Tutorials › Calculus III › Outward Normals on Piecewise Closed Surfaces

The Divergence Theorem · Tutorial 956 of 999

Outward Normals on Piecewise Closed Surfaces

Determine outward-pointing normals on common boundary pieces by checking how each piece bounds the solid, including when an inner wall faces into a cavity.

Advanced 10 min read

What You'll Learn

  • Use a defining inequality to select the outward direction of a normal.
  • Find normals to spheres and cylindrical walls from their defining equations.
  • Determine the correct normal direction on planar caps.
  • Orient the lateral surface of a cone using the solid’s inequality.
  • Reverse the normal direction on inner boundary components surrounding cavities.
  • Check an orientation using a point just inside or outside the solid.

Why the Direction of a Normal Matters

A surface has two possible normal directions at each smooth point: one points to one side of the surface, and the other points to the opposite side. For a closed surface bounding a solid, the outward normal is the direction that points from the solid into its exterior.

On a boundary made of several pieces, “outward” must be determined separately for each piece. A spherical part, a cylindrical wall, a planar cap, and a conical side do not all have the same normal formula. In addition, an inner boundary surrounding a cavity points into the cavity: that direction is outward from the material of the solid, even though it points toward the center of the overall region.

Definition: At a smooth point of the boundary of a solid, an outward unit normal is a vector of length \(1\) perpendicular to the boundary that points from the solid toward its exterior. For a piecewise smooth boundary, this direction is defined separately on each smooth piece; edges where pieces meet do not have a unique smooth normal.

A Reliable Test Using the Solid’s Inequality

The most dependable way to orient a normal is to describe which side of a boundary belongs to the solid. Suppose a smooth boundary piece is described by \(g(x,y,z)=0\), and the solid near that piece is described by \(g(x,y,z)\leq 0\). If \(\nabla g\neq \mathbf{0}\) on the piece, then \(\nabla g\) points toward increasing values of \(g\), hence out of the solid. The unit normal is \(\nabla g/\|\nabla g\|\).

If instead the solid is described locally by \(g(x,y,z)\geq 0\), the outward direction points toward decreasing values of \(g\), so the outward unit normal is \(-\nabla g/\|\nabla g\|\). The inequality, not the appearance of the surface equation alone, determines the sign.

Orientation rule: If a smooth boundary is \(g=0\) and \(\nabla g\neq\mathbf{0}\), then $$ \mathbf{n}_{\mathrm{out}}= \begin{cases} \dfrac{\nabla g}{\|\nabla g\|},&\text{if the solid is locally }g\leq 0,\\[6pt] -\dfrac{\nabla g}{\|\nabla g\|},&\text{if the solid is locally }g\geq 0. \end{cases} $$

The rule follows because the gradient is perpendicular to each level surface \(g=c\), and points in the direction in which \(g\) increases most rapidly. Moving a small distance in the \(\nabla g\) direction therefore moves toward larger \(g\)-values. If the solid is on the \(g\leq 0\) side, that is the exterior side. When using this rule, first choose \(g\) and state the inequality that describes the solid.

Common Normal Formulas

For a sphere centered at the origin with radius \(R\), take \(g=x^2+y^2+z^2-R^2\). The gradient is \((2x,2y,2z)\), so the outward unit normal on the sphere is \((x,y,z)/R\). For a sphere centered at \((a,b,c)\), replace the position vector by the displacement from its center.

For a circular cylindrical wall of radius \(R\) about the \(z\)-axis, take \(g=x^2+y^2-R^2\). Its gradient is \((2x,2y,0)\). On the wall, the outward unit normal from the solid inside the cylinder is \((x/R,y/R,0)\). A plane has a constant normal direction; its sign is selected by checking which side contains the solid. A conical side can be handled in the same way as any other implicit surface, but its gradient must be evaluated at the point in question.

Worked Example: Outer Sphere and Spherical Cavity

Worked Example: Outer Sphere and Spherical Cavity

Consider the spherical shell \(1\leq x^2+y^2+z^2\leq 9\). Find the outward unit normals on both spherical boundary components.

Write \(\rho^2=x^2+y^2+z^2\). The outer boundary is \(\rho^2=9\), and the solid lies on the side \(\rho^2-9\leq 0\). Set \(g_{\mathrm{out}}=\rho^2-9\). Then $$ \nabla g_{\mathrm{out}}=(2x,2y,2z), \qquad \|\nabla g_{\mathrm{out}}\|=2\sqrt{x^2+y^2+z^2}=6 $$ on the outer sphere. Therefore $$ \mathbf{n}_{\mathrm{out}}=\frac{(2x,2y,2z)}{6} =\frac{(x,y,z)}{3}. $$ At the point \((0,0,3)\), this gives \((0,0,1)\). Moving a small positive distance in that direction increases \(z\) beyond \(3\), so it leaves the shell.

The inner boundary is \(\rho^2=1\). Here the solid satisfies \(\rho^2-1\geq 0\), so the gradient points from the cavity into the solid, not outward from the solid. Thus $$ \nabla(\rho^2-1)=(2x,2y,2z), \qquad \|\nabla(\rho^2-1)\|=2 $$ on this sphere, and the outward normal for the shell is $$ \mathbf{n}_{\mathrm{out}}=-\frac{(2x,2y,2z)}{2}=-(x,y,z). $$ At \((0,0,1)\), the normal is \((0,0,-1)\), which points into the central cavity. Substituting any point with \(\rho=1\) into the formula gives a vector of length \(1\), confirming the unit length.

