9

This identity, which was shared on math.stackexchange and seem curious at first sight, caught my attention. Here is the equality:

$$\color {red} {\dfrac{1646-736\sqrt{5}} {2641-1181\sqrt{5}} =\color{blue}{\dfrac{17+15\sqrt{5}} {7+15\sqrt{5}}}}$$

The truth is that even the almost all students can easily prove this.

But, what I'm wondering about isn't the proof that actually uses a cross multiplication.

Goal : Knowing where these numbers come from

Maybe, here something is hiding from us. Has a complicated fraction been simplified to an easy fraction?

I have reasons not to believe this.

I will try to show that the numbers are not randomly chosen and moreover, where they come from.

Let, $$m,n,r,s \in\mathbb {Z}.$$

We have:

$$\frac{m-n\sqrt{5}} {r-s\sqrt{5}} =\frac{17+15\sqrt{5}} {7+15\sqrt{5}}$$

$$\implies ( r-s\sqrt{5})(17+15\sqrt{5})=(m-n\sqrt{5})(7+15\sqrt {5})$$

$$\implies (17r-75s)+\sqrt {5}(15r-17s)=(7m-75n)+\sqrt {5}(15m-7n)$$

Then we have,

$$\begin{cases} 15r-17s=15m-7n \\ 17r-75s=7m-75n \end{cases}$$

There are probably good methods for solving such equations in number theory. But since I don't know, I usually solve such equations by methods I know myself.

$$\begin{cases} 255r-289s= 255m-119n \\ 255 r -1125s =105 m - 1125 n \\ 1125r-1275s=1125 m - 525 n \\ 289r-1275s= 119 m - 1275 n \end{cases}$$

$$ \implies \begin{cases} r=\dfrac {503m+375n}{418} \in\mathbb Z \\ s=\dfrac {75m+503n}{418} \in\mathbb Z \end{cases}$$

First restriction

$$\begin {align} r=\dfrac {503m+375n}{418} \in\mathbb Z \Longrightarrow r=\dfrac {418m+85m+418n-43n}{418} \Longrightarrow \dfrac{85m-43n}{418}=k_1 \Longrightarrow m=\dfrac{418k_1+43n}{85}=\dfrac{425k_1-7k_1+43n}{85} \Longrightarrow \dfrac {43n-7k_1}{85}=k_2 \Longrightarrow k_1=\dfrac{42n-84k_2+(n-k_2)}{7} \Longrightarrow \dfrac {n-k_2}{7}=k_3 \Longrightarrow n=7k_3+k_2 \end {align} $$

$$\implies k_1=\dfrac{43(7k_3+k_2)-85k_2}{7}=43k_3-6k_2 $$

$$\implies m= \dfrac { 418(43k_3-6k_2)+43(7k_3+k_2)}{85}=215k_3-29k_2$$

$$\implies n=7k_3+k_2$$

$$\implies r=\dfrac {503m+375n}{418} =\dfrac {503(215k_3-29k_2)+375(7k_3+k_2)}{418}=265k_3-34k_2, ~ \text {where} ~ k_2, k_3 \in\mathbb Z.$$

$$\color {gold}{\boxed {\color{black} {r=265k_3-34k_2}}}$$

Second restriction

$$\begin{align} s=\dfrac {75m+503n}{418} \in\mathbb Z \Longrightarrow s=\dfrac {75(215k_3-29k_2)+503(7k_3+k_2)}{418}=47k_3-4k_2, ~ \text {where} ~ k_2, k_3 \in\mathbb Z. \end{align}$$

$$\color {gold}{\boxed {\color{black} {s=47k_3-4k_2}}}$$

Thus, no new restrictions are required.

