Skip to content

2.8. Polynomial Rings (for Galois Theory).

Say \((R, +, \cdot)\) is a ring with identity \(1 \ne 0\). We will always assume \(R\) is of this form.

  • 1. If \(x\) is indeterminate (i.e. a variable) a polynomial in \(x\) is a formal sum
\[ a_dx^d + \cdots + a_1x + a_0. \]

with \(a_k \in R\). * 2. We say the degree \(\deg(p) = d\) if \(a_d\). Otherwise, set \(\deg(p) = -\infty\) if \(a_d = \cdots = a_0 = 0\). * 3. Let \(R[x]\) be the collectionn of all such \(p(x)\). This is the polynomial ring over \(R\). * 4. More generally, say \(\{x_1, x_2, \dots, x_n\}\) is a collection of \(n\) indeterminates. Inductively define \(R[x_1, x_2, \dots, x_n] = S[x_n]\) where \(S = R[x_1, x_2, \dots, x_{n-1}]\). A typical element is in the form

\[ p(x_1, \cdots, x_n) = \sum_{i_1 = 1}^{d_1}\cdots\sum_{i_n = 1}^{d_n}a(i_1, \cdots, i_n)x_1^{i_1}\cdots x_n^{i_n}. \]

We then say that this \(p(x_1, \dots, x_n)\) is a polynomial in \(n\) variables. * 5. Assuming that the highest coefficient \ \(a(d_1, \dots, d_n) \ne 0\), we then say the degree is

\[ \deg(p) = \max\{i_1 + \cdots + i_n\} \text{ (fix here)}. \]

Otherwise, if all terms are zero, i.e., if all \(a(i_1, \dots, i_n) = 0\), then we set \(\deg(p) = -\infty\).

Example. Consider the polynomial ring \(R[x, y]\) as the polynomial ring in \(n = 2\) variables. Let \(p(x, y) = 1 + x^3 + y^3\). Then \(\deg(p) = 3\).

\((R[x_1, \cdots, x_n], + , \cdot)\) is a ring with \(1 \ne 0\).

We'll showt this by induction:

\[\begin{align*} P(n) = "(R[x_1, \dots, x_n]) \text{ is a ring with } 1 \ne 0". \end{align*}\]

We have already shown the base case \(P(1)\) is true. Assume that \(P(n)\) is true for some \(n_0\). We show \(P(n)\) is true for \(n = n_0 + 1\).

By definition, \(R[x_1, \dots, x_n] = S[x_n]\) where \(S = R[x_1, \dots, x_{n_0}]\). By our inductive hypothesis, we have that \ \((S, +, \cdot )\) is a ring with \(1 \ne 0\). Hence by \(P(1)\) we have that \((S[x_n], \cdot, +)\) is also a ring with \(1 \ne 0\). Hence \(P(n_0 + 1)\) is true.

Assume \(R\) is an integral domain.

  • 1. Suppose \(p, q \in R[x_1, \cdots x_n]\). Then

$$ \deg(p \cdot q) = \deg(p) + \deg(q).
$$ * 2. \(R[x_1, \dots, x_n]\) is also an integral domain. * 3. The units of \(R[x_1, \dots, x_n]\) is \(R^{\times}\).

  • 1. If either \(p = 0\) or \(q = 0\), then \(p \cdot q = 0\). Then observe that
\[ \deg(p \cdot q) = \deg(p) + \deg(q) \]

doesn't make sense unless we choose to set \(\deg(0) = -\infty\). Hence we see that

\[ \deg(p \cdot q) = -\infty. \]

We could make it positive infinity, but we assume that it is negative for subtle reasons later on.

Now assume that \(p, q \ne 0\). We write our polynomials

\[ p = \sum_{i_1, \dots, i_n}^{d_1, d_2, \cdots, d_n}a(i_1, \dots, i_n)x_1^{i_1}\cdots x_n^{i_n} \qquad q = \sum_{j_1, \dots, j_n}^{e_1, e_2, \cdots, e_n}a(j_1, \dots, j_n)x_1^{j_1}\cdots x_n^{j_n} \]

where \(\deg(p) = d_1 + d_2 + \cdots + d_n\) and \(\deg(q) = e_1 + e_2 + \cdots + e_n\). Then we see that

\[ p \cdot q = \sum_{k_1, \dots, k_n}^{d_1 + e_1, d_2 + e_2, \cdots, d_n + e_n}c(i_1, \dots, i_n)x_1^{k_1}\cdots x_n^{k_n} \]

where

\[ c(k_1, \dots, k_n) = \sum_{i_1+j_1 = k_1, \dots, i_n+j_n = k_n.} a(i_1, \dots i_n)b(j_1, \dots, j_n). \]

