Problem to start with

  1. Do gaussian elimination to put the following matrix in row echelon form. Show your work! $$ \begin{bmatrix} 2 & -4 & 0 & -1\\ 6 & -10 & 10 & -7\\ 4 & -7.5 & 3.5 & 6 \end{bmatrix} $$
  2. Using your solution to the previous problem, find a non-trivial solution to $$ \begin{bmatrix} 2 & -4 & 0 & -1\\ 6 & -10 & 10 & -7\\ 4 & -7.5 & 3.5 & 6 \end{bmatrix} \cdot \boldsymbol{x} = \boldsymbol{0} $$ in which the 4th component of $\boldsymbol{x}$ is $-1$.

    Verify that your $x_1,x_2,x_3$ (the first three components of the non-trivial solution) satisfy: $$ \begin{bmatrix} 2 & -4 & 0 \\ 6 & -10 & 10 \\ 4 & -7.5 & 3.5 \end{bmatrix} \cdot \begin{bmatrix} x_1\\ x_2\\ x_3 \end{bmatrix} = \begin{bmatrix} -1\\ -7\\ 6 \end{bmatrix} $$

Look: you solved a matrix vector product equation with a right-hand side that is not zero by solving a slightly different matrix vector product equation with a right-hand side that is zero!

An analogy

Everyone knows that the solution(s) to $a x^2 + b x + c = 0$ are $x = \frac{-b \pm \sqrt{b^2 - 4ac}}{2a}$, right? Right?

Suppose your english major friend came to you and said, "I have this equation $$ 3 x^2 + 5 x = 2 $$ but I don't know how to solve it because the quadratic equation only works when there is a zero on the right-hand side of the equals sign."

What would you recommend they do? Just move the 2 to the left side of the equals sign and multiply it by -1. Now ... read the next section!

Solving $A\cdot\boldsymbol{x}=\boldsymbol{b}$

We now know how to use Gaussian elmination and back substitution to solve matrix-vector product equations of the form $A\cdot\boldsymbol{x}=\boldsymbol{0}$. But what about $A\cdot\boldsymbol{x}=\boldsymbol{b}$? I.e. what can we do when the left-hand side is not zero? The last problem from the homework pointed the way.

Notation: If $A$ is an $m\times n$ matrix and $\boldsymbol{b}$ is an $n$-dimensional column vector, we write $[A|\boldsymbol{b}]$ to denote the $m\times(n+1)$ matrix formed by adding $\boldsymbol{b}$ as an extra column on the end of $A$.

Let $A$ be an $m\times n$ matrix and $\boldsymbol{b}$ an $n$-dimensional column vector. Vector $\boldsymbol{u}$ is a solution to $A\cdot\boldsymbol{x}=\boldsymbol{b}$ if and only if the column vector $\boldsymbol{u'} = [u_1,\ldots,u_n,-1]$ is a solution to $[A|\boldsymbol{b}]\cdot\boldsymbol{x}=\boldsymbol{0}$.
The following sequence of equivalences shows that the theorem is true. $$ A\cdot\boldsymbol{u}=\boldsymbol{b} \Longleftrightarrow \begin{array}{ccc} a_{1,1} u_1 + a_{1,2} u_2 + ... + a_{1,n} u_n & = & b_1\\ \vdots & &\\ a_{m,1} u_m + a_{m,2} u_2 + ... + a_{m,n} u_n & = & b_m \end{array} \Longleftrightarrow \begin{array}{ccc} a_{1,1} u_1 + a_{1,2} u_2 + ... + a_{1,n} u_n + b_1 (-1)& = & 0\\ \vdots & &\\ a_{m,1} u_m + a_{m,2} u_2 + ... + a_{m,n} u_n + b_m (-1)& = & 0 \end{array} \Longleftrightarrow [A|\boldsymbol{b}]\cdot \begin{bmatrix} u_1\\ \vdots\\ u_n\\ -1 \end{bmatrix} =\boldsymbol{0} \Longleftrightarrow [A|\boldsymbol{b}]\cdot\boldsymbol{u'}=\boldsymbol{0} $$

In class activity

We did the following in-class activity: activity. I'll give a solution to the first one here:

Problem: Solve the following matrix-vector product equation: $$ \begin{bmatrix} 3 & -1 & 4\\ 6 & -4 & 12\\ -1.5 & -5.5 & 12 \end{bmatrix} \cdot \boldsymbol{x} = \begin{bmatrix} 2\\ -1\\ 5 \end{bmatrix} $$ The system is equivalent to: $$ \underbrace{ \begin{bmatrix} 3 & -1 & 4 & 2\\ 6 & -4 & 12 & -1\\ -1.5 & -5.5 & 12 & 5 \end{bmatrix} }_{\text{called the "augmented matrix"}} \cdot \begin{bmatrix} x_1\\ x_2\\ x_3\\ -1 \end{bmatrix} = \boldsymbol{0} \text{, do gaussian elimination: } \begin{bmatrix} 3 & -1 & 4 & 2\\ 0 & -2 & 4 & -5\\ 0 & 0 & 2 & 21 \end{bmatrix} \text{, back sub. with "$x_4$"=-1: } \left\{ \begin{array}{rl} 2x_3=21 &\rightarrow x_3=21/2\\ -2x_2+4x_3=-5 &\rightarrow -2x_2 + 4(21/2) = -5 \rightarrow x_2=47/2\\ 3x_1 - x_2 + 4x_3 = 2 &\rightarrow 3x_1 - 47/2 + 42 = 2 \rightarrow 3x_1 = -33/2 \rightarrow x_1 = -11/2 \end{array} \right. $$ Let's check our work by multiplying $A$ by our proposed solution vector ... $$ \begin{bmatrix} 3 & -1 & 4\\ 6 & -4 & 12\\ -1.5 & -5.5 & 12 \end{bmatrix} \cdot \begin{bmatrix} -11/2\\ 47/2\\ 21/2 \end{bmatrix} = \begin{bmatrix} 2\\ -1\\ 5 \end{bmatrix} $$ Yay!

Note: In the last problem from the activity (Problem c) we find that there is no solution! This is because, after Gaussian elimination, the last row is $[0\ 0\ 0\ 25]$, which when we do back substitition (remembering that the last componant of $x$ is -1), gives us $25\cdot(-1) = 0$, which ... is never satisfied.

Christopher W Brown