Alphabeta Math
Session-authored (Fable 5 assisted)
How statement and proof provenance work

The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.

  • Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
  • AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
  • AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.

These labels describe origin, not correctness: citations and verification chips remain separate evidence.

40 results · all verified · 16 also independently AI-judged
Every result on this page is machine-checked by a proof checker and read in full and owner-audited; the judge is an additional, independent cross-model AI review of the proofs. The 24 not AI-judged were verified by owner audit (typically over a confirmed judge false positive), not failures.

Extraspecial p-Groups and Central Products

1 · Prerequisites

2 · Summary

Published notions of centre, commutator subgroup, quotient group, and semidirect product supply the ambient group theory, while the prerequisite Frattini page supplies Elementary abelian p-groups, Φ(P)=PPp for a finite p-group, The Frattini quotient is the largest elementary abelian quotient of a finite p-group, and Burnside Basis Theorem. These are the inputs that turn the three standard descriptions of an extraspecial group into equivalent ones and that let the central quotient carry a genuine Fp-linear structure.

The page defines special and extraspecial p-groups, external and internal central products, the commutator pairing, the square map for the two-group case, and the plus/minus types. It then proves the class-two commutator formulas, the central-product universal property and recognition theorem, the decomposition of every extraspecial group into order-p3 factors, the order and maximal-abelian consequences of that decomposition, the classification of the nonabelian groups of order p3, the odd- and two-primary classification theorems, the exponent dichotomy, and the two automorphism results that act on the Frattini quotient and on the commutator pairing.

3 · Logical flowchart

4 · Definitions, theorems and proofs

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-26Open item page →

Commutator identities in a group whose derived subgroup is central

Statement

Let G be a group (Group and abelian group) whose derived subgroup [G,G] is contained in the centre Z(G) (Commutators [g,h]=ghg1h1 and the commutator subgroup [G,G], The center Z(G) of a group). If [G,G]Z(G) then [xy,w]=[x,w][y,w], [x,yw]=[x,y][x,w], and [xn,y]=[x,y]n=[x,yn] for every integer n and all x,y,wG, integer powers being those of Powers gn: natural exponents in a monoid and integer exponents in a group, with g0=e.

Facts & Assumptions

Given: A group G with [G,G]Z(G), and elements x,y,wG.

[F1]

For g,hG the commutator is [g,h]:=ghg1h1, and [G,G]:={[g,h]:g,hG} is the subgroup generated by all commutators (Commutators [g,h]=ghg1h1 and the commutator subgroup [G,G]).

[F2]

Z(G):={zG:zg=gz for every gG} (The center Z(G) of a group).

Proof

technique · direct
1.1

Every commutator [g,h] lies in [G,G], hence in Z(G), so it commutes with every element of G and may be moved to any position in a product without changing that product.

F1F2given
2.1

Expanding, [xy,w]=(xy)w(xy)1w1=xywy1x1w1=x(ywy1w1)wx1w1=x[y,w]wx1w1=[y,w]xwx1w1=[y,w][x,w]=[x,w][y,w], the last two equalities moving the central factor [y,w] past x and then past [x,w].

F1L2step 1.1algebra
2.2

Expanding in the second variable, [x,yw]=x(yw)x1(yw)1=xywx1w1y1=(xyx1)(xwx1w1)y1=(xyx1)[x,w]y1=[x,w]xyx1y1=[x,y][x,w], where the central factor [x,w] is moved past y1 and then past [x,y].

F1L2step 1.1algebra
3.1

For nN the identity [xn,y]=[x,y]n follows by induction: at n=0 both sides are the identity, since x0 is the identity and [e,y]=eye1y1=e, and [xn+1,y]=[xnx,y]=[xn,y][x,y]=[x,y]n[x,y]=[x,y]n+1.

F1L1step 2.1algebra
4.1

For a negative integer n put k=n; then e=[x0,y]=[xkxk,y]=[xk,y][xk,y] by step 2.1, so [xn,y]=[xk,y]=([x,y]k)1=[x,y]k=[x,y]n, and the identity [x,yn]=[x,y]n is obtained the same way from step 2.2.

L1step 2.1step 2.2step 3.1algebra

Remarks

The hypothesis [G,G]Z(G) is exactly the vanishing of the third term of the lower central series: γ3(G)=[G,[G,G]] is trivial precisely when every commutator commutes with every element of G (Subgroup commutators and the lower central series).

The first two identities are additivity in each variable separately, and they fail without the hypothesis: in general [xy,w]=x[y,w]x1[x,w], and the conjugating factor is what the hypothesis removes.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-08-26Open item page →

In a group with central derived subgroup, (xy)n=[y,x](n2)xnyn

Statement

Let G be a group with [G,G]Z(G) (Commutators [g,h]=ghg1h1 and the commutator subgroup [G,G], The center Z(G) of a group) and let x,yG. If [G,G]Z(G) then (xy)n=[y,x](n2)xnyn for every nN, where (n2) is the binomial coefficient of The set [A]k of k-element subsets and the binomial coefficient (nk):=[n]k and the powers are those of Powers gn: natural exponents in a monoid and integer exponents in a group, with g0=e.

The commutator on the right is [y,x], not [x,y]: in the convention [g,h]=ghg1h1 fixed by Commutators [g,h]=ghg1h1 and the commutator subgroup [G,G] one has yx=[y,x]xy, and it is that factor which accumulates.

Facts & Assumptions

Given: A group G with [G,G]Z(G), elements x,yG, and nN.

[F1]

For g,hG the commutator is [g,h]:=ghg1h1 (Commutators [g,h]=ghg1h1 and the commutator subgroup [G,G]).

[F2]

Z(G):={zG:zg=gz for every gG} (The center Z(G) of a group).

[F3]

(nk):=[n]k, the number of k-element subsets of n; in particular (n1)=n, and (nk)=0 whenever k>n (The set [A]k of k-element subsets and the binomial coefficient (nk):=[n]k).

[L1]

If [G,G]Z(G) then [xy,w]=[x,w][y,w], [x,yw]=[x,y][x,w], and [xn,y]=[x,y]n=[x,yn] for every integer n (Commutator identities in a group whose derived subgroup is central).

Proof

technique · induction
1.1

At n=0 the exponent (02) is zero because 2>0, and the asserted identity reads e=e; this is the base case.

F3base
1.2

For every nN, Pascal's rule at k=1 gives (n+12)=(n1)+(n2)=n+(n2).

F3L2algebra
1.3

For all g,hG one has [g,h]hg=(ghg1h1)hg=gh, hence gh=[g,h]hg; applied to the pair y,x this reads yx=[y,x]xy.

F1algebra
1.4

Assume the identity at a given nN, that is (xy)n=[y,x](n2)xnyn.

ih
2.1

Applying step 1.3 to the pair yn,x gives ynx=[yn,x]xyn, and [yn,x]=[y,x]n, so ynx=[y,x]nxyn.

L1step 1.3algebra
2.2

Multiplying the assumption on the right by xy gives (xy)n+1=(xy)n(xy)=[y,x](n2)xnynxy.

step 1.4algebra
3.1

Substituting step 2.1 into step 2.2 and moving the central factor [y,x]n to the front gives (xy)n+1=[y,x](n2)xn[y,x]nxyny=[y,x](n2)+nxn+1yn+1.

F2L3step 2.1step 2.2algebra
4.1

By step 1.2 the exponent (n2)+n equals (n+12), so the identity holds at n+1 and therefore at every natural number.

step 1.2step 3.1discharge-induction

Remarks

The formula is the reason the p-th power map behaves differently at p=2: the coefficient (22) equals 1, so the commutator factor survives, whereas for odd p the coefficient (p2) is a multiple of p.

CorollaryStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-08-26Open item page →

For an odd prime p, the p-th power map is a homomorphism on a finite group whose derived subgroup is central of exponent dividing p

Statement

Let p be an odd prime (Prime and composite integers: p is prime when p>1 and its only positive divisors are 1 and p) and let G be a finite group with [G,G]Z(G) whose derived subgroup has exponent dividing p (The exponent of a finite group, Commutators [g,h]=ghg1h1 and the commutator subgroup [G,G], The center Z(G) of a group). Then

(xy)p=xpypfor all x,yG,

so xxp is a group homomorphism from G to G.

Facts & Assumptions

Given: An odd prime p and a finite group G with [G,G]Z(G) and exp([G,G]) dividing p; elements x,yG.

[F1]

For a finite group H, exp(H)=min{nN:n>0 and gn=e for every gH} (The exponent of a finite group).

[F2]

(nk):=[n]k, the number of k-element subsets of n (The set [A]k of k-element subsets and the binomial coefficient (nk):=[n]k).

[L1]

If [G,G]Z(G) then (xy)n=[y,x](n2)xnyn for every nN (In a group with central derived subgroup, (xy)n=[y,x](n2)xnyn).

Proof

technique · direct
1.1

Taking n=p in the product formula gives (xy)p=[y,x](p2)xpyp.

L1
1.2

Since p is odd and p3, the closed formula at n=p and k=2 gives (p2)2=p(p1), and writing p1=2m with mN turns this into (p2)2=2pm, so (p2)=pm.

F2F3L2algebra
1.3

The element c=[y,x] lies in [G,G], and exp([G,G]) divides p, say p=exp([G,G])t; hence cp=(cexp([G,G]))t=e.

F1L3algebra
2.1

Combining, [y,x](p2)=(cp)m=e, so step 1.1 reads (xy)p=xpyp; as this holds for all x,yG, the map xxp is a homomorphism.

step 1.1step 1.2step 1.3algebra

Remarks

Only the oddness of p is used, in step 1.2; primality enters through the hypothesis on the derived subgroup rather than through the arithmetic. At p=2 the conclusion fails at the first step: (22)=1, so (xy)2=[y,x]x2y2 and the commutator factor survives.

DefinitionDefinition: Literature-sourcedProof: Not applicablejudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-26Open item page →

Special and extraspecial p-groups

Definition

Let p be a prime and let P be a finite p-group (A finite p-group has order pn for a prime p and some nN). Write P=[P,P] for its derived subgroup (Commutators [g,h]=ghg1h1 and the commutator subgroup [G,G]), Z(P) for its centre (The center Z(G) of a group) and Φ(P) for its Frattini subgroup (The Frattini subgroup Φ(G) as the intersection of the maximal subgroups of a finite group).

A finite p-group P is special when Z(P)=P=Φ(P) is elementary abelian, and extraspecial when in addition P is nonabelian and this common subgroup has order p. Elementary abelian p-groups are those of Elementary abelian p-groups, and the trivial group is one of them.

Remarks

The nonabelian clause is stated rather than left implicit. It is not redundant for every formulation in the literature: the informal description "P/Z(P) is elementary abelian and Z(P)=p" is satisfied by the cyclic group of order p, whose centre is the whole group, so a definition phrased that way must exclude the abelian case by hand. With the clause Z(P)=P in force the exclusion is automatic, since an abelian group has trivial derived subgroup.

Two source conventions for extraspecial groups are in circulation and agree under the order-p centre hypothesis. Craven asks that Z(P)=P=Φ(P) be elementary abelian and then that this common subgroup have order p; van Beek asks that P=Z(P) have order p and that P/Z(P) have exponent p. Their equivalence for nonabelian P is the content of Three equivalent descriptions of an extraspecial p-group.

PropositionStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-08-26Open item page →

Three equivalent descriptions of an extraspecial p-group

Statement

For a finite p-group P the following are equivalent: P is extraspecial; P is nonabelian, Z(P)=p and P/Z(P) is elementary abelian; P is nonabelian and Z(P)=P=Φ(P) has order p.

Here P=[P,P], Φ(P) is the Frattini subgroup, and quotients are those of The quotient group G/N and coset product (gN)(hN)=ghN.

Facts & Assumptions

Given: A prime p and a finite p-group P (A finite p-group has order pn for a prime p and some nN).

[F1]

A finite p-group P is special when Z(P)=P=Φ(P) is elementary abelian, and extraspecial when in addition P is nonabelian and this common subgroup has order p (Special and extraspecial p-groups).

[L1]

For a finite p-group P, the quotient P/Φ(P) is elementary abelian, and for NP the quotient P/N is elementary abelian if and only if Φ(P)N (The Frattini quotient is the largest elementary abelian quotient of a finite p-group).

[L2]

For every finite p-group P, Φ(P)=PPp, where Pp=gp:gP (Φ(P)=PPp for a finite p-group, The pth-power subgroup Gp).

[L3]

For NG, the quotient G/N is abelian if and only if [G,G]N (G/N is abelian if and only if [G,G]N).

[L4]

An elementary abelian p-group is a finite abelian p-group in which every nonidentity element has order p; the trivial group is permitted (Elementary abelian p-groups).

[L5]

For every group G, the center Z(G) is a normal subgroup of G (The center of a group is a normal subgroup).

[L6]

For a finite group G and HG, G=[G:H]H (Lagrange's theorem: G=[G:H]H for every subgroup H of a finite group G).

[L7]

In a finite group whose order is prime, every ge has order G and generates G (A finite group of prime order is cyclic and every nonidentity element generates it).

Proof

technique · direct
1.1

Suppose P is extraspecial. Then P is nonabelian, Φ(P)=Z(P) has order p, and P/Φ(P) is elementary abelian; since Φ(P)=Z(P) this says P/Z(P) is elementary abelian, the quotient being formed along a normal subgroup. So the second description holds.

F1L1L4L5
1.2

Suppose P is nonabelian with Z(P)=p and P/Z(P) elementary abelian. Applying the elementary abelian criterion to the normal subgroup Z(P) gives Φ(P)Z(P), and an elementary abelian quotient is abelian, so PZ(P). As P is nonabelian, P1; a subgroup of the group Z(P) of order p has order 1 or p, so P=Z(P). Then Φ(P)=PPp contains P=Z(P) and is contained in Z(P), so Φ(P)=P=Z(P) has order p, which is the third description.

L1L2L3L5L6
1.3

Suppose P is nonabelian with Z(P)=P=Φ(P) of order p. A group of prime order is cyclic, hence abelian, and each of its nonidentity elements has order p, so this common subgroup is elementary abelian; it is a finite p-group because its order is p. Thus P meets the definition and is extraspecial.

F1L4L7
2.1

The three implications close a cycle, so the three descriptions are equivalent.

step 1.1step 1.2step 1.3

Remarks

The nonabelian hypothesis does real work exactly once, in step 1.2, where it supplies P1. Dropping it leaves the cyclic group of order p satisfying the second description with Z(P)=P of order p and trivial quotient, while its derived subgroup is trivial and the third description fails.

CorollaryStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-26Open item page →

An extraspecial p-group is nilpotent of class exactly two and its derived subgroup has order p

Statement

Let P be an extraspecial p-group. Then P is nilpotent of nilpotency class exactly two, its derived subgroup satisfies P=Z(P) and has order p, and every nonidentity commutator of P has order p.

Facts & Assumptions

Given: An extraspecial p-group P (Special and extraspecial p-groups).

[L1]

For a finite p-group P the following are equivalent: P is extraspecial; P is nonabelian, Z(P)=p and P/Z(P) is elementary abelian; P is nonabelian and Z(P)=P=Φ(P) has order p (Three equivalent descriptions of an extraspecial p-group).

[L2]

For subgroups A,BG the subgroup commutator is [A,B]=[a,b]:aA, bB, and the lower central series is γ1(G)=G, γr+1(G)=[G,γr(G)] (Subgroup commutators and the lower central series).

[L3]

G is nilpotent exactly when its lower central series reaches 1, and the least c with γc+1(G)=1 is its nilpotency class (Nilpotence via central series, the upper central series, and the lower central series, Nilpotent groups and nilpotency class).

[L4]

In a finite group whose order is prime, every ge has order G (A finite group of prime order is cyclic and every nonidentity element generates it).

[F1]

Z(G):={zG:zg=gz for every gG} (The center Z(G) of a group).

Proof

technique · direct
1.1

By the third description in the characterisation, P=Z(P) has order p and P is nonabelian.

L1
1.2

The second term of the lower central series is γ2(P)=[P,γ1(P)]=[P,P]=P, and the third is γ3(P)=[P,P].

L2
2.1

Every element of P=Z(P) commutes with every element of P, so each generator [g,z] of [P,P] is the identity and γ3(P)=1.

F1step 1.1step 1.2
3.1

Hence P is nilpotent of class at most two; the class is not zero or one, since class at most one would give γ2(P)=P=1 and P is nonabelian, so the class is exactly two.

L3step 1.1step 1.2step 2.1
4.1

Every commutator lies in P, a group of order p, so a nonidentity commutator has order p.

L4step 1.1

Remarks

Both conclusions are hypotheses of the class-two commutator calculus: the derived subgroup is central, which is what class two says, and it has exponent p, which is what the order-p conclusion says.

CorollaryStatement: AI-adaptedProof: AI-adaptedprecheck passjudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-26Open item page →

The centre of an extraspecial p-group has no complement

Statement

Let P be an extraspecial p-group. Then Z(P) has no complement in P: there is no subgroup HP with P=Z(P)H and Z(P)H=1.

Facts & Assumptions

Given: An extraspecial p-group P (Special and extraspecial p-groups).

[F1]

For subgroups N,H of G with NG, G=NH and NH={1}, the group G is the internal semidirect product of N by H, and H is called a complement to N in G (An internal semidirect product and a complement to a normal subgroup).

[F2]

Z(G):={zG:zg=gz for every gG} (The center Z(G) of a group).

[L1]

For a finite p-group P the following are equivalent: P is extraspecial; P is nonabelian, Z(P)=p and P/Z(P) is elementary abelian; P is nonabelian and Z(P)=P=Φ(P) has order p (Three equivalent descriptions of an extraspecial p-group).

[L2]

For every homomorphism f:GH, the rule gkerff(g) is an isomorphism from G/kerf onto imf (First isomorphism theorem for groups: G/kerfimf).

[L3]

An elementary abelian p-group is a finite abelian p-group in which every nonidentity element has order p (Elementary abelian p-groups).

Proof

technique · contradiction
1.1

Suppose HP is a complement to Z(P), so that P=Z(P)H and Z(P)H=1; the centre is normal, so the quotient P/Z(P) is defined.

F1L1assume-contra
2.1

Let π:PP/Z(P) be the quotient map and restrict it to H. Its kernel is HZ(P)=1, and its image is all of P/Z(P) because every gP is zh with zZ(P) and hH, whence π(g)=π(h); so HP/Z(P).

L2step 1.1
3.1

The quotient P/Z(P) is elementary abelian, hence abelian, so H is abelian.

L1L3step 2.1
4.1

Every element of P is zh with z central and hH, and (z1h1)(z2h2)=z1z2h1h2=z2z1h2h1=(z2h2)(z1h1), so P is abelian; this contradicts the nonabelianness of an extraspecial group.

F2step 1.1step 3.1discharge-contradiction

Remarks

The argument uses no bound on the order of P, so the conclusion holds for every extraspecial group and not only for those of order p3. What fails is not that a complement is hard to find but that its existence would make the group abelian, which the definition forbids.

PropositionStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-08-26Open item page →

Every conjugacy class of an extraspecial p-group outside the centre has exactly p elements

Statement

Every conjugacy class of an extraspecial p-group whose representative is not central has exactly p elements. That is, for an extraspecial p-group P and xPZ(P),

ClP(x)=p.

Facts & Assumptions

Given: An extraspecial p-group P (Special and extraspecial p-groups) and an element xP with xZ(P).

[F1]

ClG(x):={gxg1:gG} and CG(x):={gG:gx=xg}={gG:gxg1=x} (The conjugacy class ClG(x) and centralizer CG(x) of an element).

[F2]

Z(G):={zG:zg=gz for every gG} (The center Z(G) of a group).

[F3]

For g,hG the commutator is [g,h]:=ghg1h1 (Commutators [g,h]=ghg1h1 and the commutator subgroup [G,G]).

[F4]

For HG and gG, the right coset is Hg:={hg:hH} (Left and right cosets gH and Hg of a subgroup).

[L1]

An extraspecial p-group has derived subgroup P=Z(P) of order p (An extraspecial p-group is nilpotent of class exactly two and its derived subgroup has order p).

[L3]

For a finite group G and HG, G=[G:H]H (Lagrange's theorem: G=[G:H]H for every subgroup H of a finite group G).

[L4]

[G:H]:=G/H, the number of left cosets of H in G (The coset set G/H and the index [G:H] of a subgroup).

[L5]

A finite p-group is a finite group whose order has the form P=pn for some nN (A finite p-group has order pn for a prime p and some nN).

Proof

technique · direct
1.1

For every gP one has [g,x]x=(gxg1x1)x=gxg1, so each conjugate of x has the form [g,x]x.

F3algebra
1.2

The derived subgroup of P is Z(P) and has order p, and P is a finite p-group, so P=pn for some n.

L1L5
1.3

The size of the class of x is the index [P:CP(x)], and by Lagrange that index divides P.

F1L2L3L4
2.1

Each [g,x] lies in P=Z(P), so every conjugate of x lies in the right coset Z(P)x; the map zzx is a bijection from Z(P) onto that coset, so the coset has p elements and the class of x has at most p.

F4step 1.1step 1.2algebra
2.2

Since xZ(P), some gP fails to commute with x, so CP(x)P, its index is greater than one, and the class of x has more than one element.

F1F2step 1.3
3.1

The class size divides pn, so it is a power of p; it lies strictly between 1 and p inclusive, and the only such power of p is p itself.

step 1.2step 1.3step 2.1step 2.2

Remarks

The hypothesis xZ(P) is used only at step 2.2, and it is used to rule out the class of size one. A central x runs through the same computation and comes out with the class {x}, which is consistent with step 2.1 and shows that the two cases exhaust the group.

CorollaryStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-26Open item page →

A noncentral element of an extraspecial p-group has centraliser of index p

Statement

Let P be an extraspecial p-group and let xPZ(P). Then

[P:CP(x)]=p,equivalentlyCP(x)=Pp.

Facts & Assumptions

Given: An extraspecial p-group P and an element xP with xZ(P).

