betaFeedback

WorksMathematics

Degree-one cohomology of metacyclic 22-groups: a low-bit dichotomy

Abstract

We study the subalgebra generated by degree-one mod-22 cohomology for metacyclic 22-groups G=⟨a,b∣a2α=1, b2m=ac, b−1ab=ar⟩G=\langle a,b\mid a^{2^α}=1,\ b^{2^m}=a^c,\ b^{-1}ab=a^r\rangle with unique normal forms, where α,m≥2α,m\ge2 and r≡−1(mod4)r\equiv-1\pmod4. Presentation compatibility forces c=0c=0 or 2α−12^{α-1} and Gab≅C2×C2mG^{\mathrm{ab}}\cong C_2\times C_{2^m}. If c≡0(mod4)c\equiv0\pmod4, the degree-one-generated algebra is F2[x,y]/(y2,x2+xy)\mathbb{F}_2 [x,y]/(y^2,x^2+xy), a four-dimensional algebra shared by the four nonmodular types of Aaboun–Zekhnini. The unique exceptional case α=2α=2, c=2c=2 is modular and has infinite-dimensional degree-one algebra F2[x,y]/(y2,x2y)\mathbb{F}_2[x,y]/(y^2,x^2y). The regular algebra is PBW, Koszul, universally Koszul, and strongly Koszul, whereas the modular algebra is not Koszul. We realize the regular algebra by inflation from the common quotient C4⋊−1C4C_4\rtimes_{-1}C_4 and identify its quadratic dual with the characteristic-two Jordan plane. For the inversion tower C4⋊−1C2mC_4\rtimes_{-1}C_{2^m}, we give uniform presentations of the full mod-22 cohomology rings and determine the kernels and images of inflation in every degree. Their direct limit yields the continuous cohomology ring of C4⋊−1Z2C_4\rtimes_{-1}\mathbb Z_2. We also record an arithmetic consequence for Galois groups arising over cyclotomic Z2\mathbb Z_2-extensions of real quadratic fields.

Introduction

The mod-2 cohomology of metacyclic 2-groups has been studied at the level of the full cohomology ring. Diethelm treated nonabelian split metacyclic pp-groups [12], Rusin computed the mod-2 rings of metacyclic 2-groups together with their Steenrod operations [27], and Huebschmann subsequently treated the mod-pp cohomology of general metacyclic groups [17]. These calculations reveal substantial higher-degree structure even for groups with relatively simple presentations. For standard background on finite-group cohomology, see [2, 4, 6]. In particular, the theorem of Chebolu–Mináč–Okay–Schultz–Ure [8 Theorem 1.1] implies that the full mod-2 cohomology ring of a nonabelian finite 2-group is not generated by degree-one classes. For a graded F2\mathbb{F}_2-algebra RR, write ⟨R1⟩\langle R_1\rangle for the subalgebra generated by its degree-one component, and set

Dec⁡(H∗(G,F2)):=⟨H1(G,F2)⟩.\operatorname{Dec}(H^*(G,\mathbb{F}_2)):=\langle H^1(G,\mathbb{F}_2)\rangle.

This subalgebra records the cup products of degree-one characters and provides a natural object on which to study how the presentation of a metacyclic group controls its cohomology. The relation between degree-one cohomology, Koszul algebras, and quadratic duality has also been investigated in Galois cohomology and pro-pp group theory [22, 26]. Aaboun and Zekhnini classify metacyclic-nonmodular 2-groups with abelianization of type (2,2m)(2,2^m) into four refined presentation types [1 Proposition 2.4]. The types differ in their conjugation and power relations, yet their degree-one-generated cohomology algebras have the same form. We explain this uniformity directly from the lowest 2-adic information in the presentations, before distinguishing the individual types. The relevance of the congruence modulo 4 comes from the Bockstein itself. For a degree-one mod-2 character, the Bockstein can be computed using a set-theoretic lift to Z/4Z\mathbb{Z}/4\mathbb{Z}, so the resulting cocycle is determined by the corresponding low-bit multiplication law. This provides a direct route from the presentation to the degree-one cup products, without first computing the full cohomology ring. Consider a metacyclic group with unique normal forms

G=⟨a,b∣a2α=1, b2m=ac, b−1ab=ar⟩,α,m≥2,0≤c<2α,G=\langle a,b\mid a^{2^\alpha}=1,\ b^{2^m}=a^c,\ b^{-1}ab=a^r\rangle,\qquad\alpha,m\ge2,\quad0\le c<2^\alpha,

and assume r≡−1(mod4)r\equiv-1\pmod{4}. The power relation records the carry that occurs when the exponent of bb passes through 2m2^m. The compatibility conditions force

c=0orc=2α−1,Gab≅C2×C2m.c=0\quad\text{or}\quad c=2^{\alpha-1},\qquad G^{\mathrm{ab}}\cong C_2\times C_{2^m}.

Thus the carry is divisible by 44 except when α=2\alpha=2 and c=2c=2. This singles out the modular boundary directly from the presentation. Let x,y∈H1(G,F2)x,y\in H^1(G,\mathbb{F}_2) be the characters adapted to the generators a,ba,b. In the regular branch, the normal-form multiplication modulo 44 gives the Bockstein relations

β(y)=0,β(x)=xy≠0.\beta(y)=0,\qquad\beta(x)=xy\ne0.

As β(z)=z2\beta(z)=z^2 for a degree-one mod-22 class, these identities determine the entire degree-one-generated algebra:

Dec⁡(H∗(G,F2))≅F2[x,y]/(y2,x2+xy).\operatorname{Dec}(H^*(G,\mathbb{F}_2))\cong\mathbb{F}_2[x,y]/(y^2,x^2+xy).

The algebra is four-dimensional, with basis 1,x,y,xy1,x,y,xy. At the unique modular boundary MmM_m, the degree-one algebra instead has the presentation

Dec⁡(H∗(Mm,F2))≅F2[x,y]/(y2,x2y).\operatorname{Dec}(H^*(M_m,\mathbb{F}_2))\cong\mathbb{F}_2[x,y]/(y^2,x^2y).

In this case the powers of xx remain nonzero in every degree. Thus the divisibility of the carry by 44 distinguishes two different algebraic behaviors. The uniformity in the regular branch also has a group-theoretic explanation. Reducing the two generators modulo their fourth powers gives a surjection onto C4⋊−1C4C_4\rtimes_{-1}C_4. Inflation along this quotient induces an isomorphism of degree-one-generated cohomology algebras. Hence the common algebra of the four nonmodular types comes from a single group of order 1616 through an explicit homomorphism. We next examine the homological properties of these two algebras. The regular algebra is PBW, Koszul, universally Koszul, and strongly Koszul; its quadratic dual is the characteristic-two Jordan plane. The modular algebra has a minimal cubic relation and is not Koszul. Consequently, the same low-bit condition that determines the degree-one products also determines Koszulity. The common quotient C4⋊−1C4C_4\rtimes_{-1}C_4 is the first level of the inversion tower

Gm=C4⋊−1C2m,m≥2.G_m=C_4\rtimes_{-1}C_{2^m},\qquad m\ge2.

This family permits a comparison between the stable degree-one core and the higher-degree classes of the full cohomology ring. We determine H∗(Gm,F2)H^*(G_m,\mathbb{F}_2) at every finite level and show that inflation kills precisely the ideal generated by the periodicity class coming from the cyclic quotient, while preserving the class restricting nontrivially to C4C_4. Passing to the direct limit gives

Hcts∗(C4⋊−1Z2,F2)≅F2[x,y,ω]/(y2,x2+xy),∣ω∣=2.H^*_{\mathrm{cts}}(C_4\rtimes_{-1}\mathbb{Z}_2,\mathbb{F}_2)\cong\mathbb{F}_2[x,y,\omega]/(y^2,x^2+xy),\qquad|\omega|=2.

Finally, an arithmetic application shows how the group-theoretic classification determines the degree-one algebra of certain Galois groups over cyclotomic Z2\mathbb{Z}_2-extensions of real quadratic fields. Once the arithmetic hypotheses select a metacyclic type, the cohomological and Koszul conclusions follow from the presentation.

Metacyclic presentations and low-bit compatibility

Compatibility and abelianization

A metacyclic group is built from two cyclic pieces: a cyclic normal subgroup and a cyclic quotient. We consider finite metacyclic 22-groups with a cyclic normal subgroup

A=⟨a⟩≅C2αA=\langle a\rangle\cong C_{2^\alpha}

and cyclic quotient

G/A≅C2m.G/A\cong C_{2^m}.

where α,m≥2\alpha,m\ge2. Choose an element b∈Gb\in G whose image generates G/AG/A. Since AA is normal, conjugation by bb induces an automorphism of AA. As AA is cyclic, there is an odd integer rr such that b−1ab=arb^{-1}ab=a^r. The parameter rr therefore describes how the quotient generator acts on the normal cyclic subgroup.

Although the image of bb has order 2m2^m in G/AG/A, the element bb itself need not have order 2m2^m in GG. What is required is that b2mb^{2^m} belongs to AA, so there is a uniquely determined integer cc, with 0≤c<2α0\le c<2^\alpha, such that

b2m=ac.b^{2^m}=a^c.

Thus cc records the possible discrepancy between the order of bb and the order of its image in the quotient. More concretely, when products are expressed in terms of powers of aa and bb, the relation b2m=acb^{2^m}=a^c determines the additional factor that appears whenever the exponent of bb passes through 2m2^m. It is therefore convenient to regard aca^c as the carry in the power relation.

Every element of GG can be expressed in the form aibja^i b^j. We assume throughout that these expressions are unique for

0≤i<2α,0≤j<2m.0\le i<2^\alpha,\qquad0\le j<2^m.

In particular, ∣G∣=2α+m|G|=2^{\alpha+m}, and we work with the unified presentation

G=⟨a,b∣a2α=1, b2m=ac, b−1ab=ar⟩.(1)G=\langle a,b\mid a^{2^\alpha}=1,\ b^{2^m}=a^c,\ b^{-1}ab=a^r\rangle. \tag*{(1)}

The parameters rr and cc describe different aspects of the extension: the first determines the conjugation action, while the second determines the power relation. These parameters are nevertheless constrained by the requirement that both relations hold simultaneously.

