nLab basis of a vector space





For 𝕂\mathbb{K} a field and VV a 𝕂\mathbb{K}-vector space, hence a 𝕂\mathbb{K}-module, a linear basis for VV is

  1. a set BB (of basis elements)

  2. a 𝕂\mathbb{K}-linear isomorphism

    B𝕂V \underset{B}{\oplus} \mathbb{K} \xrightarrow{\;\; \simeq \;\;} V

to VV from the BB-indexed direct sum (the coproduct in 𝕂\mathbb{K}Vect) of copies of 𝕂\mathbb{K} (canonically regarded as a vector space over itself), also known as the free 𝕂\mathbb{K}-linear span of BB, hence the vector space of free 𝕂\mathbb{K}-linear combinations of elements of BB.

Hence if a basis for VV exists it means in particular that VV is a free module over 𝕂\mathbb{K}.


(finiteness of linear combinations)
Beware that a certain finiteness-condition is hidden in Def. : Since a linear combination is defined to be a sum of finitely many vectors, a basis of a vector space must be such that every vector in the space is the (unique) combination of finitely many basis elements – even if there are infinitely many elements in the basis.

More abstractly this is to do with the appearance of the coproduct ⨿ b= B\amalg_b = \oplus_B (direct sum) in Def. instead of the product B\prod_B. Many vector spaces in practice arise as (subspaces) of products W𝕂\prod_W \mathbb{K} (function spaces W𝕂W \to \mathbb{K}), but if WW here is not a finite set then it is not going to be a basis set.

(On the other hand, if WW is a finite set, then we have a biproduct W W W\prod_W \,\simeq\, \coprod_W \,\simeq\, \oplus_W in the additive category 𝕂\mathbb{K}Vect, see there.)

Related to this are the following phenomena:


(basis and dimension of finitely-generated spaces)
For every finitely generated vector space VV (Def. ) it is straightforward to construct a linear basis, and to see that the cardinality of all bases is the same finite natural number dim(V)dim(V) \in \mathbb{N}, called the dimension of the vector space (whence a finite-dimensional vector space).

On the other hand:


(Hamel-bases of infinite-dimensional vector spaces)
While the definition applies also to not-necessarily finitely generated vector spaces – such as for instance the space of (continuous) functions from a non-finite (topological) space to the (topological) ground field – it turns out to be subtle and somewhat ill-behaved in this generality.

In fact, in practice infinite-dimensional vector spaces tend to appear and to be understood with extra structure (typically that of topological vector spaces such as Banach spaces or Hilbert spaces) in which cases there are more appropriate notions of linear bases for them (such as that of Schauder bases, which allow infinite-linear combinations subject to a condition of convergence of a sequence).

In order to distinguish the plain notion of basis (Def. ) from these more refined notions, one also speaks of Hamel bases here.

(This is in honor of Hamel 1905 pp. 460 who considered this notion for the special case of V=V = \mathbb{R} the real numbers regarded as a rational vector space, hence over the ground field 𝕂=\mathbb{K} = \mathbb{Q} of rational numbers).

Hence, in principle, also a linear basis of a finitely generated vector space is thus a Hamel basis, but rarely called this way unless in the context of infinite-dimensional vector spaces.


(basis theorem and dimension)
For an infinitely-generated vector space it is not in general possible to construct a (Hamel-)basis, but the existence of such a basis is nevertheless implied, in classical mathematics, by Zorn's lemma (essentially a form of the axiom of choice): This is the content of the basis theorem.

With this classical context understood, it follows that every vector vector space admits a linear basis (even if non-constructible in general) and that each basis is of the same cardinality, then called the dimension of the vector space.

Background definitions


(generated vector space)
For VV a vector space, a subset GVG \subset V of its underlying set is called a generating set or spanning set if every element of VV can be expressed as a linear combination of elements of GG in VV, hence if the linear span of GG inside VV is all of VV.

The vector space VV is called finitely generated if it admits a generating set (spanning set) which is a finite set.


Lecture notes with much conceptual exposition:

  • Karen E Smith, Bases for infinite-dimensional vector spaces [pdf, pdf]

Lecture notes with the proofs concisely spelled out:

See also:

The original discussion (for \mathbb{R} regarded as a rational vector space) after which Hamel bases are named:

  • Georg Hamel, pp. 460 of: Eine Basis aller Zahlen und die unstetigen Lösungen der Funktionalgleichung: f(x+y)=f(x)+f(y)f(x + y) = f(x) + f(y), Mathematische Annalen 60 (1905) 459–462 [doi:10.1007/BF01457624]

Last revised on August 26, 2023 at 17:24:41. See the history of this page for a list of all contributions to it.