Math 540 - Solutions to Selcted HW Problems



Problem # A.1: (a) Check that the vectors \(\mathbf{v}=(1,2,1)\) and \(\mathbf{w}=(8,4,-1)\) are solutions to the linear equation \[ 2x_1 - 3x_2 + 4x_3 = 0. \qquad\text{(1)} \]
(b) The vector \(\mathbf{u}=(-37,-14,8)\) is also a solution to equation (1). Find real numbers \(a,b\in\mathbf{R}\) so that \[ \mathbf{u} = a\mathbf{v}+b\mathbf{w}. \]
(c) Prove the following general result. If the vector \(\mathbf{z}=(z_1,z_2,z_3)\in\mathbf{R}^3\) is a solution to equation (1), then there are scalars \(a,b\in\mathbf{R}\) so that \[ \mathbf{z} = a\mathbf{v}+b\mathbf{w}. \qquad\text{(2)} \]
(d) In (c), prove that for a given vector \(\mathbf{z}\), there is only one choice for \(a\) and \(b\) that makes equation (2) true.
Solution:
(b) We need to find \(a\) and \(b\) so that \[\begin{aligned} \mathbf{u} &= a \mathbf{v} + b \mathbf{w} \\ (-37,-14,8) &= a(1,2,1) + b(8,4,-1) \\ (-37,-14,8) &= (a+8b,2a+4b,a-b). \\ \end{aligned} \] So we need to solve the simultaneous equations \[ -37=a+8b,\qquad -14=2a+4b,\qquad 8=a-b. \] There are lots of ways to do this. For example, subtracting the third equation from the first one, we get \[ \begin{aligned} -37-8 &= (a+8b)-(a-b) \\ -45 &= 9b \\ -5 &= b. \\ \end{aligned} \] Substituting \(b=-5\) into the last equation gives \(a=3\). Then one should check that \((a,b)=(3,-5)\) is a solution to the three simultaneous equations. It is, so the solution to (b) is \[ \mathbf{u} = 3 \mathbf{v} - 5 \mathbf{w}. \]
(c) We are given that \(\mathbf{z}=(z_1,z_2,z_3)\) satisfies equation (1), so we know that \[ 2z_1 - 3z_2 + 4z_3 = 0. \] We want to find \(a\) and \(b\) so that \(\mathbf{z} = a\mathbf{v}+b\mathbf{w}\). This is more-or-less the same problem as in (b), we need to solve \[\begin{aligned} \mathbf{z} &= a \mathbf{v} + b \mathbf{w} \\ (z_1,z_2,z_3) &= a(1,2,1) + b(8,4,-1) \\ (z_1,z_2,z_3) &= (a+8b,2a+4b,a-b). \\ \end{aligned} \] So we need to solve \[\begin{aligned} a+8b &= z_1 \\ 2a+4b &= z_2 \\ a-b &= z_3.\\ \end{aligned} \] Proceeding as in (b), we subtract the third equation from the first to get \[ 9b = z_1-z_3, \quad\text{so}\quad b = \frac{z_1-z_3}{9}. \] Then substituting in to the first equation gives \[ a = z_1 - 8b = z_1 - \frac{8z_1-8z_3}{9} = \frac{z_1+8z_3}{9}. \] We now need to check that these values of \(a\) and \(b\) actually work. (Note that so far, we haven't used the fact that \(\mathbf{z}\) satisfies equation (1).) So we compute \[\begin{align*} a \mathbf{v} + b \mathbf{w} &= \frac{z_1+8z_3}{9}(1,2,1) + \frac{z_1-z_3}{9}(8,4,-1) \\ &= \left( z_1, \frac{2z_1+4z_3}{3}, z_3 \right) \\ &= \left( z_1, \frac{3z_2}{3}, z_3 \right) \quad\text{because \(\mathbf{z}\) satisfies equation (1).} \\ &= (z_1,z_2,z_3) \; \checkmark \end{align*} \]
(d) The computation that we did in (c) essentially shows that the only possible choices for \(a\) and \(b\) are \(a = \frac{z_1+8z_3}{9}\) and \(b = \frac{z_1-z_3}{9}\).
