aiwiki.page
English

Abelian Group

A group whose operation is commutative, central to algebra, number theory, and topology.

72 keywords33 linked from25 not yet writtenWritten by AI
IntegerAbstract AlgebraModule (mathemat…Cyclic GroupCommutatorRational NumberReal NumberComplex NumberAbelian Gr…

An abelian group is a group in which the order of combining two elements does not affect the result: a+b=b+aa+b=b+a for all elements aa and bb. Its operation also satisfies the group axioms of associativity, an identity element, and inverses. Addition of integers is a basic example. Abelian groups form a fundamental class of objects in abstract algebra. (jmilne.org)

The theory connects group operations with arithmetic and linear methods. Every abelian group is naturally a module over the integers, and every finitely generated abelian group can be decomposed into cyclic groups. Beyond finite generation, torsion, divisibility, and freeness distinguish substantially different kinds of infinite groups. These structures also supply the groups used in homology, linking algebraic classification to the study of topological spaces. (arxiv.org)

Definition, notation, and commutativity

An abelian group, or commutative group, is a nonempty set AA equipped with a binary operation ∗:A×A→A\ast:A\times A\to A satisfying the group axioms and an additional commutative law. The operation is closed: a∗ba\ast b belongs to AA whenever a,b∈Aa,b\in A. It is associative, so (a∗b)∗c=a∗(b∗c)(a\ast b)\ast c=a\ast(b\ast c). There is an identity element ee satisfying e∗a=a∗e=ae\ast a=a\ast e=a, and every aa has an inverse a−1a^{-1} satisfying a∗a−1=a−1∗a=ea\ast a^{-1}=a^{-1}\ast a=e. The additional requirement is

a∗b=b∗afor all a,b∈A.a\ast b=b\ast a\qquad\text{for all }a,b\in A.

Associativity concerns parentheses; commutativity concerns order. Neither replaces the other in the definition. (ocw.mit.edu)

Abelian groups are commonly written in additive notation: the operation is a+ba+b, the identity is 00, and the inverse of aa is −a-a. Thus a+0=aa+0=a and a+(−a)=0a+(-a)=0; subtraction means a−b=a+(−b)a-b=a+(-b). These symbols describe the abstract operation, not necessarily ordinary numerical addition. For a positive integer nn, the expression nana denotes the sum of nn copies of aa; one defines 0a=00a=0 and (−n)a=−(na)(-n)a=-(na). In multiplicative notation, the corresponding expressions are abab, 11, a−1a^{-1}, and ana^n, with a0=1a^0=1 and a−n=(a−1)na^{-n}=(a^{-1})^n. Multiplicative notation does not imply that a group is noncommutative. (math.berkeley.edu)

The adjective “abelian” commemorates the Norwegian mathematician Niels Henrik Abel (1802–1829). It specifies commutativity, not the stronger requirement of being a cyclic group, meaning a group generated by a single element. Every cyclic group is abelian because gmgn=gm+n=gn+m=gngmg^mg^n=g^{m+n}=g^{n+m}=g^ng^m, but the converse fails. For example, Z×Z\mathbb Z\times\mathbb Z, with componentwise addition, is abelian but not cyclic: the multiples of one pair (r,s)(r,s) cannot include both (1,0)(1,0) and (0,1)(0,1). (www2.math.uconn.edu)

Commutativity can also be expressed through the center of a group GG, defined in multiplicative notation by

Z(G)={z∈G:zg=gz for every g∈G}.Z(G)=\{z\in G:zg=gz\text{ for every }g\in G\}.

The center is a subgroup consisting of the elements that commute with every element of GG. Consequently, GG is abelian exactly when Z(G)=GZ(G)=G. Merely having a nonidentity element in the center is a weaker condition. (jmilne.org)

A second characterization uses commutators. With the convention [a,b]=aba−1b−1[a,b]=aba^{-1}b^{-1}, the equality [a,b]=1[a,b]=1 holds exactly when ab=baab=ba. Thus GG is abelian if and only if every commutator is the identity, equivalently if its commutator subgroup—the subgroup generated by all commutators—is {1}\{1\}. Some authors instead use a−1b−1aba^{-1}b^{-1}ab; the criterion that all commutators are trivial is unchanged. (jmilne.org)

Elementary examples and nonexamples

The integers (Z,+)(\mathbb Z,+), rational numbers (Q,+)(\mathbb Q,+), real numbers (R,+)(\mathbb R,+), and complex numbers (C,+)(\mathbb C,+) are abelian groups under ordinary addition. In each case, sums remain in the specified set, the identity is 00, and the inverse of aa is −a-a; associativity and commutativity are inherited from numerical addition. The operation is essential to the specification: Z\mathbb Z under multiplication is not a group, since 22 has no integer multiplicative inverse. Likewise, none of these four complete number systems forms a group under multiplication, because 00 has no multiplicative inverse. (math.mit.edu)

For every positive integer nn, the residues Z/nZ\mathbb Z/n\mathbb Z form an abelian group under addition in modular arithmetic. Writing a‾\overline a for the residue class of aa, the operation is a‾+b‾=a+b‾\overline a+\overline b=\overline{a+b}, with identity 0‾\overline0 and inverse −a‾=−a‾-\overline a=\overline{-a}. This is a cyclic group: repeatedly adding 1‾\overline1 produces all nn elements. For example, in Z/5Z\mathbb Z/5\mathbb Z, 3‾+4‾=2‾\overline3+\overline4=\overline2. The case n=1n=1 is the one-element group. More generally, every cyclic group is abelian, because any two powers of a generator satisfy grgs=gr+s=gs+r=gsgrg^rg^s=g^{r+s}=g^{s+r}=g^sg^r. (jmilne.org)

The Klein four-group supplies an abelian example that is not cyclic. It can be represented as

V4=Z/2Z×Z/2Z={(0,0),(1,0),(0,1),(1,1)},V_4=\mathbb Z/2\mathbb Z\times\mathbb Z/2\mathbb Z =\{(0,0),(1,0),(0,1),(1,1)\},

with componentwise addition modulo 22. Thus (1,0)+(0,1)=(1,1)(1,0)+(0,1)=(1,1), in either order. Every nonidentity element added to itself gives (0,0)(0,0), so each generates only a two-element group; no element generates all four elements. This distinguishes V4V_4 from the cyclic group Z/4Z\mathbb Z/4\mathbb Z, despite their equal sizes. A multiplicative realization consists of the four diagonal real matrices diag⁡(ε1,ε2)\operatorname{diag}(\varepsilon_1,\varepsilon_2), where each εi\varepsilon_i is 11 or −1-1. (ocw.mit.edu)

For any field FF, the nonzero elements F×=F∖{0}F^\times=F\setminus\{0\} form an abelian group under multiplication. Its identity is 11, and the inverse of aa is a−1a^{-1}; excluding zero ensures that every element is invertible. Examples include Q×\mathbb Q^\times, R×\mathbb R^\times, and C×\mathbb C^\times. In the finite field F5\mathbb F_5, the multiplicative group has elements 1,2,3,41,2,3,4: the successive powers of 22 modulo 55 are 2,4,3,12,4,3,1, so it is cyclic of order 44. These multiplicative groups must not be confused with the additive groups of the same fields. (math.mit.edu)

By contrast, the symmetric group S3S_3, consisting of the six permutations of three objects under composition, is not abelian. With the rightmost permutation applied first,

(12)(23)=(123),(23)(12)=(132).(12)(23)=(123),\qquad (23)(12)=(132).

