Calculus › Section N · Functions of several variables › Topic N.7
Tangent planes and normal lines
A curve has a tangent line; a surface deserves a tangent plane. This topic finds that plane, the normal line perpendicular to it, and the gradient's starring role as the normal direction.
From tangent lines to tangent planes
In one-variable calculus, the tangent line at a point is the straight line that best matches the curve there. A surface deserves the same treatment, but a single line cannot capture a surface that bends in two directions at once: a hillside rises differently walking north than walking east. The matching object is a plane. The tangent plane to a surface at a point is the flat plane that touches the surface there and shares its slopes in every direction. It gives the best linear model of the surface nearby, and the direction perpendicular to it is the normal direction, which physics constantly needs for forces and flows acting across a surface.
The tangent plane to a graph
For a surface \(z = f(x, y)\), slice it with the vertical plane \(y = y_0\). The slice is a curve, and its slope at \(x_0\) is the partial derivative \(f_x(x_0, y_0)\). Slicing with \(x = x_0\) instead gives a curve with slope \(f_y(x_0, y_0)\) in the y-direction. The tangent plane is the plane through \((x_0, y_0, z_0)\) with exactly those two slopes in those two directions:
When the partial derivatives exist and are continuous near the point, this plane matches the surface in every direction, not just the two slicing directions.
Worked example: a paraboloid
Find the tangent plane to \(z = x^2 + y^2\) at \((1, 2, 5)\). First check that the point is on the surface: \(1^2 + 2^2 = 5\). Next the partials: \(f_x = 2x\), which is 2 at the point, and \(f_y = 2y\), which is 4. The plane is
which simplifies to \(z = 2x + 4y - 5\). Checking the answer: at \((1, 2)\) the formula gives \(2 + 8 - 5 = 5\), the correct height. Checking it a second way, at \((1.1, 2)\) the plane gives \(5.2\) while the surface gives \(1.21 + 4 = 5.21\), very close, exactly as a tangent plane should be.
Level surfaces and the gradient
Many surfaces arrive as level sets: \(F(x, y, z) = c\), such as the sphere \(x^2 + y^2 + z^2 = 49\). For these, the normal direction comes from the gradient \(\nabla F = \langle F_x, F_y, F_z \rangle\), which is perpendicular to the surface at every point where it is not the zero vector. With the normal direction in hand, the tangent plane at \((x_0, y_0, z_0)\) follows at once:
This is the dot product of the normal vector with the vector from the point to \((x, y, z)\), set to zero, exactly the pattern for a plane in space.
Why the gradient is normal
Take any smooth curve \(\mathbf{r}(t) = \langle x(t), y(t), z(t) \rangle\) that lies entirely on the level surface and passes through the point. Along the curve, \(F\) never changes, because every point of the curve has the same value \(c\). Differentiating the constant equation \(F(\mathbf{r}(t)) = c\) with the chain rule for several variables gives \(\nabla F \cdot \mathbf{r}'(t) = 0\). The velocity \(\mathbf{r}'(t)\) can point along any direction tangent to the surface, so the gradient has zero dot product with every tangent direction. A vector perpendicular to all of the tangent plane's directions is a normal vector.
Worked example: a sphere
Find the tangent plane and normal line to the sphere \(x^2 + y^2 + z^2 = 49\) at \(P(2, 3, 6)\). Check \(P\) first: \(4 + 9 + 36 = 49\). With \(F = x^2 + y^2 + z^2\), the gradient is \(\nabla F = \langle 2x, 2y, 2z \rangle\), so at \(P\) the normal vector is \(\langle 4, 6, 12 \rangle\). The tangent plane is \(4(x - 2) + 6(y - 3) + 12(z - 6) = 0\), which simplifies to \(2x + 3y + 6z = 49\). Two checks: the point satisfies it, and the distance from the origin to the plane is \(\frac{49}{\sqrt{4 + 9 + 36}} = 7\), exactly the radius, so the plane touches the sphere without cutting through it. The normal line is \(x = 2 + 4t\), \(y = 3 + 6t\), \(z = 6 + 12t\); at \(t = -\frac{1}{2}\) it reaches the origin, the centre, as a sphere's normal line should.
The normal line to a graph
A graph \(z = f(x, y)\) fits the level-surface pattern with \(F(x, y, z) = f(x, y) - z = 0\). The gradient is \(\nabla F = \langle f_x, f_y, -1 \rangle\), evaluated at the point of contact. Notice the minus sign: \(z\) enters \(F\) with a coefficient of \(-1\). For the paraboloid example above, the normal vector at \((1, 2, 5)\) is \(\langle 2, 4, -1 \rangle\), and the normal line is \(x = 1 + 2t\), \(y = 2 + 4t\), \(z = 5 - t\). Every tangent-plane problem is really a normal-vector problem: find \(\nabla F\) at the point, and the plane writes itself.
The tangent plane as an approximation
The tangent plane is also the best flat model of the surface near the point of contact, the two-variable version of linear approximation. Moving a short step from \((x_0, y_0)\),
Suppose a drone's altitude sensor is calibrated against a hill modelled by \(h(x, y) = x^2 + y^2\) metres, with the drone hovering above \((1, 2)\), where the hill is 5 metres high. The tangent plane there is \(z = 2x + 4y - 5\). If the drone drifts to \((1.1, 2.05)\), the plane predicts a height of \(5 + 2(0.1) + 4(0.05) = 5.4\) metres; the true height is \(1.21 + 4.2025 = 5.4125\) metres. Within a few metres of the contact point, the flat model is good enough to navigate by.
Common mistakes to avoid
The most common slip is the sign of the z-component: for a graph, the normal vector is \(\langle f_x, f_y, -1 \rangle\), not \(\langle f_x, f_y, 1 \rangle\). Second, evaluate every partial derivative at the point of contact before building the plane; an \(x\) still sitting in your final equation means you forgot to substitute \(x_0\). Third, the formulas need the partials to exist and be continuous near the point. At a corner nothing works: for the cone \(z^2 = x^2 + y^2\) at the origin, \(\nabla F = \langle 2x, 2y, -2z \rangle\) is the zero vector, which points nowhere, so the cone has no tangent plane there.
Key takeaways
For a graph \(z = f(x, y)\) at \((x_0, y_0, z_0)\), the tangent plane is \(z = z_0 + f_x(x_0, y_0)(x - x_0) + f_y(x_0, y_0)(y - y_0)\). For a level surface \(F(x, y, z) = c\), the gradient \(\nabla F\) is normal to the surface at every point where it is nonzero, because \(F\) is constant along every curve on the surface; the tangent plane sets \(\nabla F \cdot \langle x - x_0, y - y_0, z - z_0 \rangle = 0\). The normal line runs along the normal vector, and near the point of contact the tangent plane serves as the surface's linear approximation.
Flashcards
Flashcard
z = z0 + fx(x0, y0)(x - x0) + fy(x0, y0)(y - y0)
Test your understanding
Question 1 of 5
What is the tangent plane to z = 3x² - y at the point (2, 1, 11)?
Question 2 of 5
A normal vector to the surface z = x²y at the point (1, 2, 2) is:
Question 3 of 5
What is the tangent plane to the sphere x² + y² + z² = 14 at the point (1, 2, 3)?
Question 4 of 5
A hill's height is modelled by h(x, y) = x² + y² metres. At (1, 2) the hill is 5 metres high. Using the tangent plane, what height does the model predict at (1.1, 2.05)?
Question 5 of 5
Why is ∇F perpendicular to the level surface F(x, y, z) = c at every point where ∇F is not zero?
Standards: Beyond AP
