4

Suppose $f: X \rightarrow Y$, and is one-to-one, and let $A \subseteq X$, prove that $f^{-1}[f[A]] = A$.

EDIT: Actually, this identity should hold even if $f$ is not one-to-one (injective), right?

This is completely intuitive and logical, but I can't think of a proof strategy to use to prove it? I'm familiar with induction and element chasing, but they don't seem to work here.

Here is the beginning of my element chasing proof to prove that $f^{-1}[f[A]] \subseteq A$.

Start with arbitrary element: $a \in f^{-1}[f[A]]$ \

Apply definition of preimage: $a \in \{x \in X | f(x) \in f[A] \}$ \

Simplify and convert to boolean expression: $a \in X \land f(a) \in f[A]$ \

Apply definition of image: $a \in X \land f(a) \in \{y \in Y | y = f(x), x \in A \}$ \

I'm stuck here...

Git Gud
  • 31,356
clay
  • 2,713
  • 23
  • 34
  • 2
    Can you type your element chasing try into the question, please? – Git Gud Jul 10 '14 at 14:25
  • 1
    Try to prove $f^{-1}[f(A)] \subset A$ and $f^{-1}[f(A)] \supset A$. – hrkrshnn Jul 10 '14 at 14:28
  • $f^{-1}\circ f= I_d$ – Shine Jul 10 '14 at 14:33
  • @Shine: $f$ isn't a bijection, $f^{-1}$ isn't a function... It takes subsets of $Y$ to subsets of $X$. – Najib Idrissi Jul 10 '14 at 14:34
  • If you look into this overview, you can find several related posts: http://math.stackexchange.com/questions/347880/proving-a-set-is-a-subset, http://math.stackexchange.com/questions/261157/show-s-f-1fs-for-all-subsets-s-iff-f-is-injective, http://math.stackexchange.com/questions/78110/is-this-set-theory-statement-true, http://math.stackexchange.com/questions/368990/why-f-1fa-not-a – Martin Sleziak Aug 09 '14 at 08:42

2 Answers2

3

"Apply definition of image: $a \in X \land f(a) \in \{y \in Y \mid y = f(x), x \in A \}$".

The set should be written as $\{y\in Y\colon \exists x\in A(y=f(x))\}$, as you've written, it doesn't make sense.

You wish to conclude that $a\in A$.

Consider $f(a)$. One has $f(a)\in \{y\in Y\colon \exists x\in A(y=f(x))\}=\{f(x)\colon x\in A\}$, so there exists $x\in A$ such that $f(x)=f(a)$.

"Actually, this identity should hold even if $f$ is not one-to-one (injective), right?"

Can you now answer your own question?

The other inclusion holds universally (i.e., it doesn't require $f$'s injectivity).

Git Gud
  • 31,356
3

If $f$ is not injective, it is not true.
Example: $X=\{0,1\},Y=\{0\},A=\{0\}$ and $f=$ the only possibility; then $f^{-1}(f(A))=X$.

In the general case, the only you can say is $f^{-1}(f(A))\supseteq A$: $f^{-1}(f(A))=\{x\in X \mid f(x)\in f(A)\}$ and this clearly contains $A$.

Assume $f$ is injective.
Take $x\in f^{-1}(f(A))$; by definition, $f(x)\in f(A)$, so there must exist an $a\in A$ such that $f(x)=f(a)$; by injectivity $x=a\in A$ and you get the reverse inclusion.

brad
  • 166