Does the choice of generator matter?

Yes. Replacing by ^j with (j,s)=1 generally gives a different algebra; the standard relation is (K/F,^j,a) ≅ (K/F,,a^j) in the Brauer group. In particular (K/F,^-1,a) is the opposite algebra of (K/F,,a).

Cyclic Algebras | KEVOS® Mathematics

Q: Why must the Galois group be cyclic?

Because a single adjoined symbol x implements a single automorphism by conjugation, and its powers then implement . For a non-cyclic Galois group one needs one symbol per group element together with a 2-cocycle recording how they multiply — that is Noether's crossed product. The cyclic case is where the cocycle collapses to the single scalar a.

Q: How much does the algebra depend on the choice of a?

Only on the class of a in F^×/N_K/F(K^×). Replacing x by dx for d ∈ K^× replaces a by a,N_K/F(d) and gives an isomorphic algebra. This is why the splitting criterion is a statement about norms and not about a itself.

Q: Is every division algebra a cyclic algebra?

No. Every central simple algebra of degree 2 or 3 over any field is cyclic, and every central simple algebra over a local or global field is cyclic. But Amitsur produced generic division algebras that are not even crossed products, so cyclicity fails in general.

Q: What does it mean that K is a splitting field?

Extending scalars, D ⊗_F K ≅ M_s(K). Concretely, the left-regular representation of D on itself as a K-space realises D inside M_s(K), and this embedding becomes an isomorphism after tensoring. Any maximal subfield of a central division algebra splits it.

Q: Why is a = 0 excluded?

With a = 0 the element x is nilpotent rather than a unit, and Kx + Kx^2 + … is a nonzero proper two-sided ideal. Simplicity fails immediately — for s = 2 the set Kx is already an ideal.

Executive Summary

A cyclic algebra is a field extension with one extra generator bolted on. Start with a cyclic Galois extension $K / F$ of degree $s$ , let $σ$ generate $Gal (K / F)$ , pick $a \in F^{\times}$ , and adjoin a symbol $x$ subject to two relations: $x^{s} = a$ , and $x b = σ (b) x$ for $b \in K$ . The result $(K / F, σ, a)$ is an $F$ -algebra of dimension $s^{2}$ .

Three facts make it useful. It is always simple with centre exactly $F$ , so it is a central simple algebra of degree $s$ . $K$ sits inside as a maximal subfield, self-centralising. And whether it is a division algebra or a matrix ring is decided by arithmetic in $F$ : it splits precisely when $a$ is a norm from $K$ .

$s^{2}$

\dim_{F} (K / F, σ, a)

$F$ Centre

$a \in N_{K / F} (K^{\times})$ Splits iff

1906Dickson

Overview

The finite-order case of the twisted Laurent series construction produces a division ring $D$ containing a cyclic extension $K$ of its centre $F$ , with $D = ⨁_{i < s} K x^{i}$ , $x b = σ (b) x$ and $x^{s}$ central. Dickson's observation in 1906 was that these data — a cyclic extension, its Galois generator, and one central scalar — can be taken as input rather than read off as output.

D = K \cdot 1 \oplus K \cdot x \oplus \dots \oplus K \cdot x^{s - 1}, x^{s} = a \in F^{\times}, x b = σ (b) x (b \in K)

(14.5)

The definition of $(K / F, σ, a)$ . Products are computed by pushing every $x$ to the right, applying $σ$ each time it passes a coefficient, and replacing $x^{s}$ by $a$ .

Consistency is not an accident of the presentation: the algebra can be realised as a quotient of a skew polynomial ring, which supplies associativity for free.

(K / F, σ, a) ≅ K [t; σ] / (t^{s} - a),

(14.5*)

$t^{s} - a$ is a central element of $K [t; σ]$ , so the ideal it generates is two-sided and the quotient is an $F$ -algebra of dimension $s^{2}$ .

Everything here is the degree- $s$ generalisation of a familiar object: taking $s = 2$ recovers the generalised quaternion algebras treated on Generalised Quaternion Algebras, and taking $F = ℝ$ , $K = ℂ$ , $a = - 1$ recovers Hamilton's $ℍ$ .

Learning Objectives

State the defining data $(K / F, σ, a)$ and the two relations $(14.5)$ .
Verify $\dim_{F} (K / F, σ, a) = s^{2}$ and identify $x$ as a unit.
Prove simplicity by the minimal-length argument on elements of an ideal.
Prove $C_{D} (K) = K$ , hence $Z (D) = F$ and $K$ is a maximal subfield.
Prove the splitting criterion $D ≅ M_{s} (F) \Leftrightarrow a \in N_{K / F} (K^{\times})$ .
Apply the prime-degree corollary to decide division-algebra status in a concrete case.

