5

Let the multiplication graph $n:m$ be the graph with $m$ points equally distributed on a circle and a line between points $a$ and $b$ when $n\cdot a \equiv b\operatorname{mod} m$.

Looking at the multiplication graph $3:64$ with coprime $n, m$ reveals not so much at first sight – except of the ($n- 1$)-foil pattern that is to be expected for every "nominator" $n$:

enter image description here

But when highlighting permutation cycles – the shorter the stronger – something interesting appears:

enter image description here

What one sees are two $4$-cycles $(4,12,36,44)$ and $(20,60,52,28)$ which describe two perfect rectangles with integral side lengths $8 : 24 = 1 : 3$

I wonder which properties of $n$ and $m$ are responsible for the appearance of two such rectangles and their corresponding side lengths? Is it by sheer coincidence that $8$ is the square root of $64$?

  • What software are you using? Can you share your code? It may help others to better answer your questions and collaborate with you. Did you look at Dan Shanks's book yet? – Bill Dubuque Dec 12 '18 at 15:30
  • @BillDubuque: I've written the software by myself - quick and dirty and not in a shape to share, alas. But hopefully I'll release an online tool you can play around with. If you personally are interested in it and like to be a beta tester, send an email to stricker@syspedia.de. Shanks is still waiting, he comes next. Thanks again for the hint! – Hans-Peter Stricker Dec 12 '18 at 15:40
  • @BillDubuque: If you are interested: the picture in this question was also created with my software. – Hans-Peter Stricker Dec 12 '18 at 15:43
  • What programming language and/or software are you currently using? – Bill Dubuque Dec 12 '18 at 15:48
  • @BillDubuque: It's just plain Javascript and d3.js, no other libraries - everything is done by the browser. – Hans-Peter Stricker Dec 12 '18 at 15:49

3 Answers3

2

Note that $3^4=81\equiv 17 \pmod {64}$ so multiplying a number by $3$ four times is equivalent to multiplying it by $17$. Now $4 \cdot 17 \equiv 4 \pmod {64}$ so multiplying any multiple of $4$ by $3$ four times will bring us back and we will have at most a $4-$cycle. The other cycle is just the negative of the first because $60 \equiv -4 \pmod {64}$. You have accounted for all the numbers that are equivalent to $4 \pmod 8$ in our system. The heavy lines in your figure connect the multiples of $8$ except $0,32$ in pairs because $3^2 \cdot 8 \equiv 8 \pmod {64}$ so multiplying a multiple of $8$ twice by $3$ will bring us back. Finally $0,32$ are the solutions to $3x \equiv x \pmod {64}$ so they are $1-$cycles.

Ross Millikan
  • 383,099
  • To sum it up: We have $3^4\cdot 2^2 \equiv 2^2\pmod {2^6}$, $3^2\cdot 2^3 \equiv 2^3\pmod {2^6}$, $3^1\cdot 2^5 \equiv 2^5\pmod {2^6}$. What's the general pattern? Why are equations with $3^3$ and $2^4$ missing? Is it essential that $2$ and $3$ are primes, or is it essential that $2 = 3-1$? I cannot see. – Hans-Peter Stricker Dec 12 '18 at 08:46
  • $3^3 \cdot 2^4 \equiv 3 \cdot 2^4 \pmod {2^6}$. How does this fit? – Hans-Peter Stricker Dec 12 '18 at 08:56
  • $3^3$ is missing because there is nothing special relating $27$ and $64$. I don't think $2,3$ being prime matters, it is about multiplicative order of $3$ modulo factors of $64$. $3^2$ and $3^4$ are both one more than a factor of $64$ – Ross Millikan Dec 12 '18 at 14:17
2

Hint $ $ It's simply $\,3^{\large 4}\equiv 1\pmod{\!16}\,$ multiplied by $\,\color{#0a0}4,\,$ i.e.

$\qquad\qquad \begin{align} &\bmod 16\!:\ \ \ \ \ \ 3^n = \color{#c00}1,\ \ 3,\ \ 9,\,11,\, \color{#c00}1\ \ldots\\ \Rightarrow\ &\bmod 64\!:\ \color{#0a0}4\cdot 3^n = \color{#0a0}4,12,36,\,44,\ \color{#0a0}4\ \ldots \end{align}$

i.e. $\,\ 4\cdot 3^{\large k+4n}\!\bmod 64 = 4\,(3^{\large k}81^{\large n}\!\bmod 16) = 4\,(3^{\large k}1^{\large n}\!\bmod 16) = 4\cdot 3^{\large k}\bmod 64$

The second one is its negative, i.e. using $\,\color{#0a0}{{-}4}\,$ vs. $\,\color{#0a0}4\,$ i.e.

$\qquad\qquad \begin{align} &\bmod 16\!:\ \ \ \ \ \ \ \ \ 3^n = \ \ \ \color{#c00}1,\ \ \ \ \ 3,\ \ \ \ \ 9,\ \ \ \ 11,\ \ \ \ \color{#c00}1\ \ldots\\ \Rightarrow\ &\bmod 64\!:\, \color{#0a0}{{-}4}\cdot 3^n = \color{#0a0}{{-}4},-12,-36,\,-44,\ \color{#0a0}{{-}4}\ \ldots\\ &\qquad\qquad\qquad\quad\ \ \equiv\ 60,\ \ \ 52,\ \ \ 28,\ \ \ \ 20 \end{align}$

Bill Dubuque
  • 282,220
0

Note that in order to form a rectangle of the points ($p, p\times q, p\times q^2, p \times q^3$) to form a "rectangle" we need the following to occur. First, $p \times (q^2-1) \equiv \frac{m}{2} \mod m$, in order for the first and third elements to be diametrically opposite (and this the angle that forms that connects them (specifically at the $p \times q$ point) would be a right angle. Note this implies $m$ must be even, as there isn't a diametrically opposite point in an odd modulus. Note this must be true for the second and fourth elements as well, so $pq \times (q^2-1) \equiv \frac{m}{2} \mod m$. Finally, we need this to actually be a cycle of $4$, so $q^4 \equiv 1 \mod m$. From the first and second equations we essentially get that $q$ must be odd (in order for the multiplication of $q$ from the first to the second to not change the remainder mod $m$). Additionally, we need $m$'s totient function to either be $2$ or a multiple of $4$ in order for there to be solutions to the last question. Then, we can construct rectangles.

For an $m$, consider all the solutions to $q^4 \equiv 1 \mod m, q \neq 1$. Note $q$ must be odd. Now, note that $q^2-1 \equiv 0 \mod 4$, so the only way that this could be remainder of half a modulus is if the modulus is a divisor of 8. Assuming all of this is true, we now have $$2 * p \times (q^2-1) \equiv 0 \mod m,$$ now, we just find the solutions to this, and then ensure that $p \times (q^2-1) \equiv \frac{p}{2} \mod m $ and not $0 \mod m$, and then we have our solutions.

EDIT: To answer your second question, note that you are essentially asking to solve $$x \times (pq-p) \equiv (pq^2-pq) \mod m, $$ a solution to which is $q$.