The two products differ, although composition remains associative and every permutation has an inverse. Thus associativity does not imply commutativity. (jmilne.org)

Invertible matrices under multiplication provide another nonexample. The general linear group GL2(R)GL_2(\mathbb R) contains

A=(1101),B=(1011),A=\begin{pmatrix}1&1\\0&1\end{pmatrix}, \qquad B=\begin{pmatrix}1&0\\1&1\end{pmatrix},

both with determinant 11, but direct multiplication gives

AB=(2111)≠(1112)=BA.AB=\begin{pmatrix}2&1\\1&1\end{pmatrix} \ne \begin{pmatrix}1&1\\1&2\end{pmatrix}=BA.

Matrix-based groups are not necessarily nonabelian—the diagonal realization of V4V_4 above is abelian—but invertibility alone does not make multiplication commutative. (jmilne.org)

Subgroups, quotients, and homomorphisms

A subgroup HH of an abelian group AA is itself abelian, since its operation is inherited from AA. It is also a normal subgroup: in additive notation, conjugation has the form a+h−a=ha+h-a=h, so every element of AA leaves HH invariant under conjugation. Consequently, every subgroup defines a quotient group A/HA/H. Its elements are cosets a+H={a+h:h∈H}a+H=\{a+h:h\in H\}, with addition

(a+H)+(b+H)=(a+b)+H.(a+H)+(b+H)=(a+b)+H.

This operation is independent of the representatives: replacing aa and bb by a+h1a+h_1 and b+h2b+h_2 changes their sum by h1+h2∈Hh_1+h_2\in H. The identity is H=0+HH=0+H, the inverse of a+Ha+H is −a+H-a+H, and the quotient remains abelian. For example, Z/5Z\mathbb Z/5\mathbb Z identifies integers whose difference is divisible by 55, giving five cosets. (math.mit.edu)

A homomorphism f:A→Bf:A\to B between additively written abelian groups preserves the operation:

f(a+a′)=f(a)+f(a′).f(a+a')=f(a)+f(a').

It follows that f(0)=0f(0)=0, f(−a)=−f(a)f(-a)=-f(a), and f(na)=nf(a)f(na)=nf(a) for every integer nn. Its kernel is ker⁡f={a∈A:f(a)=0}\ker f=\{a\in A:f(a)=0\}, and its image is im⁡f={f(a):a∈A}\operatorname{im}f=\{f(a):a\in A\}; these are subgroups of AA and BB, respectively. The kernel records precisely which differences are lost: f(a)=f(a′)f(a)=f(a') if and only if a−a′∈ker⁡fa-a'\in\ker f. Thus ff is injective exactly when ker⁡f={0}\ker f=\{0\}, and surjective exactly when im⁡f=B\operatorname{im}f=B. An isomorphism is a bijective homomorphism; its inverse also preserves addition. (jmilne.org)

The first isomorphism theorem expresses the relationship between these constructions:

A/ker⁡f  ≅  im⁡f,a+ker⁡f⟼f(a).A/\ker f\;\cong\;\operatorname{im}f, \qquad a+\ker f\longmapsto f(a).

The displayed map is well-defined because elements of the same coset have the same image, injective because equal images imply equal cosets, and surjective onto im⁡f\operatorname{im}f by definition. For a concrete example, let f:Z→Z/12Zf:\mathbb Z\to\mathbb Z/12\mathbb Z send kk to 8k+12Z8k+12\mathbb Z. Its kernel is 3Z3\mathbb Z, while its image consists of the three residue classes 0,4,80,4,8. Hence this image is isomorphic to Z/3Z\mathbb Z/3\mathbb Z, although the stated codomain has twelve elements. (math.columbia.edu)

For an arbitrary group GG, now written multiplicatively, the commutator subgroup [G,G][G,G] is generated by all elements xyx−1y−1xyx^{-1}y^{-1}. It is normal, and the quotient

Gab=G/[G,G]G^{\mathrm{ab}}=G/[G,G]

is called the abelianization of GG. In this quotient every commutator becomes the identity, forcing all cosets to commute. More precisely, [G,G][G,G] is the smallest normal subgroup NN for which G/NG/N is abelian. If GG is already abelian, then [G,G]={1}[G,G]=\{1\}, so its abelianization is naturally isomorphic to GG. (jmilne.org)

Abelianization has a universal property: every homomorphism f:G→Bf:G\to B into an abelian group factors uniquely through the quotient map q:G→Gabq:G\to G^{\mathrm{ab}}. Indeed, f(xyx−1y−1)=1f(xyx^{-1}y^{-1})=1, so [G,G]⊆ker⁡f[G,G]\subseteq\ker f, and the required map is

f‾:Gab→B,f‾(g[G,G])=f(g),f=f‾∘q.\overline f:G^{\mathrm{ab}}\to B,\qquad \overline f(g[G,G])=f(g),\qquad f=\overline f\circ q.

Its uniqueness follows from the surjectivity of qq. Thus “largest abelian quotient” describes a factorization property, not a comparison of cardinalities: every homomorphism from GG to an abelian group is completely determined by a homomorphism from GabG^{\mathrm{ab}}. (gt.postech.ac.kr)

Direct sums, products, and limiting constructions

For a family of abelian groups (Ai)i∈I(A_i)_{i\in I}, the direct product ∏i∈IAi\prod_{i\in I}A_i consists of all tuples a=(ai)i∈Ia=(a_i)_{i\in I}, with ai∈Aia_i\in A_i. Its addition is componentwise: (ai)+(bi)=(ai+bi)(a_i)+(b_i)=(a_i+b_i); the zero tuple and componentwise negatives supply the identity and inverses. The direct sum is the subgroup

⨁i∈IAi={(ai)∈∏i∈IAi:{i∈I:ai≠0} is finite}.\bigoplus_{i\in I}A_i = \left\{(a_i)\in\prod_{i\in I}A_i: \{i\in I:a_i\ne0\}\text{ is finite}\right\}.

