Commutativity and associativity of the double (GPR 18.14(a),(b)) #
Cay(B, μ) is associative if and only if B is commutative and associative with involutive
conjugation (GPR 18.14(b)), and commutative if and only if B is commutative with trivial
conjugation (GPR 18.14(a)). This file proves the directions that build algebras, and packages them
as Ring, CommRing and Algebra structures on the double so that the Cayley–Dickson tower can
be iterated.
Associativity of Cay(B, μ) supplies the associative input needed for a further application of
ConicAlgebra.isMultiplicative_cay or ConicAlgebra.isAlternative_cay. To prove that a double is
associative, mul_assoc_cay asks its input to be commutative. The base ring starts this process,
but it is not the only possible commutative input: in characteristic two, iterated doubles with
trivial conjugation can remain commutative (GPR 18.17(b)).
For iteration, ConicAlgebra.isMultiplicative_cay (the associative special case of
GPR Exc. 18.18) and ConicAlgebra.isAlternative_cay (GPR 18.14(c)) both require their input
B to be associative, which for a double Cay(B₀, μ₀) is what this file supplies.
The computation #
Write x = (a, b), y = (c, d), z = (e, f). Both components of (xy)z and of x(yz) expand
into four terms each, and they match term by term using only
- anti-multiplicativity of the conjugation of
B, to pushconjthrough the products it meets (conj (fc) = (conj c) (conj f),conj (d (conj e)) = e (conj d)), and - commutativity and associativity of
B, to reorder the resulting monomials.
The conjugates conj a, conj c, conj e, … are never expanded via conj z = t(z)·1 - z; they behave
as atoms throughout. This is unlike Conic/CayleyDicksonAlternative.lean, where eliminating the
conjugation first is what makes the cancellations linear — here commutativity does that job instead,
and ring closes both components.
Main results #
ConicAlgebra.mul_assoc_cay:GPR 18.14(b),⇐.ConicAlgebra.mul_comm_cay:GPR 18.14(a),⇐.ConicAlgebra.exists_mul_ne_mul_cay,ConicAlgebra.exists_mul_mul_ne_mul_mul_cay: explicit witnesses for the failure of commutativity and of associativity.ConicAlgebra.Cay.ring,ConicAlgebra.Cay.algebra,ConicAlgebra.Cay.commRing: the resulting structures.
Implementation notes #
Cay.ring and Cay.algebra are defs, not instances: associativity of the double depends on
hypotheses on B (CommRing, and involutivity of the conjugation) that instance search cannot
discharge, and a global instance would be unsound. Use them with letI at the point where the next
rung of the tower is built. Cay.algebra is Algebra.ofModule, which needs exactly the
SMulCommClass/IsScalarTower pair the double already carries.
⚠ The involutivity hypothesis does not follow from commutativity and associativity over an
arbitrary base — see the trace-defect counterexample over ℤ/4 — but it does as soon as B is
finitely generated projective, or the base is a field.
References #
- [GPR] Garibaldi, Petersson, Racine, Albert Algebras over Commutative Rings, 18.13, 18.14(a),(b).
GPR 18.14(b), ⇐. If B is commutative associative and its conjugation is an
involution, then Cay(B, μ) is associative, for every scalar μ.
The double of a commutative associative conic algebra with involutive conjugation, as a
Ring. A def, not an instance: see the module docstring.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The double as a k-algebra, given the Ring structure of ConicAlgebra.Cay.ring. Built by
Algebra.ofModule from the SMulCommClass/IsScalarTower instances the double already has.
Equations
- ConicAlgebra.Cay.algebra C μ hconj = Algebra.ofModule ⋯ ⋯
Instances For
Failure of commutativity and associativity (GPR 18.14(a),(b), ⇒) #
The contrapositives, in the explicit form the tower needs: a witness that B's conjugation is not
the identity produces a witness that the double is not commutative, and a witness that B is not
commutative produces a witness that the double is not associative. These are what show the
Cayley–Dickson tower really does lose a property at each step rather than merely failing to
preserve it.
Cay(B, μ) is not commutative as soon as the conjugation of B moves some element:
(u, 0)(0, 1) = (0, u) while (0, 1)(u, 0) = (0, conj u).
Cay(B, μ) is not associative as soon as B is not commutative. With x = (0, 1),
y = (conj a, 0) and z = (conj c, 0), the second components of (xy)z and x(yz) are ac and
ca.
Commutativity (GPR 18.14(a)) #
If the conjugation of a commutative B is the identity then Cay(B, μ) is commutative.
Both products come out as (ac + μ db, da + bc); only commutativity of B is used, since with
conj = id there is nothing left for the conjugation to do. The base ring satisfies the hypothesis
(ConicAlgebra.baseRing_conj_apply). More generally, GPR 18.14(a) characterizes commutativity of
Cay(B, μ) by commutativity of B with trivial conjugation.
With the conjugation of B the identity, the conjugation is automatically an involution: the
condition conj (xy) = conj y * conj x reduces to xy = yx.
The double of a commutative conic algebra whose conjugation is the identity, as a CommRing.
This is the instance that gets the Cayley–Dickson tower off the ground: it makes Cay(k, μ₁)
commutative, which is what ConicAlgebra.mul_assoc_cay needs in order to make Cay(Cay(k, μ₁), μ₂)
associative. A def, not an instance — see the module docstring.
Equations
- ConicAlgebra.Cay.commRing C μ hconj = { toRing := ConicAlgebra.Cay.ring C μ ⋯, mul_comm := ⋯ }