MHB Gaussian Quadrature Formula for Integrating Polynomials of Degree 6

AI Thread Summary
The discussion focuses on calculating the integral of the function 1/(x+3) from 0 to 1 using the Gaussian quadrature formula, which is capable of integrating polynomials of degree up to 7 when n=4. Participants explore the possibility of transforming the integral to the interval [-1, 1] to utilize standard weights and nodes from Gaussian quadrature. They discuss the selection of a weight function, suggesting either w(x)=1 or w(x)=1/(x+3) for simplification. The conversation also touches on the challenges of calculating the roots of the associated Legendre polynomials and emphasizes the importance of understanding weight functions and their impact on integration accuracy. The overall goal is to effectively apply Gaussian quadrature to solve the integral while learning about the underlying mathematical principles.
mathmari
Gold Member
MHB
Messages
4,984
Reaction score
7
Hey! :o

I want to calculate the integral $$\int_0^1\frac{1}{x+3}\, dx$$ with the Gaussian quadrature formula that integrates exactly all polynomials of degree $6$.

The gaussian quadrature integrates exactly polynomials $\Phi (x)$ with maximum degree $2n-1$. In this case we consider $n=4$.

The formula is \begin{equation*}\int_{0}^1\frac{1}{x+3}\, dx\approx \sum_{i=1}^nf(x_i)\cdot w_i=f(x_1)\cdot w_1+f(x_2)\cdot w_2+f(x_3)\cdot w_3+f(x_4)\cdot w_4\end{equation*}

For the calculations of $w_i$ we need the following:
\begin{align*}&P_n(x)=(a_n+x)P_{n-1}(x)+c_nP_{n-2}(x) \\ &P_0=1, \ P_{-1}=0\end{align*} with \begin{equation*}a_n=-\frac{\langle x\cdot P_{n-1}, P_{n-1}\rangle_w}{\langle P_{n-1}, P_{n-1}\rangle_w}, \ \ \ , c_n=-\frac{\langle P_{n-1}, P_{n-1}\rangle_w}{\langle P_{n-2}, P_{n-2}\rangle_w}\end{equation*} right? What is in this case the weight function $w(x)$ ?

(Wondering)
 
Mathematics news on Phys.org
Hey mathmari!

Can't we do a substitution on the integral so that it's on [-1,1]?
Then the standard weights apply, and we have the points and weights as given in Gaussian quadrature.
If we want to, afterwards we can do a reverse substitution to find the points on [0,1]. (Thinking)
 
Klaas van Aarsen said:
Can't we do a substitution on the integral so that it's on [-1,1]?
Then the standard weights apply, and we have the points and weights as given in Gaussian quadrature.
If we want to, afterwards we can do a reverse substitution to find the points on [0,1]. (Thinking)

Ah ok, so we get \begin{equation*}\int_0^1\frac{1}{t+3}\, dt=\int_{-1}^1\frac{1}{\frac{1}{2}+\frac{x}{2}+3}\, \frac{1}{2}dx=\int_{-1}^1\frac{1}{x+7}\, dx\end{equation*}

Since we want that the Gaussian quadrature formula integrates exactly all polynomials of degree 6, do we consider $n=4$ ? (Wondering)
 
mathmari said:
Ah ok, so we get \begin{equation*}\int_0^1\frac{1}{t+3}\, dt=\int_{-1}^1\frac{1}{\frac{1}{2}+\frac{x}{2}+3}\, \frac{1}{2}dx=\int_{-1}^1\frac{1}{x+7}\, dx\end{equation*}

Since we want that the Gaussian quadrature formula integrates exactly all polynomials of degree 6, do we consider $n=4$ ?

Yes.
And with $n=4$ we get exact integration of polynomials up to degree $2n-1=7$, so that includes indeed degree $6$. (Nod)
 
Klaas van Aarsen said:
Yes.
And with $n=4$ we get exact integration of polynomials up to degree $2n-1=7$, so that includes indeed degree $6$. (Nod)

Ah ok! (Nerd)

If the values of $x_i$ and $w_i$ are not given and we have to calcualte them by ourself, how can we do that? In Wiki there is a formula for $w_i$, do we use this one? But what about $x_i$ ? (Wondering)
 
mathmari said:
Ah ok!

If the values of $x_i$ and $w_i$ are not given and we have to calcualte them by ourself, how can we do that? In Wiki there is a formula for $w_i$, do we use this one? But what about $x_i$ ?

Looks like it yes.

Wiki also mentions:
the associated orthogonal polynomials are Legendre polynomials, denoted by Pn(x). With the n-th polynomial normalized to give Pn(1) = 1, the i-th Gauss node, xi, is the i-th root of Pn

(Thinking)
 
mathmari said:
I want to calculate the integral $$\int_0^1\frac{1}{x+3}\, dx$$ with the Gaussian quadrature formula that integrates exactly all polynomials of degree $6$.

