Skip to main content

Section 16.6 Separability

“I’m trying to be very aware of not repeating myself.”
―Christoph Waltz

Subsection Ring Characteristic

Definition 16.78. Characteristic.

Let \(R\) be a commutative ring. The characteristic of \(R\text{,}\) written \(\char(R)\text{,}\) is the unique non-negative generator of the kernel of the unique ring homomorphism \(\varphi: \Z \to R\text{.}\) (Recall \(\varphi(n) = n \cdot 1_R\text{.}\))
Equivalently, \(\char(R)\) is the smallest positive integer \(n\) such that \(n\cdot 1_R=\underbrace{1_R+\cdots+1_R}_{n}=0_R\text{,}\) if such and integer exists, and \(\char(R) = 0\) otherwise.

Remark 16.79.

Observe that for any integer \(n\) and commutative ring \(R\text{,}\) we have \(n = 0\) in \(R\) (i.e., \(n \cdot 1_R = 0\)) if and only if \(\char(R) \mid n\text{.}\)

Example 16.80. \(\char(\Z)\).

\(\char(\Z)=0\) and \(\char(\Z/n)=n\) for any \(n \geq 0\text{.}\)

Definition 16.81. Prime Field.

For a field \(F\) its prime field is the smallest subfield of \(F\text{;}\) i.e., it is the intersection of all subfields of \(F\text{.}\)

Proof.

For the first assertion, consider the unique ring homomorphism \(\varphi:\Z \to F\text{.}\) Since \(F\) is a domain, the kernel of \(\varphi\) is a prime ideal (since \(\Z/\ker(\varphi) \cong \im(\varphi)\) and \(\im(\varphi)\) is a subring of \(F\)). The result holds since the only prime ideals of \(\Z\) are \((0)\) and \((p)\) for a prime integer \(p\text{.}\) (Note that this proof shows that, more generally, the characteristic of an integral domain must be either \(0\) or a prime.)
For the second assertion, observe that the smallest {} of \(F\) is the image of the ring map \(\varphi: \Z \to F\text{,}\) and by the first assertion, this image is isomorphic to either \(\Z\) or \(\Z/p\text{.}\) The latter is already a field and hence it is the prime field of \(F\text{.}\) In the former case, the prime subfield is isomorphic to the field of fractions of \(\Z\text{,}\) which is \(\Q\text{.}\)

Proof.

Suppose \(F\) and \(E\) are fields and \(\s: F \to E\) is a ring homomorphism. Let \(\phi_F: \Z \to F\) and \(\phi_E: \Z \to E\) be the unique ring maps from \(\Z\) to \(F\) and \(E\text{.}\) Since \(\s \circ \phi_F\) is a ring map from \(\Z\) to \(E\text{,}\) we have \(\s \circ \phi_F = \phi_E\) by the uniqueness of \(\phi_E\text{.}\) Since \(F\) is a field and \(E\) is not the zero ring, the map \(\s\) is injective. Since \(\s\) is injective, it follows that \(\ker(\s \circ \phi_F) = \ker(\phi_F)\text{,}\) and hence we obtain that \(\ker(\phi_F) = \ker(\phi_E)\text{.}\) By definition of characteristic, we conclude \(\ker(E) = \ker(F)\text{.}\)

Subsection Separable Polynomials

Definition 16.84. Root Multiplicity.

For a field \(F\) and a polynomial \(f(x) \in F[x]\text{,}\) let \(\ov{F}\) be an algebraic closure of \(F\) and \(\a \in \ov{F}\) a root of \(f(x)\text{.}\) The multiplicity of \(\a\) in \(f\) is the number of times \((x - \a)\) appears in the factorization \(f(x) = c \prod_i (x - \a_i)\) of \(f\) in \(\ov{F}[x]\text{.}\) (This number is independent of choice of algebraic closure by uniqueness of such closures up to isomorphism.)

Definition 16.85. Separable (Polynomial).

Let \(F\) be a field and \(f(x)\) a polynomial in \(F[x]\text{.}\) If the multiplicity of every root is \(1\text{,}\) we say \(f(x)\) is a separable polynomial; i.e. \(f\) is separable provided it has no repeated roots.

