Group Homomorphisms and Quotient Groups

A group records a symmetry, but symmetries almost never live alone: one symmetry group maps onto another by forgetting some structure while keeping the multiplication law intact. The device that captures "keeps the multiplication law" is the group homomorphism, a map φ:G→H with φ(g₁g₂)=φ(g₁)φ(g₂). Once such a map exists, two complementary questions organize everything — what does it crush to the identity, and what does it reach — and the answers are the kernel and the image. The central discovery of elementary group theory is that these are rigidly linked: the source, divided by everything the map forgets, is isomorphic to everything the map hits. That single sentence is the First Isomorphism Theorem, G/ker φ ≅ im φ, the workhorse this article is built around. To divide a group we need a quotient, and to form a quotient the subgroup we factor out must be normal — invariant under conjugation, gNg⁻¹=N. The quiet miracle is that the subgroups arising as kernels are exactly the normal subgroups. Normality, which can look like a technical hypothesis, is revealed here as the precise condition for the coset space G/N to carry a well-defined multiplication. In physics this language is not optional: a linear representation is literally a homomorphism into a matrix group, the determinant and the permutation sign are one-dimensional homomorphisms, and the spin double cover SU(2)→SO(3) is a two-to-one homomorphism whose kernel {±I} is why a 2π rotation flips the sign of an electron's wavefunction. The running theme is that a homomorphism is a controlled loss of information — the kernel measures what is lost, the quotient repackages the source so the loss is undone, and the image is what survives. Mastering that triangle is the prerequisite for character theory, the classification of Lie groups, and the representation theory of particle physics.

Key takeaways

A group homomorphism φ:G→H satisfies φ(g₁g₂)=φ(g₁)φ(g₂); it automatically sends identity to identity and inverses to inverses.
The kernel ker φ (everything mapped to the identity) is always a normal subgroup of G; the image im φ is a subgroup of H.
Kernels and normal subgroups are the same objects: every normal N is the kernel of its projection π:G→G/N, and the coset product (aN)(bN)=(ab)N is well-defined precisely because N is normal.
The First Isomorphism Theorem, G/ker φ ≅ im φ, reads off a quotient concretely; for finite groups it forces |G| = |ker φ|·|im φ|.
Representations are homomorphisms ρ:G→GL(V); quotients explain covering maps like SU(2)→SO(3) and the 4π periodicity of spinors.
Key idea: a homomorphism's kernel is a normal subgroup, its image is a subgroup, and quotienting by the kernel recovers the image — G/ker φ ≅ im φ — so normal subgroups and kernels are one and the same class of objects.

The core insight

A homomorphism's kernel is a normal subgroup and its image is a subgroup, and the quotient by the kernel is isomorphic to the image: G/ker φ ≅ im φ. Normal subgroups and kernels are exactly the same class of objects.

Where this lives today

The kernel–image–quotient triangle is the daily language of anywhere symmetry meets physics. A linear representation is a homomorphism ρ:G→GL(V), and which elements act trivially — the kernel — decides whether a field is a faithful representation of the gauge group or secretly a representation of a quotient. The Standard Model's electroweak gauge group is such a quotient, $(SU(2)\times U(1))/\mathbb{Z}_2$ . The double cover SU(2)→SO(3), and its relativistic parent Spin(3,1)→SO(3,1), are why fermions exist as spinor fields and why a neutron interferometer shows a measurable phase reversal after a 2π rotation — first observed in 1975, a laboratory sighting of the kernel element −I. In condensed matter and cosmology, when a symmetry group $G$ breaks to $H$ the vacuum manifold is the coset space $G/H$ , whose topology classifies defects: vortices in superfluid helium, disclinations in liquid crystals, cosmic strings and monopoles in grand unified theories. The three-line theorem Noether polished a century ago is, quite literally, still doing physics.

The physical picture

Think of a homomorphism as a lossy compression of a symmetry: it keeps the multiplication structure — products go to products — but may blur distinct elements together. The set blurred all the way down onto the identity is the kernel, the "resolution" the map throws away. Everything in a single coset of the kernel is sent to one image element, so the cosets are the fibers of the map. Counting those fibers is the whole content of the First Isomorphism Theorem: the number of kernel cosets equals the size of the image, and the coset arithmetic is literally the same group as the image. The first figure makes this fibration concrete with real numbers.