Definitions

Definition(14.5)The cyclic algebra

(K / F, σ, a)

Let $K / F$ be a cyclic Galois extension — finite, separable and normal, with $Gal (K / F) = ⟨ σ ⟩$ cyclic of order $s = \dim_{F} K$ . Fix $a \in F^{\times}$ and a symbol $x$ . Define

(K / F, σ, a) = ⨁_{i = 0}^{s - 1} K x^{i}

as a left $K$ -vector space, with multiplication extended distributively from $x^{s} = a$ and $x b = σ (b) x$ for $b \in K$ . This is an associative $F$ -algebra with $\dim_{F} = s^{2}$ , and $F \subseteq Z ((K / F, σ, a))$ .

Degree: $s = \dim_{F} K = | Gal (K / F) |$ . The algebra has dimension $s^{2}$ , so its degree as a central simple algebra is $s$ .
$N_{K / F}$: The norm $N_{K / F} (c) = c σ (c) \dots σ^{s - 1} (c) \in F$ , a multiplicative map $K^{\times} \to F^{\times}$ .
$C_{D} (S)$: The centraliser ${d \in D : d s = s d for all s \in S}$ .
$K [t; σ]$: The skew polynomial ring of left polynomials $\sum b_{i} t^{i}$ with $t b = σ (b) t$ ; a principal left ideal domain.
Split: $D ≅ M_{s} (F)$ . The opposite extreme from being a division algebra.
Crossed product: The generalisation in which $Gal (K / F)$ is any finite group and the relations are governed by a $2$ -cocycle; the cyclic case is the one with cyclic group.

The notation $(K / F, σ, a)$ is Lam's. Elsewhere one meets $(K / F, σ, a)$ written as a cyclic crossed product, and in degree 2 as a quaternion symbol.

Core Concepts

Why the two relations suffice

Any product of basis elements reduces using only $(14.5)$ : $(b x^{i}) (c x^{j}) = b σ^{i} (c) x^{i + j}$ , and if $i + j \geq s$ one writes $x^{i + j} = a x^{i + j - s}$ , legitimate because $a \in F$ is central. Associativity is then a finite check, or is inherited for free from the presentation $({14.5}^{*})$ as a quotient of $K [t; σ]$ — the polynomial $t^{s} - a$ is central there because $σ^{s} = id$ and $σ (a) = a$ .

(b x^{i}) (c x^{j}) = b σ^{i} (c) x^{i + j}, x^{i + j} = {\begin{cases} x^{i + j}, & i + j < s, \\ a x^{i + j - s}, & i + j \geq s . \end{cases}

(M)

$x$ is a unit, and that is the engine of simplicity

Since $x^{s} = a \in F^{\times}$ , the element $x$ is invertible with $x^{- 1} = a^{- 1} x^{s - 1}$ . Hence every basis element $b x^{i}$ with $b \neq 0$ is a unit. A nonzero ideal that contains a single-term element therefore contains a unit and is everything; the simplicity proof is entirely devoted to shrinking a general element down to a single term.

Distinct powers of $σ$ act differently

Because $σ$ generates a Galois group of order exactly $s$ , the automorphisms $id, σ, \dots, σ^{s - 1}$ are pairwise distinct as maps $K \to K$ . So for $0 \leq i < j \leq s - 1$ there exists $b \in K$ with $σ^{i} (b) \neq σ^{j} (b)$ . This tiny observation is used twice: once to shrink ideal elements, once to force centralisers into $K$ .

Key Results

Theorem(14.6)Structure of a cyclic algebra

Let $K / F$ be cyclic Galois of degree $s$ with $Gal (K / F) = ⟨ σ ⟩$ , let $a \in F^{\times}$ , and set $D = (K / F, σ, a)$ . Then:

$D$ is a simple $F$ -algebra with $Z (D) = F$ ; in particular $D$ is central simple over $F$ of dimension $s^{2}$ .
$C_{D} (K) = K$ .
$K$ is a maximal subfield of $D$ .

Proof

(1) Simplicity. Let $𝔄 \neq 0$ be a two-sided ideal. Among the nonzero elements of $𝔄$ choose one with the fewest nonzero coordinates,

α = b_{i_{1}} x^{i_{1}} + \dots + b_{i_{r}} x^{i_{r}}, 0 \leq i_{1} < \dots < i_{r} \leq s - 1, b_{i_{j}} \neq 0 .