Indeed, since b2m=acb^{2^m}=a^c lies in the abelian subgroup AA, conjugation by b2mb^{2^m} must act trivially on AA. At the same time, the element ac=b2ma^c=b^{2^m} must be fixed by conjugation by bb. These two compatibility conditions impose congruences on rr and cc, which we establish before considering the refined classification of metacyclic 2-groups.

Proposition 2.1. For a presentation (2.1) with the stated normal forms, one has

r2m≡1(mod2α),c(r−1)≡0(mod2α).r^{2^m}\equiv1\pmod{2^\alpha},\qquad c(r-1)\equiv0\pmod{2^\alpha}.

If, in addition, r≡−1(mod4)r\equiv-1\pmod4, then necessarily c=0c=0 or c=2α−1c=2^{\alpha-1}. Under this additional hypothesis, moreover,

G′=⟨a2⟩,Gab≅C2×C2m.G'=\langle a^2\rangle,\qquad G^{\mathrm{ab}}\cong C_2\times C_{2^m}.

Proof. For the first congruence, we compare two ways in which b2mb^{2^m} acts on the normal subgroup AA. Since b2m=ac∈Ab^{2^m}=a^c\in A and AA is abelian, conjugation by b2mb^{2^m} acts trivially on AA. In particular,

(b2m)−1ab2m=a.\left(b^{2^m}\right)^{-1}ab^{2^m}=a.

On the other hand, after 2m2^m conjugations by bb and using b−1ab=arb^{-1}ab=a^r, we obtain

(b2m)−1ab2m=ar2m.\left(b^{2^m}\right)^{-1}ab^{2^m}=a^{r^{2^m}}.

Thus ar2m=aa^{r^{2^m}}=a, and since aa has order 2α2^\alpha, this gives the first congruence.

For the second congruence, note that

ac=b2m=b−1b2mb=b−1acb=(b−1ab)c=acr,a^c=b^{2^m}=b^{-1}b^{2^m}b=b^{-1}a^c b=(b^{-1}ab)^c=a^{cr},

and we hence

c(r−1)≡0(mod2α).c(r-1)\equiv0\pmod{2^\alpha}.

Now if r≡−1(mod4)r\equiv-1\pmod4, then r−1≡2(mod4)r-1\equiv2\pmod4, so v2(r−1)=1v_2(r-1)=1. Thus the remaining factor 2α−12^{\alpha-1} must divide cc, implying 2α−1∣c2^{\alpha-1}\mid c. Because cc has been chosen in the range 0≤c<2α0\le c<2^\alpha, there are only two possibilities: c=0c=0 or c=2α−1c=2^{\alpha-1}.

We next determine the commutator subgroup. Since G/AG/A is cyclic hence abelian, it is abelian, G′⊆AG'\subseteq A. Because AA itself is cyclic and therefore abelian, the only nontrivial commutation is [a,b]=a−1b−1ab=ar−1[a,b] = a^{-1}b^{-1}ab = a^{r-1}, and so G′=⟨ar−1⟩G' = \langle a^{r-1}\rangle. But v2(r−1)=1v_2(r-1) = 1, so r−1=2ur-1 = 2u for some odd integer uu, which then gives

⟨ar−1⟩=⟨a2u⟩=⟨a2⟩.\langle a^{r-1}\rangle= \langle a^{2u}\rangle= \langle a^2\rangle.

Therefore G′=⟨a2⟩G' = \langle a^2\rangle. It remains to identify the abelianization. Since a2∈G′a^2 \in G', the image of aa in Gab=G/G′G^{\mathrm{ab}} = G/G' has order at most 2. It is not trivial, because a∉⟨a2⟩a \notin\langle a^2\rangle. Thus the image of aa has order exactly 2.

The defining relation b2m=acb^{2^m} = a^c becomes bˉ2m=aˉc\bar b^{2^m} = \bar a^c in the abelianization. We have already shown that cc is either 0 or 2α−12^{\alpha-1}, and in either case cc is even. Since aˉ\bar a has order 2, this gives aˉc=1\bar a^c = 1. Hence bˉ2m=1\bar b^{2^m} = 1. Thus the order of bˉ\bar b divides 2m2^m. In fact its order is exactly 2m2^m. If bˉk=1\bar b^k = 1, then bk∈G′=⟨a2⟩⊆Ab^k \in G' = \langle a^2\rangle\subseteq A. Projecting to G/A≅C2mG/A \cong C_{2^m} shows that the image of bkb^k is trivial, and since the image of bb generates G/AG/A, this forces 2m∣k2^m \mid k. Therefore bˉ\bar b has order 2m2^m.

Finally, the two cyclic factors are independent. Suppose aˉεbˉj=1\bar a^\varepsilon\bar b^j = 1, ε∈{0,1}\varepsilon\in\{0,1\}. Then aεbj∈G′⊆Aa^\varepsilon b^j \in G' \subseteq A. Projecting to G/AG/A gives j≡0(mod2m)j \equiv0 \pmod{2^m}. The relation therefore reduces to aε∈G′=⟨a2⟩a^\varepsilon\in G' = \langle a^2\rangle, which forces ε=0\varepsilon= 0. Consequently

Gab≅C2×C2m.G^{\mathrm{ab}} \cong C_2 \times C_{2^m}.

The preceding proposition explains why the carry condition that will be used later is almost automatic. Under the hypothesis r≡−1(mod4)r \equiv-1 \pmod4, compatibility leaves only two possible values of the carry: c=0c = 0 or c=2α−1c = 2^{\alpha-1}. The first gives b2m=1b^{2^m} = 1, while the second gives b2m=a2α−1b^{2^m} = a^{2^{\alpha-1}}, the unique element of order 2 in AA. If α≥3\alpha\ge3, then 2α−1≡0(mod4)2^{\alpha-1} \equiv0 \pmod4, so both compatible values of cc are divisible by 4. Thus the condition c≡0(mod4)c \equiv0 \pmod4 does not need to be imposed separately once α≥3\alpha\ge3.

There is only one exception: When α=2\alpha= 2, the nonzero compatible carry is c=2c = 2. This is the unique compatible case in which the carry is not divisible by 4. This exceptional situation is the modular boundary treated separately below.

Product convention and low-bit multiplication

Proposition 2.2. For 0≤j,v<2m0 \le j,v < 2^m,

(aibj)(aubv)=ai+r−ju+δcbj+v−δ2m.(a^i b^j)(a^u b^v) = a^{i+r^{-j}u+\delta c}b^{j+v-\delta2^m}.

where δ=0\delta= 0 if j+v<2mj+v < 2^m and δ=1\delta= 1 if j+v≥2mj+v \ge2^m.

Proof. The relation b−1ab=arb^{-1}ab = a^r implies bjaub−j=ar−jub^j a^u b^{-j} = a^{r^{-j}u}, and hence (aibj)(aubv)=ai+r−jubj+v(a^i b^j)(a^u b^v) = a^{i+r^{-j}u}b^{j+v}. Because 0≤j,v<2m0 \le j,v < 2^m, at most one use of b2m=acb^{2^m} = a^c is required to return to normal form, giving the stated formula. In particular, with the convention b−1ab=arb^{-1}ab = a^r, the exponent is r−jr^{-j}.

Lemma 2.3. Assume r≡−1(mod4)r \equiv-1 \pmod4 and either c=0c = 0 or α≥3\alpha\ge3. If (aibj)(aubv)=awbt(a^i b^j)(a^u b^v) = a^w b^t is in normal form, then

w≡i+(−1)ju(mod4)w \equiv i + (-1)^j u \pmod4

and

t≡j+v(mod2).t \equiv j+v \pmod2.

Proof. Since r−1≡−1(mod4)r^{-1} \equiv-1 \pmod4, we have that r−j≡(−1)j(mod4)r^{-j} \equiv(-1)^j \pmod4. By Proposition 2.1, the carry is either 0 or 2α−12^{\alpha-1}. In either case, the carry vanishes modulo 4, yielding the first congruence. The second follows since reduction to normal form changes the exponent of bb only by multiples of 2m2^m, which is even.

The four refined nonmodular types

We recall the classification of metacyclic-nonmodular 2-groups with abelianization of type (2,2m)(2,2^m) given by Aaboun–Zekhnini [1 Proposition 2.4]. Their four types are distinguished by the conjugation action and the power relation. In Types 1 and 2, the action is inversion, so r=−1r=-1, whereas Types 3 and 4 have r=−1+2sr=-1+2^s for an integer ss subject to the restrictions below. The power relation is split in Types 1 and 3, where c=0c=0, and nonsplit in Types 2 and 4, where c=2α−1c=2^{\alpha-1}.

The four types, together with their parameter restrictions, are summarized in the following table.

TyperrccParameter restrictions
1−1-100α≥2\alpha\ge2
2−1-12α−12^{\alpha-1}α≥3\alpha\ge3
3−1+2s-1+2^s00Either s=α−1s=\alpha-1 with α≥3\alpha\ge3, or 2≤s≤α−22 \le s \le\alpha-2 with α≥4\alpha\ge4
4−1+2s-1+2^s2α−12^{\alpha-1}2≤s≤α−22 \le s \le\alpha-2, α≥4\alpha\ge4

Table 1.

Thus, the classification separates the groups according to the conjugation action and the possible values of the power parameter cc. Notably, the case s=α−1s=\alpha-1 occurs only in Type 3, while Types 3 and 4 both admit 2≤s≤α−22 \le s \le\alpha-2.

In every row, r≡−1(mod4)r\equiv-1\pmod{4}. Types 1 and 3 are split. In Types 2 and 4 the classification imposes α≥3\alpha\ge3, so the nonsplit carry is divisible by 4. Hence all four types lie in the regular branch of the low-bit dichotomy without any additional cohomological hypothesis.

The unique low-bit boundary

The only compatible nonsplit presentation with r≡−1(mod4)r\equiv-1\pmod{4} and carry not divisible by 4 is

Mm=⟨a,b∣a4=1, b2m=a2, b−1ab=a−1⟩,m≥2.M_m=\langle a,b\mid a^4=1,\ b^{2^m}=a^2,\ b^{-1}ab=a^{-1}\rangle,\quad m\ge2.

Proposition 2.4. The group MmM_m is modular. More precisely, bb has order 2m+12^{m+1} and, with t=ab2m−1t=ab^{2^{m-1}}, one has

t2=1,tbt−1=b1+2m.t^2=1,\quad tbt^{-1}=b^{1+2^m}.