Figure

A homomorphism as a fibration: G of order 12 splits into 4 cosets of the kernel, each of size 3, and each coset is crushed to one point of the image. The kernel coset (green, containing the identity e) maps to the identity of H. The order of G equals the kernel order times the image order, that is 12 = 3 × 4

Normality is the geometric fine print. The kernel must look the same no matter where you stand in the group — gNg⁻¹=N — because conjugation by g is the group's own way of "changing viewpoint," and a subgroup whose blurring is consistent across every viewpoint is exactly one that can be quotiented out. Everything physical about representations lives between two extremes: a faithful representation has trivial kernel and stores the whole group, while one with a nontrivial kernel is secretly a faithful representation of the smaller quotient. The most consequential physical example is a two-to-one map, drawn next.

Figure

The spin double cover SU(2)→SO(3) is a 2-to-1 homomorphism with kernel {+I, −I}. A rotation traced from 0 to 2π in SO(3) lifts to a path in SU(2) from +I to −I, so a spinor picks up ψ → −ψ; only a 4π rotation returns to +I with ψ → +ψ. By the First Isomorphism Theorem SU(2)/{±I} ≅ SO(3)

The core equations

The starting point is the defining property. A homomorphism respects multiplication — it need not be one-to-one, onto, or size-preserving, only structure-preserving. Everything else in this article is squeezed out of this one line.

\varphi(g_1 g_2)=\varphi(g_1)\,\varphi(g_2) \qquad (1)

where: $\varphi$ — a map $G\to H$ between two groups; $g_1,g_2$ — elements of the source group $G$ ; juxtaposition — the group product (in $G$ on the left, in $H$ on the right). Reading: the image of the product equals the product of the images, so multiplication before or after applying $\varphi$ gives the same answer. Two facts that look like extra axioms are in fact forced by (1) alone: a homomorphism carries identity to identity and inverses to inverses — you never have to require φ(e)=e.

\varphi(e_G)=e_H,\qquad \varphi(g^{-1})=\varphi(g)^{-1} \qquad (2)

where: $e_G,e_H$ — the identity elements of $G$ and $H$ ; $g^{-1}$ — the inverse of $g$ in $G$ . Reading: identity and inverse structure survive the map automatically, so the image is again a genuine group. The preimage of the identity is the kernel, and it is never just any subgroup — it is always a normal one, closed under conjugation by every element of $G$ . This is the fact that ties homomorphisms to the previous article's normal subgroups.

\ker\varphi=\{\,g\in G:\varphi(g)=e_H\,\}\ \trianglelefteq\ G \qquad (3)

where: $\ker\varphi$ — the kernel, the set sent to $e_H$ ; $\trianglelefteq$ — "is a normal subgroup of." Reading: the kernel measures exactly what the map forgets, and its normality (Derivation A) makes the quotient below possible. Dually, the set of values actually attained is the image, a subgroup of the target — generally a proper one, since $\varphi$ need not be onto.

\mathrm{im}\,\varphi=\{\,\varphi(g):g\in G\,\}\ \le\ H \qquad (4)

where: $\mathrm{im}\,\varphi$ — the image of $\varphi$ ; $\le$ — "is a subgroup of." Reading: what survives the compression is itself a group, so we may legitimately compare it to $G$ . Given a normal subgroup $N$ , cosets multiply representative-by-representative, and the answer does not depend on which representatives you pick. This product turns the set of cosets $G/N$ into a group — the quotient group.

(aN)(bN)=(ab)N \qquad (5)

where: $N\trianglelefteq G$ — a normal subgroup; $aN,bN$ — left cosets, the elements of the quotient $G/N$ ; $ab$ — the product of representatives in $G$ . Reading: multiply the labels and read off the coset of the product — well-defined precisely because $N$ is normal (Derivation A). Every normal subgroup is realized as a kernel by the map that sends each element to its own coset. This canonical projection is the universal homomorphism out of $G$ that kills exactly $N$ .

