Parametric Differentiation

A circle of radius r centred at the origin satisfies x^2 + y^2 = r^2. That is one equation in two variables — an implicit description. You have already seen how to differentiate it implicitly.

But there is another way to describe the same circle. A point walks around it at a steady pace, and at time t its coordinates are

This is a parametric description: instead of one equation tying x and y together, you have two separate equations, each giving one coordinate as a function of a third variable t — the parameter. As t runs from 0 to 2\pi, the point traces out the entire circle.

Parametric descriptions are natural whenever motion is involved. A ball thrown at an angle, a planet orbiting a star, a point sliding along a spiral — in each case, you have x(t) and y(t) separately, and together they trace a curve. The parameter t is often time, but it does not have to be. It could be an angle, an arc length, or any variable that drives the point along the curve.

The question is: can you still find the slope dy/dx of such a curve? The answer is yes, and the method is an elegant application of the chain rule.

Why parametric curves need their own derivative formula

With an explicit equation like y = x^2, finding dy/dx is straightforward — differentiate directly. With an implicit equation like x^2 + y^2 = r^2, you use implicit differentiation — differentiate both sides with respect to x, treating y as a function of x.

But parametric equations present a different situation. You have x as a function of t, and y as a function of t, but you do not have y written as a function of x. In principle, you could eliminate t — from x = r\cos t and y = r\sin t you can recover x^2 + y^2 = r^2 — but this elimination is often messy or impossible. For the cycloid (which you will see in Example 1), there is no neat closed-form relationship between x and y that does not involve t.

So you need a way to compute dy/dx directly from the parametric equations, without eliminating t first. The chain rule provides exactly that.

The key idea: dividing two rates

Both x and y depend on t. So dx/dt is the rate at which x changes with respect to t, and dy/dt is the rate at which y changes with respect to t.

Think of it concretely. Suppose at some moment, x is increasing at 3 units per second and y is increasing at 6 units per second. Then for each unit x moves, y moves twice as much. The slope dy/dx is 6/3 = 2.

That reasoning generalises immediately. If you know the rate at which y moves with respect to t, and the rate at which x moves with respect to t, then the rate at which y moves with respect to x is simply their ratio — just as kilometres per hour divided by litres per hour gives kilometres per litre.

To make this precise: over a tiny change \Delta t in the parameter, the point moves by \Delta x \approx (dx/dt)\,\Delta t horizontally and \Delta y \approx (dy/dt)\,\Delta t vertically. The slope of the curve over this tiny step is

The \Delta t's cancel, leaving a ratio that depends only on the derivatives with respect to t. In the limit, this becomes exact.

Why this works — the chain rule argument. By the chain rule, \dfrac{dy}{dt} = \dfrac{dy}{dx} \cdot \dfrac{dx}{dt}. Divide both sides by dx/dt:

That is the entire derivation. The formula is not a new rule — it is the chain rule, rearranged. The Leibniz notation makes it look like you are cancelling the dt's, and in this case that cancellation gives the correct answer. (It does not always work so cleanly, which is why we derive it from the chain rule rather than treating dy/dx as a literal fraction.)

The first worked example

The second derivative in parametric form

Knowing the slope is one thing. Knowing how the slope is changing — whether the curve is bending toward or away from the axis — requires the second derivative d^2y/dx^2.

The temptation is to compute \dfrac{d^2 y}{dx^2} by differentiating the numerator and denominator of dy/dx separately with respect to t and dividing again. That is wrong. Here is the correct approach.

You already have dy/dx as a function of t. Call it \phi(t) for brevity. The second derivative is the derivative of dy/dx with respect to x:

But dy/dx = \phi(t) is written in terms of t, not x. So use the parametric formula again — to differentiate a function of t with respect to x, divide its t-derivative by dx/dt:

Derivation of the expanded form. Start with dy/dx = \dot{y}/\dot{x}. Differentiate this with respect to t using the quotient rule:

Now divide by dx/dt = \dot{x}:

The \dot{x}^2 from the quotient rule and the extra \dot{x} from the parametric formula combine to give \dot{x}^3 in the denominator.

