Motivation
Knowing the order of a finite group seldom determines it. Even among abelian groups, cyclic pieces can be assembled in different ways: a pair of elements of order eight behaves differently from elements of orders eight, four and two, despite the same total size. Classification needs information that does not depend on a convenient choice of generators.
Smith form converts a presentation into cyclic summands. Primary subgroups and their successive quotients then explain why the summands are intrinsic. We will separate these two tasks: existence comes from a proved matrix theorem, while uniqueness is recovered from subgroups defined directly inside the original group. This also supplies practical invariants such as the minimum number of generators and the exponent.
We write abelian groups additively. The PID module structure theorem and its Smith-form proof are named prior results used here, rather than being assumed without an available proof. The chapter on Euclidean domains, PIDs and UFDs established the required PID facts for the integers. The present chapter specializes that theorem and develops the finite-group consequences in detail; this cross-reference is a mathematical dependency, not an official course-prerequisite assertion.
Definitions
Definition 7.1.1
(Two forms of a finite abelian decomposition)
For a prime , let be the subgroup of elements whose orders are powers of , including the identity. A primary cyclic decomposition writes as a direct sum of groups with positive exponents. These prime-power cyclic factors are called elementary divisors.
An invariant-factor decomposition is
For the trivial group the list is empty. A direct sum of finitely many groups is also their direct product; we use the sum notation to emphasize additive coordinates.
An abelian group becomes a -module by integer repeated addition. Negative integers act by taking additive inverses, and the module distributive laws use commutativity of the group. A finite group is finitely generated, for example by all its elements. These observations are the precise bridge to the PID theorem, not an identification of arbitrary nonabelian groups with integer modules.
The exponent of a finite group is the least positive integer killing every element. Lagrange shows that the group order is one such integer. The exponent need not equal the order, and having a particular element order does not by itself describe how many cyclic directions the group contains.
Theorems
Theorem 7.1.2
(Finite abelian classification)
Every finite abelian group has a unique invariant-factor decomposition, up to isomorphism of the displayed cyclic factors. Equivalently, it is the direct sum of its intrinsic primary subgroups, and each has a unique multiset of positive cyclic exponents .
The two forms are converted by sorting each prime's exponents increasingly, padding shorter lists on the left with zeros, and multiplying the prime powers in corresponding positions.
Theorem 7.1.3
(Intrinsic filtration and numerical consequences)
If , then
Thus counts summands of exact order . The minimum number of generators of a nontrivial finite abelian group is , and its exponent is its largest invariant factor .
Concept lensCombinatorial
Rows and columns of one Young diagram
For a fixed prime, sort the exponents of as . They form a partition of , where , drawn as a Young diagram with row lengths . The numbers are its column lengths, so the filtration reads the same diagram by columns and computes the conjugate partition. For , rows give columns ; for , rows give . Either list determines the other, so abelian groups of order correspond exactly to partitions of .
Proof sketch or proof idea
From the matrix theorem to existence
Proof
(Existence of the invariant-factor and primary forms)
Apply the finitely generated PID module theorem over . It gives a free part and a finite list of cyclic torsion factors in a divisibility chain. Finiteness forces , because any nonzero free coordinate contains infinitely many integers. Normalizing each nonzero integer factor to be positive and omitting unit factors gives the displayed existence statement.
For a more intrinsic primary description, set . Every element is killed by , and its order divides , so is exactly the kernel of multiplication by . It is therefore a subgroup. CRT supplies integers that are one modulo and zero modulo every other prime-power factor of . For , put .
Since , the element belongs to . Also , so . If with each , multiply by . Its congruences make it act as identity on and zero on the other primary groups, yielding . Hence the sum is direct. This construction also shows that primary subgroups are characteristic: their definition depends only on element orders, which every automorphism preserves.
Apply the integer module theorem to . There is again no free part. Every cyclic factor order divides because that scalar kills the entire group, so each nontrivial factor is a positive power of . This proves existence of the primary cyclic form without assuming it from the subgroup's name.
Quotients recover the exponents uniquely
Proof
(Filtration dimensions force uniqueness)
Multiplication by kills , so integer scalar multiplication descends to and makes it a finite-dimensional vector space. On a cyclic summand of order , the quotient has dimension one for and zero for . Direct sums preserve these quotients coordinatewise, proving the formula for .
An isomorphism of groups preserves , its scalar multiples, and these quotient spaces. Therefore all the numbers are intrinsic. Their successive differences recover exactly how many exponents equal each positive integer, proving primary uniqueness. The dimensions eventually vanish because a fixed power of kills .
Now let be the largest number of primary cyclic summands for any prime. Sort each prime's exponent list increasingly and pad with initial zeros to length . If these exponents are , define . Then . At least one prime has a full positive list of length , so every . Integer CRT combines the coprime prime-power groups in each column into , yielding the invariant-factor form.
Conversely, decomposing any invariant-factor chain into its prime powers gives increasingly ordered exponent lists. Their positive parts have already been recovered intrinsically, and the requirement forces the same number of columns and the same padding. Thus the invariant factors too are unique. Arbitrary pairings of prime powers need not respect divisibility; the padding convention records the unique normalized arrangement.
