Is it true that $a+bi$ is prime in $\mathbb{Z}[i]$ if and only if $a^2+b^2$ is prime in $\mathbb{Z}$?

How can I prove this? Can anybody help me please?

  • $\begingroup$ $a^2+b^2=(a+ib)(a-ib)$ $\endgroup$ – Mathematician Apr 21 '13 at 4:35
  • $\begingroup$ this is obvious.but I could not understand how should I use this result $\endgroup$ – habuji Apr 21 '13 at 4:40
  • 13
    $\begingroup$ False. Three is prime in $\mathbb{Z}[i]$, yet $3^2+0^2=9$ is not prime in $\mathbb{Z}$. $\endgroup$ – Jyrki Lahtonen Apr 21 '13 at 4:43
  • 1
    $\begingroup$ If the norm is prime, it can't be the product of Norms of two other numbers unless one is a unit. However the converse isn't always true. $\endgroup$ – Macavity Apr 21 '13 at 4:57
  • 2
    $\begingroup$ you would need restrictions on your statements, such as both a, b are nonzero, as @Jyrki has shown the hole there. Also, $a^2+b^2$ can not be of the form $4k+3$ (again, as is 3). $\endgroup$ – Eleven-Eleven Apr 21 '13 at 5:01

This is not true when $b=0$ and $a$ is an ordinary prime of the form $4k+3$. And for the same reason it is not true if $a=0$ and $b$ is an ordinary prime of the form $4k+3$.

Added: If $a^2+b^2$ is an ordinary prime, then $a+bi$ is a Gaussian prime. For suppose that $a+bi=(s+ti)(u+vi)$. By a norm calculation we have $a^2+b^2=(s^2+t^2)(u^2+v^2)$. So since $a^2+b^2$ is prime, one of $s+ti$ or $u+vi$ is a unit.

For the other direction, suppose $a+bi$ is a Gaussian prime, where neither $a$ nor $b$ is equal to $0$. We show that $a^2+b^2$ is an ordinary prime. Proof would be easy if we assume standard results that characterize the Gaussian primes. So we try not to use much machinery.

If $a$ and $b$ are both odd, then $a+bi$ is divisible by $1+i$. Then $a+bi$ is an associate of $1+i$, and $a^2+b^2=2$.

So we may assume that $a$ and $b$ have opposite parities. In that case, $a+bi$ and $a-bi$ are relatively prime. For any common divisor $\delta$ must divide $2a$ and $2b$. Since $a$ and $b$ have opposite parity, any common divisor divides $a$ and $b$, so must have norm $1$ if $a+bi$ are prime.

Now suppose that $p$ is a prime that divides $a^2+b^2$. Then $p$ divides $(a+bi)(a-bi)$. Note that $p$ cannot be a Gaussian prime, else it would divide one of $a+bi$ or $a-bi$,

Let $\pi$ be a Gaussian prime that divides $p$. Then $\pi$ divides one of $a+bi$ or $a-bi$. So $\pi$ must be an associate of one of these, and the conjugate of $\pi$ is an associate of the other. Since $a+bi$ and $a-bi$ arer relatively prime, we conclude that $(a+bi)(a-bi)$ divides $p$, which forces $p=a^2+b^2$.


Here is something that is true: $a+bi$ is prime in $\def\Z{\Bbb Z}\Z[i]$ implies that $(a+bi)\Z[i]\cap\Z=p\Z$ for some (ordinary) prime $p$ of $\Z$. This is because, (1) since $a+bi$ is irreducible and $\Z[i]$ is a Euclidean and therefore unique factorisation domain, $(a+bi)\Z[i]$ is a prime ideal of $\Z[i]$, so (2) its intersection with $\Z$ is a prime ideal of $\Z$ (as is always true for the intersection of a prime ideal and a subring), and (3) the intersection is not reduced to$~\{0\}$ because $(a+bi)(a-bi)=a^2+b^2\in\Z\setminus\{0\}$. (Here, like in the question, one does not have "if and only if": here the converse fails for $a+bi$ equal or associated to a prime number $p\not\equiv3\pmod4$, such as $p=2$ or $p=5$; then $(a+bi)\Z[i]\cap\Z=p\Z$, but $p$ and therefore $a+bi$ are composite in $\Z[i]$.)

The case $a+bi$ is prime in $\Z[i]$ splits into two subcases. By the above there exists a prime number $p$ and $z\in\Z[i]$ with $p=(a+bi)z$; then $p^2=N(p)=N(a+bi)N(z)$, and either $z$ is non-invertible, in which case $N(a+bi)=p$ and $z=a-bi$, or $z$ is invertible, in which case $a+bi\in\{p,ip,-p,-ip\}$ and it can be shown that $p\equiv3\pmod4$. Indeed, the irreducibility of $p$ in the UFD $\Z[i]$ means that $\Z[i]/p\Z[i]$ is an integral domain (it is a field), so the kernel $(X^2+1)$ of the ring morphism $(\Z/p\Z)[X]\to\Z[i]/p\Z[i]$ sending $X\mapsto i$ is a prime ideal, so $X^2+1\in(\Z/p\Z)[X]$ is irreducible, which excludes both $p=2$ (for which $X^2+1=(X+1)^2$) and $p\equiv1\pmod4$ (in which case $X^2+1=(X+a)(X-a)$ for some element $a$ of order $4$ in the cyclic group $(\Z/p\Z)^\times$ of order $p-1$.

