\documentclass[a4paper]{article} \usepackage{amssymb} \title{An application for hyperbolic numbers} \author{Z.} \begin{document} \maketitle While generating solutions to a Diophantine equation, I stumbled upon a fascinating use for the hyperbolic numbers, one that displays a deep connection to algebraic number theory. To define the hyperbolic numbers, let $j^2=1 \quad (j \ne \pm 1)$. The hyperbolic numbers are then all numbers of the form $a+bj \quad (a,b \in \mathbb{R})$. Note that this is the same as $\mathbb{R}[j]$. The equation I was working on when I discovered this connection was the following: $$a^2+b^2+c^2=d^2 \quad (a,b,c,d \in \mathbb{Z})$$ To generate solutions, I found $(1,4,8,9)$ by hand, then rearranged the equation into $a^2+b^2=d^2-c^2$, which I factored over the complex numbers to get $(a+bi)(a-bi) = (d+c)(d-c)$. This is where I hit a roadblock. The usual method for generating solutions to an equation like this is quite simple. Since $a$ and $b$ are integers, $(a+bi) \in \mathbb{Z}[i]$. Now use the fact that $\mathbb{Z}[i]$ is closed under $(a+bi)^n \quad (1 \le n, n \in \mathbb{Z})$, and that if $(a+bi)^n=u+vi$ then $(a-bi)^n=u-vi$. The same behavior applies to $\mathbb{Z}[\sqrt{k}]$ in general, so $(d+c)^n$ follows the same rules. This allowed me to find that $(a+bi)^n=(a_n+b_ni)$. The issue came when I tried to apply it to the other half: $(d+c)^n=(d_n+c_n)$ does indeed hold; however, there is no separation between the two integer parts, so they combine into the single integer $d_n+c_n$ when we need the pair $(d_n, c_n)$. On its own, this one number is not enough, because there are many ways to reverse an addition ($10=5+5, 6+4, 7+3 \dots$). The information is not truly lost: the other factor gives $(d-c)^n=(d_n-c_n)$, and from the sum and the difference we can recover both parts, $$d_n=\frac{(d+c)^n+(d-c)^n}{2} \qquad c_n=\frac{(d+c)^n-(d-c)^n}{2}$$ But this means carrying two separate integers around and untangling them at the end, where the left side needed only the single number $a+bi$. It is simple to see why this happens: when you factor a difference of squares, you are essentially factoring it over $\mathbb{Z}[\sqrt1]$, and because $\sqrt{1} \in \mathbb{Z}$, it follows that $\mathbb{Z}[\sqrt1]=\mathbb{Z}$. This is where the awkwardness comes from. If only there existed a number which was a square root of 1, but was explicitly \textit{not} an integer. $\mathbb{Z}[j]$ is the direct solution to this problem. In $\mathbb{Z}[j]$, $(d^2-c^2)=(d+cj)(d-cj)$. Unlike with the ordinary integers, however, $(d+cj)^n = (d_n+c_nj)$ does not collapse into a single number, meaning you can easily separate $d_n$ and $c_n$ by taking the real and hyperbolic parts. Now we can try to find nontrivial solutions to the equation. First, let's look at where the first attempt breaks: $$(1+4i)^2 = (-15+8i)$$ $$a_2=-15 \qquad b_2=8$$ $$(9+8)^2=289$$ Uh oh, we have a single number when we actually need a pair of numbers that we can split apart, as with the Gaussian integers. This is where the hyperbolic method finally shines: $$a_2=-15 \qquad b_2=8$$ $$(9+8j)^2=81+64+72j+72j = (145+144j)$$ Wonderful! Now, recalling our original identity, $145+144j=(d_2+c_2j)$, so $$c_2=144 \qquad d_2=145$$ Our new ``Pythagorean quadruple'' is $(-15,8,144,145)$. Now we can check that the method holds (I will replace $-15$ with $15$, since $15^2=(-15)^2$): $$15^2+8^2+144^2=145^2$$ This identity holds! \end{document}