Solve Reduction of Order Legendre Polynomials Equation

  • Context: MHB 
  • Thread starter Thread starter Dustinsfl
  • Start date Start date
  • Tags Tags
    Reduction
Click For Summary

Discussion Overview

The discussion revolves around solving the differential equation for Legendre polynomials using the method of reduction of order. Participants explore the formulation of solutions, particularly focusing on deriving the second solution \(y_2\) given the first solution \(y_1 = P_n(x)\). The conversation includes mathematical derivations, recurrence relations, and the application of the Frobenius method.

Discussion Character

  • Technical explanation
  • Mathematical reasoning
  • Debate/contested

Main Points Raised

  • One participant presents a solution to the Legendre polynomial equation and derives relationships between coefficients \(a_k\) based on the differential equation.
  • Another participant suggests using the method of reduction of order to find the second solution \(y_2\) and provides a formula for the Legendre functions of the second kind.
  • Several participants express concerns about the complexity of taking derivatives of the series and the potential messiness of the calculations involved.
  • There is a discussion about the correct form of the first solution, with some participants asserting that it should be \(y_1 = P_n(x)\) where \(\lambda^2 = n(n+1)\).
  • One participant questions the presence of \(x\) in the recurrence relations derived, suggesting that they should not depend on \(x\).
  • Another participant emphasizes that the second solution can be expressed in terms of \(P_n(x)\) without needing to substitute for it, indicating a potential simplification in the approach.

Areas of Agreement / Disagreement

Participants generally agree on the formulation of the first solution as \(y_1 = P_n(x)\), but there is disagreement regarding the approach to finding the second solution and the validity of the derived recurrence relations. The discussion remains unresolved with multiple competing views on the method and its implementation.

Contextual Notes

Participants note the potential complexity of the calculations involved in deriving the second solution and the need for careful handling of the series and derivatives. There are also concerns about the dependency of recurrence relations on \(x\), which may complicate the solution process.

Dustinsfl
Messages
2,217
Reaction score
5
Legendre polynomials
I have one solution to $(1-x^2)y''-2xy'+\lambda^2 y=0$ and that is
$$
\lambda^2a_0 - (2 - \lambda^2)xa_1 + 2a_2 + 6xa_3 + \sum_{k = 2}^{\infty}\left[a_k\left(\lambda^2 - k - k^2\right) + a_{k + 2}\left(k^2 + 3k + 2\right)\right]x^k = 0.
$$
From this equation, we can conclude $\lambda^2a_0 + 2a_2 = 0$ and $-(2 - \lambda^2)xa_1 + 6xa_3 = 0$.
Therefore,
$$
a_2 = -\frac{\lambda^2}{2}a_0\quad\text{and}\quad a_3 = \frac{2 - \lambda^2}{6}a_1.
$$
Finally, we have $a_k\left(\lambda^2 - k - k^2\right) + a_{k + 2}\left(k^2 + 3k + 2\right) = 0$, i.e.
$$
a_{k + 2} = \frac{k(k + 1) - \lambda^2}{(k + 2)(k + 1)}a_k.
$$
Use the method of reduction of order to derive the following formula for the Legendre functions of the 2nd kind:
$$
Q_n(x) = P_n(x)\int\frac{1}{[P_n(x)]^2(1 - x^2)}dx, \quad n = 0,1,2,\ldots
$$
I basically need to solve
$$
y_2 = y_1\int\frac{\exp\left(\int\frac{2x}{1-x^2}dx\right)}{y_1^2}
$$
where $y_1$ is my solution above, correct?
 