[F1]

CG(x):={gG:gx=xg}={gG:gxg1=x} (The conjugacy class ClG(x) and centralizer CG(x) of an element).

[F2]

[G:H]:=G/H, the number of left cosets of H in G (The coset set G/H and the index [G:H] of a subgroup).

[L1]

Every conjugacy class of an extraspecial p-group whose representative is not central has exactly p elements (Every conjugacy class of an extraspecial p-group outside the centre has exactly p elements).

[L3]

For a finite group G and HG, G=[G:H]H (Lagrange's theorem: G=[G:H]H for every subgroup H of a finite group G).

Proof

technique · direct
1.1

The class of x has exactly p elements, because x is not central.

L1
1.2

The size of that class is the index of the centraliser of x in P.

F1F2L2
2.1

Hence [P:CP(x)]=p, and Lagrange turns this into P=pCP(x).

L3step 1.1step 1.2

Remarks

Every centraliser named here is proper, since x is noncentral, and maximal in the order sense: index p is the smallest index a proper subgroup of a finite p-group can have.

DefinitionDefinition: Literature-sourcedProof: Not applicablejudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-26Open item page →

The central product GαH of two groups along an isomorphism of central subgroups

Definition

Let G and H be groups, let Z1Z(G) and Z2Z(H) be subgroups of their centres (The center Z(G) of a group, Subgroup), and let α:Z1Z2 be an isomorphism (Group isomorphisms, automorphisms and the set Aut(G)). Inside the external direct product G×H (The external direct product G×H with componentwise multiplication, G×H is a group with identity (eG,eH), coordinatewise inverses, and homomorphic coordinate projections) put

N:={(z,α(z)1):zZ1}.

For groups G,H with central subgroups Z1Z(G), Z2Z(H) and an isomorphism α:Z1Z2, the central product GαH is the quotient of G×H by N={(z,α(z)1):zZ1}, formed as in The quotient group G/N and coset product (gN)(hN)=ghN:

GαH:=(G×H)/N.

That N is a normal subgroup, so that the quotient is defined (Normal subgroup: invariance under conjugation), is proved in The identified subgroup used to form a central product is central, hence normal .

Write π:G×HGαH for the quotient map and gˉ:=π(g,e), hˉ:=π(e,h) for the canonical images of gG and hH.

Remarks

The identification is along α and reverses the second coordinate: killing (z,α(z)1) is exactly what makes zˉ=α(z) hold in the quotient, so the two identified central subgroups become one. Killing (z,α(z)) instead would identify z with α(z)1, which is the central product along α1 composed with inversion rather than along α.

Craven writes GH for a central product and van Beek writes GH; the notation α is used here because the isomorphism is part of the data and different choices of α can give non-isomorphic quotients.

Taking Z1=Z2=1 gives N=1 and recovers the direct product, so the construction is a genuine generalisation and not a separate object.

LemmaStatement: AI-adaptedProof: AI-adaptedprecheck passaudited 2026-08-26Open item page →

The identified subgroup used to form a central product is central, hence normal

Statement

Let G and H be groups with central subgroups Z1Z(G) and Z2Z(H) and an isomorphism α:Z1Z2. The subgroup N={(z,α(z)1):zZ1} of G×H is central, hence normal, so the quotient (G×H)/N of The central product GαH of two groups along an isomorphism of central subgroups is defined.

Facts & Assumptions

Given: Groups G,H, central subgroups Z1Z(G) and Z2Z(H), and an isomorphism α:Z1Z2.

[F1]

For groups G,H with central subgroups Z1Z(G), Z2Z(H) and an isomorphism α:Z1Z2, the central product GαH is the quotient of G×H by N={(z,α(z)1):zZ1} (The central product GαH of two groups along an isomorphism of central subgroups).

[F2]

A subset HG is a subgroup when eH, H is closed under the operation, and H is closed under inverses (Subgroup).

[F3]

Z(G):={zG:zg=gz for every gG} (The center Z(G) of a group).

[F4]

A subgroup NG is normal in G when gNg1=N for every gG, where gNg1:={gng1:nN} (Normal subgroup: invariance under conjugation).

[L2]

The external direct product G×H:={(g,h):gG, hH} carries the componentwise operation (g,h)(g,h)=(gg,hh) (The external direct product G×H with componentwise multiplication).

[L3]

The componentwise operation makes G×H a group with identity (eG,eH) and (g,h)1=(g1,h1) (G×H is a group with identity (eG,eH), coordinatewise inverses, and homomorphic coordinate projections).

Proof

technique · direct
1.1

The isomorphism α is in particular a homomorphism, so α(e)=e and α(z1)=α(z)1; moreover Z2Z(H), so any two elements of Z2 commute and α(z1z2)1=α(z1)1α(z2)1.

F3L1algebra
2.1

Hence (e,e)=(e,α(e)1) lies in N; the product (z1,α(z1)1)(z2,α(z2)1)=(z1z2,α(z1)1α(z2)1)=(z1z2,α(z1z2)1) lies in N; and (z,α(z)1)1=(z1,α(z))=(z1,α(z1)1) lies in N. So N is a subgroup of G×H.

F1F2L2L3step 1.1
3.1

Every element of N has first coordinate in Z1Z(G) and second coordinate in Z2Z(H), and the operation is componentwise, so (z,α(z)1)(g,h)=(zg,α(z)1h)=(gz,hα(z)1)=(g,h)(z,α(z)1) for every (g,h); thus NZ(G×H).

F1F3L2step 2.1
4.1

For nN and xG×H centrality gives xnx1=nxx1=n, so xNx1=N and N is normal; the quotient (G×H)/N is therefore defined.

F4step 3.1

Remarks

Centrality is used twice. It makes the inverse-coordinate rule z(z,α(z)1) multiplicative, so that N is a subgroup, and it then makes that subgroup central and hence normal. For merely isomorphic subgroups the displayed antidiagonal need be neither a subgroup nor a normal subset, so the quotient construction does not apply.

PropositionStatement: AI-adaptedProof: AI-adaptedprecheck passjudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-26Open item page →

The two canonical maps into a central product are injective homomorphisms whose images commute, generate it, and meet in the identified centre

Statement

Let G and H be groups with central subgroups Z1Z(G) and Z2Z(H) and an isomorphism α:Z1Z2, and let π:G×HGαH be the quotient map. The canonical maps GGαH and HGαH are injective homomorphisms whose images commute elementwise, generate GαH, and meet in the image of Z1. They are ggˉ=π(g,e) and hhˉ=π(e,h), and the intersection of their images is {zˉ:zZ1}={α(z):zZ1}.

Facts & Assumptions

Given: Groups G,H, central subgroups Z1Z(G) and Z2Z(H), an isomorphism α:Z1Z2, and the quotient map π:G×HGαH.

[F1]

For groups G,H with central subgroups Z1Z(G), Z2Z(H) and an isomorphism α:Z1Z2, the central product GαH is the quotient of G×H by N={(z,α(z)1):zZ1} (The central product GαH of two groups along an isomorphism of central subgroups).

[F2]

The quotient group G/N has the left cosets gN as elements with product (gN)(hN):=ghN (The quotient group G/N and coset product (gN)(hN)=ghN).

[F3]

For a group homomorphism f:GH, kerf:={gG:f(g)=eH} and imf:={f(g):gG} (The kernel and image of a group homomorphism).

[L1]

The subgroup N={(z,α(z)1):zZ1} of G×H is central, hence normal (The identified subgroup used to form a central product is central, hence normal).

[L2]

The external direct product G×H:={(g,h):gG, hH} carries the componentwise operation (g,h)(g,h)=(gg,hh) (The external direct product G×H with componentwise multiplication).

[L3]

S is the smallest subgroup of G containing S, namely {K:KG and SK} (The subgroup S generated by a subset, the cyclic subgroup g, and cyclic groups).

[L4]

An isomorphism is a bijective group homomorphism (Group isomorphisms, automorphisms and the set Aut(G)).

Proof

technique · direct
1.1

The coordinate maps g(g,e) and h(e,h) are homomorphisms into G×H, because the operation there is componentwise, and π is a homomorphism onto the quotient; so both canonical maps are homomorphisms.

F1F2L1L2
2.1

The kernel of ggˉ is {gG:(g,e)N}; an equality (g,e)=(z,α(z)1) gives g=z and α(z)=e, so z=e because α is injective, and the kernel is trivial. Likewise (e,h)=(z,α(z)1) gives z=e and then h=α(e)1=e. Both canonical maps are therefore injective.

F1F3L4step 1.1
2.2

In G×H one has (g,e)(e,h)=(g,h)=(e,h)(g,e), so gˉhˉ=hˉgˉ for all gG and hH: the two images commute elementwise.

L2step 1.1
2.3

Every element of GαH is π(g,h)=π((g,e)(e,h))=gˉhˉ, so the two images together generate GαH.

F2L2L3step 1.1
3.1

If gˉ=hˉ then (g,e)(e,h)1=(g,h1) lies in N, so g=zZ1 and h1=α(z)1, that is h=α(g); conversely zˉ=α(z) for every zZ1, since (z,α(z)1)N. Hence the two images meet exactly in {zˉ:zZ1}.

F1step 2.1step 2.2

Remarks

Injectivity is what makes the central product an honest amalgam: each factor embeds, and the only collapsing is the prescribed identification of Z1 with Z2. If α were merely a surjective homomorphism, N would meet the first coordinate copy of G in kerα×1 and that copy would not embed.

PropositionStatement: AI-adaptedProof: AI-adaptedprecheck passjudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-26Open item page →

Order, centre and derived subgroup of a central product

Statement

Let G and H be groups with central subgroups Z1Z(G) and Z2Z(H) and an isomorphism α:Z1Z2, and write gˉ,hˉ for the canonical images in P=GαH. Then:

  1. if G and H are finite, P=GHZ1;
  2. Z(P)={gˉhˉ:gZ(G), hZ(H)}, the image of Z(G)×Z(H);
  3. [P,P]={cd:c[G,G], d[H,H]}, the image of [G,G]×[H,H].

Facts & Assumptions

Given: Groups G,H, central subgroups Z1Z(G) and Z2Z(H), an isomorphism α:Z1Z2, the quotient map π:G×HP, and the canonical images gˉ=π(g,e), hˉ=π(e,h).

[F1]

For groups G,H with central subgroups Z1Z(G), Z2Z(H) and an isomorphism α:Z1Z2, the central product GαH is the quotient of G×H by N={(z,α(z)1):zZ1} (The central product GαH of two groups along an isomorphism of central subgroups).

[F2]

Z(G):={zG:zg=gz for every gG} (The center Z(G) of a group).

[F3]

For g,hG the commutator is [g,h]:=ghg1h1, and [G,G] is the subgroup generated by all commutators (Commutators [g,h]=ghg1h1 and the commutator subgroup [G,G]).

[L1]

The canonical maps GGαH and HGαH are injective homomorphisms whose images commute elementwise, generate GαH, and meet in the image of Z1 (The two canonical maps into a central product are injective homomorphisms whose images commute, generate it, and meet in the identified centre).

[L2]

If G and H are finite groups, then their external direct product is finite and has order G×H=GH (For finite groups G and H, G×H=GH).

[L3]

For a finite group G and HG, G=[G:H]H (Lagrange's theorem: G=[G:H]H for every subgroup H of a finite group G).

[L5]

The subgroup N={(z,α(z)1):zZ1} of G×H is central, hence normal (The identified subgroup used to form a central product is central, hence normal).

Proof

technique · direct
1.1

The map z(z,α(z)1) is a bijection from Z1 onto N, so N=Z1; with G×H=GH and Lagrange applied to NG×H, the number of cosets is GH/Z1, which is the order of P.

F1L2L3L5algebra
1.2

The images of G and of H in P commute elementwise, generate P, and each canonical map is injective.

L1
2.1

Let gˉhˉ be central in P and let gG. Then hˉ commutes with g, so gˉ commutes with g, hence [g,g]=ggg1g1 is trivial; injectivity of the canonical map gives [g,g]=e, so gZ(G), and symmetrically hZ(H).

F2F3step 1.2
2.2

Conversely, if gZ(G) and hZ(H) then gˉhˉ commutes with every g and with every h, and those elements generate P, so gˉhˉ is central.

F2step 1.2
2.3

Because the two images commute elementwise, [gˉ1hˉ1,gˉ2hˉ2]=[gˉ1,gˉ2][hˉ1,hˉ2]=[g1,g2] [h1,h2] for all giG and hiH.

F3step 1.2
3.1

Every element of P is gˉhˉ by step 1.2, so steps 2.1 and 2.2 identify Z(P) as the image of Z(G)×Z(H); and by step 2.3 the commutators of P are exactly the products [g1,g2] [h1,h2], whose generated subgroup is the image of [G,G]×[H,H].

F3L4step 1.1step 2.1step 2.2step 2.3

Remarks

Clause 1 needs both factors finite; clauses 2 and 3 do not, since they use only that the two images commute and generate. The order formula divides by Z1 and not by Z12: one copy of the identified subgroup survives inside the product, and it is the image described in the intersection clause of the canonical-maps proposition.

TheoremStatement: AI-adaptedProof: AI-adaptedprecheck passjudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-26Open item page →

Homomorphisms out of a central product

Statement

Let G and H be groups with central subgroups Z1Z(G) and Z2Z(H) and an isomorphism α:Z1Z2, and let K be a group. If φ:GK and ψ:HK are homomorphisms with commuting images and φZ1=ψα, then there is a unique homomorphism GαHK restricting to φ and ψ along the canonical maps. Explicitly it sends gˉhˉ to φ(g)ψ(h).

Facts & Assumptions

Given: Groups G,H,K, central subgroups Z1Z(G) and Z2Z(H), an isomorphism α:Z1Z2, and homomorphisms φ:GK, ψ:HK with φ(g)ψ(h)=ψ(h)φ(g) for all g,h and φ(z)=ψ(α(z)) for all zZ1.

[F1]

The external direct product G×H:={(g,h):gG, hH} carries the componentwise operation (g,h)(g,h)=(gg,hh) (The external direct product G×H with componentwise multiplication).

[F2]

For groups G,H with central subgroups Z1Z(G), Z2Z(H) and an isomorphism α:Z1Z2, the central product GαH is the quotient of G×H by N={(z,α(z)1):zZ1} (The central product GαH of two groups along an isomorphism of central subgroups).

[F3]

For a group homomorphism f:GH, kerf:={gG:f(g)=eH} (The kernel and image of a group homomorphism).

[F4]

The quotient group G/N has the left cosets gN as elements, with product (gN)(hN):=ghN (The quotient group G/N and coset product (gN)(hN)=ghN).

[L2]

For HG the rule (aH)(bH):=abH on left cosets is independent of the representatives a and b if and only if HG (Coset multiplication (gH)(hH)=ghH is well defined if and only if H is normal).

[L3]

The subgroup N={(z,α(z)1):zZ1} of G×H is central, hence normal (The identified subgroup used to form a central product is central, hence normal).

[L4]

The canonical maps GGαH and HGαH are injective homomorphisms whose images commute elementwise, generate GαH, and meet in the image of Z1 (The two canonical maps into a central product are injective homomorphisms whose images commute, generate it, and meet in the identified centre).

Proof

technique · direct
1.1

The assignment θ(g,h):=φ(g)ψ(h) satisfies θ((g,h)(g,h))=φ(g)φ(g)ψ(h)ψ(h)=φ(g)ψ(h)φ(g)ψ(h)=θ(g,h)θ(g,h), the middle equality being the commuting-images hypothesis; so θ:G×HK is a homomorphism.

F1givenalgebra
1.2

For zZ1, θ(z,α(z)1)=φ(z)ψ(α(z)1)=φ(z)ψ(α(z))1=φ(z)φ(z)1=e, so Nkerθ.

F2F3L1given
2.1

If xN=yN in G×H then x1yN, so θ(x)1θ(y)=θ(x1y)=e and θ(x)=θ(y); hence θˉ(xN):=θ(x) is a well-defined function on GαH, and it is a homomorphism because N is normal and (xN)(yN)=xyN.

F4L2L3step 1.1step 1.2
3.1

On the canonical images, θˉ(gˉ)=θ(g,e)=φ(g) and θˉ(hˉ)=θ(e,h)=ψ(h); and any homomorphism agreeing with φ and ψ on the two images agrees with θˉ on a generating set of GαH, hence everywhere.

L1L4step 2.1

Remarks

Both hypotheses are needed and neither is implied by the other. Without commuting images the assignment of step 1.1 is not a homomorphism on the direct product; without the agreement on Z1 the subgroup N need not lie in the kernel, so nothing descends to the quotient.

DefinitionDefinition: Literature-sourcedProof: Not applicablejudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-26Open item page →

Internal central products of a finite family of subgroups

Definition

Let G be a group and let G1,,Gr be subgroups of G, where rN (Subgroup). Subgroups G1,,Gr of G form an internal central product when they generate G and [Gi,Gj]=1 for ij, where is the generated subgroup of The subgroup S generated by a subset, the cyclic subgroup g, and cyclic groups and [Gi,Gj] is the subgroup commutator of Subgroup commutators and the lower central series. The empty family is an internal central product of the trivial group.

Since the factors commute pairwise, G=G1G2Gr as a set of products, and for ij every element of GiGj commutes with both Gi and Gj, so GiGjZ(Gi)Z(Gj) (The center Z(G) of a group).

Remarks

The condition differs from that of an internal direct product (Internal direct products of finitely many normal subgroups) in exactly one place: there the factors are required to intersect trivially, here they are allowed to share a central subgroup. An internal direct product of normal subgroups is in particular an internal central product, since distinct factors of a direct product commute elementwise.

No hypothesis is placed on the intersections beyond what the commuting condition already forces. That is deliberate: the intersections are what the recognition theorem computes, rather than data prescribed in advance.

TheoremStatement: AI-adaptedProof: AI-adaptedprecheck passaudited 2026-08-26Open item page →

Internal central products are the images of external ones

Statement

Let G be a group and G1,,GrG. Subgroups G1,,Gr form an internal central product of G if and only if the multiplication map G1××GrG, (g1,,gr)g1gr, is a surjective homomorphism; each factor then meets its kernel trivially.

For two factors this identifies the internal notion with the external one: if G1,G2 form an internal central product of G and D=G1G2, then DZ(G1) and DZ(G2), and

G    G1idDG2,

the external central product of The central product GαH of two groups along an isomorphism of central subgroups taken along the identity isomorphism of D.

Facts & Assumptions

Given: A group G and subgroups G1,,GrG; in the second half, r=2 and D=G1G2.

[F1]

Subgroups G1,,Gr of G form an internal central product when they generate G and [Gi,Gj]=1 for ij (Internal central products of a finite family of subgroups).

[F2]

The external direct product G×H:={(g,h):gG, hH} carries the componentwise operation (The external direct product G×H with componentwise multiplication).

[F3]

For groups G,H with central subgroups Z1Z(G), Z2Z(H) and an isomorphism α:Z1Z2, the central product GαH is the quotient of G×H by N={(z,α(z)1):zZ1} (The central product GαH of two groups along an isomorphism of central subgroups).

[F4]

For a group homomorphism f:GH, kerf:={gG:f(g)=eH} and imf:={f(g):gG} (The kernel and image of a group homomorphism).

[L1]

For every homomorphism f:GH, the rule gkerff(g) is an isomorphism from G/kerf onto imf (First isomorphism theorem for groups: G/kerfimf).

[L3]

Z(G):={zG:zg=gz for every gG} (The center Z(G) of a group).

Proof

technique · direct
1.1

Suppose the subgroups generate G and commute pairwise. Writing μ(g1,,gr)=g1gr, the commuting hypothesis lets the factors of μ(x)μ(y) be sorted by index, so μ(xy)=μ(x)μ(y) and μ is a homomorphism; its image is a subgroup containing every Gi, hence equals G1,,Gr=G, so μ is surjective.

F1F2F4L2algebra
1.2

For the converse, suppose μ is a surjective homomorphism. Surjectivity gives G=G1GrG1,,Gr, so the subgroups generate. For ij, aGi and bGj, the tuples x with a in place i and y with b in place j commute in the direct product, so ab=μ(x)μ(y)=μ(xy)=μ(yx)=μ(y)μ(x)=ba; hence [Gi,Gj]=1.

F1F2L2algebra
1.3

In either case a tuple with a single nonidentity entry gi has μ-value gi, so it lies in kerμ only if gi=e: each factor meets the kernel trivially.

F4
2.1

Now let r=2 and let G1,G2 form an internal central product with D=G1G2. An element dD lies in G2, so it commutes with every element of G1, giving DZ(G1); symmetrically DZ(G2).

F1L3step 1.1
2.2

The kernel of μ:G1×G2G is {(g1,g2):g1g2=e}={(d,d1):dD}, since g1=g21 lies in both subgroups.

F4step 1.1
3.1

That kernel is exactly the subgroup N used to build G1idDG2, so the first isomorphism theorem gives G(G1×G2)/N=G1idDG2.

F3L1step 1.1step 2.1step 2.2

Remarks

The commuting condition is imposed only for ij. A single factor is not required to be abelian, which is what allows a nonabelian group to be an internal central product of one factor, namely itself.

The identity isomorphism of D is forced here rather than chosen: the kernel of the multiplication map is {(d,d1)}, and that is the identified subgroup of the external product along idD and along no other map.

TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-26Open item page →

A central product of extraspecial p-groups identified along their centres is extraspecial

Statement

Let E1 and E2 be extraspecial p-groups and let α:Z(E1)Z(E2) be an isomorphism. Then P=E1αE2 is extraspecial, of order E1E2/p, and its centre and derived subgroup are the common image of Z(E1) and Z(E2).

Facts & Assumptions

Given: Extraspecial p-groups E1,E2, an isomorphism α:Z(E1)Z(E2), and P=E1αE2 with canonical images gˉ for gE1 and hˉ for hE2.

[F1]

A finite p-group P is special when Z(P)=P=Φ(P) is elementary abelian, and extraspecial when in addition P is nonabelian and this common subgroup has order p (Special and extraspecial p-groups).

