4

My calculator and I were arguing one day about the cosine of some number.

The calculator said "$\cos(\frac x2)=\sqrt{2}+\sqrt{6}$".

I said "That's absurd because $\cos(\frac x2)=\sqrt{\frac{1+\cos(x)}2}$, which evaluates to $2\sqrt{2+\sqrt{3}}$ for this particular $x$!"

So I finally decided to do this the non-radical way and found the decimal approximations to be equivalent.

Which is weird and probably not a coincidence. So why is this, and does this have other values that work out like this?

6 Answers6

9

Just note that $$2\sqrt{2+\sqrt{3}}=\sqrt{8+4\sqrt{3}}=\sqrt{(\sqrt{6})^2+(\sqrt{2})^2+2(\sqrt2)(\sqrt6)}=\sqrt{(\sqrt{2}+\sqrt{6})^2}=\sqrt{2}+\sqrt{6}.$$

Jack Frost
  • 1,246
7

I must confess I've seen a while ago the method that takes care of these problems but failed to appreciated its usefulness, and choose altogether to ignore the result and the method. Till I was put in front of a concrete example that gave me some trouble..

It's the method of "simplifying" nested radicals of the form

$$\sqrt{ a \pm \sqrt{b}}$$.

One tries to write it in the form $$\sqrt{ a \pm \sqrt{b}} = \sqrt{x} \pm \sqrt{y}$$

From experience : in many cases it is much simpler to solve the system that gets $a$ and $b$ then to guess them.

For, by squaring the above we get $$a + \sqrt{b} = x + y + 2 \sqrt{x y}$$ and so we want $$ x+y= a \\ 4 x y = b$$ from where we get $(x-y)^2 = a^2 - b$ and therefore $$x = \frac{a + \sqrt{a^2 - b} }{2} \\ y = \frac{a - \sqrt{a^2 - b} }{2} $$ and so $$\sqrt{ a \pm \sqrt{b}} = \sqrt{ \frac{a + \sqrt{a^2 - b} }{2}} \pm \sqrt{ \frac{a - \sqrt{a^2 - b} }{2}}$$

Hardly anything simpler... except when $a^2 - b$ is a square, and we get the nested radical as a sum/difference of two radicals!

Side note: but what if not? Notice that in general the newly obtained radicals are again nested. What if we apply to each this transformation? Well, one should check that we simply get back the original expression...

Anyhow, whenever one is faced with this kind of nested radicals, there is a method, or a formula, that will take care of it.

Note: This formula also shows how to take square roots of complex numbers.

Obs: What about radicals of the form $\sqrt[3]{a + \sqrt{b}}$? Yes, there is a method even here... It deals with solution of equations of third degree...

orangeskid
  • 53,909
  • Oh, great. This explains the solution to the general equation. Thank you. It even hints at the solution to the third degree equation, which now that you've mentioned it, the solution to the third degree does look like a nested radical. Thank you! – Simply Beautiful Art Dec 23 '15 at 16:28
  • 1
    You probably know this, but for others who might not I'll mention that this method is a standard topic in 1800s algebra textbooks that cover radical manipulations, and there are probably over 100 different such books (from the 1800s) digitized by google that include this topic. However, the topic was slowly written out of such books so that by the early 1930s very few college algebra textbooks included it anymore. – Dave L. Renfro Dec 23 '15 at 16:38
  • @Simple Art: You are very welcome! It turned out to be an inspiring question for me. – orangeskid Dec 23 '15 at 16:38
  • 1
    @DaveL.Renfro Wow. That's kind of disappointing, math being taken out of books. – Simply Beautiful Art Dec 23 '15 at 16:40
  • @DaveL.Renfro: Yes, it used to be in older books. Still find it very interesting. I remember there was once a nested radical $\sqrt[3]{a + c\sqrt{d}}$ that in fact was $p + q \sqrt{d}$ in a non-obvious way...All my Galois theory did not help at that time. – orangeskid Dec 23 '15 at 16:45
  • You might be interested in this 4 December 1999 sci.math post and this 22 December 2006 sci.math post. For a real monster, the identity $\sqrt[3]{16+12\sqrt[3]{7}+9\sqrt[3]{49}} = 2\sqrt[3]{3\sqrt[3]{7}-5}+\sqrt[3]{2+3\sqrt[3]{49}}$ is on p. 13 of Roman Wituła's paper Ramanujan type trigonometric formulas: The general form for the argument $\frac{2\pi}{7}$. – Dave L. Renfro Dec 23 '15 at 17:28
  • @Dave L. Renfro: I see in the first post you mentioned a paper from early 1900's by Salvatore Composto... can't find it on scholar.google or anywhere else... the identities there are neat... as for the Ramanujan one with $\cos$, that seems very tough... the links are great, thanks! And for the one above... oh my.. i'll try it nevertheless – orangeskid Dec 23 '15 at 18:41
  • The first paper is freely available here (I had a typo in the title I posted) and the second paper is freely available here. To be "freely available", you might have to have a google account for google-books. I'm generally loathe to register for anything on the internet (including Facebook, LinkedIn, and many other things) -- the only accounts I have are stackexchange, google, Math Forum, and yahoo email -- but I've found the google account invaluable for my math "research". – Dave L. Renfro Dec 23 '15 at 19:23
  • I almost never use google scholar because virtually everything is behind a paywall that I don't have access to, and besides I generally get better results with a general google search (such as .pdf view of the paper's first page, which is useful for rooting out typos that often show up in other paper's bibliographies). I use a nearby university library for checking out journal volumes to get (photo-)copies of the papers themselves. In the case of older stuff, besides google-books searches, I make very frequent use of the on-line JFM archive. – Dave L. Renfro Dec 23 '15 at 19:28
  • @Dave L. Renfro: Hi! Thanks for the links. For some reason they don't seem to work for me. I will try on some other sites. For now will struggle a bit with some (easier ) Ramanujan stuff from wikipedia page. and then move to the link with the paper from Poland you provided. I had a feeling that anything that can be gotten by summation of radicals is ( in the end) for a trivial reason ( like something being an exact power) but that is not quite so. Interesting... – orangeskid Dec 23 '15 at 23:07
5

