Parametric Functions

Parametric functions describe a curve by expressing each coordinate as a separate function of an independent third variable (the parameter), usually written

x = f (t)

and

y = g (t)

. This lets you trace curves that fail the vertical line test (circles, loops, spirals) and naturally encode motion — where you are and when. Used everywhere from physics (projectile paths) to computer graphics (Bézier curves). Formalized through the work of Euler, Newton, and Leibniz in the 17th–18th centuries.

A parametric function (or parametric curve) is a pair (or triple) of functions sharing a common input variable $t$ , the parameter, so that as $t$ runs over an interval the point $(f (t), g (t))$ sweeps out a curve in the plane. Instead of asking "for a given $x$ , what is $y$ ?", parametric form asks "at parameter value $t$ , where am I?". This matters because many important curves — circles, ellipses, cycloids, helices, self-intersecting loops — simply cannot be written as a single function $y = h (x)$ , yet they are trivially parametric. The surprising fact: the same geometric curve has infinitely many parametrizations, and the choice of parametrization carries extra information (speed, direction) that the bare curve does not — which is exactly why physics loves it.

The Core Insight

A function $y = h (x)$ is a constraint — one number determines another. A parametric curve is a journey — a hidden clock $t$ ticks, and at each tick you are told both your horizontal position and your vertical position independently. Decoupling $x$ and $y$ through a shared parameter is what frees you from the vertical-line-test prison and lets a single pen-stroke loop, cross itself, and retrace.

Formal Statement

Parametric Curve in the Plane

A parametric curve in $R^{2}$ is given by two functions of a single real parameter $t$ :

x = f (t), y = g (t), t \in [a, b]

⟳This is a definition, not a derived result — it simply names

t

the parameter and

f, g

the coordinate functions; the interval

[a, b]

is the parameter domain.

Here:

$t$ is the parameter (often time, an angle, or arc length);
$f$ and $g$ are the coordinate functions;
$[a, b]$ is the parameter interval;
$(f (a), g (a))$ is the initial point and $(f (b), g (b))$ the terminal point;
the orientation is the direction of travel as $t$ increases.

The set of all points ${(f (t), g (t)) : t \in [a, b]}$ is the trace (the geometric curve); the functions together are the parametrization.

Eliminating the Parameter

To recover a Cartesian relation, solve one equation for $t$ and substitute into the other:

t = f^{- 1} (x) \Rightarrow y = g (f^{- 1} (x))

⟳Invert

x = f (t)

to express

t

in terms of

x

(only possible where

f

is one-to-one), then plug into

y = g (t)

; algebraic identities can sometimes bypass explicit inversion.

Slope of a Parametric Curve

If $f$ and $g$ are differentiable and $\frac{d x}{d t} \neq = 0$ , the slope of the tangent line is

\frac{d y}{d x} = \frac{d y / d t}{d x / d t} = \frac{g ^{'} ( t )}{f ^{'} ( t )}

⟳Apply the chain rule

\frac{d y}{d t} = \frac{d y}{d x} \cdot \frac{d x}{d t}

and divide through by

\frac{d x}{d t}

The second derivative (for concavity) is

\frac{d ^{2} y}{d x ^{2}} = \frac{\frac{d}{d t} ( \frac{d y}{d x} )}{d x / d t}

⟳Treat

\frac{d y}{d x}

as a new function of

t

, differentiate it with respect to

t

, then divide by

d x / d t

again — NOT

g^{''} / f^{''}

Arc Length of a Parametric Curve

The length of the curve traced once as $t$ goes from $a$ to $b$ is

L = \int_{a}^{b} (\frac{d x}{d t})^{2} + (\frac{d y}{d t})^{2} d t

⟳A tiny step has horizontal change

d x = f^{'} (t) d t

and vertical change

d y = g^{'} (t) d t

; the Pythagorean length of that step is

d x^{2} + d y^{2}

, and summing gives the integral.

Why It Holds / Derivation

The Slope Formula via the Chain Rule

We justify equation [3].

Step 1. Suppose near a parameter value

t_{0}

the function

x = f (t)

is invertible (guaranteed where

f^{'} (t_{0}) \neq = 0

). Then

t = f^{- 1} (x)

and

y = g (f^{- 1} (x))

is genuinely a function of

x

. Step 2. By the chain rule applied to

y

as a function of

t

through

x

\frac{d y}{d t} = \frac{d y}{d x} \cdot \frac{d x}{d t}

⟳

y

depends on

x

, and

x

depends on

t

; the chain rule links the rates.

