JOVANA
Library Glossary Getting Started Three Levels Fields How it works Mission
Join the mission
All guides

The First Isomorphism Theorem: The Payoff

Everything in this track converges here. A map's kernel and image are not independent: the domain modulo the kernel is isomorphic to the image. This one theorem reorganizes rank-nullity, exact sequences, and the very idea of 'the same up to relabeling.'

Isomorphism: same up to relabeling

An invertible linear map between two spaces is an isomorphism — a perfect dictionary translating V to W with nothing lost or invented. When such a map exists we write V ≅ W and say the spaces are isomorphic. For finite-dimensional spaces the test is trivial: V ≅ W if and only if dim V = dim W. Dimension is the one and only invariant.

Now recall the quotient construction. Given a subspace U of V, the quotient V/U glues together any two vectors differing by an element of U, collapsing U to a single point. We can push a map down to a quotient: if T kills U, there is a well-defined induced map on V/U. These two ideas — isomorphism and quotient — are about to collide productively.

The theorem itself

Here is the centerpiece of the whole track. For any linear map T: V -> W, the first isomorphism theorem states V / ker T ≅ im T. In words: if you first collapse everything T kills, what remains is a perfect copy of the image. The map becomes invertible the moment you stop distinguishing inputs it could never tell apart.

  1. Start with T: V -> W. Its kernel is the ambiguity — distinct inputs T cannot separate.
  2. Form V / ker T, which treats two inputs as equal whenever they differ by a kernel vector.
  3. Define T-bar on the quotient by T-bar(v + ker T) = T(v); this is well-defined because the kernel ambiguity has been quotiented away.
  4. T-bar is now injective (no ambiguity left) and surjective onto im T, hence an isomorphism V / ker T ≅ im T.

And rank-nullity falls out for free. Take dimensions of both sides: dim(V / ker T) = dim(im T). Since dim(V / ker T) = dim V - dim(ker T), rearranging gives dim(ker T) + dim(im T) = dim V — the abstract rank-nullity of guide two, now a one-line corollary instead of a separate theorem.

A worked split

T: R^3 -> R^2,  T(x, y, z) = (x + y,  y + z)

Kernel:  x + y = 0 and y + z = 0  ->  (t, -t, t)
  ker T = span( (1, -1, 1) )      dim 1

Image:  T(e1)=(1,0), T(e2)=(1,1), T(e3)=(0,1)
  these span all of R^2            dim 2  (so T is onto)

First isomorphism theorem:
  R^3 / ker T  ≅  im T = R^2
  dim:  3 - 1   =   2            consistent

So collapsing the line span((1,-1,1)) inside R^3
leaves a 2-dim quotient that is a perfect copy of R^2.
Quotient by the kernel; what is left is an exact copy of the image.

The matrix was a shadow

Look back at the arc. We began by demoting the matrix to a coordinate snapshot of a map. We found the space of all maps, read injectivity and surjectivity off the kernel and image, met projections and nilpotents, and saw that basis changes are mere conjugation. The first isomorphism theorem is where it pays off: it shows the kernel and image are two halves of one structural truth, glued by an isomorphism that no choice of basis can disturb.

From here the road forks into the structure theorems. The same kernel-image thinking, applied to powers of an operator, gives generalized eigenspaces, the Jordan form, and the rational canonical form — each a way of choosing the basis in which an operator reveals its true shape. The matrix was always a shadow; you now know how to read the object casting it.