I'm looking at Theorem 21.22 here, and specifically the proof of statement 1 implying statement 2, where:

  1. $E$ is a finite extension of $F$
  1. There exists a finite number of algebraic elements $\alpha_1, \dots, \alpha_n \in E$, such that $E=F(\alpha_1, \dots, \alpha_n)$.

The given proof of this goes (emphasis mine):

Let $E$ be a finite algebraic extension of $F$. Then $E$ is a finite dimensional vector space over $F$ and there exists a basis consisting of elements $\alpha_1, \dots ,\alpha_n$ in $E$ such that $E=F(\alpha_1, \dots, \alpha_n)$ . Each $\alpha_i$ is algebraic over $F$ by [a previous theorem].

My confusion arises in the statement that $\alpha_1, \dots ,\alpha_n$, are a basis for $E$ over $F$. This statement is a bit too quick for me.

Previously, for the case of the algebraic extension $F(\beta)$, we see that a basis for $F(\beta)$ is spanned by $\{1, \beta, \beta^2, \dots , \beta^{d-1}\}$, where $d = [F(\beta):F]$.

From this, I would then expect that for $F(\alpha_1, \dots, \alpha_n)$, we have a basis spanned by $\{1, \alpha_1, \dots, \alpha_1^{d_1 -1}, \alpha_2, \dots, \alpha_2^{d_2 -1}, \dots, \alpha_n, \dots, \alpha_n^{d_n -1} \}$, where $d_i$ is the degree of $\alpha_i$ over $F$. Presumably some of the $\alpha_i^{p_i}$ s are linearly dependent, so there are fewer than $1 + \sum_{i=1}^n (d_i -1)$ basis vectors.

So am I missing something in the statement that the algebraic elements $\alpha_1, \dots ,\alpha_n$ form a basis for $E$ over $F$ (should it be containing rather than consisting in the highlighted statement?)? Or is this just shorthand for the statement in the previous paragraph.

  • 1
    $\begingroup$ Note that $F(\beta)=F(1,\beta,\beta^2,\ldots,\beta^{d-1})$. You can always extend a set that generates $E$ as a field into one that generates $E$ as a vector space, and then pare it down to a basis. $\endgroup$ Sep 14 at 18:37
  • 1
    $\begingroup$ It seems like the following is true. Since $E$ is a finite algebraic extension, it is therefore a field. Since it is also finite vector space over $F$, it is spanned by some set of elements $\alpha_i$. Since $E$ is a field, all powers of $\alpha_i$ must also be in $E$. I think a way to think about this is that $E$ is the same as, say, $E(\beta,\gamma)$, which is the same as $E(\beta,\beta^2, \dots, \gamma, gamma^2, \dots$, and we're just calling the $\alpha_i$ the set of powers of $\beta$ and $\gamma$ that are linearly independent. $\endgroup$
    – march
    Sep 14 at 18:52
  • $\begingroup$ OK, so it seems that this is just an issue of the condensed notation hiding the full picture somewhat, which is understandable. $\endgroup$ Sep 14 at 18:57

Since $E$ is a field, it is closed under multiplication. In particular, all powers $\alpha_j^m$ of the the basis elements $\alpha_1,\dots\alpha_n$ (of the vector space $E$ over the field $F$) are also in $E$. Consider the extension $$ F(\alpha_1,\dots,\alpha_n). $$ The field $E$ is a subfield of this extension because all elements of $E$ can be written as a linear combination of the $\alpha_i$. The extension is a subfield of $E$ because any finite linear combination in $\{\alpha_j^k\}$ is inside $E$. Thus, they are equivalent.


Let us observe that if $a_1,a_2,\dots,a_n$ are elements of $E$ then $F(a_1,a_2,\dots,a_n) \subseteq E$. Why?

Because we have $F\subseteq E$ and $a_i\in E $. Hence any element which is made up of elements of $F$ and $a_i$ using field operations lies in $E$. All these elements are precisely the members of $F(a_1,\dots,a_n) $.

Now assume that $E$ is finite dimensional extension of $F$ with $[E:F] =n$. Then we have a basis $\{a_1,\dots,a_n\}\subseteq E$ and every element of $E$ can be written as a linear combination of $a_i$ with coefficients in $F$. And since this linear combination lies in $F(a_1,\dots,a_n)$ we have $E\subseteq F(a_1,\dots,a_n)$ and thus we get $E=F(a_1,\dots,a_n)$.

If $a$ is algebraic over $F$ with degree $n$ then it can be proved that $\{1,a,\dots,a^{n-1}\} $ is a basis of $F(a) $ over $F$ but we can also find another set of $n$ members $a_1,\dots,a_n$ in $F(a) $ such that $a_i$ is a not a power of $a_j$ and yet the $a_i$ form a basis for $F(a) $ over $F$.

For example let $F=\mathbb {Q},a=\sqrt{2}+\sqrt{3}$ and $E=F(a) $. Then $\{1,a,a^2,a^3\} $ as well as $\{1,\sqrt{2},\sqrt {3},\sqrt{6}\}$ are bases of $E$ over $F$.


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.