Tutorials › Calculus III › Common Errors in Tangent Planes and Normal Lines

Home  /  Tutorials  /  Directional Derivatives and Gradients

Directional Derivatives and Gradients · Tutorial 397 of 999

Common Errors in Tangent Planes and Normal Lines

Use the gradient at the specified point to construct tangent planes and normal lines, and recognize when a zero gradient makes the usual formulas inconclusive.

Intermediate 10 min read

What You'll Learn

  • State the hypotheses needed for tangent-plane and normal-line formulas on a level surface.
  • Evaluate the gradient at the specified point before using it in an equation.
  • Track signs consistently when rewriting an implicit equation.
  • Check a proposed tangent plane and normal line against the point and surface geometry.
  • Interpret a zero gradient without assuming that the surface has no tangent plane.

Why These Errors Matter

A tangent plane and a normal line are built from the gradient of an equation defining a surface. The construction is compact, but it is easy to use a partial derivative with the wrong sign, evaluate it at the wrong point, or apply the formulas when the gradient is zero. Such errors can produce equations that look plausible while describing the wrong geometry.

The central habit is to keep three things distinct: the equation defining the surface, the point where the geometry is being examined, and the gradient evaluated at that point. We will use these distinctions to identify common mistakes and to check the resulting equations.

The Regular-Level Formulas

Definition: Let \(F:\mathbb{R}^3\to\mathbb{R}\) be continuously differentiable on a neighborhood of \(P=(x_0,y_0,z_0)\). The level surface through \(P\) is given by \(F(x,y,z)=F(P)\). The point \(P\) is a regular point of this level surface if \(\nabla F(P)\ne\mathbf{0}\).

At a regular point, the gradient is a normal vector to the level surface. Consequently, the tangent plane and a normal line are given by

$$ \nabla F(P)\mathbin{\cdot}\bigl((x,y,z)-P\bigr)=0, \qquad \boldsymbol{\ell}(t)=P+t\nabla F(P),\quad t\in\mathbb{R}. $$

Here \(\nabla F(P)=(F_x(P),F_y(P),F_z(P))\). Each partial derivative must be evaluated at \(P\) before it is used as a coefficient. The line may use any nonzero scalar multiple of \(\nabla F(P)\) as its direction vector.

Condition: These regular-level formulas apply when \(F\) is continuously differentiable near \(P\), \(F(P)\) is the level value, and \(\nabla F(P)\ne\mathbf{0}\). The nonzero-gradient condition is essential to this theorem. If it fails, the formulas do not provide a tangent plane or normal direction.

The continuous differentiability condition matters: it is the hypothesis used by the regular-level surface theorem to ensure that the level set is locally a smooth surface near a regular point. Merely knowing that \(F\) is differentiable at \(P\) is not enough to invoke that theorem.

Why the Gradient Gives the Normal Direction

The reason behind the formulas follows from the chain rule. Suppose a differentiable curve \(\mathbf{r}(t)\) lies on the level surface and passes through \(P\) at \(t=0\). Then \(F(\mathbf{r}(t))=F(P)\) along the curve. Differentiating at \(t=0\) gives

$$ \frac{d}{dt}F(\mathbf{r}(t))\bigg|_{t=0} = \nabla F(P)\mathbin{\cdot}\mathbf{r}'(0) = 0. $$

Thus every tangent vector \(\mathbf{r}'(0)\) is perpendicular to \(\nabla F(P)\). At a regular point of a smooth level surface, the tangent directions form a plane, and the nonzero gradient is normal to that plane. A plane through \(P\) with that normal has the equation \(\nabla F(P)\cdot((x,y,z)-P)=0\). The normal line goes through \(P\) in the gradient direction.

Worked Example: Keeping Signs Consistent

Worked Example: Keeping Signs Consistent

Find the tangent plane and a normal line to \(x^2+2y^2-z=0\) at \(P=(1,1,3)\).

