3D Surfaces
In a previous lesson, we saw that 3D curves represent one-dimensional paths described by a single parameter $t$. In this lesson, instead, we will study surfaces, which have two degrees of freedom and are therefore described by two independent parameters.
Just as we did for curves, let’s begin by defining the concept of parametrization.
Let $D \subseteq \mathbb{R}^2$ a parametric domain, that is, a set of ordered pairs $(u,v)$. Moreover, let $$ r : D \to \mathbb{R}^3 $$ be a continuously differentiable function defined by: $$ r(u,v) = (x(u,v), y(u,v), z(u,v)) $$
If:
$$ \Sigma = \operatorname{Im}(r) = \left\{ r(u,v) : (u,v) \in D \right\} $$
then $r$ is called a parametrization of $\Sigma$. When $r$ is also a homeomorphism onto its image, it is called a topological chart.
Hence, we can give the following definition of parametric surface.
Let $\Sigma \subseteq \mathbb{R}^3$ and let $r : D \to \mathbb{R}^3$ be a parametrization of $\Sigma$. The pair $(\Sigma, r)$ is called a parametric surface in $\mathbb{R}^3$.
For example, if we consider the lateral surface of a cylinder of radius $R$, with its axis coincident with the $z$-axis and height equal to $H$, then we can parameterize it as follows:
$$ \begin{cases} x(u,v) = R \cos u\\ y(u,v) = R \sin u\\ z(u,v) = v \end{cases}, \qquad D = [0, 2\pi] \times [0, H] $$
The parametrization is:
$$ r(u,v) = (R \cos u, R \sin u, v) $$
Derivative of a Surface
The notion of tangent vector that we studied for curves can be extended to surfaces as well. However, since we now have two independent parameters, we need to evaluate two partial derivatives:
$$ r_u(u,v) = \frac{\partial r}{\partial u}, \qquad r_v(u,v) = \frac{\partial r}{\partial v} $$
which correspond to:
$$ r_u = \begin{pmatrix} \dfrac{\partial x}{\partial u}, &\dfrac{\partial y}{\partial u}, & \dfrac{\partial z}{\partial u} \end{pmatrix}, \quad r_v = \begin{pmatrix} \dfrac{\partial x}{\partial v}, &\dfrac{\partial y}{\partial v}, & \dfrac{\partial z}{\partial v} \end{pmatrix} $$
By fixing one of the parameters and allowing the other to vary, we obtain a curve contained within the surface. The vectors $r_u$ and $r_v$ are therefore tangent to these two coordinate curves.
Tangent Plane
Once the vectors $r_u$ and $r_v$ have been introduced, we notice that if they are linearly independent, they generate a tangent plane to the surface at a given point.
More formally, let $p = r(u_0, v_0)$. Then, the tangent plane at $p$ can be defined as:
$$ \Pi_p = \left\{ r(u_0, v_0) + \lambda r_u (u_0, v_0) + \mu r_v (u_0, v_0) : \lambda, \mu \in \mathbb{R} \right\} $$
This equation described the plane passing through $p = r(u_0, v_0)$ and generated by the two tangent vectors.
Furthermore, since the vector product of the two tangent vectors is orthogonal to both of them, we can introduce the normal vector $N$ defined as follows:
$$ \mathbf{N}(u,v) = r_u(u,v) \times r_v(u,v) $$
This vector will be fundamental when we study the concept of surface orientation.
Now we can define the concept of a regular surface.
A parametric surface $(\Sigma, r)$ with $r : D \to \mathbb{R}^3$ is regular if: $$ r_u(u,v) \times r_v(u,v) \neq 0, \quad \forall (u,v) \in D $$
We notice that this condition is satisfied precisely when the two tangent vectors are linearly independent.
In other words, if we can defined a tangent plane at every point, then there exists a nonzero normal vector and the surface has no singularities on the corresponding parameter domain. If this condition holds throughout the domain, the surface is regular.
Theory of Measure
Just as we calculate the infinitesimal length element of a curve using $ds = \Vert r'(t) \Vert \, dt$, we can do something similar for surfaces.
Let’s consider an infinitesimal area element $d\sigma$ and two infinitesimal variations $du$ and $dv$ of the parameters $u$ and $v$, respectively. The variation of the position on the surface, to first order, is determined by:
$$ d \mathbf r = r_u \, du + r_v \, dv $$
This results in a small parallelogram whose sides are approximately determined by $r_u \, du$ and $r_v \, dv$. Therefore, since the area of the parallelogram generated by two vectors is the magnitude of their vector product, we have:
$$ d\sigma = \Vert (r_u \, du) \times (r_v \, dv) \Vert $$
Hence:
$$ d\sigma = \Vert r_u \times r_v \Vert \, du \, dv $$
Therefore, the area of $\Sigma$ is given by:
$$ A(\Sigma) = \iint_{\Sigma} d \sigma = \iint_D \Vert r_u \times r_v \Vert \, du \, dv $$
where $D$ denotes the parameter domain of the parametrization $r$. Here, we assume that the parametrization covers the surface once, apart from possible repetitions along its boundary.
