If you find any mistakes, please make a comment! Thank you.

## Chapter 4 Exercise

1. Empty

2. Solution: False. Consider $1=(z^m+1)+(-z^m)\notin \{0\}\cup\{p\in\ca P(\mb F):\deg p=m\}$. Note that $(z^m+1)\in \{0\}\cup\{p\in\ca P(\mb F):\deg p=m\}$and$-z^m\in \{0\}\cup\{p\in\ca P(\mb F):\deg p=m\},$it follows that $\{0\}\cup\{p\in\ca P(\mb F):\deg p=m\}$ is not closed under addition. Hence it is not a subspace of $\ca P(\mb F)$.

3. Solution: False. Consider $z=(z^{2}+z)+(-z^2)\notin \{0\}\cup\{p\in\ca P(\mb F):\deg p\text{~is even}\}$. Note that $(z^2+z)\in \{0\}\cup\{p\in\ca P(\mb F):\deg p\text{~is even}\}$and$-z^2\in \{0\}\cup\{p\in\ca P(\mb F):\deg p\text{~is even}\},$it follows that $\{0\}\cup\{p\in\ca P(\mb F):\deg p\text{~is even}\}$ is not closed under addition. Hence it is not a subspace of $\ca P(\mb F)$.

4. Solution: Define $p\in\ca P(\mb F)$ by $p(z)=(z-\lambda_1)^{n-m+1}(z-\lambda_2)\cdots(z-\lambda_m).$Then $p$ is a polynomial of degree $n$ such that $0=p(\lambda_1)=\cdots=p(\lambda_{m})$ and such that $p$ has no other zeros.

5. Solution: See Linear Algebra Done Right Solution Manual Chapter 4 Problem 2.

6. Solution: See Linear Algebra Done Right Solution Manual Chapter 4 Problem 4.

7. Solution: See Linear Algebra Done Right Solution Manual Chapter 4 Problem 5.

8. Solution: First we show that $T$ is a linear map. Then we show $Tp \in\ca P(\R)$ for a basis of $p \in\ca P(\R)$, then by linearity of $T$, we have $Tp \in\ca P(\R)$ for every polynomial $p \in\ca P(\R)$. For any $\lambda\in\R$ and $p,q\in\ca P(\R)$, we have \begin{align*} T(p+q)=&\dfrac{(p+q)-(p+q)(3)}{x-3}=\dfrac{(p+q)-p(3)-q(3)}{x-3}\\ =&\dfrac{p-p(3)}{x-3}+\dfrac{q-q(3)}{x-3}=Tp+Tq, \end{align*} if $x\ne 3$. Similarly, $T(\lambda p)=\dfrac{(\lambda p)-(\lambda p)(3)}{x-3}=\dfrac{\lambda p-\lambda p(3)}{x-3}=\lambda\dfrac{p-p(3)}{x-3}=\lambda Tp.$If $x=3$, then $T$ is a composition of the differentiation map and evaluation map. Both of them are linear, hence $T$ is also linear. We can show it directly $T(\lambda p+q)=(\lambda p+q)'(3)=(\lambda p’+q’)(3)=\lambda p'(3)+q'(3)=\lambda Tp+Tq.$Therefore $T$ is a linear map. Let us consider $Tx^n$ for $n\in \mb N^+$, if $x\ne 3$, $T(x^n)=\frac{x^n-3^n}{x-3}=x^{n-1}+3x^{n-2}+\cdots+3^kx^{n-1-k}+\cdots+3^{n-1}\in\ca P(\R).$ Moreover, if $x=3$, we have $T(x^n)=3^{n-1}n$. Note that when $x=3$, it is true that$x^{n-1}+3x^{n-2}+\cdots+3^kx^{n-1-k}+\cdots+3^{n-1}=3^{n-1}n.$We get $T(x^n)=x^{n-1}+3x^{n-2}+\cdots+3^kx^{n-1-k}+\cdots+3^{n-1}\in\ca P(\R)$for $x\in \R$. Similarly, we can show $T(1)=0\in \ca P(\R)$.

Since any polynomial of $\ca P(\R)$ is a linear combination of $1$, $x$, $x^2$, $\cdots$, it follows that $Tp \in\ca P(\R)$ for every polynomial $p \in\ca P(\R)$.

I am not sure the textbook indicates that $1$, $x$, $x^2$, $\cdots$ is a basis of $\ca P(\R)$, so I use some easier arguments such as any polynomial of $\ca P(\R)$ is a linear combination of $1$, $x$, $x^2$, $\cdots$.

9. Solution: If $f(z)=a_nz^n+\cdots+a_1z+a_0$, where $a_n,\cdots,a_0\in\C$, then $\overline{f(\bar{z})}=\overline{a_n}z^n+\cdots+\overline{a_1}z+\overline{a_0}.$That implies $\overline{f(\bar{z})}$ is a polynomial. As the product of polynomials is a polynomial as well, we conclude $q$ is a polynomial.

