6
$\begingroup$

In this Wikipedia article it is mentioned that primes of the form $p \bmod 3 = 1$ split in the ring of Eisenstein integers $\Bbb E$. By analogy to the Gaussian primes I tried to prove that for such primes there exists $a, b \in \Bbb Z$ such that $p = a^2-ab+b^2$. But I got stuck somewhere. Let's take following ingredients : $\omega, \omega^2$ the usual $3$thd degree roots of unity, $u$ a primitive root of $\Bbb Z_p$, $\mathfrak{p}$ the ideal $\Bbb Fp$. I am interested in the quotient ring $\Bbb F/\mathfrak{p}$. I chose $p^2$ elements of the form $a\omega+b\omega^2$ with $0 \leq a,b < p$ as representatives of the cosets and multiplication is executed as complex numbers and applying $\bmod p$ afterwards. It is not hard to see that $r = u^{\frac{p-1}{3}}$ is a root of $X^3-1 = (X-1)(X^2+X+1)$ in $\Bbb Z_p$. It is not a root of $X-1$ so $-r$ is a root of $X^2-X+1$. From this we can conclude that for $x = \omega +(p-r)\omega^2$ we must have that its norm $N(x) = 0 \bmod p$. Unluckily I don't see how this allows me to find an element with norm = $p$.

  • Example 1: $p = 19$

We have $u = 2$ and $r = 2^6 \bmod 19 = 7$, $x = \omega + 12\cdot\omega^2$ but $N(x) = 7 \cdot 19$. I know that the ideal generated by $x$ in $\Bbb F/\mathfrak{p}$ is $x, 2x, 3x, \ldots, 18x$ and that (by luck?) $N(3x) = 19$ so that $19 = 3^2 - 3\cdot5 + 5^2$;

  • Example 2: $p = 37$

We have $u = 2$ and $r = 2^{12} \bmod 37 = 26$, $x = \omega + 11\cdot\omega^2$ but $N(x) = 3 \cdot 37$. I know that the ideal generated by $x$ in $\Bbb F/\mathfrak{p}$ is $x, 2x, 3x, \ldots, 36x$ and that (by luck?) $N(4x) = 37$ so that $37 = 4^2 - 4\cdot7 + 7^2$;

  • Conclusion

I am able to construct an ideal of elements all of which have norm that is a multiple of $p$, I strongly suspect that one of these has exactly norm $p$ but can't prove it.

$\endgroup$
  • $\begingroup$ I assume that $\Bbb{F}=\Bbb{E}$? Also, I think it would be easier to choose representatives of the form $a+b\omega$. And in your examples, the ideals should include $0$. And note that the norm is multiplicative; if $N(x)=7\cdot19$ then $N(3x)=N(3)N(x)=3^2\cdot7\cdot19$. The norm does not factor to a map on $\Bbb{E}/\mathfrak{p}$. $\endgroup$ – Inactive - avoiding CoC Dec 17 '16 at 21:54
2
$\begingroup$

There seem to be some misunderstandigs in your current approach; see my comment to your question. Here's an alternative approach that for the most part generalizes to the context of rings of integers in number fields:

