Recession cone

In mathematics, especially convex analysis, the recession cone of a set ${\displaystyle A}$ is a cone containing all vectors such that ${\displaystyle A}$ recedes in that direction. That is, the set extends outward in all the directions given by the recession cone.^[1]

Given a nonempty set ${\displaystyle A\subset X}$ for some vector space ${\displaystyle X}$, then the recession cone ${\displaystyle \operatorname {recc} (A)}$ is given by

${\displaystyle \operatorname {recc} (A)=\{y\in X:\forall x\in A,\forall \lambda \geq 0:x+\lambda y\in A\}.}$^[2]

If ${\displaystyle A}$ is additionally a convex set then the recession cone can equivalently be defined by

${\displaystyle \operatorname {recc} (A)=\{y\in X:\forall x\in A:x+y\in A\}.}$^[3]

If ${\displaystyle A}$ is a nonempty closed convex set then the recession cone can equivalently be defined as

${\displaystyle \operatorname {recc} (A)=\bigcap _{t>0}t(A-a)}$ for any choice of ${\displaystyle a\in A.}$^[3]

• If ${\displaystyle A}$ is a nonempty set then ${\displaystyle 0\in \operatorname {recc} (A)}$.
• If ${\displaystyle A}$ is a nonempty convex set then ${\displaystyle \operatorname {recc} (A)}$ is a convex cone.^[3]
• If ${\displaystyle A}$ is a nonempty closed convex subset of a finite-dimensional Hausdorff space (e.g. ${\displaystyle \mathbb {R} ^{d}}$), then ${\displaystyle \operatorname {recc} (A)=\{0\}}$ if and only if ${\displaystyle A}$ is bounded.^[1][3]
• If ${\displaystyle A}$ is a nonempty set then ${\displaystyle A+\operatorname {recc} (A)=A}$ where the sum denotes Minkowski addition.

The asymptotic cone for ${\displaystyle C\subseteq X}$ is defined by

${\displaystyle C_{\infty }=\{x\in X:\exists (t_{i})_{i\in I}\subset (0,\infty ),\exists (x_{i})_{i\in I}\subset C:t_{i}\to 0,t_{i}x_{i}\to x\}.}$^[4][5]

By the definition it can easily be shown that ${\displaystyle \operatorname {recc} (C)\subseteq C_{\infty }.}$^[4]

In a finite-dimensional space, then it can be shown that ${\displaystyle C_{\infty }=\operatorname {recc} (C)}$ if ${\displaystyle C}$ is nonempty, closed and convex.^[5] In infinite-dimensional spaces, then the relation between asymptotic cones and recession cones is more complicated, with properties for their equivalence summarized in.^[6]

• Dieudonné's theorem: Let nonempty closed convex sets ${\displaystyle A,B\subset X}$ a locally convex space, if either ${\displaystyle A}$ or ${\displaystyle B}$ is locally compact and ${\displaystyle \operatorname {recc} (A)\cap \operatorname {recc} (B)}$ is a linear subspace, then ${\displaystyle A-B}$ is closed.^[7][3]
• Let nonempty closed convex sets ${\displaystyle A,B\subset \mathbb {R} ^{d}}$ such that for any ${\displaystyle y\in \operatorname {recc} (A)\backslash \{0\}}$ then ${\displaystyle -y\not \in \operatorname {recc} (B)}$, then ${\displaystyle A+B}$ is closed.^[1][4]

• Barrier cone