The countable Nullstellensatz over ℂ #
The classical Nullstellensatz produces a ℂ-point of a nonzero
commutative ℂ-algebra of finite type. Deligne's Proposition 4.5
needs a ℂ-point of the algebra over which the fibre functor is
defined, and that algebra is not presented as a finite-type one; it
is instead built from countably much data, so its dimension as a
ℂ-vector space is at most countable. The countable-dimension
hypothesis is a perfectly good replacement for the finite-type
hypothesis, because the obstruction is uncountable: a field
extension of ℂ containing a transcendental element x already
contains the ℂ-linearly independent family (x - a)⁻¹, a : ℂ, of
cardinality the continuum.
Three results are recorded. A field extension of ℂ of at most countable dimension is algebraic; a nonzero commutative ℂ-algebra of at most countable dimension maps onto ℂ; and an algebra generated by a countable set has at most countable dimension.
The finite-type predecessor is RS.exists_algHom_complex in
NullPoint.lean, which runs the same last two steps
— quotient by a maximal ideal, lift along IsAlgClosed.lift — over
Zariski's lemma instead of the dimension count.
A field extension of ℂ of at most countable dimension is
algebraic. A transcendental element x would give the ℂ-linearly
independent family (x - a)⁻¹ indexed by a : ℂ, forcing the
dimension to be at least the continuum.
The countable Nullstellensatz: a nonzero commutative ℂ-algebra of at most countable dimension admits a ℂ-point. Quotient by a maximal ideal, observe that the residue field again has at most countable dimension, hence is algebraic over ℂ, and lift along the algebraically closed base.
A field extension of the complex numbers of countable dimension is the complex numbers: every element is a complex multiple of the unit. This is the shape the scalar computation for a simple algebra consumes.