[F2]

A finite p-group is a finite group whose order has the form P=pn for some nN (A finite p-group has order pn for a prime p and some nN).

[L1]

For a finite p-group P the following are equivalent: P is extraspecial; P is nonabelian, Z(P)=p and P/Z(P) is elementary abelian; P is nonabelian and Z(P)=P=Φ(P) has order p (Three equivalent descriptions of an extraspecial p-group).

[L2]

For a central product of finite groups, GαH=GH/Z1; without a finiteness hypothesis, its centre is the image of Z(G)×Z(H) and its derived subgroup is the image of [G,G]×[H,H] (Order, centre and derived subgroup of a central product).

[L3]

The canonical maps GGαH and HGαH are injective homomorphisms whose images commute elementwise, generate GαH, and meet in the image of Z1 (The two canonical maps into a central product are injective homomorphisms whose images commute, generate it, and meet in the identified centre).

[L4]

For NG, the quotient G/N is abelian if and only if [G,G]N (G/N is abelian if and only if [G,G]N).

[L5]

An elementary abelian p-group is a finite abelian p-group in which every nonidentity element has order p (Elementary abelian p-groups).

Proof

technique · direct
1.1

Each Ei is nonabelian, has Z(Ei)=[Ei,Ei] of order p, and has elementary abelian central quotient Ei/Z(Ei).

L1
1.2

The canonical maps embed E1 and E2 in P; their images commute elementwise and generate P.

L3
2.1

The central-product formulas give P=E1E2/p, which is a power of p; the centre of P is the image of Z(E1)×Z(E2) and the derived subgroup of P is the image of [E1,E1]×[E2,E2]. Those two subgroups of E1×E2 coincide by step 1.1, so Z(P)=[P,P]; the subgroup Z(E1)×Z(E2) has order p2 and contains the identified subgroup of order p, so its image has order p.

F2L2step 1.1algebra
2.2

The group P is nonabelian, because the canonical map embeds the nonabelian group E1 into it.

step 1.1step 1.2
3.1

Since [P,P]=Z(P), the quotient P/Z(P) is abelian; and for gE1, hE2 the commuting images give (gˉhˉ)p=gˉphˉp=gp hp, where gpZ(E1) and hpZ(E2) because the central quotients are elementary abelian, so (gˉhˉ)p lies in Z(P). Every element of P has this form, so every element of the finite abelian p-group P/Z(P) has order dividing p and P/Z(P) is elementary abelian.

L4L5step 1.1step 1.2step 2.1
4.1

So P is a nonabelian finite p-group with Z(P)=p and elementary abelian central quotient, which is the second description in the characterisation; hence P is extraspecial.

F1L1step 2.1step 2.2step 3.1

Remarks

The identification must be along the full centres: if a proper subgroup of Z(E1) were identified it would be trivial, the product would be the direct product, and its centre would have order p2. That is the case recorded on the companion page as a special group which is not extraspecial.

DefinitionDefinition: AI-adaptedProof: Not applicablejudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-26Open item page →

The commutator pairing of an extraspecial p-group relative to a chosen generator of its centre

Definition

Let P be an extraspecial p-group (Special and extraspecial p-groups) and fix a generator z of its centre, so that Z(P)=z has order p. Write

V:=P/Z(P),

which is elementary abelian (Three equivalent descriptions of an extraspecial p-group, Elementary abelian p-groups, The quotient group G/N and coset product (gN)(hN)=ghN), and give it its canonical Fp-vector-space structure (An elementary abelian p-group has a canonical Fp-vector-space structure), where Fp=Z/p is the field of For every prime p, the two operations on Z/p make it a field and For every natural n, (Z/n,+) is an abelian group, multiplication is a commutative monoid operation, and both distributive laws hold, the group operation of V is vector addition and the scalar action is aˉxˉ=xˉa. Elements of V are written multiplicatively, xˉ denoting the coset xZ(P); scalars are written additively.

The commutator pairing of P relative to z is the map

bz:V×VFp,[x,y]=zbz(xˉ,yˉ).

Why the exponent exists and is unique. Every commutator of P lies in [P,P]=Z(P)=z (An extraspecial p-group is nilpotent of class exactly two and its derived subgroup has order p, Commutators [g,h]=ghg1h1 and the commutator subgroup [G,G], The center Z(G) of a group), so [x,y]=zk for some integer k. Since z has order p (The order G of a finite group and the order ord(g) of an element, with ord(g)= when no positive power of g is the identity, A finite group of prime order is cyclic and every nonidentity element generates it), zk depends only on the class of k in Z/p, and zk=zk forces kk; so the class kˉFp is determined by [x,y], and that class is bz(xˉ,yˉ).

That the value depends only on the cosets xˉ and yˉ, and not on the representatives x and y, is proved in The commutator pairing is well defined on the central quotient, is bilinear over Fp, and is alternating .

Remarks

The pairing depends on the choice of z, and only on it: replacing z by zc with c0 replaces bz by c1bz. So the radical, the orthogonality relation and every statement about a subspace being self-orthogonal are independent of the choice, while the individual values are not. The companion page records the false statement that no choice is needed.

Nothing here is imported from a theory of bilinear forms. The target Fp is the field Z/p, the vector-space structure on V is the canonical scalar action of an elementary abelian p-group, and every property of bz used below is proved from the commutator identities of a group whose derived subgroup is central.

LemmaStatement: AI-adaptedProof: AI-adaptedprecheck passjudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-26Open item page →

The commutator pairing is well defined on the central quotient, is bilinear over Fp, and is alternating

Statement

Let P be an extraspecial p-group with Z(P)=z and V=P/Z(P). The commutator pairing bz is well defined on V×V: the value of bz(xˉ,yˉ) does not depend on the representatives x and y. It is Fp-bilinear,

bz(xˉyˉ,wˉ)=bz(xˉ,wˉ)+bz(yˉ,wˉ),bz(wˉ,xˉyˉ)=bz(wˉ,xˉ)+bz(wˉ,yˉ),bz(xˉa,yˉ)=abz(xˉ,yˉ)=bz(xˉ,yˉa),

and alternating: bz(xˉ,xˉ)=0 for every xˉV. Consequently bz(yˉ,xˉ)=bz(xˉ,yˉ).

Facts & Assumptions

Given: An extraspecial p-group P with Z(P)=z of order p, the quotient V=P/Z(P) with its canonical Fp-structure, and the pairing bz defined by [x,y]=zbz(xˉ,yˉ).

[F1]

The commutator pairing of P relative to z is the map bz:V×VFp determined by [x,y]=zbz(xˉ,yˉ), where V=P/Z(P) carries the canonical scalar action aˉxˉ=xˉa (The commutator pairing of an extraspecial p-group relative to a chosen generator of its centre).

[F2]

For g,hG the commutator is [g,h]:=ghg1h1 (Commutators [g,h]=ghg1h1 and the commutator subgroup [G,G]).

[F3]

The rule aˉx=xa gives every elementary abelian p-group its canonical Fp-vector-space structure, with the group operation as vector addition and the identity as zero (An elementary abelian p-group has a canonical Fp-vector-space structure).

[L1]

An extraspecial p-group is nilpotent of class exactly two, so its derived subgroup is central, and every nonidentity commutator has order p (An extraspecial p-group is nilpotent of class exactly two and its derived subgroup has order p).

[L2]

If [G,G]Z(G) then [xy,w]=[x,w][y,w], [x,yw]=[x,y][x,w], and [xn,y]=[x,y]n=[x,yn] for every integer n (Commutator identities in a group whose derived subgroup is central).

[L3]

Z(G):={zG:zg=gz for every gG} (The center Z(G) of a group).

[L4]

For every prime p, the operations of addition and multiplication on Z/p make it a field (For every prime p, the two operations on Z/p make it a field).

Proof

technique · direct
1.1

The derived subgroup of P is central, so the two expansion identities and the power identity are available for all elements of P.

L1L2
1.2

Alternation: [x,x]=xxx1x1=e=z0, so bz(xˉ,xˉ)=0.

F1F2
2.1

If xˉ=x then x=xu with uZ(P), and [xu,y]=[x,y][u,y]=[x,y] because u is central makes [u,y]=uyu1y1=e; the second variable is the same computation with the other expansion identity. So bz(xˉ,yˉ) depends only on the two cosets.

F1F2L3step 1.1
2.2

Additivity in the first variable: zbz(xˉyˉ,wˉ)=[xy,w]=[x,w][y,w]=zbz(xˉ,wˉ)zbz(yˉ,wˉ)=zbz(xˉ,wˉ)+bz(yˉ,wˉ), and exponents of z are determined modulo p; the second variable is symmetric.

F1F3step 1.1
2.3

Compatibility with scalars: for an integer a, zbz(xˉa,yˉ)=[xa,y]=[x,y]a=zabz(xˉ,yˉ), and both sides depend only on a modulo p because z has order p; the scalar action on V is aˉxˉ=xˉa, so this is exactly bz(axˉ,yˉ)=abz(xˉ,yˉ), and likewise in the second variable.

F1F3L4step 1.1
3.1

Expanding 0=bz(xˉyˉ,xˉyˉ) by steps 2.2 and 1.2 gives 0=bz(xˉ,yˉ)+bz(yˉ,xˉ), so bz(yˉ,xˉ)=bz(xˉ,yˉ).

step 2.2step 1.2

Remarks

Alternation is the primitive property and skew symmetry is derived from it, not the other way round. At p=2 the two are not interchangeable: there 1=1, so skew symmetry says only that the pairing is symmetric, and it is the vanishing of bz(xˉ,xˉ) that carries content.

LemmaStatement: AI-adaptedProof: AI-adaptedprecheck passaudited 2026-08-26Open item page →

The commutator pairing of an extraspecial p-group has trivial radical

Statement

Let P be an extraspecial p-group with Z(P)=z and V=P/Z(P), and let bz be its commutator pairing. The radical of bz is trivial: if xˉV satisfies bz(xˉ,yˉ)=0 for every yˉV, then xˉ is the identity of V. Conversely the identity of V pairs to zero with every element.

Facts & Assumptions

Given: An extraspecial p-group P with Z(P)=z, the quotient V=P/Z(P), and the commutator pairing bz.

[F1]

The commutator pairing of P relative to z is the map bz:V×VFp determined by [x,y]=zbz(xˉ,yˉ), where V=P/Z(P) (The commutator pairing of an extraspecial p-group relative to a chosen generator of its centre).

[F2]

Z(G):={zG:zg=gz for every gG} (The center Z(G) of a group).

[F3]

The quotient group G/N has the left cosets gN as elements (The quotient group G/N and coset product (gN)(hN)=ghN).

[L1]

The commutator pairing is well defined on V×V, is Fp-bilinear, and is alternating (The commutator pairing is well defined on the central quotient, is bilinear over Fp, and is alternating).

[L2]

For a finite p-group P the following are equivalent: P is extraspecial; P is nonabelian, Z(P)=p and P/Z(P) is elementary abelian; P is nonabelian and Z(P)=P=Φ(P) has order p (Three equivalent descriptions of an extraspecial p-group).

Proof

technique · direct
1.1

Let xP represent xˉ and suppose bz(xˉ,yˉ)=0 for every yˉV. Since every element of P represents some coset, this says [x,y]=z0=e for every yP, that is xy=yx for every yP.

F1L1
2.1

Hence xZ(P), so xˉ=xZ(P) is the identity coset of V.

F2F3L2step 1.1
3.1

Conversely, if xˉ is the identity of V then xZ(P), so [x,y]=e and bz(xˉ,yˉ)=0 for every yˉ.

F1F2F3step 2.1

Remarks

Triviality of the radical, rather than merely its smallness, is exactly the statement that the centre is the whole kernel of the quotient map. It is what lets a single element of V be detected by pairing it against the others, and it is used in that form by every counting argument below.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-08-26Open item page →

Two elements of an extraspecial p-group with nontrivial commutator generate an extraspecial subgroup of order p3

Statement

Let P be an extraspecial p-group with Z(P)=z, and let x,yP satisfy [x,y]e. Then Q=x,y contains Z(P), has order p3, is nonabelian, and is extraspecial with Z(Q)=Z(P).

Facts & Assumptions

Given: An extraspecial p-group P with Z(P)=z of order p, the quotient V=P/Z(P) with its commutator pairing bz, and elements x,yP with [x,y]e.

[F1]

The commutator pairing of P relative to z is the map bz:V×VFp determined by [x,y]=zbz(xˉ,yˉ) (The commutator pairing of an extraspecial p-group relative to a chosen generator of its centre).

[F2]

Z(G):={zG:zg=gz for every gG} (The center Z(G) of a group).

[F3]

A subset S of an elementary abelian p-group is independent when sSsas=e with finite support forces every as=0, and it spans when every element is such a product (Fp-spanning sets, independence, and bases in an elementary abelian p-group).

[F4]

An elementary abelian p-group is a finite abelian p-group in which every nonidentity element has order p (Elementary abelian p-groups).

[L1]

For a finite p-group P the following are equivalent: P is extraspecial; P is nonabelian, Z(P)=p and P/Z(P) is elementary abelian; P is nonabelian and Z(P)=P=Φ(P) has order p (Three equivalent descriptions of an extraspecial p-group).

[L2]

The commutator pairing is well defined, Fp-bilinear and alternating, and satisfies bz(yˉ,xˉ)=bz(xˉ,yˉ) (The commutator pairing is well defined on the central quotient, is bilinear over Fp, and is alternating).

[L3]

In a finite group whose order is prime, every ge has order G and generates G (A finite group of prime order is cyclic and every nonidentity element generates it).

[L4]

For a finite group G and HG, G=[G:H]H (Lagrange's theorem: G=[G:H]H for every subgroup H of a finite group G).

[L5]

If the quotient group G/Z(G) is cyclic, then G is abelian (If G/Z(G) is cyclic, then G is abelian).

[L7]

The rule aˉx=xa gives every elementary abelian p-group its canonical Fp-vector-space structure (An elementary abelian p-group has a canonical Fp-vector-space structure).

[L8]

If P is a finite p-group and HP, then H=pk for some kN (Every subgroup of a finite p-group has order a power of p).

Proof

technique · direct
1.1

The element [x,y] lies in [P,P]=Z(P) and is not the identity, so it generates the group Z(P) of order p; since [x,y]Q, this gives Z(P)Q.

F2L1L3L6
1.2

Write c=bz(xˉ,yˉ); then zc=[x,y]e, so c0 in Fp.

F1
2.1

The pair xˉ,yˉ is independent in V: if xˉayˉd is the identity of V, pairing with yˉ gives ac=0 and hence a=0, and pairing with xˉ gives dc=0 and hence d=0.

F3L2L7step 1.2
3.1

Since Z(P)Q, the image of Q in V is Q/Z(P), and it equals xˉ,yˉ={xˉayˉd:a,dFp}; independence makes the p2 displayed products pairwise distinct, so Q/Z(P)=p2 and Lagrange gives Q=pp2=p3.

F3L4L6L7step 1.1step 2.1
4.1

Q is nonabelian because [x,y]e, and Z(P)Z(Q) because an element central in P is central in the subgroup Q containing it. The order of Z(Q) is a power of p dividing p3 and is not p3; were it p2, the quotient Q/Z(Q) would have order p and hence be cyclic, forcing Q abelian. So Z(Q)=p and Z(Q)=Z(P).

F2L3L4L5L8step 1.1step 3.1
5.1

The quotient Q/Z(Q)=Q/Z(P) is a subgroup of the elementary abelian group V, hence is itself a finite abelian p-group all of whose nonidentity elements have order p; so Q is a nonabelian finite p-group with centre of order p and elementary abelian central quotient, and the characterisation makes it extraspecial.

F4L1step 3.1step 4.1

Remarks

The hypothesis is on the pair, not on either element separately: x and y are automatically noncentral, since a central element commutes with everything, but two noncentral elements can commute and then generate an abelian subgroup.

TheoremStatement: AI-adaptedProof: AI-adaptedprecheck passaudited 2026-08-26Open item page →

Every extraspecial p-group is an internal central product of nonabelian subgroups of order p3

Statement

Let P be an extraspecial p-group and Z(P)=z.

Splitting. If x,yP satisfy [x,y]e and F=x,y, then P=FCP(F) and FCP(F)=Z(P), so F and CP(F) form an internal central product of P (Internal central products of a finite family of subgroups, The centralizer CG(H) of a subgroup). Here F is extraspecial of order p3 with Z(F)=Z(P); and CP(F) has order P/p2 with Z(CP(F))=Z(P), and is extraspecial when P>p3.

Decomposition. There are n1 subgroups P1,,Pn of P, each nonabelian of order p3 with Z(Pi)=Z(P), which form an internal central product of P; call such a family admissible. Moreover P=p1+2n.

Peeling. For an admissible family, PiPj=Z(P) whenever ij; and when n2, for each index i the subgroup Ci=Pj:ji satisfies [Pi,Ci]=1, P=PiCi, PiCi=Z(P) and Z(Ci)=Z(P), is extraspecial of order p1+2(n1), and has (Pj)ji as an admissible family for itself.

Facts & Assumptions

Given: An extraspecial p-group P with Z(P)=z of order p, the quotient V=P/Z(P) and its commutator pairing bz.

[F1]

Subgroups G1,,Gr of G form an internal central product when they generate G and [Gi,Gj]=1 for ij (Internal central products of a finite family of subgroups).

[F2]

The commutator pairing of P relative to z is the map bz:V×VFp determined by [x,y]=zbz(xˉ,yˉ) (The commutator pairing of an extraspecial p-group relative to a chosen generator of its centre).

[F3]

Z(G):={zG:zg=gz for every gG} (The center Z(G) of a group).

[F4]

CG(H):={gG:gh=hg for every hH}, and CG(H)H=Z(H) (The centralizer CG(H) of a subgroup).

[L1]

For a finite p-group P the following are equivalent: P is extraspecial; P is nonabelian, Z(P)=p and P/Z(P) is elementary abelian; P is nonabelian and Z(P)=P=Φ(P) has order p (Three equivalent descriptions of an extraspecial p-group).

[L2]

The commutator pairing is well defined, Fp-bilinear and alternating, with bz(yˉ,xˉ)=bz(xˉ,yˉ) (The commutator pairing is well defined on the central quotient, is bilinear over Fp, and is alternating).

[L3]

The radical of the commutator pairing is trivial (The commutator pairing of an extraspecial p-group has trivial radical).

[L4]

If [x,y]e in an extraspecial p-group P, then x,y contains Z(P), has order p3, is nonabelian, and is extraspecial with centre Z(P) (Two elements of an extraspecial p-group with nontrivial commutator generate an extraspecial subgroup of order p3).

[L5]

Subgroups G1,,Gr form an internal central product of G if and only if the multiplication map G1××GrG is a surjective homomorphism; for two factors this identifies the internal product with the external central product along the identity on their intersection (Internal central products are the images of external ones).

[L6]

For a finite group G and HG, G=[G:H]H (Lagrange's theorem: G=[G:H]H for every subgroup H of a finite group G).

[L7]

For every group G and xG, the centralizer CG(x) is a subgroup of G (CG(x) and NG(H) are subgroups of G).

[L8]

An elementary abelian p-group is a finite abelian p-group in which every nonidentity element has order p (Elementary abelian p-groups).

[L9]

For every homomorphism f:GH, the rule gkerff(g) is an isomorphism from G/kerf onto imf (First isomorphism theorem for groups: G/kerfimf).

[L11]

Proof

technique · induction
1.1

If P=p3 the one-member family P1=P is admissible, and p3=p1+21; moreover any F=x,y with [x,y]e then has order p3 and equals P, so CP(F)=Z(P) has order p=P/p2 and the splitting clause holds.

F1F3F4L4L6base
1.2

Assume all three clauses of the theorem hold for every extraspecial p-group of order smaller than P.

ih
1.3

Suppose A,BP satisfy [A,B]=1, P=AB, AB=Z(P) and A=p3. An element of Z(B) commutes with B and, lying in B, with A, hence with AB=P, so Z(B)Z(P); and Z(P)=AB lies in B and is central in P, so Z(B)=Z(P). The multiplication map μ:A×BP is a surjective homomorphism. If μ(a,b)=e, then a=b1AB, while every (z,z1) with zAB lies in the kernel. Thus kerμ is the antidiagonal of AB and has order p, so P=p3B/p and B=P/p2. If B were abelian it would equal Z(B)=Z(P) and P would be p3; so for P>p3 the group B is nonabelian, and B/Z(B)=B/Z(P) is a subgroup of the elementary abelian group V, hence elementary abelian, so B is extraspecial.

F1F3F4L1L5L6L8L9
1.4

If [x,y]e then F=x,y is extraspecial of order p3 with Z(F)=Z(P), and β=bz(xˉ,yˉ) is a nonzero element of the field Fp, hence invertible.

F2L4L10L11
1.5

Let (P1,,Pn) be an admissible family and fix an index i, writing Ci=Pj:ji. Each generator of Ci commutes with every element of Pi, and the centraliser of an element is a subgroup, so [Pi,Ci]=1 and CiCP(Pi); then PiCi is a subgroup containing every member of the family, so P=PiCi; and PiCiPiCP(Pi)=Z(Pi)=Z(P), while Z(P)=Z(Pj)PjCi gives the reverse inclusion. Taking Ci to be a single Pj shows PiPj=Z(P).

F1F3F4L7L10
2.1

For gP put s=β1bz(gˉ,yˉ) and r=β1bz(gˉ,xˉ) and u=xsyr. Bilinearity and alternation give bz(uˉ,xˉ)=rbz(yˉ,xˉ)=rβ=bz(gˉ,xˉ) and bz(uˉ,yˉ)=sbz(xˉ,yˉ)=sβ=bz(gˉ,yˉ), so bz(gu1,xˉ)=0 and bz(gu1,yˉ)=0.

F2L2L11step 1.4
2.2

Applying step 1.3 with A=Pi and B=Ci gives Z(Ci)=Z(P) and Ci=P/p2; when n2 the subgroup Ci contains the nonabelian Pj, so P>p3 and Ci is extraspecial. The members Pj with ji generate it, commute pairwise and have centre Z(Ci), so they form an admissible family for it; since Ci has smaller order, the induction hypothesis applied to that (n1)-member family gives Ci=p1+2(n1).