\pi:G\to G/N,\qquad \pi(g)=gN,\qquad \ker\pi=N \qquad (6)

where: $\pi$ — the canonical (quotient) projection; $gN$ — the coset of $g$ , an element of $G/N$ . Reading: because $\ker\pi=N$ for every normal $N$ , kernels and normal subgroups are the same class of objects. The First Isomorphism Theorem is the payoff. Quotienting the source by what the map forgets reproduces, on the nose, what the map sees.

G/\ker\varphi\ \cong\ \mathrm{im}\,\varphi \qquad (7)

where: $\cong$ — a group isomorphism (a bijective homomorphism). Reading: the abstract quotient $G/\ker\varphi$ and the concrete subgroup $\mathrm{im}\,\varphi\le H$ are the same group; this is how quotients are identified in practice (Derivation B). For finite groups, taking sizes in (7) and combining with Lagrange's theorem gives a rigid counting law relating the orders of the three players.

|G|=|\ker\varphi|\cdot|\mathrm{im}\,\varphi| \qquad (8)

where: $|\cdot|$ — the order (number of elements) of a finite group. Reading: the image order and the kernel order multiply to $|G|$ , so both must divide $|G|$ — a fast consistency check on any proposed homomorphism (Derivation B, final step). Two companions round out the toolkit: the Second Isomorphism Theorem, $NH/N\cong H/(N\cap H)$ , and the Third, $(G/N)/(K/N)\cong G/K$ for $N\subseteq K$ , which says a quotient of a quotient is again a single quotient — the fact that keeps iterated symmetry breaking and composition series well-behaved.

Derivation AWhy the kernel is normal and the coset product is well-defined

The goal is to prove the two structural facts that make quotient groups exist at all: first, that ker φ is a subgroup and is conjugation-invariant, so it is normal; and second, that the coset multiplication (aN)(bN)=(ab)N gives the same answer regardless of which representatives are chosen — and that this holds exactly when N is normal, not merely a subgroup. Together these show that "kernel of a homomorphism" and "normal subgroup" are one and the same class of objects, and that normality is the precise price of admission to a well-defined quotient.

Step 1. Take $a,b\in\ker\varphi$ . Then $\varphi(ab^{-1})=\varphi(a)\varphi(b)^{-1}=e_H$ , so $ab^{-1}\in\ker\varphi$ ; by the one-step subgroup criterion the kernel is a subgroup. Why: closure under "multiply-by-inverse" is all a nonempty subset needs to be a subgroup, and it follows straight from (1) and (2).

Step 2. Take any $g\in G$ and $n\in\ker\varphi$ . Then $\varphi(gng^{-1})=\varphi(g)\,e_H\,\varphi(g)^{-1}=e_H$ , so $gng^{-1}\in\ker\varphi$ ; hence $g(\ker\varphi)g^{-1}\subseteq\ker\varphi$ for every $g$ , the definition of a normal subgroup. Why: conjugation cannot move an element out of the kernel because $\varphi$ conjugates the identity to itself — normality is automatic, never an extra hypothesis.

Step 3. Now let $N\trianglelefteq G$ and suppose $aN=a'N$ and $bN=b'N$ , so $a'=an_1$ and $b'=bn_2$ with $n_1,n_2\in N$ . Compute $a'b'=an_1bn_2=ab\,(b^{-1}n_1b)\,n_2$ , inserting $b^{-1}b$ between the two factors. Why: to test whether the product of cosets depends on the representatives, we must express the "primed" product in terms of the original ab times something we hope lands in N.

Step 4. Normality says $b^{-1}Nb=N$ , so $b^{-1}n_1b\in N$ and hence $(b^{-1}n_1b)\,n_2\in N$ . Thus $a'b'\in(ab)N$ : the coset product is independent of representatives. Why: the lone place normality is used is to slide $n_1$ past $b$ and keep it inside N — for a non-normal subgroup $b^{-1}n_1b$ escapes N and the product (5) is genuinely ill-defined, the whole reason only normal subgroups can be quotiented.