     Alternatively, here is a direct proof, if we wanted to prove (d) without having first done (c). We assume that \[ \mathbf{z} = a\mathbf{v}+b\mathbf{w} = a'\mathbf{v}+b'\mathbf{w}, \] and we need to show that \(a=a'\) and \(b=b'\). Subtracting, we have \[ (a-a')\mathbf{v}+(b-b')\mathbf{w} = \mathbf{z}-\mathbf{z}= \mathbf{0}, \] so what we really need to do is show that \(\mathbf{v}\) and \(\mathbf{w}\) are linearly independent. So we suppose that \[ c_1\mathbf{v}+c_1\mathbf{w} = \mathbf{0}, \] and we will prove that \(c_1=c_2=0\). Thus \[\begin{aligned} c_1\mathbf{v}+c_1\mathbf{w} &= \mathbf{0} \quad\text{by assumption,} \\ c_1(1,2,1)+c_1(8,4,-1) &= (0,0,0) \quad\text{these are the values of \(\mathbf{v}\) and \(\mathbf{w}\),} \\ (c_1+8c_2,2c_1+4c_2,c_1-c_2) &= (0,0,0).\\ \end{aligned} \] So \[ c_1+8c_2=0,\qquad 2c_1+4c_2=0,\qquad c_1-c_2=0. \] Subtracting the third equation from the first gives \(9c_2=0\), so \(c_2=0\), and then substituting \(c_2=0\) into the first equation gives \(c_1=0\). This completes the proof that \(\mathbf{v}\) and \(\mathbf{w}\) are linearly independent, so there is at most one way to write any given vector \(\mathbf{z}\) as a linear combination of \(\mathbf{v}\) and \(\mathbf{w}\).


Problem # A.2: An \(m\)-by-\(n\) matrix with coefficients in a field \(\mathbb{F}\) is defined to be an \(m\)-by-\(n\) array of elements of \(\mathbb{F}\). We write \(M_{m,n}(\mathbb{F})\) for the set of all such matrices, so an element \(A\in M_{m,n}(\mathbb{F})\) looks like \[ A = \begin{pmatrix} a_{11} & a_{12} & \cdots & a_{1n} \\ a_{21} & a_{22} & \cdots & a_{2n} \\ \vdots & & \ddots & \vdots \\ a_{m1} & a_{m2} & \cdots & a_{mn} \\ \end{pmatrix} \] We make \(M_{m,n}(\mathbb{F})\) into a vector space in the obvious way: \[ \begin{pmatrix} a_{11} & \cdots & a_{1n} \\ \vdots & \ddots & \vdots \\ a_{m1} & \cdots & a_{mn} \\ \end{pmatrix} + \begin{pmatrix} b_{11} & \cdots & b_{1n} \\ \vdots & \ddots & \vdots \\ b_{m1} & \cdots & b_{mn} \\ \end{pmatrix} = \begin{pmatrix} a_{11}+b_{11} & \cdots & a_{1n}+b_{1n} \\ \vdots & \ddots & \vdots \\ a_{m1}+b_{m1} & \cdots & a_{mn}+b_{mn} \\ \end{pmatrix} \] and \[ c \begin{pmatrix} a_{11} & \cdots & a_{1n} \\ \vdots & \ddots & \vdots \\ a_{m1} & \cdots & a_{mn} \\ \end{pmatrix} = \begin{pmatrix} ca_{11} & \cdots & ca_{1n} \\ \vdots & \ddots & \vdots \\ ca_{m1} & \cdots & ca_{mn} \\ \end{pmatrix} \]
(a) Write down a basis for the \(\mathbb{F}\)-vector space \(M_{3,2}(\mathbb{F})\) of 3-by-2 matrices. What is the dimension of \(M_{3,2}(\mathbb{F})\)?
(b) More generally, what is the dimension of \(M_{m,n}(\mathbb{F})\)?