F1F3ihstep 1.2step 1.3step 1.5
3.1

Hence gu1 commutes with x and with y; the centraliser of gu1 is a subgroup containing both, hence contains F, so gu1CP(F) and g=(gu1)uCP(F)F. Therefore P=FCP(F), and FCP(F)=Z(F)=Z(P), so F and CP(F) form an internal central product of P; step 1.3 with A=F and B=CP(F) supplies the order of CP(F), its centre, and that it is extraspecial when P>p3.

F1F4L7L10step 1.3step 1.4step 2.1
4.1

If P>p3, choose xZ(P), which exists because P is nonabelian, and then y with bz(xˉ,yˉ)0, which exists because the radical is trivial; put F=x,y and C=CP(F). By step 3.1 the group C is extraspecial of order P/p2, so the induction hypothesis gives an admissible family P2,,Pn for C with C=p1+2(n1). Then F,P2,,Pn generate P, commute pairwise, are nonabelian of order p3 and have centre Z(C)=Z(P), so they form an admissible family for P, and P=p2C=p1+2n; with step 1.1 this completes the induction.

F1F2F3L1L3L10ihstep 1.1step 1.2step 3.1discharge-induction

Remarks

The factors are not canonical: the subgroup F depends on the choice of x and of a partner y, and the companion page records two decompositions of one group with different factors. What the order formula does fix is the number of factors.

The splitting and peeling clauses are stated for an arbitrary noncommuting pair and an arbitrary admissible family because the classification arguments need to remove a factor of a prescribed isomorphism type, not the one this proof happens to construct.

CorollaryStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-08-26Open item page →

An extraspecial p-group has order p1+2n for some n1

Statement

Let P be an extraspecial p-group. Then P=p1+2n for some integer n1, and P/Z(P)=p2n. In particular no extraspecial group has order p2m, and none has order p.

Facts & Assumptions

Given: An extraspecial p-group P.

[F1]

Z(G):={zG:zg=gz for every gG} (The center Z(G) of a group).

[L1]

Every extraspecial p-group is an internal central product of n1 nonabelian subgroups of order p3 with pairwise intersections Z(P), and P=p1+2n (Every extraspecial p-group is an internal central product of nonabelian subgroups of order p3).

[L2]

For a finite p-group P the following are equivalent: P is extraspecial; P is nonabelian, Z(P)=p and P/Z(P) is elementary abelian; P is nonabelian and Z(P)=P=Φ(P) has order p (Three equivalent descriptions of an extraspecial p-group).

[L3]

For a finite group G and HG, G=[G:H]H (Lagrange's theorem: G=[G:H]H for every subgroup H of a finite group G).

[L4]

[G:H]:=G/H, the number of left cosets of H in G (The coset set G/H and the index [G:H] of a subgroup).

Proof

technique · direct
1.1

The decomposition theorem writes P as an internal central product of n1 nonabelian subgroups of order p3 and gives P=p1+2n.

L1
1.2

The centre of P has order p.

F1L2
2.1

Lagrange applied to Z(P)P gives P/Z(P)=P/p=p2n.

L3L4step 1.1step 1.2
3.1

Since 1+2n is odd, no extraspecial group has order an even power of p; and n1 excludes the order p, which corresponds to n=0.

step 1.1step 2.1

Remarks

The exponent n is determined by the order and therefore by the group, so it can be used as an invariant even though the decomposition producing it is not unique. That n1 is what the nonabelian clause of the definition buys: an abelian group with a centre of order p would be the cyclic group of order p, of order p1+20.

CorollaryStatement: AI-adaptedProof: AI-adaptedprecheck passjudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-26Open item page →

An extraspecial p-group of order p1+2n has generator rank 2n

Statement

Let P be an extraspecial p-group of order p1+2n. Then Φ(P)=Z(P) has order p, the Frattini quotient P/Φ(P) has order p2n, the generator rank is d(P)=2n, and every minimal generating set of P has exactly 2n elements.

Facts & Assumptions

Given: An extraspecial p-group P of order p1+2n.

[F1]

For a finite p-group P, the generator rank d(P) is the common size of a basis of P/Φ(P) (The generator rank d(P) of a finite p-group).

[F2]

A subset S of an elementary abelian p-group spans when every element is a product sSsas with coefficients in Fp, and is independent when such a product is the identity only for zero coefficients; a basis is an independent spanning subset (Fp-spanning sets, independence, and bases in an elementary abelian p-group).

[L1]

For a finite p-group P the following are equivalent: P is extraspecial; P is nonabelian, Z(P)=p and P/Z(P) is elementary abelian; P is nonabelian and Z(P)=P=Φ(P) has order p (Three equivalent descriptions of an extraspecial p-group).

[L2]

An extraspecial p-group has P=p1+2n with n1 and P/Z(P)=p2n (An extraspecial p-group has order p1+2n for some n1).

[L3]

Every finite elementary abelian p-group has a basis; every independent subset extends to a basis, every spanning subset contains a basis, and all bases have the same finite size (Finite elementary abelian p-groups have bases, basis extension, and a well-defined dimension).

[L4]

A subset X of a finite p-group P is a minimal generating set if and only if the quotient map restricts to a bijection from X onto a basis of P/Φ(P) (Burnside Basis Theorem).

[L5]

The rule aˉx=xa gives every elementary abelian p-group its canonical Fp-vector-space structure (An elementary abelian p-group has a canonical Fp-vector-space structure).

Proof

technique · direct
1.1

By the third description in the characterisation, Φ(P)=Z(P) has order p, so the Frattini quotient is the central quotient.

L1
1.2

The central quotient has order p2n and is elementary abelian.

L1L2
2.1

It therefore has a basis, and all of its bases have the same size, say k; independence and spanning make the map sending a coefficient family (a1,,ak) to the product of the corresponding powers a bijection from the pk coefficient families onto the group, so pk=p2n and k=2n.

F2L3L5step 1.1step 1.2
3.1

Hence d(P)=2n by the definition of the generator rank, and by the Burnside basis theorem a minimal generating set of P is carried bijectively onto a basis of P/Φ(P), so it has 2n elements.

F1L4step 2.1

Remarks

The two clauses say different things. The first is about the quotient and is a count of a basis; the second is about P itself and needs the Burnside basis theorem, because a generating set of P of size 2n could a priori collapse in the quotient. It is the restricted-bijection clause of that theorem which rules that out.

LemmaStatement: AI-adaptedProof: AI-adaptedprecheck passaudited 2026-08-26Open item page →

A subgroup of the central quotient and its orthogonal complement have orders multiplying to the order of the quotient

Statement

Let P be an extraspecial p-group of order p1+2n, let V=P/Z(P) and let bz be its commutator pairing. For a subgroup UV put

U:={vV:bz(v,u)=0 for every uU}.

Then U is a subgroup of V and

UU=V=p2n.

This is a statement about this pairing on this quotient, proved by counting inside V; no theory of bilinear forms on a vector space is used.

Facts & Assumptions

Given: An extraspecial p-group P of order p1+2n, the quotient V=P/Z(P) with its commutator pairing bz, and a subgroup UV.

[F1]

The commutator pairing of P relative to z is the map bz:V×VFp determined by [x,y]=zbz(xˉ,yˉ) (The commutator pairing of an extraspecial p-group relative to a chosen generator of its centre).

[F2]

A subset S of an elementary abelian p-group spans when every element is a product sSsas with coefficients in Fp, and is independent when such a product is the identity only for zero coefficients; a basis is an independent spanning subset (Fp-spanning sets, independence, and bases in an elementary abelian p-group).

[F3]

An elementary abelian p-group is a finite abelian p-group in which every nonidentity element has order p (Elementary abelian p-groups).

[L1]

The commutator pairing is well defined, Fp-bilinear and alternating (The commutator pairing is well defined on the central quotient, is bilinear over Fp, and is alternating).

[L2]

The radical of the commutator pairing is trivial (The commutator pairing of an extraspecial p-group has trivial radical).

[L3]

An extraspecial p-group has P=p1+2n with n1 and P/Z(P)=p2n (An extraspecial p-group has order p1+2n for some n1).

[L4]

Every finite elementary abelian p-group has a basis; every independent subset extends to a basis, every spanning subset contains a basis, and all bases have the same finite size (Finite elementary abelian p-groups have bases, basis extension, and a well-defined dimension).

[L5]

The rule aˉx=xa gives every elementary abelian p-group its canonical Fp-vector-space structure (An elementary abelian p-group has a canonical Fp-vector-space structure).

[L6]

For every homomorphism f:GH, the rule gkerff(g) is an isomorphism from G/kerf onto imf (First isomorphism theorem for groups: G/kerfimf).

[L7]

For a finite group G and HG, G=[G:H]H (Lagrange's theorem: G=[G:H]H for every subgroup H of a finite group G).

Proof

technique · direct
1.1

The group V is elementary abelian of order p2n, and U is a subgroup of it, hence also a finite abelian p-group whose nonidentity elements have order p.

F3L3L5
1.2

For each fixed uV the map vbz(v,u) is a homomorphism from V to the additive group of Fp, by additivity of bz in its first variable; U is the intersection of the kernels of these maps as u runs over U, hence a subgroup of V.

F1L1
2.1

Choose a basis u1,,uk of U and extend it to a basis u1,,um of V. Unique representation in a basis makes the assignment of coefficient families to elements a bijection, so U=pk and pm=V=p2n, whence m=2n.

F2L4L5step 1.1
3.1

Define Ψ:VV by Ψ(v)=i=1muibz(v,ui). It is a homomorphism, again by additivity in the first variable. If Ψ(v) is the identity then independence of the basis forces bz(v,ui)=0 for every i, and then additivity in the second variable gives bz(v,w)=0 for every wV, so v is the identity by triviality of the radical. Thus Ψ is injective, hence bijective because V is finite.

F1F2L1L2step 2.1
4.1

Let π:VU send i=1muiai to i=1kuiai; unique representation makes π a well-defined homomorphism, and it is surjective onto U because it fixes each ui with ik. The composite Φ=πΨ is therefore a surjective homomorphism from V onto U, and Φ(v) is the identity exactly when bz(v,ui)=0 for every ik, which by additivity in the second variable is exactly vU.

F2step 2.1step 3.1
5.1

The first isomorphism theorem and Lagrange applied to Φ give V=UU, that is UU=p2n.

L6L7step 2.1step 4.1

Remarks

The two ends of the range behave as the formula predicts and are worth naming. For the trivial subgroup the formula reads V=V, since the perpendicular of the trivial subgroup is all of V; for U=V it reads V=1, which is triviality of the radical.

PropositionStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-08-26Open item page →

In an extraspecial p-group of order p1+2n every maximal abelian subgroup has order p1+n

Statement

Let P be an extraspecial p-group of order p1+2n. Call an abelian subgroup of P maximal abelian when it is not properly contained in any abelian subgroup of P. Then maximal abelian subgroups exist, every one of them contains Z(P), and every one of them has order p1+n. Under the correspondence AA/Z(P) they are exactly the subgroups A with Z(P)AP whose image U=A/Z(P) in V=P/Z(P) satisfies U=U.

Facts & Assumptions

Given: An extraspecial p-group P of order p1+2n, the quotient V=P/Z(P) with its commutator pairing bz, and for UV the subgroup U={vV:bz(v,u)=0 for all uU}.

[F1]

The commutator pairing of P relative to z is the map bz:V×VFp determined by [x,y]=zbz(xˉ,yˉ) (The commutator pairing of an extraspecial p-group relative to a chosen generator of its centre).

[F2]

Z(G):={zG:zg=gz for every gG} (The center Z(G) of a group).

[L1]

The commutator pairing is well defined, Fp-bilinear and alternating, with bz(yˉ,xˉ)=bz(xˉ,yˉ) (The commutator pairing is well defined on the central quotient, is bilinear over Fp, and is alternating).

[L2]

An extraspecial p-group has P=p1+2n with n1 and P/Z(P)=p2n (An extraspecial p-group has order p1+2n for some n1).

[L3]

For a subgroup U of V, UU=V=p2n (A subgroup of the central quotient and its orthogonal complement have orders multiplying to the order of the quotient).

[L4]

For NG the maps HH/N and Kπ1(K) are inverse inclusion-preserving bijections between subgroups H with NHG and subgroups KG/N (Correspondence theorem: subgroups of G/N correspond to subgroups of G containing N, with normality preserved).

[L5]

For a finite group G and HG, G=[G:H]H (Lagrange's theorem: G=[G:H]H for every subgroup H of a finite group G).

[L6]

For a finite p-group P the following are equivalent: P is extraspecial; P is nonabelian, Z(P)=p and P/Z(P) is elementary abelian; P is nonabelian and Z(P)=P=Φ(P) has order p (Three equivalent descriptions of an extraspecial p-group).

Proof

technique · direct
1.1

If BP is abelian then BZ(P) is a subgroup, because Z(P) is normal, and it is abelian, because (bu)(bu)=bbuu=(bu)(bu) for u,uZ(P); it contains B. So a maximal abelian subgroup equals BZ(P) and contains Z(P).

F2algebra
1.2

For Z(P)AP with image U=A/Z(P): two elements a,a of A commute exactly when [a,a]=e, that is exactly when bz(aˉ,aˉ)=0; so A is abelian exactly when UU.

F1L1
1.3

The correspondence AA/Z(P) is an inclusion-preserving bijection between the subgroups of P containing Z(P) and the subgroups of V, and A=pA/Z(P) by Lagrange.

F2L4L5L6
1.4

If UU then U2UU=p2n, so Upn.

L2L3
1.5

The trivial subgroup satisfies UU, and V is finite, so among the subgroups with UU there is one of largest order, and it is maximal with that property.

L2
2.1

Combining the previous three observations, AA/Z(P) carries the maximal abelian subgroups of P bijectively onto the subgroups U of V that are maximal subject to UU.

step 1.1step 1.2step 1.3
2.2

Let U be maximal subject to UU and suppose UU. Pick vU with vU and set U=U,v, whose elements are the products uva. For such elements, bz(uva,uva)=bz(u,u)+abz(u,v)+abz(v,u)+aabz(v,v)=0, using UU, the choice of v, skew symmetry and alternation. So UU and U properly contains U, contradicting maximality. Hence U=U and U2=p2n, so U=pn.

F1F3L1L3step 1.4
3.1

Therefore maximal abelian subgroups exist, each contains Z(P), each corresponds to a subgroup U with U=U of order pn, and each has order ppn=p1+n.

step 1.5step 2.1step 2.2

Remarks

Maximality is under inclusion, not merely maximality of order, and the two agree here only because step 2.2 shows every maximal self-orthogonal subgroup has the same order. That is what makes the conclusion a statement about every maximal abelian subgroup rather than about a largest one.

CorollaryStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-26Open item page →

An extraspecial p-group is the product of two maximal abelian subgroups meeting in its centre

Statement

Let P be an extraspecial p-group of order p1+2n. Then there are maximal abelian subgroups A and B of P, each of order p1+n, with

P=AB,AB=Z(P),[A,B]=Z(P).

Facts & Assumptions

Given: An extraspecial p-group P of order p1+2n with Z(P)=z, the quotient V=P/Z(P) and its commutator pairing bz.

[F1]

The commutator pairing of P relative to z is the map bz:V×VFp determined by [x,y]=zbz(xˉ,yˉ) (The commutator pairing of an extraspecial p-group relative to a chosen generator of its centre).

[F2]

Z(G):={zG:zg=gz for every gG} (The center Z(G) of a group).

[F4]

For subgroups A,BG, [A,B]=[a,b]:aA, bB (Subgroup commutators and the lower central series).

[L1]

Every extraspecial p-group is an internal central product of n1 nonabelian subgroups P1,,Pn of order p3 with Z(Pi)=Z(P) and PiPj=Z(P) for ij (Every extraspecial p-group is an internal central product of nonabelian subgroups of order p3, Internal central products of a finite family of subgroups).

[L2]

The commutator pairing is well defined, Fp-bilinear and alternating (The commutator pairing is well defined on the central quotient, is bilinear over Fp, and is alternating).

[L3]

For a subgroup U of V, UU=V=p2n (A subgroup of the central quotient and its orthogonal complement have orders multiplying to the order of the quotient).

[L4]

Maximal abelian subgroups of P contain Z(P), have order p1+n, and under the correspondence AA/Z(P) are exactly the subgroups containing Z(P) whose image U satisfies U=U (In an extraspecial p-group of order p1+2n every maximal abelian subgroup has order p1+n).

[L5]

In a finite group whose order is prime, every ge has order G and generates G (A finite group of prime order is cyclic and every nonidentity element generates it).

[L6]

For NG the maps HH/N and Kπ1(K) are inverse inclusion-preserving bijections between subgroups H with NHG and subgroups KG/N (Correspondence theorem: subgroups of G/N correspond to subgroups of G containing N, with normality preserved).

[L7]

If HG and NG, then H/(HN)HN/N (Second isomorphism theorem for groups: H/(HN)HN/N).

[L8]

For every group G and xG, the centralizer CG(x) is a subgroup of G (CG(x) and NG(H) are subgroups of G).

[L9]

For a finite group G and HG, G=[G:H]H (Lagrange's theorem: G=[G:H]H for every subgroup H of a finite group G).

[L10]

For a finite p-group P the following are equivalent: P is extraspecial; P is nonabelian, Z(P)=p and P/Z(P) is elementary abelian; P is nonabelian and Z(P)=P=Φ(P) has order p (Three equivalent descriptions of an extraspecial p-group).

Proof

technique · direct
1.1

Fix an internal central product decomposition P=P1,,Pn with each Pi nonabelian of order p3, Z(Pi)=Z(P), [Pi,Pj]=1 for ij, and PiPj=Z(P).

F3L1
2.1

Each Pi is nonabelian, so it contains elements with nontrivial commutator; choose xi,wiPi with [xi,wi]e. That commutator lies in Z(P) and is not the identity, so it generates Z(P) and equals zci with ci0; putting yi=witi with tici=1 in Fp gives [xi,yi]=z.

F1F2L2L5L10step 1.1
3.1

The pairing values are bz(xˉi,yˉi)=1, while bz(xˉi,yˉj)=0, bz(xˉi,xˉj)=0 and bz(yˉi,yˉj)=0 whenever the two elements lie in different factors or are equal, because distinct factors commute elementwise and the pairing is alternating.

F1L2step 1.1step 2.1
4.1

Put A=Z(P),x1,,xn and B=Z(P),y1,,yn. Their generators commute pairwise, so the centraliser of each generator is a subgroup containing all of them and hence contains A (respectively B); therefore each generator is central in A (respectively B), and A and B are abelian.

F2F3L8step 2.1step 3.1
5.1

The images Aˉ=xˉ1,,xˉn and Bˉ=yˉ1,,yˉn consist of the products ixˉiai and iyˉici. If ixˉiai is the identity, pairing with yˉk gives ak=0, so the xˉi are independent and Aˉ=pn; the same argument with the roles exchanged gives Bˉ=pn. If ixˉiai=jyˉjcj, pairing with yˉk gives ak=0 and pairing with xˉk gives ck=0, so AˉBˉ is trivial.

F3L2step 3.1step 4.1
6.1

Since A is abelian its image satisfies AˉAˉ, and the counting formula gives Aˉ=p2n/pn=pn=Aˉ, so Aˉ=Aˉ and A is maximal abelian of order p1+n; likewise for B.

L3L4L9step 4.1step 5.1
7.1

In the abelian group V the product AˉBˉ is a subgroup and the second isomorphism theorem gives AˉBˉ=AˉBˉ/AˉBˉ=p2n=V, so AˉBˉ=V; lifting along the correspondence, every gP has gˉ=aˉbˉ, hence gabZ(P)AB because Z(P)A, and P=AB. The correspondence also gives (AB)/Z(P)=AˉBˉ trivial, so AB=Z(P).

L6L7L9step 5.1step 6.1
8.1

Finally [A,B] is generated by commutators of elements of P, so it lies in [P,P]=Z(P); and it contains [x1,y1]=z, which generates Z(P). Hence [A,B]=Z(P).

F4L5L10step 2.1step 7.1

Remarks

The two subgroups are built from a chosen decomposition and are not canonical; another decomposition may yield the same pair or a different one. What the statement fixes is that a factorisation of this shape exists, with both factors as large as an abelian subgroup can be.

DefinitionDefinition: Literature-sourcedProof: Not applicablejudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-26Open item page →

The Heisenberg group of order p3 over Z/p

Definition

Let p be a prime (Prime and composite integers: p is prime when p>1 and its only positive divisors are 1 and p) and let Z/p carry the addition and multiplication of For every natural n, (Z/n,+) is an abelian group, multiplication is a commutative monoid operation, and both distributive laws hold, which make it a field (For every prime p, the two operations on Z/p make it a field). The Heisenberg group of order p3 is the set

Hp:={(a,b,c):a,b,cZ/p}

with the multiplication

(a,b,c)(a,b,c):=(a+a,b+b,c+c+ab).

That this is a group law (Group and abelian group), that it is not commutative, and that Hp has p3 elements are proved in The Heisenberg multiplication is a group law, nonabelian, on a set of p3 elements .

Remarks

The same multiplication is written on the published example The finite Heisenberg group is the unique Sylow p-subgroup of its coordinate upper-triangular group, where the triple (a,b,c) records the three entries above the diagonal of a unipotent upper-triangular 3×3 matrix over Z/p; the two constructions produce the same group, and the notation is kept identical so that a reader meets one object rather than two.

The asymmetry of the third coordinate, c+c+ab rather than c+c+ab, is a choice of convention: the opposite choice gives the group with the roles of the first two coordinates exchanged, and the map (a,b,c)(b,a,c) carries one to the other. Nothing below depends on which is taken, provided one is taken throughout.

At p=2 the construction does not produce a group of exponent two: the element (1,1,0) squares to (0,0,1), so H2 has an element of order four.