Suppose $r \geq 2$ . Since $σ^{i_{1}} \neq σ^{i_{r}}$ , pick $b \in K$ with $σ^{i_{1}} (b) \neq σ^{i_{r}} (b)$ . The ideal contains both $α b = \sum_{j} b_{i_{j}} σ^{i_{j}} (b) x^{i_{j}}$ and $σ^{i_{1}} (b) α = \sum_{j} σ^{i_{1}} (b) b_{i_{j}} x^{i_{j}}$ , hence their difference

α b - σ^{i_{1}} (b) α = \sum_{j = 2}^{r} b_{i_{j}} (σ^{i_{j}} (b) - σ^{i_{1}} (b)) x^{i_{j}} .

Its $x^{i_{1}}$ -coordinate has vanished while its $x^{i_{r}}$ -coordinate $b_{i_{r}} (σ^{i_{r}} (b) - σ^{i_{1}} (b))$ is nonzero, so it is a nonzero element of $𝔄$ with fewer terms — contradiction. Hence $r = 1$ and $α = b_{i_{1}} x^{i_{1}}$ , a unit; so $𝔄 = D$ .

**(2) Centraliser of $K$ .** Let $d = \sum_{i = 0}^{s - 1} b_{i} x^{i}$ centralise $K$ . For $b \in K$ , comparing coordinates in $d b = b d$ gives $b_{i} σ^{i} (b) = b b_{i}$ for every $i$ . If $b_{i} \neq 0$ for some $1 \leq i \leq s - 1$ , then $σ^{i} (b) = b$ for all $b \in K$ , i.e. $σ^{i} = id$ , contradicting $| σ | = s$ . So $d = b_{0} \in K$ , and $K \subseteq C_{D} (K)$ is trivial.

**(3) Maximality of $K$ .** If $L$ is a subfield of $D$ with $K \subseteq L$ , then $L$ is commutative, so $L \subseteq C_{D} (K) = K$ , giving $L = K$ .

Centre. $F \subseteq Z (D)$ holds by construction. Conversely $b \in Z (D)$ lies in $C_{D} (K) = K$ , and $b x = x b = σ (b) x$ forces $σ (b) = b$ ; since $⟨ σ ⟩ = Gal (K / F)$ , Galois theory gives $b \in F$ . Hence $Z (D) = F$ .

Example(14.6a)Simple but not a division algebra

Take $a = 1$ , so $x^{s} = 1$ and $(1 - x) (1 + x + \dots + x^{s - 1}) = 0$ . For $s > 1$ both factors are nonzero, so $(K / F, σ, 1)$ has zero divisors. Simplicity is therefore strictly weaker than being a division algebra — as it must be, since $M_{s} (F)$ is simple.

Theorem(14.7)Splitting criterion

With $K / F$ , $σ$ , $s$ , $a$ as above and $D = (K / F, σ, a)$ , and with $N = N_{K / F}$ the field norm:

D ≅ M_{s} (F) as F -algebras ⟺ a \in N (K^{\times}) .

Proof

**( $\Leftarrow$ ) Reduce to $a = 1$ .** Suppose $a = N (d)^{- 1}$ for some $d \in K^{\times}$ , i.e. $N (d) a = 1$ . Put $y = d x$ . Repeatedly pushing $x$ past the coefficients gives $y^{s} = d σ (d) \dots σ^{s - 1} (d) x^{s} = N (d) a = 1$ , and for $b \in K$ we have $y b = d x b = d σ (b) x = σ (b) y$ . Since $y^{i} \in K^{\times} x^{i}$ , the elements $1, y, \dots, y^{s - 1}$ form a left $K$ -basis, so $D ≅ (K / F, σ, 1)$ .

**( $\Leftarrow$ ) The case $a = 1$ splits.** Write $D' = (K / F, σ, 1) = B / (t^{s} - 1)$ with $B = K [t; σ]$ . Because the coefficients of $t^{s} - 1 = (t^{s - 1} + \dots + t + 1) (t - 1)$ lie in the prime field, the factorisation is valid in $B$ , so $(t^{s} - 1) \subseteq B (t - 1)$ . The left ideal $B (t - 1)$ is maximal, since $\dim_{F} B / B (t - 1) = \deg (t - 1) \cdot \dim_{F} K = s$ and $B$ is a principal left ideal domain in which degree-one factors are irreducible. Hence $M = B / B (t - 1)$ is a simple left $D'$ -module with $\dim_{F} M = s$ , and the action gives an $F$ -algebra map $D' \to {End}_{F} (M) ≅ M_{s} (F)$ . It is injective because $D'$ is simple, and both sides have $F$ -dimension $s^{2}$ , so it is an isomorphism.

