Let δ=(1+19)/2\delta = (1 + \sqrt{-19})/2 and let R=Z[δ].R = \mathbb{Z}[\delta]. Define norm N(a+b19)=a2+19b2.N(a + b\sqrt{-19}) = a^2 + 19b^2.

Lemma L2.2

N(a+bδ)=a2+ab+5b2N(a + b\delta) = a^2 + ab + 5b^2

Proof

N(a+bδ)=N(a+b1+192)=N((a+b2)+b219)=(a+b2)2+19(b2)2=a2+ab+b24+19b24=a2+ab+5b2\begin{align*} N(a + b\delta) &= N\left(a + b\frac{1+\sqrt{-19}}{2}\right) \\ &= N\left(\left(a + \frac{b}{2}\right) + \frac{b}{2}\sqrt{-19}\right) \\ &= \left(a + \frac{b}{2}\right)^2 + 19\left(\frac{b}{2}\right)^2 \\ &= a^2 + ab + \frac{b^2}{4} + \frac{19b^2}{4} \\ &= a^2 + ab + 5b^2 \\ \end{align*}

\blacksquare

Proposition A

The 44 least values taken by N(a+bδ)N(a + b\delta) are 0,1,4,0, 1, 4, and 5.5. Moreover, N(a+bδ)5N(a + b\delta) \geq 5 if b0.b \neq 0.

Proof

Let f(a,b)=a2+ab+5b2.f(a,b) = a^2 + ab + 5b^2. For a fixed b,b, the extremum of this parabola is at a=b/2.a = -b/2. We use this to find a lower bound for the value of f(a,b).f(a,b).

f(b2,b)=(b2)2+(b2)b+5b2=b242b24+20b24=19b24\begin{align*} f\left(-\frac{b}{2}, b\right) &= \left(-\frac{b}{2}\right)^2 + \left(-\frac{b}{2}\right)b + 5b^2 \\[1em] &= \frac{b^2}{4} - \frac{2b^2}{4} + \frac{20b^2}{4}\\[1em] &= \frac{19b^2}{4} \end{align*}

Now, if b=±1,b = \pm 1, then f(a,b)f(a,b) is at least 4.75,4.75, and since f(a,b)f(a,b) is an integer for integer values of a,a, it must be greater than or equal to 5.5.

Observe that the lower bound grows monotonically with b.|b|. This forces f(a,b)5f(a,b) \geq 5 whenever b1|b| \geq 1 and a,bZ.a,b \in \mathbb{Z}. Hence, N(a+bδ)5N(a + b\delta) \geq 5 whenever a,bZa,b \in \mathbb{Z} and b0.b \neq 0.

We observe that f(a,0)=a2,f(a, 0) = a^2, and the minimal integer values of a2a^2 are 0,1,4,9, .0, 1, 4, 9, \ldots \ .

Observe also that f(1,1)=5,f(1,-1) = 5, attaining the lower bound for non-zero integer b.b.

Combining our results, we see that indeed, the minimal values attained by f(a,b)f(a,b) with integer arguments (and hence by N(a+bδ)N(a + b\delta)) are 0,1,4,5.0,1,4,5.

\blacksquare

Proposition B

The units of RR are ±1.\pm 1.

Proof

Analogous to the arguments given in 1.17A3 and 1.17A4.

\blacksquare

Proposition C

Suppose there exists a non-zero non-unit element cRc \in R such that every aRa \in R can be written as a=qc+ra = qc + r for some q,rRq, r \in R with r=0r = 0 or rR×.r \in R^{\times}. Then cc divides 22 or 3,3, and thus is equal to ±2\pm 2 or ±3.\pm 3.

Proof

Consider 2=qc+r,2 = qc + r, which can be written as 2r=qc.2 - r = qc. If r=0,r = 0, then c2.c \mid 2. If r=1,r = -1, then c3.c \mid 3. Since cc is not a unit, it cannot be the case that r=1,r = 1, as that would force 1=qc.1 = qc.

Observe that if 4=N(2)=N(qc)=N(q)N(c),4 = N(2) = N(qc) = N(q)N(c), then N(q)=1N(q) = 1 necessarily.

Observe that if 9=N(3)=N(qc)=N(q)N(c),9 = N(3) = N(qc) = N(q)N(c), but N(x)N(x) cannot be equal to 3,3, so N(c)=9,N(c) = 9, and hence N(q)=1.N(q) = 1.

This forces c{±2,±3}c \in \{\pm 2, \pm 3\}

\blacksquare

Proposition D

There does not exist qRq \in R such that δ=qc+r\delta = qc + r with c{±2,±3}c \in \{\pm 2, \pm 3\} and r{0,±1}.r \in \{0, \pm 1\}.

Proof

Let q=a+δb.q = a + \delta b. Then N(qc)=c2(a2+ab+5b2).N(qc) = c^2 \cdot (a^2 + ab + 5b^2). If c=±2,c = \pm 2, then N(qc)N(qc) is a multiple of 4.4. If c=±3,c = \pm 3, then N(qc)N(qc) is a multiple of 9.9.

Observe that N(δ)=N(δ1)=5,N(\delta) = N(\delta - 1) = 5, N(δ+1)=7.N(\delta + 1) = 7.

So, if δ=qc+r\delta = qc + r held, then we would have δr=qc,\delta - r = qc, which requires N(δr)=N(qc).N(\delta - r) = N(qc). But N(δr){5,7}N(\delta - r) \in \{5,7\} is prime, while N(qc)N(qc) is composite. This is an impossibility.

\blacksquare

Proposition E

RR is not a Euclidean domain.

Proof

We know from 2.17 that a Euclidean domain must have an element cc such that the remainder of Euclidean division by cc is either a unit or 0.0. We have demonstrated in Proposition C and Proposition D that no such element exists in R.R. Therefore, RR is not a Euclidean domain.

\blacksquare