Definition 16.86. Derivative.

For any field \(F\) and \(f(x) = a_n x^n + \dots + a_1 x + a_0 \in F[x]\text{,}\) define its derivative to be
\begin{equation*} f'(x) = n a_n x^{n-1} + (n-1) a_{n-1} x^{n-2} + \cdots + 2 a_2 x + a_1. \end{equation*}

Example 16.87. Derivatives in Characteristic \(p\).

If \(\char(F) = p\) and \(f(x) = x^p +c\) for \(c \in F\text{,}\) then \(f'(x) = 0\text{.}\) So beware that non-constant polynomials can have \(0\) derivatives! Observe, however, that this cannot occur in characteristic \(0\text{.}\)

Proof.

Let \(h = \gcd_{F[x]}(f, g)\text{.}\) To prove (1), we note that \(h\) is the unique monic polynomial such that - \(h = f \cdot p + g \cdot q\) for some \(p,q \in F[x]\text{,}\) - \(f = h \cdot \a\) for soem \(\a \in F[x]\text{,}\) and - \(g = h \cdot \beta\) for soem \(\beta \in F[x]\text{.}\) Since \(F[x]\) is a subring of \(L[x]\text{,}\) these three properties also hold when we regard \(h,f,g\) as belonging to \(L[x]\) and thus by the uniqueness property, we have \(h = \gcd_{L[x]}(f,g)\text{.}\)
  1. is a consequence of (1), since if two polynomials factor completely into linear factors, then they are relatively prime if and only if they have no linear factors in common, which is equivalent to their having no roots in common.

Proof.

For (1), suppose \(\a\) is a root of \(f(x)\) of multiplicity at least two. Then \(f(x) = (x-\a)^2g(x)\) in \(\ov{F}[x]\) and hence \(f'(x) = 2(x - \a) g(x) + (x-\a)^2 g(x)\text{,}\) by the Product Rule. It follows that \(f'(\a) = 0\text{.}\) Conversely, suppose \(f(\a) = f'(\a) =0\text{.}\) Since \(f(\a) = 0\text{,}\) we have \(f(x) = (x-\a)h(x)\) and hence \(f'(x) = h(x) + (x-\a)h'(x)\text{.}\) Since \(f'(\a) = 0\) it follows \(h(\a) = 0\) and thus \(\a\) has multiplicity at least two.
For the second assertion, by (1), we have that \(f\) is separable if and only if \(f\) and \(f'\) has no common roots in \(\ov{F}\text{.}\) The result thus follows from the Lemma [provisional cross-reference: cite]
For the final assertion, assume \(f(x)\) is irreducible. Since the degree of \(f'(x)\) is strictly less than the degree of \(f(x)\text{,}\) we have that \(\gcd(f(x),f'(x)) \ne 1\) if and only if \(f'(x) = 0\text{.}\)

Example 16.90. Separability and \((x^4-x^3-x+1)\).

\(x^4 -x^3 -x + 1\) is not separable sine \(x-1\) is a double root (it factors as \((x^3-1)(x-1) = (x^2+x+1)(x-1)^2\)). As predicted by the Theorem [provisional cross-reference: cite], it fails to be relatively prime to its derivative, which is \(4x^3 - 3x^2 -1\text{,}\) since each are divisible by \(x-1\text{.}\)
\(x^3-1\) is separable in \(\R[x]\) because it has \(3\) distinct roots in \(\C\text{,}\) namely \(1,\zeta_3,\zeta_3^2\text{.}\) As predicted by the Theorem, it is relatively prime to its derivative \(3x^2\text{.}\)
Now interpret \(x^3 - 1\) as belonging to \((\Z/3)[x]\text{.}\) Then \(x^3 - 1 = (x-1)^3\) is not separable. As predicted by the Theorem, it is not relatively prime to its derivative, which is \(0\text{.}\)

Example 16.91. \(\Z/p(y)\).