Generators, exponent, and counts
Proof
(Minimum generators, exponent and kernel counts)
The invariant-factor form gives generators, one for each cyclic coordinate. Choose a prime with . The canonical projection is onto. Any generating family maps to a spanning family of this -dimensional vector space, so it has at least elements. This proves the minimum, including the reason that an apparently shorter clever generating list cannot work.
Every divides , so kills all coordinates. A generator of the last summand has order , forcing any common killing integer to be divisible by . Thus the exponent is exactly . The group order is , a different invariant. For a nontrivial group, cyclicity is equivalent to having only one invariant factor, or to at most one cyclic summand in every primary group.
There is another useful direct count. In a cyclic factor, exactly elements are killed by . Consequently
The successive differences of the base- logarithms of these sizes are the same numbers . Thus one can recover the decomposition from kernel counts as well as from quotient dimensions. This is an exact finite count, not an asymptotic estimate.
Proof X-ray
Existence and uniqueness use different information
Smith reduction provides some cyclic coordinates. The primary filtrations then recover their lengths without trusting that chosen presentation. This is why a second relation matrix cannot produce a genuinely different normalized answer: its quotient has the same intrinsic primary spaces. A successful diagonalization proves existence, while the invariant calculation prevents hidden dependence on the route taken.
Worked examples
Worked example 7.1.4
(Aligning three different cyclic factors)
Take . Its two-primary exponents are , from factors of orders two, four and eight. The three-primary exponents are , from orders three and nine. Pad the latter list to , then multiply columns: the invariant factors are . Their product is , and their divisibility chain is valid. The exponent is seventy-two, and three generators are necessary because has dimension three. Aligning the largest prime powers in different columns could give the same total order without giving the normalized chain.
Worked example 7.1.5
(All abelian groups of order seventy-two)
The two-primary order is eight, whose exponent partitions are , , and . The three-primary order is nine, with partitions and . Independent choices give exactly six isomorphism types:
| Invariant factors | Exponent | Minimum generators |
|---|---|---|
Each row has product seventy-two and a divisibility chain. Primary uniqueness proves the rows are distinct and exhaustive. Generally, for , the number of abelian isomorphism types is the product of the numbers of integer partitions of the , because those partitions are exactly the independent primary exponent lists.
Worked example 7.1.6
(From relations to an explicit group)
Let generate an abelian group with relations and . Its relation matrix, with columns as relations, is . The gcd of the entries is three and the determinant is seventy-two. Smith's minor characterization therefore gives factors three and twenty-four.
An explicit map sends and in . Both relations vanish; the images generate because maps to . Conversely every source element has a representative with and : reduce using , then reduce . Thus the source has at most seventy-two elements and maps onto a target of exactly that size, proving the map is an isomorphism. The presentation has exponent twenty-four, not seventy-two.
Counterexample mode
Order and exponent together still do not classify
The abelian groups and both have order sixty-four and exponent eight. The first has four elements killed by two, while the second has eight. Their intrinsic first two-primary dimensions are two and three, so they are not isomorphic. The repair is the full sequence of primary filtration dimensions, not one additional scalar invariant.
Common mistakes
Common mistake
(Normalize before comparing decompositions)
A displayed direct product need not already be an invariant-factor chain. One may split coprime factors by CRT and regroup them, but must normalize the result before comparing lists. Factors of order one carry no information and are omitted; the trivial group's empty list is a genuine boundary case.
Common mistake
(The classification is only for abelian groups)
The classification is abelian. The nonabelian group also has order six, but is not an overlooked row in a list classifying abelian groups. Nor is an arbitrary group presentation an integer-linear presentation: the relations and commutativity must justify treating its generators as a -module.
Summary
Finite generation and the integer PID theorem supply cyclic decompositions. Canonical primary projections and their filtration quotients recover every exponent independently of the presentation. Sorting and padding convert elementary divisors into invariant factors, which determine the exponent, minimum generator count, and complete finite abelian isomorphism type.
Exercises
Quick check
(A finite abelian two-group has order , with elements killed by , respectively. Recover its cyclic factors.)
Solution · Answer
The logarithms of the kernel sizes are , with initial value zero. Their successive differences are , followed by zero because sixteen kills the group. The counts of exact exponent sizes are therefore one at exponent one, one at exponent two, none at exponent three, and one at exponent four. The group is .
Quick check
(List all abelian groups of order one hundred in invariant-factor form, with a proof of exhaustiveness.)
Solution · Answer
Each of the primary orders four and twenty-five has two partitions, and . Combining them gives , , , and . Primary uniqueness proves these four choices distinct and complete.
Quick check
(If , count the subgroups of order . Deduce that a finite abelian group is cyclic exactly when it has a unique subgroup of each order dividing its order.)
Solution · Answer
Every subgroup of order is a one-dimensional subspace of . Its nonzero elements belong to no other such line, so the number is . If the group is cyclic, the cyclic subgroup theorem gives one subgroup for every divisor. Conversely, uniqueness at each prime divisor forces , hence each primary component is cyclic by the classification. Their coprime orders combine by CRT into a cyclic group. The trivial group also satisfies the assertion, with only its order-one subgroup.
Quick check
(Prove for positive , omitting any order-one factor.)
Solution · Answer
For each prime, write the valuations of as , including zero. The two primary factors have exponents ; sorting gives . These are precisely the valuations of the gcd and lcm. Aligning the prime lists and using the proved classification gives the isomorphism. This also covers coprime inputs, when the gcd factor is trivial and the product is cyclic of order .