Thus

Mm≅⟨b,t∣b2m+1=1, t2=1, tbt−1=b1+2m⟩.M_m\cong\langle b,t\mid b^{2^{m+1}}=1,\ t^2=1,\ tbt^{-1}=b^{1+2^m}\rangle.

Proof. The relation b2m=a2≠1b^{2^m}=a^2\ne1 shows that bb has order 2m+12^{m+1}. Since 2m−12^{m-1} is even, b2m−1b^{2^{m-1}} centralizes aa, and therefore t2=a2b2m=a2a2=1t^2=a^2b^{2^m}=a^2a^2=1. From b−1ab=a−1b^{-1}ab=a^{-1} we obtain aba−1=ba−2=ba2=b1+2maba^{-1}=ba^{-2}=ba^2=b^{1+2^m}. The factor b2m−1b^{2^{m-1}} commutes with bb, so the same conjugation formula holds with aa replaced by tt. □

Theorem 2.5. Let m≥2m\ge2 and let

Mm=⟨a,b∣a4=1, b2m=a2, b−1ab=a−1⟩.M_m=\langle a,b\mid a^4=1,\ b^{2^m}=a^2,\ b^{-1}ab=a^{-1}\rangle.

Define x,y∈H1(Mm,F2)x,y\in H^1(M_m,\mathbb{F}_2) by

x(aibj)=i(mod2),y(aibj)=j(mod2).x(a^i b^j)=i\pmod2,\quad y(a^i b^j)=j\pmod2.

Then

Dec⁡(H∗(Mm,F2))≅F2[x,y](y2,x2y),∣x∣=∣y∣=1.\operatorname{Dec}(H^*(M_m,\mathbb{F}_2))\cong\frac{\mathbb{F}_2[x,y]}{(y^2,x^2y)},\quad|x|=|y|=1.

In particular, xy≠0xy\ne0, xn≠0x^n\ne0 for every n≥1n\ge1, and y2=x2y=0y^2=x^2y=0. The algebra has F2\mathbb{F}_2-basis {xn:n≥0}∪{y,xy}\{x^n:n\ge0\}\cup\{y,xy\} and Hilbert series 11−t+t+t2\frac{1}{1-t}+t+t^2.

Proof. By Proposition 2.4, with t=ab2m−1t=ab^{2^{m-1}} we have

Mm=K⋊T,K=⟨b⟩≅C2m+1,T=⟨t⟩≅C2,M_m=K\rtimes T,\quad K=\langle b\rangle\cong C_{2^{m+1}},\quad T=\langle t\rangle\cong C_2,

where conjugation by tt acts on KK as σ(b)=b1+2m\sigma(b)=b^{1+2^m}. Since m≥2m\ge2,

x(t)=1,y(t)=0,x(b)=0,y(b)=1.x(t)=1,\quad y(t)=0,\quad x(b)=0,\quad y(b)=1.

Thus xx restricts to the generator of H1(T,F2)H^1(T,\mathbb{F}_2), while yy restricts to the generator of H1(K,F2)H^1(K,\mathbb{F}_2). Consequently, xn≠0x^n\ne0 for every n≥1n\ge1, by restriction to TT.

Moreover, yy lifts to a homomorphism Mm→Z/4ZM_m\to\mathbb{Z}/4\mathbb{Z} sending aa to 00 and bb to $1: the defining relations are preserved because m≥2m\ge2. Therefore the mod-2 Bockstein of yy vanishes, and y2=0y^2=0.

The restriction to TT shows that the powers of xx persist, while the lift of yy gives y2=0y^2=0. To determine the remaining products, we first compute the dimension of H2(Mm,F2)H^2(M_m,\mathbb{F}_2). This will constrain the differentials in the mod-2 Lyndon–Hochschild–Serre spectral sequence and allow us to identify the first relation involving both generators. Consider the integral homology Lyndon–Hochschild–Serre spectral sequence associated with 1→K→Mm→T→11\to K\to M_m\to T\to1 (see [19, 15]):

Ep,q2=Hp(T,Hq(K,Z))⟹Hp+q(Mm,Z).E^2_{p,q}=H_p(T,H_q(K,\mathbb{Z}))\Longrightarrow H_{p+q}(M_m,\mathbb{Z}).

Since H2(K,Z)=H2(T,Z)=0H_2(K,\mathbb{Z})=H_2(T,\mathbb{Z})=0, the only potentially nonzero term in total degree two is E1,12=H1(T,K)E^2_{1,1}=H_1(T,K). Under the identification K=Z/2m+1ZK=\mathbb{Z}/2^{m+1}\mathbb{Z}, the action of the generator of TT is multiplication by σ=1+2m\sigma=1+2^m. The standard cyclic homology formula [6] gives

H1(T,K)=ker⁡(σ−1)(σ+1)K.H_1(T,K)=\frac{\ker(\sigma-1)}{(\sigma+1)K}.

Since σ−1=2m\sigma-1=2^m, multiplication by σ−1\sigma-1 on K=Z/2m+1ZK=\mathbb{Z}/2^{m+1}\mathbb{Z} has kernel 2K2K. On the other hand, 1+σ=2(1+2m−1)1+\sigma=2(1+2^{m-1}), and 1+2m−11+2^{m-1} is odd because m≥2m\ge2. Hence

(1+σ)K=2K,H1(T,K)=2K2K=0.(1+\sigma)K=2K,\quad H_1(T,K)=\frac{2K}{2K}=0.

As H2(K,Z)=H2(T,Z)=0H_2(K,\mathbb{Z})=H_2(T,\mathbb{Z})=0, all terms of total degree two on the E2E_2-page vanish. Thus H2(Mm,Z)=0H_2(M_m,\mathbb{Z})=0. Since Mmab≅C2×C2mM_m^{\mathrm{ab}}\cong C_2\times C_{2^m}, the universal coefficient theorem yields

dim⁡F2H2(Mm,F2)=2.\dim_{\mathbb{F}_2}H^2(M_m,\mathbb{F}_2)=2.

We next use the mod-2 cohomology spectral sequence; for the general theory of extension differentials see [16]. Here

E2p,q=Hp(T,Hq(K,F2))⟹Hp+q(Mm,F2).E_2^{p,q}=H^p(T,H^q(K,\mathbb{F}_2))\Longrightarrow H^{p+q}(M_m,\mathbb{F}_2).

The action of TT on H∗(K,F2)H^*(K,\mathbb{F}_2) is trivial, because σ\sigma is an odd-power automorphism of KK. Hence

E2∗,∗≅F2[X,yˉ,v](yˉ2),E_2^{*,*}\cong\frac{\mathbb{F}_2[X,\bar{y},v]}{(\bar{y}^2)},

where ∣X∣=(1,0)|X|=(1,0), ∣yˉ∣=(0,1)|\bar{y}|=(0,1), ∣v∣=(0,2)|v|=(0,2). The classes XX and yˉ\bar{y} are represented by the global characters xx and yy; in particular, they are permanent cycles. Their product XyˉX\bar{y} is nonzero in E21,1E_2^{1,1} and is also a permanent cycle: there are no incoming differentials, and every outgoing differential vanishes by multiplicativity. Thus

xy≠0in H2(Mm,F2).xy\ne0\quad\text{in }H^2(M_m,\mathbb{F}_2).

The classes x2x^2 and xyxy are linearly independent, since x2x^2 restricts nontrivially to TT whereas xyxy restricts to zero. They therefore form a basis of H2(Mm,F2)H^2(M_m,\mathbb{F}_2).

It follows that the class v∈E20,2v\in E_2^{0,2} must be killed by a differential. Its only possible outgoing differentials are

d2:E20,2⟶E22,1,d_2:E_2^{0,2}\longrightarrow E_2^{2,1},

and

d3:E30,2⟶E33,0.d_3 : E_3^{0,2} \longrightarrow E_3^{3,0}.

The second is impossible, since X3∈E23,0X^3 \in E_2^{3,0} survives to E∞E_\infty, because x3x^3 restricts nontrivially to TT. Consequently,

d2(v)=X2yˉ≠0,d_2(v) = X^2\bar{y} \ne0,

and

E∞2,1=0.E_\infty^{2,1} = 0.

The cohomology class x2yx^2y belongs to filtration at least two in H3(Mm,F2)H^3(M_m,\mathbb{F}_2), and the vanishing of E∞2,1E_\infty^{2,1} places it in filtration three. This last filtration step is spanned by x3x^3. Since x2yx^2y restricts to zero on TT, while x3x^3 does not, we conclude that x2y=0x^2y = 0.

We have therefore obtained a surjection

F2[x,y](y2,x2y)⟶Dec⁡(H∗(Mm,F2)).\frac{\mathbb{F}_2[x,y]}{(y^2,x^2y)} \longrightarrow\operatorname{Dec}(H^*(M_m,\mathbb{F}_2)).

The source has basis {xn:n≥0}∪{y,xy}\{x^n:n\ge0\}\cup\{y,xy\}. Its images are linearly independent; indeed, in degree one this follows from the independence of x,yx,y; in degree two from the independence of x2,xyx^2,xy; and in every degree at least three from the nonvanishing of xnx^n. The surjection is therefore an isomorphism. □

The modular boundary shows how the power relation can change the degree-one algebra without changing the dimension of H1H^1. The classes xnx^n persist in every degree, in contrast with the finite-dimensional algebra that occurs when the carry is divisible by 4. We now turn to that regular branch and derive its relations directly from the normal-form multiplication law.

The degree-one cohomology dichotomy

We first consider the regular branch, where the power-relation carry is divisible by 4. In this case, the degree-one relations can be obtained directly from the Bockstein, using the normal-form multiplication law established in Section 2. The exceptional modular branch has already been determined in Theorem 2.5, and the two calculations will be combined in the main dichotomy below.

Assume that (2.1) has unique normal forms, r≡−1(mod4)r\equiv-1\pmod{4}, and either c=0c=0 or α≥3\alpha\ge3. Proposition 2.1 gives

Gab≅C2×C2m.G^{\mathrm{ab}}\cong C_2\times C_{2m}.

Define

x(aibj)=i(mod2),y(aibj)=j(mod2).x(a^i b^j)=i\pmod{2},\qquad y(a^i b^j)=j\pmod{2}.

