Gradient
The gradient in the direction of \(\vec{e_{q_1}}\) is
\[\label{eq:dirgradcurvi} \nabla_{q_i} =
\frac{\partial}{\partial s_i} = \frac{1}{h_{q_i}}
\frac{\partial}{\partial q_i}\]
thus, the total gradient operator is
\[\label{eq:gradincurvi}
\nabla = \sum_i \vec{e_{q_i}} \frac{\partial}{\partial s_i} = \sum_i \vec{e_{q_i}} \frac{1}{h_{q_i}} \frac{\partial}{\partial q_i}\]
Divergence
We can define the divergence from the gradient, since
\[\nabla \cdot \vec{A} = \nabla \cdot \left( \sum_j A_j \vec{e_{q_j}}
\right)\]
clearly we need \(\nabla \cdot \vec{e_{q_j}}\) to continue; from equation ([eq:dirgradcurvi]),
\[\nabla \cdot q_j = \sum_i \frac{1}{h_{q_i}} \frac{\partial
q_j}{\partial q_i} = \sum_i \vec{e_{q_i}} \frac{1}{h_{q_i}}
\delta_{ij} = \vec{e_{q_j}} \frac{1}{h_{q_j}}\]
Thus,
\[\vec{e_{q_j}} = h_{q_j} \nabla q_j\]
The basis vectors form a right-handed set, so,
\[e_{q_1} \times e_{q_2} = e_{q_3}\]
and
\[\nabla q_1 \times \nabla q_2 =
\frac{\vec{e_{q_3}}}{h_{q_1}h_{q_2}}\]
Considering that \(\curl(\grad) = 0\),
\[\begin{split}\begin{aligned}
\nabla \cdot (\nabla_{q_1} \times \nabla_{q_2}) &=
\nabla q_2 \cdot (\nabla \times \nabla q_1) - \nabla q_1 \cdot (
\nabla \times \nabla q_2) \\ &= 0\end{aligned}\end{split}\]
So
\[\nabla \cdot \left( \frac{\vec{e_{q_3}}}{h_{q_1} h_{q_2}} \right) =
0\]
and the same argument applies for cyclic permutations,
\[\nabla \cdot \left( \frac{\vec{e_{q_3}}}{h_{q_1} h_{q_2}} \right) =
\nabla \cdot \left( \frac{\vec{e_{q_1}}}{h_{q_2} h_{q_3}} \right) =
\nabla \cdot \left( \frac{\vec{e_{q_2}}}{h_{q_3} h_{q_1}} \right) =
0\]
Then returning to
\[\begin{split}\begin{aligned}
\nabla \cdot \vec{A} &= \nabla \cdot \left( \sum_j A_j \vec{e_{q_j}} \right) &\\
&= \nabla \cdot \left\{ [h_{q_1}h_{q_2}A_1] \left[
\frac{\vec{e_{q_1}}}{h_{q_2} h_{q_3}} \right] \right.
&+& [h_{q_3}h_{q_1}A_2] \left[ \frac{\vec{e_{q_2}}}{h_{q_3} h_{q_1}} \right] \\
&&+&\left.[h_{q_1} h_{q_2}A_3] \left[ \frac{\vec{e_{q_3}}}{h_{q_1} h_{q_2}} \right] \right\} \\
&= \left[ \frac{\vec{e_{q_1}}}{h_{q_2} h_{q_3}} \right] \cdot \nabla
[h_{q_1}h_{q_2}A_1]
&+&\left[ \frac{\vec{e_{q_2}}}{h_{q_3} h_{q_1}} \right] \cdot \nabla [h_{q_3}h_{q_1}A_2]\\
&&+&\left[ \frac{\vec{e_{q_3}}}{h_{q_1} h_{q_2}} \right]\cdot \nabla
[h_{q_1}h_{q_2}A_3]\end{aligned}\end{split}\]
Since
\[\vec{e_{q_i}} \cdot \nabla = \frac{1}{h_{q_i}}
\frac{\partial}{\partial q_i}\]
,
\[\begin{split}\label{eq:divincurvi}
\begin{split}
\nabla \cdot \vec{A} = \frac{1}{h_{q_1}h_{q_2}h_{q_3}} \left( \frac{\partial}{\partial q_1} (h_{q_2} h_{q_3} A_1) \right. \\
+ \left. \frac{\partial}{\partial q_2} (h_{q_3} h_{q_1} A_2) +
\frac{\partial}{\partial q_3} (h_{q_2} h_{q_1} A_3) \right)
\end{split}\end{split}\]
Curl
The curl can be derived from earlier results,
\[\begin{split}\begin{aligned}
\nabla \times \vec{A} &=&& \nabla \times \left( \sum_j A_j \vec{e_{q_j}} \right) \\
&=&& \nabla \times \left(
\begin{matrix}
[h_{q_1}A_1] \qty[\frac{\vec{e_{q_1}}}{h_{q_1}} ] &+
[h_{q_2}A_2] \qty[ \frac{\vec{e_{q_2}}}{h_{q_2}} ] \\ &+
[h_{q_3}A_3] \qty[ \frac{\vec{e_{q_3}}}{h_{q_3}} ]
\end{matrix}
\right) \\
&= && - \left[ \frac{\vec{e_{q_1}}}{h_{q_1}} \right] \times \nabla
(h_{q_1}A_1)
- \left[ \frac{\vec{e_{q_2}}}{h_{q_2}} \right] \times \nabla (h_{q_2}A_2) \\
&&& - \left[ \frac{\vec{e_{q_3}}}{h_{q_3}} \right] \times \nabla
(h_{q_3}A_3)\end{aligned}\end{split}\]
so,
\[\begin{split}\label{eq:curlincurvi}
\nabla \times \vec{A} =
\frac{1}{h_{q_1}h_{q_2}h_{q_3}}
\begin{vmatrix}
h_{q_1} \vec{e_1} & h_{q_2} \vec{e_2} & h_{q_3} \vec{e_{3}} \\
\frac{\partial}{\partial q_1} & \frac{\partial}{\partial q_2} & \frac{\partial}{\partial q_3} \\
h_{q_1} A_1 & h_{q_2} A_2 & h_{q_3} A_3
\end{vmatrix}\end{split}\]