The common mistake. Many students write \dfrac{d^2 y}{dx^2} = \dfrac{d^2y/dt^2}{d^2x/dt^2} = \dfrac{\ddot{y}}{\ddot{x}}. This is incorrect. The second derivative d^2y/dx^2 is not the ratio of the second derivatives with respect to t. The ratio formula only works for the first derivative.

Why the naive formula fails — a quick check. For the circle x = \cos t, y = \sin t, the naive formula would give \ddot{y}/\ddot{x} = (-\sin t)/(-\cos t) = \tan t. But the correct second derivative (computed by the proper formula) is -1/\sin^3 t, which is not \tan t. At t = \pi/2, the naive formula gives \tan(\pi/2) = \text{undefined}, while the correct value is -1. The naive formula is not just slightly off — it gives qualitatively wrong answers.

A second example, with curvature

Applications to curves

The parametric derivative formulas unlock several important geometric computations.

Tangent and normal lines

At the point corresponding to parameter value t = t_0, the curve passes through (x_0, y_0) = (f(t_0), g(t_0)) with slope m = dy/dx evaluated at t_0. The tangent line is

and the normal line (perpendicular to the tangent) is

provided m \neq 0 and m is finite.

For the parabola x = at^2, y = 2at, the tangent at the point (at_0^2, 2at_0) has slope 1/t_0, so the tangent line is:

Multiply through by t_0: t_0 y - 2at_0^2 = x - at_0^2, which simplifies to t_0 y = x + at_0^2. This is the standard tangent equation for a parabola at the point t_0 — a result you would find much harder to derive without parametric differentiation.

Horizontal and vertical tangents

A tangent is horizontal when dy/dx = 0, which happens when dy/dt = 0 and dx/dt \neq 0.

A tangent is vertical when dy/dx is undefined (in the sense of \pm\infty), which happens when dx/dt = 0 and dy/dt \neq 0.

When both dx/dt = 0 and dy/dt = 0 at the same parameter value, the point is called a singular point of the parametric curve. The behaviour there requires more careful analysis — typically using L'Hopital's rule on the ratio dy/dx = (dy/dt)/(dx/dt) as t approaches the value in question.

Concavity

The sign of d^2y/dx^2 tells you the concavity of the curve:

• d^2y/dx^2 > 0: the curve is concave upward (bending like a cup)
• d^2y/dx^2 < 0: the curve is concave downward (bending like a cap)
• d^2y/dx^2 = 0: a possible inflection point, where the concavity might change

This is identical to the concavity test for explicit functions y = f(x), just computed through the parametric machinery.

The circle, revisited

Apply the formula to the circle x = r\cos t, y = r\sin t:

At t = \pi/2 (the top of the circle), \cot(\pi/2) = 0, so dy/dx = 0 — a horizontal tangent. At t = 0 (the rightmost point), \cot 0 is undefined, so dy/dx blows up — a vertical tangent. Both match the geometry perfectly.

For the second derivative:

At the top of the circle (t = \pi/2), this gives -1/r. The sign is negative (concave down, as expected — the top of a circle curves downward). The magnitude is 1/r, which is the curvature of the circle. A smaller circle bends more sharply and has a larger curvature — exactly what 1/r says.

Common confusions

• "dy/dx = (dy/dt)/(dx/dt) means you can just cancel the dt's." It looks like cancellation, and the answer is correct, but this is not algebraic cancellation. It is the chain rule in Leibniz notation. The distinction matters when you try to extend the same reasoning to second derivatives — naive cancellation gives the wrong answer there.

• "The second derivative is (d^2y/dt^2)/(d^2x/dt^2)." This is the most common mistake in parametric differentiation. The second derivative d^2y/dx^2 is not the ratio of second derivatives with respect to t. You must differentiate dy/dx with respect to t first, then divide by dx/dt. The expanded formula (\ddot{y}\dot{x} - \dot{y}\ddot{x})/\dot{x}^3 has three terms in the numerator, not just \ddot{y}.