**( $\Rightarrow$ ).** If $D ≅ M_{s} (F)$ then $D$ has a simple module of $F$ -dimension $s$ . Writing $D = B / (t^{s} - a)$ , such a module is $B / B f$ for a principal left ideal with $B f \supseteq (t^{s} - a)$ , and $s = \dim_{F} B / B f = (\dim_{F} K) \deg f = s \deg f$ , so $\deg f = 1$ . After scaling on the left, $f = t - c$ with $c \in K$ . Then $t^{s} - a = β (t - c)$ for some $β = \sum_{i} b_{i} t^{i} \in B$ ; multiplying out and comparing coefficients from the top down yields successively

b_{s - 1} = 1, b_{s - 2} = σ^{s - 1} (c), b_{s - 3} = σ^{s - 1} (c) σ^{s - 2} (c), \dots, b_{0} = σ^{s - 1} (c) \dots σ (c),

and finally, from the constant term, $a = b_{0} c = σ^{s - 1} (c) \dots σ (c) c = N (c) \in N (K^{\times})$ .

Corollary(14.8)Prime degree

Suppose in addition that $s$ is a prime number. Then $D = (K / F, σ, a)$ is a division algebra if and only if $a \notin N_{K / F} (K^{\times})$ .

Proof

If $a \in N (K^{\times})$ then $D ≅ M_{s} (F)$ by $(14.7)$ , which is not a division algebra as $s \geq 2$ . Conversely suppose $D$ is not a division algebra. $D$ is a simple $F$ -algebra of dimension $s^{2}$ , so Wedderburn–Artin gives $D ≅ M_{r} (E)$ with $E$ a division $F$ -algebra and necessarily $r > 1$ . Comparing dimensions, $s^{2} = r^{2} \dim_{F} E$ , so $r^{2} ∣ s^{2}$ . As $s$ is prime, the square divisors of $s^{2}$ are $1$ and $s^{2}$ ; since $r > 1$ we get $r = s$ and $\dim_{F} E = 1$ , i.e. $E = F$ and $D ≅ M_{s} (F)$ . By $(14.7)$ , $a \in N (K^{\times})$ .

Remark(14.9)Beyond prime degree

When $s$ is composite, $a \notin N (K^{\times})$ is no longer sufficient — $D$ can be a proper matrix ring over a smaller division algebra. Wedderburn's 1914 sufficient condition asks instead that the image of $a$ in the abelian group $F^{\times} / N (K^{\times})$ have order exactly $s$ . That refinement, its proof by a norm on $K [t; σ]$ , and an example showing it is not necessary are the subject of The Reduced Norm of a Cyclic Algebra.

Proof Techniques and Method

How these proofs work, and which move to reuse.

Move 1

Shortest element in an ideal

To prove simplicity of a graded-looking algebra, take an ideal element with fewest nonzero components and kill its bottom component by a commutator-style difference. This is the standard proof for crossed products and skew group rings alike.

Move 2

Separate points with the Galois action

Distinct powers of $σ$ differ at some $b \in K$ . This single fact drives both simplicity and the centraliser computation, and is the reason the Galois group must act faithfully.

Move 3

Change of generator absorbs a norm

Replacing $x$ by $y = d x$ multiplies $a$ by $N (d)$ . So $(K / F, σ, a)$ depends on $a$ only through its class in $F^{\times} / N (K^{\times})$ — the whole splitting theory in one line.

Move 3 deserves to be stated as a lemma in its own right: for $d \in K^{\times}$ there is an $F$ -algebra isomorphism $(K / F, σ, a) ≅ (K / F, σ, a N_{K / F} (d))$ . It reduces every question about the algebra to a question about the class of $a$ modulo norms, and it is why the norm group appears at all.

The dimension count in the converse half of $(14.7)$ is also worth isolating: over a principal left ideal domain $B$ that is free of rank $\deg$ over its coefficient field, $\dim_{F} B / B f = \deg (f) \cdot \dim_{F} K$ . Turning module dimensions into polynomial degrees is what makes the norm appear from nothing.

Fix a cyclic extensionChoose

K / F

finite, separable, normal, with cyclic Galois group, and name a generator

σ

Compute the norm groupDetermine

N_{K / F} (K^{\times}) \subseteq F^{\times}

— usually the hardest step, and the only one that is genuinely arithmetic.

