5

Suppose we have $f(x)=x^3-x^2-4x-1$ and $\alpha \in \mathbb{C}$ a root of $f(x)$. (We do not know the value of $\alpha$.) It is easy to see that $f(x)$ is irreducible in $\mathbb{Q}[x]$.

It is also easy to prove $-\frac{1}{1+\alpha}$ is also a root of $f(x)$. I need to show $\mathbb{Q}(\alpha)$ is the decomposition field of $f(x)$. Since $f(x)$ is separable, we have three distinct roots $\alpha,-\frac{1}{1+\alpha}, \gamma$. So the decomposition field is $\mathbb{Q}(\alpha,-\frac{1}{1+\alpha}, \gamma)=\mathbb{Q}(\alpha, \gamma)$. How do I show that this is actually equal to $\mathbb{Q}(\alpha)$ without actually calculating the roots of $f(x)?$

usere5225321
  • 1,309

4 Answers4

6

Notice that the sum of all the roots equals the coefficient at $x^2$, e.g. $-1$. So $\gamma$ can be expressed in polynomial terms of $\alpha$.

Jef
  • 3,821
4

You already have two excellent answers by Jef and Will, but I really do think it is worth mentioning a method that you can just memorise and apply mindlessly in future situations.

Let $\alpha$, $\beta$ and $\gamma$ be the three roots of your irreducible cubic. The Galois group is a subgroup of the permutation group on $ \{ \alpha, \beta, \gamma \}$. Furthermore, the Galois group acts transitively on the roots, since the cubic is irreducible. So there are two possibilities:

  • The Galois group is $S_3$, in which case the degree of the extension is six, so the splitting field is larger than $\mathbb Q(\alpha)$.
  • The Galois group is $A_3$, in which case the degree of the extension is three, so the splitting field is equal to $\mathbb Q(\alpha)$.

To distinguish between the two cases, consider the discriminant of the cubic: $$ \Delta = (\alpha - \beta)^2(\beta - \gamma)^2(\gamma - \alpha)^2.$$

  • If the Galois group is $S_3$, then the odd permutations in $S_3$ send $\sqrt{\Delta} \mapsto - \sqrt{\Delta}$, so $\sqrt{\Delta}$ is not fixed by the Galois group, hence $\sqrt{\Delta} \notin \mathbb Q$.
  • If the Galois group is $A_3$, then the Galois group only contains even permutations, which all send $\sqrt{\Delta} \mapsto + \sqrt{\Delta}$, so $\sqrt{\Delta}$ is fixed by the Galois group, hence $\sqrt{\Delta} \in \mathbb Q$.

There is a formula for the discriminant in terms of the coefficients of the cubic. If the cubic is $$ f(x) = x^3 + bx^2 + cx + d,$$ then the discriminant is $$ \Delta = b^2 c^2 - 4c^3 -4b^3 d - 27d^2 + 18bcd. $$ [Note that this simplifies to $\Delta = -4c^3 - 27 d^2$ if $b = 0$, which is easy to remember.]

In your example, $\Delta = 169$, so $\sqrt{\Delta} = 13 \in \mathbb Q$, hence the Galois group is $A_3$, hence the splitting field is $\mathbb Q(\alpha)$.

Needless to say, your method and the answers by Will and Jef show far more ingenuity than what I have suggested, but still I think it's good to have a reliable tool that you can always fall back upon.