The finite set of nonzero coordinates is the tuple’s support. Adding two finite-support tuples again gives finite support, since the support of their sum is contained in the union of their supports. Neither construction requires a topology or a notion of convergence. (lib.ysu.am)

When II is finite, every tuple has finite support, so the direct sum and product coincide canonically. Their distinction for infinite families is also expressed by their universal properties: homomorphisms B→∏iAiB\to\prod_iA_i correspond to families of homomorphisms B→AiB\to A_i, whereas homomorphisms ⨁iAi→B\bigoplus_iA_i\to B correspond to families Ai→BA_i\to B. In the latter case, the induced map sends (ai)(a_i) to ∑ifi(ai)\sum_i f_i(a_i), a well-defined finite sum. Thus the direct sum is the coproduct in the category of abelian groups; no general infinite summation operation is needed. (lib.ysu.am)

For example, ⨁n≥1Z\bigoplus_{n\ge1}\mathbb Z contains precisely the integer sequences that are eventually zero, while ∏n≥1Z\prod_{n\ge1}\mathbb Z contains every integer sequence, including (1,1,1,…)(1,1,1,\ldots). These groups are not merely different subgroups of the same ambient product: they cannot be isomorphic. Indeed, the sum is countable, being the union of the countable sets of sequences supported in {1,…,m}\{1,\ldots,m\}. The product is uncountable because it contains all binary sequences {0,1}N\{0,1\}^{\mathbb N}; its cardinality is 2ℵ02^{\aleph_0}. This cardinality argument directly establishes the failure of the finite sum–product identification in this example. (lib.ysu.am)

A direct limit assembles groups connected by forward homomorphisms. For a sequence A1→f1A2→f2⋯A_1\xrightarrow{f_1}A_2\xrightarrow{f_2}\cdots, it can be constructed as

lim→⁡nAn=(⨁n≥1An)/⟨ιn(a)−ιn+1(fn(a)):n≥1, a∈An⟩,\varinjlim_n A_n = \left(\bigoplus_{n\ge1}A_n\right)\Big/ \left\langle \iota_n(a)-\iota_{n+1}(f_n(a)): n\ge1,\ a\in A_n \right\rangle,

where ιn\iota_n denotes the inclusion into the direct sum. These relations identify each element with its image at the next stage. Every limit element has a representative at one stage, and two representatives agree exactly when their images agree at some common later stage. For inclusions A1⊆A2⊆⋯A_1\subseteq A_2\subseteq\cdots, the direct limit is the increasing union ⋃nAn\bigcup_nA_n. In particular, adjoining a zero coordinate gives lim→⁡nZn≅⨁n≥1Z\varinjlim_n\mathbb Z^n\cong\bigoplus_{n\ge1}\mathbb Z. (arxiv.org)

An inverse limit instead uses backward maps gn:An+1→Ang_n:A_{n+1}\to A_n and retains compatible sequences:

lim←⁡nAn={(an)∈∏n≥1An:gn(an+1)=an for every n}.\varprojlim_n A_n = \left\{(a_n)\in\prod_{n\ge1}A_n: g_n(a_{n+1})=a_n\text{ for every }n\right\}.

Addition remains componentwise. For a prime pp, the reduction maps Z/pn+1Z→Z/pnZ\mathbb Z/p^{n+1}\mathbb Z\to\mathbb Z/p^n\mathbb Z yield the additive group of the pp-adic integers,

Zp=lim←⁡n≥1Z/pnZ.\mathbb Z_p=\varprojlim_{n\ge1}\mathbb Z/p^n\mathbb Z.

An element therefore records a residue at every precision, each reducing to the preceding residue. For p=2p=2, the compatible sequence (1 mod 2, 3 mod 4, 7 mod 8, 15 mod 16,…)(1\bmod2,\,3\bmod4,\,7\bmod8,\,15\bmod16,\ldots) represents −1-1. Ordinary integers embed by taking their residues at every level; injectivity follows because an integer divisible by every pnp^n must be zero. (math.mit.edu)

Integer modules and free abelian groups

For an abelian group AA, integer scalar multiplication is determined by its addition: for n>0n>0, define nana as the sum of nn copies of aa, set 0a=00a=0, and put (−n)a=−(na)(-n)a=-(na). Commutativity ensures that

n(a+b)=na+nb,(m+n)a=ma+na,m(na)=(mn)a,1a=a.n(a+b)=na+nb,\qquad (m+n)a=ma+na,\qquad m(na)=(mn)a,\qquad 1a=a.

These are precisely the axioms of a module over the ring Z\mathbb Z. Conversely, the additive group of any Z\mathbb Z-module is abelian, and its scalar multiplication must be this repeated-addition operation. Every group homomorphism f:A→Bf:A\to B satisfies f(na)=nf(a)f(na)=nf(a), so it is automatically Z\mathbb Z-linear. Thus abelian groups and Z\mathbb Z-modules describe the same objects and maps, not merely analogous structures; their subgroups are exactly their submodules. (faculty.niu.edu)

A free abelian group is an abelian group FF with a subset BB, called an integer basis, such that every element has a unique expression

x=∑b∈Bnbb,nb∈Z,x=\sum_{b\in B}n_b b,\qquad n_b\in\mathbb Z,

with only finitely many nonzero coefficients. Equivalently, BB generates FF and is linearly independent over Z\mathbb Z: any finite integer relation among distinct basis elements has all coefficients zero. Given an arbitrary set BB, the free abelian group on BB can be constructed as the group of finitely supported functions B→ZB\to\mathbb Z, with pointwise addition. Its basis elements are the functions taking value 11 at one specified point and 00 elsewhere, and it is denoted Z(B)=⨁b∈BZ\mathbb Z^{(B)}=\bigoplus_{b\in B}\mathbb Z. The finite-support condition remains essential when BB is infinite. (faculty.niu.edu)

This construction has a universal mapping property: every function u:B→Au:B\to A, where AA is any abelian group, extends uniquely to a homomorphism u~:Z(B)→A\widetilde u:\mathbb Z^{(B)}\to A. Explicitly,

u~ ⁣(∑b∈Bnbb)=∑b∈Bnbu(b).\widetilde u\!\left(\sum_{b\in B}n_b b\right)=\sum_{b\in B}n_bu(b).

Finiteness makes the sum meaningful, and uniqueness of coordinates makes it well-defined. No independence condition is imposed on the chosen images u(b)u(b). For example, specifying the images of the two standard basis elements of Z2\mathbb Z^2 determines a homomorphism to any abelian group. More generally, taking BB to be the underlying set of AA gives a surjective homomorphism from a free abelian group onto AA. (faculty.niu.edu)

An integer basis differs from a vector-space basis because integers do not form a field: division by a nonzero scalar is not generally available. For example, (2,0)(2,0) and (0,1)(0,1) are integer-linearly independent in Z2\mathbb Z^2, but they do not generate it, since their first coordinates are always even. The same pair is a basis of the vector space Q2\mathbb Q^2, where multiplication by 1/21/2 is permitted. Thus an independent set need not extend to an integer basis. Nor does every abelian group possess one: in Z/nZ\mathbb Z/n\mathbb Z, for n>1n>1, every element is annihilated by nn, whereas a nonzero element of a free abelian group cannot satisfy such a relation, by coordinate uniqueness. (faculty.niu.edu)

Every subgroup of a free abelian group is itself free abelian, including groups with infinite bases under the usual axiom of choice. In finite rank, an induction illustrates the mechanism. For H≤ZrH\leq\mathbb Z^r, project onto the last coordinate. Its image is either zero or dZd\mathbb Z for some d>0d>0. The kernel is a subgroup of Zr−1\mathbb Z^{r-1}, hence free by induction. If the image is nonzero, choose h∈Hh\in H with last coordinate dd; subtracting suitable multiples of hh gives H=ker⁡(π∣H)⊕ZhH=\ker(\pi|_H)\oplus\mathbb Zh. The subgroup’s basis need not extend to one of the ambient group: 2Z2\mathbb Z has basis {2}\{2\}, but neither 22 nor −2-2 generates Z\mathbb Z. (arxiv.org)

Classification of finitely generated abelian groups

An abelian group is finitely generated if some finite list a1,…,ama_1,\ldots,a_m expresses every element as an integer combination n1a1+⋯+nmamn_1a_1+\cdots+n_ma_m. The fundamental classification theorem states that every such group has an isomorphism

A≅Zr⊕Z/d1Z⊕⋯⊕Z/dkZ,1<d1∣d2∣⋯∣dk.A\cong \mathbb Z^r\oplus \mathbb Z/d_1\mathbb Z\oplus\cdots\oplus \mathbb Z/d_k\mathbb Z, \qquad 1<d_1\mid d_2\mid\cdots\mid d_k.

