ConceptComplete

Linear Combinations and Span

The notion of linear combination is the engine of linear algebra. It tells us what vectors we can "reach" using addition and scalar multiplication.


Definitions

Definition1.5Linear combination

Let VV be a vector space over FF. A linear combination of vectors v1,v2,,vnVv_1, v_2, \ldots, v_n \in V is any vector of the form

a1v1+a2v2++anvna_1 v_1 + a_2 v_2 + \cdots + a_n v_n

where a1,a2,,anFa_1, a_2, \ldots, a_n \in F are scalars.

Definition1.6Span

The span (or linear span) of a set S={v1,,vn}VS = \{v_1, \ldots, v_n\} \subseteq V is the set of all linear combinations of vectors in SS:

span(S)={a1v1++anvnaiF}.\text{span}(S) = \{a_1 v_1 + \cdots + a_n v_n \mid a_i \in F\}.

By convention, span()={0}\text{span}(\varnothing) = \{0\}.

We say SS spans VV (or SS is a spanning set for VV) if span(S)=V\text{span}(S) = V.

Theorem1.4Span is a subspace

For any subset SVS \subseteq V, span(S)\text{span}(S) is a subspace of VV. Moreover, it is the smallest subspace containing SS: if WW is any subspace with SWS \subseteq W, then span(S)W\text{span}(S) \subseteq W.

Proof

span(S)\text{span}(S) contains 00 (take all coefficients to be 00). If u=aiviu = \sum a_i v_i and w=biviw = \sum b_i v_i are in span(S)\text{span}(S), then u+w=(ai+bi)viu + w = \sum (a_i + b_i)v_i and cu=(cai)vicu = \sum (ca_i)v_i are in span(S)\text{span}(S). By the subspace test, span(S)\text{span}(S) is a subspace. Since every subspace containing SS must contain all linear combinations of elements of SS, the minimality follows.


Examples

ExampleSpanning R^2

span{(1,0),(0,1)}=R2\text{span}\{(1, 0), (0, 1)\} = \mathbb{R}^2 because every (a,b)=a(1,0)+b(0,1)(a, b) = a(1, 0) + b(0, 1).

Also span{(1,0),(0,1),(1,1)}=R2\text{span}\{(1, 0), (0, 1), (1, 1)\} = \mathbb{R}^2: the third vector is redundant since (1,1)=(1,0)+(0,1)(1, 1) = (1, 0) + (0, 1).

ExampleSpan of a single vector

span{(2,3)}={t(2,3)tR}={(2t,3t)tR}\text{span}\{(2, 3)\} = \{t(2, 3) \mid t \in \mathbb{R}\} = \{(2t, 3t) \mid t \in \mathbb{R}\}, which is the line through the origin with slope 3/23/2.

ExampleSpan of two independent vectors in R^3

span{(1,0,1),(0,1,1)}={(a,b,a+b)a,bR}\text{span}\{(1, 0, 1), (0, 1, 1)\} = \{(a, b, a + b) \mid a, b \in \mathbb{R}\}, which is the plane z=x+yz = x + y in R3\mathbb{R}^3.

ExampleSpan of parallel vectors

span{(1,2),(3,6)}=span{(1,2)}\text{span}\{(1, 2), (3, 6)\} = \text{span}\{(1, 2)\} because (3,6)=3(1,2)(3, 6) = 3(1, 2). Two parallel vectors span only a line, not a plane. Geometrically, adding a redundant vector does not enlarge the span.

ExampleSpanning polynomials

span{1,x,x2}=P2(R)\text{span}\{1, x, x^2\} = P_2(\mathbb{R}), the space of polynomials of degree at most 2.

Alternatively, span{1,1+x,1+x+x2}=P2(R)\text{span}\{1, 1+x, 1+x+x^2\} = P_2(\mathbb{R}) as well, since: 1=11 = 1, x=(1+x)1x = (1+x) - 1, and x2=(1+x+x2)(1+x)x^2 = (1+x+x^2) - (1+x).