Lemma 2.3 shows that xx and yy are homomorphisms. Their values on aa and bb show that they are independent, and Proposition 2.1 shows that they form a basis of H1(G,F2)H^1(G,\mathbb{F}_2).

We use the Bockstein associated to

0⟶F2→⋅2Z/4Z⟶F2⟶0,0\longrightarrow\mathbb{F}_2 \xrightarrow{\cdot2} \mathbb{Z}/4\mathbb{Z}\longrightarrow\mathbb{F}_2\longrightarrow0,

for which β(u)=u2\beta(u)=u^2 on H1(−,F2)H^1(-,\mathbb{F}_2).

Lemma 3.1. The character yy lifts to Z/4Z\mathbb{Z}/4\mathbb{Z}, and hence y2=β(y)=0y^2=\beta(y)=0.

Proof. Set y~(a)=0\widetilde{y}(a)=0 and y~(b)=1(mod4)\widetilde{y}(b)=1\pmod{4}. Since m≥2m\ge2, the left side of b2m=acb^{2^m}=a^c maps to 0 modulo 4, and the right side maps to 0 because y~(a)=0\widetilde{y}(a)=0. The remaining relations are automatic. □

Lemma 3.2. The character xx does not lift to Z/4Z\mathbb{Z}/4\mathbb{Z}. Consequently

x2=β(x)≠0.x^2=\beta(x)\ne0.

Proof. If x~\tilde{x} were a lift, then x~(a)\tilde{x}(a) would be odd. Applying x~\tilde{x} to b−1ab=arb^{-1}ab=a^r gives

x~(a)=rx~(a)(mod4).\tilde{x}(a)=r\tilde{x}(a)\pmod{4}.

Since r≡−1(mod4)r\equiv-1\pmod{4}, this would imply 2x~(a)=0(mod4)2\tilde{x}(a)=0\pmod{4}, impossible for an odd element. □

The failure of xx to lift establishes that its Bockstein is nonzero, but does not yet identify the resulting degree-two class. To determine it, we use the multiplication law modulo 4 to compute an explicit representative of β(x)\beta(x).

Lemma 3.3. One has β(x)=xy\beta(x)=xy.

Proof. Use the set-theoretic lift x~(aibj)=i(mod4)\tilde{x}(a^ib^j)=i\pmod{4}. For g=aibjg=a^ib^j and h=aubvh=a^ub^v, write gh=awbtgh=a^wb^t in normal form. The connecting cocycle is

(g,h)⟼x~(g)+x~(h)−x~(gh)2(mod2).(g,h)\longmapsto\frac{\tilde{x}(g)+\tilde{x}(h)-\tilde{x}(gh)}{2}\pmod{2}.

By Lemma 2.3,

w≡i+(−1)ju(mod4).w\equiv i+(-1)^ju\pmod{4}.

Hence the numerator is 0 modulo 4 when jj is even and 2u2u modulo 4 when jj is odd. The resulting cocycle is

(j mod 2)(u mod 2)=y(g)x(h),(j\bmod2)(u\bmod2)=y(g)x(h),

which represents y⌣xy\smile x. Mod-2 group cohomology is graded-commutative, and in characteristic 2 the degree-one sign disappears, so

y⌣x=x⌣y=xy.y\smile x=x\smile y=xy.

Therefore β(x)=xy\beta(x)=xy. □

Theorem 3.4. Let GG have a metacyclic normal-form presentation (1), with α,m≥2\alpha,m\ge2 and r≡−1(mod4)r\equiv-1\pmod{4}. Then

c=0orc=2α−1,c=0\quad\text{or}\quad c=2^{\alpha-1},

and the degree-one-generated cohomology algebra Dec⁡(H∗(G,F2))\operatorname{Dec}(H^*(G,\mathbb{F}_2)) is determined by the following dichotomy.

(1) If c≡0(mod4)c\equiv0\pmod{4}, equivalently if c=0c=0 or α≥3\alpha\ge3, then

Dec⁡(H∗(G,F2))≅F2[x,y](y2,x2+xy).\operatorname{Dec}(H^*(G,\mathbb{F}_2))\cong\frac{\mathbb{F}_2[x,y]}{(y^2,x^2+xy)}.

This algebra is four-dimensional, with F2\mathbb{F}_2-basis {1,x,y,xy}\{1,x,y,xy\}.

(2) If c≢0(mod4)c\not\equiv0\pmod{4}, then necessarily α=2\alpha=2 and c=2c=2. In this case GG is the modular group MmM_m, and

Dec⁡(H∗(G,F2))≅F2[x,y](y2,x2y).\operatorname{Dec}(H^*(G,\mathbb{F}_2))\cong\frac{\mathbb{F}_2[x,y]}{(y^2,x^2y)}.

This algebra is infinite-dimensional, with F2\mathbb{F}_2-basis {xn:n≥0}∪{y,xy}\{x^n:n\ge0\}\cup\{y,xy\} and Hilbert series

11−t+t+t2.\frac{1}{1-t}+t+t^2.

In particular, the degree-one-generated cohomology algebra depends only on whether the power-relation parameter cc vanishes modulo 4, and is independent of the remaining admissible presentation parameters.

Proof. Proposition 2.1 shows that presentation compatibility forces c=0c=0 or c=2α−1c=2^{\alpha-1} and gives

Gab≅C2×C2m.G^{\mathrm{ab}}\cong C_2\times C_{2^m}.

Consequently, the presentation-adapted characters x,yx,y form a basis of H1(G,F2)H^1(G,\mathbb{F}_2).

Suppose first that c≡0(mod4)c \equiv0 \pmod{4}. By the compatibility condition, this occurs precisely when c=0c=0 or α≥3\alpha\ge3. Lemmas 3.1, 3.2, and 3.3 give y2=0y^2=0, and x2=xy≠0x^2=xy\ne0. These relations induce a surjective homomorphism of graded F2\mathbb{F}_2-algebras

F2[x,y](y2,x2+xy)⟶Dec⁡(H∗(G,F2)).\frac{\mathbb{F}_2[x,y]}{(y^2,x^2+xy)}\longrightarrow\operatorname{Dec}(H^*(G,\mathbb{F}_2)).

In the source algebra, every monomial of degree at least 3 vanishes, since x3=x(xy)=x2y=xy2=0x^3=x(xy)=x^2y=xy^2=0. Moreover, the degree-two component is spanned by xyxy. Since xx and yy are linearly independent in H1(G,F2)H^1(G,\mathbb{F}_2) and xy≠0xy\ne0, the images of 1,x,y,xy1,x,y,xy are linearly independent. The homomorphism is therefore an isomorphism, giving the first assertion.

Now suppose that c≢0(mod4)c\not\equiv0 \pmod{4}. Since c=0c=0 or c=2α−1c=2^{\alpha-1}, this forces α=2\alpha=2 and c=2c=2. By Proposition 2.4, the corresponding group is the modular group MmM_m. The exact calculation in Theorem 2.5 gives

Dec⁡(H∗(Mm,F2))≅F2[x,y](y2,x2y).\operatorname{Dec}(H^*(M_m,\mathbb{F}_2))\cong\frac{\mathbb{F}_2[x,y]}{(y^2,x^2y)}.

The defining relations show that this algebra has basis {xn:n≥0}∪{y,xy}\{x^n:n\ge0\}\cup\{y,xy\}. In particular, it is infinite-dimensional, and its Hilbert series is

∑n≥0tn+t+t2=11−t+t+t2.\sum_{n\ge0}t^n+t+t^2=\frac{1}{1-t}+t+t^2.

The two cases exhaust all compatible values of cc. Their algebra presentations depend only on the residue of cc modulo 4, establishing the asserted dichotomy. □

The uniformity in Theorem 3.4 suggests a common group-theoretic source for the degree-one algebra. Since the relevant multiplication law depends only on the presentation modulo 4, we retain the fourth-power residues of the two generators. The resulting quotient makes the cohomological identification compatible with inflation.

Proposition 3.5. Let GG have a metacyclic normal-form presentation (1), with α,m≥2\alpha,m\ge2, r≡−1(mod4)r\equiv-1\pmod{4}, and c≡0(mod4)c\equiv0\pmod{4}. Let

H=⟨A,B∣A4=B4=1, B−1AB=A−1⟩≅C4⋊−1C4.H=\langle A,B\mid A^4=B^4=1,\ B^{-1}AB=A^{-1}\rangle\cong C_4\rtimes_{-1}C_4.

Then the assignments

a⟼A,b⟼Ba\longmapsto A,\qquad b\longmapsto B

define a surjective homomorphism

ρG:G↠H\rho_G:G\twoheadrightarrow H

with

ker⁡(ρG)=⟨ ⁣⟨a4,b4⟩ ⁣⟩G.\ker(\rho_G)=\langle\!\langle a^4,b^4\rangle\!\rangle_G.

Moreover, the induced inflation homomorphism restricts to an isomorphism of graded F2\mathbb{F}_2-algebras

ρG∗:Dec⁡(H∗(H,F2))→∼Dec⁡(H∗(G,F2)).\rho_G^*:\operatorname{Dec}(H^*(H,\mathbb{F}_2))\xrightarrow{\sim}\operatorname{Dec}(H^*(G,\mathbb{F}_2)).

Thus every group in the regular branch has the same degree-one-generated cohomology algebra as the fixed group HH of order 16, and this identification is induced by the presentation-defined quotient map.

Proof. We first verify that the proposed assignments respect the defining relations of GG. Since α≥2\alpha\ge2, the relation a2α=1a^{2^\alpha}=1 maps to

A2α=1.A^{2^\alpha}=1.

Likewise, m≥2m\ge2 implies

B2m=1.B^{2^m}=1.

Because c≡0(mod4)c \equiv0 \pmod{4}, we also have Ac=1A^c = 1, so the power relation b2m=acb^{2^m} = a^c is preserved. Finally, since r≡−1(mod4)r \equiv-1 \pmod{4},

B−1AB=A−1=Ar.B^{-1}AB = A^{-1} = A^r.

Hence the assignments extend to a homomorphism ρG:G→H\rho_G : G \to H. Its image contains both AA and BB, so ρG\rho_G is surjective.

Let

N=⟨⟨a4,b4⟩⟩G.N = \langle\langle a^4,b^4\rangle\rangle_G.