Here r,k≥0r,k\geq0, and ⊕\oplus denotes the direct sum. The integer rr, called the free rank, and the ordered sequence d1,…,dkd_1,\ldots,d_k, called the invariant factors, are uniquely determined by AA. Consequently, two finitely generated abelian groups are isomorphic exactly when these data agree. Empty lists are allowed: k=0k=0 gives Zr\mathbb Z^r, while r=k=0r=k=0 gives the trivial group. Uniqueness concerns the isomorphism types of the factors, not a preferred choice of generators or internal summands. (jmilne.org)

The finite cyclic summands together constitute the subgroup of all finite-order elements, whereas the quotient by that subgroup is isomorphic to Zr\mathbb Z^r. Thus AA is finite precisely when r=0r=0; in that case its order is d1⋯dkd_1\cdots d_k, with the empty product equal to 11. A finitely generated group with no nonzero finite-order elements is therefore free. For example, Z2⊕Z/6Z\mathbb Z^2\oplus\mathbb Z/6\mathbb Z has free rank 22 and six finite-order elements, but is itself infinite. The finite-generation hypothesis is essential to these conclusions. (crypto.stanford.edu)

An equivalent elementary-divisor decomposition replaces the finite cyclic summands by cyclic groups of prime-power order:

A≅Zr⊕⨁p ⨁j=1tpZ/pep,jZ,ep,j≥1.A\cong\mathbb Z^r\oplus \bigoplus_p\ \bigoplus_{j=1}^{t_p} \mathbb Z/p^{e_{p,j}}\mathbb Z, \qquad e_{p,j}\geq1.

Only finitely many summands occur, and the multiset of prime powers is unique. Conversion from invariant factors uses the Chinese remainder theorem: if d=∏ppvp(d)d=\prod_p p^{v_p(d)}, reduction modulo the prime-power factors gives

Z/dZ≅⨁p∣dZ/pvp(d)Z.\mathbb Z/d\mathbb Z\cong \bigoplus_{p\mid d}\mathbb Z/p^{v_p(d)}\mathbb Z.

To reverse the conversion, arrange each prime’s exponents in nondecreasing order, pad shorter lists on the left with zeros, and multiply the prime powers in each column. The resulting integers form a divisibility chain. For instance, elementary divisors 2,4,3,3,52,4,3,3,5 give invariant factors 6,606,60, so Z/12Z⊕Z/30Z≅Z/6Z⊕Z/60Z\mathbb Z/12\mathbb Z\oplus\mathbb Z/30\mathbb Z \cong\mathbb Z/6\mathbb Z\oplus\mathbb Z/60\mathbb Z. (jmilne.org)

For order 8=238=2^3, the theorem yields exactly three abelian groups up to isomorphism:

Z/8Z,Z/4Z⊕Z/2Z,(Z/2Z)3.\mathbb Z/8\mathbb Z,\qquad \mathbb Z/4\mathbb Z\oplus\mathbb Z/2\mathbb Z,\qquad (\mathbb Z/2\mathbb Z)^3.

These correspond to the exponent lists 33, 2,12,1, and 1,1,11,1,1. Their largest element orders are respectively 88, 44, and 22, which also directly distinguishes their isomorphism types. This list classifies only the abelian groups of that order, not all groups of order 88. (crypto.stanford.edu)

More generally, for a fixed prime pp, each integer partition n=λ1+⋯+λtn=\lambda_1+\cdots+\lambda_t, with λ1≥⋯≥λt≥1\lambda_1\geq\cdots\geq\lambda_t\geq1, determines

⨁j=1tZ/pλjZ.\bigoplus_{j=1}^{t}\mathbb Z/p^{\lambda_j}\mathbb Z.

Its order is pnp^n; existence and uniqueness of elementary divisors show that every abelian group of this order occurs exactly once. Hence the number of isomorphism classes is the partition number P(n)P(n), independent of pp. For an arbitrary positive order N=∏ppnpN=\prod_p p^{n_p}, the choices for distinct primes are independent, giving ∏p∣NP(np)\prod_{p\mid N}P(n_p) classes. For example, 360=23⋅32⋅5360=2^3\cdot3^2\cdot5 gives P(3)P(2)P(1)=3⋅2⋅1=6P(3)P(2)P(1)=3\cdot2\cdot1=6 abelian groups up to isomorphism. (crypto.stanford.edu)

Presentations and Smith normal form

An abelian presentation specifies generators a1,…,ana_1,\ldots,a_n and integer relations ∑i=1nMijai=0\sum_{i=1}^{n}M_{ij}a_i=0, for 1≤j≤m1\leq j\leq m. Commutativity is understood; these are relations in a free abelian group, not an unrestricted group presentation. Using the relations as columns produces an n×mn\times m integer matrix representing a homomorphism M:Zm→ZnM:\mathbb Z^m\to\mathbb Z^n. The presented group is its cokernel:

A=coker⁡M=Zn/im⁡M.A=\operatorname{coker}M=\mathbb Z^n/\operatorname{im}M.

Thus two integer combinations of generators represent the same element exactly when their coefficient vectors differ by an integer combination of the relation columns. Every finitely generated abelian group admits such a finite presentation. (ocw.mit.edu)

The permitted changes of presentation are integral changes of basis. A square integer matrix is unimodular if its determinant is 11 or −1-1, equivalently if its inverse also has integer entries. Replacing MM by UMVUMV, with unimodular matrices UU and VV, preserves its cokernel up to isomorphism: UU changes coordinates in the generator lattice, while VV changes the basis used to express the relations. Elementary operations are swapping rows or columns, multiplying a row or column by −1-1, and adding an integer multiple of one row or column to another. The Euclidean algorithm supplies the repeated division steps underlying this reduction. (math.mit.edu)

The resulting Smith normal form is a rectangular diagonal matrix

UMV=D=diag⁡(d1,…,dr,0,…,0),di>0,d1∣d2∣⋯∣dr,UMV=D=\operatorname{diag}(d_1,\ldots,d_r,0,\ldots,0), \qquad d_i>0,\quad d_1\mid d_2\mid\cdots\mid d_r,

where rr is the rank of MM over Q\mathbb Q. The positive entries are uniquely determined. Consequently,

A≅Z n−r⊕⨁i=1rZ/diZ.A\cong\mathbb Z^{\,n-r}\oplus \bigoplus_{i=1}^{r}\mathbb Z/d_i\mathbb Z.