Step 3. Provided

\frac{d x}{d t} \neq = 0

, divide both sides by

\frac{d x}{d t}

\frac{d y}{d x} = \frac{d y / d t}{d x / d t}

⟳Pure algebra — isolate

\frac{d y}{d x}

from [6].

Step 4. Where

\frac{d x}{d t} = 0

but

\frac{d y}{d t} \neq = 0

, the tangent is vertical; where both vanish, the point may be a cusp or corner and requires limits.

■

Leonhard Euler | 1707–1783

Euler was the great systematizer of the parameter idea. While Newton and Leibniz used "fluents" flowing in time, Euler in his Introductio in analysin infinitorum (1748) treated curves analytically as coordinates expressed through an auxiliary variable, and his notation $f (x)$ for functions made the parametric viewpoint — two functions, one input — clean and general. His study of the cycloid, catenary, and motion problems repeatedly used time-parametrized coordinates, cementing parametric description as a standard tool of analysis.

Building Intuition

Picture a bug walking on graph paper while a stopwatch runs. At every instant the stopwatch reads $t$ . You separately record the bug's east–west position $x (t)$ and its north–south position $y (t)$ in two columns. Plot those recorded points and connect them in time order — you have re-created the bug's path. The graph $y$ vs $x$ throws away the timestamps; it cannot tell whether the bug walked slowly or sprinted, nor whether it backtracked across its own trail. Parametric form keeps the timestamps. That is why a circle works: at $t = 0$ the bug is at the right edge, by $t = π /2$ it is at the top — the same $x$ -value occurs twice (top and bottom of the circle) but at different times, which is forbidden for an ordinary function but perfectly fine when the clock distinguishes them.

A parametric circle: the clock value t determines a point on the curve

Worked Examples

Eliminating the Parameter and Finding a Tangent

Given:

x = t^{2} - 1

y = 2 t + 3

, for

t \in [- 2, 2]

. Step 1 — Eliminate $t$ . From the second equation,

t = \frac{y - 3}{2}

. Step 2 — Substitute.

x = (\frac{y - 3}{2})^{2} - 1 = \frac{( y - 3 ) ^{2}}{4} - 1

⟳Plug

t = \frac{y - 3}{2}

into

x = t^{2} - 1

This is a sideways parabola opening in the $+ x$ direction with vertex at $(- 1, 3)$ .

Step 3 — Slope at $t = 1$ . Using [3]:

\frac{d y}{d x} = \frac{d y / d t}{d x / d t} = \frac{2}{2 t} = \frac{1}{t}

⟳

d x / d t = 2 t

d y / d t = 2

; take the ratio.

At $t = 1$ : point is $(0, 5)$ , slope $= 1/1 = 1$ .

y = 5 + 1 \cdot (x - 0) = x + 5 (tangent line, dimensionless)

Projectile Motion — A Physical Parametrization

Given: A ball launched at speed

v_{0} = 20 m/s

at angle

θ = 3 0^{\circ}

from the origin, gravity

g = 9.8 m/s^{2}

. Time

t

is the parameter. Step 1 — Write the parametric equations.

x (t) = v_{0} cos θ t, y (t) = v_{0} sin θ t - \frac{1}{2} g t^{2}

⟳Horizontal motion is constant-velocity; vertical motion is constant-acceleration from the kinematics equations.

With numbers: $v_{0} cos 3 0^{\circ} = 20 (0.866) = 17.32 m/s$ , $v_{0} sin 3 0^{\circ} = 20 (0.5) = 10 m/s$ .

x (t) = 17.32 t, y (t) = 10 t - 4.9 t^{2}

Step 2 — Time of flight. Set

y = 0

t (10 - 4.9 t) = 0 \Rightarrow t = \frac{10}{4.9} = 2.04 s

. Step 3 — Range (horizontal distance).

R = x (2.04) = 17.32 \times 2.04 = 35.3 m

⟳Evaluate

x (t)

at the time the ball lands.

R \approx 35.3 m, flight time \approx 2.04 s

Eliminating $t$ recovers the familiar parabolic trajectory $y = (tan θ) x - \frac{g}{2 v _{0}^{2} cos ^{2} θ} x^{2}$ .

