2

Background

Proposition 4.33 In the ring $K[X]$ of polynomials over a field $K$ every ideal is principal. If $J$ is a non-zero ideal of $K[X]$ and $g$ in $J$ is a non-zero polynomial of minimal degree among the non-zero polynomials of $J$, then $g$ generates $J$. Any two generators of $J$ are constant multiples of each other.

The substitutino map $\phi_1:\mathbb{R}[X]\to \mathbb{C}$ given by $f(X)\mapsto f(i)$ is surjectibe. Indeed, if $a,b\in \mathbb{R}$, then $\phi_1(a+bX)=a+bi$, which covers all complex numbers. According to Proposition 4.33, $\text{ker }\phi_1$ is a pirincipal ideal generated by a polynomial in $\text{ker }\phi_1$ of minimal degree. Well, $X^2 +1$ is such a polynomial, since $\phi_1(X^2 +1)=i^2 +1=0$ and since no polynomial of lower degree linear or constant, lies in $\text{ker }\phi_1$. By the first isomorphism theorem,

$$\mathbb{C}\cong\mathbb{R}[x]/\langle X^2+1 \rangle$$.

Question

I would like to know suppose I am simply given the map $\phi_1:\mathbb{R}[X]\to \mathbb{C}$ defined by $\phi_1(a+bX)=a+bi$, then how would i determine the kernel of $\phi_1$ as $\langle X^2+1 \rangle$ without resorting to Proposition 4.33 above. I mean how does it go from $a+bX$ to $X^2 +1$ in the domain of $\phi_1$. In the end, we want $\phi_1(a+bX)=a+bi=0$, and the only information we have is $x\mapsto i$. But again the domain of $\phi_1$ seem to be asking for a polynomial of specific degree, but it starts with a linear polynomial.

Thank you in advance.

Seth
  • 4,043
  • 1
    The linear polynomial in the first part of the proof is only used to show that the evaluation map is onto. That polynomial is not related to the rest of the proof (kernel computation). Why do you wish to avoid 4.33? This fundamental result - that ideals in a (Euclidean $\Rightarrow$) PID are generated by minimal elements - has a very simple proof. It is used in many places in algebra so you should master it it - not try to avoid it. – Bill Dubuque May 30 '24 at 21:24
  • @BillDubuque I was hoping to get an answer in this similar to what jgon stated in this post, via some sort of algebraic derivation. – Seth May 30 '24 at 21:42
  • Ok, I will explain how to do it that way. – Bill Dubuque May 30 '24 at 23:12
  • @BillDubuque the reason I ask to be that restrictive is suppose that a student is only introduced to the isomorphism theorems and basics of quotient rings. Like what you see in Hungerford's undergrad algebra text, and they are not even at maximal or prime ideals. They are only given the map $\phi_1(a+bX)=a+bi$ from my post.. How would they be able to compute the kernel without resorting to proposition 4.33, to show that the image of $\phi_1$ is isomorphic to $f(i)/\mathbb{C}$. The thing is if we ask students to describe the quotient ring $\mathbb{R}[X]/<X^2+1>$,.. – Seth May 30 '24 at 23:19
  • @BillDubuque ... I see in such example that it is assume that $<X^2+1>$ is already a kernel of some homomorphic map. – Seth May 30 '24 at 23:22

3 Answers3

1

Let $z = a + bi$. The map $ \phi_1$ is simply evaluating a polynomial at $x = z$ it therefore follows that the kernel of $\phi_1$ is exactly the set of polynomials $f$ such that $f(z) = 0$. We deduce that $\overline z$ is also a root of $f \in \ker \phi_1$ by using

If $z \in \mathbb C$ is a root of a real polynomial then so is $\overline z \in \mathbb C$.

