# Linear Algebra exam problem: is my answer correct?

This was a problem on a midterm I just took.

## Homework Statement

Let f and g be linear transformations such that f: C^n ---> C^m is injective and g: C^m ---> C^n is onto (C = complex numbers).

Prove or disprove with a counterexample: g composed with f is an isomorphism.

## The Attempt at a Solution

I claim the proposition is false.

Let g: C^1 --> C^1 be such that g(z) ={ 0 if z=0, z -(1+i) if z does not = 0}.
Let f: C^1 --> C^1 be such that f(z) = z.

Clearly, f is injective. I claim that g is onto, since if y is complex, there exists a complex z such that g(z) = y. Namely, z = y + (1+i). However, since both g(0)=g(1+i)=0, g is not injective.

Hence, g composed with f is not injective, since g(f(0))=g(f(1+i)), and thus cannot be an isomorphism since it is not bijective (not invertible).

How does that look?

Related Calculus and Beyond Homework Help News on Phys.org
Dick
Homework Helper
This was a problem on a midterm I just took.

## Homework Statement

Let f and g be linear transformations such that f: C^n ---> C^m is injective and g: C^m ---> C^n is onto (C = complex numbers).

Prove or disprove with a counterexample: g composed with f is an isomorphism.

## The Attempt at a Solution

I claim the proposition is false.

Let g: C^1 --> C^1 be such that g(z) ={ 0 if z=0, z -(1+i) if z does not = 0}.
Let f: C^1 --> C^1 be such that f(z) = z.

Clearly, f is injective. I claim that g is onto, since if y is complex, there exists a complex z such that g(z) = y. Namely, z = y + (1+i). However, since both g(0)=g(1+i)=0, g is not injective.

Hence, g composed with f is not injective, since g(f(0))=g(f(1+i)), and thus cannot be an isomorphism since it is not bijective (not invertible).

How does that look?
Pretty bad. There's no value of z such that g(z)=(-(1+i)), so g isn't onto. And g certainly isn't linear.

oh damn, yeah that's brutal.

so without providing a proof, is the proposition true or false?

In my head I convinced myself that if g isn't injective it ruins the isomorphism but i'm starting to think that's a serious mind-lapse.

Dick
Homework Helper
If g is onto then m<=n. What does f injective tell you about m and n?

g is onto, but in general if we restrict the domain of g to some subset of C^m, there's no reason to suspect that it remains onto as a map to C^n. f isn't necessarily onto, so it's image may be some subset of C^m, and therefore the image of g composed with f will not necessarily be all of C^n. Therefore I suspect the composition is not necessarily an isomorphism. One example you could consider, I think would be if f mapped from C^1 to C^2 where z mapped to (z,0). Then you could consider a g that maps from C^2 to C^1 mapping (z1,z2) to z2. The composition of these will map everything in C^1 to 0.

f inj -> ker(f)=0 -> n = m - 0 = m ---> lin trans between vect spaces of same dim is isomorphism.

ugh.

Dick
Homework Helper
f inj -> ker(f)=0 -> n = m - 0 = m ---> lin trans between vect spaces of same dim is isomorphism.

ugh.
Sure, once you conclude m=n then both f and g are isomorphisms.

g is onto, but in general if we restrict the domain of g to some subset of C^m, there's no reason to suspect that it remains onto as a map to C^n. f isn't necessarily onto, so it's image may be some subset of C^m, and therefore the image of g composed with f will not necessarily be all of C^n. Therefore I suspect the composition is not necessarily an isomorphism. One example you could consider, I think would be if f mapped from C^1 to C^2 where z mapped to (z,0). Then you could consider a g that maps from C^2 to C^1 mapping (z1,z2) to z2. The composition of these will map everything in C^1 to 0.
but we can't restrict domain of g, since 1) the domain is given as C^m and still, if we argue that f may restrict g's domain in the composition, see above for proof no.

uhggggg nothing worse than absolutely butchering a pretty simple problem.