1. ## Substitution of differentials

I have never thought that differentials could be manipulated as numbers. I do not see why that is, in any case, justified.

Can someone prove that the product of dy/du and 1/(dx/du) is equal to dy/dx ? I have seen that if you manipulate the differentials you will get the result easily, but is that justified and what are the domain restrictions in any case that it is justified?

Sorry if my question isn't well phrased.

2. Most people prefer to prove it from the chain rule alone, which doesn't depend on the fractional notation.

Just in case a picture helps...

... where...

... is the chain rule. Straight continuous lines differentiate downwards (integrate up) with respect to the main variable (in this case x), and the straight dashed line similarly but with respect to the dashed balloon expression (the inner function of the composite which is subject to the chain rule).

From the bottom row, dx/du = one over du/dx.

Or maybe

See here for an interesting historical perspective on (and reaction against) nervousness about taking the fractional notation literally...

Non-standard analysis - Wikipedia, the free encyclopedia

_________________________________________

Don't integrate - balloontegrate!

Balloon Calculus; standard integrals, derivatives and methods

Balloon Calculus Drawing with LaTeX and Asymptote!