Kenny Wong
  • 32,192
  • 1
    page 15 in https://books.google.com/books?id=wt7lgfeYqMQC&pg=PR1&lpg=PR1&dq=reuschle++tafeln+complexer+primzahlen&source=bl&ots=VGZFPrfUBn&sig=MlQ667PqXaQ9rAvLWkG3_F1rwsk&hl=en&sa=X&ved=0ahUKEwiIwtSvm9TQAhUJ-2MKHXJIA_kQ6AEIODAE#v=onepage&q=reuschle%20%20tafeln%20complexer%20primzahlen&f=false I bought a cheap paperback reprint print-on-demand better to see how the book is organized He does all the primes up to 97, then various composite numbers. Method was introduced in Gauss Disquitiones – Will Jagy Mar 21 '17 at 17:38
  • @WillJagy I'm not familiar with this stuff - did Gauss come up with a super-smart method for solving equations with roots in cyclotomic fields? And how are you meant to realise that the roots are contained in $\mathbb Q(\zeta_{13})$ in the first place? I'm still amazed that you solved that equation! – Kenny Wong Mar 21 '17 at 18:01
  • Kenny, thanks. Gauss did make a super smart method. There is a modern treatment in Galois Theory by David A. Cox, chapter 9. Very helpful. I implemented as computer programs for polynomial degrees, 3,5,7. Let me find where I posted some... – Will Jagy Mar 21 '17 at 18:13
  • raw list of cubics: http://math.stackexchange.com/questions/1987168/does-this-trigonometric-pattern-continue-for-all-primes-p-6m1/2021528#2021528 – Will Jagy Mar 21 '17 at 18:16
  • quintics, I asked, then found out and programmed the Gauss method and put the answers in the question as edit: http://math.stackexchange.com/questions/1996552/any-more-cyclic-quintics – Will Jagy Mar 21 '17 at 18:23
  • degree 7 : http://math.stackexchange.com/questions/1996552/any-more-cyclic-quintics/2033565#2033565 – Will Jagy Mar 21 '17 at 18:24
  • 1
    Oh, I did not solve this question, I remembered it... – Will Jagy Mar 21 '17 at 18:25
3

Just for laughs, the roots are $$ 2 \cos \left( \frac{\pi}{13} \right) + 2 \cos \left( \frac{5\pi}{13} \right), $$ $$ 2 \cos \left( \frac{3\pi}{13} \right) + 2 \cos \left( \frac{11\pi}{13} \right), $$ $$ 2 \cos \left( \frac{7\pi}{13} \right) + 2 \cos \left( \frac{9\pi}{13} \right). $$

Reuschle uses $x^3 + x^2 - 4 x + 1$ which is what you get by the method of Gauss, see page 15, jpeg below. I recommend Galois Theory by David A. Cox, chapter 9 for a modern exposition. Method introduced by Gauss in Section VII of the Disquisitiones, some 30 years before Galois Theory. Gauss did few examples, it does take some practice.

enter image description here

Why not: here is a jpeg from the AbeBooks site, showing how to order Reuschle(1875) as a paperback on-demand reprint. The reprint is from a free online source, they don't have a big library of actual old books there at the printing location. I really prefer having actual books to read rather than just on a computer screen.

enter image description here

I wrote programs in November and December (2016) to do this. So, here is the whole song and dance for degree 7, prime 29:

jagy@phobeusjunior:~$ ls -l | grep septic | grep greedy
-rwxrwxr-x  1 jagy jagy   370501 Dec  4 14:11 septic_cyclic_gauss_greedy
-rw-rw-r--  1 jagy jagy    23990 Dec  4 14:12 septic_cyclic_gauss_greedy.cc
-rw-rw-r--  1 jagy jagy    23989 Dec  4 14:12 septic_cyclic_gauss_greedy.cc~
jagy@phobeusjunior:~$ 



jagy@phobeusjunior:~$  ./septic_cyclic_gauss_greedy  29
 g to the e  12
    1   12    generator  
    2   28
    3   17    generator  
    1   12   17   28
   a   2

 h1  
    1   12   17   28
 h2  
    2    5   24   27
 h4  
    4   10   19   25
 h8  
    8    9   20   21
 h16  
   11   13   16   18
 h32  
    3    7   22   26
 h64  
    6   14   15   23


=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=

  h1_1     4    0    1    0    0    2    0    0
  h1_2     0    1    0    1    0    0    1    1
  h1_4     0    0    1    0    1    1    1    0
  h1_8     0    0    0    1    2    0    1    0
 h1_16     0    2    0    1    0    0    0    1
 h1_32     0    0    1    1    1    0    0    1
 h1_64     0    0    1    0    0    1    1    1

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=

   h1_1     4    0    1    0    0    2    0    0
   h2_2     4    0    0    1    0    0    2    0
   h4_4     4    0    0    0    1    0    0    2
   h8_8     4    2    0    0    0    1    0    0
 h16_16     4    0    2    0    0    0    1    0
 h32_32     4    0    0    2    0    0    0    1
 h64_64     4    1    0    0    2    0    0    0