For the calculations of $w_i$ we need the following:
\begin{align*}&P_n(x)=(a_n+x)P_{n-1}(x)+c_nP_{n-2}(x) \\ &P_0=1, \ P_{-1}=0\end{align*} with \begin{equation*}a_n=-\frac{\langle x\cdot P_{n-1}, P_{n-1}\rangle_w}{\langle P_{n-1}, P_{n-1}\rangle_w}, \ \ \ , c_n=-\frac{\langle P_{n-1}, P_{n-1}\rangle_w}{\langle P_{n-2}, P_{n-2}\rangle_w}\end{equation*}

mathmari said:
If the values of $x_i$ and $w_i$ are not given and we have to calcualte them by ourself, how can we do that? In Wiki there is a formula for $w_i$, do we use this one? But what about $x_i$ ?

Wiki also mentions:
The integration problem can be expressed in a slightly more general way by introducing a positive weight function ω into the integrand, and allowing an interval other than [−1, 1]. That is, the problem is to calculate
$$\int _{a}^{b}\omega (x)\,f(x)\,dx$$

I believe we can freely pick $\omega(x)$. Whatever makes the calculations the easiest. (Wondering)

In our case we have $a=0$, $b=1$, and $\omega (x)\,f(x)=\frac{1}{x+3}$.
So I think we should either pick $\omega (x)=1$ or $\omega (x)=\frac{1}{x+3}$.

Suppose we pick $\omega (x)=1$.
Then:
\begin{equation*}a_1=-\frac{\langle x\cdot P_0, P_0\rangle_\omega}{\langle P_0, P_0\rangle_\omega}
= -\frac{\langle x\cdot 1, 1\rangle_\omega}{\langle 1, 1\rangle_\omega}
= -\frac{\int_0^1 \omega(x)x\,dx}{\int_0^1 \omega(x)\,dx}
= -\frac{\int_0^1 x\,dx}{\int_0^1 dx}
= -\frac {\frac 12}1 \\
P_1(x)=(a_1+x)P_0(x)+c_1P_{-1}(x)
= \left(-\frac 12+x\right)1+c_1\cdot 0 = x-\frac 12
\end{equation*}
(Thinking)

If we would pick $n=1$, then $x_i$ would be the $i$-th root of $P_1(x)$. That is, we would have $x_1=\frac 12$.
For $n=4$ we need to do a bit more work. (Sweating)
 
Klaas van Aarsen said:
Wiki also mentions:
The integration problem can be expressed in a slightly more general way by introducing a positive weight function ω into the integrand, and allowing an interval other than [−1, 1]. That is, the problem is to calculate
$$\int _{a}^{b}\omega (x)\,f(x)\,dx$$

I believe we can freely pick $\omega(x)$. Whatever makes the calculations the easiest. (Wondering)

In our case we have $a=0$, $b=1$, and $\omega (x)\,f(x)=\frac{1}{x+3}$.
So I think we should either pick $\omega (x)=1$ or $\omega (x)=\frac{1}{x+3}$.
Ahh ok! So, do we not have to shift an arbitrary interval into $[-1,1]$ ? I thought that these formulas would hold only at $[-1,1]$. (Wondering)
 
mathmari said:
Ahh ok! So, do we not have to shift an arbitrary interval into $[-1,1]$ ? I thought that these formulas would hold only at $[-1,1]$.

No, the interval [-1,1] is merely a convenient choice, and tables are readily available for the resulting $P_n(x)$, $x_i$, and $w_i$.
Also, it's not so easy to find the roots of $P_n(x)$ for $n=4$, which is a 4-th degree polynomial after all. (Thinking)
 
  • #10
Klaas van Aarsen said:
No, the interval [-1,1] is merely a convenient choice, and tables are readily available for the resulting $P_n(x)$, $x_i$, and $w_i$.

So, do you suggest to let the interval of the integral at $[0,1]$ or to shift it to $[-1,1]$ ? (Wondering)
Klaas van Aarsen said:
Also, it's not so easy to find the roots of $P_n(x)$ for $n=4$, which is a 4-th degree polynomial after all. (Thinking)

But in this case we have to calculate the roots of the $4$th degree polynomial, or is there also an other way? (Wondering)
 
  • #11
mathmari said:
So, do you suggest to let the interval of the integral at $[0,1]$ or to shift it to $[-1,1]$ ?

It's easiest to approximate the integral with a Gaussian quadrature by shifting to $[-1,1]$ and using the tabulated data.
You had already almost done that. It makes sense to finish it, so that we can use it for comparisons.

