The book I'm reading introduces polynomials over a field and proves the statement that a polynomial of degree $n$ has at most $n$ zeros. They do this by using division algorithm and induction.

Then they make the following remark:

This is not true for polynomial over arbitrary rings. For instance $x^2 + 7 \in \mathbb{Z}_8$ has roots $1,3,5,$ and $7$.

My question: what fails in the previous proof for arbitrary rings? They just made that remark and moved on.

Edit -- Proof (from Gallian):

We proceed by induction on $n$. Clearly, a polynomial of degree $0$ over a field has no zeros. Now suppose that $f(x)$ is a polynomial of degree $n$ over a field and $a$ is a zero of $f(x)$ of multi­plicity $k$. Then, $f(x)=(x-a)^kq(x)$ and $q(a) \neq 0$. Note we have $\text{deg }f = n = k + \text{deg }q$. If $f(x)$ has no zeros other than $a$, we are done. On the other hand, if $b \neq a$ and $b$ is a zero of $f(x)$, then $0=f(b)=(b-a)^kq(b)$ so that $b$ is a zero for $q(x)$ with the same multiplicity it has for $f(x)$. By the Second Principle of Mathematical Induction, we know that $q(x)$ has at most deg $q(x)=n-k$ zeros, counting multiplicity. Thus, $f(x)$ has at most $k + n -k = n$ zeros, counting multiplicity.

  • $\begingroup$ Can you sketch the proof you are referring to? $\endgroup$
    – Florian R
    Jun 9 '18 at 14:55
  • 1
    $\begingroup$ Why don't you go through the proof line by line and see what hypotheses are used and whether they hold in $\mathbb{Z}/8\mathbb{Z}$? $\endgroup$
    – Alex B.
    Jun 9 '18 at 14:55
  • $\begingroup$ well I have a hunch, I'm not sure though. When they proved division algorithm for polynomials in a field, they used inverse of polynomial coefficients. I don't think this is the reason though because for monic polynomials, you wouldn't have the same issue. (I realize I haven't put the proofs in, I'm doing that now) $\endgroup$
    – yoshi
    Jun 9 '18 at 15:01

The issue is that the polynomial division is not unique in $\mathbb Z_8$. This is the consequence of existence of zero divisors.

You’ll be able to verify that $$x^2+7=(x-1)(x-7)=(x-3)(x-5)$$

providing evidence that $1, 3, 5,7$ are roots.

  • $\begingroup$ Nailed it. This is exactly what this is about: zero divisors. Not multiplicative inverses. A polynomial $f$ over an integral domain always has at most $\deg(f)$ roots. $\endgroup$
    – freakish
    Jun 9 '18 at 16:10
  • 3
    $\begingroup$ In the case of noncommutative rings, it’s not just about zero divisors. $x^2+1=0$ is satisfied by infinitely many elements of $\mathbb H$, for example. (Probably the OP had commutative tings in mind, but this is worth knowing.) $\endgroup$
    – rschwieb
    Jun 9 '18 at 16:23

Usually, you can go from a factorization like $(x-1)(x-2)=0$ to the pair of equations $x-1=0$ and $x-2=0$ to get your solutions. This uses the zero product property, which says that if $AB=0$ then either $A$ or $B$ are equal to zero. This fails over arbitrary rings. For example, in $\mathbb{Z}/8\mathbb{Z}$, the zero product property fails because $2\cdot 4=4\cdot 4=4\cdot 6=0$. So if we want to solve $(x-1)(x-7)=0$, then there are six things that can happen. Either: $$ (1)\;\;\;\;x-1 = 0\;\;\textrm{or}\;\; x-7=0\\ (2)\;\;\;\;x-1 = 2\;\;\textrm{and}\;\; x-7=4\\ (3)\;\;\;\;x-1 = 4\;\;\textrm{and}\;\; x-7=2\\ (4)\;\;\;\;x-1 = 4\;\;\textrm{and}\;\; x-7=4\\ (5)\;\;\;\;x-1 = 4\;\;\textrm{and}\;\; x-7=6\\ (6)\;\;\;\;x-1 = 6\;\;\textrm{and}\;\; x-7=4 $$ The only ones that can happen are (1), (2) and (5) using $x=1,7$ and $x=3$ and $x=5$ respectively. Therefore, the solutions are $1,7,3,5$. Note that the extra zeros correspond both to zero divisors of the base-ring, as well as two different factorizations in the polynomial ring.

In general, if $R$ has zero divisors, then $R[x]$ will not have unique factorization since there is a correspondence between zeros and factors of a polynomial. The zero divisors can ensure that there are "too many" factors.

For instance, let $a,b$ be zero divisors and let $c,d$ be any other values so that $a-b=c-d$. Set $s=a+d=b+c$. Consider the polynomial $p(x)=(x-c)(x-d)$. Clearly $x=c,d$ are roots. But, furthermore, $p(s)=(b+c-c)(a+d-d)=ab=0$. So $x=s$ is also a root, and you can check that $x=t$, where $t=c+d-s$, is also a root. We then get $p(x)=(x-c)(x-d)=(x-s)(x-t)$.


The problem is the existence of zeros divisors in the ring.

Polynomial division does not always work over rings that are not fields. For instance, $x^2-5$ cannot be divided by $2x+1$ over $\mathbb Z$.

Polynomial division by monic polynomials does work over all rings. So, if $a$ is a root of a polynomial $p$ over a ring $R$, then $p(x)=(x-a)q(x)$. If $b\ne a$ is another root of $p$, then $(b-a)q(b)=0$. Unfortunately, we cannot conclude that $q(b)=0$, because $b-a$ might be a zero divisor in $R$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.