The leading term is \(c(d_1 + e_1, \dots, d_n + e_n) = a(d_1, \dots, d_n )b(e_1, \dots, e_n)\). Hence, the leading term is also nonzero, so the degree must be

\[ d_1 + e_1 + \dots + d_n + e_n = \deg(p) + \deg(q). \]

There are actually many ways to define the degree of a polynomial in \(R[x_1, \dots, x_n]\) when\(n \ge 2\). * 2., 3. We can show both (2) and (3) at the same time via induction. Let

\[ P(n) = "R[x_1, \dots, x_n] \text{ is an integral domain and } R[x_1, \dots, x_n]^{\times} = R^{\times}." \]

We've done this in the one-variable case, so that \(P(1)\) is true. We can invoke our inductive hypothesis to suppose that \(P(n_0)\) is true for some \(n_0\). We next show that \(n = n_0 + 1\) is true.

By definition, our ring \(R[x_1, \dots, x_{n_0}] = S[x_n]\) for

\[ S = R[x_1, \dots, x_{n_0}]. \]

By our inductive hypothesis, we know that (1) \(S\) is an integral domain. Since \(P(1)\) is true, we know that \(S[x_{n_0}]\) is an integral domain. By (2), we know that \(S^\times = R^\times\). By \(P(1)\), we see that \(S[x_n] = R^\times\). This show that \(P(n)\) is true for all \(n\).

Say \((R, +, \cdots)\) is a ring with \(1 \ne 0\) and \(I \subsetneqq R\) is a ideal. Denote \(S = R[x_1, \dots, x_n]\).

  • 1. \(J = I[x_1, \dots, x_n]\) is a proper ideal of \(S\)
  • 2. \(S/J = (R/J)[x_1, \dots, x_n]\)
  • 3. Moreover, say \(R\) is commutative. If \(I \subset R\) is a prime ideal of \(R\), then \(J\) is a prime ideal of \(S\).

  • 1., 2. Denote \(\overline{R} = R/I\) as a ring with identity \(1 \ne 0\), which holds since we are working with a proper ideal. We know both \(S\) and \(\overline{R}[x_1, \dots, x_n]\) is also a ring with \(1 \ne 0\).

Consider the following "reduction mod \(I\)" map:

\[ \phi: R[x_1,\dots, x_n] \to \overline{R}[x_1,\dots, x_n] \]

where if \(\displaystyle p = \sum_{i_1, \dots, i_n}^{d_1, d_2, \cdots, d_n}a(i_1, \dots, i_n)x_1^{i_1}\cdots x_n^{i_n}\) then

\[ \overline{p} = \sum_{i_1, \dots, i_n}^{d_1, d_2, \cdots, d_n}\overline{a(i_1, \dots, i_n)}x_1^{i_1}\cdots x_n^{i_n} \]

where \(\overline{a} = a + I\) in \(R/I\). \ Claim: This is a ring homomorphism. In fact, this map is surjective. \ Neither of these are difficult to show.

Observe that

\[ \ker(\phi) = \Big\{p = \sum_{i_1, \dots, i_n}^{d_1, d_2, \cdots, d_n}a(i_1, \dots, i_n)x_1^{i_1}\cdots x_n^{i_n} mid a(i_1, \dots, i_n)\in I \Big\} = I[x_1, \dots, x_n]. \]

The First Isomorphism Theorem for Rings states that

\[\begin{align*} S/J &\cong R[x_1, \dots, x_n]/\ker(\phi)\\ &\cong \im(\phi)\\ &= (R/I)[x_1, \dots, x_n]. \end{align*}\]

At this point, we've shown (2). Now observe that \(J \subsetneqq R\) since \((R/I)[x_1, \dots, x_n]\) is a ring with identity \(1 \ne 0\). This proves (1). To show (3), say \(R\) is a commutative ring and \(I \subset R\) is a prime ideal. Since \(I\) is prime, we see that \((R/I)\) is an integral domain. Hence we see that \((R/I)[x_1, \dots, x_n]\) is an integral domain. Since \(S/J \cong (R/I)[x_1, \dots, x_n]\), and we know that \(R[x_1, \dots, x_n]\) is also a commutative ring, we see that \(J \subset S\) must be a prime ideal as well.