Choose $a$ outside itFor prime degree, any

a \notin N (K^{\times})

works. For composite degree, aim for

a

whose class has order

s

Assemble the algebraTake

⨁_{i < s} K x^{i}

with

x^{s} = a

and

x b = σ (b) x

, or equivalently

K [t; σ] / (t^{s} - a)

Verify and record invariantsDegree

s

, centre

F

K

a maximal subfield and a splitting field, class in

Br (F)

of order dividing

s

Worked Example

A quadratic example over a rational function field

Let $E$ be a field with $char E \neq 2$ , let $t$ be transcendental over $E$ , and set $F = E (t)$ , $K = F (\sqrt{t})$ . The extension $K / F$ is cyclic of degree $2$ ; let $σ$ be the $F$ -automorphism with $σ (\sqrt{t}) = - \sqrt{t}$ . For $a \in E^{\times}$ we claim

D = (K / F, σ, a) is a division algebra ⟺ a \notin (E^{\times})^{2} .

(E.1)

So a single non-square in $E$ produces a $4$ -dimensional division algebra over $E (t)$ .

Because $s = 2$ is prime, $(14.8)$ reduces the claim to: $a \in N_{K / F} (K^{\times})$ if and only if $a$ is a square in $E^{\times}$ .

One direction is immediate. If $a = b^{2}$ with $b \in E^{\times} \subseteq F^{\times}$ then $a = b σ (b) = N (b)$ .

The other direction is a degree count. Suppose $a = N (f + \sqrt{t} g) = f^{2} - t g^{2}$ with $f, g \in E (t)$ . Clear denominators: write $f = f_{0} / h$ , $g = g_{0} / h$ with $f_{0}, g_{0}, h \in E [t]$ and $h \neq 0$ . Then

a h (t)^{2} = f_{0} (t)^{2} - t g_{0} (t)^{2} .

(E.2)

The polynomial $f_{0}^{2}$ has even degree and $t g_{0}^{2}$ has odd degree, so no cancellation of leading terms is possible and the degree of the right-hand side is $\max {2 \deg f_{0}, 1 + 2 \deg g_{0}}$ . The left-hand side has even degree $2 \deg h$ , so the maximum must be attained by $f_{0}^{2}$ : thus $\deg f_{0} = \deg h$ and, comparing leading coefficients, $a (lc h)^{2} = (lc f_{0})^{2}$ . Hence $a = (lc f_{0} / lc h)^{2} \in (E^{\times})^{2}$ , as claimed.

This example is reused inside Lam's construction of a degree- $4$ cyclic division algebra $(K / F, σ, - 1)$ , where it identifies a centraliser as a division algebra; it is also the pattern behind most quaternion examples over function fields.

Comparison and Classification

Behaviour of $(K / F, σ, a)$ as $a$ varies
Choice of $a$	Structure	Why
$a = 1$	$≅ M_{s} (F)$ , split	$1 = N (1)$ ; explicit zero divisors $(1 - x)$ and $1 + x + \dots + x^{s - 1}$
$a = N (d)$ for some $d \in K^{\times}$	$≅ M_{s} (F)$ , split	$(14.7)$ ; substitute $y = d^{- 1} x$ to reduce to $a = 1$
$a \notin N (K^{\times})$ , $s$ prime	division algebra	$(14.8)$
$a \notin N (K^{\times})$ , $s$ composite	$≅ M_{r} (E)$ for some division $E$ ; may or may not be a division algebra	Wedderburn–Artin permits $1 < r < s$
image of $a$ has order $s$ in $F^{\times} / N (K^{\times})$	division algebra	Wedderburn's theorem $(14.9)$
$a = 0$ (not permitted)	not simple	$K x$ becomes an ideal; the construction requires $a \in F^{\times}$

What each hypothesis buys
	$\dim_{F} D = s^{2}$	$D$ simple	$Z (D) = F$	$K$ maximal subfield	$D$ division
$K / F$ finite separable normal	yes	yes	yes	yes	no
$Gal (K / F)$ cyclic, generated by $σ$	yes	yes	yes	yes	no
$a \in F^{\times}$ (nonzero)	yes	yes	yes	yes	no
$a \notin N (K^{\times})$ and $s$ prime	yes	yes	yes	yes	yes
$a$ of order $s$ in $F^{\times} / N (K^{\times})$	yes	yes	yes	yes	yes

What each hypothesis buys

Is $(K / F, σ, a)$ a division algebra?

$a \in N_{K / F} (K^{\times})$ No — it is

M_{s} (F)

, by

(14.7)

. Reduce to

a = 1

by replacing

x

with

d^{- 1} x

$s$ prime and $a \notin N (K^{\times})$ Yes, by

(14.8)

. This settles quaternion algebras and all degree-

3

examples.

$s$ composite, class of $a$ of order $s$ Yes, by Wedderburn's theorem

(14.9)

— sufficient but not necessary.

$s$ composite, class of $a$ of smaller orderUndecided by these criteria. Over a number field the converse of

(14.9)

holds, so the answer is no; over general

F

examples exist both ways.