Solution: (a) The following six vectors form a basis for \(M_{3,2}(\mathbb{F})\), so \(M_{3,2}(\mathbb{F})\) has dimension 6. \[ \begin{pmatrix} 1 & 0 \\ 0 & 0 \\ 0 & 0 \\ \end{pmatrix},\quad \begin{pmatrix} 0 & 1 \\ 0 & 0 \\ 0 & 0 \\ \end{pmatrix},\quad \begin{pmatrix} 0 & 0 \\ 1 & 0 \\ 0 & 0 \\ \end{pmatrix},\quad \begin{pmatrix} 0 & 0 \\ 0 & 1 \\ 0 & 0 \\ \end{pmatrix},\quad \begin{pmatrix} 0 & 0 \\ 0 & 0 \\ 1 & 0 \\ \end{pmatrix},\quad \begin{pmatrix} 0 & 0 \\ 0 & 0 \\ 0 & 1 \\ \end{pmatrix}. \] To see this, note that \[ \begin{pmatrix} a & b \\ c & d \\ e & f \\ \end{pmatrix}= a\begin{pmatrix} 1 & 0 \\ 0 & 0 \\ 0 & 0 \\ \end{pmatrix}+ b\begin{pmatrix} 0 & 1 \\ 0 & 0 \\ 0 & 0 \\ \end{pmatrix}+ c\begin{pmatrix} 0 & 0 \\ 1 & 0 \\ 0 & 0 \\ \end{pmatrix}+ d\begin{pmatrix} 0 & 0 \\ 0 & 1 \\ 0 & 0 \\ \end{pmatrix}+ e\begin{pmatrix} 0 & 0 \\ 0 & 0 \\ 1 & 0 \\ \end{pmatrix}+ f\begin{pmatrix} 0 & 0 \\ 0 & 0 \\ 0 & 1 \\ \end{pmatrix}. \] This shows that every matrix in \(M_{3,2}(\mathbb{F})\) is a linear combination of the given 6 matrices, so the given 6 matrices span. Further, it shows that the only linear combination of the 6 matrices that equals the 0 matrix is by taking the 0 multiple of each of them, so the 6 matrices are linearly indepedent.
(b) Let \(E_{ij}\) be the matrix that has a \(1\) for its entry in the \(i\)'th row and \(j\)'th column, and every other entry is 0. Then \[ \{ E_{ij} : 1\le i\le m,\; 1\le j\le n \} \] is a basis for \(M_{m,n}(\mathbb{F})\), so \(M_{m,n}(\mathbb{F})\) has dimension \(mn\). To see that this is a basis, we note that \[ \sum_{i=1}^m \sum_{j=1}^n a_{ij}E_{ij} = \begin{pmatrix} a_{11} & a_{12} & \cdots & a_{1n} \\ a_{21} & a_{22} & \cdots & a_{2n} \\ \vdots & & \ddots & \vdots \\ a_{m1} & a_{m2} & \cdots & a_{mn} \\ \end{pmatrix}, \] so we get every matrix in \(M_{m,n}(\mathbb{F})\) as a linear combination of the \(E_{ij}\) matrices, so the \(E_{ij}\) matrices span, and the only linear combination of the \(E_{ij}\) matrices that equals the zero matrix is taking all of the \(a_{ij}\) to be 0, so the \(E_{ij}\) matrices are linearly independent.


Problem # A.3: The transpose of a (square) matrix \(A\), denoted \(A^*\), is obtained by flipping the entries across the main diagonal. So for example \[ \begin{pmatrix} 1&2&3\\ 4&5&6\\ 7&8&9\\ \end{pmatrix}^* = \begin{pmatrix} 1&4&7\\ 2&5&8\\ 3&6&9\\ \end{pmatrix}. \] A matrix \(A\) is symmetric if \(A^*=A\) and it is anti-symmetric if \(A^*=-A\).
(a) Prove that the set of \(n\)-by-\(n\) symmetric matrices is a vector subspace of \(M_{n,n}(\mathbb{F})\).
(b) Find a basis for the space of 2-by-2 symmetric matrices. What is its dimension?
(c) Generalize by describing a basis for the space of \(n\)-by-\(n\) symmetric matrices and computing its dimension. (It might help to start with 3-by-3.)
(d) Prove that the set of \(n\)-by-\(n\) anti-symmetric matrices is also a vector subspace of \(M_{n,n}(\mathbb{F})\), describe a basis, and compute its dimension.
