of W_{\epsilon}(y) onto A_{\epsilon}(y) . Penultimately, we construct a neighborhood basis of open sets for the fiber (f|_{K})^{-1}(y) . Define U_{\epsilon}(y):=g^{-1}\big(W_{\epsilon}(y)\big). Since g is continuous and W_{\epsilon}(y) is open in \alpha(\overline{B}) , the set U_{\epsilon}(y) is open in K . Furthermore, \bigcap_{\epsilon>0}U_{\epsilon}(y)=g^{-1}\left(\bigcap_{\epsilon>0}W_{\epsilon}(y)\right)=g^{-1}\big(\alpha(y)\big)=f^{-1}(y)\cap K, so \{U_{\epsilon}(y)\}_{\epsilon>0} forms a neighborhood basis of open sets for the fiber (f|_{K})^{-1}(y)=f^{-1}(y)\cap K . Finally, we show U_{\epsilon}(y) is contractible. Define \tilde{g}_{\epsilon}:=r_{\epsilon}\circ g\big|_{U_{\epsilon}(y)}:U_{\epsilon}(y)\rightarrow A_{\epsilon}(y). Because r_{\epsilon} is a retraction and g|_{A_{\epsilon}(y)}=\mathrm{id} , it follows that \tilde{g}_{\epsilon} restricts to the identity on A_{\epsilon}(y) . Hence A_{\epsilon}(y) is a retract of the open set U_{\epsilon}(y) . Since A_{\epsilon}(y) is a compact absolute neighborhood retract, Lemma 6 implies that A_{\epsilon}(y) is a strong deformation retract of U_{\epsilon}(y) . Therefore U_{\epsilon}(y) is contractible. Consequently, for all k\geq 0 , \check{\mathrm{H}}^{k}\big(U_{\epsilon}(y)\big)\cong\check{\mathrm{H}}^{k}(\{*\}) . Since the sets U_{\epsilon}(y) form a cofinal system of open neighborhoods of the compact set f^{-1}(y)\cap K in K , the continuity property of Čech cohomology ( Lemma 7 ) yields \check{\mathrm{H}}^{k}\big(f^{-1}(y)\cap K\big)=\underset{f^{-1}(y)\cap K\subset U\subset K\text{ open}}{\underset{\longrightarrow}{\lim}}\check{\mathrm{H}}^{k}(U)=\underset{\longrightarrow}{\lim}\check{\mathrm{H}}^{k}\big(U_{\epsilon}(y)\big)\cong\check{\mathrm{H}}^{k}(\{*\}). ∎ From this, we may use the Vietoris-Begle mapping theorem to great effect. The product of this is the main result of the paper. Note that here we use the closed version of \Sigma_{\epsilon} , unlike the open analogue Coron uses in Theorem 3 . That is, we use \overline{\Sigma}_{\epsilon}:=\Big(\big(\overline{\mathbb{B}_{\epsilon}^{\mathbb{R}^{n}}(0)}\times\overline{\mathbb{B}_{\epsilon}^{\mathbb{R}^{m}}(0)}\big)\cap\Omega\Big)\setminus f^{-1}(0). Theorem 9 (Coron’s Condition-Strong Version) . Suppose \dot{x}=f(x,u) with f\in C(\Omega,\mathbb{R}^{n}) and f(0,0)=0 is locally asymptotically stabilizable by continuous stationary feedback. Then, for all sufficiently small \epsilon>0 , the set \overline{\Sigma}_{\epsilon} is a Čech cohomology (n-1) –sphere (with integer coefficients). Moreover, for all k\geq 0 , the restriction f|_{\overline{\Sigma}_{\epsilon}}:\overline{\Sigma}_{\epsilon}\rightarrow f(\overline{\Sigma}_{\epsilon}) induces an isomorphism (f|_{\overline{\Sigma}_{\epsilon}})^{*}:\check{H}^{k}\big(f(\overline{\Sigma}_{\epsilon})\big)\rightarrow\check{H}^{k}(\overline{\Sigma}_{\epsilon}). Proof. Fix \epsilon>0 sufficiently small that \overline{\mathbb{B}_{\epsilon}^{\mathbb{R}^{n}}(0)}\times\overline{\mathbb{B}_{\epsilon}^{\mathbb{R}^{m}}(0)}\subset\Omega , so that \overline{\Sigma}_{\epsilon} is a compact subset of \Omega . By Brockett’s necessary condition for continuous stationary stabilization, for every neighborhood W of (0,0) the set f(W) contains a neighborhood of 0 in \mathbb{R}^{n} . In particular, for \epsilon>0 sufficiently small there exists \rho>0 such that \mathbb{B}_{\rho}(0)\subset f\left((\overline{\mathbb{B}_{\epsilon}^{\mathbb{R}^{n}}(0)}\times\overline{\mathbb{B}_{\epsilon}^{\mathbb{R}^{m}}(0)})\cap\Omega\right) . Since f\big(W\setminus f^{-1}(0)\big)\supset f(W)\setminus\{0\} , it follows that \mathbb{B}_{\rho}(0)\setminus\{0\}\subset f(\overline{\Sigma}_{\epsilon}) . Thus, f(\overline{\Sigma}_{\epsilon}) is a punctured neighborhood of the origin in \mathbb{R}^{n} and therefore deformation retracts onto S^{n-1} . By Lemma 8 , after shrinking if necessary there exists a compact neighborhood U\subset\Omega of (0,0) such that f|_{U}:U\rightarrow f(U) is a continuous surjection and every fiber f^{-1}(y)\cap U is Čech-acyclic. Fix \epsilon>0 so small that \overline{\Sigma}_{\epsilon}\subset U . Then, f|_{\overline{\Sigma}_{\epsilon}}:\overline{\Sigma}_{\epsilon}\rightarrow f(\overline{\Sigma}_{\epsilon}) is a continuous surjection between compact metric spaces, hence a closed map. Moreover, for each y\in f(\overline{\Sigma}_{\epsilon})\subset f(U) , \big(f|_{\overline{\Sigma}_{\epsilon}}\big)^{-1}(y)=f^{-1}(y)\cap\overline{\Sigma}_{\epsilon} is Čech-acyclic by Lemma 8 . Therefore Theorem 5 applies and yields that f^{*}:\check{H}^{k}\big(f(\overline{\Sigma}_{\epsilon})\big)\rightarrow\check{H}^{k}(\overline{\Sigma}_{\epsilon}) is an isomorphism for all k\geq 0 . Since f(\overline{\Sigma}_{\epsilon}) is a punctured neighborhood of the origin, it deformation retracts onto S^{n-1} , and hence \check{H}^{k}\big(f(\overline{\Sigma}_{\epsilon})\big)\cong\check{H}^{k}(\mathbb{R}^{n}\setminus\{0\}) for all k . This proves that \overline{\Sigma}_{\epsilon} has the Čech cohomology of S^{n-1} and that f^{*}:\check{H}^{k}\big(f(\overline{\Sigma}_{\epsilon})\big)\rightarrow\check{H}^{k}(\overline{\Sigma}_{\epsilon}) is an isomorphism for all k . ∎ It’s important to note that this doesn’t mean that \overline{\Sigma}_{\epsilon} must be all that much like a sphere. For example, the Poincaré homology 3 -sphere has the same Čech cohomology as S^{3} but nontrivial homotopy groups, so it is not homeomorphic to S^{3} . Indeed, while Theorem 9 does say that \overline{\Sigma}_{\epsilon} must be connected, it need not even be the case that \overline{\Sigma}_{\epsilon} is simply connected. Example 2 . As in Example 1 , the system on \mathbb{R}^{3}\times\mathbb{R} given by \dot{x}=f(x,u)=\begin{bmatrix}x_{3}^{2}(x_{1}-x_{2})&x_{3}^{2}(x_{2}-x_{3})&u\end{bmatrix}^{T} satisfies Coron’s necessary condition ( Theorem 3 ) but is not stabilizable via a continuous feedback. Coron used this to demonstrates that Theorem 3 is not sufficient 10 . However, it also serves as a nice example showing that Theorem 9 is necessary: \Sigma_{\epsilon} is not a Čech cohomology 2 –sphere (with integer coefficients), which implies that the system violates Theorem 9 and hence cannot be stabilized by a continuous stationary feedback. To see that the strong condition of Theorem 9 fails, we show that \Sigma_{\epsilon} has nontrivial first cohomology, which is impossible for a Čech cohomology 2 –sphere. To do so, first observe that (x_{3},u)\neq(0,0) for every (x,u)\in\overline{\Sigma}_{\epsilon} . Indeed, if x_{3}=0 and u=0 , then f(x,u)=(0,0,0) , so (x,u)\in f^{-1}(0) , contradicting (x,u)\in\overline{\Sigma}_{\epsilon} . Hence the map \pi:\overline{\Sigma}_{\epsilon}\rightarrow S^{1},\qquad\pi(x,u):=\frac{(x_{3},u)}{\sqrt{x_{3}^{2}+u^{2}}} is well-defined and continuous. Now, fix \rho\in(0,\epsilon/4) and consider the loop \gamma:[0,2\pi]\rightarrow\overline{\Sigma}_{\epsilon},\qquad\gamma(t):=\big((0,0,\rho\cos t),\ \rho\sin t\big). We check that \gamma(t)\in\overline{\Sigma}_{\epsilon} for all t : clearly \|\gamma(t)\|<\epsilon by choice of \rho . Also, f(\gamma(t))=\bigl(0,\ -(\rho\cos t)^{3},\ \rho\sin t\bigr) , which is never zero because (\rho\cos t,\rho\sin t)\neq(0,0) . Therefore, \gamma avoids f^{-1}(0) and lies in \overline{\Sigma}_{\epsilon} . Finally, (\pi\circ\gamma)(t)=\frac{(\rho\cos t,\rho\sin t)}{\sqrt{\rho^{2}\cos^{2}t+\rho^{2}\sin^{2}t}}=(\cos t,\sin t), which has winding number 1 in S^{1} . Thus (\pi\circ\gamma)_{*}:H_{1}(S^{1})\rightarrow H_{1}(S^{1}) is the identity on \mathbb{Z} , and in particular \pi_{*}:H_{1}(\overline{\Sigma}_{\epsilon})\rightarrow H_{1}(S^{1}) is surjective. Equivalently, \pi^{*}:H^{1}(S^{1})\rightarrow H^{1}(\overline{\Sigma}_{\epsilon}) is injective, so H^{1}(\overline{\Sigma}_{\epsilon})\neq 0 . Since \overline{\Sigma}_{\epsilon} is a compact metric space, the same conclusion holds for Čech cohomology: Since \pi\circ\gamma:S^{1}\rightarrow S^{1} has degree 1 , the induced map (\pi\circ\gamma)^{*}:\check{H}^{1}(S^{1})\rightarrow\check{H}^{1}(S^{1}) is the identity. Thus \gamma^{*}\circ\pi^{*} is the identity on \check{H}^{1}(S^{1}) , so \pi^{*} is injective and \check{H}^{1}(\overline{\Sigma}_{\epsilon})\neq 0 . But a Čech cohomology 2 –sphere has a trivial first Čech cohomology group, so \overline{\Sigma}_{\epsilon} is emphatically not a Čech cohomology 2 -sphere. Therefore, the strong necessary condition of Theorem 9 fails for this system, even though Coron’s condition holds. Example 3 . The condition of Theorem 9 is not sufficient. To see this, it is enough to look at the system \dot{x}=f(x,u) with f(x,u)=x , which is certainly not locally asymptotically stabilizable. However, \overline{\Sigma}_{\epsilon}=\left(\overline{\mathbb{B}}_{\epsilon}^{\mathbb{R}^{n}}(0)\times\overline{\mathbb{B}}_{\epsilon}^{\mathbb{R}^{m}}(0)\right)\setminus f^{-1}(0)=\left(\overline{\mathbb{B}}_{\epsilon}^{\mathbb{R}^{n}}(0)\setminus\{0\}\right)\times\overline{\mathbb{B}}_{\epsilon}^{\mathbb{R}^{m}}(0), so, via the contractibility of \overline{\mathbb{B}}_{\epsilon}^{\mathbb{R}^{m}}(0) , \check{\mathrm{H}}^{k}(\overline{\Sigma}_{\epsilon})\cong\check{\mathrm{H}}^{k}(\overline{\mathbb{B}}_{\epsilon}^{\mathbb{R}^{n}}(0)\setminus\{0\})\cong\check{\mathrm{H}}^{k}(S^{n-1}). Worse still, f|_{\overline{\Sigma}_{\epsilon}} is just the projection \pi:\left(\overline{\mathbb{B}}_{\epsilon}^{\mathbb{R}^{n}}(0)\setminus\{0\}\right)\times\overline{\mathbb{B}}_{\epsilon}^{\mathbb{R}^{m}}(0)\rightarrow\overline{\mathbb{B}}_{\epsilon}^{\mathbb{R}^{n}}(0)\setminus\{0\} , which induces an isomorphism in Čech cohomology in all degrees. So, the conditions of Theorem 9 all hold, yet the system fails miserably to be stabilizable by feedback. 5 Conclusion We have shown that Coron’s classical homological obstruction to feedback stabilization can be strengthened by using Čech cohomology instead and applying Begle’s theorem. While Coron’s condition requires surjectivity of the induced map on the top singular homology group of \Sigma_{\epsilon} , our refinement demonstrates that the entire Čech cohomology of the closed analogue \overline{\Sigma}_{\epsilon} must coincide with that of the sphere S^{n-1} . This, like Brockett’s original necessary condition, gives further credence to the intuitive sense one gets that a control system must be able to ’see’ in all directions anywhere sufficiently near an equilibrium point for stabilization to be possible. This strengthened condition imposes significant additional topological constraints on \overline{\Sigma}_{\epsilon} and, with it, on the form a system must have if it is to be stabilizable. Several directions for future work following this theme suggest themselves: For one, it would be interesting to find if there are additional conditions that, if imposed, render this cohomology-sphere necessary condition a sufficient one instead. Likewise, producing any homological or cohomological obstructions that are stronger yet still might be of great value. In light of the fact that mappings satisfying Brockett’s necessary condition satisfy a local analogue to the characteristic property of quotient maps ( 9 , Section 4.1) and the fact that Smale provided an analogue of the Vietoris mapping theorem in homotopy 25 which was extended by Calcut et. al. and applies directly to quotient maps 7 , it would also be interesting to know if there is an analogue to Theorem 9 in homotopy. Overall, this work demonstrates that the topology underlying feedback stabilization is rich and that Coron’s original condition captures a genuinely deep property of stabilizable systems. As we have shown here, Čech cohomology provides a robust context for expressing some of these constraints stabilization requires, but the search for a complete homological or cohomological characterization of stabilizability (if such a thing is even possible) remains ongoing. Declarations • Funding: The authors did not receive support from any organization for the submitted work. • Competing interests: All authors certify that they have no affiliations with or involvement in any organization or entity with any financial interest or non-financial interest in the subject matter or materials discussed in this manuscript.