Consider the splitting field $E$ of $X^4 -2$.

Compute $T_{E|\mathbb{Q}}(\alpha)$ and $N_{E|\mathbb{Q}}(\alpha)$ for every $\alpha \in E$.

My attempt:

The splitting field $E = \mathbb{Q}(i,\sqrt{2},\sqrt[4]{2})$ which is a degree 8 extension with basis $\mathcal{B} = \{1,i,\sqrt{2},\sqrt[4]{2},i\sqrt{2},i\sqrt[4]{2},\sqrt[4]{8}\}$

We can use the definition of the trace in terms of the roots of the minimal polynomial over $\mathbb{Q}$ which is $X^4 -2$, since its an anihilating polynomial of order 4, which is monic. This polynomial is irreducible over $\mathbb{Q}$ by Einsestein criterion, then its the minimal polynomial.

Since the trace is the sum of all the zeros, and every zero is a conjugate of the other, the trace is Also the norm, can be obtained similarly.

I am right ?


2 Answers 2


Note: This answer comes after a comment in a related post, 2819799, although it was already "green" as i started. Well, the dialog is important, this is my only interest...

Let us denote by $$t=\sqrt[4]2$$ the generator of the subfield $$K=\Bbb Q(t)=\Bbb Q(\sqrt[4]2) $$ of the given splitting field, $$ E=K(i)=\Bbb Q(\sqrt[4]2,i)\ , $$ and since we will write this $t$ very often, so a short-hand was imperative above.

We will also identify it (agains my rules of always working algebraically) with the only complex root of the polynomial $X^4-2$, which is real, and positive. The conjugates of $t$ in $\Bbb C$ are $$ t_1=t\ ,\ t_2=it\ ,\ t_3=i^2t\ ,\ t_4=i^3t\ , $$ and corresponding to them we get embeedings $\sigma_j:\Bbb Q(t)\to\Bbb C$, $t\to t_j$.

The system $$ \{\ 1,\ t,\ t^2,\ t^3\ \} $$ is a base of $K=\Bbb Q(t)$ over $\Bbb Q$, and computing the trace and the norm for the extension $K:\Bbb Q$ can be done in terms of it.

The trace is linear, so for computing it, it is enough to compute $$ \begin{aligned} \operatorname{Trace}(1) &= \sum_{1\le 4\le j}\sigma_j(1) &&=1+1+1+1 &&=4\ ,\\ \operatorname{Trace}(t) &= \sum_{1\le 4\le j}\sigma_j(t) &&=t+it+-t-it&&=0\ ,\\ \operatorname{Trace}(t^2) &= \sum_{1\le 4\le j}\sigma_j(t^2) &&=t^2-t^2+t^2-t^2 &&=0\ ,\\ \operatorname{Trace}(t^3) &= \sum_{1\le 4\le j}\sigma_j(t^3) &&=t^3-it^3-t^3+it^3 &&=0\ . \end{aligned} $$ So for a general $$ \xi = a+bt+ct^2+dt^3\in K=\Bbb Q(t)\ ,\qquad a,b,c,d\in \Bbb Q\ , $$ we have $$ \operatorname{Trace}_{K:\Bbb Q}\xi = 4a\ . $$ The norm is relatively simple, too. First, we need an identity: $$ \begin{aligned} &(X+Y+Z+T) (X+iY-Z-iT) (X-Y+Z-T) (X-iY-Z+iT) \\[3mm] &\qquad= \Big(\ (X+Z)^2-(Y+T)^2\ \Big) \Big(\ (X-Z)^2+(Y-T)^2\ \Big) \\[3mm] &\qquad= \Big(\ (X^2+Z^2-2YT)-(Y^2+T^2-2XZ)\ \Big) \\ &\qquad\ \phantom{=} \Big(\ (X^2+Z^2-2YT)+(Y^2+T^2-2XZ)\ \Big) \\[3mm] &\qquad= (X^2+Z^2-2YT)^2-(Y^2+T^2-2XZ)^2 \\[3mm] &\qquad= X^4-Y^4+Z^4-T^4 \\ &\qquad\qquad -2X^2Z^2 + 2Y^2T^2 \\ &\qquad\qquad\qquad - 4 X^2 YT + 4Y^2XZ - 4 Z^2 YT + 4T^2XZ\ . \end{aligned} $$ Now we can simply compute the norm for the "smaller extension" of an element of the form $\xi=a+bt+ct^2+dt^3$ $$ \begin{aligned} \operatorname{Norm}_{K:\Bbb Q}\xi &= \operatorname{Norm}_{K:\Bbb Q}(a+bt+ct^2+dt^3) \\[3mm] &= (a+bt+ct^2+dt^3) (a+ibt-ct^2-idt^3) (a-bt+ct^2-dt^3) (a-ibt-ct^2+idt^3) \\[3mm] &=a^4-b^4t^4+c^4t^8-d^4t^{12}\\ &\qquad -2a^2c^2t^4 + 2b^2d^2t^8\\ &\qquad\qquad-4a^2bdt^4 + 4b^2act^4-4c^2bdt^8+4d^2act^8 \\[3mm] &=a^4-2b^4+4c^4-8d^4\\ &\qquad -4a^2c^2 + 8b^2d^2\\ &\qquad\qquad-8a^2bd + 8b^2ac - 16c^2bd + 16d^2ac\ . \end{aligned} $$ We can ask the computer for a confirmation of the above:

t = 2^(1/4)
prod( [ a + b*i^k*t + c*i^(2*k)*t^2 + d*i^(3*k)*t^3 for k in [0,1,2,3] ] ).expand()

and we get the result:

a^4 - 2*b^4 + 8*a*b^2*c - 4*a^2*c^2 + 4*c^4 - 8*a^2*b*d 
- 16*b*c^2*d + 8*b^2*d^2 + 16*a*c*d^2 - 8*d^4

(Result was manually broken in two lines.) In LaTeX: $$ a^{4} - 2 \, b^{4} + 8 \, a b^{2} c - 4 \, a^{2} c^{2} + 4 \, c^{4} - 8 \, a^{2} b d - 16 \, b c^{2} d + 8 \, b^{2} d^{2} + 16 \, a c d^{2} - 8 \, d^{4} $$ Now each element in $E=K(i)$ can be written uniquely in the form $$ \xi+i\xi'\ . $$ We can use the above computation and the "transitivity of the related norms". The norm for the extension $E:K$ for the above element is $$ (\xi+i\xi')(\xi-i\xi')\ . $$ This brings in scene $\xi^2$ and ${\xi'}^2$. We have to isolate the many components of them w.r.t. the basis $1,t,t^2,t^3$, then apply the above formula.

(For the trace, the game is simpler, the trace over $E:K$ of $\xi+i\xi'$ is the sum of the elements $\xi\pm i\xi'$, so it is $2\xi$. From here we apply the already computed trace.

It is hard to say more in detail for "generic elements".

  • $\begingroup$ I love how you transform the problem to avoid the roots. Thanks. $\endgroup$ Jun 16, 2018 at 20:02

First of all you say that "Hence, all the basis elements are the roots", which isn't true, as the basis has $8$ elements, as noted above, while the polynomial has $4$ roots. Nevertheless the trace can be easily calculated in terms of the representation in basis elements. So we have that

$$T(\alpha) = T(a + b\sqrt[4]{2} + c\sqrt{2} + d\sqrt[4]{8} + ei + fi\sqrt[4]{2} + gi\sqrt{2} + fi\sqrt[4]{8}) = aT(1) = 8a$$

This is true as the trace function is linear and the traces of the basis elements are $0$, except for the element $1$.

On the other side calculating the norm is substantially harder, as the norm is multiplicative, instead of linear. In particular you need to compute:

$$N(\alpha) = \prod_{\sigma \in \text{Gal}(E/\mathbb{Q})}\sigma(\alpha)$$

I guess doing this by hand is almost impossible, as there would be $8^8$ terms if you want to get an expression with respect to the basis representation of $\alpha$.


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.