4
$\begingroup$

Find an example of functions $f:A\to B$ and $g:B\to C$ such that $f$ and $g\circ f$ are both injective, but $g$ is not injective.

So If I understand this correctly,

  1. Need a function $f$ that is injective and that will also make $g$ injective when plugged in during $g\circ f$.
  2. Need a function $g$ that is not injective on its own
  3. The range of $f$ must be a subset of the domain of $g$

I tried thinking along the lines of using variations of $f(x)=x$ and $g(x)=x^2$ but all those leave my composite function as non injective

I've also been using $x\in\Bbb R$ so as to keep the range of $f$ and domain of $g$ the same.

Any suggestions of where to go with this? Thank you

Edit: Thank you everyone the answers were very helpful in understanding the problem and concepts better

$\endgroup$
  • 6
    $\begingroup$ The range of f must be the domain of g No. Range of $f$ needs only be a subset of the domain of $g$. $\endgroup$ – dxiv Dec 4 '17 at 5:20
  • $\begingroup$ ah so the range of $f$ could be just natural numbers or something like that while the domain of $g$ is the real numbers? $\endgroup$ – mathminutemaid Dec 4 '17 at 5:22
  • 1
    $\begingroup$ Yes. See more at Domain, Co-Domain & Range of a Function for example. $\endgroup$ – dxiv Dec 4 '17 at 5:24
  • $\begingroup$ It's worth noting that if "the range of $f$ is the domain of $g$" then $g$ being non-injective is enough to make the composite function non-injective. This is easy to see, take $a,b\in \text{Dom}(g)$ such that $g(a)=g(b)$. They are witnesses to the fact that $g$ is not injective. Since range of $f$ is the domain of $g$ there exist, $x,y$ such that $f(x)=a$ and $f(y)=b$ than $(g\circ f)(x)=g(f(x))=g(a)=g(b)=g(f(y))=(g\circ f)(y)$ and so $g\circ f$ is not injective. $\endgroup$ – DRF Dec 4 '17 at 7:33
7
$\begingroup$

Hint. Let's take $g:{\Bbb R}\to{\Bbb R}$ with $g(x)=x^2$ as you suggested - a good, simple, well-known function which is not one-to-one. Suppose we want $f:{\Bbb R}\to{\Bbb R}$ also.

To say $g\circ f$ is one-to-one means: if you know the value of $g(f(x))$ then you know the value of $x$ - that is, know it for sure, with only one possibility. So, suppose the value of $g(f(x))$ is given. We have $$g(f(x))=(f(x))^2\ ;$$ a good way to proceed from here would be [1] find the value of $f(x)$, then [2] find the value of $x$. Step [2] is easy as $f$ is supposed to be one-to-one. The problem is that step [1] would usually give you two values of $f(x)$.

But now suppose, for example, that $f(x)$ is always positive. Then step [1] gives a definite value of $f(x)$ and all is OK.

So - can you think of a (well known) one-to-one function $f:{\Bbb R}\to{\Bbb R}$ for which $f(x)$ is always a positive number?

Hope this doesn't sound too long-winded but I am hoping to illustrate how you can think about this kind of problem.

$\endgroup$
  • $\begingroup$ Good answer. And now I shall spoil it: Most people will think of the function $f:x\mapsto e^x$, and then the composition becomes $x\mapsto (e^x)^2 = e^{2x}$ which is a familiar injective function. $\endgroup$ – Jeppe Stig Nielsen Dec 4 '17 at 8:31
  • $\begingroup$ Additionally, I like the fact that we take $A=B=C=\mathbb{R}$ here. Some other examples may make you think that such examples occur only because you pick one of the sets "artificially" small (when in fact there is a natural continuation to all of $\mathbb{R}$). $\endgroup$ – Jeppe Stig Nielsen Dec 4 '17 at 8:38
  • $\begingroup$ @JeppeStigNielsen what do you mean by $A=B=C=\mathbb{R}$? The trick is that the range of $e^x$ is $(0,\infty)$ if it were $\mathbb{R}$ it wouldn't work. $\endgroup$ – DRF Dec 4 '17 at 13:00
  • $\begingroup$ @DRF The sets from the question, where $f:A\to B$ and $g:B\to C$. I know $f$ cannot be surjective, because then the failure of $g$ to be injective would lead to the composition $g\circ f$ failing to be injective as well. I think David already explained that in his answer. But if you compare to Parcly Taxel's answer, where $f(x)=x$, one could ask why this $f$ is restricted to the domain $A = \left[ 0,\infty \right)$. We could also just take $A=\{ 0\}$ and $B=\{ 0,1\}$, you know. I know every example is essentially the same. I just think it is more elegant to have $A=B=C=\mathbb{R}$. $\endgroup$ – Jeppe Stig Nielsen Dec 4 '17 at 21:33
6
$\begingroup$

Take $f:[0,\infty)\to\Bbb R$ with $f(x)=x$ and $g:\Bbb R\to[0,\infty)$ with $g(x)=|x|$. Then $f$ is injective and $g$ is not, while $g\circ f:[0,\infty)\to[0,\infty)$ is the identity function and is therefore injective.

$\endgroup$
  • 1
    $\begingroup$ So in this case $g$(x) isn't injective because $R$ can't map fully onto [0,∞) right $\endgroup$ – mathminutemaid Dec 4 '17 at 5:28
  • 1
    $\begingroup$ @mathminutemaid Actual reason: $g(-1)=g(1)=1$. $\endgroup$ – Parcly Taxel Dec 4 '17 at 5:29
  • $\begingroup$ R has plenty of injective maps to $[0, \infty)$, it's just that g is injective on the subset which f characterizes and not all of R $\endgroup$ – Jason Carr Dec 4 '17 at 5:30
  • 1
    $\begingroup$ Yeah, I deleted my comment and downvote again. Sorry for the rubbish. $\endgroup$ – Jeppe Stig Nielsen Dec 4 '17 at 8:44
2
$\begingroup$

$g:B\to C$ being non-injective means that there are distinct $b,b'\in B$ with $g(b)=g(b')$. All you need is to ensure that the image of $f$ never contains both values $b$ and $b'$ at the same time (maybe neither).

Example.

Choose $g(x)=x^2$, then the injectivity violating pairs are $(x,-x)$ for $x>0$. Just make sure that if $f(x)=y$, then there is no $x'$ with $f(x')=-y$, e.g. choose

$$f(x)=e^x$$

which is injective and will only assume positive values. Note that here $A=C=\Bbb R$ and $B=(0,\infty)$.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.