Go deeper · Derivation B · 1 minThe First Isomorphism Theorem

Derivation BThe First Isomorphism Theorem

The goal is to construct an explicit isomorphism from the quotient by the kernel onto the image, proving G/ker φ ≅ im φ. Given φ:G→H we define a candidate map on cosets, then verify the four things an isomorphism demands: well-defined (independent of representative), a homomorphism, injective, and surjective onto the image. The construction also yields the finite counting law |G| = |ker φ|·|im φ| as a corollary through Lagrange's theorem.

Step 1. Write $K=\ker\varphi$ and define $\bar\varphi:G/K\to\mathrm{im}\,\varphi$ by $\bar\varphi(gK)=\varphi(g)$ . If $gK=g'K$ then $g'=gn$ with $n\in K$ , so $\varphi(g')=\varphi(g)\varphi(n)=\varphi(g)\,e_H=\varphi(g)$ . Why: the value must depend only on the coset, not on the label we picked for it; the kernel absorbs the ambiguity because everything in it maps to $e_H$ .

Step 2. For cosets $aK,bK$ , $\bar\varphi\big((aK)(bK)\big)=\bar\varphi(abK)=\varphi(ab)=\varphi(a)\varphi(b)=\bar\varphi(aK)\,\bar\varphi(bK)$ , so $\bar\varphi$ is a homomorphism. Why: the quotient multiplication (5) was defined precisely so that projecting and then multiplying agrees with multiplying and then projecting — $\bar\varphi$ simply inherits (1).

Step 3. Suppose $\bar\varphi(gK)=e_H$ . Then $\varphi(g)=e_H$ , so $g\in K$ , hence $gK=K$ , the identity coset. The only coset in the kernel of $\bar\varphi$ is the trivial one, so $\bar\varphi$ is injective. Why: injectivity of a homomorphism is equivalent to a trivial kernel, and G/K is arranged so that everything φ forgot has already been divided out.

Step 4. Every element of $\mathrm{im}\,\varphi$ is $\varphi(g)=\bar\varphi(gK)$ , so $\bar\varphi$ is onto its codomain. Being a bijective homomorphism, it is an isomorphism: $G/K\cong\mathrm{im}\,\varphi$ . Taking orders for finite $G$ , $|G/K|=|\mathrm{im}\,\varphi|$ , and Lagrange gives $|G|=[G:K]\,|K|=|\mathrm{im}\,\varphi|\cdot|\ker\varphi|$ . Why: surjectivity is true by construction, and the order relation (8) drops out for free — the index of the kernel is the size of the image.

Go deeper · Derivation C · 2 minThe spin double cover SU(2) to SO(3)

Derivation CThe spin double cover SU(2) to SO(3)

The goal is to build, by hand, a surjective homomorphism Φ:SU(2)→SO(3) whose kernel is exactly {+I, −I}, then feed it to the First Isomorphism Theorem to conclude SU(2)/{±I} ≅ SO(3). The construction identifies a real three-vector with a traceless Hermitian 2×2 matrix and lets SU(2) act by conjugation; because conjugation preserves the determinant, which is minus the squared length of the vector, the action is a rotation. This one homomorphism is the algebraic reason spin-½ states have 4π periodicity and a 2π rotation multiplies a spinor by −1.

Step 1. Map a vector $\mathbf{x}=(x_1,x_2,x_3)$ to $X=\mathbf{x}\cdot\boldsymbol\sigma=\begin{pmatrix}x_3 & x_1-ix_2\\ x_1+ix_2 & -x_3\end{pmatrix}$ with the Pauli matrices $\boldsymbol\sigma$ . This $X$ is traceless and Hermitian, and $\det X=-x_3^2-(x_1^2+x_2^2)=-|\mathbf{x}|^2$ . Why: encoding the vector as a matrix turns its length into the algebraic invariant $-\det X$ , which the group action will preserve.

