Lecture 28 - The Null Space of a Matrix

Learning Objectives

Null Space

Definition. Let \( A \) be an \( m\times n \) matrix. The null space of \( A \), written \( \Nul A \), is the set of all vectors \( \bbm x \in \mathbb R^n \) such that \( A \bbm x= \bbm 0 \).

In set-builder notation, we write \( \Nul A = \{ \bbm x\in \mathbb R^n : A\bbm x = \bbm 0 \} \). The elements of \( \Nul A \) are the vectors \( \bbm x\) with the property that \( A \bbm x = \bbm 0 \).

Example 1. Let \( A = \begin{bmatrix} 1 & -3 & -2 \\ -5 & 9 & 1 \end{bmatrix} \). Is \( \bbm u = \vecthree 5 3 {-2} \) in \( \Nul A \)?

We compute \( A \bbm u = \begin{bmatrix} 1 & -3 & -2 \\ -5 & 9 & 1 \end{bmatrix} \vecthree 5 3 {-2} = \vectwo 00 \). Since \( A \bbm u = \bbm 0\), yes, \( \bbm u \) is in \( \Nul A \). \( \Box \)

Example 2. Let \( B = \begin{bmatrix} -3 & 6 \\ 1 & -2 \\ -2 & 4 \end{bmatrix} \). For what values of \( h \) is \( \bbm v = \vectwo {-4} h \) in \( \Nul B \)?

We compute \( B\bbm v = \begin{bmatrix} -3 & 6 \\ 1 & -2 \\ -2 & 4 \end{bmatrix}\vectwo {-4} h = \vecthree {12+6h} {-4-2h} {8+4h} \). In order for \( B\bbm v = \bbm 0\), we must have \( 12+6h = 0\), \(-4-2h = 0\), and \( 8+4h =0\) for the same value of \( h \). The only solution is \( h = -2 \). \( \Box \)

Definition. Let \( T : \mathbb R^n \to \mathbb R^m \) be a linear transformation. The kernel of \( T \), written \( \mbox{ker } T \), is the set of all vectors \( \bbm x \in \mathbb R^n \) such that \( T(\bbm x) = \bbm 0 \).

Note that, if \( A \) is the standard matrix for \( T \), then \( \mbox{ker } T \) is the same as \( \Nul A \) since \( T(\bbm x)=A\bbm x \).

The Null Space Is a Subspace

Theorem (Null Space). Let \( A \) be an \( m\times n\) matrix. Then \( \Nul A \) is a subspace of \( \mathbb R^n \).

Using the subspace defintion from Lecture 27, we must show that \( \Nul A \) contains the zero vector, is closed under addition, and is closed under scalar multiplication.

Spanning the Null Space

Our definition for null space is said to be "implicit," since it is define by a condition that must be checked. There is no obvious way to generate elements of \( \Nul A \) just from the definition. We would like to find a spanning set for \( \Nul A \):

Defintion. Given a subspace \( H \) of \( \mathbb R^n \), a spanning set for \( H \) is a finite set of vectors \( \{ \bbm v_1, \ldots, \bbm v_p \} \) such that \( \mbox{Span} \{ \bbm v_1, \ldots, \bbm v_p \} = H \).

Once you have a spanning set for a subspace, it is easy to generate new elements of that subspace: just form linear combinations of vectors in the spanning set. To find a spanning set for \( \Nul A \), we start by solving the matrix equation \( A \bbm x = \bbm 0 \) in the normal way.

Example 3. Find a spanning set for \( \Nul A \), where \( A = \begin{bmatrix} -3 & 6 & -1 & 1 & -7 \\ 1 & -2 & 2 & 3 & -1 \\ 2 & -4 & 5 & 8 & -4 \end{bmatrix} \).