Here for simplicity, I am changing $k_2 = v, k_3 = u$, where $u,v\in\mathbb Z$. Then we have,

$$\color {gold}{\boxed {\color{black} {\begin {align} r=265u-34v\\ s=47u-4v \\m=215u-29v \\n=7u+v \end{align}}}}$$

Now we have to solve any of the equations we obtain. Therefore, solving the first equation will be sufficient:

$$r=265u-34v.$$

$$\begin{align} v=\dfrac{265u-r}{34}=8-\dfrac{7u+r}{34}\Longrightarrow \dfrac {7u+r}{34}=k_4 \Longrightarrow u=5k_4-\dfrac{k_4+r}{7} \Longrightarrow k_4+r=7k_5 \Longrightarrow k_4=7k_5-r, ~ \text{where} ~ k_4, k_5 \in\mathbb Z \end{align}$$

$$\implies u=34k_5-5r$$

$$\implies v=\dfrac {265(34k_5-5r)-r}{34}=265k_5-39r$$

For simplicity, I accept $k_5 \longmapsto k$. So, we have

$$\begin{align} u=34k-5r \\ v=265k-39r \end{align}$$

Finally we have,

$$\color {gold}{\boxed {\color{black} {\begin{align} \color{blue}{r=r} \\ \color {red}{s=-79 r+538k} \\ \color{green} {m=56 r- 375 k} \\ \color{blueviolet}{n=- 74 r+503k} \end{align}}}}$$

Let's go back to the original problem.

If $r=2641$, then we have

$$\begin{align} r=2641 \\ s=-208639+538k \\ m=147896- 375 k \\ n=−195434+503k \end{align}$$

Now we can choose $k$ as we wish.

Let, $k=390$, then we have

$$\begin {align} m=1646 \\ n=736 \\ r=2641 \\ s=1181.\end{align}$$

Thus, we obtain the required equality:

$$\color {red}{\boxed {\color{black} {\dfrac{1646-736\sqrt{5}} {2641-1181\sqrt{5}} =\dfrac{17+15\sqrt{5}} {7+15\sqrt{5}}}}}$$

And now here we can make the numbers as difficult as we want. Let's try one example.

If $m,n,r,s \in\mathbb {Z^{+}}$, then we must solve the following system of equations:

$$\begin{cases} r>0 \\ -79 r+538k>0 \\ 56 r- 375 k>0 \\- 74 r+503k >0 \end{cases} \Longrightarrow k>0, \dfrac {375k}{56}<r<\dfrac {503k}{74}$$

Let, $k=17639735$ and $r=118123483$ ( a prime number), then we have

$$\begin{align} r=118123483 \\ s = 158422273 \\m=14423 \\ n=131648963 \end{align}$$

Finally, our new "magical" identity will be as follows:

$$\color {gold}{\boxed {\color{blue} {\dfrac{14423-131648963\sqrt{5}} {118123483-158422273\sqrt{5}} \color{red}{=\dfrac{17+15\sqrt{5}} {7+15\sqrt{5}}}}}}$$


Short proof

$$\dfrac{1646-736\sqrt{5}} {2641-1181\sqrt{5}} =\dfrac{17+15\sqrt{5}} {7+15\sqrt{5}}$$

$$\Longleftrightarrow \dfrac{445\sqrt 5-995} {1646-736\sqrt{5}} =\dfrac{10} {17+15\sqrt{5}}$$

$$\Longleftrightarrow \dfrac{89\sqrt 5-199} {1646-736\sqrt{5}} =\dfrac{2} {17+15\sqrt{5}}$$

$$ \Longleftrightarrow \dfrac{1845-825\sqrt 5} {1646-736\sqrt{5}} =\dfrac{15+15\sqrt{5}} {17+15\sqrt{5}}$$

$$\Longleftrightarrow \dfrac{123-55\sqrt 5} {1646-736\sqrt{5}} =\dfrac{1+\sqrt 5} {17+15\sqrt{5}}$$

$$\Longleftrightarrow \dfrac {1646-736\sqrt{5}} {123-55\sqrt 5}=\dfrac {17+15\sqrt{5}}{1+\sqrt 5}=15+\dfrac{2}{1+\sqrt 5}$$

$$\Longleftrightarrow \dfrac{89\sqrt 5-199} {123-55\sqrt{5}} =\dfrac{2} {1+\sqrt{5}}$$

$$\Longleftrightarrow \dfrac {246-110\sqrt 5}{89\sqrt 5-199} =1+\sqrt 5$$

$$\Longleftrightarrow \dfrac {246-110\sqrt 5-89\times 5+199 \sqrt 5}{89\sqrt 5-199}=\dfrac {89\sqrt 5-199}{89\sqrt 5-199}=1.$$