Step 2. For $U\in SU(2)$ define the action $X\mapsto UXU^\dagger$ . The result is again traceless (trace is cyclic) and Hermitian, and $\det(UXU^\dagger)=\det X$ since $|\det U|=1$ . So the action sends vectors to vectors of the same length, defining a rotation $R(U)\in O(3)$ ; continuity from $U=I$ (where $R=\mathbb{1}$ ) forces $\det R=+1$ , so $R(U)\in SO(3)$ . Why: preserving $-\det X=|\mathbf{x}|^2$ is exactly the statement that $R(U)$ is orthogonal, and connectedness pins down the proper (determinant $+1$ ) branch.

Step 3. The map $\Phi:U\mapsto R(U)$ is a homomorphism: $(U_1U_2)X(U_1U_2)^\dagger=U_1(U_2XU_2^\dagger)U_1^\dagger$ , so $R(U_1U_2)=R(U_1)R(U_2)$ . Why: conjugation by a product is the composition of conjugations, so the matrix action respects multiplication and $\Phi$ is a genuine homomorphism, not just a map of sets.

Step 4. $U\in\ker\Phi$ means $UXU^\dagger=X$ for every traceless Hermitian $X$ , so $U$ commutes with every $2\times2$ matrix; by Schur's argument $U=\lambda I$ , and $U\in SU(2)$ forces $\det U=\lambda^2=1$ , so $\lambda=\pm1$ and $\ker\Phi=\{\pm I\}$ . Why: only the two scalar matrices act trivially, which makes the map two-to-one — every rotation has exactly two SU(2) preimages differing by a sign.

Step 5. $\Phi$ is surjective (every rotation is $R(U)$ for some $U=e^{-i\theta\,\hat{\mathbf{n}}\cdot\boldsymbol\sigma/2}$ ), so the First Isomorphism Theorem gives $SU(2)/\{\pm I\}\cong SO(3)$ . Since a $2\pi$ rotation corresponds to $U=-I$ , a spinor transforms as $\psi\mapsto-\psi$ , returning to $+\psi$ only after $4\pi$ . Why: the nontrivial kernel element $-I$ projects to the identity rotation yet acts as $-1$ on spinors — the observable 4π periodicity of spin-½.

Limiting cases

The extremes and special cases sharpen what the kernel, image, and quotient actually measure. Each row is a homomorphism whose kernel or image is pushed to a boundary.

Case	Result	Meaning
Injective (faithful)	$\ker\varphi=\{e\}$	No information lost; $\varphi$ is one-to-one and $G\cong\mathrm{im}\,\varphi$ . A faithful representation stores the entire group.
Surjective (onto)	$\mathrm{im}\,\varphi=H$	Then $G/\ker\varphi\cong H$ , so $H$ is realized concretely as a quotient of $G$ by a normal subgroup.
Trivial homomorphism	$\varphi(g)=e_H\ \forall g$	Kernel is all of $G$ , image is a point; $G/G\cong\{e\}$ . The maximally lossy map — the trivial representation.
Isomorphism	$\ker\varphi=\{e\},\ \mathrm{im}\,\varphi=H$	Both injective and surjective: $G\cong H$ , a perfect dictionary translating one group into the other.
Abelian target	$[G,G]\subseteq\ker\varphi$	Any map to an abelian $H$ kills all commutators, so it factors through the abelianization $G/[G,G]$ — the largest abelian quotient.

Common trapAny subgroup can be divided out to form a quotient group.

It cannot — only normal subgroups give a well-defined quotient. The coset product $(aN)(bN)=(ab)N$ is independent of the representatives precisely when $gNg^{-1}=N$ for all $g$ ; for a non-normal subgroup the factor $b^{-1}n_1b$ escapes $N$ and the product depends on which coset labels you wrote down, so $G/N$ carries no consistent multiplication. Concretely, in $S_3$ the subgroup $H=\{e,(12)\}$ is not normal, so " $S_3/H$ " is only a set of three cosets with no group structure, whereas $A_3=\{e,(123),(132)\}$ is normal and $S_3/A_3\cong\mathbb{Z}_2$ is a genuine group. Cosets always partition the group, but they multiply as a group only when the subgroup is normal. (A related trap: $\ker\varphi$ is a subgroup of the source $G$ , while $\mathrm{im}\,\varphi$ is a subgroup of the target $H$ — they live in different groups and should never be compared directly.)

