universal property of the tensor product

The universal property is the real definition of the tensor product. Forget for a moment how you build V (x) W out of symbols; what truly pins it down is the way it behaves toward every bilinear map at once. The data is a space T together with a fixed bilinear map t : V x W -> T, written t(v, w) = v (x) w.

Here is the property. For any vector space U and any bilinear map B : V x W -> U, there exists one and only one linear map L : T -> U such that B(v, w) = L(v (x) w) for all v, w. In a diagram, B factors as B = L composed with t. The phrase 'one and only one' is doing all the work: existence says T is big enough to record every bilinear map; uniqueness says T has no extra junk beyond what bilinearity forces.

This single statement makes the tensor product unique up to a unique isomorphism. If two spaces T and T' both satisfy the property, the universal maps produce inverse linear maps between them, so any two honest constructions of V (x) W are canonically the same. That is why mathematicians say 'the' tensor product even though there are many ways to build it.

Why work this way? A universal property is a behavioral specification: it tells you what the object DOES rather than what it is made of. Proving facts about (x) then reduces to defining a bilinear map and invoking the property, which is far cleaner than wrestling with bases. This style — define by universal property — recurs throughout exterior and symmetric algebra.