Two competing definitions of the direct sum of vector spaces

116 Views Asked by At

There seem to be two competing definitions of the direct sum of vector spaces. The first one characterises it as the same as the Cartesian product for a finite number of vector spaces, and for an infinite of vector spaces it imposes the restriction that no element of the direct sum may have an infinite number of non-zero entries. Sic Wikipedia: "If the index set is finite, the direct sum is the same as the direct product".

A second definition is that the direct sum differs from the Cartesian product even for a finite number of vector spaces. The Cartesian product of $X$ and $Y$ is identified with the pairs $(x,y)$ s.t. $x \in X$ and $y \in Y$, while the direct sum is also endowed with the following linear structure:

$$ (x_1,y_1)+_{X \oplus Y}(x_2,y_2):=(x_1+_X x_2,y_1+_Y y_2) \\ \alpha \cdot_{X \oplus Y}(x,y):=(\alpha \cdot_X x,\alpha \cdot_Y y), $$ where $+_{X},\cdot_X$ are the vector sum and scalar multiplications on $X$ (similarly for $+_{Y},\cdot_Y$). Additionally, this definition carries too the restriction that no element of the direct sum may have an infinite number of non-zero entries.

My question: is the restriction that elements of the direct sum may not have an infinite number of non-zero entries motivated by the fact that the direct sum is equipped with a vector space structure? I suspect some or another vector space axiom or nice property may fail if we allow $(x_j)_{j\in I}\in \oplus_{i\in I} X_i$ with $|I| = \infty$ and $X_i$ being a vector space $\forall i \in I$, but $|\{j \in I | x_j \neq 0 \}| = \infty$.

1

There are 1 best solutions below

0
On

The direct sum of a finite number of vector spaces is (can be defined to be (1)) the Cartesian product (as a set) with the vector space operations defined componentwise.

This vector space structure can be defined on infinite Cartesian products of vector spaces, and yields a vector space. All the axioms are satisfied.

The subset of that infinite Cartesian product consisting of tuples in which all but finitely many entries are $0$ is a vector subspace. Sometimes that is the one you want to work with in a particular context.

What you name these vector spaces (direct sum, direct product, ...) is a matter of convention.

(1) In category theory you define the direct sum using the existence and uniqueness of maps with certain properties.