Functorial affinization of Nash's manifold

Let M be a singular irreducible complex manifold of dimension n. There are Q divisors D[-1], D[0], D[1],...,D[n+1] on Nash's manifold U -> M such that D[n+1] is relatively ample on bounded sets, D[n] is relatively eventually basepoint free on bounded sets, and D[-1] is canonical with the same relative plurigenera as a resolution of M. The divisor D=D[n] is the supremum of divisors (1/i)D_i. An arc g containing one singular point of M lifts to U if and only if the generating number of oplus_i O_g(D_i) is finite. When it is finite it equals 1+(K_U-K) .g where O_U(K) is the pullback mod torsion of Lambda^n Omega_M. If C is a complete curve in U then (-1/(n+1))K_U .C=D_1 .C + D_n+2 .C + D_(n+2)^2 .C +... When there are infinitely many nonzero terms the sum should be taken formally or p-adically for a prime divisor p of n+2. There are finitely many nonzero terms if and only if C. D=0. The natural holomorphic map U -> M factorizes through the contracting map U -> Y_0. If M is bounded, the Grauert-Riemenschneider sheaf of M is Hom(O_M(D_{(n+2)^i - 1}), O_M(D_{(n+2)^i})) for large i. If M is projective, singular foliations on M such that K+(n+1)H is a finitely-generated divisor of Iitaka dimension one are completely resolvable, where K is the canonical divisor of the foliation and H is a hyperplane. There are some precise open questions in the article. According to a question of [7] it is not known whether Y_0 has canonical singularities.