1

According to cyclic relation in calculus, we have $ \frac{\partial x}{ \partial z} * \frac{\partial z}{ \partial x}=1 $, which makes sense as it resembles $\frac{a}{b}* \frac{b}{a}=1$.

Then for three variables we have $\frac{\partial x}{ \partial y} * \frac{\partial y}{ \partial z}*\frac{\partial z}{ \partial x}=-1$. How can we explain this conceptually? why the relation for three variables is not equal to 1? (I have read the proof in textbooks, however, I am unable to find a conceptual or geometrical explanation for this).

Ebi
  • 115
  • 1
  • 6

2 Answers2

1

There is not a simple geometric derivation of this fact. I will slightly reshape the discussion I wrote in the posts I already linked to you. The key idea is the implicit differentiation formula that introduces a negative sign: If $f(x,y)=c$ defines $y$ locally as a function $\phi$ of $x$, then $$\frac{dy}{dx} = -\frac{\frac{\partial f}{\partial x}}{\frac{\partial f}{\partial y}}.$$ You can think of this geometrically in terms of the tangent line to $y=\phi(x)$ at $(a,b)$. Working with linear approximation along that graph, we have $$0=\Delta f = \frac{\partial f}{\partial x}(a,b)\Delta x + \frac{\partial f}{\partial y}(a,b)\Delta y,$$ and so, indeed, $$\phi'(a) \approx \frac{\Delta y}{\Delta x} = -\frac{\frac{\partial f}{\partial x}(a,b)}{\frac{\partial f}{\partial y}(a,b)}.$$

Now, consider a level surface of a function of three variables. We start with $f(x,y,z) = c$ and assume this (locally) defines each of the variables as a differentiable function of the remaining two. It will now be important for clarity to introduce the notation — common in thermodynamics — of indicating with a subscript what variable is held fixed when we compute a partial derivative. From the two-variable formula I derived above, we get \begin{align*} \left(\frac{\partial z}{\partial x}\right)_y &= -\frac{\frac{\partial f}{\partial x}}{\frac{\partial f}{\partial z}}, \\ \left(\frac{\partial x}{\partial y}\right)_z &= -\frac{\frac{\partial f}{\partial y}}{\frac{\partial f}{\partial x}}, \\ \left(\frac{\partial y}{\partial z}\right)_x &= -\frac{\frac{\partial f}{\partial z}}{\frac{\partial f}{\partial y}}. \end{align*} Multiplying the three together, we get the result you want. Not very geometric, but it's all about implicit differentiation and tangent lines (planes).

I will remark that with partial derivatives it is generally false that $\dfrac{\partial z}{\partial x} = \dfrac 1{\dfrac{\partial x}{\partial z}}$. For example: Consider cartesian and polar coordinates, with $$x=r\cos\theta,\ y=r\sin\theta \qquad\text{and}\qquad r=\sqrt{x^2+y^2}, \ \theta = \arctan(y/x).$$ Then $\dfrac{\partial x}{\partial\theta} = -r\sin\theta = -y$, whereas $\dfrac{\partial\theta}{\partial x} = -\dfrac y{x^2+y^2}$. However, it is true, as you can check, that $\left(\dfrac{\partial\theta}{\partial x}\right)_y = \dfrac 1{\left(\dfrac{\partial x}{\partial\theta}\right)_y}$.

Ted Shifrin
  • 125,228
  • Thanks Ted, it is very clear now. Also, especially thanks for the point you made at the end of your answer, I had misunderstood the concept and now it makes a lot of sense. – Ebi May 01 '24 at 02:40
0

This is not a geometric proof. As you have asked to show it,

We assume that $z(x,y)$ is differentiable and injective $$dx=\bigg(\frac{\partial{x}}{\partial y}\bigg)dy + \bigg(\frac{\partial{x}}{\partial z}\bigg)dz$$ Similarly $$dy=\bigg(\frac{\partial{y}}{\partial x}\bigg)dx + \bigg(\frac{\partial{y}}{\partial z}\bigg)dz$$ Substituting $dy$ in $dx$ gives $$dx=\bigg(\frac{\partial{x}}{\partial y}\bigg)\bigg[\bigg(\frac{\partial{y}}{\partial x}\bigg)dx + \bigg(\frac{\partial{y}}{\partial z}\bigg)dz\bigg] + \bigg(\frac{\partial{x}}{\partial z}\bigg)dz$$ You can clearly see that using the cyclic relation you mentioned above, which also made sense for you, coefficient of $dx$ is $1$ in both sides of the equation. Thus

$$\bigg(\frac{\partial x}{\partial y}\bigg)\bigg(\frac{\partial{y}}{\partial z}\bigg)dz+\bigg(\frac{\partial{x}}{\partial z}\bigg)dz=0$$

Assuming $dz≠0$, $$\bigg(\frac{\partial x}{\partial y}\bigg)\bigg(\frac{\partial{y}}{\partial z}\bigg)+\bigg(\frac{\partial{x}}{\partial z}\bigg)=0$$ $$\implies\bigg(\frac{\partial x}{\partial y}\bigg)\bigg(\frac{\partial{y}}{\partial z}\bigg)\bigg(\frac{\partial{z}}{\partial x}\bigg)=-1$$

Gwen
  • 3,910
  • 6
  • 19
  • What is the codomain of $z$? If $\mathbb{R}$, then how can $z$ be simultaneously differentiable and injective? – Tom Apr 30 '24 at 03:16
  • @Tom I think I have a problem in my concept. We have started pd this year only. I thought if a function is differentiable, it should be bijective, and if it's bijective then it's injective. Maybe I am wrong. It will be really helpful if you explain my mistake – Gwen Apr 30 '24 at 03:29