Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I'm having difficulties with this exercise (from Elon LIMA's Curso de Análise, Vol. 2):

$f:U\longrightarrow\mathbb{R}$ is function, differentiable on the open set $U\subset\mathbb{R}^n$. Let $\{v_1,...,v_n\}$ be an arbitrary basis of $\mathbb{R}^n$ and $g^{ij}:=\left<v_i,v_j\right>$. Show that grad$f(x)$ in this basis is $$\textrm{grad} f(x) = \sum_i(\sum_j g^{ij}\frac{\partial f}{\partial v_j})v_i$$ where $\frac{\partial f}{\partial v_j}$ is the directional derivative of $f$ along the vector $v_j$.

I tried starting with the RHS with $\frac{\partial f}{\partial v_j}=\left<\textrm{grad}f(x),v_j\right>$ and arriving at $\sum_i \left<\textrm{grad}f(x),e_i\right>e_i $ which is the "canonical" (see Note below) expression for the grad; writing each $v_i$ as $v_i=\sum_j \left<v_i,e_j\right>$ I come down to a sum of the form $\sum_{i,j}\left<v_j,e_k\right>g^{ji}\left<v_i,e_l\right>$ which I don't recognize and doesn't seem to simplify.

Am I approaching it incorrectly? How must this problem be tackled? I'm obviously missing something, or misinterpreting things.

NOTE: In the textbook, grad$f(x)$ is defined as $\sum_i \frac{\partial f}{\partial x_i}(x)e_i$ where $\frac{\partial f}{\partial x_i}(x)$ is the usual i-th partial derivative at the point $x$.

share|cite|improve this question
Let $w$ be the graduate vector (clearly this equation has nothing to do with derivatives), it's just linear algebra. You want to show that $w=\sum_i v_i \sum_j \langle v_i,v_j\rangle \langle w, v_j\rangle $. Can this be true? No, because if all vectors $v_i$ are multiplied by the same constant $\lambda$, the right hand side gets multiplied by $\lambda^4$. Hence, you have a wrong expression for $g^{ij}$, probably missing matrix inverse. – user31373 Jul 14 '12 at 18:12
I meant the gradient vector, of course (too late to edit). The graduate committee work is getting to me. :( – user31373 Jul 14 '12 at 19:50

Your definition of $g^{ij}$ is incorrect. We have that $$g_{ij} = \langle v_i, v_j \rangle$$ and then $g^{ij}$ is the $(i,j)$-entry of the inverse matrix $(g_{ij})^{-1}$. With this correction, your approach is fine, except you should have $$v_i = \sum_{j = 1}^n \langle v_i, e_j \rangle e_j$$ (you didn't have the last $e_j$, perhaps it was a typo). Starting with the right-hand side, the computation goes like this: \begin{align*} \sum_{i, j = 1}^n g^{ij} \frac{\partial f}{\partial v_j} v_i & = \sum_{i, j = 1}^n g^{ij} \langle \operatorname{grad} f, v_j \rangle v_i \\ & = \sum_{i,j = 1}^n g^{ij} \left\langle \operatorname{grad} f, \sum_{k = 1}^n \langle v_j, e_k \rangle e_k \right\rangle \sum_{l = 1}^n \langle v_i, e_l \rangle e_l \\ & = \sum_{i,j,k,l = 1}^n g^{ij} \left\langle \operatorname{grad} f, \langle v_j, e_k \rangle e_k \right\rangle \langle v_i, e_l \rangle e_l \\ & = \sum_{i,j,k,l = 1}^n g^{ij} \langle v_i, e_l \rangle \langle v_j, e_k \rangle \langle \operatorname{grad} f, e_k \rangle e_l \\ & = \sum_{i,j,k,l = 1}^n g^{ij} \delta_{kl} (v_i)_l (v_j)_k \langle \operatorname{grad} f, e_k \rangle e_l \\ & = \sum_{i,j,k = 1}^n g^{ij} (v_i)_k (v_j)_k \langle \operatorname{grad} f, e_k \rangle e_k \\ & = \sum_{i,j = 1}^n g^{ij} g_{ij} \sum_{k = 1}^n \langle \operatorname{grad} f, e_k \rangle e_k \\ & = \sum_{i,j = 1}^n \delta_{ij} \sum_{k = 1}^n \langle \operatorname{grad} f, e_k \rangle e_k \\ & = \sum_{k = 1}^n \langle \operatorname{grad} f, e_k \rangle e_k \\ & = \operatorname{grad} f. \end{align*}

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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