Worked examples

1 · The sign homomorphism on S₃ (confidence builder)

Let $\mathrm{sgn}:S_3\to\{+1,-1\}$ send each permutation to $+1$ if it is even and $-1$ if it is odd. Identify the kernel and image, form the quotient, and verify the counting law explicitly.

Step 1. List the six elements of $S_3$ with their parities: the identity $e$ (even), the three transpositions $(12),(13),(23)$ (odd), and the two 3-cycles $(123),(132)$ (even). So $\mathrm{sgn}$ sends $\{e,(123),(132)\}\mapsto+1$ and $\{(12),(13),(23)\}\mapsto-1$ ; it is a homomorphism because parity adds under composition.

Step 2. The kernel is the set mapped to $+1$ , namely $\ker(\mathrm{sgn})=\{e,(123),(132)\}=\mathbf{A_3}$ , the alternating group, of order 3. The image is $\{+1,-1\}\cong\mathbb{Z}_2$ , of order 2. As guaranteed, $A_3\trianglelefteq S_3$ — it has index 2, and every index-2 subgroup is normal.

Step 3. The quotient $S_3/A_3$ has two cosets: $A_3$ itself (the even permutations) and its complement (the odd permutations). Multiplying cosets follows even·even = even, even·odd = odd, odd·odd = even — exactly the $\mathbb{Z}_2$ table. So $S_3/A_3\cong\mathbb{Z}_2$ , matching the First Isomorphism Theorem $S_3/\ker(\mathrm{sgn})\cong\mathrm{im}(\mathrm{sgn})$ .

Step 4. Check the counting law (8): $|S_3|=|\ker|\cdot|\mathrm{im}|=3\times2=6$ . Kernel A₃ of order 3, image ℤ₂ of order 2, and S₃/A₃ ≅ ℤ₂ with 6 = 3 × 2. Physically this is the exchange-statistics dichotomy in miniature: the sign character's two values $+1$ and $-1$ are the bosonic and fermionic behaviours under particle exchange.

2 · Reduction modulo n: Z/nZ ≅ Zₙ

Let $\varphi:\mathbb{Z}\to\mathbb{Z}_n$ send an integer $m$ to its remainder $m\bmod n$ , where $\mathbb{Z}_n=\{0,1,\dots,n-1\}$ under addition mod $n$ . Find the kernel and apply the First Isomorphism Theorem.

Step 1. $\varphi$ is a homomorphism of additive groups: $(a+b)\bmod n=(a\bmod n)+(b\bmod n)$ , and it is surjective since every residue $0,\dots,n-1$ is hit.

Step 2. The kernel is the set of integers with remainder 0, the multiples of $n$ : $\ker\varphi=n\mathbb{Z}$ — automatically normal since $\mathbb{Z}$ is abelian.

Step 3. The First Isomorphism Theorem gives $\mathbb{Z}/n\mathbb{Z}\cong\mathrm{im}\,\varphi=\mathbb{Z}_n$ : it proves that the abstract quotient of the infinite group $\mathbb{Z}$ by the subgroup $n\mathbb{Z}$ is the finite cyclic group $\mathbb{Z}_n$ . Setting $n=12$ recovers clock arithmetic — the hours form $\mathbb{Z}/12\mathbb{Z}\cong\mathbb{Z}_{12}$ , and " $7+8=3$ o'clock" is the coset identity $(7+8)+12\mathbb{Z}=3+12\mathbb{Z}$ . The infinite source collapses onto the n-element image, and the kernel nℤ is exactly what is divided out.

3 · Constraining a homomorphism by orders

A homomorphism $\varphi:G\to H$ has $|G|=12$ and $|\mathrm{im}\,\varphi|=4$ . What is $|\ker\varphi|$ , and which image orders are even possible for a group of order 12?

Step 1. By the counting law (8), $|\ker\varphi|=|G|/|\mathrm{im}\,\varphi|=12/4=\mathbf{3}$ . So the kernel is a normal subgroup of order 3 — necessarily cyclic, $\cong\mathbb{Z}_3$ — and $G/\ker\varphi\cong\mathrm{im}\,\varphi$ is a group of order 4.