Common Misconceptions

The second derivative is NOT $\frac{d ^{2} y / d t ^{2}}{d ^{2} x / d t ^{2}}$ . A very common error. You must differentiate the slope

\frac{d y}{d x}

(already a quotient) with respect to

t

and then divide by

\frac{d x}{d t}

, as in equation [4]. Blindly dividing second derivatives gives wrong concavity. Also: the curve (trace) and the parametrization are different objects. Two parametrizations can produce the identical picture yet differ in speed, direction, and how many times the curve is traced.

Same Trace, Different Parametrizations

Consider these two parametrizations:

$x = cos t, y = sin t, t \in [0, 2 π]$
$x = cos 2 t, y = sin 2 t, t \in [0, 2 π]$

Both have trace = the unit circle, visually identical. But the second traverses the circle twice and moves at double the speed: its arc length integral [5] gives $L = \int_{0}^{2 π} 4 sin^{2} 2 t + 4 cos^{2} 2 t d t = \int_{0}^{2 π} 2 d t = 4 π$ , whereas the first gives $2 π$ . The geometric circle is one thing; the journey along it is another. This is why arc length is computed only when the curve is traced exactly once.

Connections

Parametric functions sit at a crossroads. They generalize ordinary Functions (any $y = h (x)$ is the parametrization $x = t, y = h (t)$ ) and specialize to Polar Coordinates (set $x = r (θ) cos θ, y = r (θ) sin θ$ ). The slope and arc-length formulas come straight from the Chain Rule and Definite Integrals. The position-velocity viewpoint is the doorway to Vector-Valued Functions, where $r (t) = ⟨ f (t), g (t)⟩$ packages both coordinates into one vector, and onward to Kinematics in physics. The derivative pair $(f^{'} (t), g^{'} (t))$ is the Tangent Vector.

Real-World Applications

Projectile and orbital motion — physics describes trajectories as position vs time, inherently parametric (see equation [10]).
Computer graphics & font design — Bézier curves and B-splines, the backbone of Adobe Illustrator, TrueType fonts, and CAD, are cubic parametric curves with control points.
Animation — an object's path is keyframed as $(x (t), y (t))$ so motion (timing, easing) is controllable independently of shape.
GPS and robotics path planning — a robot arm or vehicle follows a parametrized trajectory ensuring smooth velocity profiles.
Roller coaster and road design — track geometry uses parametric clothoid (Euler spiral) curves so curvature changes linearly, limiting jerk.
Signal & Lissajous figures — oscilloscopes display $x = sin (a t), y = sin (b t + δ)$ to analyze phase and frequency ratios.
Epidemiology & dynamical systems — phase-plane plots trace $(infected (t), susceptible (t))$ to visualize disease spread.

Historical Development

~250 BCE — Archimedes studies the spiral $r = a θ$ , an implicitly parameter-driven curve (angle as parameter).
1638 — Galileo investigates the cycloid, the path of a point on a rolling wheel — a quintessential parametric curve.
1659 — Christiaan Huygens proves the cycloid is the tautochrone (isochronous pendulum), using time-parametrized motion.
1665–1687 — Newton develops fluxions, treating coordinates as quantities "flowing" with time $t$ — the physical root of parametrization.
1696 — Johann Bernoulli poses the brachistochrone problem; the cycloid solution is naturally expressed parametrically.
1748 — Euler's Introductio in analysin infinitorum systematizes analytic curves and the functional viewpoint, generalizing the parameter idea.
1731 — Alexis Clairaut studies skew (3D) curves $x (t), y (t), z (t)$ , extending parametrization to space.
1959–1962 — Pierre Bézier (Renault) and Paul de Casteljau (Citroën) independently develop parametric polynomial curves for car-body design — birth of modern CAD.

Subtleties & Limits

Several deep issues lurk beneath the introductory surface. (1) Smoothness vs. appearance: a parametrization can be perfectly differentiable yet produce a cusp (sharp corner) where both $f^{'} (t) = 0$ and $g^{'} (t) = 0$ — the curve looks non-smooth though the equations are smooth. (2) Pathological curves: continuous parametrizations exist (the Peano and Hilbert space-filling curves, 1890–1891) that fill an entire 2D square — a "curve" with positive area, shattering the naive notion that curves are one-dimensional. (3) Reparametrization invariance: which quantities (length, curvature) are intrinsic to the trace and which (speed, time) belong to the parametrization is a subtle and important distinction studied rigorously in differential geometry. (4) Constant-speed parametrization by arc length always exists for nice curves but is rarely expressible in elementary functions — even the ellipse's arc length requires elliptic integrals, an unsolved-in-closed-form classic.