Therefore each polynomial $f$ in the kernel is divisible by $(X-z)$ and $(X- \overline z)$ when viewed as a polynomial in $\mathbb C [X]$which means that $f$ is divisible by

$$(X - \overline z)(X- z) = X^2 -( z + \overline z)X + \overline zz = X^2-2a X + \vert z \vert^2$$ in $\mathbb C[X].$

Since $f$ is divisible by $X^2-2a X + \vert z \vert^2$ in $\mathbb C[X]$ we can use the long division algorithm to find $q \in \mathbb C[X]$ such that

$$ f(X) = q(X) (X^2 - 2aX + \vert z \vert^2).$$

since $f$ and $X^2 - 2aX + \vert z \vert^2$ all have real coefficients the long division algorithm will give a $q$ with real coefficients. This proves that $f$ is divisible by $X^2 - 2aX + \vert z \vert^2$ in $\mathbb{R}[X]$.

Therefore each polynomial in the kernel is divisible by $X^2 - 2aX + \vert z \vert^2$

It follows that

$$ \ker \phi_1 = \langle X^2 - 2a X + \vert z \vert^2 \rangle $$

Digitallis
  • 4,246
1

I was hoping to get an answer in this similar to what jgon stated in this post, via some sort of algebraic derivation

Below I explain how to do it that way. First I clarify that way - highlighting the use of the norm (multiplying by the conjugate) to $\rm\color{#90F}{rationalize}$ the denominator, which simplifies division by complex integers to division by integers. OP works the same way, except here we $\rm\color{#90F}{realize}$ the denominator, which reduces division by complex polynomials to real polynomials. More generally see the method of $\rm\color{#90F}{simpler\ multiples}$.


$\bbox[4px,border:1px solid #c00]{A=\Bbb Z[i]/(3\!+\!i)\equiv \Bbb Z/10}\ $ $\ \ \Bbb Z \overset{h}\to A\,\color{#0a0}{ \ {\rm is\ onto}}\,$ by $\!\bmod\, 3\!+\!i\!:\ \, i\equiv -3\,\Rightarrow\, a\!+\!bi\equiv a\!-\!3b\color{#0a0}{\in\Bbb Z}$

$\color{#c00}{I := \ker h = 10\,\Bbb Z}\ $ follows immediately by $\,\rm\color{#90F}{rationalizing}\,$ a denominator

$$ n\in I\!\iff\! 3\!+\!i\mid n\ \, {\rm in}\, \ \Bbb Z[i]\!\iff\! \dfrac{n}{3\!+\!i} \color{#90f}{\overset{\large \rm\ rat}=} \dfrac{n(3\!-\!i)}{10}=\frac{3n}{10}\!-\!\color{#c00}{\frac{n}{10}}\:\!i\in\Bbb Z[i]\!\iff\! \color{#c00}{10\mid n}\qquad\ $$

Thus $\, \color{#0a0}{A = {\rm Im}\ h} \,\cong\, \Bbb Z/\color{#c00}{\ker h} \,=\, \Bbb Z/\color{#c00}{10\,\Bbb Z}\ $ by the First Isomorphism Theorem.


$\bbox[4px,border:1px solid #c00]{\Bbb R[x]/(x^2\!+\!1)\equiv \Bbb C}\ $ $\ \ \Bbb R[x]\overset{h}\to \Bbb C,\ f(x)\mapsto f(i)\,\color{#0a0}{ \ {\rm is\ onto}}\,$ by $\,h(a\!+\!bx) = a\!+\!bi$.

$\color{#c00}{I := \ker h = (x^2\!+\!1)\,\Bbb R[x]}\ $ follows immediately by $\,\rm\color{#90F}{realizing}\,$ a denominator

$\ f\!\in\! I\!\iff\! x\!-\!i\mid f\ \, {\rm in}\, \ \Bbb C[x]\!\iff\! \dfrac{f}{x\!-\!i}\color{#90f}{\overset{\!\large \rm\ rea}=} \dfrac{(x+i)f\!}{x^2+1}=\underbrace{\dfrac{xf}{x^2+1}+\color{#c00}{\dfrac{f}{x^2+1}}\:\!i\:\!\in\:\!\Bbb C[x]}_{\textstyle\ \ \ \ \iff \color{#c00}{{x^2\!+\!1}\mid f\ \,{\rm in}\,\ \Bbb R[x]} }$

Thus $\, \color{#0a0}{\Bbb C = {\rm Im}\ h} \,\cong\, \Bbb R[x]/\color{#c00}{\ker h} \,=\, \Bbb R[x]/\color{#c00}{(x^2\!+\!1)}\ $ by the First Isomorphism Theorem.