Relationship Map

Cyclic algebras are the first rung of a ladder that reaches all central simple algebras.

Cyclic algebra

(K / F, σ, a)

Crossed product

(K / F, G, c)

Central simple

F

-algebraBrauer class in

Br (F)

$(K / F, σ, a)$ — what the construction specialises to
- $s = 2$
  - generalised quaternion algebra
  - $ℍ = (ℂ / ℝ, conjugation, - 1)$
- $a$ a norm
  - $M_{s} (F)$
  - trivial class in the Brauer group
- $K = k ((x^{s}))$ , $F = k_{0} ((x^{s}))$
  - the finite-order twisted Laurent series division ring
  - $a = x^{s}$
- cyclic Kummer case
  - $K = F (\sqrt[s]{b})$ with $ζ_{s} \in F$
  - the symbol algebra with $x y = ζ_{s} y x$

Emmy Noether's crossed products replace $⟨ σ ⟩$ by an arbitrary finite Galois group and the single scalar $a$ by a $2$ -cocycle; every central simple algebra split by a Galois extension is such a crossed product. Whether every division algebra is a cyclic algebra is a much harder question — the answer is no in general, by Amitsur's non-crossed-product examples, though it is yes in degree $2$ and $3$ and over local and global fields.

Applications and Industry Use

Applications here means where this structure is used — inside mathematics and in the engineering and computing disciplines that consume it.

Wireless communications

Space–time block codes

Perfect space–time codes for multiple-antenna channels are built from cyclic division algebras: the algebra's left-regular matrix representation supplies a codebook whose nonzero differences are invertible, giving full diversity. The non-norm condition is exactly the design constraint.

Number theory

Brauer groups and class field theory

Over a local or global field every central simple algebra is cyclic, and the Brauer group is computed by local invariants. Cyclic algebras are the concrete models on which the Hasse–Brauer–Noether theorem is stated.

Algebraic geometry

Severi–Brauer varieties

A cyclic algebra of degree $s$ has an associated Severi–Brauer variety, a form of projective space with a rational point exactly when the algebra splits. Splitting criteria become rational-point questions.

Symbolic computation

Explicit algebra libraries

Magma and Sage represent cyclic and quaternion algebras by structure constants derived from $(14.5)$ , and implement splitting tests via norm equations over number fields.

Coding theory

Skew-cyclic and MRD codes

Maximum rank distance codes and skew-cyclic codes over finite fields use $K [t; σ]$ and its quotients — the finite-field case of the presentation $({14.5}^{*})$ , where $σ$ is a Frobenius power.

Quantum theory

Symbol algebras and root-of-unity relations

When $F$ contains a primitive $s$ th root of unity, the cyclic algebra becomes a symbol algebra with $x y = ζ_{s} y x$ — the finite-dimensional relative of the Weyl relation used in quantum tori.

Wireless coding is the clearest case where the division property does engineering work: a code drawn from a division algebra can never have two distinct codewords whose difference is singular, which is precisely full-diversity transmission.

Standards and Notation

Standards here covers notation, symbol and markup standards, and reference implementations, rather than material or design codes.

Lam's notation

(K / F, σ, a)

Crossed-product notation

(K / F, G, c)

with

c

2

-cocycle; the cyclic case is

c

determined by

a

Symbol notation, degree $2$

(\frac{b, a}{F})

(b, a)_{F}

for the quaternion algebra

Symbol notation, degree $s$

(b, a)_{ζ}

when

ζ_{s} \in F

and

K = F (\sqrt[s]{b})

Degree vs dimensionDegree

s

, dimension

s^{2}

. Some sources say rank for

s^{2}

— check before quoting

MagmaCyclicAlgebra, QuaternionAlgebra, IsMatrixRing

Sage / GAPQuaternionAlgebra in Sage; AlgebraByStructureConstants in GAP for general degree

MarkupPresentation MathML per ISO/IEC 40314; symbols per ISO 80000-2

Computational Notes

Computational notes cover algorithms, cost and library behaviour rather than manufacturing process.

A cyclic algebra is finite dimensional and given by structure constants, so all its ring-theoretic invariants are computable once arithmetic in $K$ is. The bottleneck is arithmetic in $F$ , not in the algebra.