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=

 constant               1              0              0              0              0              0              0              0
 linear               0              1              0              0              0              0              0              0
 fifth                 0            100             11             50             15              5             25             50

 sixth               400             21            225             56            105            300            126             91

 seventh                84           1225            294            756            392            189            477            742


7  sofar               84           1225            294            756            392            189            477            742


6  sofar              484           1246            519            812            497            489            603            833

 fourth                36              0             16              1              4             24              6              4

 cubed                 0              9              0              3              0              0              1              3

 squared               4              0              1              0              0              2              0              0
5  sofar              484             46            387            212            317            429            303            233

 table  
       1       3/4         0       1/4    -233/4

       1       1/6         0       1/6    -101/2

       1         0      1/12      1/24    -143/8

       1         0         0       1/4    -317/4

       1         3         0         1      -212

       1         0      1/16      1/16   -387/16


=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=

       1         0         0         0        -7

       0         1         0         0        28

       0         0         1         0        14

       0         0         0         1      -289

       0         0         0         0         0

       0         0         0         0         0

4  sofar              232             46            275            205            289            261            261            205

3  sofar              232            298            275            289            289            261            289            289

2  sofar              288            298            289            289            289            289            289            289

1  sofar              288            289            289            289            289            289            289            289

0  sofar              289            289            289            289            289            289            289            289

 confirm               0              0              0              0              0              0              0              0

  x^7 + x^6 - 12 x^5 - 7 x^4 + 28 x^3 + 14 x^2 - 9 x + 1
 constant 289
   p  29 p.root  2 exps 12^k 
 list of the 4 exponents 
       1      12      17      28


   $$   x^7 + x^6 - 12 x^5 - 7 x^4 + 28 x^3 + 14 x^2 - 9 x + 1, \; \;  p = 29, \; \;  r = 2, \; \; 12^k  $$


gp-pari: 

  x^7 + x^6 - 12 * x^5 - 7 * x^4 + 28 * x^3 + 14 * x^2 - 9 * x + 1


 x = t + (1/t)  + t^12 + (1/t^12)

jagy@phobeusjunior:~$ 

enter image description here

parisize = 4000000, primelimit = 500509
? f = x^7 + x^6 - 12 * x^5 - 7 * x^4 + 28 * x^3 + 14 * x^2 - 9 * x + 1
%1 = x^7 + x^6 - 12*x^5 - 7*x^4 + 28*x^3 + 14*x^2 - 9*x + 1
? polroots(f)
%2 = [
-3.347297326211866604824677822 + 0.E-28*I, 
-1.453219237250277575521353021 + 0.E-28*I, 
-1.063840303785358166816481464 + 0.E-28*I, 
 0.1723984388388905398234384116 + 0.E-28*I, 
 0.2395267590849948773703028220 + 0.E-28*I, 
 1.700463948582122544295969145 + 0.E-28*I, 
 2.751967720741494385672801928 + 0.E-28*I]~
? 

some of these roots are $$ 2 \cos \left( \frac{2 \pi}{29} \right) + 2 \cos \left( \frac{24 \pi}{29} \right) = 2 \cos \left( \frac{2 \pi}{29} \right) - 2 \cos \left( \frac{5 \pi}{29} \right) \approx 0.239526759 $$ $$ 2 \cos \left( \frac{4 \pi}{29} \right) + 2 \cos \left( \frac{48 \pi}{29} \right) = 2 \cos \left( \frac{4 \pi}{29} \right) + 2 \cos \left( \frac{10 \pi}{29} \right) \approx 2.75196772 $$ $$ 2 \cos \left( \frac{8 \pi}{29} \right) + 2 \cos \left( \frac{96 \pi}{29} \right) = 2 \cos \left( \frac{8 \pi}{29} \right) - 2 \cos \left( \frac{9 \pi}{29} \right) \approx 0.1723984 $$ $$ 2 \cos \left( \frac{16 \pi}{29} \right) + 2 \cos \left( \frac{192 \pi}{29} \right) = -2 \cos \left( \frac{13 \pi}{29} \right) - 2 \cos \left( \frac{11 \pi}{29} \right) \approx -1.06384 $$ $$ 2 \cos \left( \frac{32 \pi}{29} \right) + 2 \cos \left( \frac{384 \pi}{29} \right) = -2 \cos \left( \frac{3 \pi}{29} \right) - 2 \cos \left( \frac{7 \pi}{29} \right) \approx -3.347297326 $$ $$ 2 \cos \left( \frac{64 \pi}{29} \right) + 2 \cos \left( \frac{768 \pi}{29} \right) = 2 \cos \left( \frac{6 \pi}{29} \right) + 2 \cos \left( \frac{14 \pi}{29} \right) \approx 1.70046 $$ $$ 2 \cos \left( \frac{128 \pi}{29} \right) + 2 \cos \left( \frac{1536 \pi}{29} \right) = 2 \cos \left( \frac{12 \pi}{29} \right) - 2 \cos \left( \frac{ \pi}{29} \right) \approx -1.4532 $$