An entry di=1d_i=1 contributes a trivial group; an entry greater than 11 contributes a finite cyclic factor; each remaining coordinate in the target contributes a free generator. This makes the classification theorem computational rather than merely existential. Henry John Stephen Smith’s treatment of integer systems appeared in his 1861 paper “On systems of linear indeterminate equations and congruences.” (people.reed.edu)

For example, M=diag⁡(2,3)M=\operatorname{diag}(2,3) presents generators a,ba,b with 2a=02a=0 and 3b=03b=0. Although already diagonal, it is not in Smith normal form because 2∤32\nmid3. Applying the allowed operations gives the following explicit calculation:

(2003)⟶(2303)⟶(2103)⟶(1230)⟶(120−6)⟶(1006).\begin{pmatrix}2&0\\0&3\end{pmatrix} \longrightarrow \begin{pmatrix}2&3\\0&3\end{pmatrix} \longrightarrow \begin{pmatrix}2&1\\0&3\end{pmatrix} \longrightarrow \begin{pmatrix}1&2\\3&0\end{pmatrix} \longrightarrow \begin{pmatrix}1&2\\0&-6\end{pmatrix} \longrightarrow \begin{pmatrix}1&0\\0&6\end{pmatrix}.

The steps are R1←R1+R2R_1\leftarrow R_1+R_2, C2←C2−C1C_2\leftarrow C_2-C_1, swapping columns, R2←R2−3R1R_2\leftarrow R_2-3R_1, and finally C2←C2−2C1C_2\leftarrow C_2-2C_1 followed by a row-sign change. Hence the cokernel is Z/6Z\mathbb Z/6\mathbb Z, consistent with the Chinese remainder theorem. (math.mit.edu)

Determinantal data provide an independent check. If Δk\Delta_k is the positive greatest common divisor of all k×kk\times k minors and Δ0=1\Delta_0=1, then dk=Δk/Δk−1d_k=\Delta_k/\Delta_{k-1} for 1≤k≤r1\leq k\leq r. In the example, Δ1=gcd⁡(2,3)=1\Delta_1=\gcd(2,3)=1 and Δ2=6\Delta_2=6, giving d1=1,d2=6d_1=1,d_2=6. For a nonsingular square matrix, the cokernel is finite of order ∣det⁡M∣|\det M|. Unlike Gaussian elimination over a field, integral reduction cannot divide a row by an arbitrary nonzero pivot: dividing by 22, for example, is not an automorphism of the integer lattice. Over Q\mathbb Q, the example reduces to the identity and has zero cokernel, losing precisely the finite-group information that Smith reduction retains. (math.mit.edu)

Torsion groups and primary decomposition

An element aa of an abelian group AA is a torsion element if na=0na=0 for some positive integer nn; its order is the least such nn. The torsion subgroup is

T(A)={a∈A:na=0 for some n>0}.T(A)=\{a\in A:na=0\text{ for some }n>0\}.

Commutativity ensures closure: if ma=0ma=0 and nb=0nb=0, then mn(a−b)=0mn(a-b)=0. The quotient group A/T(A)A/T(A) is torsion-free: if n(a+T(A))=0n(a+T(A))=0, then nana is torsion, so mna=0mna=0 for some m>0m>0, forcing a∈T(A)a\in T(A). A group is a torsion group when T(A)=AT(A)=A; this does not imply finiteness. For example, Q/Z\mathbb Q/\mathbb Z is infinite, but every coset r/s+Zr/s+\mathbb Z is annihilated by ss. (arxiv.org)

For each prime pp, define the pp-primary component

Ap={a∈A:pka=0 for some integer k≥0}.A_p=\{a\in A:p^ka=0\text{ for some integer }k\geq0\}.

Primary decomposition gives the canonical internal direct sum

T(A)=⨁p primeAp.T(A)=\bigoplus_{p\text{ prime}}A_p.

Thus a torsion group equals the sum of its primary components, and every element has a unique decomposition with only finitely many nonzero components. The mechanism is the Chinese remainder theorem: factoring an annihilator nn into coprime prime powers supplies integer multiples of the element that isolate each primary part. For example, in Z/12Z\mathbb Z/12\mathbb Z, the 22-primary component is generated by 33, the 33-primary component by 44, and 1=9+41=9+4. The components are intrinsic subgroups, preserved by every homomorphism. (arxiv.org)

A group has bounded exponent if a single positive integer NN annihilates all its elements; the least such NN is its exponent. This uniform bound is stronger than torsion, where the annihilator may depend on the element. Direct calculations illustrate the distinction: ⨁j≥1Z/pZ\bigoplus_{j\geq1}\mathbb Z/p\mathbb Z is infinite with exponent pp, whereas ⨁j≥1Z/pjZ\bigoplus_{j\geq1}\mathbb Z/p^j\mathbb Z is torsion with unbounded element orders. In the latter group, every element has finite support and hence finite order, although the orders of the summand generators grow without bound. These examples follow directly from the finite-support definition of a direct sum and the torsion criterion. (arxiv.org)

Within a pp-primary group GG, element height measures divisibility inside GG, rather than element order. Define a decreasing filtration, indexed by ordinals, by

p0G=G,pα+1G=p(pαG),pλG=⋂α<λpαGp^0G=G,\qquad p^{\alpha+1}G=p(p^\alpha G),\qquad p^\lambda G=\bigcap_{\alpha<\lambda}p^\alpha G

for a limit ordinal λ\lambda. A nonzero element xx has height α\alpha when x∈pαG∖pα+1Gx\in p^\alpha G\setminus p^{\alpha+1}G; an element surviving every stage is assigned height ∞\infty. For finite nn, membership in pnGp^nG means that pny=xp^ny=x is solvable. In particular, pωG=⋂n<ωpnGp^\omega G=\bigcap_{n<\omega}p^nG can contain nonzero elements even in a reduced group, so finite heights alone need not capture its structure. (math.berkeley.edu)

Writing H[p]={x∈H:px=0}H[p]=\{x\in H:px=0\}, the Ulm invariant at stage α\alpha is

uα(G)=dim⁡Fp((pαG)[p]/(pα+1G)[p]).u_\alpha(G)= \dim_{\mathbb F_p}\bigl((p^\alpha G)[p]/(p^{\alpha+1}G)[p]\bigr).

The quotient is a vector space over the field Fp\mathbb F_p. These dimensions generalize cyclic-summand multiplicities: for G=⨁j≥1(Z/pjZ)(κj)G=\bigoplus_{j\geq1}(\mathbb Z/p^j\mathbb Z)^{(\kappa_j)}, one has un(G)=κn+1u_n(G)=\kappa_{n+1}. Helmut Ulm’s 1933 classification theorem states that two countable reduced abelian pp-groups are isomorphic if and only if all their Ulm invariants agree. Here “reduced” means having no nonzero divisible subgroup, where divisibility requires solvability of ny=xny=x for every n>0n>0 and every xx. Countability is essential to this statement: it does not assert classification of arbitrary uncountable pp-groups by these invariants. Without reducedness, the divisible part must also be specified, since its Ulm invariants all vanish. (arxiv.org)

Divisible groups and reduced complements

An abelian group DD is divisible if, for every a∈Da\in D and every positive integer nn, the equation nx=anx=a has a solution in DD. Equivalently, every multiplication map n:D→Dn:D\to D is surjective; solutions need not be unique. The additive group of rational numbers Q\mathbb Q is divisible because a/n∈Qa/n\in\mathbb Q. Its quotient group Q/Z\mathbb Q/\mathbb Z is also divisible: r/n+Zr/n+\mathbb Z maps to r+Zr+\mathbb Z under multiplication by nn. Unlike Q\mathbb Q, this quotient consists entirely of finite-order elements, showing that divisibility does not imply torsion-freeness. More generally, quotients of divisible groups are divisible. (math.ucla.edu)

For each prime pp, the Prüfer pp-group, denoted Cp∞C_{p^\infty} or Z(p∞)\mathbb Z(p^\infty), has the concrete realization

Cp∞=Z[1/p]/Z=⋃k≥1⟨1pk+Z⟩.C_{p^\infty}=\mathbb Z[1/p]/\mathbb Z =\bigcup_{k\geq 1}\left\langle \frac{1}{p^k}+\mathbb Z\right\rangle .

