Subgroups, Cosets and Lagrange's Theorem
Source lecture(s): PHY622 Ch. 6
Intuition
Lagrange's theorem is the first genuinely powerful result in group theory, and it is startlingly restrictive: the order of a subgroup must divide the order of the group. No dynamics, no computation — just counting. A group of order 12 simply cannot contain a subgroup of order 5, and you know that before examining a single element.
Cosets
Given a subgroup \(H \le G\) and \(g \in G\), the left coset is
Three facts do all the work:
- Every coset has \(|H|\) elements — the map \(h\mapsto gh\) is a bijection.
- Two cosets are either identical or disjoint. If \(g_1H \cap g_2H \neq \emptyset\) then \(g_1H = g_2H\).
- The cosets cover \(G\) — every \(g\) lies in \(gH\), since \(e \in H\).
So the cosets partition \(G\) into equal-sized blocks, and immediately:
with \([G:H]\) the index, the number of cosets. That is Lagrange's theorem, and the proof is just "equal-sized blocks tile the set".
Immediate corollaries
- The order of any element divides \(|G|\). The powers of \(g\) form a cyclic subgroup of order equal to the order of \(g\).
- Every group of prime order is cyclic, and has no proper subgroups. There is essentially only one group of order 7.
- \(g^{|G|} = e\) for all \(g\). Specialised to modular arithmetic this is Fermat's little theorem — the same counting argument underlies RSA.
Normal subgroups and quotients
Left and right cosets need not agree. When they do, \(gH = Hg\) for all \(g\), the subgroup is normal (\(H \trianglelefteq G\)), and something valuable happens: the cosets themselves form a group, the quotient group \(G/H\), under \((g_1H)(g_2H) = g_1g_2H\). This is well defined only for normal subgroups.
Normal subgroups are exactly the kernels of homomorphisms, which is why they matter to physics:
- \(SU(2) \to SO(3)\) has kernel \(\{\pm I\} \cong \mathbb{Z}_2\), so \(SO(3) \cong SU(2)/\mathbb{Z}_2\). That \(\mathbb{Z}_2\) is the whole content of the 720° rotation of a spinor — the double cover is a quotient by a normal subgroup.
- Gauge theories are built on quotients: the physical configuration space is field space modulo the gauge group.
- Crystallography classifies space groups by their translation subgroups, which are normal.
Worked example: \(S_3\)
The symmetric group on 3 objects — equivalently the symmetries of an equilateral triangle — has order 6: the identity, three transpositions (reflections), two 3-cycles (rotations).
By Lagrange, subgroups can only have order 1, 2, 3 or 6. And indeed:
| Order | Subgroup | Normal? |
|---|---|---|
| 1 | \(\{e\}\) | yes |
| 2 | \(\{e, (12)\}\) and two others | no |
| 3 | \(\{e, (123), (132)\}\) — the rotations | yes (index 2) |
| 6 | \(S_3\) | yes |
No subgroup of order 4 or 5 — forbidden by divisibility, without any search.
The order-3 subgroup is normal because any subgroup of index 2 is (left and right cosets are both "\(H\) and everything else"), and \(S_3/A_3 \cong \mathbb{Z}_2\) — the even/odd permutation grading, which is where the sign of a determinant comes from.
The order-2 subgroups are not normal: \(S_3\) is the smallest non-abelian group, and its non-normal subgroups are the reason.
Common mistakes
- Reading Lagrange backwards. It says a subgroup's order divides \(|G|\). It does not say that every divisor has a subgroup — the converse is false (\(A_4\) has order 12 and no subgroup of order 6).
- Forming \(G/H\) for non-normal \(H\). The coset multiplication is not well defined; you get a set of cosets, not a group.
- Assuming left and right cosets coincide. True only for normal subgroups — i.e. always in an abelian group, and not in general otherwise.
Related concepts
- Symmetry groups — the definitions
- Group representations — where the structure gets used
- SU(2) → SO(3) — a quotient by a normal \(\mathbb{Z}_2\)
- Lie groups — the continuous case
Knowledge graph position
Prerequisites: symmetry groups. Leads to: quotient groups, representations, SU(2)→SO(3), crystallographic classification.
Quiz
Q1 (conceptual). A group has 15 elements. What subgroup orders are possible?
Answer
Only divisors of 15: 1, 3, 5 and 15. Orders 2, 4, 6, … are impossible by Lagrange. (In fact every group of order 15 is cyclic, but that needs more than Lagrange.)
Q2 (conceptual). Why can you only form a quotient group \(G/H\) when \(H\) is normal?
Answer
The product \((g_1H)(g_2H) = g_1g_2H\) must not depend on which representatives you pick. That well-definedness requires \(gH = Hg\) for all \(g\) — precisely normality. Without it, the cosets are a set with no consistent multiplication.
Q3 (MCQ). Lagrange's theorem implies that a group of prime order \(p\):
- (a) is abelian and cyclic, with no proper subgroups
- (b) has exactly \(p\) subgroups
- (c) is non-abelian
- (d) has a subgroup of every order less than \(p\)
Answer
(a). Subgroup orders must divide \(p\), so only 1 and \(p\) are available. Any non-identity element therefore generates the whole group, making it cyclic and hence abelian.