Observation

  • We always have the chance to choose at least one of the numbers $m, n, r, s$ as a prime number. According our formulas, we have

$$\color {red}{\boxed {\color{black} {\begin{align} r=r \\ s=-79 r+538k \\ m=56 r- 375 k \\ n=- 74 r+503k \end{align}}}}$$

So, we can choose the number $\color{blue}{r}$ always as a prime.

Let's look at the following equality, which is completely randomly selected.

$$\color {red} {\dfrac{179277361-8263\sqrt{5}} {9273317-5193\sqrt{5}} =\color{blue}{\dfrac{a+b\sqrt{5}} {c+d\sqrt{5}}}}$$

We have no good reason for solving this equation that "small" values ​​of $ a, b, c, d$ exist.

Therefore, I came to the following conclusion.

The expression $\color {red}{\dfrac{17+15\sqrt{5}} {7+15\sqrt{5}}}$ is not taken from the expression $\color{blue}{\dfrac{1646-736\sqrt{5}} {2641-1181\sqrt{5}}}.$ On the contrary, the expression $\color{blue}{\dfrac{1646-736\sqrt{5}} {2641-1181\sqrt{5}}}$ is taken from the expression $\color{red}{\dfrac{17+15\sqrt{5}} {7+15\sqrt{5}}}.$

Generalization:

For any $p\in\mathbb Z^{+}$ and small $a,b,c,d$, there exist infinitely many enough large pozitive integers $m,n,r,s$ (we can choose at least one as a prime number) such that,

$$\frac{m-n\sqrt{p}} {r-s\sqrt{p}} =\frac{a+b\sqrt{p}} {c+d\sqrt{p}}$$

Questions

  • Could these numbers have come from a completely different place that has nothing to do with the method I use?

  • Does mathematics allow for the techniques I have applied above? Are the things I've done correct?

Thank you.

lone student
  • 14,709
  • 3
    Why not start with the origin of these numbers on MSE? https://math.stackexchange.com/questions/1092215/how-was-this-approximation-of-pi-involving-sqrt5-arrived-at/2285072#2285072 – Jean-Claude Arbaut Feb 02 '21 at 19:49
  • 1
    @Jean-Claude Arbaut I can't get out of school knowledge much.. – lone student Feb 02 '21 at 19:55
  • 2
    I can understand that, but you asked: "Could these numbers have come from a completely different place that has nothing to do with the method I use?" And the answer seems to be yes. Another suggestion, maybe (but I didn't explore this), would be to look at the arithmetic in $\Bbb Z[\sqrt5]$ (not high school stuff, but not that complicated). – Jean-Claude Arbaut Feb 02 '21 at 20:02
  • 1
    @JeanClaudeArbaut Thank you. Well, now I understood. – lone student Feb 02 '21 at 20:03
  • 5
    If you wish to deeply understand this and related matters you should study (algebraic) number theory of quadratic number fields. In this case the algebraic integers in $\Bbb Q(\sqrt 5),$ are a UFD so their arithmetic is similar to that of $\Bbb Z,,$ but complicated by the presence of nontrivial units. – Bill Dubuque Feb 02 '21 at 20:07
  • 1
    @BillDubuque I cannot imagine I will have such a chance. I'm grateful for you saying what it's all about. – lone student Feb 02 '21 at 20:17
  • 1
    You can generate infinitely many such fractions simply by multiplying by $1$, if you write it as $\frac{a+b\sqrt{5}}{a+b\sqrt{5}}$ and expand the numerator and denominator. For example, for $a=-2,b=1$ this gives $\frac{17+15\sqrt{5}} {7+15\sqrt{5}}=\frac{41-13\sqrt{5}}{61-23\sqrt{5}}$. Choosing slightly bigger numbers $a=-100,b=51$ we get $\frac{17+15\sqrt{5}} {7+15\sqrt{5}}=\frac{2125-633\sqrt{5}}{3125-1143\sqrt{5}}$, and so on... But I guess that is not enough "magical" :). – Sil Feb 02 '21 at 20:30
  • 1
    @Sil Maybe. Of course, I did not find this equality myself.. I don't call something magical anyway if found it myself :) I just thought about it.. – lone student Feb 02 '21 at 20:38
  • 5
    The origin of this ratio is entirely due to Ramanujan (with details in my answer linked by @Jean-ClaudeArbaut). The non-trivial part however is the observation that complicated expressions like these can be simplified to an expression which involves small numbers. Try to imagine the calculations in my answer being done by a man in colonial India of 1910-1914 using slate and chalk. Nothing beats it. – Paramanand Singh Feb 03 '21 at 02:21
  • 1
    @ParamanandSingh Can you simplify the example I gave in my question? – lone student Feb 03 '21 at 02:38
  • I don't think we can randomly take large integers $p, q, r, s$ and hope that $(r-s\sqrt {5})/(p-q\sqrt{5})$ can be simplified into a form involving small numbers. – Paramanand Singh Feb 03 '21 at 03:08
  • @ParamanandSingh Yes, this is exactly what I said in the observation part. We can always convert small numbers to large numbers or even prime numbers. But we cannot always do otherwise. – lone student Feb 03 '21 at 03:12
  • 1
    But you know it's also $=\frac{29+\sqrt{5}}{34-4\sqrt{5}}$? As several people have said, this boils down to arithmetic in the ring of integers of $\mathbb Q(\sqrt5)$, which is a cool motivation to learn algebraic number theory! Basically, you have to find prime factors of the numerator and denominator, and after that, you can expand as much as you want. Have you noticed that e.g. for the denominators $a+b\sqrt5$, we have that $a^2-5b^2$ is divisible by $269$? That's no coincidence! It's because your original denominator lies above that prime. – Torsten Schoeneberg Feb 03 '21 at 04:47
  • @TorstenSchoeneberg Are the $m,n,r,s$ values ​​I found wrong?Is the method I applied correct?I believed such things were just algebra precalculus.. – lone student Feb 03 '21 at 06:27
  • 2
    I have not checked all your calculations, but as far as I can tell, your formula for $r,s$ indeed gives that $r^2-5s^2$ is always divisible by $269$, and that is consistent with what algebraic number theory tells me. I'm just saying maybe you are interested in learning that, which will give more understanding of what is going on here, and maybe help avoid some of those calculations. (It's GREAT you do those calculations, but higher math is all about avoiding calculations ...) – Torsten Schoeneberg Feb 03 '21 at 07:20
  • 1
    Amazing work, although I did not take the time to read it – Some Guy Feb 04 '21 at 03:56
  • @TorstenSchoeneberg: unfortunately higher math has taken a highly uninspiring approach to avoid calculations and that is use of math software. The focus should have been to develop theories which reduce calculations possible via hand. – Paramanand Singh Feb 05 '21 at 04:15
  • I posted an answer to my question which gives another way: https://math.stackexchange.com/a/4013577/72031 – Paramanand Singh Feb 05 '21 at 15:13

1 Answers1

4

We go from the blue fraction to the red one by multiplying the numerator & denominator by $\tfrac{-199+89\sqrt{5}}{2}$. More generally $\tfrac{17+15\sqrt{5}}{7+15\sqrt{5}}=\tfrac{(17+15\sqrt{5})(a+b\sqrt{5})}{(7+15\sqrt{5})(a+b\sqrt{5})}=\tfrac{17a+75b+(15a+17b)\sqrt{5}}{7a+75b+(15a+7b)\sqrt{5}}$. Now take $a=-\tfrac{199}{2},\,b=\tfrac{89}{2}$. These satisfy $a^2-5b^2=-1$, but we can do the above with arbitrary $a,\,b$.

J.G.
  • 115,835