We form the augmented matrix corresponding to \( A \bbm x = \bbm 0 \) and row-reduce: \[ \begin{bmatrix} -3 & 6 & -1 & 1 & -7 & 0 \\ 1 & -2 & 2 & 3 & -1 & 0 \\ 2 & -4 & 5 & 8 & -4 & 0 \end{bmatrix} \longrightarrow \begin{bmatrix} 1 & -2 & 0 & -1 & 3 & 0 \\ 0 & 0 & 1 & 2 & -2 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 \end{bmatrix} \]

Now, write the solution in parametric vector form as we learned in Lecture 10: \[ \bbm x = \vecfive {x_1} {x_2} {x_3} {x_4} {x_5} = \vecfive {2x_2+x_4-3x_5} {x_2} {-2x_4+2x_5} {x_4} {x_5} = x_2 \vecfive 21000 + x_4 \vecfive 10{-2}10 + x_5 \vecfive {-3}0201. \]

From this, we can see that every solution of \( A \bbm x = \bbm 0 \) can be written as a linear combination of \( \bbm v_1 = \vecfive 21000, \bbm v_2 = \vecfive 10{-2}10 \), and \( v_3 = \vecfive {-3}0201 \). So, \( \{ \bbm v_1, \bbm v_2, \bbm v_3 \}\) is a spanning set for \( \Nul A \). \( \Box \)

Example 4. Suppose that \( B = \begin{bmatrix} -4 & 0 & 1 & 4 \\ 2 & 0 & 3 & 5 \\ -6 & 0 & -2 & -1 \end{bmatrix} \), which row-reduces to \( \begin{bmatrix} 1 & 0 & 0 & -1/2 \\ 0 & 0 & 1 & 2 \\ 0 & 0 & 0 & 0 \end{bmatrix} \). Find a spanning set for \( \Nul B \).

Since the row reduction has been done for us, we can now write the parametric solution to \( B \bbm x = \bbm 0 \): \[ \bbm x = \vecfour {x_1} {x_2} {x_3} {x_4} = x_2 \vecfour 0 1 0 0 + x_4 \vecfour {1/2} 0 {-2} 1. \]

So, a spanning set for \( \Nul B \) is \( \left\{ \vecfour 0 1 0 0, \vecfour {1/2} 0 {-2} 1 \right\} \). \( \Box \)

From these examples, you should notice that the spanning set we construct for \( \Nul A \) contains one vector for each free variable in the equation \( A \bbm x = \bbm 0\). The Linearly Independent Columns Theorem implies that \( \Nul A = \{ \bbm 0 \} \) exactly when \( A \) has a pivot in every column.

Challenge Question. What can you say about an \( m\times n\) matrix \( A \) if \( \Nul A = \mathbb R^n \) ?

Reasoning About Abstract Matrices

Sometimes we have limited information about a matrix, but we can still sometimes draw conclusions about its null space.

Example 5. Suppose that \( A \) is a \( 3\times 4\) matrix whose third column is all zeroes. Find a nonzero vector in \( \Nul A \).

We are looking for a nonzero vector \( \bbm x \) for which \( A \bbm x = \bbm 0 \). Remember that \( A \bbm x \) is just a linear combination of the columns of \( A\). Now, \( A \vecfour 0010 \) will equal the third column of \( A \), which we know is all zeroes. So, \( \bbm x= \vecfour 0010 \) (or any nonzero multiple of this vector) is a nonzero vector that is guaranteed to be in \( \Nul A \). \( \Box \)

Example 6. Suppose that \( A \) is a \( 2\times 5 \) matrix whose second column equals 3 times its fourth column. Find a nonzero vector in \( \Nul A \).

Write \( \bbm a_1, \ldots, \bbm a_5 \) for the columns of \( A \). We are given that \( \bbm a_2 = 3 \bbm a_4 \). We can rewrite this as \( \bbm a_2 - 3\bbm a_4 = \bbm 0 \), or \( A \vecfive 0 1 0 {-3} 0 = \bbm 0 \). Thus \( \vecfive 0 1 0 {-3} 0 \in \Nul A \). \( \Box \)

« Lecture 27 Back to Top Lecture 29 »