Physics news on Phys.org
Hi dwsmith, :)

dwsmith said:
Legendre polynomials
I have one solution to $(1-x^2)y''-2xy'+\lambda^2 y=0$ and that is
$$
\lambda^2a_0 - (2 - \lambda^2)xa_1 + 2a_2 + 6xa_3 + \sum_{k = 2}^{\infty}\left[a_k\left(\lambda^2 - k - k^2\right) + a_{k + 2}\left(k^2 + 3k + 2\right)\right]x^k = 0.
$$

From this equation, we can conclude $\lambda^2a_0 + 2a_2 = 0$ and $-(2 - \lambda^2)xa_1 + 6xa_3 = 0$.
Therefore,
$$
a_2 = -\frac{\lambda^2}{2}a_0\quad\text{and}\quad a_3 = \frac{2 - \lambda^2}{6}a_1.
$$
Finally, we have $a_k\left(\lambda^2 - k - k^2\right) + a_{k + 2}\left(k^2 + 3k + 2\right) = 0$, i.e.
$$
a_{k + 2} = \frac{k(k + 1) - \lambda^2}{(k + 2)(k + 1)}a_k.
$$

Note that,

\[\lambda^2a_0 - (2 - \lambda^2)xa_1 + 2a_2 + 6xa_3 + \sum_{k = 2}^{\infty}\left[a_k\left(\lambda^2 - k - k^2\right) + a_{k + 2}\left(k^2 + 3k + 2\right)\right]x^k = 0\]

\[\Rightarrow \sum_{k = 0}^{\infty}\left[a_k\left(\lambda^2 - k - k^2\right) + a_{k + 2}\left(k^2 + 3k + 2\right)\right]x^k = 0\]

Therefore,

\[a_{k + 2} = \frac{k(k + 1) - \lambda^2}{(k + 2)(k + 1)}a_k\mbox{ for }k\geq 0\]

dwsmith said:
Use the method of reduction of order to derive the following formula for the Legendre functions of the 2nd kind:
$$
Q_n(x) = P_n(x)\int\frac{1}{[P_n(x)]^2(1 - x^2)}dx, \quad n = 0,1,2,\ldots
$$

You have the differential equation,

\[(1-x^2)y''-2xy'+\lambda^2 y=0\]

with the solution,

\[y_{1}(x)=P_{n}(x)\]

>>Here<< you can find an example of how to perform the method of reduction of order. Try to find the second solution \(y_{2}(x)=Q_{n}(x)\) using this method. Hope you can continue. :)

Kind Regards,
Sudharaka.
 
Sudharaka said:
>>Here<< you can find an example of how to perform the method of reduction of order. Try to find the second solution \(y_{2}(x)=Q_{n}(x)\) using this method. Hope you can continue. :)

Kind Regards,
Sudharaka.

I am looking through Zill's ODE book and it tells me to find the second solution using that formula I listed.
 
