Consistency checking the relation between total differentials using the Jacobian matrix

32 Views Asked by At

I have the following relationship between variables: $$\varepsilon = \Psi -\frac{v^2}{2}\tag{1}$$

A while back, I had an integral with respect to $v$ and I wanted to convert it to an integral with respect to $\varepsilon$. I used the Jacobian method of converting differential elements.

  1. $$\mathrm{d}v = \det\left[\frac{\partial v}{\partial \varepsilon}\right]\mathrm{d}\varepsilon$$ $$\frac{\partial \varepsilon}{\partial v} = -v \implies \frac{\partial v}{\partial \varepsilon} = -\frac{1}{v} \implies \mathrm{d}v = -\frac{1}{v}\mathrm{d}\varepsilon$$ $$\implies \mathrm{d}\varepsilon = -v\mathrm{d}v\tag{2}$$

I was satisfied with equation (2) and used it to prove an expression in a research paper. Now, I wonder what would have happened if I need to change $\mathrm{d}\varepsilon$ to $\mathrm{d}\Psi$. In such a case,

  1. $$\mathrm{d}\varepsilon = \det\left[\frac{\partial \varepsilon}{\partial \Psi}\right]\mathrm{d}\Psi$$ $$\frac{\partial \varepsilon}{\partial \Psi} = 1$$ $$\implies \mathrm{d}\varepsilon = \mathrm{d}\Psi\tag{3}$$

Then, the prediction for the relationship between $\mathrm{d}v$ and $\mathrm{d}\Psi$ from eqn (2) and eqn (3) would be: $$-v\mathrm{d}v = \mathrm{d}\varepsilon = \mathrm{d}\Psi$$ $$\text{i.e.,}\ \ \ \ \ \ -v\mathrm{d}v = \mathrm{d}\Psi\tag{4}$$

However, if I directly use the Jacobian method to find out the relation between $\mathrm{d}v$ and $\mathrm{d}\Psi$,

  1. $$\mathrm{d}\Psi = \det\left[\frac{\partial \Psi}{\partial v}\right]\mathrm{d}v$$ $$\frac{\partial \Psi}{\partial v} = v$$ $$\implies \mathrm{d}\Psi = v\mathrm{d}v\tag{5}$$

Clearly, eqn (4) and eqn (5) don't match. So, there is some inconsistency here in the way I am relating the differentials. Could someone shed some light on what I'm doing wrong, because I was able to prove the expression in the research paper using eqn (2).

1

There are 1 best solutions below

1
On

Assume that all quantities depend on each other. This amounts to not making the assumption that any partial derivatives are $0$. Since you seem interested in the differential $\mathrm{d}\Psi$, first solve for $\Psi$: $$ \Psi = \varepsilon + \frac{v^2}{2} $$ Now, use the (multivariable) chain rule: $$ \mathrm{d}\Psi = \frac{\partial\Psi}{\partial\varepsilon}\, \mathrm{d}\varepsilon + \frac{\partial\Psi}{\partial v}\, \mathrm{d}v = 1\, \mathrm{d}\varepsilon + v\, \mathrm{d}v $$ Your various formulas arise from assuming certain partial derivatives vanish. For $(2)$, (perhaps without realizing it) you assume that $\mathrm{d}\Psi = 0$, so $\mathrm{d}\varepsilon = -v\, \mathrm{d}v$. For $(3)$, you assume that $\mathrm{d}v = 0$, so $\mathrm{d}\Psi = \mathrm{d}\varepsilon$, etc.

Once you assume that more than one of these differentials vanish, then they all vanish, so your equations amount to the tautology $0 = 0$, which is of course equivalent to $0 = -0$.