3.2-final-2013¶
Final Exam, 2013 · Section 3.2: Computing the Nullspace by Elimination: A = CR Source: finalexam_2.pdf, pp. 97-98
Problem¶
6 (14 pts.) Choose the best choice from "must", "might", "can't" (and explain briefly.)
(a) (2 pts.) A 3x3 matrix \(M\) with rank \(r = 2\) ________ have a non-zero solution to \(Mx = 0\).
(b) (2 pts.) A 3x3 matrix \(M\) with rank \(r = 3\) ________ have a non-zero solution to \(Mx = 0\) with \(x_1 = 0\).
(c) (2 pts.) A 3x4 matrix \(M\) with rank \(r = 3\) ________ have a non-zero solution to \(Mx = 0\) with \(x_1 = 1\).
(d) (2 pts.) A permutation matrix ________ be singular.
(e) (2 pts.) A projection matrix ______ be singular.
(f) (2 pts.) \(M - \lambda I\) ______ be singular, if \(\lambda\) is an eigenvalue of \(M\).
(g) (2 pts.) We recall the number of columns of an incidence matrix for a graph is the number of nodes and every row has one entry +1, one entry -1, and the remaining entries 0. Such an incidence matrix ______ have full column rank.
Solution
(a) MUST. An \(n \times n\) matrix with \(\mathrm{rank}(M) = 2\) has a 1-dimensional nullspace, so there is a nonzero \(x\) with \(Mx = 0\). (b) CAN'T. \(\mathrm{rank}(M) = 3\) means \(M\) is invertible, so \(Mx = 0\) has only the zero solution (in particular no nonzero solution with \(x_1 = 0\)). (c) MIGHT. A \(3 \times 4\) matrix of rank 3 has a 1-dimensional nullspace; that nullspace may or may not contain a vector whose first component is 1 (e.g. the nullspace vector could be \((0, 1, 0, 0)\) or \((1, 0, 0, 0)\)). (d) CAN'T. A permutation matrix has exactly one nonzero term in the big determinant formula, equal to \(\pm 1\) (the sign of the permutation), so \(\det \neq 0\). (e) MIGHT. The zero matrix \(\begin{bmatrix} 0 & 0 \\ 0 & 0 \end{bmatrix}\) is a singular projection, while the identity matrix is an invertible projection. (f) MUST. This is the definition of an eigenvalue: \(\lambda\) is an eigenvalue of \(M\) exactly when \(M - \lambda I\) is singular, i.e. when \((M - \lambda I)x = 0\) has a nonzero solution. (g) CAN'T. Every row of an incidence matrix contains one +1 and one -1, so the sum of all columns is the zero vector and the columns are linearly dependent. Hence an incidence matrix of a graph with \(n \ge 2\) nodes can never have full column rank (the all-ones vector is always in its nullspace).Keywords: true/false, rank, nullspace, incidence matrix