How to Find a Unit Tangent Vector

The unit tangent vector is T̂(t) = r′(t)/‖r′(t)‖: differentiate the position vector to get the velocity, then divide by its length. What you get is a vector of length 1 that points along the direction of travel, with the speed stripped out. Once you have r′(t) in hand, you can normalize it in the calculator and check the length in one step.

That is the whole formula. The part that actually stalls people is not the dividing, it is the object being divided, so this guide spends most of its time on setting up r(t) and reading r′(t) correctly.

What the vector points at

Imagine driving along a winding road. At every instant your car has a heading, the direction the hood is pointing. That heading is the unit tangent. It ignores how fast you are going and reports only which way you are moving right now.

The velocity r′(t) also points that way, but its length is your speed. Divide the velocity by its own length and the speed cancels, leaving a clean direction of length 1. Two cars taking the same corner, one crawling and one flooring it, share the exact same unit tangent at each point on the curve. That is the useful thing about it: it isolates shape from pace.

Reading the parameterization first

If r(t) is new to you, start here. A position vector is just three ordinary functions bundled together, one for each coordinate, all sharing the same input t. Feed in a value of t and you get a point in space; let t run and the point traces the curve.

Write r(t) = ⟨x(t), y(t), z(t)⟩. For a circle you might have r(t) = ⟨cos t, sin t⟩, where t is the angle swept. To differentiate, you differentiate each slot on its own: r′(t) = ⟨x′(t), y′(t), z′(t)⟩. Nothing couples the components together, so this is easier than it looks the first time.

The method, step by step

1. Write the curve as a position vector r(t) = ⟨x(t), y(t), z(t)⟩.
2. Differentiate each component to get the velocity r′(t).
3. Compute the speed ‖r′(t)‖, the square root of the sum of the squared components.
4. Divide the velocity by the speed: T̂ = r′(t)/‖r′(t)‖.
5. Confirm the components of T̂ square to 1.

A clean example: the helix

Take the helix r(t) = ⟨cos t, sin t, t⟩, a circle in the first two coordinates that climbs steadily in the third. Differentiating each component gives the velocity r′(t) = ⟨−sin t, cos t, 1⟩.

Now the speed. It is ‖r′(t)‖ = √(sin²t + cos²t + 1) = √2, and the pleasant surprise is that the t drops out entirely. The helix has constant speed, so the same magnitude works at every point. Evaluate the velocity at t = 0 to get ⟨0, 1, 1⟩, and normalize that.

In general the unit tangent is T̂(t) = (1/√2)⟨−sin t, cos t, 1⟩. The third component stays fixed at 1/√2 ≈ 0.7071 no matter where you are on the helix, which says the climb angle never changes. The first two components rotate as t advances, tracking the circular part of the motion.

A messier example, because problem sets are messy

The helix is friendly because its speed is constant. Most curves are not, and the honest way to show that is a curve where the speed does not simplify. Take r(t) = ⟨t, t², t³⟩. Differentiating gives r′(t) = ⟨1, 2t, 3t²⟩.

Here the speed ‖r′(t)‖ = √(1 + 4t² + 9t⁴) is a genuine function of t that does not collapse into anything tidy. Most tutorials quietly avoid this. You handle it by evaluating at the specific point you care about. At t = 1 the velocity is ⟨1, 2, 3⟩, and its speed is √14 ≈ 3.7417.

The lesson is that you almost never need the tangent as a formula in t. You need it at a point, so plug the value in before you normalize and the arithmetic stays manageable. If you did want the general expression, it is T̂(t) = ⟨1, 2t, 3t²⟩ / √(1 + 4t² + 9t⁴), correct but rarely worth writing out.

Why the speed cancels, and what it means

The division by ‖r′(t)‖ is not a formality. Speed and direction are two separate pieces of the velocity, and normalizing surgically removes the speed. Whatever units the velocity carried, meters per second or otherwise, the unit tangent is dimensionless and simply points.

This is exactly why the unit tangent, not the raw velocity, is the starting block for curvature. Curvature measures how fast the direction turns, so you first have to erase how fast the particle moves. Differentiating T̂ gives the change in pure direction, which is where the TNB frame and the principal normal come from.