(e) (Bonus) Let's write \(M_{n,n}(\mathbb{F})^{\text{sym}}\) for the space of symmetric matrices and \(M_{n,n}(\mathbb{F})^{\text{anti-sym}}\) for the space of anti-symmetric matrices. Prove that \[ M_{n,n}(\mathbb{F}) = M_{n,n}(\mathbb{F})^{\text{sym}} + M_{n,n}(\mathbb{F})^{\text{anti-sym}}. \] Is this a direct sum of vector spaces?
Solution: What happens if we add two matrices and then take their transpose? We compute \[ \begin{align*} (A+B)^* &= \left(\begin{pmatrix} a_{11} & \cdots & a_{1n} \\ \vdots & \ddots & \vdots \\ a_{m1} & \cdots & a_{mn} \\ \end{pmatrix} + \begin{pmatrix} b_{11} & \cdots & b_{1n} \\ \vdots & \ddots & \vdots \\ b_{m1} & \cdots & b_{mn} \\ \end{pmatrix}\right)^* \\ &= \begin{pmatrix} a_{11}+b_{11} & \cdots & a_{1n}+b_{1n} \\ \vdots & \ddots & \vdots \\ a_{m1}+b_{m1} & \cdots & a_{mn}+b_{mn} \\ \end{pmatrix}^* \\ &= \begin{pmatrix} a_{11}+b_{11} & \cdots & a_{m1}+b_{m1}\\ \vdots & \ddots & \vdots \\ a_{1n}+b_{1n} & \cdots & a_{mn}+b_{mn} \\ \end{pmatrix} \\ &= \left(\begin{pmatrix} a_{11} & \cdots & a_{m1} \\ \vdots & \ddots & \vdots \\ a_{1n} & \cdots & a_{mn} \\ \end{pmatrix} + \begin{pmatrix} b_{11} & \cdots & b_{m1} \\ \vdots & \ddots & \vdots \\ b_{1n} & \cdots & b_{mn} \\ \end{pmatrix}\right) \\ &= A^* + B^* \end{align*} \] and similarly \((cA)^* = c A^*\). So we have the useful formulas \[ (A+B)^*=A^*+B^*\quad\text{and}\quad(cA)^* = c A^*. \]
(a) Suppose that \(A\) and \(B\) are symmetric matrices, so \(A^*=A\) and \(B^*=B\). Then using the formulas that we just proved, we have \[ (A+B)^* = A^*+B^* = A+B \quad\text{and}\quad (cA)^* = cA^* = cA. \] This shows that the set of symmetric matrices \(M_{n,n}^{\text{sym}}(\mathbb{F})\) is closed under addition and scalar multiplication, so it is a vector subspace of \(M_{n,n}(\mathbb{F})\).
(b) Consider the three matrices \[ E_{11} = \begin{pmatrix} 1&0\\ 0&0\\ \end{pmatrix},\quad E_{22} = \begin{pmatrix} 0&0\\ 0&1\\ \end{pmatrix},\quad F_{12} = \begin{pmatrix} 0&1\\ 1&0\\ \end{pmatrix}. \] Then any 2-by-2 symmetric matrix can be written as \[ \begin{pmatrix} a&b\\ b&d\\ \end{pmatrix} = \begin{pmatrix} a&0\\ 0&0\\ \end{pmatrix} + \begin{pmatrix} 0&0\\ 0&d\\ \end{pmatrix} + \begin{pmatrix} 0&b\\ b&0\\ \end{pmatrix} = aE_{11} + dE_{22} + bF_{12} \] in exactly one way, so \(\{E_{11},E_{22},F_{12}\}\) is a basis for \(M_{n,n}^{\text{sym}}(\mathbb{F})\). In particular, \(\dim M_{n,n}^{\text{sym}}(\mathbb{F})=3\).
