1 Definition and Basic Notation

1.1 Kernel of a general function

Let \(f:X\to Y\) be a function between sets, and fix an element \(y\in Y\). The kernel at \(y\) is the collection of inputs that produce \(y\): \[ \ker_y(f)=\{x\in X: f(x)=y\}. \] In many algebraic settings the distinguished value is \(0\in Y\), so the “kernel of \(f\)” is shorthand for the preimage of \(0\).

1.2 Kernel of a homomorphism

When \(f\) is a homomorphism of algebraic structures (such as groups, rings, modules, or vector spaces), the codomain contains a distinguished identity element (written \(0\) for additive notation, or \(1\) for multiplicative notation). The kernel is the set of elements mapped to the identity:

  • For additive structures: \(\ker(f)=\{x: f(x)=0\}\).
  • For multiplicative structures, one often uses \(\ker(f)=\{x: f(x)=1\}\); this can be translated to additive form when the structure supports it.

Common conventions include:

  • \(\ker(f)\) for the kernel of a homomorphism into an additive object.
  • \(\ker(f)\) or \(\{x: f(x)=0\}\) in contexts where \(0\) is understood.
  • \(\ker_y(f)\) when the target value needs to be specified explicitly.

In linear algebra and module theory, \(\ker(f)\) is frequently used without further qualification.

2 Kernels in Linear Algebra

2.1 Kernel of a linear transformation

Let \(T:V\to W\) be a linear transformation between vector spaces over a field. The kernel is \[ \ker(T)=\{v\in V: T(v)=0\}. \] It is a subspace of \(V\). The kernel measures exactly which vectors are “collapsed” to the zero vector by \(T\).

2.2 Geometric and algebraic interpretations

2.2.1 Dimension of the kernel

The dimension \(\dim(\ker(T))\) is an invariant of the transformation. Geometrically, the kernel is the set of directions in \(V\) that get sent to the origin in \(W\); algebraically, it encodes the degrees of freedom in solutions to \(T(v)=0\).

2.2.2 Rank–nullity perspective

If \(V\) is finite-dimensional, the rank–nullity theorem states \[ \dim(V)=\operatorname{rank}(T)+\dim(\ker(T)). \] Thus, knowing the kernel dimension determines how much “information” the map retains (rank) versus discards (nullity).

2.3 Relationship to solutions of linear systems

2.3.1 Homogeneous systems and kernel vectors

If \(T\) is represented by a matrix \(A\) (so \(T(x)=Ax\)), then \[ \ker(T)=\{x: Ax=0\}, \] which is precisely the solution space of the homogeneous linear system.

2.3.2 Particular parameterizations via basis

Once a basis \(\{v_1,\dots,v_k\}\) for \(\ker(T)\) is chosen, every kernel element can be written uniquely as a linear combination \[ x=\alpha_1 v_1+\cdots+\alpha_k v_k. \] This provides a parameterization of all solutions to the homogeneous system and often serves as the starting point for solving nonhomogeneous equations.

3 Kernels in Abstract Algebra

3.1 Kernels of group homomorphisms

Let \(\varphi:G\to H\) be a group homomorphism. The kernel is \[ \ker(\varphi)=\{g\in G:\varphi(g)=e_H\}, \] where \(e_H\) is the identity in \(H\). The kernel is always a normal subgroup of \(G\), reflecting that elements with the same “effect” under \(\varphi\) form a compatible quotient structure.

3.1.1 Normal subgroups and the kernel

Normality guarantees that \(G/\ker(\varphi)\) is a well-defined group. This quotient often captures the image of \(\varphi\) via the isomorphism theorems.

3.2 Kernels of ring homomorphisms

For a ring homomorphism \(\psi:R\to S\) (with additive identity preserved), the kernel is \[ \ker(\psi)=\{r\in R:\psi(r)=0_S\}. \] It is an ideal of \(R\), reflecting compatibility with ring multiplication: if \(r\) maps to \(0\), then so do all products \(ar\) and \(ra\).

3.2.1 Ideals as kernels

Many ideals arise as kernels of suitable homomorphisms. This correspondence underpins the standard technique of studying ring quotients through kernels.