Step 2. Since $|\mathrm{im}\,\varphi|=|G|/|\ker\varphi|$ with the kernel a subgroup, the image order must divide 12 — it can only be $1,2,3,4,6,12$ . An image of order 5 or 7 is impossible for any homomorphism out of a group of order 12, a one-line sanity check. The endpoints are the familiar extremes: order 12 means a trivial kernel (faithful, $G\cong H$ ), order 1 means the whole group is the kernel (trivial map), and our case sits strictly between. |ker φ| = 3, and the image order must be one of 1, 2, 3, 4, 6, 12.

Physics of the portrait

The homomorphism theorems reached their clean modern form with Emmy Noether, who insisted that the right objects of study are not a group's elements but the maps between groups.

Origins · 1 minEmmy Noether and how this was discovered

Emmy Noether

1882–1935, German mathematician · the isomorphism theorems and the homomorphism view of abstract algebra

Emmy Noether is remembered by physicists chiefly for Noether's theorem — every continuous symmetry yields a conservation law — but her deeper legacy is the abstract, map-centered style of algebra that makes this article possible. Working at Göttingen in the 1920s, often without salary or an official position because she was a woman, she reorganized group and ring theory around homomorphisms and their kernels rather than explicit elements and generators. The isomorphism theorems, in the crisp form G/ker φ ≅ im φ and its companions, are essentially her formulation; van der Waerden's influential 1930–31 textbook Moderne Algebra, which taught the subject to a generation, was built from her lectures. The seeds are older — Évariste Galois had introduced normal subgroups around 1831, and Jordan and Hölder had proved the composition-series theorem using the quotient construction — but it was Noether who made "pass to the quotient by the kernel" a reflex rather than a trick, and who saw that the same three-line theorem governs abelian groups, matrix groups, rings, and modules alike. The physics payoff is everywhere her algebra is: a representation is a homomorphism, a superselection rule is a kernel, a broken symmetry leaves a quotient group, and the spinor sign that puzzled the founders of quantum mechanics is the nontrivial element of a kernel {±I}. Einstein, writing after her death, called her the most significant creative mathematical genius produced since the higher education of women began.

History

1831

— Galois introduces normal subgroups to characterize solvable equations

1870

— Jordan systematizes permutation groups and the germ of quotient constructions

1889

— Hölder completes the Jordan–Hölder theorem on composition factors

1927

— Noether formulates the isomorphism theorems in their modern homomorphism form

1931

— van der Waerden's Moderne Algebra spreads the map-centered viewpoint worldwide

Going further

Texts. Stone & Goldbart, Mathematics for Physics, group-theory chapter · Arfken, Weber & Harris, Mathematical Methods for Physicists, 7th ed., group-theory chapter · Hassani, Mathematical Physics, 2nd ed., group theory · Riley, Hobson & Bence, Mathematical Methods for Physics and Engineering, 3rd ed., Ch. 28 · Tinkham, Group Theory and Quantum Mechanics, Ch. 3 · Hamermesh, Group Theory and Its Application to Physical Problems. Key results throughout: $\varphi(e_G)=e_H$ from the homomorphism property (1); the kernel is closed under $ab^{-1}$ and conjugation, and the coset product is well-defined iff $b^{-1}Nb=N$ (Derivation A); the four-part construction establishes $G/\ker\varphi\cong\mathrm{im}\,\varphi$ with corollary $|G|=|\ker\varphi||\mathrm{im}\,\varphi|$ (Derivation B); $\det X=-|\mathbf{x}|^2$ and $\ker\Phi=\{\pm I\}$ (Derivation C). Worked numbers: $S_3$ has $|A_3|=3$ , image order 2, and $6=3\times2$ ; $\mathbb{Z}/n\mathbb{Z}\cong\mathbb{Z}_n$ ; and $12/4=3$ .

Prerequisites. Groups, Subgroups, and Cosets · Conjugacy Classes and Normal Subgroups. Leads to. Group Representations · Schur's Lemmas · SU(2) and the Spin Double Cover · Lie Groups.