(c) More generally, as in Problem A.2, let \(E_{ij}\) be the matrix that has a \(1\) for its entry in the \(i\)'th row and \(j\)'th column, and every other entry is 0. Also let \[ F_{ij} = E_{ij} + E_{ji}, \] so \(F_{ij}\) has two 1's and the rest of its entries are 0. Since it's clear that \[ E_{ij}^* = E_{ji}, \quad\text{we have}\quad F_{ij}^* = (E_{ij} + E_{ji})^* = E_{ij}^* + E_{ji}^* = E_{ji} + E_{ij} = F_{ij}, \] so \(F_{ij}\) is a symmetric matrix. It's also clear that \(E_{ii}\) is a symmetric matrix. Then generalizing the proof in (b), one sees that \[ \{E_{ii} : 1\le i\le n\} \cup \{F_{ij} : 1\le i\lt j\le n\} \] is a basis for \(M_{n,n}^{\text{sym}}(\mathbb{F})\). (Note that \(F_{ij}=F_{ji}\), so we only need one of them.) More precisely, since a symmetric matrix has \(a_{ij}=a_{ji}\), we have \[ \begin{pmatrix} a_{11} & a_{12} & \cdots & a_{1n} \\ a_{21} & a_{22} & \cdots & a_{2n} \\ \vdots & & \ddots & \vdots \\ a_{n1} & a_{n2} & \cdots & a_{nn} \\ \end{pmatrix} = \sum_{i=1}^n a_{ii}E_{ii} + \sum_{i=1}^{n-1} \sum_{j=i+1}^n a_{ij}F_{ij}. \] Finally, \[ \dim M_{n,n}^{\text{sym}}(\mathbb{F}) = n + \frac{n(n-1)}{2} = \frac{n^2+n}{2}. \]
(d) This is similar, but instead of using the \(E_{ij}\) and the \(F_{ij}\), use the matrices \(G_{ij}=E_{ij}-E_{ji}\). Note that \[ G_{ij}^* = (E_{ij}-E_{ji})^* = E_{ij}^*-E_{ji}^* = E_{ji}-E_{ij} = -G_{ij}, \] so \(G_{ij}\in M_{n,n}^{\text{anti-sym}}(\mathbb{F})\). Further, it's not hard to show that \(\{G_{ij} : 1\le i\lt j\le n\}\) is a basis for \(M_{n,n}^{\text{anti-sym}}(\mathbb{F})\). Thus a matrix is anti-symmetric if and only if \(a_{ij}=-a_{ji}\), which means that the matrix can be written as \[ \begin{pmatrix} a_{11} & a_{12} & \cdots & a_{1n} \\ a_{21} & a_{22} & \cdots & a_{2n} \\ \vdots & & \ddots & \vdots \\ a_{n1} & a_{n2} & \cdots & a_{nn} \\ \end{pmatrix} = \sum_{i=1}^{n-1} \sum_{j=i+1}^n a_{ij}G_{ij}. \] (Notice in particular that diagonal entries satisfy \(a_{ii}=0\), since they equal their own negatives.) Hence \[ \dim M_{n,n}^{\text{anti-sym}}(\mathbb{F})= \frac{n^2-n}{2}. \]
(e) Given any matrix \(A\in M_{n,n}(\mathbb{F})\), write \(A\) as \[ A = \frac{1}{2}(A+A^*) + \frac{1}{2}(A-A^*). \] Then it's easy to check that \(\frac{1}{2}(A+A^*)\) is symmetric and \(\frac{1}{2}(A-A^*)\) is anti-symmetric, which shows the desired sum formula. In order to show that it is a direct sum, it's enough to show (by a result in the book) that the only matrix that is both symmetric and anti-symmetric is the zero matrix. But if \(A\) is both symmetric and anti-symmetric, then \[ A = A^* = -A,\quad\text{so}\quad 2A=0,\quad\text{so}\quad A=0. \]


ProblemPage 59 # 2: Give an example of a function \(f:\mathbf{R}^2\to\mathbf{R}\) such that \[ f(a v) = af(v) \] for all \(a\in\mathbf{R}\) and all \(v\in\mathbf{R}^2\), but \(f\) is not linear.
Solution: An interesting observation is that if \(f\) has a Taylor series expansion around \((0,0)\), then \(f\) has to be linear. (You might try proving this.) Anyway, it shows that we need to look for a function whose derivative fails to exist.