3.3 Kernels of module homomorphisms

Let \(f:M\to N\) be a homomorphism of modules over a ring \(R\). Then \[ \ker(f)=\{m\in M: f(m)=0\}. \] The kernel is a submodule of \(M\). As in the vector space case, it measures the solution space of the homogeneous “equation” \(f(m)=0\), but now in the module setting.

3.3.1 Submodules and induced structure

Because the kernel is a submodule, quotient constructions \(M/\ker(f)\) become meaningful in the category of \(R\)-modules and connect to the structure of maps through induced morphisms.

4 Kernel Properties and Structure

4.1 Substructure results (subspace/submodule/ideal)

Depending on the algebraic category, kernels inherit structure:

  • Vector spaces: kernels are subspaces.
  • Modules: kernels are submodules.
  • Groups: kernels are normal subgroups.
  • Rings: kernels are ideals.

These results follow from the defining homomorphism properties and the presence of a distinguished identity element in the codomain.

4.2 Basic algebraic identities involving kernels

Kernels interact with preimages in systematic ways. For instance, for a homomorphism \(f\) and \(0\) the target identity: \[ \ker(f)=f^{-1}(\{0\}). \] Moreover, for any two homomorphisms \(f,g\) where the difference makes sense, kernel conditions can often be rewritten as equations, e.g. \(f(x)=g(x)\) corresponds to membership in \(\ker(f-g)\) in additive contexts.

4.3 Monotonicity under composition

If \(f:X\to Y\) and \(g:Y\to Z\) are homomorphisms, then \[ \ker(f)\subseteq f^{-1}(\ker(g\circ f)). \] A common useful consequence is:

  • Elements annihilated by \(f\) certainly get annihilated by the composite \(g\circ f\).

So the kernel of the composite contains the kernel of the first map when identities align appropriately.

4.4 Behavior under restriction and quotient constructions

Kernels behave predictably under restriction:

- If \(U\subseteq X\), then \(\ker(f_U)=U\cap \ker(f)\).

Kernels also fit naturally with quotients:

  • If \(K\subseteq \ker(f)\), then \(f\) factors through the quotient map \(X\to X/K\), producing an induced homomorphism with kernel related to the original one.

5 Kernel and Isomorphism Theorems

5.1 First Isomorphism Theorem (homomorphism form)

For a homomorphism \(f:A\to B\), the First Isomorphism Theorem gives \[ A/\ker(f)\cong \operatorname{im}(f), \] where \(\operatorname{im}(f)\) is the image. The kernel identifies which elements of \(A\) become indistinguishable under \(f\), and the quotient captures exactly the part of \(B\) that \(f\) can reach.

5.2 Quotients by kernels

5.2.1 Induced maps and well-definedness

Define the quotient projection \(\pi:A\to A/\ker(f)\). Under the homomorphism theorem, there is a unique induced map \(\tilde f:A/\ker(f)\to \operatorname{im}(f)\) satisfying \(\tilde f(\pi(a))=f(a)\). Well-definedness relies precisely on the fact that \(a\) and \(a'\) differ by an element of \(\ker(f)\) exactly when \(f(a)=f(a')\).

5.3 Computing quotients using kernels

In practice, one can:

  1. Determine \(\ker(f)\).
  2. Form the quotient \(A/\ker(f)\).
  3. Identify \(\operatorname{im}(f)\) with that quotient via the induced isomorphism.

This method often simplifies classification problems by replacing complicated maps with structural quotients.

6 Computing Kernels

6.1 Techniques for linear maps (row reduction)

6.1.1 Solving \(Ax=0\)

Given a matrix \(A\) representing \(T(x)=Ax\), compute \(\ker(T)\) by solving the homogeneous system \(Ax=0\). Row reduction (or equivalent elimination techniques) transforms the system into a form where free variables and pivot variables are clear.

6.1.2 Extracting a basis for the kernel

After row reduction, one typically expresses solutions as \[ x = \text{(linear combination of parameters)}. \] The kernel basis is obtained by setting each free parameter to \(1\) while others are \(0\), producing vectors spanning all solutions. The number of basis vectors equals \(\dim(\ker(T))\) when working over a field.

6.2 Kernels over modules and rings

6.2.1 Presentations and generators

For module homomorphisms, kernels are often computed from presentations. If \(M\) is given by generators and relations and \(f\) is described on generators, then kernel elements correspond to relations that map to zero in \(N\). Generating sets for the kernel can be extracted by solving the induced constraints.

