I'm self-studying linear algebra and don't want to post too many details here in case someone is using the same textbook. I ran across this simple exercise where I think there is an extra hypothesis.
To prove:
Suppose $ A \in \mathbb{R}^n, A \neq 0, c \in \mathbb{R}$.
Then the set $ E = \{ x : A \cdot x \ge c, x \in \mathbb{R}^n \} $ is convex.
My Attempt:
We need to show that for all $ 0 \le \alpha \le 1 $ and $x, y \in E $ that $ w = \alpha x + (1-\alpha) y \in E.$
Consider $ A \cdot w = A \cdot (\alpha x + (1 - \alpha) y) = \alpha (A \cdot x) + (1 - \alpha)(A \cdot y)$ by the linearity of the dot product. This last expression is at least $c$ since $A \cdot x \ge c$ and $A \cdot y \ge c$, and the expression is the weighted average of these two real numbers. Hence $w \in E$ as desired.
My Question:
Why is the $A \ne 0$ hypothesis necessary? Suppose $A = 0$. If $ c \leq 0$, then $E = \mathbb{R}^n$, and the entire vector space is convex since it is closed under addition and scalar multiplication. If $ c \gt 0 $, then $E = \varnothing$ and is "vacuously" convex.