Statement

Let 0<a<b0<a<b, d>0d>0, eRe\in\mathbb R, and let f(R,y)f(R,y) be smooth and supported in [a,b]×Tn[a,b]\times\mathbb T^n. Put U=RdU=R^d and

F(U,y)=Ref(R,y)dRd1,Tef(R,y)=ReIχF(Rd,y),F(U,y)=\frac{R^e f(R,y)}{dR^{d-1}},\qquad \mathcal T_e f(R,y)=R^{-e}I_\chi F(R^d,y),

where IχI_\chi is the and its transition lies in a compact subset of U>0U>0. For

DR=R+MdRd1vy,\mathcal D_R=\partial_R+MdR^{d-1}v\cdot\nabla_y,

the exact weighted identity is

(DR+e/R)Tef=fRedRd1χ(Rd)JMF(Rd,y).(\mathcal D_R+e/R)\mathcal T_e f =f-R^{-e}dR^{d-1}\chi'(R^d)J_MF(R^d,y).
Consequences

This follows by the chain rule and cancellation of the derivative of ReR^{-e} with e/Re/R. All radial powers and the coordinate change have bounded derivatives on the fixed positive shell. The output is compactly supported, and fixed coefficient derivatives have no loss in powers of MM by the fixed-shift formula.

The zero weighted mean condition RefydR=0\int R^e\langle f\rangle_y\,dR=0 is exactly FydU=0\int\langle F\rangle_y\,dU=0. Under a Diophantine drift, it gives arbitrarily high inverse-power bounds for the defect when the source has the required derivative bounds. For an auxiliary-independent source the defect is zero. Parameter derivatives of a moving coordinate map or of the drift require a separate chain-rule calculation.