eleven9Silicon  /  Calc III  •  E&M Bridge

Surface Normal
Orientation

Converting \(z = f(x,y)\) to a level surface — and choosing the right sign

Multivariable Calc Gradient Flux Integrals E&M
The Setup

What Are We Even Doing?

You've got a surface defined as \(z = f(x,y)\) — a height function sitting over the \(xy\)-plane. But the gradient theorem wants a level surface of the form \(F(x,y,z) = 0\).

The move is simple: just rearrange. But the rearrangement you choose determines which way your normal vector points — and that sign controls the sign of flux.

⚡ Core Claim
You can always convert \(z = f(x,y)\) into a level surface. The two natural choices give normals pointing in opposite directions.
Level Surface Form

Two Ways to Write the Same Surface

Starting from:

$$z = f(x,y)$$
Option A — Downward
$$g(x,y,z) = f(x,y) - z = 0$$

Move \(z\) to the right. The \(z\)-coefficient in \(\nabla g\) will be \(-1\).

Option B — Upward ✓
$$h(x,y,z) = z - f(x,y) = 0$$

Move \(f\) to the right. The \(z\)-coefficient in \(\nabla h\) will be \(+1\).

Term Interrogation

Why call these "level surfaces"? A level surface is all the points where some function \(F(x,y,z)\) hits a constant value. Here that constant is zero — so the surface \(z = f(x,y)\) is the zero-level set of either \(g\) or \(h\). This is the bridge from "height functions" to the machinery of gradients.

The Math

Computing the Gradients

Recall: the gradient of \(F(x,y,z)\) is just the vector of partial derivatives. Each component asks "how fast does \(F\) change if I nudge in that direction?"

Option A: \(g = f(x,y) - z\)

Taking partials term by term:

$$\nabla g = \left( \frac{\partial g}{\partial x},\ \frac{\partial g}{\partial y},\ \frac{\partial g}{\partial z} \right) = \left( \frac{\partial f}{\partial x},\ \frac{\partial f}{\partial y},\ -1 \right)$$
Term Interrogation — the \(-1\)
Why is \(\partial g/\partial z = -1\)? Because \(g = f(x,y) - z\), and \(f\) doesn't depend on \(z\) at all. So the only \(z\)-dependence is the lone \(-z\) term. Differentiate that: \(\partial(-z)/\partial z = -1\). That's it.

Option B: \(h = z - f(x,y)\)

$$\nabla h = \left( -\frac{\partial f}{\partial x},\ -\frac{\partial f}{\partial y},\ +1 \right)$$

Now notice:

$$\nabla h = -\nabla g$$
🔑 Key Insight

The two level surfaces are identical as sets — they're the same curve in space. But their gradient vectors are exact negatives. This is like describing the same wall from the inside vs. the outside — same wall, opposite "outward" direction.

Geometric Intuition

See It: Normal Vectors on a Paraboloid

This canvas shows the surface \(z = x^2 + y^2\) (a paraboloid) viewed from the side. Toggle between the two gradient choices to see the normals flip.

ORIENTATION:
Why It Matters

The Flux Connection

Electric flux through a surface is:

$$\Phi = \iint_S \mathbf{E} \cdot d\mathbf{A}$$
Term Interrogation
  • \(\mathbf{E}\) — the electric field vector at each point on the surface
  • \(d\mathbf{A}\) — a tiny area element with a direction; it equals \(\hat{n}\,dA\)
  • \(\hat{n}\) — the unit normal, the thing we're choosing right now
  • \(\Phi\) — counts how much \(\mathbf{E}\) "punches through" the surface

The dot product \(\mathbf{E} \cdot d\mathbf{A}\) is sensitive to the direction of \(\hat{n}\). If you flip \(\hat{n}\), you flip \(\Phi\). So choosing between Option A and Option B isn't cosmetic — it changes the physical answer.

$$\hat{n}\text{ flips} \implies d\mathbf{A} \to -d\mathbf{A} \implies \Phi \to -\Phi$$
The Workhorse Formula

Which Form to Use (and Why)

For a surface \(z = f(x,y)\) with upward orientation, the standard area element is:

$$d\mathbf{A} = \left(-\frac{\partial z}{\partial x},\ -\frac{\partial z}{\partial y},\ 1\right)dx\,dy$$

This comes directly from \(\nabla h = \nabla(z - f)\) — Option B.

Geometric Reason

"Upward" means the normal has a positive \(k\)-component (pointing away from the \(xy\)-plane, toward \(+z\)). Option B gives a \(+1\) in the \(z\)-slot of the gradient, so it's the upward choice. Option A gives \(-1\), which is downward. Pick based on the problem's stated orientation.

The Rule of Thumb

Worked Example

Locking It In: \(z = x^2 + y^2\)

Write the two zero-forms:

Option A (Downward)
$$x^2 + y^2 - z = 0$$
$$\nabla g = (2x,\;2y,\;-1)$$

The \(z\)-component is \(-1\) → points downward.

Option B (Upward) ✓
$$z - x^2 - y^2 = 0$$
$$\nabla h = (-2x,\;-2y,\;+1)$$

The \(z\)-component is \(+1\) → points upward.

Physical Picture
On the paraboloid \(z = x^2 + y^2\), if you're standing on the surface and the normal points away from the origin (upward and outward), that's the upward orientation. This is the default for most E&M problems unless told otherwise.
Quick Sanity Check

Dimensional Analysis

Let's verify the normal vector components make dimensional sense.

$$[z] = \text{length}, \qquad [f(x,y)] = \text{length}$$
$$\left[\frac{\partial f}{\partial x}\right] = \frac{\text{length}}{\text{length}} = \text{dimensionless}$$
$$\nabla h = \underbrace{(-f_x,\;-f_y,\;1)}_{\text{all dimensionless}}$$
Why This Matters
Normal vectors are direction vectors — they should be dimensionless or have consistent units before normalization. The fact that all three components of \(\nabla h\) are dimensionless checks out. If you got a component with units of meters or meters\(^{-1}\), something went wrong.
The Big Picture

Mental Model — Lock This In

🧠 The Two-Step Mental Model
  1. Convert to a level surface so you can use the gradient: \(z = f(x,y) \;\longrightarrow\; F(x,y,z) = 0\)
  2. Choose the sign based on orientation. Gradient points away from the level set — so pick the sign that makes the \(z\)-component match the required orientation (\(+1\) for up, \(-1\) for down).

This same move shows up everywhere in E&M — closed surfaces for Gauss's Law, open surfaces for Stokes' theorem, and flux calculations via the Divergence Theorem. The sign discipline you build here pays off massively later.

🚀 What's Next

Full curved surface flux problem — apply this formula to a real \(\mathbf{E}\) field over a paraboloid or hemisphere.

Divergence Theorem connection — this orientation choice is exactly what makes the outward normal convention in the closed-surface integral: $$\oint\!\!\!\oint_S \mathbf{E} \cdot d\mathbf{A} = \frac{Q_{\text{enc}}}{\varepsilon_0}$$