Clearly N⊆ker⁡(ρG)N \subseteq\ker(\rho_G). In the quotient G/NG/N, the images of aa and bb have orders dividing 44. The relation b−1ab=arb^{-1}ab = a^r becomes b−1ab=a−1b^{-1}ab = a^{-1}, since r≡−1(mod4)r \equiv-1 \pmod{4}. Moreover, the power relation becomes trivial because c≡0(mod4)c \equiv0 \pmod{4}. Consequently, G/NG/N is generated by the images of aa and bb subject to the relations defining HH. These relations give a surjective homomorphism H→G/NH \to G/N, while ρG\rho_G induces a surjective homomorphism G/N→HG/N \to H. Since HH is finite, both maps are isomorphisms. Therefore

G/N≅H,ker⁡(ρG)=N.G/N \cong H,\qquad\ker(\rho_G) = N.

We now consider the induced map on cohomology. Let xH,yH∈H1(H,F2)x_H,y_H \in H^1(H,\mathbb{F}_2) be the characters determined by

xH(A)=1,xH(B)=0,yH(A)=0,yH(B)=1.x_H(A) = 1,\qquad x_H(B) = 0,\qquad y_H(A) = 0,\qquad y_H(B) = 1.

Let xG,yG∈H1(G,F2)x_G,y_G \in H^1(G,\mathbb{F}_2) be the corresponding presentation-adapted characters. By construction,

ρG∗(xH)=xG,ρG∗(yH)=yG.\rho_G^*(x_H) = x_G,\qquad\rho_G^*(y_H) = y_G.

Since both pairs form bases of their respective first cohomology groups, ρG∗\rho_G^* induces an isomorphism

H1(H,F2)→∼H1(G,F2).H^1(H,\mathbb{F}_2) \xrightarrow{\sim} H^1(G,\mathbb{F}_2).

By Theorem 3.4, both HH and GG belong to the regular branch, and hence

Dec⁡(H∗(H,F2))≅F2[xH,yH](yH2,xH2+xHyH).\operatorname{Dec}(H^*(H,\mathbb{F}_2)) \cong\frac{\mathbb{F}_2[x_H,y_H]}{(y_H^2,x_H^2+x_Hy_H)}.

and

Dec⁡(H∗(G,F2))≅F2[xG,yG](yG2,xG2+xGyG).\operatorname{Dec}(H^*(G,\mathbb{F}_2)) \cong\frac{\mathbb{F}_2[x_G,y_G]}{(y_G^2,x_G^2+x_Gy_G)}.

In particular, both algebras are four-dimensional, with respective bases

{1,xH,yH,xHyH}and{1,xG,yG,xGyG}.\{1,x_H,y_H,x_Hy_H\}\quad\text{and}\quad\{1,x_G,y_G,x_Gy_G\}.

Naturality of the cup product gives

ρG∗(xHyH)=xGyG.\rho_G^*(x_Hy_H) = x_Gy_G.

Thus ρG∗\rho_G^* maps the displayed basis of the first algebra onto the displayed basis of the second. It follows that its restriction to the degree-one-generated subalgebras is an isomorphism, as claimed.

Corollary 3.6. For every metacyclic-nonmodular 2-group in Types 1–4 of Aaboun–Zekhnini,

Dec⁡(H∗(G,F2))≅F2[x,y]/(y2,x2+xy).\operatorname{Dec}(H^*(G,\mathbb{F}_2)) \cong\mathbb{F}_2[x,y]/(y^2,x^2+xy).

Thus the degree-one-generated algebra is independent of the refined type and all admissible parameters.

Proof. Each of the four source presentations has r≡−1(mod4)r \equiv-1 \pmod{4}, and none is the exceptional α=2\alpha= 2, c=2c = 2 boundary. Theorem 3.4 applies. □\square

Remark 3.7. Theorem 3.4 isolates the low-bit mechanism behind the degree-one relations across the refined metacyclic types. The full cohomology calculations of Rusin and Huebschmann describe the corresponding ambient rings. The normal-form calculation used here makes the uniformity transparent: the congruence r≡−1(mod4)r \equiv-1 \pmod{4} fixes the conjugation contribution, while the compatibility condition on cc controls the power-relation carry modulo 44.

Corollary 3.8. In the regular branch,

β(y)=0,β(x)=xy,β(x+y)=xy.\beta(y)=0,\quad\beta(x)=xy,\quad\beta(x+y)=xy.

Among the three nonzero elements of H1(G,F2)H^1(G,\mathbb{F}_2), exactly one has vanishing Bockstein.

Koszulity and Quadratic Duality

The explicit algebra presentations allow us to compare homological properties in the two branches. Related enhanced-Koszul questions have been studied for right-angled Artin groups and in Galois cohomology [7, 21]. The regular algebra is defined by quadratic relations, while the modular boundary has a minimal cubic relation. We first establish the enhanced Koszul properties of the regular algebra and then show that the same low-bit distinction determines Koszulity.

We use the colon criterion for universal Koszulity as developed by Conca [10] and discussed by Conca–De Negri–Rossi [11]. For strong Koszulity we use the notion introduced by Herzog–Hibi–Restuccia [14], also used in the enhanced-Koszul literature [21].

Proposition 4.1. The algebra F2[x,y]/(y2,x2+xy)\mathbb{F}_2[x,y]/(y^2,x^2+xy) is PBW and therefore Koszul. It is also universally Koszul and strongly Koszul.

Proof. With degree-lexicographic order x>yx>y, the two quadrics x2+xyx^2+xy, y2y^2 form a Gröbner basis: their only SS-polynomial is xy3xy^3, which reduces to zero. Thus the algebra is PBW.

For universal Koszulity, set z=x+yz=x+y. The nonzero proper ideals generated by degree-one elements are (x)(x), (y)(y), (z)(z), m=(x,y)\mathfrak{m}=(x,y). Using

x2=xy,y2=0,z2=xy,xz=0,yz=xy,x^2=xy,\quad y^2=0,\quad z^2=xy,\quad xz=0,\quad yz=xy,

one obtains

(0:x)=(z),(0:y)=(y),(0:z)=(x),(0:x)=(z),\quad(0:y)=(y),\quad(0:z)=(x),

and, whenever ℓ\ell is a nonzero linear form outside the degree-one part of one of (x),(y),(z)(x),(y),(z),

((x):ℓ)=((y):ℓ)=((z):ℓ)=m((x):\ell)=((y):\ell)=((z):\ell)=\mathfrak{m}

in the applicable cases. Hence the standard colon criterion gives universal Koszulity.

For strong Koszulity use the basis {x,z}\{x,z\}. Then

F2[x,y]/(y2,x2+xy)≅F2[x,z]/(xz,x2+z2),\mathbb{F}_2[x,y]/(y^2,x^2+xy)\cong\mathbb{F}_2[x,z]/(xz,x^2+z^2),

and

(0:x)=(z),(0:z)=(x),(x):z=(x,z),(z):x=(x,z).(0:x)=(z),\quad(0:z)=(x),\quad(x):z=(x,z),\quad(z):x=(x,z).

Every required colon is generated by a subset of the same basis. □

Corollary 4.2. Let GG have a metacyclic normal-form presentation (1), with α,m≥2\alpha,m\geq2 and r≡−1(mod4)r\equiv-1\pmod{4}. Then the degree-one-generated cohomology algebra Dec⁡(H∗(G,F2))\operatorname{Dec}(H^*(G,\mathbb{F}_2)) is Koszul if and only if c≡0(mod4)c\equiv0\pmod{4}.

More precisely, if c≡0(mod4)c\equiv0\pmod{4}, then Dec⁡(H∗(G,F2))\operatorname{Dec}(H^*(G,\mathbb{F}_2)) is PBW, Koszul, universally Koszul, and strongly Koszul. If c≢0(mod4)c\not\equiv0\pmod{4}, then necessarily α=2\alpha=2 and c=2c=2, and the corresponding algebra is not quadratic and hence is not Koszul. In particular, all four nonmodular metacyclic types of Abou-Zekhnini lie in the Koszul branch.

Proof. By Theorem 3.4, the degree-one-generated cohomology algebra has precisely two possible presentations, depending on whether cc vanishes modulo 4.

Suppose first that c≡0c \equiv0 (mod 4). Then

Dec⁡(H∗(G,F2))≅F2[x,y]/(y2,x2+xy).\operatorname{Dec}(H^{\ast}(G,\mathbb{F}_2)) \cong\mathbb{F}_2[x,y]/(y^2,x^2+xy).

The preceding proposition shows that this algebra is PBW, Koszul, universally Koszul, and strongly Koszul.

Now suppose that c≠0c \ne0 (mod 4). Presentation compatibility forces α=2\alpha=2 and c=2c=2, and Theorem 3.4 gives

S:=Dec⁡(H∗(G,F2))≅F2[x,y]/(y2,x2y).S := \operatorname{Dec}(H^{\ast}(G,\mathbb{F}_2)) \cong\mathbb{F}_2[x,y]/(y^2,x^2y).

The relation x2y=0x^2y=0 is a minimal cubic relation. Indeed, the only quadratic relation is y2=0y^2=0, and x2yx^2y remains nonzero in F2[x,y]/(y2)\mathbb{F}_2[x,y]/(y^2). Consequently, the defining ideal of SS cannot be generated by quadratic relations, so SS is not quadratic. Since every connected standard graded Koszul algebra is quadratic (see [25]), SS is not Koszul.

The failure of universal Koszulity can also be seen directly from the colon criterion. In SS one has (0:x)=(xy)(0:x)=(xy). Since xy≠0xy \ne0 and (xy)(xy) contains no nonzero element of degree one, this annihilator is not generated by linear forms. Thus SS fails the universal Koszul colon criterion.

Finally, the PBW, universal Koszul, and strong Koszul properties each imply Koszulity. Hence none of these enhanced properties holds in the exceptional branch. This establishes the dichotomy.

Koszulity makes the quadratic dual a natural object to consider, since it describes the Yoneda algebra of the degree-one core; see [3, 24] for general treatments of quadratic duality. In the regular branch, the dual admits a familiar noncommutative presentation, connecting the cohomology calculation with the characteristic-two Jordan plane [18].

Proposition 4.3. Let

R=F2[x,y]/(y2,x2+xy),∣x∣=∣y∣=1.R=\mathbb{F}_2[x,y]/(y^2,x^2+xy),\quad|x|=|y|=1.

