11
$\begingroup$

I know there is a nice way of getting the continued fraction expansion of quadratic irrationals mainly because they recur after a point, and if they recur after a point they are quadratic irrationals. When constructing the expansion you can multiply by conjugates (kind of), e.g.

$\sqrt 3 =1+\sqrt 3 -1 = 1+\frac {1}{\frac {\sqrt 3 +1}{2}} $

Where you use $(\sqrt 3 - 1)(\sqrt 3 +1)=2$.

Are there identities that would help with the construction for $ \sqrt[3]{2} $?

One I thought was useful in the first step to get [1; 3,...] was

$ (\sqrt[3]{2}-1)( \sqrt[3]{4} + \sqrt[3]{2}+1 )=1$,

So you get:

$ \sqrt[3]{2}=1+( \sqrt[3]{2}-1 )=1+\frac {1}{ \sqrt[3]{4} + \sqrt[3]{2}+1 }= 1+\frac {1}{3+ (\sqrt[3]{4} + \sqrt[3]{2}-2)} $

Thanks for the help.

$\endgroup$
  • 2
    $\begingroup$ As far as I am aware very little is known about the continued fraction of any algebraic number with degree greater than $2$. However your computations can be simplified by working with polynomials, starting with the fact that $\root3\of2$ is a root of $x^3-2$. I have an exposition using this very example here starting at page 80. $\endgroup$ – David May 3 '14 at 13:21
  • $\begingroup$ @David Yeh, that was mentioned in our number theory class. Thanks for the link to the ps. It's really helpful (+1). Its pretty much an answer, that polynomial method and the identity that comes after it, for the more complicated way. Thanks for the post. $\endgroup$ – snulty May 3 '14 at 17:43
  • $\begingroup$ There are rules to derive CF for higher roots. For more info look at this Wikipedia page: en.wikipedia.org/wiki/Generalized_continued_fraction $\endgroup$ – Stefan Gruenwald Sep 9 '17 at 3:34
5
$\begingroup$

Starting from the column vector $(1,0,0,-2)$, consider the following steps:

Step a) Repeat multiplication by the matrix $A$ $$A=\begin{bmatrix} 1&0&0&0\\ 3&1&0&0\\ 3&2&1&0\\ 1&1&1&1 \end{bmatrix}$$ while the coefficients of the resulting vector have different signs.

Step b) Reverse the coefficients of the vector, or equivalently multiply by $$B=\begin{bmatrix} 0&0&0&1\\ 0&0&1&0\\ 0&1&0&0\\ 1&0&0&0 \end{bmatrix}$$

Then the number of times you multiply by $A$ in step a gives the partial quotients of continued fraction of $\sqrt[3]{2}$.


For, starting from $(1,0,0,-2)$, successive multiplication by $A$ gives: \begin{align} (1,0,0,-2) &\xrightarrow A\color{red}{(1,3,3,-1)}\\ &\xrightarrow A(1,6,12,6) \end{align} hence in step a we multiply by $A$ one time only, because $(1,6,12,6)$ have positive coefficients only, hence the first partial quotient is $1$: $$\sqrt[3]{2}=1+\cdots$$

Apply step b to $(1,3,3,-1)$ we get $(-1,3,3,1)$. Then applying step a to $(-1,3,3,1)$, successive multiplication by $A$ gives: \begin{align} (-1,3,3,1) &\xrightarrow A(-1,0,6,6)\\ &\xrightarrow A(-1,-3,3,11)\\ &\xrightarrow A\color{red}{(-1,-6,-6,10)}\\ &\xrightarrow A(-1,-9,-21,-3)\\ \end{align} hence the second partial quotient is $3$: $$\sqrt[3]{2}=1+\frac 1{3+}\cdots$$ and so on...


This algorithm holds for every algebraic number of third degree which is the only positive root of it minimal polynomial. For higher degree the matrix $A$ is enlarged as in the Tartaglia-Pascal triangle; for example for fourth degree: $$A=\begin{bmatrix} 1&0&0&0&0\\ 4&1&0&0&0\\ 6&3&1&0&0\\ 4&3&2&1&0\\ 1&1&1&1&1 \end{bmatrix}$$


For the intuition behind this algorithm. Then vector $(1,0,0,-2)$ corresponds to the polynomial $p(x)=x^3-2$. Multyplication by $A$ corresponds to $p(x)\mapsto p(x+1)$, while revesing in step b corresponds to $p(x)\mapsto x^3p(1/x)$. Finally Descartes's signs rule provide the stopping criterion in step a.

$\endgroup$
  • $\begingroup$ So just to see if I understand a bit why this works. $(1,0,0,-2)$ is representing $1x^3+0x^2+0x-2$. Then $A$ and $B$ are some kind of linear maps on this polynomial ring/vector space. $A$ looks like $p\mapsto p+p’+p’’/2!+p’’’/3!$, which is a kind of Taylor expansion, while B I suppose can be seen like the map $p(x)->x^3*p(1/x)$. I have no idea why this works though. Is there some intuition for it? It seems really cool. $\endgroup$ – snulty Oct 26 '19 at 10:37
  • $\begingroup$ @snulty: I added explanation in answer, thank'you $\endgroup$ – Fabio Lucchini Oct 26 '19 at 11:09
  • $\begingroup$ @FabioLucchini Can this method be used to bound the continued fraction coefficients of $\sqrt[3]{2}$? $\endgroup$ – Franklin Pezzuti Dyer Nov 10 '19 at 18:49
  • $\begingroup$ @franklin I am not sure the ciefficients are bounded in the first place. $\endgroup$ – Oscar Lanzi Nov 10 '19 at 21:57
  • $\begingroup$ @OscarLanzi No, they’re probably not bounded, but I meant bounding them using something like big-O notation (showing that they have sub-quadratic growth, or sub cubic growth, or sub-exponential growth, or something like that). $\endgroup$ – Franklin Pezzuti Dyer Nov 10 '19 at 22:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.