PropositionStatement: AI-adaptedProof: AI-adaptedprecheck passaudited 2026-08-26Open item page →

The Heisenberg multiplication is a group law, nonabelian, on a set of p3 elements

Statement

Let p be a prime. The multiplication (a,b,c)(a,b,c)=(a+a,b+b,c+c+ab) makes the set Hp a group with identity (0,0,0) and inverse (a,b,c)1=(a,b,c+ab); the group is not abelian; and Hp=p3. Moreover (1,0,0)a=(a,0,0), (0,1,0)b=(0,b,0) and (0,0,1)c=(0,0,c) for all a,b,c, so those three elements generate Hp (The subgroup S generated by a subset, the cyclic subgroup g, and cyclic groups) and each has order p (The order G of a finite group and the order ord(g) of an element, with ord(g)= when no positive power of g is the identity, Powers gn: natural exponents in a monoid and integer exponents in a group, with g0=e).

Facts & Assumptions

Given: A prime p and the set Hp with the multiplication above.

[F1]

The Heisenberg group of order p3 is the set Hp={(a,b,c):a,b,cZ/p} with (a,b,c)(a,b,c):=(a+a,b+b,c+c+ab) (The Heisenberg group of order p3 over Z/p).

[F2]

A group is a set with an associative operation having a two-sided identity and two-sided inverses (Group and abelian group).

[L1]

For every nN, (Z/n,+,[0]n) is an abelian group, (Z/n,,[1]n) is a commutative monoid, and multiplication distributes over addition on both sides (For every natural n, (Z/n,+) is an abelian group, multiplication is a commutative monoid operation, and both distributive laws hold).

[L2]

If A and B are finite then A×B is finite and A×B=AB (The product rule: A×B=AB, and i<mAi=i<mAi).

[L3]

A is the unique natural number n with An (The cardinality A of a finite set).

[L4]

Proof

technique · direct
1.1

Associativity: both ((a,b,c)(a,b,c))(a,b,c) and (a,b,c)((a,b,c)(a,b,c)) have first coordinate a+a+a, second coordinate b+b+b, and third coordinate c+c+c+ab+ab+ab, the two computations differing only in the order in which the three products are formed.

F1L1algebra
1.2

The triple (0,0,0) is a two-sided identity: (a,b,c)(0,0,0)=(a,b,c+0+a0)=(a,b,c) and (0,0,0)(a,b,c)=(a,b,0+c+0b)=(a,b,c).

F1L1algebra
1.3

The triple (a,b,c+ab) is a two-sided inverse of (a,b,c): the product in one order is (0,0,c+(c+ab)+a(b))=(0,0,0), and in the other it is (0,0,(c+ab)+c+(a)b)=(0,0,0).

F1L1algebra
1.4

The quotient set Z/p has the p distinct classes [0]p,,[p1]p, so it has p elements. The underlying set of Hp is the threefold product of those p-element sets, and [L2] therefore gives Hp=p3.

F1L2L3algebra
1.5

The three displayed power formulas hold because (a,0,0)(1,0,0)=(a+1,0,0+0+a0)=(a+1,0,0), (0,b,0)(0,1,0)=(0,b+1,0+0+01)=(0,b+1,0) and (0,0,c)(0,0,1)=(0,0,c+1), so each power is obtained from the previous one by adding one in the relevant coordinate.

F1L1algebra
2.1

By steps 1.1 to 1.3 the multiplication makes Hp a group.

F2step 1.1step 1.2step 1.3
2.2

It is not abelian: (1,0,0)(0,1,0)=(1,1,1) while (0,1,0)(1,0,0)=(1,1,0), and these differ because 10 in Z/p for every prime p.

F1L4step 1.4
3.1

Each of (1,0,0), (0,1,0) and (0,0,1) has p-th power (0,0,0) by step 1.5, hence order p since each is not the identity; and (a,0,0)(0,b,0)(0,0,cab)=(a,b,ab)(0,0,cab)=(a,b,c), so the three elements generate Hp.

L1step 1.5step 2.1step 2.2

Remarks

The verification of associativity is where the third coordinate earns its shape: the two bracketings produce the cross terms ab+ab and ab+ab respectively together with the common term ab, and they agree because multiplication in Z/p distributes over addition.

PropositionStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-26Open item page →

The Heisenberg group of order p3 is extraspecial, and for odd p it has exponent p

Statement

Let p be a prime and let Hp be the Heisenberg group of order p3. Then

Z(Hp)=[Hp,Hp]={(0,0,c):cZ/p},

a subgroup of order p, and Hp is extraspecial. For odd p the exponent of Hp is p. At p=2 the exponent is 4, since (1,1,0)2=(0,0,1).

Facts & Assumptions

Given: A prime p and the Heisenberg group Hp.

[F1]

The Heisenberg group of order p3 is the set Hp={(a,b,c):a,b,cZ/p} with (a,b,c)(a,b,c):=(a+a,b+b,c+c+ab) (The Heisenberg group of order p3 over Z/p).

[F2]

Z(G):={zG:zg=gz for every gG} (The center Z(G) of a group).

[F3]

For g,hG the commutator is [g,h]:=ghg1h1, and [G,G] is the subgroup generated by all commutators (Commutators [g,h]=ghg1h1 and the commutator subgroup [G,G]).

[F4]

For a finite group G, exp(G)=min{nN:n>0 and gn=e for every gG} (The exponent of a finite group).

[L1]

The Heisenberg multiplication is a group law with identity (0,0,0) and inverse (a,b,c)1=(a,b,c+ab); the group is not abelian, Hp=p3, and (1,0,0)a=(a,0,0), (0,1,0)b=(0,b,0), (0,0,1)c=(0,0,c) (The Heisenberg multiplication is a group law, nonabelian, on a set of p3 elements).

[L2]

For a finite p-group P the following are equivalent: P is extraspecial; P is nonabelian, Z(P)=p and P/Z(P) is elementary abelian; P is nonabelian and Z(P)=P=Φ(P) has order p (Three equivalent descriptions of an extraspecial p-group).

[L3]

For NG, the quotient G/N is abelian if and only if [G,G]N (G/N is abelian if and only if [G,G]N).

[L4]

If the quotient group G/Z(G) is cyclic, then G is abelian (If G/Z(G) is cyclic, then G is abelian).

[L5]

For a finite group G and HG, G=[G:H]H (Lagrange's theorem: G=[G:H]H for every subgroup H of a finite group G).

[L6]

An elementary abelian p-group is a finite abelian p-group in which every nonidentity element has order p (Elementary abelian p-groups).

[L7]

For an odd prime p and a finite group G with [G,G]Z(G) of exponent dividing p, (xy)p=xpyp for all x,y (For an odd prime p, the p-th power map is a homomorphism on a finite group whose derived subgroup is central of exponent dividing p).

[L9]

A finite p-group is a finite group whose order has the form pn (A finite p-group has order pn for a prime p and some nN).

Proof

technique · direct
1.1

A triple (a,b,c) commutes with (a,b,c) exactly when ab=ab. Taking (a,b,c)=(0,1,0) gives a=0, and taking (a,b,c)=(1,0,0) gives b=0; conversely every (0,0,c) commutes with everything. So Z(Hp)={(0,0,c)}, a subgroup of order p.

F1F2L1
1.2

For g=(a,b,c) and h=(a,b,c) one has gh=(a+a,b+b,c+c+ab) and hg=(a+a,b+b,c+c+ab), so gh=hg(0,0,abab) and therefore [g,h]=(gh)(hg)1=(0,0,abab).

F1F3L1
1.3

Hp is a finite p-group of order p3 and is not abelian.

L1L9
2.1

Every commutator lies in {(0,0,t)}, and taking a=b=1 with a=b=0 gives the commutator (0,0,1), which generates that subgroup; hence [Hp,Hp]=Z(Hp), of order p.

F3L8step 1.1step 1.2
3.1

By Lagrange the quotient Hp/Z(Hp) has order p2; it is abelian because [Hp,Hp]Z(Hp), and it is not cyclic, since a cyclic central quotient would make Hp abelian. An abelian group of order p2 that is not cyclic has no element of order p2, so every one of its nonidentity elements has order p and it is elementary abelian.

L3L4L5L6step 2.1step 1.3
4.1

By the second description in the characterisation, Hp is extraspecial.

L2step 1.1step 1.3step 3.1
5.1

For odd p, the derived subgroup is central of order p, so the p-th power map is a homomorphism; the three generators (1,0,0), (0,1,0) and (0,0,1) have p-th power the identity, so the p-th power map is trivial on a generating set and hence on Hp. Since Hp is nontrivial, its exponent is p.

F4L1L7L8step 2.1step 4.1
6.1

At p=2 the exponent is not 2: (1,1,0)2=(1+1,1+1,0+0+11)=(0,0,1), which is not the identity, while (0,0,1)2=(0,0,0), so (1,1,0) has order 4 and the exponent is 4.

F1F4L1L5step 1.3

Remarks

The group is extraspecial at every prime, p=2 included; only the exponent depends on the parity of p. The step that fails at p=2 is the p-th power homomorphism, whose hypothesis is that p be odd, and the failure is visible in the element (1,1,0) of order four.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-08-26Open item page →

Raising to the power 1+p is an automorphism of order p of a cyclic group of order p2

Statement

Let p be a prime. In Z/p2 the class of 1+p is a unit of multiplicative order p, and (1+p)k=1+kp for every kN. Consequently, if A is a cyclic group of order p2, the map xx1+p is an automorphism of A of order p (Group isomorphisms, automorphisms and the set Aut(G)).

Facts & Assumptions

Given: A prime p, the ring Z/p2, and a cyclic group A of order p2.

[F1]

For every nN, (Z/n,+,[0]n) is an abelian group, (Z/n,,[1]n) is a commutative monoid, and multiplication distributes over addition on both sides (For every natural n, (Z/n,+) is an abelian group, multiplication is a commutative monoid operation, and both distributive laws hold).

[F2]
[F3]

p is prime when p>1 and its only positive divisors are 1 and p (Prime and composite integers: p is prime when p>1 and its only positive divisors are 1 and p).

[L1]

Aut(Cn)(Z/n)×, and if Cn=g the unit class [a] corresponds to the automorphism gga ( Aut(Cn)(Z/nZ)×).

[L2]

A cyclic group with a generator of finite order n is isomorphic to (Z/n,+) (Every cyclic group is isomorphic to (Z,+) or to (Z/n,+) for its finite order n1).

Proof

technique · induction
1.1

At k=0 the claim reads (1+p)0=1=1+0p in Z/p2.

F1base
1.2

Assume (1+p)k=1+kp in Z/p2 for a given kN.

ih
1.3

In Z/p2 the class of p2 is zero, so kp=0 holds exactly when p2 divides kp, that is exactly when p divides k.

F1F3algebra
1.4

An automorphism of a cyclic group of order p2 is xxa for a unit class [a] of Z/p2, and this correspondence is an isomorphism of groups, so it preserves orders.

L1L2
2.1

Then (1+p)k+1=(1+kp)(1+p)=1+(k+1)p+kp2=1+(k+1)p, the last equality because p2=0 in Z/p2.

F1step 1.2algebra
3.1

Hence (1+p)k=1+kp for every kN; by step 1.3 this equals 1 exactly when p divides k, so the least positive such k is p and the class of 1+p is a unit of order p. Under the correspondence of step 1.4 it is the automorphism xx1+p, which therefore has order p.

F2step 1.3step 1.4step 2.1discharge-induction

Remarks

The computation holds at p=2 as well: there 1+p=3 and 32=9=1 in Z/4, so the automorphism xx3=x1 of a cyclic group of order four is inversion and has order two.

DefinitionDefinition: Literature-sourcedProof: Not applicablejudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-26Open item page →

The modular group of order p3 as a semidirect product Cp2Cp

Definition

Let p be a prime, let A=a be a cyclic group of order p2 and let B=s be a cyclic group of order p (Every cyclic group is isomorphic to (Z,+) or to (Z/n,+) for its finite order n1). Define

α:BAut(A),αsi(x)=x(1+p)i.

Why this is well defined and is an action by automorphisms. Each xx1+p is an automorphism of A of order p (Raising to the power 1+p is an automorphism of order p of a cyclic group of order p2), so αsi is the i-th power of that automorphism and is an automorphism. If si=sj then p divides ij, and the p-th power of that automorphism is the identity, so αsi=αsj. Finally αsiαsj=αsi+j, so α is a homomorphism into Aut(A) and hence an action by automorphisms (An action of a group H on a group N by automorphisms, Group isomorphisms, automorphisms and the set Aut(G)).

The modular group of order p3 is the external semidirect product ( The external semidirect product NαH, The semidirect-product multiplication makes N×H a group)

Mp:=AαB.

Writing a and s for the canonical images of the two generators (The canonical copy of N is normal, the canonical copy of H is a complement, and conjugation induces the action), the group is generated by a and s subject to

ap2=1,sp=1,sas1=a1+p.

Remarks

Craven writes Modn(p) for the analogous group of order pn and reserves p+1+2 for the exponent-p group; only the case n=3 is built here. Its relation uses the standard order-p power automorphism aa1+p, the first nonidentity power automorphism congruent to the identity modulo p.

At p=2 the relation reads sas1=a3=a1, so the action is inversion and M2 is the generalized dihedral group of a cyclic group of order four ( Dih(Cn)=CnC2 with inversion action has order 2n and the dihedral relations).

PropositionStatement: AI-adaptedProof: AI-adaptedprecheck passaudited 2026-08-26Open item page →

The modular group of order p3 is extraspecial, of exponent p2 when p is odd

Statement

Let p be a prime and let Mp=AαB be the modular group of order p3, with A=a of order p2, B=s of order p and sas1=a1+p. Then Mp=p3, the group is nonabelian,

Z(Mp)=[Mp,Mp]=ap

has order p, and Mp is extraspecial of exponent p2. At p=2 the group is the generalized dihedral group Dih(C4).

Facts & Assumptions

Given: A prime p and the modular group Mp with its generators a and s.

[F1]

The modular group of order p3 is Mp=AαB with A=a of order p2, B=s of order p, and sas1=a1+p (The modular group of order p3 as a semidirect product Cp2Cp).

[F2]

Z(G):={zG:zg=gz for every gG} (The center Z(G) of a group).

[F3]

For g,hG the commutator is [g,h]:=ghg1h1 (Commutators [g,h]=ghg1h1 and the commutator subgroup [G,G]).

[F4]

For a finite group G, exp(G)=min{nN:n>0 and gn=e for every gG} (The exponent of a finite group).

[L1]

In NαH the sets Nˉ={(n,1)} and Hˉ={(1,h)} are subgroups, Nˉ is normal, NˉHˉ={(1,1)}, every element has a unique factorisation (n,1)(1,h), and (1,h)(n,1)(1,h)1=(αh(n),1) (The canonical copy of N is normal, the canonical copy of H is a complement, and conjugation induces the action).

[L2]

The class of 1+p in Z/p2 is a unit of multiplicative order p, and (1+p)k=1+kp (Raising to the power 1+p is an automorphism of order p of a cyclic group of order p2).

[L3]

For a finite p-group P the following are equivalent: P is extraspecial; P is nonabelian, Z(P)=p and P/Z(P) is elementary abelian; P is nonabelian and Z(P)=P=Φ(P) has order p (Three equivalent descriptions of an extraspecial p-group).

[L4]

For a finite group G and HG, G=[G:H]H (Lagrange's theorem: G=[G:H]H for every subgroup H of a finite group G).

[L5]

If the quotient group G/Z(G) is cyclic, then G is abelian (If G/Z(G) is cyclic, then G is abelian).

[L6]

For NG, the quotient G/N is abelian if and only if [G,G]N (G/N is abelian if and only if [G,G]N).

[L7]

An elementary abelian p-group is a finite abelian p-group in which every nonidentity element has order p (Elementary abelian p-groups).

[L8]

For n1, Dih(Cn)=CnC2 where the nonidentity element of C2 acts by inversion ( Dih(Cn)=CnC2 with inversion action has order 2n and the dihedral relations).

Proof

technique · direct
1.1

Every element of Mp has a unique factorisation aisj with 0i<p2 and 0j<p, so Mp=p2p=p3 and Mp is a finite p-group generated by a and s.

F1L1L9
1.2

The relation gives [s,a]=sas1a1=a1+pa1=ap, and ap has order p because a has order p2.

F1F3L2L10
1.3

The element ap is central: it commutes with a, and saps1=(sas1)p=a(1+p)p=ap+p2=ap.

F1F2L1L2L10
2.1

Mp is nonabelian, since sas1=a1+pa: equality would give ap=1, contradicting that a has order p2.

F1L2step 1.1
2.2

Modulo ap the images of a and s commute, by step 1.2, and they generate; so Mp/ap is abelian and [Mp,Mp]ap. With ap=[s,a][Mp,Mp] this gives [Mp,Mp]=ap, of order p.

F3L6L9step 1.1step 1.2step 1.3
3.1

The centre contains ap and is not all of Mp; its order divides p3, and an order of p2 would leave a quotient of order p, necessarily cyclic, forcing Mp abelian. So Z(Mp)=ap has order p and equals [Mp,Mp].

F2L4L5step 1.3step 2.1step 2.2
3.2

The exponent is p2: the element a has order p2, so the exponent is a multiple of p2 dividing p3; it is not p3, because an element of order p3 in a group of order p3 would generate it and make it cyclic, hence abelian.

F4L4L9step 1.1step 2.1
4.1

The quotient Mp/Z(Mp) has order p2, is abelian because [Mp,Mp]=Z(Mp), and is not cyclic, so all of its nonidentity elements have order p and it is elementary abelian; by the second description in the characterisation Mp is extraspecial.

L3L4L5L6L7step 2.1step 3.1
5.1

At p=2 the relation reads sas1=a3=a1, so the action of B on A is inversion and M2 is by definition the semidirect product of a cyclic group of order four by a cyclic group of order two acting by inversion, which is Dih(C4).

F1L8L2step 1.1

Remarks

The group is extraspecial at every prime and its exponent is p2 at every prime; what changes at p=2 is only that the resulting group already has a name, since inversion is the unique nontrivial power automorphism of a cyclic group of order four.

PropositionStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-08-26Open item page →

Dih(C4) and Q8 are extraspecial of order 8, with six and two solutions of x2=1 respectively

Statement

Both Dih(C4) and Q8 are extraspecial groups of order 8: each is nonabelian, each has centre equal to its derived subgroup of order two, and each has elementary abelian central quotient. In Dih(C4) there are exactly six solutions of x2=1, and in Q8 exactly two.

Facts & Assumptions

Given: The generalized dihedral group Dih(C4)=rs with r of order four, and the quaternion group Q8={1,1,i,i,j,j,k,k}.

[F1]

Z(G):={zG:zg=gz for every gG} (The center Z(G) of a group).

[F2]

For g,hG the commutator is [g,h]:=ghg1h1, and [G,G] is the subgroup generated by all commutators (Commutators [g,h]=ghg1h1 and the commutator subgroup [G,G]).

[F3]

The generalized dihedral group of an abelian group A is Dih(A)=AC2 with the nonidentity element of C2 acting by inversion ( The generalized dihedral group Dih(A)=AC2 for an abelian group A).

[F4]

Q8:={1,1,i,i,j,j,k,k} inside the nonzero quaternions, where i2=j2=k2=1, ij=k, jk=i, ki=j, ji=k, kj=i and ik=j (The quaternion group Q8={±1,±i,±j,±k} inside the nonzero quaternions, The quaternions H: real quadruples with componentwise addition and an explicit multiplication formula matching the table on 1,i,j,k).

[L1]

For n1, Dih(Cn)=CnC2 has order 2n, and with Cn=r and C2=s one has rn=s2=1, srs1=r1, and every element has a unique form ri or ris with 0i<n ( Dih(Cn)=CnC2 with inversion action has order 2n and the dihedral relations).

[L2]

Q8 is a subgroup of the nonzero quaternions with Q8=8; the element 1 is the only element of order 1, 1 is the only element of order 2, and each of ±i,±j,±k has order 4 (Q8 is a subgroup of H× with eight elements, and 1 is its only element of order 2).

[L3]

For a finite p-group P the following are equivalent: P is extraspecial; P is nonabelian, Z(P)=p and P/Z(P) is elementary abelian; P is nonabelian and Z(P)=P=Φ(P) has order p (Three equivalent descriptions of an extraspecial p-group).

[L4]

For a finite group G and HG, G=[G:H]H (Lagrange's theorem: G=[G:H]H for every subgroup H of a finite group G).

[L5]

If the quotient group G/Z(G) is cyclic, then G is abelian (If G/Z(G) is cyclic, then G is abelian).

[L6]

For NG, the quotient G/N is abelian if and only if [G,G]N (G/N is abelian if and only if [G,G]N).

[L7]

An elementary abelian p-group is a finite abelian p-group in which every nonidentity element has order p (Elementary abelian p-groups).

Proof

technique · direct
1.1

In D=Dih(C4) the relation srs1=r1 gives sr1=rs, so r(ris)r1=ri+1sr1=ri+2s, which differs from ris because r21; and ri is central exactly when ri=ri, that is when r2i=1, that is when i is 0 or 2. So Z(D)={1,r2}, of order two, and D is nonabelian of order eight.

F1F3L1
1.2

In D one has [s,r]=srs1r1=r2=r2; modulo r2 the images of r and s commute and generate, so [D,D]r2 and therefore [D,D]=r2=Z(D).

F2L1L6L8
1.3

In D the squares are 12=1, (r2)2=1, (ris)2=ri(sris1)s2=riri=1 for each of the four values of i, while r and r3 have order four. So exactly six elements satisfy x2=1.

L1L9
1.4

In Q8 the element 1 is central, and i is not, since ij=k while ji=k and these are distinct elements of the eight-element set; the same computation excludes ±i,±j,±k, since x is central exactly when x is. So Z(Q8)={1,1}, of order two, and Q8 is nonabelian of order eight.

F1F4L2
1.5

In Q8 one has [i,j]=iji1j1=k(i)(j)=(j)(j)=j2=1; modulo {1,1} the images of i and j commute and generate, so [Q8,Q8]={1,1}=Z(Q8).