Multiplication of two general elements costs $O (s^{2})$ multiplications in $K$ plus $O (s^{2})$ evaluations of powers of $σ$ ; precomputing $σ^{i}$ on a basis of $K$ reduces the latter to matrix–vector products.
Regular representation. The map sending $d$ to the matrix of right multiplication on the left $K$ -basis $1, x, \dots, x^{s - 1}$ embeds $D$ into $M_{s} (K)$ ; this is how a machine stores the algebra and is the source of the reduced norm.
Splitting test. Deciding $a \in N_{K / F} (K^{\times})$ is a norm equation. Over a number field this is solvable — by class field theory the norm group is described by local conditions at finitely many places — and Magma implements it. Over a general field there is no algorithm.
**Degree- $2$ shortcut.** For quaternion algebras splitting reduces to solvability of the conic $b u^{2} + a v^{2} = w^{2}$ , decidable over $ℚ$ by Legendre's theorem plus factorisation of $a$ and $b$ .
Zero-divisor search is the wrong algorithm. Searching for zero divisors by brute force is exponential in $\dim_{F} D$ and settles nothing when none are found; use the norm criterion instead.

Failure Modes and Common Mistakes

Do not drop separability or normality: without them $\dim_{F} K \neq | Gal (K / F) |$ and the dimension count $s^{2}$ fails.
Do not allow $a = 0$ : then $K x$ is a nonzero proper ideal and the algebra is not simple.
Do not assume $K$ is the only maximal subfield. Non-isomorphic maximal subfields coexist; Lam's exercises for §14 exhibit them in Dickson's degree- $3$ example.
Do not confuse the degree $s$ with the dimension $s^{2}$ , or the order of $a$ in $F^{\times} / N (K^{\times})$ with the order of the Brauer class.
Do not expect $(K / F, σ, a) ≅ (K / F, σ, a')$ to force $a = a'$ — the algebra sees $a$ only modulo $N_{K / F} (K^{\times})$ .

Historical Notes and Lessons Learned

1899Hilbert's twisted seriesThe finite-order case of $k ((x; σ))$ exhibits, implicitly, the relations later abstracted as $(14.5)$ .
1906Dickson defines cyclic algebrasDickson isolates the data $(K / F, σ, a)$ and proves the basic structure theory, launching the systematic construction of division algebras of arbitrary degree.
1914Wedderburn's norm criterionA sufficient condition for a cyclic algebra of any degree to be a division algebra, in terms of the order of $a$ in $F^{\times} / N (K^{\times})$ .
1929Noether's crossed productsEmmy Noether generalises to arbitrary finite Galois groups with a $2$ -cocycle, recasting the theory cohomologically.
1932Hasse–Brauer–NoetherOver an algebraic number field every central simple algebra is cyclic, and the Brauer group is determined by local invariants.
1972Amitsur: non-crossed productsGeneric division algebras of suitable degree are shown not to be crossed products at all, ending the hope that cyclic algebras exhaust the subject.

The lesson is that a construction outlives the classification it was invented for. Cyclic algebras do not exhaust central simple algebras, but they remain the only family in which one can write down a basis, multiply two elements by hand, and decide the division property by an arithmetic condition.

Quick Reference

Data

K / F

cyclic of degree

s

Gal (K / F) = ⟨ σ ⟩

a \in F^{\times}

Relations

x^{s} = a

and

x b = σ (b) x

for

b \in K

Presentation

K [t; σ] / (t^{s} - a)

Dimension

\dim_{F} D = s^{2}

; degree

s

Structuresimple,

Z (D) = F

C_{D} (K) = K

Splits iff

a \in N_{K / F} (K^{\times})

Prime degreedivision

\Leftrightarrow a \notin N_{K / F} (K^{\times})

Change of generator

(K / F, σ, a) ≅ (K / F, σ, a N_{K / F} (d))

Results at a glance
Statement	Hypotheses	Reference
$\dim_{F} D = s^{2}$ , $F \subseteq Z (D)$	$K / F$ cyclic of degree $s$ , $a \in F^{\times}$	(14.5)
$D$ simple, $Z (D) = F$	same	(14.6)(1)
$C_{D} (K) = K$	same	(14.6)(2)
$K$ is a maximal subfield	same	(14.6)(3)
$D ≅ M_{s} (F) \Leftrightarrow a \in N (K^{\times})$	same	(14.7)
division $\Leftrightarrow a \notin N (K^{\times})$	same, and $s$ prime	(14.8)

Frequently Asked Questions

Why must the Galois group be cyclic?

Because a single adjoined symbol $x$ implements a single automorphism by conjugation, and its powers then implement $⟨ σ ⟩$ . For a non-cyclic Galois group one needs one symbol per group element together with a $2$ -cocycle recording how they multiply — that is Noether's crossed product. The cyclic case is where the cocycle collapses to the single scalar $a$ .