Square both expressions.

$$(2\sqrt{2+\sqrt3})^2=4(2+\sqrt3)=8+4\sqrt3$$ $$(\sqrt2+\sqrt6)^2=2+2\sqrt{2\cdot 6}+6=8+4\sqrt3.$$

3

A way of showing this that has not yet been given by anyone is to obtain a "nice equation" (i.e. with integer coefficients) having $\;\sqrt{2}+\sqrt{6}\;$ as a solution and then solving this equation by standard methods.

Start by setting $\;x = \sqrt{2}+\sqrt{6}.\;$ Subtracting $\sqrt 2$ from both sides and then squaring both sides gives $\;x^2 - 2\sqrt{2}x + 2 = 6.\;$ Isolate the radical term and square again, to get $\;x^2 - 4 = 2\sqrt{2}x,\;$ followed by $\;x^4 - 8x^2 + 16 = 8x^2,\;$ or $\;x^4 - 16x^2 + 16 = 0.$

This equation is quadratic in $x^2,$ so using the quadratic formula gives $$x^2 \; =\; \frac{16 \pm \sqrt{256 - 64}}{2} \;=\; \frac{16 \pm 8\sqrt{4 - 1}}{2} \;=\; 4(2 \pm \sqrt{3})$$

Therefore, we get

$$x \; = \; \pm2\sqrt{2 \pm \sqrt{3}}$$

Since $\;\sqrt{2} + \sqrt{6}\;$ is clearly greater than $3$ (note that $\sqrt{2} > 1$ and $\sqrt{6} > 2$), it follows that the value of $x$ that corresponds to $\;\sqrt{2} + \sqrt{6}\;$ is the expression $\;2\sqrt{2 \pm \sqrt{3}}.$

2

HINT:

$$2+\sqrt3=\dfrac{(\sqrt3+1)^2}2$$

1

$$\color{gray}{2+\sqrt 3=\frac 1 2+\sqrt 3+\frac 3 2=\frac{1+2\sqrt3 +3}{2}=\frac{1+2\sqrt 3+(\sqrt 3)^2}{2}=\frac{(1+\sqrt3)^2}{2}}\\2\sqrt{\frac{(1+\sqrt 3)^2}{2}}\\ \color{gray}{\sqrt{\frac 1 2(1+\sqrt 3)^2}=\frac{\sqrt{(1+\sqrt 3)}}{\sqrt 2}}\\ 2\frac{\sqrt{(1+\sqrt 3)^2}}{\sqrt 2}\\ \color{gray}{\sqrt{(1+\sqrt 3)^2}=1+\sqrt 3:}\\ \frac{2}{\sqrt 2}1+\sqrt 3\\ \color{gray}{\frac{2(1+\sqrt 3)}{\sqrt 2}=\frac{2(1+\sqrt 3)}{\sqrt 2}\times \frac{\sqrt 2}{\sqrt 2}=\frac{1(1+\sqrt 3)\sqrt 2}{2}:}\\ \frac{2(1+\sqrt 3)\sqrt 2}{2}\\ \color{gray}{\frac{2(1+\sqrt 3)\sqrt 2}{2}=\frac 2 2\times(1+\sqrt 3)\sqrt 2=(1+\sqrt3)\sqrt 2}\\ (1+\sqrt 3)\sqrt 2\\ \color{gray}{\sqrt 2(1+\sqrt 3)}=\boxed{\color{blue}{\sqrt 2+\sqrt 6}}$$

3SAT
  • 7,512