Linear combinations, span, and basis vectors | Chapter 2, Essence of linear algebra

This video explores the foundational concepts of linear algebra: basis vectors, linear combinations, and span. It begins by reinterpreting standard Cartesian coordinates (x, y) as scalars that stretch or flip the unit basis vectors $\hat{i}$ and $\hat{j}$. The core idea is extended to arbitrary pairs of non-collinear vectors, demonstrating that their linear combinations can reach any point in a 2D plane, defining the 'span' of those vectors. The visualizations then transition to 3D space, showing how two non-parallel vectors span a flat plane, while adding a third vector outside that plane expands the span to fill all 3D space. Finally, it introduces linear dependence (redundancy) versus independence (each vector adds a new dimension), culminating in the formal definition of a basis.

Before you watch

• Basic understanding of vectors as directed line segments with magnitude and direction.
• Familiarity with 2D Cartesian coordinate systems and plotting points.
• Conceptual grasp of scalar multiplication (stretching/shrinking) and vector addition (tip-to-tail method).

Chapters

Learning script

Generated from the video's visuals and explanation; not verbatim speech.

0:00We usually think of a coordinate pair like $(3, -2)$ simply as a location on a grid. However, in linear algebra, it is far more powerful to view these numbers as scaling factors. Consider the standard unit vectors pointing right and up, denoted as $\hat{i}$ and $\hat{j}$. A coordinate pair tells us exactly how much to stretch or shrink these specific arrows. For instance, the first number scales $\hat{i}$ horizontally, and the second scales $\hat{j}$ vertically. When you place these scaled vectors tip-to-tail and add them together, the resulting diagonal arrow represents the original coordinate.

1:27These starting unit vectors are known as the 'basis' of the coordinate system. But what happens if we choose entirely different vectors as our foundation? Suppose we pick two arbitrary vectors, let's call them $\vec{v}$ and $\vec{w}$, which point in different directions and do not lie on the same line. By multiplying $\vec{v}$ by some scalar $a$ and $\vec{w}$ by another scalar $b$, and then adding them together ($a\vec{v} + b\vec{w}$), we create what is called a 'linear combination'. If we allow both scalars $a$ and $b$ to vary continuously across all real numbers, the tips of these combined vectors sweep out every possible point on the 2D plane.

4:43Showing vector endpoints as points reveals the set of all linear combinations. Two nonzero, non-collinear vectors span the plane. Nonzero collinear vectors span a line, and an all-zero set spans only the origin. The span uses every real coefficient, not just the points visible in the animation.

5:56Let us elevate this concept into three-dimensional space. Imagine selecting two non-parallel vectors in 3D. Their individual spans form lines, but when you combine them via $a\vec{v} + b\vec{w}$, the resulting endpoint traces out a tilted, infinitely extending flat sheet cutting through the origin. This plane is the exact span of those two vectors. Now, introduce a third vector, $\vec{u}$. There are two distinct scenarios depending on where this third arrow points relative to our existing plane.

8:16If the third vector lies perfectly flat on the plane already created by the first two, it provides no new directional freedom; mathematically, it is redundant because it can be written as a combination of the others. We say such a set is 'linearly dependent.' Conversely, if the third vector pokes out of the plane at an angle, scaling it allows the entire 2D sheet to slide up and down through space. With three freely varying scalars ($a\vec{v} + b\vec{w} + c\vec{u}$), the span now encompasses every conceivable point in 3D volume. Because each vector contributed a brand-new dimension, they are 'linearly independent,' forming a perfect basis for the space.

Knowledge cards