      Someone suggested the function \(f(x,y)=\sqrt{xy}\). This almost works, but it has a couple of problems. First, it's value isn't in \(\mathbf{R}\) if \(xy<0\). Second, it satisfies \[ f(ax,ay) = \sqrt{a^2xy} = |a|f(x,y). \] However, this example suggests an approach. The problem with taking square roots is that \(\sqrt{x}\) is only defined if \(x\ge0\) and the values all satisfy \(\sqrt{x}\ge0\). But if we take cube roots, then \(\sqrt[3]{x}\) is defined for all real values of \(x\), and the cube root of \(x^3\) is always equal to \(x\). So here are some functions that have the desired property: \[ f_1(x,y) = \sqrt[3]{x^3+y^3},\qquad f_2(x,y) = \sqrt[3]{xy^2+x^2y},\qquad f_3(x,y) = \sqrt[7]{x^7 + xy^6}. \] And you can make up many others of a similar nature. Let's check that \(f_1\) is not linear. We have \[ f_1\bigl((1,0)+(0,1)\bigr) = f(1,1) = \sqrt[3]{2} \quad\text{and}\quad f_1(1,0)+f_1(0,1) = \sqrt[3]{1}+\sqrt[3]{1}=2. \]
      One can also create weirder sorts of non-linear functions that have the desired property. Here's an example: \[ f(x,y) = \begin{cases} x &\text{if \(y=0\) or if \(x/y\) is a rational number,} \\ 0 &\text{if \(y\ne0\) and \(x/y\) is an irrational number,} \\ \end{cases} \] I'll leave you to check that \(f(ax,ay)=f(x,y)\). To see that \(f\) is not linear, we use the fact that \(\pi=3.14159\ldots\) is not rational. Then \[ f(\pi,1) + f(1-\pi,1) = 0 + 0 = 0 \quad\text{and}\quad f\bigl((\pi,1)+(1-\pi,1)\bigr) = f(1,1) = 1. \] Pretty weird!


Problem Page 59 # 4: Suppose that \(T\) is a linear map from \(V\) to \(\mathbf{F}\). Prove that if \(u\in V\) is not in null\((T)\), then \[ V = \text{null}(T) \oplus \{au : a\in\mathbf{F}\}. \]
Solution: The fact that \(T\) is a map from \(V\) to \(\mathbf{F}\) means that the values of \(T\) are in \(\mathbf{F}\), i.e., they are scalars. So \(T(u)\in\mathbf{F}\) is a scalar, and since we are given that \(u\) is not in null\((T)\), that scalar is not \(0\). To emphasize that it is a scalar, we'll call it \(b\). In other words, \[ b = T(u) \in \mathbf{F}\quad\text{and}\quad b \ne 0. \] Now let \(v\in V\) be any vector. We want to show that \(v\) is the sum of a vector in null\((T)\) and a vector in \(\{au:a\in\mathbf{F}\}\). The idea is to try to subtract a multiple of \(u\) from \(v\) to get something that's in null\((T)\). So we look for a scalar \(t\) so that \[ T(v-tu) = 0. \] By linearity, this is \[ T(v) - tT(u) = 0. \] But remember that \(T(u)=b\) is a non-zero scalar, i.e., \(b\) is a non-zero number, so it has a multiplicative inverse \(b^{-1}\). So we should set \(t=b^{-1}T(v)\), where remember that \(T(v)\) is also a scalar. To emphasize that \(T(v)\) is a scalar, let's call it \(c\), so \(c=T(v)\). With this notation, we take \(t=b^{-1}c\).
      This means that if we write \(v\) as \[ v = (v - b^{-1}cu) + b^{-1}cu, \] then the vector in parentheses is in null\((T)\), since \[ T(v-b^{-1}cu) = T(v) - b^{-1}cT(u) = c - b^{-1}cb = 0, \] while the vector \(b^{-1}cu\) is clearly in \(\{au:a\in\mathbf{F}\}\), since it is a scalar multiple of \(u\). This proves that every vector in \(V\) is a sum of a vector in null\((T)\) and a vector in \(\{au:a\in\mathbf{F}\}\), so this proves that \[ V = \text{null}(T) + \{au:a\in\mathbf{F}\}. \]
Finally, to show that it is a direct sum, we need to show that \[ \text{null}(T) \cap \{au:a\in\mathbf{F}\} = \{0\}. \] So let \(v\) be a vector in the intersection. Then \(v=au\) for some scalar \(a\). But also \(v\) is in null\((T)\), so \[ 0 = T(v) = T(au) = aT(u). \] We are given that \(T(u)\ne0\), so we conclude that \(a=0\), and hence that \(v=au=0\).


Return to the Math 540 Homework Page

Return to the Math 540 Home Page