Doing $y_2 = y_1v$. I would have to take the derivative of that series twice? That is going to be bit messy I imagine.
$$
y_2 = v\sum_{k = 0}^{\infty}\left[a_k\left(\lambda^2 - k - k^2\right) + a_{k + 2}\left(k^2 + 3k + 2\right)\right]x^k
$$
$$
y_2' = v'\sum_{k = 0}^{\infty}\left[a_k\left(\lambda^2 - k - k^2\right) + a_{k + 2}\left(k^2 + 3k + 2\right)\right]x^k+
v\sum_{k = 1}^{\infty}\left[a_k\left(\lambda^2k - k^2 - k^3\right) + a_{k + 2}\left(k^3 + 3k^2 + 2k\right)\right]x^{k-1}
$$
\begin{alignat}{3}
y_2'' & = & v''\sum_{k = 0}^{\infty}\left[a_k\left(\lambda^2 - k - k^2\right) + a_{k + 2}\left(k^2 + 3k + 2\right)\right]x^k\\
& + & 2v'\sum_{k = 1}^{\infty}\left[a_k\left(\lambda^2k - k^2 - k^3\right) + a_{k + 2}\left(k^3 + 3k^2 + 2k\right)\right]x^{k-1}\\
& + \ &v\sum_{k = 2}^{\infty}\left[a_k\left(\lambda^2(k^2-k) - k^4 + k^2\right) + a_{k + 2}\left(k^4 + 2k^3 -k^2-2k\right)\right]x^{k-2}
\end{alignat}
Is there a better way to do this? I mean plugging this into the Legendre equation is going to be ridiculous.I plugged it into Mathematica and obtained
\begin{alignat}{2}
(1-x^2)\left(\sum_{k = 2}^{\infty}\left[a_k\left(\lambda^2(k^2-k) - k^4 + k^2\right) + a_{k + 2}\left(k^4 + 2k^3 -k^2-2k\right)\right]x^{k-2}\right) & -\\
2\left(\sum_{k = 1}^{\infty}\left[a_k\left(\lambda^2k - k^2 - k^3\right) + a_{k + 2}\left(k^3 + 3k^2 + 2k\right)\right]x^{k-1}\right)(xv+(x^2-1)v' & +\\
\left(\sum_{k = 0}^{\infty}\left[a_k\left(\lambda^2 - k - k^2\right) + a_{k + 2}\left(k^2 + 3k + 2\right)\right]x^k\right)(\lambda^2v-2xv'+(1-x^2)v'')& =\\
& 0
\end{alignat}
However, Paul's notes says the v parameter should be eliminated. I let Mathematica do all the work and the v parameter is still present.
 
Last edited:
I think you are in a confusion as to what the solution to the differential equation is. :) It should be,

\[y_{1}=P_{n}(x)=\sum_{k=0}^{\infty}a_{k}x^k\mbox{ where }\lambda^2=n(n+1)\]

You will find >>this<< helpful in understanding how to solve the Legendre's differential equation using Frobenius method.

So in the method of reduction of order you have to take,

\[y_{2}(x)=v(x)\sum_{k=0}^{\infty}a_{k}x^k\]
 
Sudharaka said:
I think you are in a confusion as to what the solution to the differential equation is. :) It should be,

\[y_{1}=P_{n}(x)=\sum_{k=0}^{\infty}a_{k}x^k\mbox{ where }\lambda^2=n(n+1)\]

You will find >>this<< helpful in understanding how to solve the Legendre's differential equation using Frobenius method.

So in the method of reduction of order you have to take,

\[y_{2}(x)=v(x)\sum_{k=0}^{\infty}a_{k}x^k\]

That is what I did original to get my first solution. Now I am looking for a second solution.
 
dwsmith said:
That is what I did original to get my first solution. Now I am looking for a second solution.

Making that change I still have
$$
(1-x^2)\left(\sum_{k = 2}^{\infty}k(k-1)a_kx^{k-2}\right) -
2\left(\sum_{k = 1}^{\infty}ka_kx^{k-1}\right)(xv+(x^2-1)v') +
\left(\sum_{k = 0}^{\infty}a_kx^k\right)(\lambda^2v-2xv'+(1-x^2)v'') =
0
$$
 
dwsmith said:
Making that change I still have
$$
(1-x^2)\left(\sum_{k = 2}^{\infty}k(k-1)a_kx^{k-2}\right) -
2\left(\sum_{k = 1}^{\infty}ka_kx^{k-1}\right)(xv+(x^2-1)v') +
\left(\sum_{k = 0}^{\infty}a_kx^k\right)(\lambda^2v-2xv'+(1-x^2)v'') =
0
$$

Simplify this and equate the coefficients of similar powers of \(x\) to zero; similar to what you have done in your first post to find the recurrence relation.
 