enter image description here enter image description here

Will Jagy
  • 139,541
  • ahaha thanks! Worth a laugh. – usere5225321 Mar 20 '17 at 02:10
  • @usere5225321 they really are the roots, Has to do with cyclotomic fields. The method of Gauss leads most naturally to $x^3 + x^2 - 4x +1,$ the result of negating your $x.$ This particular example is not in the Disquistiones, I found it in an 1875 book; available online, I bought a cheap paperback reprint. – Will Jagy Mar 20 '17 at 02:28
  • 1
    @usere5225321: Worth a laugh and also an opportunity to get amazed:-) The fact that $\Bbb{Q}(\alpha)$ is the splitting field implies that the Galois group is cyclic of order three. By Kronecker-Weber theorem the field then has to be a subfield of some cyclotomic field. This time apparently a subfield of $\Bbb{Q}(\zeta_{13})$. – Jyrki Lahtonen Mar 21 '17 at 08:42
  • 1
    @JyrkiLahtonen I may have mentioned to you, page 15 in https://books.google.com/books?id=wt7lgfeYqMQC&pg=PR1&lpg=PR1&dq=reuschle++tafeln+complexer+primzahlen&source=bl&ots=VGZFPrfUBn&sig=MlQ667PqXaQ9rAvLWkG3_F1rwsk&hl=en&sa=X&ved=0ahUKEwiIwtSvm9TQAhUJ-2MKHXJIA_kQ6AEIODAE#v=onepage&q=reuschle%20%20tafeln%20complexer%20primzahlen&f=false I bought a cheap paperback reprint print-on-demand better to see how the book is organized – Will Jagy Mar 21 '17 at 17:32
  • @WillJagy Does that method cover this cubic as well? Sounds like it might. I don't know how to convert whatever Cardano spits out to those sums of cosines. Checking the discriminant and a bit of educated guessing gave me the zeros of that cubic, but I couldn't really help the asker. May be you can? – Jyrki Lahtonen Mar 29 '17 at 04:45
  • @JyrkiLahtonen I posted an answer at that question http://math.stackexchange.com/questions/2206564/find-splitting-field-of-a-cubic-polynomial Probably not every necessary detail in discussing the splitting field. – Will Jagy Mar 30 '17 at 18:49
1

Let f:=x^3-x^2-4x-1. It is easy to see that f is an irreducible polynomial over Q. In fact, the degree of f is 3 and so if we prove that f hasn't rational roots, we can conclude that f is irreducible (over Q). Since 1 and -1 are not zeros of f, f is then irreducible.

Now, a is a root of f. You can easy proved that b=:-1/(1+a) is also a root of f. And so, the third root is: c= -1/(1+b) or equivalently: -1/((-1)^3 * ab) = -(1+a)/a. Since the degree of f is odd, f has at least one real root. Since f is separable (f is irreducible and the characteristic of Q is zero), f has not multiple roots. And so: a, b and c are three (simple) real roots. Q(a) is then the splitting field of f over Q (because b and c are clearly elements in Q(a)) and Q(a)/Q is a Galois extension. I hope this helps you. I am very sorry for my bad English.

  • Maybe you can give a more detailed explanation for the benefit of the OP? – mlc Mar 28 '17 at 16:10
  • Yes... f in its splitting field, splits (x-a)(x-b)(x-c). So: -abc = -1 (the term of f without x). We have a and b and so we can easy find c which is clearly in Q(a). – Omar Baratelli Mar 28 '17 at 16:18