Ask the tutor: Ask me to walk through why a non-normal subgroup like {e, (12)} in S₃ fails to form a quotient, or to trace the spinor sign flip through the SU(2)→SO(3) kernel step by step.

What exactly is the difference between a homomorphism and an isomorphism?

A homomorphism is any map φ:G→H that preserves multiplication, φ(g₁g₂)=φ(g₁)φ(g₂); it may lose information by sending several elements to one, and it need not reach all of H. An isomorphism is the special case that loses nothing and misses nothing — both injective (trivial kernel) and surjective (image all of H), so it is a bijection whose inverse is also a homomorphism. In kernel–image language, φ is an isomorphism exactly when ker φ = {e} and im φ = H. An isomorphism means G and H are "the same group with different labels"; a general homomorphism only means H contains a compressed copy of G, the quotient G/ker φ.

Why must the subgroup be normal to form a quotient — what actually breaks otherwise?

The elements of the quotient are the cosets, and you multiply them through representatives: (aN)(bN)=(ab)N. For this to define anything, the answer must not depend on which representatives you chose. Replacing a by an₁ and b by bn₂ (same cosets), the product becomes ab·(b⁻¹n₁b)·n₂, and you need the middle factor b⁻¹n₁b to stay inside N — precisely the normality condition b⁻¹Nb=N. Without it, two people writing the same two cosets could multiply them and get different "products," so the operation is not a function and there is no group. Cosets of a non-normal subgroup still partition G, but only as a set, never as a group.

How can two different groups have the same kernel structure yet not be isomorphic?

Because the isomorphism type of G is fixed by both the kernel and the image together, glued by an extension. Knowing that a normal subgroup of order 3 sits inside G with quotient of order 4 tells you |G|=12, but not which group of order 12 you have — the cyclic $\mathbb{Z}_{12}$ , the dihedral $D_6$ , and the alternating $A_4$ can share such data yet differ. The First Isomorphism Theorem reconstructs the image from G and the kernel, not G from the kernel and image; rebuilding G needs extra "extension data" describing how the quotient acts on the kernel. So the kernel–image pair constrains G tightly but does not always determine it.

Is the image of a homomorphism ever a normal subgroup of the target?

Not in general. The kernel is always normal in the source, but the image is only guaranteed to be a subgroup of the target, and it can be a decidedly non-normal one — embedding $S_2=\{e,(12)\}$ into $S_3$ gives a non-normal image. The image is normal in two common situations: when φ is surjective (the image is all of H), and when the target is abelian (every subgroup is normal). Keeping the asymmetry straight — kernel normal in G, image merely a subgroup of H — prevents a great many false steps.

What does it mean physically when a representation has a nontrivial kernel?

It means the representation cannot tell certain group elements from the identity — they act as the identity operator on every state. The representation is then not faithful, and by the First Isomorphism Theorem it is really a faithful representation of the smaller quotient G/ker ρ. A physical example: the integer-spin (tensor) representations of SU(2) all contain −I in their kernel, so they are genuine representations of the quotient SU(2)/{±I}=SO(3), while half-integer-spin (spinor) representations have −I acting as −1 and so are faithful for SU(2) and only projective for SO(3). The kernel is thus the precise bookkeeping of which symmetries a given field can actually feel.

Why do physicists care about SU(2)/{±I} ≅ SO(3) rather than just using SO(3)?

Because nature uses the covering group, not the rotation group, whenever spin is half-integer. SO(3) is the group of physical rotations, but its representations exclude the spin-½ states electrons demonstrably occupy — those require the double cover SU(2), where a 2π rotation is the distinct element −I rather than the identity. The quotient statement SU(2)/{±I}≅SO(3) says SU(2) carries exactly one extra bit beyond ordinary rotations, the sign −I, and that bit is measured directly in neutron interferometry, where a spin-½ beam rotated by 2π returns 180° out of phase with an unrotated reference. Working in SU(2) and tracking the kernel {±I} is the only way to describe both bosons and fermions with one symmetry group.