6.2.2 Using relations to identify kernel elements

Because modules and rings may lack field structure, computation may rely on manipulating relations rather than performing Gaussian elimination. Strategies include:

  • Translating \(f(m)=0\) into constraints on coefficients of a chosen generating set.
  • Using syzygies (dependencies) among relations to build the kernel.

7 Special Cases and Examples

7.1 Trivial kernel and injectivity

For linear maps, \(T\) is injective precisely when \(\ker(T)=\{0\}\). More generally in homomorphism contexts, a trivial kernel means no non-identity element is sent to the target identity, so the map collapses nothing.

7.2 Full kernel and zero maps

If \(f\) is the zero homomorphism, then every element maps to the identity, so the kernel equals the entire domain (\(\ker(f)=A\)). Conversely, if the kernel is all of \(A\), the map must be zero.

7.3 Nontrivial kernels as invariants

A nonzero kernel indicates the presence of structure that is lost under the map. In algebraic classification, kernel size, shape, or additional invariants (such as submodule structure) can distinguish non-isomorphic objects or non-equivalent morphisms.

7.4 Worked example set (linear, group, and ring contexts)

  • Linear: Let \(T:\mathbb{R}^3\to\mathbb{R}^2\) be represented by

\[ A=\begin{pmatrix}1&2&0\\ 0&1&1\end{pmatrix}. \] Solving \(Ax=0\) yields the solution space spanned by one vector \(v\), so \(\ker(T)=\operatorname{span}\{v\}\).

  • Group: For \(\varphi:\mathbb{Z}\to \mathbb{Z}_n\) given by reduction mod \(n\), the kernel is \(n\mathbb{Z}\), the set of integers mapping to \(0\) in \(\mathbb{Z}_n\).
  • Ring: For \(R=\mathbb{Z}\) and the homomorphism \(\psi:\mathbb{Z}\to \mathbb{Z}_n\) by reduction mod \(n\), the kernel is the ideal \(n\mathbb{Z}\). In general, quotient rings \(\mathbb{Z}/n\mathbb{Z}\) arise as domains modulo kernels.

8 Kernel in Category-Theoretic Language

8.1 Kernels as categorical limits (zero morphisms)

In categories with a zero object, a kernel of a morphism \(f:A\to B\) can be defined as an equalizer of \(f\) and the zero morphism \(0_{A,B}\). Concretely, a kernel is a morphism \(k:K\to A\) such that \(f\circ k=0\), and it is universal among maps into \(A\) that become zero after composing with \(f\).

8.2 Kernels in abelian categories

In an abelian category, kernels exist and behave like subobjects in additive contexts. They are compatible with exact sequences and support algebraic operations such as forming quotients, images, and coimages, mirroring classical module theory.

8.3 Relation to exact sequences

Kernels are the starting point of exactness: in a sequence \[ A\xrightarrow{f}B\xrightarrow{g}C, \] exactness at \(B\) means \(\operatorname{im}(f)=\ker(g)\). Thus, category-theoretic kernels provide a systematic way to express “what maps to zero” inside compositional structures.

9 Exactness and Kernel–Image Interplay

9.1 Image and kernel in exact sequences

Given a sequence of homomorphisms \(A\xrightarrow{f}B\xrightarrow{g}C\), the containment \[ \operatorname{im}(f)\subseteq \ker(g) \] holds whenever \(g\circ f=0\). Exactness at \(B\) strengthens this by asserting equality.

9.2 Criteria for exactness via kernels

A sequence is exact at \(B\) precisely when every element of \(B\) that maps to zero under \(g\) actually comes from \(A\) via \(f\). Equivalently, the kernel of the next map captures the range of the previous one. This viewpoint turns computations of kernels into structural information about sequences.

9.3 Cycles, boundaries, and cohomology viewpoints (intuition level)

In homological algebra, one organizes data into cycles (kernel-like objects) and boundaries (image-like objects). Informally, cycles are elements annihilated by a differential, while boundaries are those produced by applying the differential one step earlier. Cohomology then measures the difference between cycles and boundaries, explaining why kernel–image relationships govern invariants computed from exactness patterns.