Polar coordinates come with their own pair of unit vectors: r̂ = ⟨cos θ, sin θ⟩, which points straight out from the origin, and θ̂ = ⟨−sin θ, cos θ⟩, which points the way the angle grows. The one thing to hold onto before anything else: these vectors change direction depending on where you are. If you ever need to confirm a component set has length 1, the calculator does the check.
That single fact is the whole difficulty of the topic, and most pages bury it. The Cartesian vectors i and j point the same way at every point in the plane. The polar pair does not.
The idea that makes polar coordinates hard
Stand at a point on the positive x-axis. There, r̂ points to the right and θ̂ points up. Now walk to a point on the positive y-axis. Now r̂ points up and θ̂ points to the left. The vectors rotated, because they are tied to your angular position, not pinned to the room.
This is why polar coordinates feel slippery at first. In Cartesian coordinates a vector’s components mean the same thing everywhere. In polar coordinates the basis itself turns as you move, so a “radial” component at one location does not point the same way as a “radial” component at another. Once that clicks, the rest is arithmetic.
The definitions, and why they are unit vectors
The radial unit vector r̂ = ⟨cos θ, sin θ⟩ is exactly the direction from the origin to your point. The angular unit vector θ̂ = ⟨−sin θ, cos θ⟩ is that direction rotated a quarter turn counterclockwise, so it runs tangent to the circle through your point.
Both are already length 1, which the Pythagorean identity guarantees. The verify run below confirms it and adds the perpendicularity check, since a good basis needs both.
Put in real angles and the rotation shows itself. At θ = 0 the pair is r̂ = ⟨1, 0⟩ and θ̂ = ⟨0, 1⟩, which is just i and j. At θ = 90° it becomes r̂ = ⟨0, 1⟩ and θ̂ = ⟨−1, 0⟩. Same formulas, different directions, because the angle changed.
Writing a vector in polar components
Say you have a vector in ordinary Cartesian form and want its radial and angular parts at a particular location. You project it onto the two unit vectors: the radial component is v · r̂ and the angular component is v · θ̂.
Take v = ⟨3, 4⟩ at a point sitting on the positive y-axis, where θ = 90°. There the basis is r̂ = ⟨0, 1⟩ and θ̂ = ⟨−1, 0⟩. The projections are:
So v = 4 r̂ − 3 θ̂ at that point. The size did not change: √(4² + (−3)²) = 5, the same as √(3² + 4²) = 5 in Cartesian form. Only the description changed, because you measured the same arrow against a rotated pair of axes.
The derivatives, and where the extra terms come from
Because the basis itself moves, differentiating in polar coordinates picks up terms that do not exist in Cartesian. Differentiate the definitions with respect to θ:
Read those two lines slowly, because they explain a lot. When you differentiate a position written in polar form to get velocity, and again to get acceleration, these dr̂/dθ = θ̂ and dθ̂/dθ = −r̂ relations feed in and produce the extra pieces. That is where the terms in circular-motion formulas come from, the ones that look like they appeared from nowhere. They come from the basis turning. For the coordinate system itself, the Wikipedia article on the polar coordinate system is a solid reference.
Where it earns its keep: circular motion
Put a point at radius r and angle θ. Its position is position = r r̂, radius times the outward direction. That is the whole location written in polar form, and it is where the payoff begins.
To get velocity, differentiate that with respect to time. The radius can change and the angle can change, so both r and r̂ are moving. Using the relations from the previous section (dr̂/dθ = θ̂ and dθ̂/dθ = −r̂) together with the chain rule, the velocity comes out as:
The first term is motion outward or inward, the radial part, and it shows up only when the radius is changing. The second term is motion around the center, the angular part. The r θ′ factor in front of θ̂ is why a point far from the center sweeps faster: same angular rate, bigger radius, more distance covered. Two clean terms, each with a plain meaning.
Those two terms only appear this neatly because the basis turns with the point. Write the same motion in Cartesian coordinates and you get a messier pair of expressions with sines and cosines tangled through both components. Differentiate the velocity one more time and the turning basis produces the centripetal and Coriolis terms, but that is a longer story than this section needs.
The pure circular case is worth pausing on, because it shows the split at its cleanest. Hold the radius fixed and only the angle changes, so r′ = 0 and the radial term drops out entirely. Velocity is then r θ′ θ̂, pointing straight along the circle with nothing pointing outward. That matches what you already know about circular motion: the velocity is tangent to the path, and its size grows with both the radius and how fast the angle turns.