• "If dx/dt = 0, the derivative doesn't exist." More precisely, dy/dx is infinite there — the tangent line is vertical. The curve itself is perfectly well-defined; it just has a vertical tangent. The only truly problematic case is when both dx/dt and dy/dt vanish simultaneously.

• "The parameter t must be time." Not at all. In the ellipse example, t was an angle. In many geometry problems, t is a slope or a ratio. The parametric derivative formula works for any smooth parameter.

Going deeper

Why parametric form exists at all

Not every curve can be written as y = f(x). A circle fails this test — for most x-values there are two y-values. A figure-eight curve fails even more dramatically. The equation y = f(x) forces the curve to pass the vertical line test, and many interesting curves do not.

Parametric form sidesteps this entirely. Instead of asking "y as a function of x," you describe both coordinates as functions of a third variable. Every curve in the plane — no matter how it twists and self-intersects — has a parametric description. That is the fundamental advantage.

Arc length from parametric equations

If a curve is given by x = f(t), y = g(t) for t \in [a, b], its arc length is

This formula comes from the Pythagorean theorem applied to infinitesimal segments of the curve: a tiny step dt in the parameter produces a horizontal displacement dx = (dx/dt)\,dt and a vertical displacement dy = (dy/dt)\,dt, and the distance along the curve is \sqrt{dx^2 + dy^2}.

For the circle x = r\cos t, y = r\sin t over [0, 2\pi]:

The circumference formula, recovered from first principles.

Curvature

The curvature \kappa of a plane curve measures how sharply it bends at each point. In parametric form:

Look at the numerator: it is |\dot{x}\ddot{y} - \dot{y}\ddot{x}|, the same cross-product-like expression that appears in the second derivative formula. This is not a coincidence — curvature and the second derivative are measuring the same underlying quantity (the rate at which the tangent direction rotates), just normalized differently.

For a circle of radius r, compute each piece. With x = r\cos t, y = r\sin t:

So \kappa = 1/r everywhere — a circle has constant curvature, and the curvature is the reciprocal of the radius. A small circle (r small, \kappa large) bends sharply; a large circle (r large, \kappa small) bends gently. A straight line is a circle of infinite radius: \kappa = 0.

For an ellipse, the curvature varies: it is largest at the ends of the shorter axis (where the curve bends most sharply) and smallest at the ends of the longer axis.

A third derivative example: the parabola's second derivative

To consolidate the second derivative formula, compute d^2y/dx^2 for the parabola x = at^2, y = 2at:

Differentiate dy/dx = 1/t = t^{-1} with respect to t:

Divide by dx/dt = 2at:

This is always negative (for t > 0), confirming that the upper branch of the parabola is concave downward when viewed as a graph of y against x — which is correct, since y^2 = 4ax means the curve bends back toward the x-axis.

You can verify this with the expanded formula. The second derivatives with respect to t are \ddot{x} = 2a and \ddot{y} = 0. Substituting:

Same answer.

Parametric form in the Indian curriculum

The NCERT and JEE syllabi frequently present curves in parametric form — particularly conics. The standard parametric forms are:

For the parabola y^2 = 4ax with parameter t:

This is a clean result: the slope of the parabola at the point (at^2, 2at) is 1/t. At t = 1 (the point (a, 2a)), the slope is 1. At t = 2 (the point (4a, 4a)), the slope is 1/2 — flatter, as you would expect since the parabola opens out. At t \to 0 (near the vertex), the slope 1/t \to \infty — a vertical tangent at the vertex.

Where this leads next

You now know how to differentiate curves given in parametric form — both the first and second derivatives. Here are the natural next steps:

• Higher Order Derivatives — the general theory of second, third, and nth derivatives, including the Leibniz theorem for differentiating products.
• Differentiation of Functions w.r.t. Functions — what happens when you differentiate one function with respect to another, not just with respect to x.
• Chain Rule — the foundational rule that makes parametric differentiation work.
• Implicit Differentiation — the alternative approach when the curve is given as F(x,y) = 0 instead of parametrically.
• Tangent and Normal — using derivatives to find equations of tangent and normal lines to curves.