Bill Dubuque
  • 282,220
0

Clearly you have $\langle x^2+1\rangle \subseteq \ker\phi_1$, since $x^2+1\in\ker\phi_1$ and $\phi_1$ is a map of rings.
Conversly, assume that $f\in\ker\phi_1$

Then we have: $\phi_1(f)=f(i)=0$.
So $i$ is a root of the polynomial $f$.
Since $f$ is a polynomial with real coefficients, it must be the case that $\bar{i}=-i$ is also a root of $f$. (Do you see why?)

As $i,-i$ are roots of $f$, there exists a polynomial $g\in\mathbb{R}[x]$ such that: $$f(x)=(x-i)(x+i)g(x)=(x^2+1)g(x)$$ But this is equivalent to $f\in\langle x^2+1\rangle$.

Since the $f\in\ker\phi_1$ we started with was arbitrary, we conclude that $\ker\phi_1\subseteq\langle x^2+1\rangle$, proving the inverse inclusion, giving us the wanted equality $\langle x^2+1\rangle=\ker\phi_1$

Fotis
  • 953
  • 1
    "Since $$ is a polynomial with real coefficients, it must be the case that $\bar{}=−i$... That has to do with the complex conjugate root theorem. – Seth May 30 '24 at 21:34
  • Likely the reason for the downvotes is that you are mixing up polynomial divisibility and ideals in $\Bbb R[x]$ vs. $\Bbb C[x].\ \ $ – Bill Dubuque May 30 '24 at 22:00
  • @BillDubuque indeed I could have written it a bit better, i.e. first seeing $f$ as a polynomial in $\mathbb{C}[x]$ factoring it there so finding a $g\in\mathbb{C}[x]$ and arguing that such $g$ should be in $\mathbb{R}[x]$. And finally conclude that the $f$ should be contained in the ideal. But downvote for that, instead of just leaving a comment when the rest is fine is a bit weird. Oh well, hopefully it answered OP's question – Fotis May 30 '24 at 22:37
  • @Fotis can i ask why $\langle X^2 +1 \rangle$ is automatticvally a subset of $\text{ker }\phi_1$? – Seth May 30 '24 at 23:10
  • @BillDubuque I respond to your earlier comment that you deleted about why I want to savoid proposition 4.33. I was hoping form something along the line of the post I referenced in that comment. Something calculuational, along the line of high school polynomial algebra. – Seth May 30 '24 at 23:11
  • @Seth I didn't delete that comment - it's still there (on the question). I added an answer which is understandable at high-school level (uses nothing deeper than rationalizing denominators to compute the kernel). – Bill Dubuque May 30 '24 at 23:58
  • @Seth as you showed, $\phi_1(x^2+1)=0$. Now an element of $\langle x^2+1\rangle$ is of the form $(x^2+1)g(x)$ where $g\in\mathbb{R}[x]$. Now, as $\phi_1$ is a ring map, we have: $$\phi_1((x^2+1)g(x))=\phi_1(x^2+1)\phi_1(g(x))=0\phi_1(g(x))=0$$ so $(x^2+1)g(x)\in\ker\phi_1$ – Fotis May 31 '24 at 00:38
  • @Fotis wait, can i do the following: $\phi_1(a+bX^2)=a+bi^2=a-b=0,$ I then let $b=a=1$ or simply $a=b$, since $X\mapsto i$, so $X^2\mapsto i^2=-1$. Is that what you mean? – Seth May 31 '24 at 00:47
  • If you are referring to the part $\phi_1(X^2+1)=0$ then yes, you take $a=b=1$ and correctly note that $X^2\mapsto i^2=-1$ – Fotis May 31 '24 at 00:49