Ma · Mathematics

What Does “Extend Linearly” Mean in Linear Algebra?

Published By the 9to5Science editorial team

At a glance

  1. 1. Start with known values

    Specify the vectors whose images are already known, often a basis or a basis for a subspace.

  2. 2. Use linear combinations

    Send each linear combination to the same combination of the specified images.

  3. 3. Check consistency

    If the starting vectors are dependent, their assigned images must preserve every linear relation.

  4. 4. Identify choices

    Values on additional basis vectors determine an extension, so different choices can give different maps.

Advertisement

“Extend linearly” means use linearity to determine the value of a map on vectors built from vectors whose images are already known. If TT is linear and v1,…,vkv_1,\ldots,v_k are known, then

T(a1v1+⋯+akvk)=a1T(v1)+⋯+akT(vk).T(a_1v_1+\cdots+a_kv_k)=a_1T(v_1)+\cdots+a_kT(v_k).

The coefficients do not change: you apply the map to each vector, then combine the results with those same coefficients. This is useful when a map is specified only on a basis or on a subspace and you need to define it on more vectors.

Extending from a basis

Suppose v1,…,vnv_1,\ldots,v_n is a basis of a vector space VV, and you choose vectors w1,…,wnw_1,\ldots,w_n in another vector space WW. Every v∈Vv\in V has exactly one expression

v=a1v1+⋯+anvn.v=a_1v_1+\cdots+a_nv_n.

Define

T(v)=a1w1+⋯+anwn.T(v)=a_1w_1+\cdots+a_nw_n.

This defines a unique linear map with T(vi)=wiT(v_i)=w_i for each basis vector. Uniqueness follows because every vector has just one set of basis coefficients: there is no conflicting way to calculate its image.

In practice, “extend linearly” often appears as an instruction after the values on basis vectors are given. You do not need a separate formula for every vector; express it in the basis and use the rule above.

Worked example

Define a linear map T:R2→R2T:\mathbb{R}^2\to\mathbb{R}^2 by

T(1,0)=(2,1),T(0,1)=(−1,3),T(1,0)=(2,1),\qquad T(0,1)=(-1,3),

and extend it linearly. The vectors (1,0)(1,0) and (0,1)(0,1) are the standard basis. Since

(x,y)=x(1,0)+y(0,1),(x,y)=x(1,0)+y(0,1),

linearity requires

T(x,y)=xT(1,0)+yT(0,1)=x(2,1)+y(−1,3)=(2x−y, x+3y).\begin{aligned} T(x,y) &=xT(1,0)+yT(0,1)\\ &=x(2,1)+y(-1,3)\\ &=(2x-y,\ x+3y). \end{aligned}

For example, (3,2)=3(1,0)+2(0,1)(3,2)=3(1,0)+2(0,1), so

T(3,2)=3(2,1)+2(−1,3)=(4,9).T(3,2)=3(2,1)+2(-1,3)=(4,9).

The map is determined on every vector in R2\mathbb{R}^2 by the two given basis values.

Extending a map from a subspace

The phrase can also mean extending a map already defined on a subspace U⊆VU\subseteq V so that it is defined on all of VV. In finite dimensions, choose a basis of UU and add vectors to obtain a basis of VV. The map’s values on the basis of UU are fixed; choose images for the added basis vectors, then extend linearly.

For example, let UU be the xx-axis in R2\mathbb{R}^2, and suppose T(x,0)=(x,0)T(x,0)=(x,0). To extend TT to the whole plane, choose an image for the additional basis vector (0,1)(0,1). If you set T(0,1)=(0,0)T(0,1)=(0,0), then T(x,y)=(x,0)T(x,y)=(x,0). If instead you set T(0,1)=(0,1)T(0,1)=(0,1), then T(x,y)=(x,y)T(x,y)=(x,y). Both extensions agree with the given map on the xx-axis, but they differ elsewhere.

Thus an extension need not be unique. The phrase “extend linearly” tells you how to calculate values once the images of a basis are chosen; it does not always tell you what those images must be.

When the rule is well-defined

A basis is linearly independent, so each vector has a unique basis representation. That makes arbitrary choices of images on a basis safe. But if you start with a dependent set, a vector may have more than one representation. The assigned images must then give the same result for every representation.

For instance, suppose v2=2v1v_2=2v_1. A linear map must satisfy T(v2)=2T(v1)T(v_2)=2T(v_1). You cannot independently assign an unrelated value to T(v2)T(v_2) and still have a well-defined linear map on the span. More generally, every linear relation among the starting vectors must also hold among their assigned images.

Common mistakes

  • Changing the coefficients: If v=av1+bv2v=av_1+bv_2, then T(v)=aT(v1)+bT(v2)T(v)=aT(v_1)+bT(v_2), not a combination with new coefficients.
  • Assuming any assignments work: Arbitrary images work for a basis, but assignments on dependent vectors must respect their linear relations.
  • Assuming the extension is unique: A map on a proper subspace usually has multiple extensions unless extra conditions fix the images outside that subspace.
  • Confusing linear with affine: A linear map satisfies T(0)=0T(0)=0 and preserves sums and scalar multiples. Adding a fixed vector to every output generally makes a map affine, not linear.

When you see “extend linearly,” identify the vectors with known images, express the target vector as a linear combination, and apply the same coefficients to those images. Then check that different representations cannot produce conflicting answers.

References

  1. Sheldon Axler, Linear Algebra Done Right

Spotted an error? Send a correction. Read how we write and check articles in our editorial policy.