Now let us show $q$ has only real coefficients. Denote $q(z)$ by $q(z)=\mu_{2n}z^{2n}+\cdots+\mu_1z+\mu_0.$Note that $\overline{q(\bar{z})}=\overline{f(\bar{z})\overline{f(z)}}=\overline{f(\bar{z})}f(z)=q(z)$, it follows $\overline{\mu_{2n}}z^{2n}+\cdots+\overline{\mu_1}z+\overline{\mu_0}=\mu_{2n}z^{2n}+\cdots+\mu_1z+\mu_0.$Hence $\overline{\mu_k}=\mu_k$, i.e $\mu_k\in \R$, for $k=0,\cdots,2n$.

Here you can also compute the coefficients $\mu_k$ in terms of $a_i$ and show $\mu_k=\overline{\mu_k}$ or use some calculus methods.

10. Note that $x_0,x_1,\cdots,x_m$ are distinct, we can define the polynomial$f(x)=\sum_{j=0}^m\frac{(x-x_0)(x-x_1)\cdots(x-x_{j-1})(x-x_{j+1})\cdots(x-x_m)}{(x_j-x_0)(x_j-x_1)\cdots(x_j-x_{j-1})(x_j-x_{j+1})\cdots(x_j-x_m)}p(x_j).$Then it is obvious that $f(x)\in\ca P_m(\C)$. Moreover, since $x_j$ and $p(x_j)$, $j=0,1,\cdots,m$, are real, it follows that the coefficients of $f(x)$ are real. Hence it suffices to show that $p(x)=f(x)$.

By plugging $x=x_i$ into $f(x)$, we have $f(x_i)=p(x_i)$ since all summands except one are zero (see the link in the remark below),
$\frac{(x_i-x_0)(x-x_1)\cdots(x_i-x_{j-1})(x_i-x_{j+1})\cdots(x_i-x_m)}{(x_j-x_0)(x_j-x_1)\cdots(x_j-x_{j-1})(x_j-x_{j+1})\cdots(x_j-x_m)}=\delta_{ij},$where $\delta_{ij}=0$ if $i\ne j$ and $\delta_{ij}=1$ if $i=j$.

This implies that $f-p$ has $m+1$ distinct zeros. Since $f-p\in\ca P_m(\C)$, it follows from 4.12 that the degree of $f-p$ can not be nonnegative. Hence $f-p$ is the zero polynomial, thus completing the proof.

The polynomial used here is called the Lagrange Interpolating Polynomial . Please see the following link for more detail.

11. By the division algorithm of polynomials in 4.8, we know that for every polynomial $f\in\ca P(\mb F)$ there exist unique polynomials $q$ and $r$ such that $f=pq+r,\quad\text{and}\quad \deg r<\deg p.$This implies that $\ca P(\mb F)=U\oplus \ca P_{\deg p-1}(\mb F)$.

Therefore$\ca P(\mb F)/U\cong\ca P_{\deg p-1}(\mb F).$It follows that $$\dim\ca P(\mb F)/U=\dim\ca P_{\deg p-1}(\mb F)=\deg p.$$Moreover, a basis of $\ca P(\mb F)/U$ is $1,x,x^2,\cdots,x^{\deg p -1}$.

Here I used the fact that if $V=U\oplus W$, then $V/U\cong W$. Please try to prove as the following alternative solution.

Another solution (explains the solution above more explicitly).

For any given polynomial $f\in\ca P(\mb F)$, let $r(f)$ be the reminder of $f$ divided by $q$. Note that $\deg r(f)<\deg q$we have a map $r:\ca P(\mb F)\to \ca P_{\deg p-1}(\mb F)$. One can check this is a linear map. Moreover, $\mathrm{Null}\, r=U$. By taking polynomials in $\ca P_{\deg p-1}(\mb F)$, we have that $\mathrm{range}\,r=\ca P_{\deg p-1}(\mb F)$.

By 3.91(d), we have that $\ca P(\mb F)/U=\ca P(\mb F)/\mathrm{Null}\, r\cong \mathrm{range}\,r=\ca P_{\deg p-1}(\mb F).$Now the problem is solved similarly.

### This Post Has 6 Comments

1. Number 8 is an interesting one. It shows that in order to calculate the derivative of a polynomial p(x) at point x_0, it's enough to
1. Take the quotient q(x) of division of p(x) by x-x_0: p(x)=q(x)(x-x_0) + r
2. Evaluate q(x) at the point x_0

This was unexpected for me. Idk if this fact has any useful applications though.

2. Problem 11 has a problem.
The element of a basis of P(F) is not polynomial, but space.

1. I have no idea what you are talking about. The space $\mathcal P(\mathbf F)$ is the set of all polynomials. You mean elements of $\mathcal P(\mathbf F)$ are spaces?

3. The another altenative solution for 10 is by deriving the result of 5.
There is a polynomial p of real coefficients with m+1 distinct values in Pm(R) and p must also be in Pm(C). And since p must be uniquely defined in Pm(C), we have arrived at what is needed to be proven.

4. Did you choose not to provide a solution to #11? I found one, though it involved proving P(F)/U isomorphic to P_(m-1) (F) by defining a linear transformation T(p)=r where r is the remainder defined by the division algorithm. I was wondering if there was a simpler method.

1. That is the essential part of this problem.