The smoothness condition

The formula quietly assumes ‖r′(t)‖ is not zero, because you cannot divide by zero. A point where r′(t) = 0 is called a singular point, and the curve can have a corner or a cusp there with no well-defined direction of travel.

A curve is called smooth on an interval when r′(t) is continuous and never zero there. On a smooth curve the unit tangent exists and turns continuously, which is what you want before doing anything further with it. When a problem says “smooth curve,” that condition, r′(t) ≠ 0, is what it is buying you. The broader setup is described well in the Wikipedia article on the tangent vector.

A plane-curve example

Not every curve lives in three dimensions, and the method does not change when one is missing. A plane curve has a two-slot position vector, you differentiate each slot the same way, and you divide by the speed the same way. There is simply one fewer number to square when you take the magnitude.

Take the parabola r(t) = ⟨t², 2t⟩. Differentiating each component gives the velocity r′(t) = ⟨2t, 2⟩. At t = 1 that is ⟨2, 2⟩, a velocity pointing up and to the right along the curve.

The equal components say the heading sits at 45 degrees at this point, which matches the picture: at t = 1 the parabola is rising as fast as it runs. Everything you did on the helix carries over unchanged, minus a coordinate.

The speed here does depend on t, since ‖r′(t)‖ = √(4t² + 4) = 2√(t² + 1) changes as you move along the parabola. That is the usual situation, so pick the value of t the problem asks about and substitute it before you take the magnitude. The calculator only needs the numeric velocity at that point, not the symbolic speed, which is why plugging in first keeps the work short.

Tangent vector versus tangent line

These two objects are related but not the same, and the words look enough alike to cause trouble. The unit tangent T̂ is a single vector: it gives the direction of the tangent line to the curve at the point, and nothing more. The tangent line itself is the whole line you get by starting at the point on the curve and adding any scalar multiple of that direction.

So the direction is one ingredient and the anchor point is the other. The unit tangent strips length out and reports a pure heading, which is why it, and not the raw velocity, is what you feed into a direction-based calculation. The velocity would work for the line too, since it points the same way, but its length carries the speed and that extra baggage gets in the way the moment you want to compare directions cleanly.

A quick way to hold the distinction: the tangent line is a set of points, while the unit tangent is a single arrow. If you evaluate the position at the point of tangency and pair it with T̂, you can write the line as a point plus t times T̂, and the length being 1 means the parameter you slide along the line measures actual distance travelled on it. That is a small bonus of normalizing that raw velocity does not give you.

Two common slips

Most wrong answers here come from one of a few predictable places, and all of them are easy to catch once you know to look.

• Normalizing r(t) itself instead of its derivative r′(t). The position tells you where the particle is, not which way it is heading, so normalizing it gives a unit vector that points from the origin to the point, not along the curve.
• Forgetting to plug the specific value of t in before computing the magnitude. Do that and you are stuck simplifying a speed like √(1 + 4t² + 9t⁴) that never collapses, when the number you needed was waiting one substitution away.
• Treating a point where r′(t) = 0 as if the tangent exists there. It does not: the division has no answer, the point is singular, and the curve has no defined direction of travel to report.

Try one yourself

Find the unit tangent to r(t) = ⟨3 cos t, 3 sin t⟩ at t = 0. This is a circle of radius 3, so watch what the radius does on the way through.

Differentiate each component to get r′(t) = ⟨−3 sin t, 3 cos t⟩. At t = 0 the sine is 0 and the cosine is 1, so the velocity is ⟨0, 3⟩, pointing straight up. Its magnitude is √(0² + 3²) = 3, and dividing gives T̂ = ⟨0, 1⟩, which checks out since 0² + 1² = 1. Notice the radius 3 divided straight out: that factor was the speed, and stripping it is exactly what normalizing does.

Notation across textbooks

Some books write the unit tangent as T̂(t) with a hat to stress that it has length 1. Others drop the hat and write T(t), treating “unit” as part of the name and expecting you to remember. A few reserve T for the velocity direction and only add the hat after normalizing.

The safe move is to check your own course’s convention once and match it. Whatever the symbol, the thing it names is the same: velocity, divided by speed, length exactly 1. The next step from here is stacking it with the principal normal and the binormal in the TNB frame, and if you are unsure which normal a problem wants, the unit normal hub sorts that out.

Speed out, direction in. Build r′(t), then let the calculator handle the division and the length check.