Assume that $\Omega$ is a smooth domain with $C^2$ boundary $\Gamma$.
# Local Extension Operator
Let $\Gamma_{\epsilon}$ be a sufficiently small neighborhood of $\Gamma$. Define a unique projection operator $\mathcal{P}$ such that for $y\in\Gamma_{\epsilon}$ we have $\mathcal{P}(y)\in\Gamma$. Let
$ \delta(y) := |y-\mathcal{P}(y)|. $
The line $y-P(y)$ is directed along the normal to $\Gamma$ at $y$. When $y$ is in the exterior of $\Omega$ it holds that
$ \nabla \delta(y) = n(\mathcal{P}(y)). $
If $y$ is in the interior of $\Omega$ it holds that
$ \nabla \delta(y) = -n(\mathcal{P}(y)). $
For any point $y\in\Gamma_{\epsilon}$ we can write $y = \mathcal{P}(y) + sn(\mathcal{P}(y))$ with $s=\pm\delta(y)$ (depending whether $y$ is in the interior or exterior of $\Omega$.
For any function $u$ defined on $\Gamma$ can define a lifting operator by
$ \tilde{u}(y) = u(\mathcal{P}(y)). $
We can define a gradient field $n(y)$ as $n(y):=\nabla s(y)$ for any $s\in\Gamma_{\epsilon}$.
Now define the family $\Gamma_s$ of parallel surfaces as
$ \Gamma_s = \{y; y = x+sn(x); x\in\Gamma\}. $
The field $n(y)$ is the field of normals to the surface $\Gamma_s$.
# Surface Differential operators of scalar functions
## Surface Gradient
The surface gradient $\nabla_{\Gamma}u$ is defined as
$ \nabla_{\Gamma}u = \nabla\tilde{u}_{\Gamma}. $
## Tangential curl
The tangential curl is defined as
$ \overrightarrow{\text{curl}}_{\Gamma} = \text{curl}\,(\tilde{u}n)_{\Gamma}. $
Since $n(y)$ is a gradient field it follows that $\text{curl}\, n = 0$ and therefore, using $\text{curl}\,(u\overrightarrow{v}) = \nabla u\times \overrightarrow{v} + u\,\text{curl}\,\overrightarrow{v}$,
$ \overrightarrow{\text{curl}}_{\Gamma} = \nabla_{\Gamma}u\times n. $
# Lifting of tangential vector fields
Let $\mathcal{R}_s(y) := \nabla n = D^2s$. This is called the curvature operator. One can show that $\mathcal{R}_s(y)n(y) = 0$. Hence, the normals are in the kernel of the curvature operator.
With this operator one can for a tangential vector field $v$ define the lifted field
$ \tilde{v}(y) = v(x) -s\mathcal{R}_s(y)v(x), $
which is tangential at each $\Gamma_s$. For the tangential derivative at $\Gamma_s$ we obtain
$ \nabla_{\Gamma_s}\tilde{u} = (I -s\mathcal{R}_s)\nabla_{\Gamma}u. $
# Surface Divergence
The surface divergence of the tangential vector field $v$ is defined as
$ \text{div}_{\Gamma} = \text{div}\,\tilde{v}|_{\Gamma}, $
that is, we take the divergence of the extension and restrict it to $\Gamma$.
# Surface curl
We define the surface curl of a tangential vector field $v$ as
$ \text{curl}_{\Gamma}\,v = (n\cdot \text{curl}\,\tilde{v})|_{\Gamma}. $
# Function Spaces
The function spaces for boundary differential operators can be found in [[De Rham Sequence for Trace Spaces]].
# References
Nédélec, _Acoustic and Electromagnetic Equations_, Springer, Chapter 2.5.6