Tangents and normals over parametric surfaces are defined very similarly to that of tangent planes for bivariate functions since a bivariate function is a type of parametric surface!
Let’s assume the parametric equation ϕ​ can be represented as:
ϕ​(s,t)=(x(s,t),y(s,t),z(s,t))
where x,y,z:R2→R are scalar functions.
We start by taking two lines c1​ and c2​, parallel to the s and t-axis respectively. Under some parametric equation ϕ​, these get mapped onto a surface S.
So the tangent vector to the curve c1​ is also tangent with respect to s, because c1​ is parallel to the s-axis. Hence, for a given value of s0​ we have:
Unlike with bivariate functions, our two tangent vectors are not parallel to the x and y axis! This means they don’t have to be orthogonal (perpendicular) to each other! That is, we don’t need to have them being at 90 degrees to each other or equivalently, we can have their dot product being non-zero: Ts​(s0​)â‹…Tt​(t0​)î€ =0
The unit normal vector is constructed the same way as the unit binormal vector in parametric curves. That is, it is constructed from the cross product of the two unit tangent vectors. However, unlike the unit binormal vector, we have two choices of normal vectors:
A cone can be a parameterised surface via the equation:$$
x = s \cos (t) \quad\quad y = s \sin(t) \quad\quad z = s \tag{t∈[0,2π],s≥0}
A torus (donut) of radius R and thickness r can be a parameterised surface via the equation:$$
x = (R + r \cos(\varphi))\cos(\theta) \quad\quad y = (R+ r \cos(\varphi))\sin(\theta) \quad\quad z = r \sin(\varphi) \tag{θ,φ∈[0,2π]}

>[!example] 1: Paraboloid as a Parametric Surface
>
> **Parameterise the surface of the paraboloid: $z = x^2 + y^2$**
>>[!info]- Solution
>> There are actually two common ways we can approach this problem:
>>
>> 1. Using [Cartesian coordinates](Cartesian%20Coordinates.md):
>> $$
>> x = s \quad\quad y = t \quad\quad z = s^2 + t^2 \tag{$s,t \in \mathbb{R}$}
>> $$
>>
>> 2. Using [cylindrical coordinates](Cylindrical%20Coordinates.md):
>> $$
>> x = \rho \cos(\phi)\quad\quad y = \rho \sin (\phi) \quad\quad z = x^2 + y^2 = \rho^2 \tag{$\rho \geq 0, \phi \in [0,2\pi)$}
>> $$
>
>>[!info]- Visuals
>>
>[!example] 2: Normal vectors to a Parametric Cone ^e2
>
> **Find a normal vector to the cone parameterised by:**
> $$
> \vec{\Phi}: x = s\cos (t) \quad\quad y = s \sin(t) \quad\quad z = s \tag{$s\geq 0, t \in [0,2\pi)$}
> $$
>>[!info]- Solution
>> The two tangent vectors (parallel to $s$ and $t$-axis) are given by:
>> $$
>> \vec{T}_{s} = \frac{ \partial \vec{\Phi} }{ \partial s } =( \cos (t), \sin(t), 1) \quad\quad \vec{T}_{t} = \frac{ \partial \vec{\Phi} }{ \partial t } = (-s \sin(t), s \cos (t), 0)
>> $$
>> (A) normal vector (well, technically a [vector field](Vector%20Field.md)) is given by:
>> $$
>> \vec{N} = \vec{T}_{s} \times \vec{T}_{t} = \det\begin{bmatrix}
>> \hat{i} & \hat{j} & \hat{k} \\
>> \cos(t) & \sin(t) & 1 \\
>> -s \sin(t) & s \cos(t) & 0 \\
>> \end{bmatrix}
>> =(-s \cos(t), -s \sin(t), s (\cos^2(t) + \sin^2 (t))
>> $$
>> Hence:
>> $$
>> \vec{N} (s,t) = (-s \cos (t), -s \sin(t), s)
>> $$
>>
>> Note: $\vec{N} = \vec{0}$ at $(s,t) = (0,0)$, which corresponds to the vertex of the cone. Here the normal vector being zero implies there is no [tangent plane](Tangent%20Plane.md), nor is the cone [differentiable](Bivariate%20Differentiability.md) at that point. Also, check the visuals to see that this choice of normal vector orients the cone 'inside'.
>
>>[!info]- Visuals
>>