First check the point: \(1^2+2(1)^2-3=1+2-3=0\), so \(P\) lies on the surface. Let \(F(x,y,z)=x^2+2y^2-z\). Then

$$ \nabla F(x,y,z)=(2x,4y,-1), \qquad \nabla F(1,1,3)=(2,4,-1). $$

The gradient is nonzero. Substitution into the tangent-plane formula gives

$$ 2(x-1)+4(y-1)-(z-3)=0, \qquad 2x+4y-z-3=0. $$

Therefore the tangent plane is \(z=2x+4y-3\). A normal line is

$$ (x,y,z)=(1,1,3)+t(2,4,-1). $$

Check the plane at \(P\): \(2+4-3=3\), as required. Also, the normal line passes through \(P\) at \(t=0\) and has direction \((2,4,-1)\), the plane's normal vector. Rewriting the defining equation as \(z-x^2-2y^2=0\) changes the gradient to \((-2,-4,1)\), which is the negative of the original gradient. That reverses the direction vector but leaves both the tangent plane and the geometric normal line unchanged.

Common Error: Changing Only Some Signs

Multiplying an entire defining equation by \(-1\) changes every component of its gradient by \(-1\). That is harmless for the tangent plane and normal line: a normal vector and its negative describe the same plane and line. The error is to change only some signs.

For the surface in the example, the correct gradient from \(F=x^2+2y^2-z\) is \((2x,4y,-1)\). A mistaken calculation might use \((2x,4y,1)\), treating the derivative of \(-z\) as \(+1\). At \(P\), this gives the incorrect normal \((2,4,1)\) and the false plane \(2(x-1)+4(y-1)+(z-3)=0\).

This false plane does pass through \(P\), so checking only the point is not enough. The surface can be written as \(z=x^2+2y^2\), and a tangent direction at \(P\) is \((1,0,2)\): moving \(x\) at unit rate changes \(z\) at rate \(2x=2\). The proposed normal is not perpendicular to that direction, since \((2,4,1)\cdot(1,0,2)=2+2=4\ne0\). The sign error has therefore produced a plane that is not tangent.

Common Error: Substituting the Wrong Point

The gradient in the formulas is not a variable vector: its components are numbers evaluated at the specified point. The following example makes the evaluation visible.

Worked Example: Evaluating the Gradient at the Given Point

Find the tangent plane and a normal line to \(xy+z^2-5=0\) at \(P=(1,2,1)\).

Check the point: \((1)(2)+1^2-5=2+1-5=-2\). This calculation reveals that the point does not lie on the stated surface. The requested tangent plane at this point is therefore not defined for this level surface.

This is an important first check: do not proceed with a formula just because a point has been supplied. To illustrate the correct calculation for a point on this surface, take \(Q=(1,2,\sqrt{3})\), since \(1\cdot2+(\sqrt{3})^2-5=2+3-5=0\). For \(F(x,y,z)=xy+z^2-5\),

$$ \nabla F(x,y,z)=(y,x,2z), \qquad \nabla F(Q)=(2,1,2\sqrt{3}). $$

The gradient is nonzero. Thus the tangent plane at \(Q\) is

$$ 2(x-1)+(y-2)+2\sqrt{3}(z-\sqrt{3})=0, \qquad 2x+y+2\sqrt{3}z-9=0. $$

A normal line is \((x,y,z)=(1,2,\sqrt{3})+t(2,1,2\sqrt{3})\). Check the plane at \(Q\): \(2+2+6-9=1\) is not zero—so the expansion must be checked. Expanding the uncombined expression correctly gives \(2x-2+y-2+2\sqrt{3}z-6=0\), hence \(2x+y+2\sqrt{3}z-10=0\). The constant is \(-10\), not \(-9\). Substituting \(Q\) gives \(2+2+6-10=0\), confirming the result.

This example illustrates two checks worth making in order: verify that the point is on the surface, and substitute it into the final plane. The gradient components must also be evaluated at that same point. For instance, \(F_x=y\), so at \(Q\), \(F_x(Q)=2\), not the first coordinate \(1\).

