logo

Gaussian Ring Norm 📂Number Theory

Gaussian Ring Norm

Theorem 1

Consider the function N:Z[i]ZN : \mathbb{Z}[i] \to \mathbb{Z} for the Gaussian Ring Z[i]\mathbb{Z}[i].

  • [1]: If defined as N(x+iy):=x2+y2N(x + iy) := x^2 + y^2, NN becomes the multiplicative norm of Z[i]\mathbb{Z}[i].
  • [2]: Z[i]\mathbb{Z}[i] is a Euclidean domain.
  • [3]: The only unit of Z[i]\mathbb{Z}[i] is 1,1,i,i1,-1,i,-i.

Description

Gaussian integers can be much more comfortably studied with the help of [abstract algebra](../../categories/abstract algebra). By proving [2] with the norm NN defined in an integral domain, it is almost equivalent to proving the fundamental theorem of arithmetic extended to Gaussian primes because ED is UFD. This is similar to explaining the fundamental theorem of arithmetic algebraically as Z\mathbb{Z} is UFD.

Proof

[1]

Definition of the multiplicative norm:

  • (i): N(α)=0    α=0N (\alpha) = 0 \iff \alpha = 0
  • (ii): N(αβ)=N(α)N(β)N ( \alpha \beta ) = N ( \alpha ) N ( \beta )

To demonstrate that NN is a multiplicative norm and that Z[i]\mathbb{Z}[i] actually becomes an ID, it is not initially necessary to have a theorem for defining the norm, so it is unnecessary first to prove that it is an ID. Let’s consider a,b,c,dZa,b,c,d \in \mathbb{Z} as α:=a+ib\alpha := a + ib and β:=c+id\beta := c + id.


Part (i). N(α)=0    α=0N (\alpha) = 0 \iff \alpha = 0

N(α)=a2+b2=0    a=b=0    α=a+ib=0 N(\alpha) = a^2 + b^2 = 0 \iff a = b = 0 \iff \alpha = a + ib = 0


Part (ii). N(αβ)=N(α)N(β)N ( \alpha \beta ) = N ( \alpha ) N ( \beta )

Since the product of Gaussian integers is calculated as (a+ib)(c+id)=(acbd)+i(ad+bc)( a + ib )( c + id) = (ac - bd) + i (ad + bc), N(αβ)=N(acbd+i(ad+bc))=a2c2+a2d2+b2c2+b2d2=(a2+c2)(b2+d2)=N(α)N(β) \begin{align*} N ( \alpha \beta ) =& N(ac - bd + i (ad + bc)) \\ =& a^2 c^2 + a^2 d^2 + b^2 c^2 + b^2 d^2 \\ =& (a^2 + c^2) (b^2 + d^2) \\ =& N ( \alpha ) N ( \beta ) \end{align*}


Part 3.Z[i]Z[i] is ID

Assuming that α\alpha and β\beta are not 0Z[i]0 \in \mathbb{Z}[i] while being αβ=0\alpha \beta = 0 according to Parts (i) and (ii), N(α)N(β)=N(αβ)=N(0)=0 N(\alpha) N(\beta) = N(\alpha \beta ) = N(0) = 0 Since N(α)N(\alpha) and N(β)N(\beta) are elements of the integral domain Z\mathbb{Z}, to satisfy N(α)N(β)=0N(\alpha) N(\beta) = 0, either N(α)N(\alpha) or N(β)N(\beta) must be 00, which contradicts the hypothesis, thus Z[i]Z[i] is also an integral domain.

[2]

Definition of Euclidean norm:

  • (i): For all a,bD(b0)a,b \in D (b \ne 0 ), there exist qq and rr such that a=bq+ra = bq + r is satisfied. In this case, either r=0r = 0 or ν(r)<ν(b)\nu (r) < \nu (b) must be true.
  • (ii): For all a,bD(b0)a,b \in D (b \ne 0 ), ν(a)ν(ab)\nu ( a ) \le \nu ( ab )

Assume ν:=N\nu := N and show that NN is the Euclidean norm of Z[i]\mathbb{Z}[i]. Let’s assume β0\beta \ne 0.


Part (i). σ,ρ:α=βσ+ρ\exists \sigma, \rho : \alpha = \beta \sigma + \rho

Let’s say r,sQr, s \in \mathbb{Q} about some αβ=r+is\displaystyle {{ \alpha } \over { \beta }} = r + is. Let’s set σ\sigma and ρ\rho as close to integer r,sr, s as possible. σ:=q1+iq2ρ:=αβσ \sigma := q_{1} + i q_{2} \\ \rho := \alpha - \beta \sigma

  • Case 1. ρ=0\rho = 0
    Since α=βσ\alpha = \beta \sigma, there’s nothing more to show. Found σ=αβ=q1+iq2\displaystyle \sigma = {{ \alpha } \over { \beta }} = q_{1} + i q_{2} and ρ=0\rho = 0.
  • Case 2. N(ρ)<N(β)N ( \rho ) < N ( \beta )
    According to the definition of q1q_{1} and q2q_{2}, rq112sq212 | r - q_{1} | \le {{1} \over {2}} \\ | s - q_{2} | \le {{1} \over {2}} this leads to N(αβσ)=N[(r+is)(q1+iq2)]=N[(rq1)+i(sq2)](12)2+(12)2=12 \begin{align*} N \left( {{ \alpha } \over { \beta }} - \sigma \right) =& N \left[ (r + is) - (q_{1} + i q_{2}) \right] \\ =& N \left[ (r - q_{1} ) + i ( s - q_{2}) \right] \\ \le & \left( {{1} \over {2}} \right)^2 + \left( {{1} \over {2}} \right)^2 \\ =& {{1} \over {2}} \end{align*} Accordingly, N(ρ)=N(αβσ)=N(β(αβσ))=N(β)N(αβσ)N(β)12N(β) \begin{align*} N ( \rho ) =& N(\alpha - \beta \sigma ) \\ =& N \left( \beta \left( {{ \alpha } \over { \beta }} - \sigma \right) \right) \\ =& N (\beta) N \left( {{ \alpha } \over { \beta }} - \sigma \right) \\ \le & N(\beta) {{1} \over {2}} \\ \le & N (\beta) \end{align*} Thus, we can ensure that ρ\rho and σ\sigma exist to satisfy either ρ=0\rho = 0 or N(ρ)N(β)N(\rho) \le N (\beta).

Part (ii). N(α)N(α)N(β)N ( \alpha ) \ge N ( \alpha ) N ( \beta )

Since β0    N(β)1\beta \ne 0 \implies N ( \beta ) \ge 1, N(α)N(α)1N(α)N(β)=N(αβ) \begin{align*} N ( \alpha ) \le & N ( \alpha) \cdot 1 \\ \le & N(\alpha) N(\beta) \\ =& N ( \alpha \beta ) \end{align*}

[3]

According to the properties of a multiplicative norm, if uZ[i]u \in \mathbb{Z}[i] is a unit, then N(u)=1| N ( u ) | = 1. Therefore, by contraposition, if N(u)=1| N(u) | = 1 is not true, then uu is not a unit. The only case that satisfies N(u)=x2+y2=1N(u) = x^2 + y^2 = 1 for u:=x+iyu := x + iy is when it is u{1,1,i,i}u \in \left\{ 1, -1, i, -i \right\}.


  1. Silverman. (2012). A Friendly Introduction to Number Theory (4th Edition): p272. ↩︎