3.3-exam1-2009-b¶
Quiz 1, 2009 · Section 3.3: The Complete Solution to Ax = b Source: exam1_1.pdf, pp. 210-212
Problem¶
Problem 3: \(A\) is a matrix with a nullspace \(N(A)\) spanned by the following three vectors:
(\(\alpha\)) Give a matrix \(B\) such that its column space \(C(B)\) is the same as \(N(A)\). (There is more than one correct answer.) [Thus, any vector \(\vec{y}\) in the nullspace of \(A\) satisfies \(B\vec{u} = \vec{y}\) for some \(\vec{u}\).]
(\(\beta\)) Give a different possible answer to (\(\alpha\)): another \(B\) with \(C(B) = N(A)\).
(\(\gamma\)) For some vector \(\vec{b}\), you are told that a particular solution to \(A\vec{x} = \vec{b}\) is
Now, your classmate Zarkon tells you that a second solution is:
while your other classmate Hastur tells you "No, Zarkon's solution can't be right, but here's a second solution that is correct:"
Is Zarkon's solution correct, or Hastur's solution, or are both correct? (Hint: what should be true of \(\vec{x} - \vec{x}_p\) if \(\vec{x}\) is a valid solution?)
Solution
(\(\alpha\)) Since the nullspace is spanned by the given three vectors, we may simply take \(B\) to consist of the three vectors as columns, i.e., \[ B = \begin{bmatrix} 1 & 0 & -1 \\ 2 & 1 & -1 \\ -1 & 1 & 3 \\ 3 & 4 & 1 \end{bmatrix}. \] \(B\) need not be square (many students insisted on square solutions). (\(\beta\)) For example, we may simply add a zero column to \(B\): \[ B = \begin{bmatrix} 1 & 0 & -1 & 0 \\ 2 & 1 & -1 & 0 \\ -1 & 1 & 3 & 0 \\ 3 & 4 & 1 & 0 \end{bmatrix}. \] Or, we could interchange two columns. Or we could multiply one of the columns by \(-1\). For example: \[ B = \begin{bmatrix} 1 & 0 & 1 \\ 2 & 1 & 1 \\ -1 & 1 & -3 \\ 3 & 4 & -1 \end{bmatrix}. \] Or we could replace one of the columns by a linear combination of that column with the other two columns (any invertible column operation). Or we could replace \(B\) by \(-B\) or \(2B\). There are many possible solutions. In any case, the solution shouldn't require any significant calculation! (\(\gamma\)) Since any solution \(\vec{x}\) to the equation \(A\vec{x} = \vec{b}\) is of the form \(\vec{x}_p + \vec{n}\) for some vector \(\vec{n}\) in the nullspace, the vector \(\vec{x} - \vec{x}_p\) must lie in the nullspace \(N(A)\). Thus, we want to look at: \[ \vec{x}_Z - \vec{x}_p = \begin{bmatrix} 0 \\ -1 \\ 0 \\ -4 \end{bmatrix}, \qquad \vec{x}_H - \vec{x}_p = \begin{bmatrix} 0 \\ -1 \\ 0 \\ -3 \end{bmatrix}. \] To determine whether a vector \(\vec{y}\) lies in the nullspace \(N(A)\), we can just check whether it is in the column space of \(B\), i.e. check whether \(B\vec{z} = \vec{y}\) has a solution. As we learned in class, we can check this just by doing elimination: if elimination produces a zero row in \(B\), it should produce a zero row in the right-hand side. In terms of \(B\) from part (\(\alpha\)) augmented by the right-hand side, this gives: \[ \begin{bmatrix} 1 & 0 & -1 & 0 \\ 2 & 1 & -1 & -1 \\ -1 & 1 & 3 & 0 \\ 3 & 4 & 1 & a \end{bmatrix} \leadsto \begin{bmatrix} 1 & 0 & -1 & 0 \\ 0 & 1 & 1 & -1 \\ 0 & 1 & 2 & 0 \\ 0 & 4 & 4 & a \end{bmatrix} \leadsto \begin{bmatrix} 1 & 0 & -1 & 0 \\ 0 & 1 & 1 & -1 \\ 0 & 0 & 1 & 1 \\ 0 & 0 & 0 & a + 4 \end{bmatrix}. \] We can get a solution if and only if \(a = -4\). So Zarkon is correct (and Hastur is wrong). REMARK: Some students found a linear combination that gave \(\vec{x}_Z - \vec{x}_p\) by inspection: \[ \vec{x}_Z - \vec{x}_p = \begin{bmatrix} 0 \\ -1 \\ 0 \\ -4 \end{bmatrix} = 1 \begin{bmatrix} 1 \\ 2 \\ -1 \\ 3 \end{bmatrix} - 2 \begin{bmatrix} 0 \\ 1 \\ 1 \\ 4 \end{bmatrix} + 1 \begin{bmatrix} -1 \\ -1 \\ 3 \\ 1 \end{bmatrix}. \] Then they stared at Hastur's solution, couldn't find such a combination, and concluded that it was not a solution. This conclusion is correct *in this case*, and was awarded full marks because you were not asked to justify your solution. However, doing elimination is much more systematic and reliable, and ensures that there isn't a linear combination that you simply missed. Use elimination next time!Keywords: nullspace, column space, complete solution, consistency