ExampleSpanning a matrix space

The four matrices

E11=(1000),E12=(0100),E21=(0010),E22=(0001)E_{11} = \begin{pmatrix} 1 & 0 \\ 0 & 0 \end{pmatrix}, \quad E_{12} = \begin{pmatrix} 0 & 1 \\ 0 & 0 \end{pmatrix}, \quad E_{21} = \begin{pmatrix} 0 & 0 \\ 1 & 0 \end{pmatrix}, \quad E_{22} = \begin{pmatrix} 0 & 0 \\ 0 & 1 \end{pmatrix}

span M2×2(F)M_{2 \times 2}(F), since any matrix (abcd)=aE11+bE12+cE21+dE22\begin{pmatrix} a & b \\ c & d \end{pmatrix} = aE_{11} + bE_{12} + cE_{21} + dE_{22}.

ExampleThree vectors in R^3

Do the vectors v1=(1,2,3)v_1 = (1, 2, 3), v2=(4,5,6)v_2 = (4, 5, 6), v3=(7,8,9)v_3 = (7, 8, 9) span R3\mathbb{R}^3?

We check: v3=2v2v1v_3 = 2v_2 - v_1 (since (7,8,9)=2(4,5,6)(1,2,3)(7,8,9) = 2(4,5,6) - (1,2,3)). So span{v1,v2,v3}=span{v1,v2}\text{span}\{v_1, v_2, v_3\} = \text{span}\{v_1, v_2\}, which is a plane in R3\mathbb{R}^3. These three vectors do not span R3\mathbb{R}^3.

ExampleSpan of trigonometric functions

In C(R)C(\mathbb{R}), span{sinx,cosx}\text{span}\{\sin x, \cos x\} is the set of all functions asinx+bcosxa\sin x + b\cos x with a,bRa, b \in \mathbb{R}. This can also be written as {Asin(x+φ)A0,φR}\{A\sin(x + \varphi) \mid A \geq 0, \varphi \in \mathbb{R}\}.

ExampleSpan of infinitely many vectors

span{1,x,x2,x3,}=P(R)\text{span}\{1, x, x^2, x^3, \ldots\} = P(\mathbb{R}), the space of all polynomials. This is an infinite-dimensional example: no finite subset spans all of P(R)P(\mathbb{R}).

ExampleSpan of exponential functions

In C(R)C^\infty(\mathbb{R}), span{ex,e2x,e3x}\text{span}\{e^x, e^{2x}, e^{3x}\} is a 3-dimensional subspace. These functions are linearly independent (as can be verified using a Wronskian determinant).

ExampleVectors that do not span

span{(1,0,0),(0,1,0)}\text{span}\{(1, 0, 0), (0, 1, 0)\} is the xyxy-plane in R3\mathbb{R}^3, not all of R3\mathbb{R}^3. The vector (0,0,1)(0, 0, 1) cannot be expressed as a linear combination of (1,0,0)(1, 0, 0) and (0,1,0)(0, 1, 0).

ExampleFinding a spanning set for a subspace

The subspace W={(x,y,z)R3x+2yz=0}W = \{(x, y, z) \in \mathbb{R}^3 \mid x + 2y - z = 0\} can be parametrized by setting y=sy = s, z=tz = t, giving x=2s+tx = -2s + t. So

W={s(2,1,0)+t(1,0,1)s,tR}=span{(2,1,0),(1,0,1)}.W = \{s(-2, 1, 0) + t(1, 0, 1) \mid s, t \in \mathbb{R}\} = \text{span}\{(-2, 1, 0), (1, 0, 1)\}.


Properties of span

RemarkMonotonicity

If S1S2S_1 \subseteq S_2, then span(S1)span(S2)\text{span}(S_1) \subseteq \text{span}(S_2). Adding more vectors can only enlarge the span.

RemarkIdempotence

span(span(S))=span(S)\text{span}(\text{span}(S)) = \text{span}(S). The span of a subspace is itself.

RemarkLooking ahead

A spanning set may contain redundant vectors. The quest for a spanning set with no redundancy leads to the notion of a basis, which in turn requires linear independence.