However, I think the purpose of the problem is to learn about weight functions, weights, and how Gaussian quadrature really works. For that I think we should try both choices for $\omega(x)$ and see what we get.
Maybe one of them works out really well. (Wait)
mathmari said:
But in this case we have to calculate the roots of the $4$th degree polynomial, or is there also an other way?

I'm not sure yet. Maybe we will find a 4-th degree polynomial that is not so hard.
And anyway, we can always solve it with for instance Wolfram. (Thinking)
 
  • #12
I have done the following:

\begin{align*}&a_1=-\frac{\langle x\cdot P_0, P_0\rangle_\omega}{\langle P_0, P_0\rangle_\omega}
= -\frac{\langle x\cdot 1, 1\rangle_\omega}{\langle 1, 1\rangle_\omega}
= -\frac{\int_0^1 \omega(x)x\,dx}{\int_0^1 \omega(x)\,dx}
= -\frac{\int_0^1 x\,dx}{\int_0^1 dx}
= -\frac {\frac 12}1 \\
&P_1(x)=(a_1+x)P_0(x)+c_1P_{-1}(x)
= \left(-\frac 12+x\right)1+c_1\cdot 0 = x-\frac 12
\end{align*}

\begin{align*}&a_2=-\frac{\langle x\cdot P_{1}, P_{1}\rangle_w}{\langle P_{1}, P_{1}\rangle_w}=-\frac{\langle x\cdot \left (x-\frac 12\right ), x-\frac 12\rangle_w}{\langle x-\frac 12, x-\frac 12\rangle_w}=-\frac{\int_0^1x\cdot \left (x-\frac 12\right )\cdot w(x)\, dx}{\int_0^1\left (x-\frac 12\right )^2\cdot w(x)\, dx}=-\frac{\int_0^1\left (x^2-\frac x2\right )\, dx}{\int_0^1\left (x^2-x+\frac 14\right )\, dx}=-\frac{\frac{1}{12}}{\frac{1}{12}}=-1 \\ & c_2=-\frac{\langle P_{1}, P_{1}\rangle_w}{\langle P_{0}, P_{0}\rangle_w}=-\frac{\langle x-\frac 12, x-\frac 12\rangle_w}{\langle 1, 1\rangle_w}=-\frac{\int_0^1\left (x-\frac 12\right )^2\cdot w(x)\, dx}{\int_0^1 w(x)\, dx}=-\frac{\int_0^1\left (x-\frac 12\right )^2\, dx}{\int_0^1 1\, dx}=-\frac{\frac{1}{12}}{1}=-\frac{1}{12} \\ & P_2(x)=(a_2+x)P_{1}(x)+c_2P_{0}(x)=(-1+x)\cdot \left (x-\frac{1}{2}\right )-\frac{1}{12}\cdot 1=x^2-\frac{3x}{2}+\frac{5}{12}
\end{align*}

\begin{align*}&a_3=-\frac{\langle x\cdot P_{2}, P_{2}\rangle_w}{\langle P_{2}, P_{2}\rangle_w}=-\frac{\langle x\cdot \left (x^2-\frac{3x}{2}+\frac{5}{12} \right ), x^2-\frac{3x}{2}+\frac{5}{12}\rangle_w}{\langle x^2-\frac{3x}{2}+\frac{5}{12}, x^2-\frac{3x}{2}+\frac{5}{12}\rangle_w}=-\frac{\int_0^1 x\cdot \left (x^2-\frac{3x}{2}+\frac{5}{12} \right )^2\, dx}{\int_0^1 \left ( x^2-\frac{3x}{2}+\frac{5}{12}\right )^2\, dx}=-\frac{\frac{11}{1440}}{\frac{19}{720}}=-\frac{11}{38} \\ & c_3=-\frac{\langle P_{2}, P_{2}\rangle_w}{\langle P_{1}, P_{1}\rangle_w}=-\frac{\langle x^2-\frac{3x}{2}+\frac{5}{12}, x^2-\frac{3x}{2}+\frac{5}{12}\rangle_w}{\langle x-\frac 12, x-\frac 12\rangle_w}=-\frac{\int_0^1\left ( x^2-\frac{3x}{2}+\frac{5}{12}\right )^2\, dx}{\int_0^1\left ( x-\frac 12\right )^2\, dx}=-\frac{\frac{19}{720}}{\frac{1}{12}}=-\frac{19}{60} \\ & P_3(x)=(a_3+x)P_{2}(x)+c_3P_{1}(x)=\left (-\frac{11}{38}+x\right )\left (x^2-\frac{3x}{2}+\frac{5}{12}\right )-\frac{19}{60}\left (x-\frac 12\right ) \\ & =x^3-\frac{34}{19}x^2+\frac{203}{380}x+\frac{43}{1140}\end{align*}