Worked Example: Cylindrical Wall, Bore, and Planar Caps

Worked Example: Cylindrical Wall, Bore, and Planar Caps

Consider the hollow cylindrical solid $$ 1\leq x^2+y^2\leq 4,\qquad 0\leq z\leq 3. $$ Determine the outward unit normal on every smooth boundary piece.

On the outer cylindrical wall, \(x^2+y^2=4\), the solid is on the side \(x^2+y^2-4\leq 0\). Its gradient is \((2x,2y,0)\), whose length on this wall is \(2\sqrt{x^2+y^2}=4\). Hence $$ \mathbf{n}_{\mathrm{out}}=\left(\frac{x}{2},\frac{y}{2},0\right). $$ For example, at \((2,0,1)\), this is \((1,0,0)\), pointing radially away from the solid.

On the inner cylindrical wall, \(x^2+y^2=1\), the solid satisfies \(x^2+y^2-1\geq 0\). The gradient \((2x,2y,0)\) points from the cavity into the material. It must be reversed. Its length is \(2\) on this wall, so $$ \mathbf{n}_{\mathrm{out}}=(-x,-y,0). $$ At \((1,0,1)\), this is \((-1,0,0)\), pointing into the hole.

The top cap is \(z=3\), with the solid on the side \(z-3\leq 0\). Since \(\nabla(z-3)=(0,0,1)\), its outward unit normal is \((0,0,1)\). The bottom cap is \(z=0\), with the solid on the side \(-z\leq 0\). Since \(\nabla(-z)=(0,0,-1)\), its outward unit normal is \((0,0,-1)\).

Each listed vector has length \(1\): on either cylindrical wall, \(x^2+y^2\) equals the square of that wall’s radius; the cap normals already have length \(1\). The outer and inner directions also pass the side test: a small radial step outward from the outer wall exits the solid, while a small radial step inward from the inner wall enters the hole. The cap directions similarly lead above \(z=3\) and below \(z=0\).

Worked Example: The Lateral Surface of a Cone

Worked Example: The Lateral Surface of a Cone

Let a solid cone be described by \(0\leq z\leq 2\) and \(x^2+y^2\leq z^2/4\). Find the outward unit normal on its lateral surface at \((1,0,2)\).

The lateral boundary is \(g(x,y,z)=x^2+y^2-z^2/4=0\), and the cone’s interior satisfies \(g\leq 0\). Therefore the outward direction is \(\nabla g\). Differentiating gives $$ \nabla g=\left(2x,2y,-\frac{z}{2}\right). $$ At \((1,0,2)\), this is \((2,0,-1)\), with length $$ \sqrt{2^2+0^2+(-1)^2}=\sqrt{5}. $$ The outward unit normal is consequently $$ \mathbf{n}_{\mathrm{out}}=\frac{(2,0,-1)}{\sqrt{5}}. $$

To check the direction, increase \(x\) slightly while holding \(y=0\) and \(z=2\). At \((1+\varepsilon,0,2)\), where \(\varepsilon>0\), $$ g=(1+\varepsilon)^2-\frac{2^2}{4} =1+2\varepsilon+\varepsilon^2-1 =2\varepsilon+\varepsilon^2>0. $$ That point is outside the cone, so the gradient points outward. The proposed normal has length \(\sqrt{4+1}/\sqrt{5}=1\), as required. The top planar cap \(z=2\), if included, has outward unit normal \((0,0,1)\) because the solid lies below it.

A Pitfall: Reusing the Outer-Wall Direction at a Cavity

A frequent error is to assign the same radial direction to both sides of a hollow cylinder. For the inner wall \(x^2+y^2=1\), one might write \((x,y,0)\) as the outward normal because the gradient of \(x^2+y^2-1\) points radially outward. But at \((1,0,z)\), that gives \((1,0,0)\), which points from the cavity into the material. It is not outward from the solid.

The correction is to use the inequality: the solid has \(x^2+y^2-1\geq 0\) near the inner wall. The outward direction is therefore the negative gradient, and the correct unit normal is \((-x,-y,0)\). A useful check is to imagine moving a short distance along the proposed normal: if that motion enters the solid rather than leaving it, the sign is wrong.

Orientation on Piecewise Boundaries

A closed boundary can combine many kinds of pieces. Determine the normal on each piece from the local inequality describing the solid, then check the direction against the solid’s interior. At a seam where two pieces meet, the boundary is not smooth and there is no single normal shared by the seam; the adjacent smooth pieces have their own normals. For flux calculations, the relevant normal is the one on the surface piece being considered, not a direction chosen by looking at the entire shape from a distance.

Takeaway: Derive the normal from a defining function and the solid-side inequality. On an outer boundary the gradient often points outward; on a cavity wall the solid-side inequality is reversed, so the normal points into the cavity.

Check Your Understanding

For each question, state the solid-side inequality before choosing the normal direction.

  1. For the ball \(x^2+y^2+z^2\leq 16\), what is the outward unit normal at \((0,0,4)\)?
  2. For a cylindrical cavity \(x^2+y^2=4\) inside a solid with \(x^2+y^2\geq 4\), which radial direction is outward from the solid?
  3. For the planar top boundary \(z=5\) of a solid lying below it, what is the outward unit normal?
  4. For the cone \(x^2+y^2\leq z^2\), \(z\geq 0\), which gradient gives the outward direction on the lateral boundary?
  5. How can a small displacement test reveal that a proposed normal has the wrong sign?