The ring of Eisenstein integers is isomorphic to the quotient $\Bbb{Z}[X]/(X^2+X+1)$. If $p\in\Bbb{Z}$ is a prime with $p\equiv1\pmod3$ then $(\Bbb{Z}/p\Bbb{Z})^{\times}$ is a group whose order is a multiple of $3$, so it has two distinct primitive third roots of unity, say $a$ and $b$. Then in $(\Bbb{Z}/p\Bbb{Z})[X]$ we have $$X^2+X+1=(X-a)(X-b),$$ from which it follows that \begin{align*} \Bbb{E}/p\Bbb{E} &\cong\Bbb{Z}[X]/(p,X^2+X+1)\\ &\cong(\Bbb{Z}/p\Bbb{Z})[X]/(X^2+X+1)\\ &\cong(\Bbb{Z}/p\Bbb{Z})[X]/(X-a)\times(\Bbb{Z}/p\Bbb{Z})[X]/(X-b). \end{align*} In particular this gives a bijection between the prime ideals of $\Bbb{E}$ containing $p$ and the prime ideals of $(\Bbb{Z}/p\Bbb{Z})[X]/(X-a)\times(\Bbb{Z}/p\Bbb{Z})[X]/(X-b)$, which is a product of two fields. So its prime ideals are $$(\Bbb{Z}/p\Bbb{Z})[X]/(X-a)\times\{0\}\qquad\text{ and }\qquad\{0\}\times(\Bbb{Z}/p\Bbb{Z})[X]/(X-b).$$ The corresponding ideals in $\Bbb{E}$ are $(p,\omega-a)\Bbb{E}$ and $(p,\omega-b)\Bbb{E}$. Their product is $p\Bbb{E}$ which is an ideal of norm $p^2$, so both ideals are of norm $p$. Because $\Bbb{E}$ is a principal ideal domain, there exist $x,y\in\Bbb{E}$ such that $$x\Bbb{E}=(p,\omega-a)\Bbb{E}\qquad\text{ and }\qquad y\Bbb{E}=(p,\omega-b)\Bbb{E},$$ and so $x$ and $y$ both have norm $p$. Because $\Bbb{E}$ is even a Euclidean domain, both $x$ and $y$ can be computed by the Euclidean algorithm.


EDIT: Let's see what this means for the given examples. For $p=19$ we find $$X^2+X+1\equiv(X-7)(X-11)\pmod{19},$$ so the prime ideals of $\Bbb{E}$ containing $p$ are $(19,\omega-7)\Bbb{E}$ and $(19,\omega-11)\Bbb{E}$. We apply the Euclidean algorithm to the generators $19$ and $\omega-7$ of the first ideal, where $$N(19)=19^2\qquad\text{ and }\qquad N(\omega-7)=57=3\times19.$$ Now we want to find $q,r\in\Bbb{E}$ such that $19=q(\omega-7)+r$ with $N(r)<57$. We know that $$57=N(\omega-7)=(\omega-7)(\omega^2-7)=(\omega-7)(-\omega-8),$$ so $\frac{1}{\omega-7}=\frac{-\omega-8}{57}$. Hence $q$ should be close to $$\frac{19}{\omega-7}=19\times\frac{-\omega-8}{57}=-\frac{8}{3}-\frac{1}{3}\omega.$$ The nearest Eisenstein integer is $-2$, and we find that $$19=(-2)\times(\omega-7)+(5+2\omega),$$ where $N(5+2\omega)=19$, so we're done. The other prime of norm $19$ is then $$5+2\omega^2=3-2\omega,$$ and indeed if we apply the same algorithm to the ideal $(19,\omega-11)\Bbb{E}$ we find that $\frac{19}{\omega-11}=\frac{-12-\omega}{7}$, which is close to $-2$, and so $$19=(-2)\times(\omega-11)+(-3+2\omega).$$

$\endgroup$
  • $\begingroup$ I know this theorethical approach, but it doesn't answer my question. The question is whether there is an explicit construction to find these numbers, for instance in the examples I gave. $\endgroup$ – Marc Bogaerts Dec 18 '16 at 0:00
  • $\begingroup$ I'll expand upon the last part of my answer, about the Euclidean algorithm. $\endgroup$ – Inactive - avoiding CoC Dec 18 '16 at 0:05
  • $\begingroup$ Ok, It's clear that the Euclidean algorithm is involved, but I'm not used to apply it in cyclotomic integers. $\endgroup$ – Marc Bogaerts Dec 18 '16 at 0:12
  • $\begingroup$ Thanks, this is the answer I was looking for. I guess I have to get used to use the Euclidean algorithm in such cases. $\endgroup$ – Marc Bogaerts Dec 18 '16 at 1:03

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.