1st Fundamental Form
We now have all the prerequisite to determine lengths and angles on a regular surface. Let’s begin by defining the surface Jacobian:
$$ J_{\Sigma}(u,v) = \Vert r_u \times r_v \Vert $$
This quantity represents the scaling factor by which the parametrization transforms a small rectangle in the $uv$-plane into a small parallelogram tangent to the surface. Therefore, we have:
$$ d\sigma = J_{\Sigma} (u,v) \, du \, dv $$
Given the two vectors $r_u$ and $r_v$, we know that the infinitesimal displacement on the surface is:
$$ d \mathbf{r} = r_u \, du + r_v \, dv $$
Let’s now calculate the square of its length:
$$ ds^2 = d \mathbf{r} \cdot d \mathbf{r} = \langle d_r, d_r \rangle $$
This corresponds to:
$$ \begin{aligned} ds^2 &= (r_u \cdot r_u) \, du^2\\ &\quad+2(r_u \cdot r_v) \, du \, dv\\ &\quad+(r_v \cdot r_v) \, dv^2 \end{aligned} $$
Let’s now define:
$$ E = \langle r_u, r_u \rangle = \Vert r_u \Vert^2, \quad F = \langle r_u, r_v \rangle, \quad G = \langle r_v, r_v \rangle = \Vert r_v \Vert^2 $$
We conclude that:
$$ ds^2 = E \, du^2 + 2F \, du \, dv + G \, dv^2 $$
This is known as the Riemannian metric or first fundamental form of the surface; it described its local intrinsic geometry, such as the lengths, angles and areas.
In fact, starting from the surface Jacobian, we can derive the following identity:
$$ \Vert r_u \times r_v \Vert^2 = \Vert r_u \Vert^2 \cdot \Vert r_v \Vert^2 - (r_u \cdot r_v)^2 $$
Thus:
$$ J_{\Sigma}^2 = EG - F^2 \iff J_{\Sigma} = \sqrt{EG - F^2} $$
Therefore, the area element can equivalently be represented by:
$$ d\sigma = \sqrt{EG - F^2} \, du \, dv $$
and the total area as:
$$ A(\Sigma) = \iint_{\Sigma} d\sigma = \iint_D \sqrt{EG - F^2} \, du \, dv $$
Let’s look at a practical example. Consider the lateral surface of a cylinder of radius $R$ and height $H$ defined by:
$$ r(u,v) = (R \cos u, R \sin u, v) $$
with $0 \leq u \leq 2\pi$ and $0 \leq v \leq H$. Let’s start by determining the two partial derivatives: $$ r_u = (-R \sin u, R \cos u, 0), \qquad r_v = (0,0,1) $$
The vector product, instead, is:
$$ r_u \times r_v = (R \cos u, R \sin u, 0) $$
while its magnitude is:
$$ \begin{aligned} J_{\sigma} &= \sqrt{R^2 \cos^2 u + R^2 \sin^2 u}\\ &= \sqrt{R^2}\\ &= R \end{aligned} $$
Thus, $d\sigma = R \, du \, dv$.
Let’s verify this result by calculating the total lateral surface area of the cylinder. If our previous formal is correct, this area must correspond to the usual geometrical formula:
$$ A=2\pi RH $$
Indeed:
$$ \begin{aligned} A &= \int_0^{2\pi} \int_0^H R \, dv \, du\\ &= \int_0^{2\pi} RH \, du\\ &= 2 \pi R H \end{aligned} $$
Exactly what we expected!
2nd Fundamental Form
Before introducing the second fundamental form, we need to discuss another important concept: the orientation of a surface, which is required to define surface integrals over vector fields.
In the previous section, we saw that if the two tangent vectors are linearly independent, they generate the tangent plane to the surface at a given point. Furthermore, we saw that their vector product defined as:
$$ \mathbf{N}(u,v) = r_u \times r_v $$ is orthogonal to both of them and thus normal. If we now divide this normal vector by its norm, we obtain a normal versor, which is defined as:
$$ \mathbf{n}(u,v) = \frac{r_u \times r_v}{\Vert r_u \times r_v \Vert} $$
We notice that, at every regular point of the surface, we can define two different unit normal vectors, that is $\mathbf{n}$ and $\mathbf{-n}$, which are both orthogonal to the tangent plane but have opposite directions.
This ambiguity leads us to the concept of orientation of a curve.
A regular surface is called orientable if it is possible to choose a continuous unit normal vector at every point of the surface. Such a continuous choice of unit normal vectors defines an orientation of the surface.
For a parametrized surface, the choice:
$$ \mathbf{n}(u,v) = \frac{r_u \times r_v}{\Vert r_u \times r_v \Vert} $$
provides a local orientation. When a surface requires multiple parametrizations, their corresponding normal vectors must agree consistently on overlapping regions for a global orientation to exist.
It goes without saying that not all surfaces are orientable. The famous Möbius strip is not orientable, since if we move continuously along its central closed curve and follow the normal vector, after one complete traversal the normal returns with the opposite direction, that is:
$$ \mathbf n(b)=-\mathbf n(a) $$
Consequently, it is impossible to define a globally continuous unit normal vector over the entire Möbius strip.
Once an orientation has been chosen, we can determine a formula describing how the surface curves in space. To do this, we assume that the parametrization is twice continuously differentiable (that is $r \in C^2$), we evaluate the second derivatives of the parametrization, namely $r_{uu}$, $r_{uv}$ and $r_{vv}$ and project them onto the unit normal vector $\mathbf{n}$. We obtain:
$$ e = r_{uu} \cdot \mathbf{n}, \qquad f = r_{uv} \cdot \mathbf{n}, \qquad g = r_{vv} \cdot \mathbf{n} $$
Hence:
$$ \operatorname{II} = e \, du^2 + 2f \, du \, dv + g \, dv^2 $$
This is called the second fundamental form. Unlike the first fundamental form, which describes intrinsic properties such as lengths and angles, the second fundamental form describes how the surface bends in the surrounding three-dimensional space.