The decomposition I = M(I) · N(I) at the stalk #
The nonmonomial part N(I) = I : M(I) is defined so that I = M(I) · N(I)
([Kol07, Definition–Lemma 110]). This file proves that identity at every stalk,
M(I)_x · N(I)_x = I_x. It is an instance of the ring-theoretic fact that for an ideal I ≤ (g)
one has (g) · (I : (g)) = I, applied with (g) = M(I)_x, which is principal
(exists_nonZeroDivisor_stalkIdeal_monomialPart) and contains I_x (stalkIdeal_le_monomialPart).
The analytic counterpart of Hironaka.BMO.monomialPart_mul_nonmonomialPart.
In a commutative ring, if I ≤ (g) then (g) · (I : (g)) = I. No regularity of g is needed:
the invertibility of the monomial part enters the decomposition I = M(I) · N(I) only through the
containment I ≤ M(I) and the local principal generator. Not in the sources.
At every stalk the monomial part times the nonmonomial part is I: M(I)_x · N(I)_x = I_x
([Kol07, Definition–Lemma 110]). The stalk M(I)_x is principal and contains I_x, so
Ideal.span_singleton_mul_colon_of_le applies to the colon N(I)_x = I_x : M(I)_x.