Zero Gradient: What You Can and Cannot Conclude

If \(\nabla F(P)=\mathbf{0}\), the regular-level formulas cannot be used: there is no nonzero gradient vector to serve as the proposed normal. This does not prove that the geometric surface has no tangent plane. It says only that this particular equation and the regular-level test do not determine one at \(P\).

Worked Example: A Zero Gradient from a Redundant Equation

Consider the level surface \((z-x^2-y^2)^2=0\) at \(P=(1,0,1)\). Define \(F(x,y,z)=(z-x^2-y^2)^2\). Since \(z-x^2-y^2=0\) at \(P\), the chain rule gives \(\nabla F(P)=\mathbf{0}\). The regular-level formula cannot supply a normal here.

But for real numbers, a square is zero exactly when its base is zero. Thus this level set is precisely the paraboloid \(z-x^2-y^2=0\). Use that simpler defining function \(G(x,y,z)=z-x^2-y^2\). Its gradient is

$$ \nabla G(x,y,z)=(-2x,-2y,1), \qquad \nabla G(1,0,1)=(-2,0,1). $$

The tangent plane to this regular level surface is \(-2(x-1)+(z-1)=0\), or \(z-1=2(x-1)\). A normal line is \((x,y,z)=(1,0,1)+t(-2,0,1)\). The plane contains \(P\), and its normal vector is \((-2,0,1)\), as required. The zero gradient arose from squaring a defining function, not from the absence of tangent geometry.

Zero gradient can also occur at a point where the level set is not a surface at all. For example, \(x^2+y^2+z^2=0\) consists only of the origin, and its gradient is zero there. A single point does not have a two-dimensional tangent plane as a regular surface. These contrasting cases show why a zero gradient requires further geometric analysis rather than an automatic conclusion.

A Reliable Error-Checking Routine

1
Check membership.
Substitute the proposed point into the defining equation. If it is not on the level surface, it is not a point where that surface has the requested tangent plane.
2
Differentiate the chosen defining function.
Keep each partial derivative's sign. If the whole defining equation is multiplied by a nonzero constant, all gradient components change by that same factor.
3
Evaluate at the point.
Turn the gradient into a fixed vector by substituting all three point coordinates into its components.
4
Check regularity and construct.
Confirm that the function is continuously differentiable near the point and that the evaluated gradient is nonzero before using the regular-level formulas.
5
Verify the result.
The plane must contain the point, and its normal must be parallel to the direction vector of the normal line. When possible, check perpendicularity against a known tangent direction.

Why the Checks Are Useful

A tangent-plane equation can pass through the correct point and still have the wrong orientation, as the sign-error example showed. Conversely, a correct gradient direction cannot repair a point that is not on the surface. The checks test separate parts of the construction: membership tests the location, gradient evaluation tests the coefficients, and perpendicularity tests the geometry.

When the gradient is zero, pause rather than forcing a zero vector into a normal-line formula. The equation may be poorly chosen, as in the squared example; the level set may instead fail to be a smooth surface. Simplify or re-express the equation when that is justified, then examine the actual level set and its local geometry. Do not claim a regular tangent plane from the gradient theorem until its hypotheses hold.

Check Your Understanding

Answer each question by identifying which condition or check is relevant.

  1. Why does a tangent-plane formula use \(\nabla F(P)\), rather than the unevaluated vector \(\nabla F(x,y,z)\)?
  2. If a defining function is multiplied throughout by \(-1\), what happens to the gradient and to the geometric tangent plane?
  3. What should you check before calculating a tangent plane at a supplied point?
  4. Does \(\nabla F(P)=\mathbf{0}\) prove that the level set has no tangent plane? Explain.
  5. What hypotheses allow the regular-level tangent-plane and normal-line formulas to be used?

Previous

Next

Back to Tutorials