The displayed cyclic subgroups have orders pkp^k and are nested. Multiplication by pp is surjective because m/pk+1+Zm/p^{k+1}+\mathbb Z maps to m/pk+Zm/p^k+\mathbb Z. Multiplication by an integer coprime to pp is an automorphism on every finite cyclic stage; together these facts establish divisibility for every positive integer. Thus Cp∞C_{p^\infty} is an infinite divisible torsion group, despite having only finite cyclic groups at its individual stages. (math.ucla.edu)

The classification theorem states that every divisible abelian group has an isomorphism

D≅Q(κ0)⊕⨁p primeCp∞(κp),D\cong \mathbb Q^{(\kappa_0)} \oplus \bigoplus_{p\ \mathrm{prime}}C_{p^\infty}^{(\kappa_p)},

where the multiplicities κ0\kappa_0 and κp\kappa_p are uniquely determined cardinals, possibly zero or infinite. Parenthesized exponents mean direct sums, not unrestricted products. If T(D)T(D) denotes the torsion subgroup, then

κ0=dim⁡Q(D/T(D)),κp=dim⁡FpD[p],D[p]={x∈D:px=0}.\kappa_0=\dim_{\mathbb Q}(D/T(D)), \qquad \kappa_p=\dim_{\mathbb F_p}D[p], \quad D[p]=\{x\in D:px=0\}.

Here D/T(D)D/T(D) is naturally a rational vector space, while each Prüfer summand contributes one dimension to D[p]D[p]. In particular, Q/Z≅⨁pCp∞\mathbb Q/\mathbb Z\cong\bigoplus_p C_{p^\infty}, with exactly one summand for each prime. (math.ucla.edu)

Every abelian group AA has a unique largest divisible subgroup D(A)D(A): it is the sum of all divisible subgroups of AA. This sum is divisible because each element is a finite sum, whose terms can be divided separately. The splitting mechanism is the equivalence between divisibility and injectivity for abelian groups: every homomorphism from a subgroup B⊆CB\subseteq C into a divisible group extends to CC. Applying this property to the identity map of D(A)D(A) gives a retraction r:A→D(A)r:A\to D(A), with r∣D(A)=idr|_{D(A)}=\mathrm{id}, and hence

A=D(A)⊕ker⁡r.A=D(A)\oplus\ker r.

Thus the divisible subgroup is always a direct summand, not merely a subgroup or quotient. (math.ucla.edu)

A group is reduced if its only divisible subgroup is zero. The complement R=ker⁡rR=\ker r above is reduced, since any divisible subgroup of RR would also lie in D(A)D(A). Although D(A)D(A) is canonical, the embedded complement need not be: only its isomorphism type, R≅A/D(A)R\cong A/D(A), is determined. For an explicit illustration, in A=Q⊕ZA=\mathbb Q\oplus\mathbb Z, the largest divisible subgroup is Q⊕0\mathbb Q\oplus0, but every subgroup

Rc={(nc,n):n∈Z},c∈Q,R_c=\{(nc,n):n\in\mathbb Z\},\qquad c\in\mathbb Q,

is a reduced complement. Indeed, (q,n)=(q−nc,0)+(nc,n)(q,n)=(q-nc,0)+(nc,n) uniquely, and distinct values of cc give distinct embedded copies of Z\mathbb Z. This directly exhibits the nonuniqueness permitted by the splitting theorem. (math.colorado.edu)

Infinite torsion-free groups and classification limits

For a torsion-free abelian group AA, the equation na=0na=0, with nn a nonzero integer, implies a=0a=0. Its torsion-free rank is

rank⁡(A)=dim⁡Q(Q⊗ZA),\operatorname{rank}(A)= \dim_{\mathbb Q}\bigl(\mathbb Q\otimes_{\mathbb Z}A\bigr),

where the tensor product is a vector space over the rational numbers. Equivalently, rank is the cardinality of a maximal subset that is linearly independent over Z\mathbb Z: every finite integer relation among its elements has all coefficients zero. Rank measures independence, not the minimum number of generators. A maximal independent subset need not generate AA; it becomes a vector-space basis only after rationalization. For torsion-free AA, the natural map a↦1⊗aa\mapsto1\otimes a is injective, so a group of finite rank rr can be represented as a subgroup of Qr\mathbb Q^r. (math.stonybrook.edu)

Every finitely generated torsion-free abelian group is free, hence isomorphic to Zr\mathbb Z^r, but finite rank alone does not imply finite generation. The additive groups Q\mathbb Q and

Z[1/p]={mpk:m∈Z, k≥0},\mathbb Z[1/p]= \left\{\frac{m}{p^k}:m\in\mathbb Z,\ k\geq0\right\},

for a prime pp, both have rank one and are not free. Indeed, any finite collection of rational numbers has a common denominator dd, so its integer span lies in (1/d)Z(1/d)\mathbb Z. This cannot exhaust Q\mathbb Q; for Z[1/p]\mathbb Z[1/p], finitely many generators have bounded powers of pp in their denominators and cannot generate the next power. Both groups require countably infinitely many generators. Since a free group of rank one is infinite cyclic, neither is free. Nevertheless, every finitely generated subgroup of either is cyclic. Thus even being an increasing union of finite-rank free groups does not guarantee freeness. (arxiv.org)

Nonzero torsion-free groups of rank one are precisely the groups isomorphic to subgroups of the additive group of rational numbers. Their classification records divisibility prime by prime. For a≠0a\neq0, define its pp-height by

hp(a)=sup⁡{k≥0:a∈pkA},h_p(a)=\sup\{k\geq0:a\in p^kA\},

allowing the value ∞\infty. Two height sequences have the same type when they have identical infinite coordinates and differ at only finitely many finite coordinates. All nonzero elements of a rank-one group have the same type, and two such groups are isomorphic exactly when their types agree. For the element 11, all heights in Z\mathbb Z are zero; in Z[1/p]\mathbb Z[1/p], the pp-height is infinite and all other prime heights are zero; in Q\mathbb Q, every prime height is infinite. These invariants distinguish groups whose rationalizations are all one-dimensional. (karlin.mff.cuni.cz)

Whitehead’s problem, attributed to J. H. C. Whitehead, asks whether an abelian group AA must be free if every short exact sequence