Then its quadratic dual is

R!≅F2⟨U,V⟩(VU+UV+U2),∣U∣=∣V∣=1.R^! \cong\frac{\mathbb{F}_2\langle U,V\rangle}{(VU+UV+U^2)},\quad|U|=|V|=1.

In particular, R!R^! is the Jordan plane in characteristic two. It has an F2\mathbb{F}_2-basis {UiVj:i,j≥0}\{U^iV^j:i,j\ge0\} and Hilbert series

HR!(t)=1(1−t)2.H_{R^!}(t)=\frac{1}{(1-t)^2}.

Since RR is Koszul, its Yoneda algebra satisfies

Ext⁡R∗(F2,F2)≅R!,dim⁡F2Ext⁡Rn(F2,F2)=n+1 (n≥0).\operatorname{Ext}^{\ast}_{R}(\mathbb{F}_2,\mathbb{F}_2)\cong R^!,\quad\dim_{\mathbb{F}_2}\operatorname{Ext}^{n}_{R}(\mathbb{F}_2,\mathbb{F}_2)=n+1\ (n\ge0).

Consequently, this description applies to the degree-one-generated cohomology algebra of every group in the regular branch of Theorem 3.4.

Proof. Let W=F2x⊕F2yW=\mathbb{F}_2x\oplus\mathbb{F}_2y. To compute the quadratic dual, write RR as a quotient of the tensor algebra T(W)T(W). Since RR is commutative, its space of quadratic relations is

J=⟨y⊗y, x⊗x+x⊗y, x⊗y+y⊗x⟩⊆W⊗W.J=\langle y\otimes y,\ x\otimes x+x\otimes y,\ x\otimes y+y\otimes x\rangle\subseteq W\otimes W.

Let U,VU,V be the dual basis of W∗W^{\ast}. An element

λU⊗U+μU⊗V+νV⊗U+κV⊗V\lambda U\otimes U+\mu U\otimes V+\nu V\otimes U+\kappa V\otimes V

belongs to the orthogonal complement J⊥J^{\perp} precisely when

κ=0,λ+μ=0,μ+ν=0.\kappa=0,\quad\lambda+\mu=0,\quad\mu+\nu=0.

Thus J⊥J^\perp is generated by U⊗U+U⊗V+V⊗UU \otimes U + U \otimes V + V \otimes U, and the definition of the quadratic dual gives

R′=T(W∗)/(J⊥)≅F2⟨U,V⟩/(VU+UV+U2).R' = T(W^*)/(J^\perp) \cong\mathbb{F}_2\langle U,V\rangle/(VU+UV+U^2).

Equivalently, its defining relation is VU=UV+U2VU=UV+U^2. Let δ\delta be the derivation of F2[U]\mathbb{F}_2[U] determined by δ(U)=U2\delta(U)=U^2. Then the same relation presents the Ore extension

R′≅F2[U][V;δ].R' \cong\mathbb{F}_2[U][V;\delta].

The Ore-extension normal form, equivalently obtained by the reduction VU↦UV+U2VU\mapsto UV+U^2, gives the F2\mathbb{F}_2-basis {UiVj:i,j≥0}\{U^iV^j:i,j\geq0\} (see [20, 5]). Hence its degree-nn component has dimension n+1n+1, and

HR′(t)=∑n≥0(n+1)tn=1(1−t)2.H_{R'}(t)=\sum_{n\geq0}(n+1)t^n=\frac{1}{(1-t)^2}.

On the other hand, the relations of RR give the basis {1,x,y,xy}\{1,x,y,xy\}, so HR(t)=(1+t)2H_R(t)=(1+t)^2 and

HR(t)HR′(−t)=1.H_R(t)H_{R'}(-t)=1.

By Proposition 4.1, RR is Koszul. The standard identification of the Yoneda algebra of a Koszul algebra with its quadratic dual therefore gives Ext⁡R∗(F2,F2)≅R′\operatorname{Ext}^{*}_{R}(\mathbb{F}_2,\mathbb{F}_2)\cong R'. Finally, Theorem 3.4 identifies the degree-one-generated cohomology algebra of every group in the regular branch with RR, so the same description applies throughout that family. □

Cohomology and inflation in the inversion family

The group C4⋊−1C4C_4\rtimes_{-1}C_4 appearing in Proposition 3.5 is also the first level of a natural inverse system of metacyclic groups. The degree-one algebra is unchanged throughout this system, while higher-degree classes exhibit different behavior under inflation. We determine the full finite-level cohomology rings and their inflation maps in order to identify the classes that persist in the continuous limit.

Let

Gm=⟨a,b∣a4=1, b2m=1, b−1ab=a−1⟩=C4⋊−1C2m,m≥2,G_m=\langle a,b\mid a^4=1,\ b^{2^m}=1,\ b^{-1}ab=a^{-1}\rangle=C_4\rtimes_{-1}C_{2^m},\qquad m\geq2,

with quotient maps

πm:Gm+1↠Gm.\pi_m:G_{m+1}\twoheadrightarrow G_m.

Then

Γ:=lim←⁡mGm≅C4⋊−1Z2.\Gamma:=\varprojlim_m G_m\cong C_4\rtimes_{-1}\mathbb{Z}_2.

Finite-level cohomology and inflation

The mod-2 cohomology of G2=C4⋊−1C4G_2=C_4\rtimes_{-1}C_4 was computed by David Benson in Appendix A of [8], in the course of the proof of Proposition A.1. In his notation, the group is

H2=⟨g,h∣g4=h4=1, gh=h−1g⟩,H_2=\langle g,h\mid g^4=h^4=1,\ gh=h^{-1}g\rangle,

which agrees with our G2G_2 under g=bg=b and h=ah=a.

Benson’s calculation gives, after renaming the generators to agree with our degree-one characters,

H∗(G2,F2)≅F2[x,y,u,w](y2,x2+xy),∣x∣=∣y∣=1,∣u∣=∣w∣=2.H^*(G_2,\mathbb{F}_2)\cong\frac{\mathbb{F}_2[x,y,u,w]}{(y^2,x^2+xy)},\qquad|x|=|y|=1,\quad|u|=|w|=2.

Here uu is inflated from the degree-two generator of the cohomology of the cyclic quotient G2/⟨a⟩≅C4G_2/\langle a\rangle\cong C_4.

The second degree-two generator ww arises in Benson’s calculation from the central extension

1⟶⟨a2⟩⟶G2⟶C4×C2⟶1.1\longrightarrow\langle a^2\rangle\longrightarrow G_2\longrightarrow C_4\times C_2\longrightarrow1.

Writing H∗(⟨a2⟩,F2)=F2[ζ]H^*(\langle a^2\rangle,\mathbb{F}_2)=\mathbb{F}_2[\zeta], with ∣ζ∣=1|\zeta|=1, the associated Lyndon–Hochschild–Serre spectral sequence has

d2(ζ)=x2+xy.d_2(\zeta)=x^2+xy.

Benson shows that ζ2\zeta^2 survives the spectral sequence and gives rise to the degree-two generator ww. In particular,

res⁡⟨a2⟩(w)=ζ2≠0.\operatorname{res}_{\langle a^2\rangle}(w)=\zeta^2\ne0.

Since ⟨a2⟩⊆⟨a⟩\langle a^2\rangle\subseteq\langle a\rangle, it follows that

res⁡⟨a⟩(w)≠0.\operatorname{res}_{\langle a\rangle}(w)\ne0.

We take ωA,2=w\omega_{A,2}=w. For each m≥2m\ge2, let

qm:Gm⟶G2q_m:G_m\longrightarrow G_2

be the quotient map fixing aa and reducing the exponent of bb modulo 44, and define

ωA,m:=qm∗(ωA,2)∈H2(Gm,F2).\omega_{A,m}:=q_m^*(\omega_{A,2})\in H^2(G_m,\mathbb{F}_2).

Because qmq_m restricts to the identity on ⟨a⟩≅C4\langle a\rangle\cong C_4, naturality of restriction gives

res⁡⟨a⟩(ωA,m)=res⁡⟨a⟩(ωA,2)≠0.\operatorname{res}_{\langle a\rangle}(\omega_{A,m})=\operatorname{res}_{\langle a\rangle}(\omega_{A,2})\ne0.

Thus the degree-two classes ωA,m\omega_{A,m} are nonzero at every finite level.

Let

ωB,m∈H2(Gm,F2)\omega_{B,m}\in H^2(G_m,\mathbb{F}_2)

be the inflation of the standard degree-two class of C2mC_{2^m}. Since the quotient Gm⟶C2mG_m\longrightarrow C_{2^m} splits, this class is nonzero.

Lemma 5.1. The inflation map

H2(C2m,F2)⟶H2(C2m+1,F2)H^2(C_{2^m},\mathbb{F}_2)\longrightarrow H^2(C_{2^{m+1}},\mathbb{F}_2)

is zero for every m≥2m\ge2.

Proof. For

1⟶C2⟶C2m+1⟶C2m⟶1,1\longrightarrow C_2\longrightarrow C_{2^{m+1}}\longrightarrow C_{2^m}\longrightarrow1,

the first inflation map in the five-term sequence is an isomorphism on H1H^1, because the parity character pulls back to the parity character. The following restriction map to H1(C2,F2)H^1(C_2,\mathbb{F}_2) is therefore zero, so the transgression

H1(C2,F2)⟶H2(C2m,F2)H^1(C_2,\mathbb{F}_2)\longrightarrow H^2(C_{2^m},\mathbb{F}_2)

is injective. Both groups are one-dimensional, hence the transgression is an isomorphism and the next inflation map in degree 22 is zero. □\square

The preceding lemma determines what happens to the quotient-side periodicity class, while the construction of ωA,m\omega_{A,m} gives a compatible class restricting nontrivially to C4C_4. To describe inflation beyond degree 22, we must identify these classes within the full cohomology ring. The following proposition provides this description and determines the kernel and image of inflation in every degree.

Proposition 5.2. For m≥2m\ge2, write xm,ym∈H1(Gm,F2)x_m,y_m\in H^1(G_m,\mathbb{F}_2) for the presentation-adapted characters and set

um=ωB,m,wm=ωA,m.u_m=\omega_{B,m},\qquad w_m=\omega_{A,m}.

