Question: Let $f,g:\mathbb{R}^2 \to \mathbb{R}^2 $ be continously differentiable functions where $f\circ g$ is defined. Let $$f = (f_1,f_2)\quad\text{and}\quad g=(g_1,g_2)$$ where all $f_1,f_2,g_1,g_2:\mathbb{R}^2\to\mathbb{R}$ are functions. What is $$\frac{\partial}{\partial x_1} [f(g(x_1,x_2))]?$$

I think the derivative should be a function from $\mathbb{R}^2$ to $\mathbb{R}^2$ as well.

Let $x =(x_1,x_2).$ By multivariate chain rule, we have $$\frac{\partial}{\partial x_1} [f(g(x_1,x_2))] = \frac{\partial}{\partial x_1} [f(g_1(x),g_2(x))] = \frac{\partial f}{\partial g_1} \cdot \frac{\partial g_1}{\partial x_1} + \frac{\partial f}{\partial g_2} \cdot \frac{\partial g_2}{\partial x_1}.$$ I am not sure whether the equation above is correct.

I notice that the multiplication in the RHS of the equation above is scalar multiplication. This means that both $$\frac{\partial f}{\partial g_1}\quad\text{and}\quad \frac{\partial f}{\partial g_2}$$ is a vector in $\mathbb{R}^2.$ But I do not understand its meaning.


Given a point $p\in{\mathbb R}^2$ the chain rule says that $$d(f\circ g)(p)=df\bigl(g(p)\bigr)\circ dg(p)\ ,$$ in Jacobian matrix terms: $$J_{f\circ g}(p)=J_f(g(p)\bigr)\cdot J_g(p)\ .$$ This implies $${\partial\over\partial x_1}(f\circ g)(p)=d(f\circ g)(p).e_1=\bigl(df\bigl(g(p)\bigr)\circ dg(p)\bigr).e_1=df\bigl(g(p)\bigr)\bigl(dg(p).e_1\bigr)\ .$$ Now $dg(p).e_1$ is the first column of the matrix $J_g(p)$; therefore we obtain $${\partial\over\partial x_1}(f\circ g)(p)=\left[\matrix{f_{1.1}&f_{1.2}\cr f_{2.1}&f_{2.2}\cr}\right]_{g(p)}\left[\matrix{g_{1.1}\cr g_{2.1}\cr}\right]_p\ ,$$ where $f_{i.k}$ means that that the component $f_i$ of $f$ is partially differentiated with respect to $x_k$.

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.