  • $\begingroup$ The first paragraph does not make clear precisely what is a consequence of said UFD property (that prime ideals are preserved under contraction is true in any ring). $\endgroup$ – Math Gems Apr 21 '13 at 16:20
  • $\begingroup$ @MathGems: The "since ... UFD" in the second sentence was awkwardly placed very early to indicate that that it is needed to justify the immediately following "is a prime ideal". And you are right, pulling back a prime ideal through a ring morphism always gives a prime ideal; maybe that is why they are so useful. $\endgroup$ – Marc van Leeuwen Apr 21 '13 at 16:45
  • $\begingroup$ I surmised what you intended. But I fear that those beginning their studies might have more difficulty inferring the intended meaning. Whenever I encounter things like that I leave comments in the hope that the author might improve the exposition to eliminate ambiguities etc. Thankfully many folks do the same for me when I too do likewise. $\endgroup$ – Math Gems Apr 21 '13 at 17:13
  • $\begingroup$ @MathGems: Fine, thank you. Did I succeed in reducing the ambiguity, or if not what would be better? $\endgroup$ – Marc van Leeuwen Apr 21 '13 at 18:39
  • $\begingroup$ Yes, that reads much more clearly. Thanks and +1. $\endgroup$ – Math Gems Apr 21 '13 at 19:05

Let me add another argument that I believe is simpler. First I need to prove a little proposition.

Let $p \in \mathbb{N}$ be a prime and $N$ be the norm function $N(\alpha) = \alpha \overline{\alpha}$.

If $\alpha \in \mathbb{Z}[\sqrt{n}]$ is prime then $N(\alpha) = p,-p,p^2,-p^2$

Furthermore, if $N(\alpha) = p^2,-p^2$ then $\alpha \sim p$ where $\sim$ is the associates relation on $\mathbb{Z}[\sqrt{n}]$.


If $\alpha$ is prime in $\mathbb{Z}[\sqrt{n}]$, then $N(\alpha) \neq 0,1,-1$ this is because $\alpha$ is not a unit and the fact that $N(\alpha) = \alpha \overline{\alpha} \neq 0$ comes because we are working in an integrity domain so that $\alpha = 0 \lor \overline{\alpha} = 0$ but since $n$ is square-free (by definition) it is necessary that if $\alpha = a+bi$ then $a = b = 0$, therefore $\alpha = 0$ and cannot be prime.

So let $N(\alpha) = \prod p_i$ be its prime factorization over the integers. Then $\alpha|p_i$ since $\alpha$ is prime. Therefore, $\alpha \beta = p \implies N(\alpha) N(\beta) = p^2 \implies N(\alpha) \in \{p,-p,p^2,-p^2\}$ as we wanted. Furthermore, if $N(\alpha) = p,-p^2$ then $N(\beta) = 1,-1$ which means that $\beta$ is a unit and therefore $\beta \sim p$. $$\tag*{$\blacksquare$}$$

So now come to our problem over $\mathbb{Z}[i]$:

$\alpha = a+bi$ is prime in $\mathbb{Z}[i] \simeq p = N(a+bi) = a^2+b^2$ is prime in $\mathbb{Z}$.

$\impliedby)$ Since, $\mathbb{Z}[i]$ is a UFD, $\alpha$ is prime if and only if it is irreducible. But if, $\alpha = \beta \gamma$ is a proper factorization of $\alpha$ then $N(\alpha) = N(\beta) N(\gamma)$ and you would have a proper factorization of $p$. Contradiction.

$\implies)$ $\alpha$ prime implies that $N(\alpha) = p,-p,p^2,-p^2$. Since the norm is positive $N(\alpha) = p,p^2$. If $N(\alpha) = p$ then we are done. If $N(\alpha) = p^2$ then $\alpha \sim p \in \mathbb{N}$ prime. Recalling the units of $\mathbb{Z}[i]$ are $U(\mathbb{Z}[i]) = \{i,-i,1,-1\}$ and multiplying you will see that necessarily $\alpha \in \{p,-p,pi,-pi\}$.

However, you cannot rule out this last case. For instance, take $3$ in $\mathbb{Z}[i]$ which following a norm-argument can be shown to be prime, yet its norm is $p^2 = 9$ which is not prime in $\mathbb{Z}$.

  • $\begingroup$ Could you explain the last line where you claim that $ ab\neq 0 $? Take $ \alpha=0+pi $, then $ a=0, b=p $, but $ ab=0 $ and we still have $ \alpha\sim p $. $\endgroup$ – user549397 Jan 27 at 14:35
  • $\begingroup$ @user549397 you're right it seems i made a mistake, indeed above they are pointing the conjecture is wrong, so let me correct it $\endgroup$ – Javier Jan 27 at 17:17

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.