Math 540 - Solutions to Selcted HW Problems - Due Feb 13



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: \[ f_1(x,y) = \sqrt[3]{x^3+y^3},\qquad f_2(x,y) = \sqrt[3]{xy^2},\qquad f_3(x,y) = \sqrt[7]{x^7 + xy^6}. \] 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