Then, with ∣xm∣=∣ym∣=1|x_m|=|y_m|=1 and ∣um∣=∣wm∣=2|u_m|=|w_m|=2,

H∗(Gm,F2)≅F2[xm,ym,um,wm](ym2,xm2+xmym).H^*(G_m,\mathbb{F}_2)\cong\frac{\mathbb{F}_2[x_m,y_m,u_m,w_m]}{(y_m^2,x_m^2+x_my_m)}.

Under the inflation homomorphism πm∗:H∗(Gm,F2)⟶H∗(Gm+1,F2)\pi_m^*:H^*(G_m,\mathbb{F}_2)\longrightarrow H^*(G_{m+1},\mathbb{F}_2), the generators satisfy

πm∗(xm)=xm+1,πm∗(ym)=ym+1,πm∗(um)=0,πm∗(wm)=wm+1.\pi_m^*(x_m)=x_{m+1},\qquad\pi_m^*(y_m)=y_{m+1},\qquad\pi_m^*(u_m)=0,\qquad\pi_m^*(w_m)=w_{m+1}.

Consequently,

ker⁡(πm∗)=(um)(2)\ker(\pi_m^*)=(u_m) \tag*{(2)}

and

im⁡(πm∗)=F2[xm+1,ym+1,wm+1](ym+12,xm+12+xm+1ym+1)\operatorname{im}(\pi_m^*)=\frac{\mathbb{F}_2[x_{m+1},y_{m+1},w_{m+1}]}{(y_{m+1}^2,x_{m+1}^2+x_{m+1}y_{m+1})}

In particular, dim⁡F2Hn(Gm,F2)=n+1\dim_{\mathbb{F}_2} H^n(G_m,\mathbb{F}_2)=n+1 for every n≥0n\geq0. The rank of πm∗\pi_m^* is 11 in degree 00 and 22 in each positive degree, and its kernel has dimension n−1n-1 in degree n≥2n\geq2.

Proof. Let A=⟨a⟩≅C4A=\langle a\rangle\cong C_4 and Qm=Gm/A=⟨bˉ⟩≅C2mQ_m=G_m/A=\langle\bar{b}\rangle\cong C_{2^m}, where bˉ=bA\bar{b}=bA. Since the quotient map has the section bˉ↦b\bar{b}\mapsto b, we have the split extension

1⟶A⟶Gm⟶Qm⟶11\longrightarrow A\longrightarrow G_m\longrightarrow Q_m\longrightarrow1

We use its Lyndon–Hochschild–Serre spectral sequence to compare the cohomology of GmG_m with that of the two cyclic groups; for resolutions adapted to group extensions, see [29]. The generator bˉ\bar{b} acts on AA by inversion, which acts trivially on both the degree-one character and the degree-two periodicity generator of H∗(A,F2)H^*(A,\mathbb{F}_2). Thus

E2p,q=Hp(Qm,Hq(A,F2))⟹Hp+q(Gm,F2)E_2^{p,q}=H^p(Q_m,H^q(A,\mathbb{F}_2))\Longrightarrow H^{p+q}(G_m,\mathbb{F}_2)

and

E2∗,∗≅F2[xˉ,yˉ,uˉ,wˉ](xˉ2,yˉ2)E_2^{*,*}\cong\frac{\mathbb{F}_2[\bar{x},\bar{y},\bar{u},\bar{w}]}{(\bar{x}^2,\bar{y}^2)}

where

∣xˉ∣=(0,1),∣wˉ∣=(0,2),∣yˉ∣=(1,0),∣uˉ∣=(2,0).|\bar{x}|=(0,1),\qquad|\bar{w}|=(0,2),\qquad|\bar{y}|=(1,0),\qquad|\bar{u}|=(2,0).

The classes xˉ,wˉ\bar{x},\bar{w} come from H∗(A,F2)H^*(A,\mathbb{F}_2), while yˉ,uˉ\bar{y},\bar{u} come from H∗(Qm,F2)H^*(Q_m,\mathbb{F}_2). To determine the differentials, we identify these four generators with classes already defined on GmG_m. The classes ymy_m and umu_m are inflated from QmQ_m and represent yˉ\bar{y} and uˉ\bar{u}. The restriction of xmx_m to AA is the nonzero degree-one character, while the restriction of wmw_m is the nonzero degree-two periodicity class. Thus xm,wmx_m,w_m represent xˉ,wˉ\bar{x},\bar{w} under the edge maps. All four generators are permanent cycles because they are represented by classes in H∗(Gm,F2)H^*(G_m,\mathbb{F}_2). Since they generate the E2E_2-page multiplicatively, every differential vanishes and the spectral sequence collapses at E2E_2.

The classes xm,ym,um,wmx_m,y_m,u_m,w_m generate H∗(Gm,F2)H^*(G_m,\mathbb{F}_2), since their images generate the associated graded algebra. By Theorem 3.4,

ym2=0,xm2=xmym.y_m^2=0,\qquad x_m^2=x_my_m.

These relations give a surjective homomorphism

F2[xm,ym,um,wm](ym2,xm2+xmym)⟶H∗(Gm,F2).\frac{\mathbb{F}_2[x_m,y_m,u_m,w_m]}{(y_m^2,x_m^2+x_my_m)}\longrightarrow H^*(G_m,\mathbb{F}_2).

The source has Hilbert series

(1+t)2(1−t2)2=1(1−t)2.\frac{(1+t)^2}{(1-t^2)^2}=\frac{1}{(1-t)^2}.

The collapse of the spectral sequence gives the same Hilbert series for H∗(Gm,F2)H^*(G_m,\mathbb{F}_2), so the surjection is an isomorphism in every degree.

The maps πm\pi_m fix aa and reduce the exponent of bb modulo 2m2^m. Consequently, the presentation-adapted characters satisfy

πm∗(xm)=xm+1,πm∗(ym)=ym+1.\pi_m^*(x_m)=x_{m+1},\qquad\pi_m^*(y_m)=y_{m+1}.

The identities qm+1=qm∘πmq_{m+1}=q_m\circ\pi_m and wm=qm∗(ωA,2)w_m=q_m^*(\omega_{A,2}) give

πm∗(wm)=wm+1.\pi_m^*(w_m)=w_{m+1}.

Likewise, umu_m is inflated from C2mC_{2^m}, and naturality together with Lemma 5.1 gives

πm∗(um)=0.\pi_m^*(u_m)=0.

Under the ring presentations just established, πm∗\pi_m^* is therefore the homomorphism obtained by setting um=0u_m=0 and identifying the other three generators with their counterparts at level m+1m+1. It follows that its kernel is (um)(u_m) and its image is the displayed subalgebra. The latter has Hilbert series

(1+t)21−t2=1+t1−t.\frac{(1+t)^2}{1-t^2}=\frac{1+t}{1-t}.

so its degree-zero component has dimension 1 and every positive-degree component has dimension 2. Comparison with 1/(1−t)21/(1-t)^2 gives the assertions about ranks and kernel dimensions. □

The continuous ring as the resulting limit

The finite-level calculation separates the two periodicity directions in every cohomological degree. The quotient-side class umu_m disappears under the next inflation map, while the subalgebra generated by xm,ym,wmx_m,y_m,w_m survives unchanged. The continuous cohomology ring is therefore obtained by retaining precisely this stable part of the finite cohomology.

For finite discrete coefficients,

Hctsq(Γ,F2)≅lim→⁡mHq(Gm,F2).H_{\mathrm{cts}}^q(\Gamma,\mathbb{F}_2)\cong\varinjlim_m H^q(G_m,\mathbb{F}_2).

with transition maps given by inflation; see [13 Proposition 4.18] and the standard treatments of continuous cohomology in [28, 23].

The continuous LHS spectral sequence for

1⟶C4⟶Γ⟶Z2⟶11\longrightarrow C_4\longrightarrow\Gamma\longrightarrow\mathbb{Z}_2\longrightarrow1

has

E2p,q≅Hctsp(Z2,F2)⊗Hq(C4,F2).E_2^{p,q}\cong H_{\mathrm{cts}}^p(\mathbb{Z}_2,\mathbb{F}_2)\otimes H^q(C_4,\mathbb{F}_2).

The inversion action is trivial on H∗(C4,F2)H^*(C_4,\mathbb{F}_2), and cd⁡2(Z2)=1\operatorname{cd}_2(\mathbb{Z}_2)=1, so the spectral sequence is concentrated in columns 0 and 1 and collapses. The continuous Bockstein calculation, as in § 3, gives

y2=0,x2=xy.y^2=0,\qquad x^2=xy.

A class ω∈Hcts2(Γ,F2)\omega\in H_{\mathrm{cts}}^2(\Gamma,\mathbb{F}_2) restricting to the nonzero periodicity class of C4C_4 has nonzero powers in every even degree. The finite-level inflation calculation identifies this class and the resulting ring structure explicitly.

Corollary 5.3. One has

Hcts∗(C4⋊−1Z2,F2)≅F2[x,y,ω](y2,x2+xy),∣x∣=∣y∣=1,∣ω∣=2.H_{\mathrm{cts}}^*(C_4\rtimes_{-1}\mathbb{Z}_2,\mathbb{F}_2)\cong\frac{\mathbb{F}_2[x,y,\omega]}{(y^2,x^2+xy)},\qquad|x|=|y|=1,\quad|\omega|=2.

Under the direct-limit identification, ω\omega is represented by the compatible family {ωA,m}m≥2\{\omega_{A,m}\}_{m\geq2}, while the classes ωB,m\omega_{B,m} represent zero in the limit.

Proof. By Proposition 5.2, each finite-level ring has the presentation

H∗(Gm,F2)≅F2[xm,ym,um,wm](ym2,xm2+xmym),H^*(G_m,\mathbb{F}_2)\cong\frac{\mathbb{F}_2[x_m,y_m,u_m,w_m]}{(y_m^2,x_m^2+x_my_m)},

and inflation preserves xm,ym,wmx_m,y_m,w_m while sending umu_m to zero. Thus every class divisible by umu_m vanishes at the next level, whereas the subalgebras generated by xm,ym,wmx_m,y_m,w_m are identified isomorphically under inflation. Taking the direct limit gives

lim→⁡mH∗(Gm,F2)≅F2[x,y,ω](y2,x2+xy),∣ω∣=2.\varinjlim_m H^*(G_m,\mathbb{F}_2)\cong\frac{\mathbb{F}_2[x,y,\omega]}{(y^2,x^2+xy)},\qquad|\omega|=2.