How much does the algebra depend on the choice of

a

Only on the class of $a$ in $F^{\times} / N_{K / F} (K^{\times})$ . Replacing $x$ by $d x$ for $d \in K^{\times}$ replaces $a$ by $a N_{K / F} (d)$ and gives an isomorphic algebra. This is why the splitting criterion is a statement about norms and not about $a$ itself.

Does the choice of generator

σ

matter?

Yes. Replacing $σ$ by $σ^{j}$ with $\gcd (j, s) = 1$ generally gives a different algebra; the standard relation is $(K / F, σ^{j}, a) ≅ (K / F, σ, a^{j})$ in the Brauer group. In particular $(K / F, σ^{- 1}, a)$ is the opposite algebra of $(K / F, σ, a)$ .

Is every division algebra a cyclic algebra?

No. Every central simple algebra of degree $2$ or $3$ over any field is cyclic, and every central simple algebra over a local or global field is cyclic. But Amitsur produced generic division algebras that are not even crossed products, so cyclicity fails in general.

What does it mean that

K

is a splitting field?

Extending scalars, $D \otimes_{F} K ≅ M_{s} (K)$ . Concretely, the left-regular representation of $D$ on itself as a $K$ -space realises $D$ inside $M_{s} (K)$ , and this embedding becomes an isomorphism after tensoring. Any maximal subfield of a central division algebra splits it.

Why is

a = 0

excluded?

With $a = 0$ the element $x$ is nilpotent rather than a unit, and $K x + K x^{2} + \dots$ is a nonzero proper two-sided ideal. Simplicity fails immediately — for $s = 2$ the set $K x$ is already an ideal.

References

T. Y. Lam, A First Course in Noncommutative Rings, Graduate Texts in Mathematics 131, Springer-Verlag, 1991, §14, (14.5)–(14.8), pp. 231–236.
L. E. Dickson, Algebras and Their Arithmetics, University of Chicago Press, 1923; and the 1906 papers on linear associative algebras in which cyclic algebras first appear.
N. Jacobson, Basic Algebra II, 2nd edition, W. H. Freeman, 1989, Chapter 4 (central simple algebras, crossed products, cyclic algebras).
R. S. Pierce, Associative Algebras, Graduate Texts in Mathematics 88, Springer-Verlag, 1982, Chapters 13–15.
P. Gille and T. Szamuely, Central Simple Algebras and Galois Cohomology, Cambridge Studies in Advanced Mathematics 101, Cambridge University Press, 2006, Chapter 2.
S. A. Amitsur, “On central division algebras”, Israel Journal of Mathematics 12 (1972), 408–420.

AI Suggested Questions

Prove that $(K / F, σ, a) \otimes_{F} (K / F, σ, a')$ is Brauer-equivalent to $(K / F, σ, a a')$ .
Show that $(K / F, σ, a)$ has order dividing $s$ in the Brauer group of $F$ , and find an example where the order is strictly smaller than $s$ .
Work out the Severi–Brauer variety of a cyclic algebra of degree $3$ and its rational points.
How is a cyclic algebra of degree $s$ used to build a full-diversity space–time code, and where does the non-norm condition enter?
Give an algorithm deciding whether $a$ is a norm from a cyclic extension of number fields, and state its complexity.
Compare cyclic algebras with symbol algebras when $F$ contains a primitive $s$ th root of unity.
Describe Amitsur's non-crossed-product division algebras and what degree is needed.

01Executive Summary

02Overview

03Learning Objectives

04Definitions

05Core Concepts

Why the two relations suffice

x is a unit, and that is the engine of simplicity

Distinct powers of σ act differently

06Key Results

07Proof Techniques and Method

Shortest element in an ideal

Separate points with the Galois action

Change of generator absorbs a norm

08Worked Example

A quadratic example over a rational function field

09Comparison and Classification

10Relationship Map

11Applications and Industry Use

Space–time block codes

Brauer groups and class field theory

Severi–Brauer varieties

Explicit algebra libraries

Skew-cyclic and MRD codes

Symbol algebras and root-of-unity relations

12Standards and Notation

13Computational Notes

14Failure Modes and Common Mistakes

15Historical Notes and Lessons Learned

16Quick Reference

17Frequently Asked Questions

18Related KEVOS Topics

19References

20AI Suggested Questions

Continue learning

Executive Summary

Overview

Learning Objectives

Definitions

Core Concepts

$x$ is a unit, and that is the engine of simplicity

Distinct powers of $σ$ act differently

Key Results

Proof Techniques and Method

Worked Example

Comparison and Classification

Relationship Map

Applications and Industry Use

Standards and Notation

Computational Notes

Failure Modes and Common Mistakes

Historical Notes and Lessons Learned

Quick Reference

Frequently Asked Questions

Related KEVOS Topics

References

AI Suggested Questions