Sudharaka said:
Simplify this and equate the coefficients of similar powers of \(x\) to zero; similar to what you have done in your first post to find the recurrence relation.

\begin{align}
a_0(\lambda^2v-2v'x+v'')+a_1(v''+2v'+vx(2+\lambda^2)+a_2(2+v'x)+6a_3+\\
\sum_{n=2}^{\infty}[a_{n+2}(n^2+3n+2)+a_{n+1}2v'(n+1)+a_n(v''+v( \lambda^2 -2n)-n(n-1))+a_{n-1}2v'n+a_{n-2}v'']x^n
\end{align}

Is this correct (^)?
If so,
$$
a_2 = \frac{2v'-v''-\lambda^2v}{2+v'x}a_0
$$
and
$$
a_3=-a_1\frac{v''+2v'+vx(\lambda^2+2)}{6}.
$$
 
  • #10
dwsmith said:
\begin{align}
a_0(\lambda^2v-2v'x+v'')+a_1(v''+2v'+vx(2+\lambda^2)+a_2(2+v'x)+6a_3+\\
\sum_{n=2}^{\infty}[a_{n+2}(n^2+3n+2)+a_{n+1}2v'(n+1)+a_n(v''+v( \lambda^2 -2n)-n(n-1))+a_{n-1}2v'n+a_{n-2}v'']x^n
\end{align}

Is this correct (^)?
If so,
$$
a_2 = \frac{2v'-v''-\lambda^2v}{2+v'x}a_0
$$
and
$$
a_3=-a_1\frac{v''+2v'+vx(\lambda^2+2)}{6}.
$$

The recurrence relations should not be dependent on \(x\). :)
 
  • #11
Sudharaka said:
The recurrence relations should not be dependent on \(x\). :)

What? That solution is following what you said.
 
  • #12
dwsmith said:
What? That solution is following what you said.

The recurrence relations you have obtained have \(x\) in them,

\[a_2 = \frac{2v'-v''-\lambda^2v}{2+v'{\color{red}x}}a_0\]

\[a_3=-a_1\frac{v''+2v'+v{\color{red}x}(\lambda^2+2)}{6}.\]

Anyway, thinking about this problem further I came to note that there is no need to substitute for \(P_{n}(x)\) since the answer is given in terms of \(P_{n}(x)\). So you can take,

\[y_{2}(x)=v(x)P_{n}(x)\]

and perform the reduction of order. Let me guide you through the first steps. The differential equation is,

\[(1-x^2)y''-2xy'+\lambda^2 y=0\]

We know that \(y_{1}(x)=P_{n}(x)\). Let, \(y_{2}(x)=Q_{n}(x)=v(x)P_{n}(x)\). Then,

\[(1-x^2)[v(x)P_{n}(x)]''-2x[v(x)P_{n}(x)]'+\lambda^2 v(x)P_{n}(x)=0\]

Using, \((1-x^2)P''_{n}(x)-2xP'_{n}(x)+\lambda^2 P_{n}(x)=0\) you can obtain,

\[2(1-x^2)v'(x)P'_{n}(x)+(1-x^2)v''(x)P_{n}(x)-2xv'(x)P_{n}(x)=0\]

\[\Rightarrow 2(1-x^2)v'(x)P_{n}(x) P'_{n}(x)+(1-x^2)v''(x)P_{n}^{2}(x)-2xv'(x)P_{n}^{2}(x)=0\]

\[\Rightarrow \frac{d}{dx}\left[(1-x^2)P_{n}^{2}(x)\right]v'(x)+[(1-x^2)P_{n}^{2}(x)]v''(x)=0\]

Now you should be able to solve for \(v(x)\) by separation of variables.
 
Last edited:
  • #13
Sudharaka said:
\[\Rightarrow 2\frac{d}{dx}\left[(1-x^2)P_{n}(x)\right]v'(x)+[(1-x^2)P_{n}(x)]v''(x)=0\]

Now you should be able to solve for \(v(x)\) by separation of variables.

$$
\left[(1-x^2)P_{n}(x)\right]' = [P - x^2P]' = P' - 2xP - x^2P' = P'(1-x^2) - 2xP
$$
If I multiply 2 and v' in, we have
$$
4P'x
$$
which isn't right.
 
