Let JJ be a nonsingular integer matrix and suppose finitely many nonnegative integer levels ii occur. Put i0=minii_0=\min i and H=pJi0(Y)H=p_{J^{i_0}}(Y). Then every level variable has the representation

pJi(Y)=pJii0(H).p_{J^i}(Y)=p_{J^{i-i_0}}(H).

Thus the torus with coordinate HH is a common torus on which all these level functions can be compared, added, and multiplied by .

Bounded levels and compatibility

If 0ii0D0\le i-i_0\le D, only finitely many matrices Jii0J^{i-i_0} occur. The chain rule therefore gives uniform constants at each fixed derivative order for these pullbacks. A function defined on the common torus need not descend to an individual higher level: that would require additional invariance under the relevant deck translations.

When the active levels vary between slow-coordinate neighborhoods, local formulas must have equal pullbacks to the original YY-torus on overlaps. This is the compatibility condition defining one global function; merely giving a smooth formula on each local common torus does not ensure it.