Angle Between Two Vectors Calculator

What the dot product means

Multiply matching components and add them up. For (3,4) and (4,3): 3×4 + 4×3 = 24.

Geometrically it measures how much the two vectors agree in direction, scaled by both lengths. Pointing the same way gives a large positive number; pointing opposite ways gives a large negative one; at right angles it gives exactly zero.

That zero is the useful part. Testing perpendicularity needs no square roots, no inverse cosine and no trigonometry at all — just a handful of multiplications and an addition. It is why the dot product appears in almost every geometry routine ever written.

Why divide by the magnitudes

The raw dot product depends on how long the vectors are, and length is not what the question is about.

(1,0) and (2,0) point the same way and have a dot product of 2. (10,0) and (20,0) also point the same way and have a dot product of 200. Same angle, wildly different numbers.

Dividing by both magnitudes strips the lengths out, leaving a number between −1 and 1 that depends only on direction. That number is the cosine of the angle, and the inverse cosine recovers the angle itself.

This is why scaling either vector never changes the answer — something the page’s test suite checks across four thousand generated pairs rather than assuming.

Perpendicular and parallel

Perpendicular: dot product exactly zero, angle exactly 90°. No other test is needed.

Parallel, same direction: cosine 1, angle 0°.

Parallel, opposite direction: cosine −1, angle 180°. Sometimes called antiparallel.

The parallel cases need care in floating point. (2,4) and (1,2) are exactly parallel, but computing their cosine gives 0.9999999999999998, and the inverse cosine of that is not zero — it is about a millionth of a degree, because arccos magnifies error near the ends of its range.

This page tests for parallelism with the cross product instead, which is exactly zero for those vectors, and reports 0° when it is. A millionth of a degree is not a rounding detail when the correct answer is nothing.

The cross product, and what it adds

Where the dot product measures agreement, the cross product measures disagreement. Its magnitude is |u||v|·sin θ, largest when the vectors are perpendicular and zero when they are parallel.

That magnitude is also the area of the parallelogram the two vectors span, which is the geometric reason it vanishes for parallel vectors: a flattened parallelogram has no area.

In three dimensions the cross product is itself a vector, perpendicular to both inputs — which is how a surface normal is computed from two edges. In two dimensions there is no room for such a vector, so what is reported is the scalar determinant.

Between them the two products give the whole picture: the dot gives cos θ, the cross gives sin θ, and their ratio gives the tangent.

Signed angles need more than two vectors

The answer here is always between 0° and 180°. It is the unsigned angle, the one you would measure with a protractor laid between them.

A signed angle — “45° anticlockwise” rather than just “45°” — needs an orientation, and two vectors alone do not supply one. In the plane you can get it from atan2 of the cross and dot products, because the plane has a conventional positive direction.

In three dimensions there is no such convention. “Anticlockwise” depends on which side you are viewing from, so a signed angle needs an axis to be nominated first. That is why this page reports the unsigned angle: it is the one that is well defined without extra information.

Where it gets used

Work in physics. Work is force times displacement times the cosine of the angle between them — which is the dot product. A force at right angles to the motion does no work, because its dot product is zero.

Lighting in graphics. Surface brightness is the dot product of the surface normal and the direction to the light. Facing the light gives 1, edge-on gives 0.

Similarity in machine learning. Cosine similarity between two embedding vectors is exactly the cosine computed here. Comparing documents or search results is this operation at scale.

Projection. The component of one vector along another is (u · v)/|v|, which is the shadow one casts on the other. Every decomposition into components uses it.

Sources and methodology

The definitions are standard linear algebra; these are the references.

• American Mathematical Society — where the inner product and its geometric interpretation are catalogued
• NIST SP 811 — plane angle conventions and the radian
• Central Angle Calculator — the same angle, measured at the centre of a circle
• How we calculate — methodology

Method. The dot product and both magnitudes are computed and displayed as separate steps rather than folded into the answer, because the dot product alone settles the question most readers arrived with. Exactly parallel vectors are detected by the cross product, which is zero for them, rather than by the inverse cosine — near ±1 the arccos magnifies rounding, and (2,4) with (1,2) returned 0.0000012° before that check was added. The suite compares every result against the law of cosines applied to the triangle the two vectors and their difference form. That engine is verified on every change against 82 hand-written assertions, including agreement with the law of cosines on three thousand random plane pairs, and that the angle is symmetric in its two arguments and unchanged by scaling either vector. The count and the per-case breakdown are published on the formula verification page.

Related calculators

Where this goes next:

More in Math, or browse all calculators.

Read the guide

For rotations in three dimensions rather than a single angle between directions, the Quaternion Calculator handles the composition that Euler angles make awkward.