A point in polar form
Take the point at radius r = 2 and angle θ = 60°. Its Cartesian location is (2 cos 60°, 2 sin 60°) = (1, 1.7321), one unit right and a bit over one and a half up. That is the same point, just written the ordinary way.
The polar basis sitting at that point is r̂ = ⟨cos 60°, sin 60°⟩ = ⟨0.5, 0.8660⟩ and θ̂ = ⟨−sin 60°, cos 60°⟩ = ⟨−0.8660, 0.5⟩. The length check on r̂ holds: 0.5² + 0.8660² = 1. Notice that r̂ points from the origin straight out toward the point, and θ̂ is a quarter turn ahead of it, running along the circle in the direction the angle grows.
The radius does not enter either unit vector. Whether the point sat at radius 2 or somewhere else along the same ray, the basis would read the same, because both vectors depend on the angle alone. Move to a different angle and the pair swings to a new orientation, which is the behavior the whole page keeps returning to.
Another Cartesian-to-polar conversion
Convert v = ⟨0, 2⟩, an arrow pointing straight up, into polar components at the point where θ = 0, out on the positive x-axis. There the basis is r̂ = ⟨1, 0⟩ and θ̂ = ⟨0, 1⟩, the plain i and j again. Project v onto each:
So v = 2 θ̂ at that point: standing on the x-axis, an upward arrow is purely angular, with no radial part at all. The very same vector would split up differently at a different point, because the basis has rotated to a new pair of directions there. The arrow never moved; the axes you measured it against did.
The reverse trip works with the same two vectors. Given radial and angular components, rebuild the Cartesian vector by scaling r̂ and θ̂ and adding them: the radial component times r̂ plus the angular component times θ̂. Both directions of the conversion lean on the same fact, that r̂ and θ̂ form a perpendicular pair of length-1 vectors at the point you care about. Fix the point, write the basis, and the projections fall out.
Try one yourself
At the point where θ = 90°, write v = ⟨1, 0⟩, an arrow pointing right, in polar components. Work out the basis at that angle first, then take the two dot products, and check your answer below.
There the basis is r̂ = ⟨0, 1⟩ and θ̂ = ⟨−1, 0⟩. The projections are v · r̂ = (1)(0) + (0)(1) = 0 and v · θ̂ = (1)(−1) + (0)(0) = −1, so v = −θ̂. Pointing right while you stand at the top of the circle is the backward angular direction, which is where the minus sign comes from.
Notation across textbooks
Two spellings are common. Many books write the polar unit vectors as r̂ and θ̂ with hats. Others, especially in physics, write them as e_r and e_θ. A few use ρ̂ for the radial direction to reserve r for a three-dimensional radius.
They all name the same two directions, one pointing out and one pointing around. Match your course and move on. When you extend this to three dimensions, cylindrical coordinates keep exactly this pair and add a fixed vertical direction, while spherical coordinates replace the whole set, all of which the cylindrical and spherical guide covers. If you would rather see where the perpendicular directions stay fixed instead of turning, the unit normal hub is the place to start.
They move with you. Fix your angle first, read off r̂ and θ̂ there, and the rest of the problem lines up.
Frequently Asked Questions
What are the polar unit vectors?
They are r̂ = ⟨cos θ, sin θ⟩, which points outward from the origin, and θ̂ = ⟨−sin θ, cos θ⟩, which points in the direction of increasing angle. Both have length 1 and they are perpendicular to each other.
Why do polar unit vectors change direction?
Because they are defined by the angle θ of your position. At a point on the positive x-axis r̂ points right, and at a point on the positive y-axis it points up. Unlike i and j, they are different vectors at different places.
What are the derivatives of the polar unit vectors?
Differentiating with respect to θ gives dr̂/dθ = θ̂ and dθ̂/dθ = −r̂. These extra terms are why polar velocity and acceleration carry pieces that Cartesian coordinates do not.
Are r̂ and θ̂ perpendicular?
Yes. Their dot product is (cos θ)(−sin θ) + (sin θ)(cos θ) = 0 at every angle, so they meet at a right angle everywhere, just like the Cartesian axes do.
Is θ̂ the same as e_θ?
Yes, they are two names for the same vector. Some books write the polar unit vectors as r̂ and θ̂, others as e_r and e_θ. The notation differs; the vector does not.