MHB For j >1 the first 2^j digits are correct

  • Thread starter Thread starter mathmari
  • Start date Start date
Click For Summary
The discussion centers on the iteration formula \(x_{j+1}=2x_j-3x_j^2\) with a starting point of \(x_0=0.3\) to approximate \(\frac{1}{3}\). It is demonstrated that for \(j \geq 1\), the first \(2^j\) digits of \(x_j\) are correct for \(\frac{1}{3}\). Induction is suggested as the method to prove this, with a base case established for \(j=1\) and an inductive hypothesis for \(j=k\). The inductive step confirms the formula holds for \(j=k+1\), leading to the conclusion that the approximation improves exponentially with each iteration. The discussion concludes with the correction of an initial calculation error, affirming the validity of the results.
mathmari
Gold Member
MHB
Messages
4,984
Reaction score
7
Hey! 😊

We have the iteration formula $$x_{j+1}=2x_j-3x_j^2$$ and using the starting point $x_0=0.3$ we get the following approximations for $\frac{1}{3}$ :

\begin{align*}&x_1=2x_0-3x_0^2=2\cdot 0.3-3\cdot 0.3^2=0.33 \\ &x_2=2x_1-3x_1^2=2\cdot 0.33-3\cdot 0.33^2=0.3333 \\ &x_3=2x_2-3x_2^2=2\cdot 0.3333-3\cdot 0.3333^2=0.33333333\end{align*}

Now I want to show that for $j\geq 1$ the first $2^j$ digits of $x_j$ are correct for $\frac{1}{3}$.

It is $\frac{1}{3}=0.33333333333333333333333333\ldots$.

The approximation $x_1$ has the first $2=2^1$ digits correct.

The approximation $x_2$ has the first $4=2^2$ digits correct.

The approximation $x_3$ has the first $8=2^3$ digits correct.To show that it holds for $j\geq 1$ do we have to use induction? :unsure:
 
Last edited by a moderator:
Mathematics news on Phys.org
Hi mathmari!

Induction seems to be the way to go yes.
So we need to write $x_j$ in terms of $2^j$. Can we find such an expression? šŸ¤”
 
Klaas van Aarsen said:
Induction seems to be the way to go yes.
So we need to write $x_j$ in terms of $2^j$. Can we find such an expression? šŸ¤”

As a power of $10$ we get \begin{align*}x_j&=\sum_{i=1}^{2^j}\frac{3}{10^i}=3\cdot \sum_{i=1}^{2^j}\frac{1}{10^i}=3\cdot \left (\sum_{i=0}^{2^j}\frac{1}{10^i}-1\right )=3\cdot \frac{\frac{1}{10^{2^j+1}}-1}{\frac{1}{10}-1}-3=3\cdot \frac{\frac{1}{10^{2^j+1}}-1}{-\frac{9}{10}}-3=-\frac{10}{9}\cdot 3\cdot \left (\frac{1}{10^{2^j+1}}-1\right )-3=-\frac{10}{3}\cdot \left (\frac{1}{10^{2^j+1}}-1\right )-3=-\frac{10}{3}\cdot \frac{1}{10^{2^j+1}}+\frac{10}{3}-3\\ & =-\frac{1}{3}\cdot \frac{1}{10^{2^j}}+\frac{1}{3}\end{align*} right?
So by induction on $j$ we get:

Base Case: For $j=1$ we get the desired result, as seen.

Inductive Hypothesis: We suppose that the formula is true for $j=k$, i.e. $\displaystyle{x_k=-\frac{1}{3}\cdot \frac{1}{10^{2^k}}+\frac{1}{3}}$.

Inductive Step: We want to show that the formula is true also for $j=k+1$ : \begin{align*}x_{k+1}=2x_k-3x_k^2\ &\overset{(IH)}{=} \ 2\left (-\frac{1}{3}\cdot \frac{1}{10^{2^k}}+\frac{1}{3}\right )-3\left (-\frac{1}{3}\cdot \frac{1}{10^{2^k}}+\frac{1}{3}\right )^2\\ & =-\frac{2}{3}\cdot \frac{1}{10^{2^k}}+2\cdot \frac{1}{3}-3\left (\frac{1}{3^2}\cdot \frac{1}{10^{2^{k+1}}}-2\cdot \frac{1}{3}\cdot \frac{1}{10^{2^k}}\cdot \frac{1}{3}+\frac{1}{3^2}\right ) \\ & =-\frac{2}{3}\cdot \frac{1}{10^{2^k}}+2\cdot \frac{1}{3}-\frac{1}{3}\cdot \frac{1}{10^{2^{k+1}}}+2\cdot \frac{1}{10^{2^k}}\cdot \frac{1}{3}-\frac{1}{3}\\ & =\frac{1}{3}-\frac{1}{3}\cdot \frac{1}{10^{2^{k+1}}} \end{align*}

So we get that this hold for every $j\geq 1$ (Happy)
 
Last edited by a moderator:
mathmari said:
As a power of $10$ we get \begin{align*}x_j&=-\frac{1}{3}\cdot \frac{1}{10^{2^j}}+\frac{10}{3}\end{align*} right?
Suppose we substitute $j=1$. Then we should get $x_1=0.33$ yes?
$$x_1=-\frac{1}{3}\cdot \frac{1}{10^{2^1}}+\frac{10}{3}=-\frac{1}{300}+\frac{10}{3}=3.33$$
That's not correct is it? :eek:
 
Klaas van Aarsen said:
Suppose we substitute $j=1$. Then we should get $x_1=0.33$ yes?
$$x_1=-\frac{1}{3}\cdot \frac{1}{10^{2^1}}+\frac{10}{3}=-\frac{1}{300}+\frac{10}{3}=3.33$$
That's not correct is it? :eek:

I found my error! I edited my previous post and the correct result came out!