F2F4L2L6L8
1.6

In Q8 the only solutions of x2=1 are 1 and 1, because every other element has order four.

L2L9
2.1

For each of the two groups the central quotient has order four by Lagrange, is abelian because the derived subgroup equals the centre, and is not cyclic, since a cyclic central quotient would force commutativity; an abelian group of order four that is not cyclic has all nonidentity elements of order two, hence is elementary abelian.

L4L5L6L7step 1.1step 1.2step 1.4step 1.5
3.1

Each group is therefore a nonabelian group of order 23 with centre of order two and elementary abelian central quotient, so the second description in the characterisation makes both extraspecial; the solution counts are those of steps 1.3 and 1.6.

L3step 1.1step 1.3step 1.4step 1.6step 2.1

Remarks

The two solution counts are what separate the two groups: an isomorphism would carry solutions of x2=1 to solutions of x2=1, and six is not two. Nothing about the centres or the derived subgroups distinguishes them, since those agree.

Both sources write the dihedral group of order eight as D8; this library writes Dn for the dihedral group of order 2n, so the group here is Dih(C4), which is D4 in that notation.

TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-08-26Open item page →

A nonabelian group of order p3 is extraspecial

Statement

Let p be a prime and let P be a nonabelian group of order p3. Then P is extraspecial: Z(P)=[P,P]=Φ(P) has order p and P/Z(P) is elementary abelian of order p2.

Facts & Assumptions

Given: A prime p and a nonabelian group P with P=p3.

[F1]

Z(G):={zG:zg=gz for every gG} (The center Z(G) of a group).

[F2]

A finite p-group is a finite group whose order has the form P=pn (A finite p-group has order pn for a prime p and some nN).

[L1]

If P is a nontrivial finite p-group then p divides Z(P) (Every nontrivial finite p-group has nontrivial center, in fact p divides Z(P)).

[L2]

If the quotient group G/Z(G) is cyclic, then G is abelian (If G/Z(G) is cyclic, then G is abelian).

[L3]

For a finite group G and HG, G=[G:H]H (Lagrange's theorem: G=[G:H]H for every subgroup H of a finite group G).

[L4]

If p is prime and G is a group of order p2, then G is abelian (Every group of order p2, for prime p, is abelian).

[L5]

An elementary abelian p-group is a finite abelian p-group in which every nonidentity element has order p (Elementary abelian p-groups).

[L6]

For a finite p-group P the following are equivalent: P is extraspecial; P is nonabelian, Z(P)=p and P/Z(P) is elementary abelian; P is nonabelian and Z(P)=P=Φ(P) has order p (Three equivalent descriptions of an extraspecial p-group).

[L7]

If P is a finite p-group and HP then H=pk for some k (Every subgroup of a finite p-group has order a power of p).

Proof

technique · direct
1.1

P is a nontrivial finite p-group, so its centre has order a power of p divisible by p; and Z(P)P because P is nonabelian. So Z(P) is p or p2.

F1F2L1L3L7
2.1

If Z(P)=p2 then P/Z(P) has order p by Lagrange, hence is cyclic, and P would be abelian. So Z(P)=p.

L2L3step 1.1
3.1

By Lagrange P/Z(P) has order p2, so it is abelian; it is not cyclic, since that would again force P abelian; and an abelian group of order p2 that is not cyclic has every nonidentity element of order p, so it is elementary abelian.

L2L3L4L5step 2.1
4.1

So P is a nonabelian finite p-group with centre of order p and elementary abelian central quotient, which is the second description in the characterisation; hence P is extraspecial and Z(P)=[P,P]=Φ(P) has order p.

L6step 2.1step 3.1

Remarks

Order p3 is the smallest order at which a nonabelian p-group exists, and the argument shows the extraspecial condition is automatic there. At larger orders it is not: a direct product of two nonabelian groups of order p3 is nonabelian of order p6 with centre of order p2.

TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-26Open item page →

For each prime there are exactly two nonabelian groups of order p3 up to isomorphism

Statement

For every prime p there are exactly two nonabelian groups of order p3 up to isomorphism. For odd p they are the Heisenberg group Hp, of exponent p, and the modular group Mp, of exponent p2. For p=2 they are Dih(C4) and Q8.

Facts & Assumptions

Given: A prime p and a nonabelian group P with P=p3.

[F1]

The Heisenberg group of order p3 is the set Hp of triples over Z/p with (a,b,c)(a,b,c):=(a+a,b+b,c+c+ab) (The Heisenberg group of order p3 over Z/p).

[F2]

The modular group of order p3 is Mp=AαB with A=a of order p2, B=s of order p and sas1=a1+p (The modular group of order p3 as a semidirect product Cp2Cp).

[F3]

Z(G):={zG:zg=gz for every gG} (The center Z(G) of a group).

[F4]

For g,hG the commutator is [g,h]:=ghg1h1 (Commutators [g,h]=ghg1h1 and the commutator subgroup [G,G]).

[F5]

For a finite group G, exp(G)=min{nN:n>0 and gn=e for every gG} (The exponent of a finite group).

[F8]

For a group G and a prime p, Gp=gp:gG (The pth-power subgroup Gp).

[L1]

A nonabelian group of order p3 is extraspecial, with Z(P)=[P,P]=Φ(P) of order p and P/Z(P) elementary abelian of order p2 (A nonabelian group of order p3 is extraspecial).

[L2]

For a finite p-group P the following are equivalent: P is extraspecial; P is nonabelian, Z(P)=p and P/Z(P) is elementary abelian; P is nonabelian and Z(P)=P=Φ(P) has order p (Three equivalent descriptions of an extraspecial p-group).

[L3]

For every finite p-group P, Φ(P)=PPp (Φ(P)=PPp for a finite p-group).

[L4]

If [x,y]e in an extraspecial p-group P, then x,y contains Z(P), has order p3, is nonabelian, and is extraspecial with centre Z(P) (Two elements of an extraspecial p-group with nontrivial commutator generate an extraspecial subgroup of order p3).

[L5]

If [G,G]Z(G) then [xy,w]=[x,w][y,w], [x,yw]=[x,y][x,w], and [xn,y]=[x,y]n=[x,yn] for every integer n (Commutator identities in a group whose derived subgroup is central).

[L6]

For an odd prime p and a finite group G with [G,G]Z(G) of exponent dividing p, (xy)p=xpyp for all x,y (For an odd prime p, the p-th power map is a homomorphism on a finite group whose derived subgroup is central of exponent dividing p).

[L7]

Z(Hp)=[Hp,Hp] is the third coordinate axis, of order p; Hp is extraspecial; and for odd p its exponent is p (The Heisenberg group of order p3 is extraspecial, and for odd p it has exponent p).

[L8]

The Heisenberg multiplication is a group law with identity (0,0,0) and inverse (a,b,c)1=(a,b,c+ab), the group is nonabelian of order p3, and (1,0,0)a=(a,0,0), (0,1,0)b=(0,b,0), (0,0,1)c=(0,0,c) (The Heisenberg multiplication is a group law, nonabelian, on a set of p3 elements).

[L9]

Mp is nonabelian of order p3, extraspecial of exponent p2, with Z(Mp)=[Mp,Mp]=ap; at p=2 it is Dih(C4) (The modular group of order p3 is extraspecial, of exponent p2 when p is odd).

[L10]

Dih(C4) and Q8 are extraspecial of order 8, with exactly six and exactly two solutions of x2=1 respectively (Dih(C4) and Q8 are extraspecial of order 8, with six and two solutions of x2=1 respectively).

[L11]

The conditions NG, G=NH, NH={1} hold if and only if conjugation restricts to an action α:HAut(N) and (n,h)nh is an isomorphism NαHG carrying the canonical factors onto N and H ( Recognition theorem: G=NH with NG, NH=1 exactly realises an external semidirect product).

[L12]

For n1, Dih(Cn)=CnC2 with rn=s2=1 and srs1=r1, of order 2n ( Dih(Cn)=CnC2 with inversion action has order 2n and the dihedral relations).

[L14]

Q8=8, the element 1 is its only element of order two, and each of ±i,±j,±k has order four (Q8 is a subgroup of H× with eight elements, and 1 is its only element of order 2).

[L15]

For a finite group G and HG, G=[G:H]H (Lagrange's theorem: G=[G:H]H for every subgroup H of a finite group G).

[L17]

A cyclic group with a generator of finite order n is isomorphic to (Z/n,+) (Every cyclic group is isomorphic to (Z,+) or to (Z/n,+) for its finite order n1).

Proof

technique · cases
1.1

P is extraspecial: Z(P)=[P,P]=Φ(P)=z has order p, and V=P/Z(P) is elementary abelian of order p2.

F3F4L1L2
1.2

Since Φ(P)=[P,P]Pp contains Pp, every gP has gpZ(P).

F8L1L3
1.3

Every element order divides p3; an element of order p3 would generate P and make it abelian, so every element has order dividing p2 and exp(P) is p or p2.

F5F6F7L15L16
1.4

For odd p the groups Hp and Mp are nonabelian of order p3 with exponents p and p2, so they are not isomorphic.

F5L7L8L9
1.5

The groups Dih(C4) and Q8 are nonabelian of order eight and have six and two solutions of x2=1, so they are not isomorphic.

L10
1.6

First case: suppose gp=e for every gP. At p=2 this makes P abelian, since gh=(gh)1=h1g1=hg; so p is odd here.

F7L16assume-case expp
1.7

Second case: suppose instead that P has an element a of order p2. Then a has order p2 and index p, and ap is a nonidentity element of Z(P), so Z(P)=ap.

F6F7L15L16assume-case exppsq
2.1

In the first case, nonabelianness gives x,y with [x,y]e; then z:=[x,y] generates Z(P) and x,y has order p3, so P=x,y.

F4F6L4step 1.1step 1.6
2.2

In the second case, choose ba; then a,b properly contains a subgroup of index p and so equals P, and [b,a]e since otherwise P would be abelian.

F4F6L15step 1.7
3.1

In the first case every element of P is uniquely xuyvzw with u,v,w in {0,,p1}: the images xˉ,yˉ generate V, so the products xuyv meet every coset of Z(P)={zw}, and there are exactly p3 such triples of exponents.

F3F6L1L15step 2.1
3.2

In the second case [b,a] lies in ap and is not the identity, so [b,a]=apk with pk; choosing k with kk1(modp) and replacing b by bk, which still lies outside a because its image in V is nontrivial, gives [b,a]=ap, that is bab1=a1+p. Moreover bpZ(P)=ap, say bp=apm.

F4L5L16step 1.2step 1.7step 2.2
4.1

In the first case the class-two identities give yvxu=[yv,xu]xuyv=zuvxuyv, so (xuyvzw)(xuyvzw)=xu+uyv+vzw+wuv, all exponents read modulo p because xp=yp=zp=e.

F4L5L16step 1.6step 3.1
4.2

In the second case with p odd, the p-th power map is a homomorphism, so c=bam has cp=bp(am)p=e; also ca, and cac1=bab1=a1+p.

L6L16step 3.2assume-case odd
4.3

In the second case with p=2, the element b2 lies in {e,a2}, and the relation reads bab1=a3=a1.

L16step 3.2assume-case two
5.1

In Hp put X=(1,0,0), Y=(0,1,0) and W=(0,0,1). Then XY=(1,1,1), YX=(1,1,0) and (YX)1=(1,1,1), so [X,Y]=(XY)(YX)1=(0,0,1)=W, which generates Z(Hp); for odd p every element of Hp has p-th power the identity, so steps 3.1 and 4.1 hold verbatim in Hp with X,Y,W in place of x,y,z.

F1F4L4L7L8step 3.1step 4.1
5.2

In the second case with p odd, a is normal because c conjugates it into itself and a normalises it, ac is trivial because ca and c has prime order, and ac=p3, so P=ac is an internal semidirect product whose conjugation action sends a to a1+p; hence PMp.

F2F6L11L15L17step 4.2
5.3

In the second case with p=2 and b2=e: a is normal of index two, ab is trivial, and b acts on a by inversion, so P is the internal semidirect product of a cyclic group of order four by a cyclic group of order two acting by inversion, that is PDih(C4).

F6L11L12L15L17step 4.3
5.4

In the second case with p=2 and b2=a2: the eight elements aubv with 0u<4 and v{0,1} are distinct and exhaust P, and the relations a4=e, b2=a2 and ba=a1b determine every product of two of them. The quaternion group satisfies the same three relations with i for a and j for b, since i4=1, j2=1=i2 and ji=k=i1j, and ji, so its eight elements have the same normal form; matching normal forms is therefore an isomorphism and PQ8.

F6L13L14L15L16step 4.3
6.1

In the first case, matching normal forms gives a bijection HpP carrying XuYvWw to xuyvzw, and both products are computed by the same rule, so it is an isomorphism and PHp.

step 3.1step 4.1step 5.1
7.1

The two cases are exhaustive, and within the second the two parities are exhaustive; so for odd p every nonabelian group of order p3 is isomorphic to Hp or to Mp, and for p=2 to Dih(C4) or to Q8. With the two non-isomorphy statements this gives exactly two isomorphism classes at every prime.

step 1.3step 1.4step 1.5step 5.2step 5.3step 5.4step 6.1cases-exhaustive

Remarks

The parity of p enters twice and in opposite directions. It rules out the exponent-p case at p=2, where it forces commutativity; and it is what allows the correction of the second generator in the exponent-p2 case, since that correction is made with the p-th power homomorphism, which is available only for odd p. At p=2 the correction is not available and the two possible values of b2 produce the two groups of order eight.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-08-26Open item page →

Q8Q8Dih(C4)Dih(C4)

Statement

Let E=Q8Q8 be the central product of two copies of the quaternion group along the unique isomorphism between their centres. Then E is also an internal central product of two subgroups isomorphic to Dih(C4) meeting in Z(E), and therefore

Q8Q8    Dih(C4)Dih(C4).

Facts & Assumptions

Given: Two copies E1,E2 of Q8, the central product E=E1αE2 along the unique isomorphism α of their centres, the canonical images x1,y1 of the generators i,j of E1 and x2,y2 of those of E2, and the common central image z, so that xi2=yi2=z, z2=1 and yixiyi1=xi1.

[F1]

For groups G,H with central subgroups Z1Z(G), Z2Z(H) and an isomorphism α:Z1Z2, the central product GαH is the quotient of G×H by N={(z,α(z)1):zZ1} (The central product GαH of two groups along an isomorphism of central subgroups).

[F2]

Subgroups G1,,Gr of G form an internal central product when they generate G and [Gi,Gj]=1 for ij (Internal central products of a finite family of subgroups).

[L2]

Q8=8, the element 1 is its only element of order two, and each of ±i,±j,±k has order four (Q8 is a subgroup of H× with eight elements, and 1 is its only element of order 2).

[L4]

The canonical maps into a central product are injective homomorphisms whose images commute elementwise, generate the product, and meet in the image of the identified subgroup (The two canonical maps into a central product are injective homomorphisms whose images commute, generate it, and meet in the identified centre).

[L5]

A central product of two extraspecial p-groups identified along their centres is extraspecial of order E1E2/p (A central product of extraspecial p-groups identified along their centres is extraspecial).

[L6]

Subgroups form an internal central product of G if and only if the multiplication map from their direct product is a surjective homomorphism; for two factors GG1idG2 along the identity of G1G2 (Internal central products are the images of external ones).

[L7]

For n1, Dih(Cn)=CnC2 with rn=s2=1 and srs1=r1, of order 2n ( Dih(Cn)=CnC2 with inversion action has order 2n and the dihedral relations).

[L8]

The conditions NG, G=NH, NH={1} hold if and only if conjugation restricts to an action and (n,h)nh is an isomorphism NαHG ( Recognition theorem: G=NH with NG, NH=1 exactly realises an external semidirect product).

[L9]

For a finite group G and HG, G=[G:H]H (Lagrange's theorem: G=[G:H]H for every subgroup H of a finite group G).

Proof

technique · direct
1.1

The two canonical images are isomorphic copies of Q8 that commute elementwise, generate E, and meet exactly in z; and E is extraspecial of order 88/2=32.

F1L2L3L4L5
1.2

Each xi has order four, since xi2=z and z2=1 with z1; likewise each yi.

F4L1L2
2.1

Put t1=x2y1 and t2=x1y2. Then t12=x2y1x2y1=x22y12=zz=1, because x2 commutes with y1; likewise t22=1.

L1L4step 1.1step 1.2
2.2

Also t1x1t11=x2(y1x1y11)x21=y1x1y11=x11, because x2 commutes with everything in the first image; likewise t2x2t21=x21.

L1L4step 1.1
2.3

Neither t1 lies in x1 nor t2 in x2: if x2y1 were a power of x1 then x2 would lie in the first canonical image, hence in z, contradicting that x2 has order four.

F3L4step 1.1step 1.2
3.1

Set H1=x1,t1 and H2=x2,t2. In H1 the cyclic subgroup x1 of order four is normalised by t1 and meets t1 trivially, and x1t1=8; so H1 is the internal semidirect product of a cyclic group of order four by a group of order two acting by inversion, that is H1Dih(C4) of order eight with Z(H1)=x12=z. The same holds for H2.

F3L7L8L9step 2.1step 2.2step 2.3
4.1

The four generators commute in pairs across the two subgroups: x1 commutes with x2 and with t2=x1y2; t1=x2y1 commutes with x2; and t1t2=x2(y1x1)y2=x11x2y1y2 while t2t1=x1(y2x2)y1=x21x1y1y2, and these agree because x12=z=x22 gives x11x2=x21x1. Hence [H1,H2]=1.

F2F3L1L4step 1.1step 3.1
4.2

The two subgroups generate E: they contain x1,x2,t1,t2, hence y1=x21t1 and y2=x11t2, hence both canonical images, which generate E.

F3L4step 1.1step 3.1
5.1

So H1 and H2 form an internal central product of E, and the recognition theorem gives EH1idH2 along the identity of D=H1H2. Comparing orders, 32=E=88/D, so D=2 and D=z=Z(H1)=Z(H2).

F2L6L9step 1.1step 3.1step 4.1step 4.2
6.1

Since H1 and H2 are isomorphic to Dih(C4) by isomorphisms carrying D to the centre, E is a central product of two copies of Dih(C4) along the unique isomorphism of their centres, which is what was claimed.

F1L3step 3.1step 5.1

Remarks

The identity x12=x22 is the whole of the computation in step 4.1, and it is where the quaternion hypothesis is spent: in a central product of two dihedral groups the corresponding squares are both trivial and the same computation succeeds for a different reason. What the statement records is that these two central products are the same group, so the number of quaternion factors in a decomposition is not an invariant of it.

LemmaStatement: AI-adaptedProof: AI-adaptedprecheck passjudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-26Open item page →

A product formula for the number of square roots of the identity in a central product of extraspecial 2-groups

Statement

Let P1 and P2 be extraspecial 2-groups and let P=P1αP2 be the central product along an isomorphism α of their centres. Write t(G)={gG:g2=1}. Then

t(P)=t(P1)t(P2)+(P1t(P1))(P2t(P2))2.

Facts & Assumptions

Given: Extraspecial 2-groups P1,P2 with Z(Pi)=zi of order two, an isomorphism α:Z(P1)Z(P2), and P=P1αP2 with quotient map π:P1×P2P.

[F1]

For groups G,H with central subgroups Z1Z(G), Z2Z(H) and an isomorphism α:Z1Z2, the central product GαH is the quotient of G×H by N={(z,α(z)1):zZ1} (The central product GαH of two groups along an isomorphism of central subgroups).

[F2]

The external direct product G×H carries the componentwise operation (The external direct product G×H with componentwise multiplication).

[F3]

The quotient group G/N has the left cosets gN as elements, with product (gN)(hN):=ghN (The quotient group G/N and coset product (gN)(hN)=ghN).

[F4]

Z(G):={zG:zg=gz for every gG} (The center Z(G) of a group).

[L1]

For a finite p-group P the following are equivalent: P is extraspecial; P is nonabelian, Z(P)=p and P/Z(P) is elementary abelian; P is nonabelian and Z(P)=P=Φ(P) has order p (Three equivalent descriptions of an extraspecial p-group).

[L2]

An elementary abelian p-group is a finite abelian p-group in which every nonidentity element has order p (Elementary abelian p-groups).

[L3]

The subgroup N={(z,α(z)1):zZ1} of G×H is central, hence normal (The identified subgroup used to form a central product is central, hence normal).

[L4]

The canonical maps into a central product are injective homomorphisms whose images commute elementwise, generate the product, and meet in the image of the identified subgroup (The two canonical maps into a central product are injective homomorphisms whose images commute, generate it, and meet in the identified centre).

[L5]

For a finite group G and HG, G=[G:H]H (Lagrange's theorem: G=[G:H]H for every subgroup H of a finite group G).

[L6]

A is the unique natural number n with An (The cardinality A of a finite set).

Proof

technique · direct
1.1

Since Pi/Z(Pi) is elementary abelian, g2Z(Pi)={1,zi} for every gPi; so Pi splits into the t(Pi) elements with g2=1 and the Pit(Pi) elements with g2=zi.

F4L1L2
1.2

The quotient map π is surjective with kernel N={(z,α(z)1):zZ(P1)}, which has two elements, so every element of P has exactly two preimages in P1×P2.

F1F3L3L5L6
1.3

Because the two canonical images commute, π(g,h)2=π(g2,h2) for all gP1 and hP2.

F2F3L4
2.1

Hence π(g,h)2=1 exactly when (g2,h2)N, that is exactly when g2=1 and h2=1, or g2=z1 and h2=α(z1)1=z2.

F1F4step 1.1step 1.3
3.1

The number of pairs (g,h) with π(g,h)2=1 is therefore t(P1)t(P2)+(P1t(P1))(P2t(P2)).

L6step 1.1step 2.1
4.1

Each element of P with square the identity is counted exactly twice in that total, so t(P) is half of it, which is the displayed formula.