0⟶Z⟶E→πA⟶00\longrightarrow\mathbb Z\longrightarrow E \xrightarrow{\pi}A\longrightarrow0

splits. Here splitting means that a homomorphism s:A→Es:A\to E satisfies πs=id⁡A\pi s=\operatorname{id}_A, yielding E≅Z⊕AE\cong\mathbb Z\oplus A compatibly with the sequence. The condition is conventionally written Ext⁡Z1(A,Z)=0\operatorname{Ext}^1_{\mathbb Z}(A,\mathbb Z)=0; groups satisfying it are called Whitehead groups. Free groups satisfy it because lifts of basis elements define a section. Conversely, every countable Whitehead group is free, and every Whitehead group is torsion-free. The difficulty therefore lies in uncountable groups. (shelah.logic.at)

In his 1974 paper Infinite abelian groups, Whitehead problem and some constructions, Saharon Shelah proved independence from ZFC already for groups of cardinality ℵ1\aleph_1, the least uncountable cardinal. Under the constructibility axiom V=LV=L, every Whitehead group of that cardinality is free; under Martin’s axiom together with the negation of the continuum hypothesis, a nonfree Whitehead group of cardinality ℵ1\aleph_1 exists. Assuming ZFC is consistent, both settings are consistent with it. Consequently, ZFC cannot prove either the unrestricted assertion that every Whitehead group is free or its negation. This is a proved limitation of the specified axioms, not an ordinary unresolved conjecture awaiting a proof within them. (shelah.logic.at)

Exact sequences, extensions, and categorical structure

A sequence of homomorphisms is exact at a group when the image of the incoming map equals the kernel of the outgoing map. Thus a short exact sequence

0⟶B→iE→qA⟶00\longrightarrow B\xrightarrow{i}E\xrightarrow{q}A\longrightarrow0

means that ii is injective, qq is surjective, and im⁡i=ker⁡q\operatorname{im}i=\ker q. It identifies BB with a subgroup of EE and AA with the quotient E/i(B)E/i(B); the group EE, together with these maps, is an extension of AA by BB. The sequence splits if there is a homomorphism s:A→Es:A\to E satisfying q∘s=id⁡Aq\circ s=\operatorname{id}_A. In that case (b,a)↦i(b)+s(a)(b,a)\mapsto i(b)+s(a) gives an isomorphism B⊕A≅EB\oplus A\cong E compatible with the sequence. (stacks.math.columbia.edu)

Exactness alone does not imply splitting. For every integer n>1n>1, reduction modulo nn gives

0⟶Z→×nZ⟶Z/nZ⟶0.0\longrightarrow\mathbb Z\xrightarrow{\times n}\mathbb Z \longrightarrow\mathbb Z/n\mathbb Z\longrightarrow0.

This sequence is exact because the kernel of reduction is nZn\mathbb Z. If a section existed, the image x∈Zx\in\mathbb Z of the residue class 11 would satisfy both nx=0nx=0 and x≡1(modn)x\equiv1\pmod n, which is impossible. By contrast, the inclusion and projection in 0→Z→Z⊕Z/nZ→Z/nZ→00\to\mathbb Z\to\mathbb Z\oplus\mathbb Z/n\mathbb Z\to\mathbb Z/n\mathbb Z\to0 give a split extension with the same end groups. (math.uchicago.edu)

In categorical terms, abelian groups and their homomorphisms form the category Ab\mathbf{Ab}. Each set Hom⁡(A,B)\operatorname{Hom}(A,B) is itself an abelian group under pointwise addition, (f+g)(a)=f(a)+g(a)(f+g)(a)=f(a)+g(a), and composition distributes over this addition. The trivial group is a zero object, and finite direct sums serve simultaneously as products and coproducts. A map f:A→Bf:A\to B has kernel f−1(0)f^{-1}(0) and cokernel B/f(A)B/f(A), while its coimage A/ker⁡fA/\ker f is naturally isomorphic to its image f(A)f(A). These properties make Ab\mathbf{Ab} an abelian category. That term describes a category with additive and exactness structure, not a single group with a commutative operation; categories of modules over rings are further examples. (math.uchicago.edu)

The tensor product A⊗ZBA\otimes_{\mathbb Z}B is generated by symbols a⊗ba\otimes b, subject to additivity in each variable. Its universal property says that every map β:A×B→C\beta:A\times B\to C additive separately in both arguments factors uniquely through a homomorphism A⊗ZB→CA\otimes_{\mathbb Z}B\to C. For example, (Z/nZ)⊗ZB≅B/nB(\mathbb Z/n\mathbb Z)\otimes_{\mathbb Z}B\cong B/nB. Tensoring is right exact: it preserves the final part B→E→A→0B\to E\to A\to0 of an exact sequence, but need not preserve the initial injection. Tensoring the multiplication-by-nn sequence above with Z/nZ\mathbb Z/n\mathbb Z turns its injective map into the zero map between two nonzero cyclic groups. (stacks.math.columbia.edu)

The functor Tor⁡1Z(A,B)\operatorname{Tor}^{\mathbb Z}_1(A,B) measures this failure of tensoring to preserve injections. It is computed by tensoring a free resolution of AA with BB and taking first homology. The resolution 0→Z→×nZ→Z/nZ→00\to\mathbb Z\xrightarrow{\times n}\mathbb Z\to\mathbb Z/n\mathbb Z\to0 yields

Tor⁡1Z(Z/nZ,B)≅B[n]={b∈B:nb=0}.\operatorname{Tor}^{\mathbb Z}_1(\mathbb Z/n\mathbb Z,B) \cong B[n]=\{b\in B:nb=0\}.

Thus it detects elements annihilated by nn, rather than the quotient B/nBB/nB produced by tensoring itself. Over Z\mathbb Z, this Tor group vanishes whenever either argument is torsion-free. (math.uchicago.edu)

The Ext functor has a complementary extension-theoretic interpretation: Ext⁡Z1(A,B)\operatorname{Ext}^1_{\mathbb Z}(A,B) classifies short exact sequences 0→B→E→A→00\to B\to E\to A\to0, up to isomorphisms of middle groups inducing the identity on both end groups. Its addition, the Baer sum, combines two extensions using the diagonal map on AA and the addition map on BB; its zero class consists precisely of split extensions. Applying Hom⁡(−,B)\operatorname{Hom}(-,B) to the same cyclic resolution gives

Ext⁡Z1(Z/nZ,B)≅B/nB.\operatorname{Ext}^1_{\mathbb Z}(\mathbb Z/n\mathbb Z,B)\cong B/nB.

In particular, Ext⁡Z1(Z/nZ,Z)≅Z/nZ\operatorname{Ext}^1_{\mathbb Z}(\mathbb Z/n\mathbb Z,\mathbb Z)\cong\mathbb Z/n\mathbb Z, whereas Ext⁡Z1(F,B)=0\operatorname{Ext}^1_{\mathbb Z}(F,B)=0 for every free abelian group FF: every extension with free quotient splits. (stacks.math.columbia.edu)

Roles in topology, number theory, and harmonic analysis

In algebraic topology, integral homology is constructed from free abelian groups Cn(X;Z)C_n(X;\mathbb Z) generated by continuous maps of nn-simplices into a space XX. Boundary homomorphisms satisfy ∂n∂n+1=0\partial_n\partial_{n+1}=0, allowing the quotient

Hn(X;Z)=ker⁡∂n/im⁡∂n+1.H_n(X;\mathbb Z)=\ker\partial_n/\operatorname{im}\partial_{n+1}.

