Skip to content

3.2-exam1-2008-strang

Quiz 1, 2008 (strang) · Section 3.2: Computing the Nullspace by Elimination: A = CR Source: exam1_1.pdf, pp. 197-198

Problem

3 (30 pts.) Suppose the nonzero vectors \(a_1, a_2, a_3\) point in different directions in \(\mathbb{R}^3\) but

\[ 3a_1 + 2a_2 + a_3 = \text{zero vector}. \]

The matrix \(A\) has those vectors \(a_1, a_2, a_3\) in its columns.

(a) Describe the nullspace of \(A\) (all \(x\) with \(Ax = 0\)).

(b) Which are the pivot columns of \(A\)?

(c) I want to show that all 3 by 3 matrices with

\[ (*) \quad 3(\text{column 1}) + 2(\text{column 2}) + (\text{column 3}) = \text{zero vector} \]

form a subspace \(S\) of the space \(M\) of 3 by 3 matrices. Now the zero matrix is certainly included.

Suppose \(B\) and \(C\) are matrices whose columns have this property \((*)\). To show that we have a subspace, we have to prove that every linear combination of \(B\) and \(C\) (finish sentence) _.

Go ahead and prove that.

Solution (a) First, the problem gives us one vector in the nullspace: the equation \[ 3a_1 + 2a_2 + a_3 = \text{zero vector} \] is the same as saying that \(Ax = 0\), where \[ x = \begin{bmatrix} 3 \\ 2 \\ 1 \end{bmatrix}. \] Thus, the nullspace \(N(A)\) contains all scalar multiples of this vector \((3, 2, 1)\). Technically, we need to argue that there is nothing else in the nullspace. You must recognize that the matrix has rank 2 either here or in part (b) to get full credit. (b) Since we have a linear relationship between the columns, column 3 cannot be a pivot column, so the rank is at most 2. The problem also tells us that the three \(a_i\) vectors point in different directions. This means that they can't all be on the same line; that is, the column space must be at least a plane. Thus the rank is exactly 2. Having rank 2 means that there are 2 pivot columns and one free column. Since column 3 is a linear combination of columns before it, it must be a free column. So columns 1 and 2 are pivots, and column 3 is free. (c) To show that we have a subspace, we have to prove that every linear combination of \(B\) and \(C\) *also has property (*)*. Suppose that \(B\) and \(C\) have property \((*)\). Consider the matrix \(D = t_1B + t_2C\) for any two numbers \(t_1, t_2\). For ease of notation, I'll denote the \(i\)th column of a matrix \(A\) by \(\mathrm{col}_i(A)\). We have \[ \mathrm{col}_i(D) = t_1\mathrm{col}_i(B) + t_2\mathrm{col}_i(C). \] We can now check that \(D\) also satisfies \((*)\): \[ \begin{aligned} 3\mathrm{col}_1(D) + 2\mathrm{col}_2(D) + \mathrm{col}_3(D) &= 3(t_1\mathrm{col}_1(B) + t_2\mathrm{col}_1(C)) \\ &\quad + 2(t_1\mathrm{col}_2(B) + t_2\mathrm{col}_2(C)) \\ &\quad + (t_1\mathrm{col}_3(B) + t_2\mathrm{col}_3(C)) \\ &= t_1(3\mathrm{col}_1(B) + 2\mathrm{col}_2(B) + \mathrm{col}_3(B)) \\ &\quad + t_2(3\mathrm{col}_1(C) + 2\mathrm{col}_2(C) + \mathrm{col}_3(C)) \\ &= 0 \end{aligned} \]

Keywords: nullspace, pivot columns, subspace, linear dependence