Skip to content

3.5-exam1-1997-strang

Quiz 1, 1997 (strang) · Section 3.5: Dimensions of the Four Subspaces Source: exam1_1.pdf, pp. 8-9

Problem

3. \(A\) is an \(m\) by \(n\) matrix of rank \(r\). Suppose \(Ax = b\) has no solution for some right sides \(b\) and infinitely many solutions for some other right sides \(b\).

(a) (9) Decide whether the nullspace of \(A\) contains only the zero vector and why.

(b) (9 points) Decide whether the column space of \(A\) is all of \(R^m\) and why.

(c) (10) For this \(A\), find all true relations between the numbers \(r\), \(m\), and \(n\).

(d) (9 points) Can there be a right side \(b\) for which \(Ax = b\) has exactly one solution? Why or why not?

Solution (a) NO — the nullspace does not contain only the zero vector. If \(Ax = b\) has infinitely many solutions \(x_1 \neq x_2\), their difference is nonzero and satisfies \(A(x_1 - x_2) = 0\). Hence \(n - r \ge 1\), i.e. \(r < n\). (b) NO — the column space is not all of \(R^m\). If \(Ax = b\) has no solution for some right side \(b\), then \(b \notin C(A)\), so \(C(A) \neq R^m\) and \(r < m\). (c) From (a) and (b): \(r < n\) and \(r < m\). No other relations are forced — in particular, nothing can be said about the relative sizes of \(m\) and \(n\). (d) NO. Since \(N(A) \neq \{0\}\) by (a), whenever \(Ax = b\) is solvable the solutions form \(x_p + N(A)\), infinitely many of them. So a system with this \(A\) never has exactly one solution.

Keywords: rank, column space, nullspace, consistency, complete solution