  • #14
dwsmith said:
$$
\left[(1-x^2)P_{n}(x)\right]' = [P - x^2P]' = P' - 2xP - x^2P' = P'(1-x^2) - 2xP
$$
If I multiply 2 and v' in, we have
$$
4P'x
$$
which isn't right.

Sorry. There was a little error. I have edited my last post (#12). :)
 
  • #15
Sudharaka said:
Sorry. There was a little error. I have edited my last post (#12). :)

Then we have
$$
\frac{v''}{v'} = -\frac{1}{2}\frac{[(1-x^2)P^2]'}{(1-x^2)P^2}\Rightarrow \ln v' = \frac{1}{2}\ln((1-x^2)P^2)
$$
I don't see how to go from this to the solution.
 
Last edited:
  • #16
dwsmith said:
Then we have
$$
\frac{v''}{v'} = -\frac{1}{2}\frac{[(1-x^2)P^2]'}{(1-x^2)P^2}\Rightarrow \ln v' = \frac{1}{2}\ln((1-x^2)P^2)
$$
I don't see how to go from this to the solution.

From where did the fraction \(\frac{1}{2}\) come from? :) Continuing from post #12 we have,

\[\frac{d}{dx}\left[(1-x^2)P_{n}^{2}(x)\right]v'(x)+[(1-x^2)P_{n}^{2}(x)]v''(x)=0\]

\[\Rightarrow -\frac{v''(x)}{v'(x)}=\frac{\frac{d}{dx}\left[(1-x^2)P_{n}^{2}(x)\right]}{[(1-x^2)P_{n}^{2}(x)]}\]

Integrate both sides and we get,

\[\ln[v'(x)]=\ln\left[(1-x^2)P_{n}^{2}(x)\right]^{-1}\]

\[\Rightarrow v'(x)=\frac{1}{(1-x^2)P_{n}^{2}(x)}\]

Integrate again to obtain,

\[v(x)=\int \frac{1}{(1-x^2)P_{n}^{2}(x)}\,dx\]

\[\therefore Q_{n}(x)=v(x)P_{n}(x)=P_{n}(x)\int \frac{1}{(1-x^2)P_{n}^{2}(x)}\,dx\]
 
  • #17
From this, how can I find $P_0(x)$?
 
  • #18
dwsmith said:
From this, how can I find $P_0(x)$?

You have,

\[P_{n}(x)=\sum_{k=0}^{\infty}a_k x^k\mbox{ where }a_{k + 2} = \frac{k(k + 1) - [n(n+1)]^2}{(k + 2)(k + 1)}a_k\]

Substitute \(n=0\) and try to obtain a solution to the recurrence relation.
 
  • #19
Sudharaka said:
You have,

\[P_{n}(x)=\sum_{k=0}^{\infty}a_k x^k\mbox{ where }a_{k + 2} = \frac{k(k + 1) - [n(n+1)]^2}{(k + 2)(k + 1)}a_k\]

Substitute \(n=0\) and try to obtain a solution to the recurrence relation.

How does one come up with $Q_0$ if $P_0$ is represented as a relation?
 
  • #20
dwsmith said:
How does one come up with $Q_0$ if $P_0$ is represented as a relation?

Recall that,

\[Q_{n}(x)=v(x)P_{n}(x)\]
 

Similar threads

  • · Replies 2 ·
Replies
2
Views
5K
  • · Replies 2 ·
Replies
2
Views
4K
  • · Replies 1 ·
Replies
1
Views
3K
  • · Replies 3 ·
Replies
3
Views
4K
  • · Replies 1 ·
Replies
1
Views
2K
  • · Replies 3 ·
Replies
3
Views
2K
  • · Replies 2 ·
Replies
2
Views
2K
  • · Replies 1 ·
Replies
1
Views
3K
  • · Replies 7 ·
Replies
7
Views
3K
  • · Replies 1 ·
Replies
1
Views
2K