L6step 1.2step 3.1

Remarks

The second summand is what makes the formula more than a product: an element of P can square to the identity because both of its coordinates square to the identity, or because both square to the identified central element and those two squares cancel in the quotient.

Writing Pi=21+2ni and t(Pi)=22ni+εi2ni with εi=±1, the formula collapses to t(P)=22n+ε1ε22n with n=n1+n2: the signs multiply.

TheoremStatement: AI-adaptedProof: AI-adaptedprecheck passaudited 2026-08-26Open item page →

For each n1 there are exactly two extraspecial groups of order 21+2n

Statement

For each n1 there are exactly two extraspecial groups of order 21+2n up to isomorphism. Writing t(G) for the number of solutions of x2=1 in G, one of them has t=22n+2n and the other has t=22n2n, and an extraspecial group of that order is determined up to isomorphism by which of the two values it takes.

Facts & Assumptions

Given: An integer n1 and an extraspecial group P of order 21+2n with Z(P)=z.

[F1]

Subgroups G1,,Gr of G form an internal central product when they generate G and [Gi,Gj]=1 for ij (Internal central products of a finite family of subgroups).

[F2]

For groups G,H with central subgroups Z1Z(G), Z2Z(H) and an isomorphism α:Z1Z2, the central product GαH is the quotient of G×H by N={(z,α(z)1):zZ1} (The central product GαH of two groups along an isomorphism of central subgroups).

[F3]

Z(G):={zG:zg=gz for every gG} (The center Z(G) of a group).

[L1]

There are n1 subgroups P1,,Pn of P, each nonabelian of order p3 with Z(Pi)=Z(P), which form an internal central product of P; such a family is admissible, P=p1+2n, and peeling one member leaves an extraspecial group of order p1+2(n1) with the induced admissible family (Every extraspecial p-group is an internal central product of nonabelian subgroups of order p3).

[L2]

For every prime p there are exactly two nonabelian groups of order p3 up to isomorphism; at p=2 they are Dih(C4) and Q8 (For each prime there are exactly two nonabelian groups of order p3 up to isomorphism).

[L3]

Dih(C4) and Q8 are extraspecial of order eight, with exactly six and exactly two solutions of x2=1 (Dih(C4) and Q8 are extraspecial of order 8, with six and two solutions of x2=1 respectively).

[L4]

Q8Q8 is an internal central product of two subgroups isomorphic to Dih(C4) meeting in its centre (Q8Q8Dih(C4)Dih(C4)).

[L5]

For extraspecial 2-groups, t(P1αP2)=(t(P1)t(P2)+(P1t(P1))(P2t(P2)))/2 (A product formula for the number of square roots of the identity in a central product of extraspecial 2-groups).

[L6]

Subgroups form an internal central product of G if and only if the multiplication map from their direct product is a surjective homomorphism; for two factors GG1idG2 along the identity of G1G2 (Internal central products are the images of external ones).

[L7]

A central product of two extraspecial p-groups identified along their centres is extraspecial of order E1E2/p (A central product of extraspecial p-groups identified along their centres is extraspecial).

[L8]

A finite p-group P is extraspecial when it is nonabelian and Z(P)=P=Φ(P) is elementary abelian of order p (Special and extraspecial p-groups).

[L9]

For a finite group G and HG, G=[G:H]H (Lagrange's theorem: G=[G:H]H for every subgroup H of a finite group G).

Proof

technique · induction
1.1

At n=1 an extraspecial group of order eight is nonabelian, hence isomorphic to Dih(C4) or to Q8; these have t=6=22+21 and t=2=2221, so there are exactly two and the value of t tells them apart.

L2L3L8base
1.2

Assume, for every m with 1m<n: both values 22m+2m and 22m2m are realised; every extraspecial group of order 21+2m has an admissible family with at most one quaternion member; if that number is k then t=22m+(1)k2m; and two such groups with equal t are isomorphic.

ih
1.3

If ϕ:GG and ψ:HH are isomorphisms carrying the identified central subgroups to the identified central subgroups compatibly with the identifying isomorphisms, then ϕ×ψ carries N onto N and induces an isomorphism GαHGαH; when all four identified subgroups have order two the compatibility is automatic, since a group of order two has only one automorphism.

F2algebra
1.4

Let P be extraspecial of order 21+2n with n2 and take an admissible family P1,,Pn; each member is nonabelian of order eight, hence isomorphic to Dih(C4) or to Q8.

F1L1L2
2.1

If two members Pi,Pj are isomorphic to Q8, then R=Pi,Pj is an internal central product of them, so RQ8Q8 and R is an internal central product of two subgroups isomorphic to Dih(C4) with the same centre Z(P); replacing Pi,Pj by those two subgroups leaves an admissible family with two fewer quaternion members. Repeating, P has an admissible family with k{0,1} quaternion members.

F1F3F4L4L6step 1.3step 1.4
3.1

Fix such a family. Since n2 and k1, some member P1 is isomorphic to Dih(C4); peeling it leaves C=P2,,Pn, extraspecial of order 21+2(n1) with an admissible family of n1 members of which k are quaternion, and PP1idC along the identity of Z(P).

F1L1L6L9step 2.1
4.1

By the induction hypothesis t(C)=22(n1)+(1)k2n1, and t(P1)=6=22+2; the counting formula then gives t(P)=22n+(1)k2n.

L5L9step 1.2step 3.1
5.1

If P and P are extraspecial of order 21+2n with t(P)=t(P), their normalised families have the same k by step 4.1, so peeling a dihedral member from each gives C and C extraspecial of order 21+2(n1) with equal t, hence isomorphic by the induction hypothesis, by an isomorphism carrying Z(C)=Z(P) onto Z(C)=Z(P); the peeled members are isomorphic too, so PP.

F3step 1.2step 1.3step 3.1step 4.1
6.1

Both values are realised: if C is extraspecial of order 21+2(n1) then Dih(C4)C is extraspecial of order 21+2n with t=22n+(1)k2n where t(C)=22(n1)+(1)k2n1, so the two groups supplied by the induction hypothesis produce one group of each value. With step 5.1 this gives exactly two isomorphism classes at order 21+2n and completes the induction.

L5L7step 1.1step 1.2step 4.1step 5.1discharge-induction

Remarks

The quaternion factors are not an invariant of the group, only their parity is: two of them can always be traded for two dihedral factors, and it is exactly that trade which leaves the count t unchanged, since the two signs multiply.

The count t is an isomorphism invariant because an isomorphism carries solutions of x2=1 to solutions of x2=1; that is what makes the two classes provably distinct rather than merely differently presented.

LemmaStatement: AI-adaptedProof: AI-adaptedprecheck passjudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-26Open item page →

For odd p, a central product of two modular groups of order p3 is a central product of a modular group with a Heisenberg group

Statement

Let p be an odd prime and let E be a central product of two copies of the modular group Mp along an isomorphism of their centres. Then E is an internal central product of a subgroup isomorphic to the Heisenberg group Hp and a subgroup isomorphic to Mp; consequently

MpMp    HpMp.

Facts & Assumptions

Given: An odd prime p, two copies of Mp with generators ai of order p2 and si of order p satisfying siaisi1=ai1+p, and the central product E of the two along an isomorphism of their centres, with canonical images xi,yi of ai,si and common central image z.

[F1]

The modular group of order p3 is Mp=AαB with A=a of order p2, B=s of order p, and sas1=a1+p (The modular group of order p3 as a semidirect product Cp2Cp).

[F2]

For groups G,H with central subgroups Z1Z(G), Z2Z(H) and an isomorphism α:Z1Z2, the central product GαH is the quotient of G×H by N={(z,α(z)1):zZ1} (The central product GαH of two groups along an isomorphism of central subgroups).

[F3]

For g,hG the commutator is [g,h]:=ghg1h1 (Commutators [g,h]=ghg1h1 and the commutator subgroup [G,G]).

[F4]

Z(G):={zG:zg=gz for every gG} (The center Z(G) of a group).

[F5]

For a finite group G, exp(G)=min{nN:n>0 and gn=e for every gG} (The exponent of a finite group).

[L1]

Mp is nonabelian of order p3, extraspecial of exponent p2, with Z(Mp)=[Mp,Mp]=ap (The modular group of order p3 is extraspecial, of exponent p2 when p is odd).

[L2]

A central product of two extraspecial p-groups identified along their centres is extraspecial of order E1E2/p (A central product of extraspecial p-groups identified along their centres is extraspecial).

[L3]

The canonical maps into a central product are injective homomorphisms whose images commute elementwise, generate the product, and meet in the image of the identified subgroup (The two canonical maps into a central product are injective homomorphisms whose images commute, generate it, and meet in the identified centre).

[L4]

For an odd prime p and a finite group G with [G,G]Z(G) of exponent dividing p, (xy)p=xpyp for all x,y (For an odd prime p, the p-th power map is a homomorphism on a finite group whose derived subgroup is central of exponent dividing p).

[L5]

An extraspecial p-group is nilpotent of class exactly two, its derived subgroup satisfies P=Z(P) and has order p, and every nonidentity commutator has order p (An extraspecial p-group is nilpotent of class exactly two and its derived subgroup has order p).

[L6]

If [G,G]Z(G) then [xy,w]=[x,w][y,w], [x,yw]=[x,y][x,w], and [xn,y]=[x,y]n=[x,yn] for every integer n (Commutator identities in a group whose derived subgroup is central).

[L7]

If [x,y]e in an extraspecial p-group P, then x,y contains Z(P), has order p3, is nonabelian, and is extraspecial with centre Z(P) (Two elements of an extraspecial p-group with nontrivial commutator generate an extraspecial subgroup of order p3).

[L8]

If x,yP satisfy [x,y]e and F=x,y, then P=FCP(F) and FCP(F)=Z(P), and CP(F) is extraspecial of order P/p2 with centre Z(P) when P>p3 (Every extraspecial p-group is an internal central product of nonabelian subgroups of order p3, The centralizer CG(H) of a subgroup).

[L9]

For every prime p there are exactly two nonabelian groups of order p3 up to isomorphism; for odd p they are Hp, of exponent p, and Mp, of exponent p2 (For each prime there are exactly two nonabelian groups of order p3 up to isomorphism).

[L10]

Subgroups form an internal central product of G if and only if the multiplication map from their direct product is a surjective homomorphism; for two factors GG1idG2 along the identity of G1G2 (Internal central products are the images of external ones).

Proof

technique · direct
1.1

E is extraspecial of order p5, its centre is the common image z of the two centres, and the two canonical images commute elementwise, are injective and generate E.

F2F4L1L2L3
1.2

The generator a2 may be replaced by a power a2c with c not divisible by p without changing the relations of the second copy, and such a replacement multiplies x2p by c in the exponent; choosing c suitably we may assume x1p=x2p=z.

F1F2L1L11
2.1

Each xi has order p2 and each yi has order p, and [x2,y2] is the image of [a2,s2]=a2p, so [x2,y2]=z1.

F1F3L1L3step 1.2
2.2

Put w=x2x11. The p-th power map on E is a homomorphism, because p is odd and [E,E]=Z(E) has order p, so wp=x2p(x1p)1=zz1=e.

L4L5L11step 1.1step 1.2
3.1

Moreover we: otherwise x1=x2 would lie in both canonical images, hence in z, contradicting that x1 has order p2. So w has order p.

L3step 1.1step 2.1step 2.2
3.2

Since x1 and y2 lie in different canonical images they commute, so [w,y2]=[x2x11,y2]=[x2,y2][x11,y2]=[x2,y2]=z1e.

F3L3L6step 1.1step 2.1
4.1

Hence F=w,y2 is extraspecial of order p3 with Z(F)=Z(E).

F6L7step 3.2
5.1

The elements of E whose p-th power is the identity form the kernel of the p-th power homomorphism, hence a subgroup; it contains w, y2 and z, so it contains F, and F has exponent p.

F5F6L4step 2.2step 3.1step 4.1
5.2

By the splitting clause, E=FCE(F) with FCE(F)=Z(E), and C=CE(F) is extraspecial of order p5/p2=p3 with Z(C)=Z(E); so F and C form an internal central product of E.

L8step 1.1step 3.2step 4.1
6.1

A nonabelian group of order p3 and exponent p is isomorphic to Hp, since the other one has exponent p2; so FHp.

L9step 4.1step 5.1
6.2

If C had exponent p then every element of E=FC would be a product of two commuting elements of p-th power the identity, so E would have exponent p, contradicting that x1 has order p2. Hence C has exponent p2 and CMp.

F5L4L9step 2.1step 5.1step 5.2
7.1

Therefore E is an internal central product of FHp and CMp meeting in Z(E), and the recognition theorem identifies it with HpMp along the identity of that centre.

L10step 6.1step 5.2step 6.2

Remarks

The construction of the exponent-p subgroup is where oddness of p is spent: the element w=x2x11 has order p only because the p-th power map is a homomorphism, and at p=2 that map is not one. The corresponding statement at p=2 is the trade of two quaternion factors for two dihedral ones, which is a different computation with a different outcome.

TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-08-26Open item page →

For odd p and each n1 there are exactly two extraspecial groups of order p1+2n, distinguished by their exponent

Statement

Let p be an odd prime. For each n1 there are exactly two extraspecial groups of order p1+2n up to isomorphism, and they are distinguished by their exponent: one has exponent p and the other has exponent p2.

Facts & Assumptions

Given: An odd prime p, an integer n1, and an extraspecial group P of order p1+2n with Z(P)=z.

[F1]

Subgroups G1,,Gr of G form an internal central product when they generate G and [Gi,Gj]=1 for ij (Internal central products of a finite family of subgroups).

[F2]

For groups G,H with central subgroups Z1Z(G), Z2Z(H) and an isomorphism α:Z1Z2, the central product GαH is the quotient of G×H by N={(z,α(z)1):zZ1} (The central product GαH of two groups along an isomorphism of central subgroups).

[F3]

For a finite group G, exp(G)=min{nN:n>0 and gn=e for every gG} (The exponent of a finite group).

[F4]

Z(G):={zG:zg=gz for every gG} (The center Z(G) of a group).

[F6]

For a group G and a prime p, Gp=gp:gG (The pth-power subgroup Gp).

[F7]

The Heisenberg group is Hp=(Z/p)3 with (a,b,c)(a,b,c)=(a+a,b+b,c+c+ab) (The Heisenberg group of order p3 over Z/p).

[L1]

There are n1 subgroups P1,,Pn of P, each nonabelian of order p3 with Z(Pi)=Z(P), which form an internal central product of P; such a family is admissible, P=p1+2n, and peeling one member leaves an extraspecial group of order p1+2(n1) with the induced admissible family (Every extraspecial p-group is an internal central product of nonabelian subgroups of order p3).

[L2]

For every prime p there are exactly two nonabelian groups of order p3 up to isomorphism; for odd p they are Hp, of exponent p, and Mp, of exponent p2 (For each prime there are exactly two nonabelian groups of order p3 up to isomorphism).

[L3]

For odd p, a central product of two copies of Mp along an isomorphism of their centres is an internal central product of a subgroup isomorphic to Hp and a subgroup isomorphic to Mp (For odd p, a central product of two modular groups of order p3 is a central product of a modular group with a Heisenberg group).

[L4]

For an odd prime p and a finite group G with [G,G]Z(G) of exponent dividing p, (xy)p=xpyp for all x,y (For an odd prime p, the p-th power map is a homomorphism on a finite group whose derived subgroup is central of exponent dividing p).

[L5]

An extraspecial p-group is nilpotent of class exactly two, its derived subgroup satisfies P=Z(P) and has order p, and every nonidentity commutator has order p (An extraspecial p-group is nilpotent of class exactly two and its derived subgroup has order p).

[L6]

For a finite p-group P the following are equivalent: P is extraspecial; P is nonabelian, Z(P)=p and P/Z(P) is elementary abelian; P is nonabelian and Z(P)=P=Φ(P) has order p (Three equivalent descriptions of an extraspecial p-group).

[L7]

For every finite p-group P, Φ(P)=PPp (Φ(P)=PPp for a finite p-group).

[L8]

Subgroups form an internal central product of G if and only if the multiplication map from their direct product is a surjective homomorphism; for two factors GG1idG2 along the identity of G1G2 (Internal central products are the images of external ones).

[L9]

A central product of two extraspecial p-groups identified along their centres is extraspecial of order E1E2/p (A central product of extraspecial p-groups identified along their centres is extraspecial).

[L10]

Hp is extraspecial and, for odd p, has exponent p (The Heisenberg group of order p3 is extraspecial, and for odd p it has exponent p).

[L12]

For a finite group G and HG, G=[G:H]H (Lagrange's theorem: G=[G:H]H for every subgroup H of a finite group G).

[L13]

The canonical maps from both factors into a central product are injective homomorphisms; their images commute and generate the central product (The two canonical maps into a central product are injective homomorphisms whose images commute, generate it, and meet in the identified centre).

Proof

technique · induction
1.1

At n=1 an extraspecial group of order p3 is nonabelian, hence isomorphic to Hp or to Mp; their exponents are p and p2, so there are exactly two and the exponent tells them apart.

L2L6L10L11base
1.2

Assume, for every m with 1m<n: both exponents p and p2 are realised by extraspecial groups of order p1+2m; every such group has an admissible family with at most one modular member; its exponent is p when that number is zero and p2 when it is one; and two such groups with the same exponent are isomorphic.

ih
1.3

If ϕ:GG and ψ:HH are isomorphisms carrying the identified central subgroups onto the identified central subgroups compatibly with the identifying isomorphisms, then ϕ×ψ carries N onto N and induces an isomorphism GαHGαH.

F2algebra
1.4

Every gP has gpPpΦ(P)=Z(P), so gp2=(gp)p=e and the exponent of P divides p2.

F3F6L6L7
1.5

Let P be extraspecial of order p1+2n with n2 and take an admissible family P1,,Pn; each member is nonabelian of order p3, hence isomorphic to Hp or to Mp.

F1L1L2
2.1

If two members Pi,Pj are isomorphic to Mp, then R=Pi,Pj is an internal central product of them, so RMpMp and R is an internal central product of a subgroup isomorphic to Hp and one isomorphic to Mp, both with centre Z(P); replacing Pi,Pj by those two leaves an admissible family with one fewer modular member. Repeating, P has an admissible family with k{0,1} modular members.

F1F4F5L3L8step 1.3step 1.5
3.1

If k=0 every member has exponent p; since the members commute and generate P and the p-th power map is a homomorphism, every element of P is a product of elements of the members and has p-th power the identity, so exp(P)=p. If k=1 the modular member contains an element of order p2, so exp(P) is a multiple of p2, and by step 1.4 it equals p2. Thus the exponent determines k.

F3F5L4L5L10L11step 1.4step 2.1
3.2

Since n2 and k1, some member P1 is isomorphic to Hp; peeling it leaves C=P2,,Pn, extraspecial of order p1+2(n1) with an admissible family of n1 members of which k are modular, and PP1idC along the identity of Z(P).

F1L1L8L12step 2.1
4.1

If P and P are extraspecial of order p1+2n with the same exponent, their normalised families have the same k by step 3.1, so the peeled subgroups C and C have the same exponent and are isomorphic by the induction hypothesis. Any such isomorphism restricts to an isomorphism Z(C)Z(C). For the peeled Heisenberg factors, the maps (a,b,c)(ra,b,rc) with rFp× preserve the multiplication of [F7] and induce every automorphism of their order-p centres; choose one whose central restriction makes the two factor isomorphisms compatible. Step 1.3 then induces PP.

F4F7L10step 1.2step 1.3step 3.1step 3.2
5.1

Both exponents are realised: if C is extraspecial of order p1+2(n1) then HpC is extraspecial of order p1+2n. When C has exponent p, the commuting generating images of [L13] and [L4] show that the product has exponent p; when C has exponent p2, its injective canonical image from [L13] still contains an element of order p2, while step 1.4 bounds the product exponent by p2. Applying this to the two groups supplied by the induction hypothesis gives one group of each exponent. With step 4.1 this gives exactly two isomorphism classes at order p1+2n and completes the induction.

L4L9L10L13step 1.1step 1.2step 1.4step 3.1step 4.1discharge-induction

Remarks

The modular factors are not an invariant of the group and, unlike the quaternion factors at p=2, not even their parity is: two of them can be traded for one Heisenberg factor and one modular factor, so the count drops by one rather than by two. What survives is the presence or absence of an element of order p2, which is the exponent.

DefinitionDefinition: Literature-sourcedProof: Not applicablejudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-26Open item page →

Plus and minus type of an extraspecial p-group

Definition

Let p be a prime and n1. By For odd p and each n1 there are exactly two extraspecial groups of order p1+2n, distinguished by their exponent for odd p and by For each n1 there are exactly two extraspecial groups of order 21+2n for p=2, there are exactly two extraspecial groups of order p1+2n up to isomorphism. They are named as follows.

For odd p, write p+1+2n for the one of exponent p and p1+2n for the one of exponent p2 (The exponent of a finite group). At n=1 these are the Heisenberg group and the modular group (The Heisenberg group of order p3 is extraspecial, and for odd p it has exponent p, The modular group of order p3 is extraspecial, of exponent p2 when p is odd).

For p=2, write 2+1+2n for the one with 22n+2n solutions of x2=1 and 21+2n for the one with 22n2n solutions. At n=1 these are Dih(C4) and Q8 (Dih(C4) and Q8 are extraspecial of order 8, with six and two solutions of x2=1 respectively).

The two names are well defined because in each case the two classification theorems supply exactly two isomorphism classes and an invariant that separates them, so the label is a property of the isomorphism class and not of a presentation.

Remarks

The two source conventions differ in scope and are both recorded here. van Beek's Definition 2.38 writes p±1+2n for every prime, the sign being read off an iterated central product, which is the convention taken above. Craven's Definition 3.3 introduces p+1+2 only for the odd exponent-p group of order p3 and names the others by their constructions. Where the two overlap they agree, and the convention in force here is van Beek's.

At p=2 the signs multiply under central products, as the counting formula shows. At odd p the notation records exponent instead: a central product is of plus type precisely when every order-p3 factor is Heisenberg. If a modular factor occurs, the absorption lemma reduces all modular factors to one, so the product is of minus type. Thus the two uses of the signs agree on the basic plus factors but obey different product rules.

CorollaryStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-08-26Open item page →

An extraspecial group of odd order has exponent p or p2, and an extraspecial 2-group has exponent 4

Statement

Let P be an extraspecial p-group of order p1+2n. If p is odd then exp(P)=p when P is of plus type and exp(P)=p2 when P is of minus type. If p=2 then exp(P)=4, whichever type P is.

Facts & Assumptions

Given: An extraspecial p-group P of order p1+2n with n1.

[F1]

For a finite group G, exp(G)=min{nN:n>0 and gn=e for every gG} (The exponent of a finite group).

[F2]

For odd p, p+1+2n is the extraspecial group of that order with exponent p and p1+2n the one with exponent p2; at p=2 the two are named by their number of solutions of x2=1 (Plus and minus type of an extraspecial p-group).

[F4]

For a group G and a prime p, Gp=gp:gG (The pth-power subgroup Gp).

[L1]

For odd p there are exactly two extraspecial groups of order p1+2n, distinguished by their exponent, which is p for one and p2 for the other (For odd p and each n1 there are exactly two extraspecial groups of order p1+2n, distinguished by their exponent).

[L2]

For each n1 there are exactly two extraspecial groups of order 21+2n, separated by the number of solutions of x2=1 (For each n1 there are exactly two extraspecial groups of order 21+2n).

[L3]

There are n1 subgroups of P, each nonabelian of order p3 with centre Z(P), forming an internal central product of P (Every extraspecial p-group is an internal central product of nonabelian subgroups of order p3).

[L4]

For every prime p there are exactly two nonabelian groups of order p3; at p=2 they are Dih(C4) and Q8 (For each prime there are exactly two nonabelian groups of order p3 up to isomorphism).

[L6]

For a finite p-group P the following are equivalent: P is extraspecial; P is nonabelian, Z(P)=p and P/Z(P) is elementary abelian; P is nonabelian and Z(P)=P=Φ(P) has order p (Three equivalent descriptions of an extraspecial p-group).

[L7]

For every finite p-group P, Φ(P)=PPp (Φ(P)=PPp for a finite p-group).

Proof

technique · direct
1.1

Every gP has gpPpΦ(P)=Z(P), a group of order p, so gp2=(gp)p=e and exp(P) divides p2.

F1F4L6L7
1.2

For odd p the classification names the two isomorphism classes by their exponents, which are p and p2, and the plus and minus labels are those names.

F2L1
2.1

At p=2, take an internal central product decomposition into subgroups of order eight; each is nonabelian, hence isomorphic to Dih(C4) or to Q8, and each contains an element of order four. So exp(P) is a multiple of four, and by step 1.1 it divides four; hence exp(P)=4 for both types.

F1F3L2L3L4L5step 1.1

Remarks

At p=2 the exponent does not separate the two types, and that is why the classification there uses the number of solutions of x2=1 instead. The two invariants are not interchangeable: for odd p the exponent separates the types and the number of solutions of xp=1 does so as well, while at p=2 only the second does.

DefinitionDefinition: Literature-sourcedProof: Not applicablejudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-26Open item page →

The square map of an extraspecial 2-group relative to a chosen generator of its centre

Definition

Let P be an extraspecial 2-group (Special and extraspecial p-groups) and fix the generator z of its centre, so Z(P)={1,z} with z2=1 (The center Z(G) of a group, The order G of a finite group and the order ord(g) of an element, with ord(g)= when no positive power of g is the identity). Write V=P/Z(P) with its canonical F2-vector-space structure (The quotient group G/N and coset product (gN)(hN)=ghN, An elementary abelian p-group has a canonical Fp-vector-space structure, For every prime p, the two operations on Z/p make it a field), and let bz be the commutator pairing (The commutator pairing of an extraspecial p-group relative to a chosen generator of its centre).

The square map of P relative to z is

qz:VF2,x2=zqz(xˉ).

Why the exponent exists and is unique. The quotient P/Z(P) is elementary abelian (Three equivalent descriptions of an extraspecial p-group), so x2Z(P)={1,z} for every xP; and z0=1z=z1, so exactly one class in F2 records which of the two values x2 takes.

That the value depends only on the coset xˉ, and the identity relating it to the commutator pairing, are proved in The square map is well defined on the central quotient and satisfies q(xˉyˉ)=q(xˉ)+q(yˉ)+b(xˉ,yˉ) .

Remarks

The map is not a homomorphism to F2: it fails additivity by exactly the value of the commutator pairing, and that failure is the whole content of the identity proved for it. Where the pairing vanishes the map is additive, and it is on such subspaces that the counting of elementary abelian subgroups is done.

The map is defined only at p=2. For odd p the corresponding assignment xxp also lands in the centre, but it is a homomorphism by the class-two power formula (For an odd prime p, the p-th power map is a homomorphism on a finite group whose derived subgroup is central of exponent dividing p) and carries no extra information beyond the type.

LemmaStatement: AI-adaptedProof: AI-adaptedprecheck passaudited 2026-08-26Open item page →

The square map is well defined on the central quotient and satisfies q(xˉyˉ)=q(xˉ)+q(yˉ)+b(xˉ,yˉ)

Statement

Let P be an extraspecial 2-group with Z(P)=z and V=P/Z(P), and let q=qz be its square map and b=bz its commutator pairing. Then q is well defined on V and

q(xˉyˉ)=q(xˉ)+q(yˉ)+b(xˉ,yˉ)for all xˉ,yˉV.

Moreover q(xˉ)=0 exactly when the elements of the coset xˉ satisfy x2=1; a subgroup UV on which q vanishes satisfies UU and has elementary abelian preimage in P of order 2U; and every maximal elementary abelian subgroup of P contains Z(P).

Facts & Assumptions

Given: An extraspecial 2-group P with Z(P)=z of order two, the quotient V=P/Z(P), the commutator pairing b and the square map q.

[F1]

The square map of P relative to z is qz:VF2 determined by x2=zqz(xˉ) (The square map of an extraspecial 2-group relative to a chosen generator of its centre).

[F2]

The commutator pairing of P relative to z is bz:V×VFp determined by [x,y]=zbz(xˉ,yˉ) (The commutator pairing of an extraspecial p-group relative to a chosen generator of its centre).

[F3]

Z(G):={zG:zg=gz for every gG} (The center Z(G) of a group).

[F4]

An elementary abelian p-group is a finite abelian p-group in which every nonidentity element has order p (Elementary abelian p-groups).

[L1]

For a finite p-group P the following are equivalent: P is extraspecial; P is nonabelian, Z(P)=p and P/Z(P) is elementary abelian; P is nonabelian and Z(P)=P=Φ(P) has order p (Three equivalent descriptions of an extraspecial p-group).

[L2]

If [G,G]Z(G) then (xy)n=[y,x](n2)xnyn for every nN (In a group with central derived subgroup, (xy)n=[y,x](n2)xnyn).

[L3]

An extraspecial p-group is nilpotent of class exactly two, its derived subgroup satisfies P=Z(P) and has order p (An extraspecial p-group is nilpotent of class exactly two and its derived subgroup has order p).

[L4]

The commutator pairing is well defined, Fp-bilinear and alternating, with bz(yˉ,xˉ)=bz(xˉ,yˉ) (The commutator pairing is well defined on the central quotient, is bilinear over Fp, and is alternating).

[L5]

The rule aˉx=xa gives every elementary abelian p-group its canonical Fp-vector-space structure (An elementary abelian p-group has a canonical Fp-vector-space structure).

[L6]

The quotient group G/N has the left cosets gN as elements (The quotient group G/N and coset product (gN)(hN)=ghN).

Proof

technique · direct
1.1

Since P/Z(P) is elementary abelian, x2Z(P)={1,z} for every xP, and z2=1; so (xz)2=x2z2=x2 and the value of q depends only on the coset xˉ, which makes q well defined on V.

F1F3L1L6
2.1

By definition q(xˉ)=0 exactly when x2=1, and since q is well defined this holds for one element of the coset exactly when it holds for both.

F1F5step 1.1
2.2

The class-two power formula at exponent two gives (xy)2=[y,x](22)x2y2=[y,x]x2y2, and [y,x]=zb(yˉ,xˉ). Over F2 one has 1=1, so b(yˉ,xˉ)=b(xˉ,yˉ)=b(xˉ,yˉ); hence zq(xˉyˉ)=zb(xˉ,yˉ)+q(xˉ)+q(yˉ) and the displayed identity holds.

F1F2L2L3L4L5step 1.1
3.1

Let UV satisfy q(u)=0 for every uU. Then for u,uU the identity gives 0=q(uu)=q(u)+q(u)+b(u,u)=b(u,u), so UU.

F2step 2.2
4.1

Let E be the preimage of such a U in P. It contains Z(P), has order 2U, is abelian because b vanishes on U, and every one of its elements x satisfies x2=1 by step 2.1 together with z2=1; so E is elementary abelian of order 2U.

F3F4L6step 2.1step 3.1
5.1

If E is elementary abelian and does not contain Z(P), then EZ(P) is trivial and EZ(P) is an abelian subgroup all of whose elements square to the identity, properly containing E; so a maximal elementary abelian subgroup contains Z(P).

F3F4step 2.1step 4.1

Remarks

The identity is not additivity, and the correction term is where the two isomorphism types differ: on a subspace where b vanishes the map q is additive and its zero set is a subspace, and it is precisely the size of the largest such subspace that separates the two extraspecial groups of a given order.

PropositionStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-08-26Open item page →

An automorphism of an extraspecial p-group acting trivially on its Frattini quotient is inner

Statement

Let P be an extraspecial p-group of order p1+2n and let ρP:Aut(P)AutFp(P/Φ(P)) be the induced-action homomorphism. Then kerρP=Inn(P): an automorphism of P acting trivially on the Frattini quotient is inner.

Facts & Assumptions

Given: An extraspecial p-group P of order p1+2n.

[F1]

Inn(G):={cg:gG} with cg(x)=gxg1 (Inner automorphisms and Inn(G)).

[F2]

For a finite p-group P, the generator rank d(P) is the common size of a basis of P/Φ(P) (The generator rank d(P) of a finite p-group).

[F3]

Φ(G) is the intersection of the maximal proper subgroups of G (The Frattini subgroup Φ(G) as the intersection of the maximal subgroups of a finite group).

[F4]

An isomorphism is a bijective group homomorphism, and Aut(G) is the set of automorphisms of G (Group isomorphisms, automorphisms and the set Aut(G)).

[L1]

For a finite p-group P the following are equivalent: P is extraspecial; P is nonabelian, Z(P)=p and P/Z(P) is elementary abelian; P is nonabelian and Z(P)=P=Φ(P) has order p (Three equivalent descriptions of an extraspecial p-group).

[L2]

An extraspecial p-group has P=p1+2n with n1 and P/Z(P)=p2n (An extraspecial p-group has order p1+2n for some n1).

[L3]

An extraspecial p-group of order p1+2n has Φ(P)=Z(P) of order p and generator rank d(P)=2n, and every minimal generating set has 2n elements (An extraspecial p-group of order p1+2n has generator rank 2n).

[L4]

Every automorphism of a finite p-group induces an Fp-linear automorphism of P/Φ(P), and these form a homomorphism ρP (Automorphisms act linearly on the Frattini quotient).

[L5]

A subset X of a finite p-group P is a minimal generating set if and only if the quotient map restricts to a bijection from X onto a basis of P/Φ(P) (Burnside Basis Theorem).

[L6]

G/Z(G)Inn(G) (G/Z(G)Inn(G)).

[L7]

If A and B are finite then A×B is finite and A×B=AB (The product rule: A×B=AB, and i<mAi=i<mAi).

[L8]

A is the unique natural number n with An (The cardinality A of a finite set).

Proof

technique · direct
1.1

The Frattini subgroup is Φ(P)=Z(P) of order p, the Frattini quotient has order p2n, and every minimal generating set of P has exactly 2n elements.

F2F3L1L2L3
2.1

Fix a minimal generating set X={g1,,g2n}, which exists because the Burnside basis theorem matches minimal generating sets with bases of the Frattini quotient. An automorphism of P is determined by its values on X, since X generates P.

F4L5step 1.1
2.2

Every inner automorphism lies in kerρP: for g,xP one has cg(x)x1=gxg1x1[P,P]=Φ(P), so cg fixes every coset of Φ(P).

F1L1L4step 1.1
2.3

The inner automorphism group has order P/Z(P)=p2n.

F1L2L6step 1.1
3.1

If θ lies in kerρP then θ(gi)Φ(P)=giΦ(P) for each i, so θ(gi)=giui with uiΦ(P)=Z(P), a set of p elements. Hence θ is determined by the tuple (u1,,u2n), and there are at most p2n such tuples, so kerρPp2n.

F3L4L7L8step 1.1step 2.1
4.1

So Inn(P) is a subset of kerρP of size p2n, while kerρP has at most p2n elements; hence the two coincide.

L8step 3.1step 2.2step 2.3

Remarks

The equality is forced by two counts that happen to agree, and each uses the extraspecial hypothesis: the upper bound uses Φ(P)=Z(P) of order p together with the generator rank 2n, and the lower bound uses P/Z(P)=p2n. For a general finite p-group the kernel can be larger than the inner automorphism group; equality is not asserted or excluded without additional hypotheses.

PropositionStatement: Literature-sourcedProof: AI-adaptedprecheck passaudited 2026-08-26Open item page →

An automorphism fixing the centre pointwise induces a pairing-preserving automorphism of the central quotient, with kernel the inner automorphisms

Statement

Let P be an extraspecial p-group with Z(P)=z and commutator pairing bz, and let αAut(P) fix Z(P) pointwise.

Then α induces an automorphism αˉ of P/Z(P) satisfying

bz(αˉ(xˉ),αˉ(yˉ))=bz(xˉ,yˉ)

for all xˉ,yˉP/Z(P). The kernel of the action of the centre-fixing automorphism subgroup on P/Z(P) is Inn(P).

Facts & Assumptions

Given: An extraspecial p-group P with Z(P)=z, its commutator pairing bz, and an automorphism α fixing Z(P) pointwise.

[F1]

For an extraspecial p-group P with Z(P)=z, the commutator pairing is the map bz(xˉ,yˉ)Z/p determined by [x,y]=zbz(xˉ,yˉ) (The commutator pairing of an extraspecial p-group relative to a chosen generator of its centre).

[L1]

For a finite p-group P the following are equivalent: P is extraspecial; P is nonabelian, Z(P)=p and P/Z(P) is elementary abelian; P is nonabelian and Z(P)=P=Φ(P) has order p (Three equivalent descriptions of an extraspecial p-group).

[L2]

Every extraspecial p-group has order p1+2n for some n1 (An extraspecial p-group has order p1+2n for some n1).

[L3]

For an extraspecial p-group of order p1+2n, an automorphism acting trivially on its Frattini quotient is inner (An automorphism of an extraspecial p-group acting trivially on its Frattini quotient is inner).

[L4]

Group isomorphisms, automorphisms and the set Aut(G). (Group isomorphisms, automorphisms and the set Aut(G)).

[L5]

Inner automorphisms and Inn(G). (Inner automorphisms and Inn(G)).

Proof

technique · direct
1.1

Because α fixes Z(P) pointwise, it sends each coset xZ(P) to α(x)Z(P); this is well defined, so α induces an automorphism αˉ of P/Z(P). Also [α(x),α(y)]=α([x,y])=α(zbz(xˉ,yˉ))=zbz(xˉ,yˉ), so bz(αˉ(xˉ),αˉ(yˉ))=bz(xˉ,yˉ) for all xˉ,yˉP/Z(P).

F1L4L6
2.1

If αˉ is the identity on P/Z(P), then α acts trivially on the central quotient. Since P is extraspecial, [L1] gives Φ(P)=Z(P) and [L2] supplies the order hypothesis of [L3], so α acts trivially on the Frattini quotient and is inner. Conversely, every inner automorphism cg(x)=gxg1 acts trivially on P/Z(P) because gxg1x1=[g,x][P,P]=Z(P) by [L1] and [L6]. Hence the kernel of the action on P/Z(P) is exactly Inn(P).

L1L2L3L5L6step 1.1
PropositionStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (deepseek-v4-pro + gpt-5.6-terra)audited 2026-08-26Open item page →

The maximal elementary abelian subgroups of the two extraspecial groups of order 21+2n have orders 2n+1 and 2n

Statement

The maximal elementary abelian subgroups of the two extraspecial groups of order 21+2n have orders 2n+1 and 2n.

Facts & Assumptions

Given: The hypotheses of the Statement.

[L1]

For an extraspecial 2-group P of order 21+2n with Z(P)=z, the square map is the function q determined by x2=zq(xˉ), with values in F2 (The square map of an extraspecial 2-group relative to a chosen generator of its centre).

[L2]

The square map is well defined on the central quotient and satisfies q(xˉyˉ)=q(xˉ)+q(yˉ)+b(xˉ,yˉ) (The square map is well defined on the central quotient and satisfies q(xˉyˉ)=q(xˉ)+q(yˉ)+b(xˉ,yˉ)).

[L3]

For a subgroup Aˉ of P/Z(P) one has AˉAˉ=P/Z(P) (A subgroup of the central quotient and its orthogonal complement have orders multiplying to the order of the quotient).

[L4]

For each n1 there are exactly two extraspecial groups of order 21+2n up to isomorphism, with 22n+2n and 22n2n solutions of x2=1 (For each n1 there are exactly two extraspecial groups of order 21+2n).

[L5]

If P1 and P2 are extraspecial 2-groups with ti solutions of x2=1, then P1P2 has (t1t2+(P1t1)(P2t2))/2 such solutions (A product formula for the number of square roots of the identity in a central product of extraspecial 2-groups).

[L6]

Every maximal abelian subgroup of an extraspecial p-group of order p1+2n has order p1+n (In an extraspecial p-group of order p1+2n every maximal abelian subgroup has order p1+n).

[L7]

An extraspecial p-group has order p1+2n for some n1 (An extraspecial p-group has order p1+2n for some n1).

[L8]

The commutator pairing is independent of the coset representatives, is Fp-bilinear on P/Z(P), and is alternating (The commutator pairing is well defined on the central quotient, is bilinear over Fp, and is alternating).

[L9]

The commutator pairing of an extraspecial p-group has trivial radical (The commutator pairing of an extraspecial p-group has trivial radical).

[L10]

For an extraspecial p-group P with Z(P)=z, the commutator pairing is the map bz(xˉ,yˉ)Z/p determined by [x,y]=zbz(xˉ,yˉ) (The commutator pairing of an extraspecial p-group relative to a chosen generator of its centre).

[L11]

An elementary abelian p-group is a finite abelian p-group in which every nonidentity element has order p; the trivial group is permitted (,, ). (Elementary abelian p-groups).

[L12]

The set S is independent when sSsas=e with finite support forces every as=0. A basis of an elementary abelian p-group is an independent spanning subset for its canonical Fp-linear structure. (Fp-spanning sets, independence, and bases in an elementary abelian p-group).

[L13]

Every finite elementary abelian p-group has a basis; every independent subset extends to a basis, every spanning subset contains a basis, and all bases have the same finite size. (Finite elementary abelian p-groups have bases, basis extension, and a well-defined dimension).

[L15]

S;:=;{H;:;HG and SH}. (The subgroup S generated by a subset, the cyclic subgroup g, and cyclic groups).

[L16]

Z(G):={zG:zg=gz for every gG}. (The center Z(G) of a group).

[L17]

An extraspecial 2-group of order 21+2n is of plus type when it has 22n+2n solutions of x2=1 and of minus type when it has 22n2n such solutions; for odd p, plus and minus mean exponent p and p2 respectively (Plus and minus type of an extraspecial p-group).

[L18]

A central product of extraspecial p-groups identified along their centres is extraspecial (A central product of extraspecial p-groups identified along their centres is extraspecial).

[L19]

Every extraspecial p-group is an internal central product of nonabelian subgroups of order p3 pairwise intersecting in its centre (Every extraspecial p-group is an internal central product of nonabelian subgroups of order p3).

Proof

technique · direct
1.1

A maximal elementary abelian subgroup E contains Z(P), so EˉP/Z(P) has order 2k with E=2k+1, and q vanishes on Eˉ.

L1L2L7L11L14L16
1.2

On a coset Eˉvˉ with vˉEˉ the map uˉb(uˉ,vˉ) is a nonzero F2-linear functional on Eˉ, so its kernel has index two and exactly 2k1 of the coset's classes have q=0.

L1L2L8L10L12L13
2.1

Maximality of E says no vˉEˉ orthogonal to Eˉ has q(vˉ)=0: the polar identity would make q vanish on Eˉvˉ, whose preimage is a strictly larger elementary abelian subgroup.

L2L10L11L15step 1.1
2.2

On the coset Eˉ itself q vanishes identically, contributing 2k classes.

L1step 1.1
3.1

On a coset Eˉvˉ with vˉEˉEˉ the correction term vanishes, so q is constantly q(vˉ)=1 by step 2.1 and the coset contributes no class.

L1L2step 2.1
4.1

With Eˉ=22nk there are 22n2k cosets inside the complement and 22nk22n2k outside, so the classes with q=0 number 2k+(22nk22n2k)2k1=22n1+2k22nk1, and the elements with x2=1 number twice that.

L1L2L3L7L9L14step 2.2step 3.1step 1.2algebra
5.1

The classification counts those elements as 22n+2n in the plus case and 22n2n in the minus case, so 2k22nk1=±2n1.

L4L5step 4.1algebra
6.1

The left side is strictly increasing in k, so k=n in the plus case and k=n1 in the minus case are the only solutions; both are attained, giving E=2n+1 and E=2n for EVERY maximal elementary abelian subgroup.

L11L13L17L18L19step 2.1step 5.1algebra
7.1

Independently, E is abelian, so the maximal-abelian bound gives E2n+1 and hence kn without the monotonicity argument; this is the free half of the plus case.

L6step 1.1

5 · Examples, counterexamples and false statements

None yet.

Sources