Let \(F\) be a field of characteristic \(p > 0\) and assume \(a \in F\) is an element such that \(a \ne b^p\) for all \(b \in F\text{.}\) Then \(x^p - a\) is irreducible but not separable. It is not separable since in \(\ov{F}\) we have \(x^p - a= (x-b)^p\) where \(b^p = a\text{.}\) Also note that its derivative is \(px^{p-1} = 0\text{.}\)
It is less obvious that it is irreducible, but we can see that this is indeed the case in a specific example: Take \(F = \Z/p(y)\) (the field of fractions of the polynomial ring \(\Z/p[y]\)) and let \(a = y\text{.}\) In this case, \(x^p - y\) is seen to be irreducible, by Eisenstein’s Criterion, but not separable.

Subsection Separable Field Extensions

“Because we are separated everything separates us, even our efforts to join each other.”
―Simone de Beauvoir

Definition 16.93. Separable (Field Extension).

An algebraic field extension \(F \subseteq L\) is called separable if for every \(\a \in L\) its minimum polynomial \(m_{\a, F}(x)\) is separable (i.e., has no repeated roots in an algebraic closure of \(F\)).

Proof.

Suppose \(F\subseteq L\) is an algebraic extension and let \(\alpha\in L\) be algebraic over \(F\text{.}\) Then the minimal polynomial \(f(x)\) of \(\alpha\) over \(F\) has coefficients in \(F\) and is of the form
\begin{equation*} f(x) = x^m + a_{m-1} x^{m-1} + \cdots + a_1 x + a_0 \end{equation*}
where \(a_i \in F\text{.}\) We need to show that \(f(x)\) is separable, i.e., has no repeated roots in its splitting field.
Suppose \(f(x)\) has a repeated root \(\beta\) in some splitting field \(K\) of \(f(x)\text{,}\) i.e., \(f(x) = (x-\beta)^2 g(x)\) for some polynomial \(g(x) \in K[x]\text{.}\) Since \(\alpha\) is also a root of \(f(x)\text{,}\) we have \((\alpha - \beta)^2 g(\alpha) = 0\text{.}\) Since \(F\subseteq L\) is an algebraic extension, \(\alpha\) is algebraic over \(F\text{,}\) so \(g(\alpha)\neq 0\text{.}\) It follows that \((\alpha-\beta)^2=0\text{,}\) i.e., \(\alpha=\beta\text{,}\) which means that \(\alpha\) is a repeated root of \(f(x)\text{,}\) contradicting the assumption that \(f(x)\) has no repeated roots. Hence \(f(x)\) is separable, and since \(\alpha\) was an arbitrary algebraic element of \(L\text{,}\) we conclude that \(F\subseteq L\) is a separable extension.

Example 16.95. \(F = (\Z/p)(y) \subseteq (\Z/p)(z) = L\).

Let \(y\) and \(z\) be indeterminants. The extension of fields \(F = (\Z/p)(y) \subseteq (\Z/p)(z)= L\) given by identifying \(y\) with \(z^p\) is not separable. Somewhat more precisely, \(F\) is isomorphism to the subfield of \(L\) consisting of elements of the form \(\frac{\sum_i a_i z^{ip}}{\sum_j b_j z^{jp}}\text{,}\) with the isomorphism given by sending \(\frac{\sum_i a_i y^{i}}{\sum_j b_j y^{j}}\) to \(\frac{\sum_i a_i z^{ip}}{\sum_j b_j z^{jp}}\text{.}\)
Then \(z \in L\) is a root of the polynomial \(x^p - y \in F[x]\text{.}\) Moreover since \(F\) is the field of fractions of the PID \(R = (\Z/p)[y]\) and \(y\) is a prime element of \(R\text{,}\) we may apply Eisenstein’s Criterion (and Gauss) to conclude that \(x^p - y\) is irreducible in \(F[x]\text{.}\) This proves that \(m_{z, F}(x) = x^p - y\text{.}\) This polynomial is not separable since in \(L[x]\) it is equal to \((x-z)^p\) and hence has a repeated root. (Or, you may use that its derivative is \(0\text{.}\))