Here ω\omega is represented by {wm}m≥2={ωA,m}m≥2\{w_m\}_{m\geq2}=\{\omega_{A,m}\}_{m\geq2}, and the quotient-side classes um=ωB,mu_m=\omega_{B,m} vanish in the limit. The asserted continuous cohomology ring follows. □

Remark 5.4. The same continuous LHS argument applies to C2α⋊rZ2C_{2^\alpha}\rtimes_r \mathbb{Z}_2 with α≥2\alpha\ge2 and an action r≡−1(mod4)r \equiv-1 \pmod{4} of 2-power order. The inversion family considered here additionally provides explicit finite-level representatives and determines the inflation maps in every degree.

An arithmetic consequence

The preceding results apply whenever arithmetic conditions identify a Galois group with one of the metacyclic families considered here. In that situation, the group-theoretic classification determines its degree-one cohomology algebra and its Koszul properties. We illustrate this with the real quadratic fields studied by Chems-Eddin and El Mamry.

Let η∈{1,2}\eta\in\{1,2\} and let q,r,sq,r,s be distinct odd primes satisfying the residue-symbol, unit, and 2-class-number conditions considered by Chems-Eddin and El Mamry [9 Theorem 3.10]. Set F=Q(ηqrs)F = \mathbb{Q}(\sqrt{\eta qrs}), and write h2(d)h_2(d) for the largest power of 2 dividing the class number of Q(d)\mathbb{Q}(\sqrt{d}). Let F∞F_\infty denote the cyclotomic Z2\mathbb{Z}_2-extension of FF, and let L(F∞)\mathcal{L}(F_\infty) be its maximal unramified pro-2-extension. We consider the Galois group

GF∞=Gal⁡(L(F∞)/F∞).G_{F_\infty}=\operatorname{Gal}(\mathcal{L}(F_\infty)/F_\infty).

Theorem 3.10 of Chems-Eddin and El Mamry identifies the metacyclic structure of this group under the stated arithmetic conditions. In particular, when the quartic-residue symbols satisfy

(rs)4≠(sr)4,\left(\frac{r}{s}\right)_4 \ne\left(\frac{s}{r}\right)_4,

their result selects a group of Type 1 with α=2\alpha=2 in the classification recalled in § 2. This allows us to apply the degree-one cohomology and Koszulity results established above.

Corollary 6.1. Assume the hypotheses of [9 Theorem 3.10], and let m≥2m \ge2 be determined by h2(ηqrs)=2m+1h_2(\eta qrs)=2^{m+1}. If

(rs)4≠(sr)4,\left(\frac{r}{s}\right)_4 \ne\left(\frac{s}{r}\right)_4,

then

GF∞≅C4⋊−1C2m.G_{F_\infty}\cong C_4\rtimes_{-1} C_{2^m}.

and

Dec⁡(H∗(GF∞,F2))≅F2[x,y](y2,x2+xy).\operatorname{Dec}(H^*(G_{F_\infty},\mathbb{F}_2))\cong\frac{\mathbb{F}_2[x,y]}{(y^2,x^2+xy)}.

In particular, this degree-one algebra is PBW, Koszul, universally Koszul, and strongly Koszul.

Proof. By [9 Theorem 3.10], the quartic-residue condition identifies GF∞G_{F_\infty} as the Type 1 group with α=2\alpha=2 and quotient exponent mm. The degree-one algebra follows from Theorem 3.4, and its enhanced Koszul properties follow from Corollary 4.2. □

References

References

  1. [1]B. Aaboun and A. Zekhnini, On the metacyclic 2-groups whose abelianizations are of type (2, 2^n), n ≥ 2, and applications, Research in Number Theory 9 (2023), Paper No. 55. https://doi.org/10.1007/s40999-023-00461-x
  2. [2]A. Adem and R. J. Milgram, Cohomology of Finite Groups, 2nd ed., Grundlehren der mathematischen Wissenschaften, Vol. 309, Springer, Berlin, 2004. https://doi.org/10.1007/978-3-662-06280-7
  3. [3]A. Beilinson, V. Ginzburg, and W. Soergel, Koszul duality patterns in representation theory, Journal of the American Mathematical Society 9 (1996), no. 2, 473–527. https://doi.org/10.1090/S0894-0347-96-00192-0
  4. [4]D. J. Benson, Representations and Cohomology II: Cohomology of Groups and Modules, Cambridge Studies in Advanced Mathematics, Vol. 31, Cambridge University Press, Cambridge, 1991. https://doi.org/10.1017/CBO9780511623622
  5. [5]G. M. Bergman, The diamond lemma for ring theory, Advances in Mathematics 29 (1978), no. 2, 178–218. https://doi.org/10.1016/0001-8708(78)90010-5
  6. [6]K. S. Brown, Cohomology of Groups, Graduate Texts in Mathematics, Vol. 87, Springer-Verlag, New York, 1982. https://doi.org/10.1007/978-1-4684-9327-6
  7. [7]A. Cassella and C. Quadrelli, Right-angled Artin groups and enhanced Koszul properties, Journal of Group Theory 24 (2021), no. 1, 17–38. https://doi.org/10.1515/jgth-2020-0049
  8. [8]S. K. Chebolu, J. Minač, C. Okay, A. Schultz, C. Ure, and D. Benson, The Bloch–Kato conjecture, decomposing fields, and generating cohomology in degree one, Journal of Number Theory 286 (2026), 301–341. https://doi.org/10.1016/j.jnt.2026.02.007
  9. [9]M. M. Chems-Eddin and H. El Mamry, On the existence of the maximal unramified pro-2-extension over the cyclotomic Z₂-extension with prescribed metacyclic Galois group, The Ramanujan Journal 70 (2026), Article 17. https://doi.org/10.1007/s11139-026-01381-x
  10. [10]A. Conca, Universally Koszul algebras, Mathematische Annalen 317 (2000), no. 2, 329–346. https://doi.org/10.1007/s002080000100
  11. [11]A. Conca, E. De Negri, and M. E. Rossi, Koszul algebras and regularity, in Commutative Algebra: Expository Papers Dedicated to David Eisenbud on the Occasion of his 65th Birthday, Springer, New York, 2013, pp. 285–315. https://doi.org/10.1007/978-1-4614-5292-8_8
  12. [12]T. Diethelm, The mod-p cohomology rings of the nonabelian split metacyclic p-groups, Archiv der Mathematik 44 (1985), no. 1, 29–38. https://doi.org/10.1007/BF019378
  13. [13]D. Harari, Galois Cohomology and Class Field Theory, Universitext, Springer, 2020. https://doi.org/10.1007/978-3-030-43901-9
  14. [14]J. Herzog, T. Hibi, and G. Restuccia, Strongly Koszul algebras, Mathematica Scandinavica 86 (2000), no. 2, 161–178. https://doi.org/10.7146/math.scand.a-14287
  15. [15]G. Hochschild and J.-P. Serre, Cohomology of group extensions, Transactions of the American Mathematical Society 74 (1953), no. 1, 110–134. https://doi.org/10.1090/S0002-9947-1953-0052438-8
  16. [16]J. Huebschmann, Automorphisms of group extensions and differentials in the Lyndon–Hochschild–Serre spectral sequence, Journal of Algebra 72 (1981), no. 2, 296–334. https://doi.org/10.1016/0021-8693(81)90296-9
  17. [17]J. Huebschmann, The mod-p cohomology rings of metacyclic groups, Journal of Pure and Applied Algebra 60 (1989), no. 1, 53–103. https://doi.org/10.1016/0022-4049(89)90107-2
  18. [18]N. K. Iyudu, Representation spaces of the Jordan plane, Communications in Algebra 42 (2014), no. 8, 3507–3540. https://doi.org/10.1080/00927872.2013.788184
  19. [19]R. C. Lyndon, The cohomology theory of group extensions, Duke Mathematical Journal 15 (1948), no. 1, 271–292. https://doi.org/10.1215/S0012-7094-48-01528-2
  20. [20]J. C. McConnell and J. C. Robson, Noncommutative Noetherian Rings, revised ed., Graduate Studies in Mathematics, Vol. 30, American Mathematical Society, Providence, RI, 2001. https://doi.org/10.1090/gsm/030
  21. [21]J. Minač, M. Palaisti, F. W. Pasini, and N. D. Tân, Enhanced Koszul properties in Galois cohomology, Research in the Mathematical Sciences 7 (2020), no. 2, Paper No. 10, 34 pp. https://doi.org/10.1007/s40687-020-00208-5
  22. [22]J. Minač, F. W. Pasini, C. Quadrelli, and N. D. Tân, Koszul algebras and quadratic duals in Galois cohomology, Advances in Mathematics 380 (2021), Article 107569. https://doi.org/10.1016/j.aim.2021.107569
  23. [23]J. Neukirch, A. Schmidt, and K. Wingberg, Cohomology of Number Fields, 2nd ed., Grundlehren der mathematischen Wissenschaften, Vol. 323, Springer, Berlin, 2008. https://doi.org/10.1007/978-3-540-37889-1
  24. [24]A. Polishchuk and L. Positselski, Quadratic Algebras, University Lecture Series, Vol. 37, American Mathematical Society, Providence, RI, 2005. https://doi.org/10.1090/ulect/037
  25. [25]S. B. Priddy, Koszul resolutions, Transactions of the American Mathematical Society 152 (1970), 39–60. https://doi.org/10.1090/S0002-9947-1970-0265437-8
  26. [26]C. Quadrelli, Pro-p groups with few relations and universal Koszulity, Mathematica Scandinavica 127 (2021), no. 1, 28–94. https://doi.org/10.7146/math.scand.a-123644
  27. [27]D. J. Rusin, The mod-2 cohomology of metacyclic 2-groups, Journal of Pure and Applied Algebra 44 (1987), 315–327. https://doi.org/10.1016/0022-4049(87)90035-1
  28. [28]J.-P. Serre, Galois Cohomology, Springer Monographs in Mathematics, Springer, Berlin, 1997. https://doi.org/10.1007/978-3-642-59141-9
  29. [29]C. T. C. Wall, Resolutions for extensions of groups, Proceedings of the Cambridge Philosophical Society 57 (1961), no. 2, 251–255. https://doi.org/10.1017/S0305004100035155

Paper details

vlnc:W_szcam2an
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 1

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents