What is the Linear Combination of Recursive Series?

  • Thread starter Thread starter ramsey2879
  • Start date Start date
  • Tags Tags
    Series
ramsey2879
Messages
841
Reaction score
3
Let there be two recursive series S and R as follows:

S_{0} = 0 \quad S_{1} = s \quad S_{n} = 6S_{n-1} - S_{n-2} + p
R_{0} = 0 \quad R_{1} = t \quad R_{n} = 6R_{n-1} - R_{n-2} + p

Prove S_{n}*(R_{n+1} + t - p) = R_{n}*(S_{n+1} + s - p)
 
Physics news on Phys.org
Sn is a linear combination of s and p, as Rn is of t and p. In fact, they can be written as
\begin{align*}<br /> S_n &amp;= s A_n + p B_n \\<br /> R_n &amp;= t A_n + p B_n<br /> \end{align*}​
where
<br /> \begin{align*}<br /> A_n &amp;= 6 A_{n-1} - A_{n-2} &amp; &quot;(Sloane A001109)&quot; \\<br /> B_n &amp;= 6 B_{n-1} - B_{n-2} + 1<br /> \end{align*}<br />​
A few things can be said about An and Bn. It is easy to prove by induction that
B_n = \sum_{i=1}^{n-1} A_i​
Thus B_{n+1} = A_n + B_n.
Since B_{n+1} is also 6 B_n - B_{n-1} + 1,
we have A_n = 5 B_n - B_{n-1} + 1.

From the above, it is possible to work by induction to prove
<br /> B_n (B_n - 1) = B_{n+1} B_{n-1}<br />​
and from here, to arrive by substitution to
<br /> A_n (B_{n+1} - 1) = B_n (A_{n+1}+1)<br />​

At this point it is possible to use these results to prove your claim. It is best to rule out first two simpler cases, (1) s=t and (2) p=0, and finally, the case where both s\neqt and p\neq0.

More specifically, you'll arrive to something like
<br /> p (s-t) \left( A_n B_{n+1} - A_n - B_n - A_{n+1}B_n \right) = 0<br />​
 
Last edited:
ramsey2879 said:
Let there be two recursive series S and R as follows:

S_{0} = 0 \quad S_{1} = s \quad S_{n} = 6S_{n-1} - S_{n-2} + p
R_{0} = 0 \quad R_{1} = t \quad R_{n} = 6R_{n-1} - R_{n-2} + p

Prove S_{n}*(R_{n+1} + t - p) = R_{n}*(S_{n+1} + s - p)
I also discovered that "p" can be different for S and R so let p1 correspond to S and p2 correspond to R. My identity becomes

S_{n}*(R_{n+1}+t -p2) = R_{n}*(S_{n+1} + s - p1)
 
Also any integer can be substituted for the 6 in the recursive formula
 
I was wondering about that. Now what about a second integer coefficient (i.e. other than -1) on the n-2 term?
 
why not look at:
T(n)=S(n)-R(n)=6(S(n-1)-R(n-1))-(S(n-2)-S(n-1))=6T(n-1)-T(n-2)
T(0)=0
T(1)=s-t
solve for T(n), and now you have a connection between S(n) and R(n), then you can write S(n)=R(n)+(solution of T(n)) and then you write each side of the equation with only one of the expressions which is much easier than both S and R, i think it should be easier, did'nt try it myself though.
 
Dodo said:
I was wondering about that. Now what about a second integer coefficient (i.e. other than -1) on the n-2 term?
I tried integers from -9 to 9 for the n-2 term multiplier but only -1 worked. I will work more on the theory now. By the way the LaTex graphic for the A(n) and B(n) definition in your first post doesn't generate on my computer. Is there a problem with it?
 
Yes, thanks. It was supposed to be
<br /> \begin{align*}<br /> A_0 = 0; \ \ \ A_1 = 1; \ \ \ &amp; A_n = 6 A_{n-1} - A_{n-2} &amp; \mbox{ (Sloane A001109) } \\<br /> B_0 = 0; \ \ \ B_1 = 0; \ \ \ &amp; B_n = 6 B_{n-1} - B_{n-2} + 1<br /> \end{align*}<br />​
It showed in the preview, but not in the post. I just thought it was some server load issue. And it was incomplete anyway. And my LaTex sucks.

BTW, did you follow the same route for the proof? Mine was too long. Maybe using the closed form of the sequences is faster, but that's beyond me.
 
Last edited:
Dodo said:
Yes, thanks. It was supposed to be
<br /> \begin{align*}<br /> A_0 = 0; \ \ \ A_1 = 1; \ \ \ &amp; A_n = 6 A_{n-1} - A_{n-2} &amp; \mbox{ (Sloane A001109) } \\<br /> B_0 = 0; \ \ \ B_1 = 0; \ \ \ &amp; B_n = 6 B_{n-1} - B_{n-2} + 1<br /> \end{align*}<br />​
It showed in the preview, but not in the post. I just thought it was some server load issue. And it was incomplete anyway. And my LaTex sucks.

BTW, did you follow the same route for the proof? Mine was too long. Maybe using the closed form of the sequences is faster, but that's beyond me.

It seems that a proof is more involved now that we added more variables to the mix
I like the A_{n} and B_{n} approach but we need to rephase it using x in place of 6.
There are two tables of coefficients for A_n and B_n each are triangular with T{i,j) = T(i-1,j-1) - T(i-1,j) for j > 0. The left hand column, (j=0) alternates between 1 and -1 for A_n and between 1 and 0 for B_n.
Table of coefficients for A_{n}
1
-1 1
1 -2 1
-1 3 -3 1
1 -4 6 -4 1
...
The powers of x increase from left to right
You use diagonal lines through this table with a slope of 1 down to 2 across

A(1) = 1
A(2) = x
A(3) = -1 + x^2
A(4) = -2x+x^3
A(5) = 1 -3x^2 + x^4
...

Table of coefficients for B_{n}

1
0 1
1 -1 1
0 2 -2 1
1 -2 4 -3 1

Here there are two adjacent diagonal lines that make up the polynomial in x

B(2) = 1
B(3) = 1 + x
B(4) = x+ x^2
B(5) = -x + x^2 + x^3
...

I got to close now.
 
Last edited:
  • #10
have you even tried my suggestion, cause it seems this appraoch gets you nowhere.

p.s
do you know how to solve recursive equations?
 
  • #11
loop quantum gravity said:
have you even tried my suggestion, cause it seems this appraoch gets you nowhere.

p.s
do you know how to solve recursive equations?
I thought of that but the solution I knew was for basic equations with the d variable equal to 0. The known solution would be for S_{n} = bS_{n-1} - S_{n-2} + d for d = 0.
Now that you mention it I went back to figure how to add a constant to each term of my series to get to a recursive series that has 0 for the d variable.
Thus I found that if I add \frac{d}{b-2} to each term of my series I obtain a new series with the same auxiliary equation x^2-bx+1 = 0 with roots r_{1} , r_{2} but with d = 0. Then A and B are found by solving the following set of equations;
S_{n} = Ar_{1}^{n} + Br_{2}^{n} - \frac{d}{b-2}

Do you agree with this?
 
Last edited:
  • #12
Dodo said:
Sn is a linear combination of s and p, as Rn is of t and p. In fact, they can be written as
\begin{align*}<br /> S_n &amp;= s A_n + p B_n \\<br /> R_n &amp;= t A_n + p B_n<br /> \end{align*}​
where
A_{0} = 0 \quad A_{1} = 1 \quad A_{n} = 6 A_{n-1} - A_{n-2}
B_{0} = B_{1} = 0 \quad B_{n} = 6 B_{n-1} - B_{n-2} + 1
Dodo said:
A few things can be said about An and Bn. It is easy to prove by induction that
B_n = \sum_{i=1}^{n-1} A_i​
Thus B_{n+1} = A_n + B_n[/itex].<br /> Since B_{n+1} is also 6 B_n - B_{n-1} + 1,<br /> we have A_n = 5 B_n - B_{n-1} + 1.<br /> <br /> From the above, it is possible to work by induction to prove<br /> <div style="margin-left: 20px">&lt;br /&gt; B_n (B_n - 1) = B_{n+1} B_{n-1}&lt;br /&gt;&#8203;</div>and from here, to arrive by substitution to<br /> <div style="margin-left: 20px">&lt;br /&gt; A_n (B_{n+1} - 1) = B_n (A_{n+1}+1)&lt;br /&gt;&#8203;</div>At this point it is possible to use these results to prove your claim. It is best to rule out first two simpler cases, (1) s=t and (2) p=0, and finally, the case where both s\neqt and p\neq0.<br /> <br /> More specifically, you&#039;ll arrive to something like<br /> <div style="margin-left: 20px">&lt;br /&gt; p (s-t) \left( A_n B_{n+1} - A_n - B_n - A_{n+1}B_n \right) = 0&lt;br /&gt;&#8203;</div>
<br /> I looked this over again and it indeed does prove my identity even for the added variables. Thanks for your effort. PS I simplified the LaTex that didn&#039;t take for some reason. I think your LaTex is much better than mine though.<br /> <br /> Again I thank you.
 
Back
Top