Its elements are cycles considered equivalent when their difference is a boundary. Cohomology instead uses cochains Hom⁡(Cn(X;Z),A)\operatorname{Hom}(C_n(X;\mathbb Z),A), with an abelian coefficient group AA, and takes cocycles modulo coboundaries. Continuous maps induce homomorphisms in the same direction on homology and the reverse direction on cohomology; both constructions are invariant under homotopy. (math.mit.edu)

For the torus T2=S1×S1T^2=S^1\times S^1,

H1(T2;Z)≅Z2.H_1(T^2;\mathbb Z)\cong\mathbb Z^2.

The two generators represent loops around its two circle factors; a class (m,n)(m,n) records their integer multiplicities. When integral homology is finitely generated, its free rank counts independent infinite-order classes—the corresponding Betti number. A torsion class of order kk is represented by a cycle that is not a boundary although its kk-fold multiple is. For example, H1(RP2;Z)≅Z/2ZH_1(\mathbb{RP}^2;\mathbb Z)\cong\mathbb Z/2\mathbb Z: twice a generating cycle bounds. Torsion therefore records information absent from rational homology. (math.mit.edu)

In algebraic number theory, let KK be a number field and OK\mathcal O_K its ring of integers. Nonzero fractional ideals form an abelian group IKI_K under multiplication, and principal fractional ideals form a subgroup PKP_K. The ideal class group

Cl⁡(K)=IK/PK\operatorname{Cl}(K)=I_K/P_K

is finite; its order is the class number. Nonzero integral ideals factor uniquely into prime ideals even when elements do not factor uniquely into irreducibles. The class group measures the obstruction to ideals being principal: it is trivial precisely when OK\mathcal O_K is a principal ideal domain, equivalently a unique factorization domain. (jmilne.org)

The unit group OK×\mathcal O_K^\times, by contrast, consists of invertible elements and describes multiplicative arithmetic within the ring. Dirichlet’s unit theorem gives

OK×≅μ(K)⊕Zr1+r2−1,\mathcal O_K^\times\cong\mu(K)\oplus\mathbb Z^{r_1+r_2-1},

where μ(K)\mu(K) is the finite cyclic group of roots of unity in KK, r1r_1 counts real embeddings, and r2r_2 counts conjugate pairs of complex embeddings. Thus real quadratic fields have unit rank 11, whereas imaginary quadratic fields have rank 00. Units also describe the ambiguity in generators: (a)=(b)(a)=(b) for nonzero elements exactly when a=uba=ub for a unit uu. (jmilne.org)

In harmonic analysis, a locally compact abelian group carries a Hausdorff topology making addition and inversion continuous, with a compact neighborhood at every point. This is additional structure, not part of the abstract group axioms. A continuous character is a homomorphism χ:G→T\chi:G\to\mathbb T, where T={z∈C:∣z∣=1}\mathbb T=\{z\in\mathbb C:|z|=1\}. Characters form the dual group G^\widehat G under pointwise multiplication, equipped with the compact-open topology. Pontryagin duality states that

G⟶G^^,g⟼(χ⟼χ(g))G\longrightarrow\widehat{\widehat G},\qquad g\longmapsto\bigl(\chi\longmapsto\chi(g)\bigr)

is an isomorphism of topological groups. In particular, it recovers both the group operation and the topology. (math.uh.edu)

The dual of discrete Z\mathbb Z is the compact circle T\mathbb T, since a character is determined by χ(1)\chi(1); conversely, the circle’s characters z↦znz\mapsto z^n give its discrete dual Z\mathbb Z. The additive real line is self-dual, with characters x↦e2πiξxx\mapsto e^{2\pi i\xi x}. More generally, duality exchanges compact and discrete abelian groups. For an integrable function and a chosen Haar measure, the Fourier transform takes the form

f^(χ)=∫Gf(x)χ(x)‾ dx.\widehat f(\chi)=\int_G f(x)\overline{\chi(x)}\,dx.

Accordingly, Fourier series on the circle have integer-indexed frequencies, while the Fourier transform on the real line has real-indexed frequencies. (people.math.harvard.edu)

References

  1. 703 Modern Algebra, Groupsocw.mit.edu
  2. Additive notation for commutative semigroups and abelian groupsmath.berkeley.edu
  3. Group Theoryjmilne.org
  4. Mathematics 315: People documentswww2.math.uconn.edu
  5. Niels Abel (1802–1829) – Biographymathshistory.st-andrews.ac.uk
  6. Commutator subgroupen.wikipedia.org
  7. Algebra fact sheetmath.mit.edu
  8. RES.18-011 (Fall 2021) Full Lecture Notes: Algebra I Student Notesocw.mit.edu
  9. Section 19.10: Grothendieck's AB conditions—The Stacks projectstacks.math.columbia.edu
  10. Lemma 13.33.6—The Stacks projectstacks.math.columbia.edu
  11. Abelian Groupsarxiv.org
  12. 782 Introduction to Arithmetic Geometry, Lecture #4math.mit.edu
  13. Modules | Department of Mathematical Sciences | Northern Illinois Universityfaculty.niu.edu
  14. Group Theory — Finitely Generated Abelian Groupscrypto.stanford.edu
  15. Finitely Generated Abelian Groupssites.millersville.edu
  16. Smith Normal Form and Combinatoricsmath.mit.edu
  17. RES.18-012 (Spring 2022) Full Lecture Notes: Algebra II Student Notesocw.mit.edu
  18. Instructions for presentationsmath.ucsd.edu
  19. Math 361 lecture for Friday, Week 8people.reed.edu
  20. Paper, 'On systems of linear indeterminate equations and congruences' by Henry J [John] Stephen Smithcatalogues.royalsociety.org
  21. The Smith normal formbruno.stonek.com
  22. Translating the class of abelian p-groups into an elementary first-order theorymath.berkeley.edu
  23. The Isomorphism Problem for Computable Abelian p-Groups of Bounded Lengtharxiv.org
  24. Classification theorems for p-groups and modules over a discrete valuation ringscispace.com
  25. Abstract Algebra — Chapter 12 Homological algebra · Romyar Sharifimath.ucla.edu
  26. Your title heremath.ucla.edu
  27. 32 also define D_max to be the problemweb.math.wisc.edu
  28. 12 MARCOS MAZARI-ARMIDAmath.colorado.edu
  29. Lecture Notes on Morse and Conley Theorymath.stonybrook.edu
  30. Abelian groups with a minimal generating setkarlin.mff.cuni.cz
  31. Infinite abelian groups, Whitehead problem and some constructionsshelah.logic.at
  32. Infinite abelian groups, whitehead problem and some constructions — The Hebrew University of Jerusalemcris.huji.ac.il
  33. Section 12.5: Abelian categories—The Stacks projectstacks.math.columbia.edu
  34. Section 12.6: Extensions—The Stacks projectstacks.math.columbia.edu
  35. Section 13.27: Ext groups—The Stacks projectstacks.math.columbia.edu
  36. Section 10.12: Tensor products—The Stacks projectstacks.math.columbia.edu
  37. WOMP 2006: Homological Algebra, Part 2math.uchicago.edu
  38. Lectures on Algebraic Topologymath.mit.edu
  39. Algebraic Number Theoryjmilne.org
  40. Pontryagin Dualitymath.uh.edu
  41. Outline of Pontrjagin dualitypeople.math.harvard.edu