\begin{align*}&a_4=-\frac{\langle x\cdot P_{3}, P_{3}\rangle_w}{\langle P_{3}, P_{3}\rangle_w}=-\frac{\int_0^1x\cdot \left (x^3-\frac{34}{19}x^2+\frac{203}{380}x+\frac{43}{1140}\right )^2\, dx}{\int_0^1\left (x^3-\frac{34}{19}x^2+\frac{203}{380}x+\frac{43}{1140}\right )^2\, dx}=-\frac{\frac{120583}{12129600}}{\frac{2087}{159600}}=-\frac{120583}{158612} \\ &c_4=-\frac{\langle P_{3}, P_{3}\rangle_w}{\langle P_{2}, P_{2}\rangle_w}=-\frac{\int_0^1 \left (x^3-\frac{34}{19}x^2+\frac{203}{380}x+\frac{43}{1140} \right )^2\, dx}{\int_0^1\left (x^2-\frac{3x}{2}+\frac{5}{12}\right )^2\, dx}=-\frac{\frac{2087}{159600}}{\frac{19}{720}}=-\frac{6261}{12635} \\ &P_4= (a_4+x)P_{3}(x)+c_4P_{2}(x)=\left (-\frac{120583}{158612}+x\right )\left (x^3-\frac{34}{19}x^2+\frac{203}{380}x+\frac{43}{1140}\right )-\frac{6261}{12635}\left (x^2-\frac{3x}{2}+\frac{5}{12}\right )\\ & =x^4-\frac{21285}{8348}x^3+\frac{1553409}{1110284}x^2+\frac{24973673}{66617040}x-\frac{15664717}{66617040}\end{align*}

But I must have done something wrong because the roots (as we see in Wolfram) are not the same as the ones that are in Wiki. Do you have an idea what I have done wrong? (Wondering)
 
  • #13
mathmari said:
I have done the following:

\begin{align*}&a_1=-\frac{\langle x\cdot P_0, P_0\rangle_\omega}{\langle P_0, P_0\rangle_\omega}
= -\frac{\langle x\cdot 1, 1\rangle_\omega}{\langle 1, 1\rangle_\omega}
= -\frac{\int_0^1 \omega(x)x\,dx}{\int_0^1 \omega(x)\,dx}
= -\frac{\int_0^1 x\,dx}{\int_0^1 dx}
= -\frac {\frac 12}1 \\
&P_1(x)=(a_1+x)P_0(x)+c_1P_{-1}(x)
= \left(-\frac 12+x\right)1+c_1\cdot 0 = x-\frac 12
\end{align*}

\begin{align*}&a_2=-\frac{\langle x\cdot P_{1}, P_{1}\rangle_w}{\langle P_{1}, P_{1}\rangle_w}
=-\frac{\langle x\cdot \left (x-\frac 12\right ), x-\frac 12\rangle_w}{\langle x-\frac 12, x-\frac 12\rangle_w}
=-\frac{\int_0^1x\cdot \left (x-\frac 12\right )\cdot w(x)\, dx}{\int_0^1\left (x-\frac 12\right )^2\cdot w(x)\, dx}=-\frac{\int_0^1\left (x^2-\frac x2\right )\, dx}{\int_0^1\left (x^2-x+\frac 14\right )\, dx}
=-\frac{\frac{1}{12}}{\frac{1}{12}}=-1
\end{align*}

Don't we have:
$$a_2=-\frac{\langle x\cdot P_{1}, P_{1}\rangle_w}{\langle P_{1}, P_{1}\rangle_w}=-\frac{\langle x\cdot \left (x-\frac 12\right ), x-\frac 12\rangle_w}{\langle x-\frac 12, x-\frac 12\rangle_w}=-\frac{\int_0^1x\cdot \left (x-\frac 12\right )^2\cdot w(x)\, dx}{\int_0^1\left (x-\frac 12\right )^2\cdot w(x)\, dx}$$
That is, with a square in the numerator? (Thinking)

mathmari said:
But I must have done something wrong because the roots (as we see in Wolfram) are not the same as the ones that are in Wiki. Do you have an idea what I have done wrong? (Wondering)

We wouldn't get exactly the same roots, since we have scaled [-1,1] to [0,1].
But we should get more or less the same distribution yes.
And I would also expect the roots to be in the interval [0,1], because otherwise we cannot use them to calculate $f(x_i)$. (Worried)
 
  • #14
Klaas van Aarsen said:
Don't we have:
$$a_2=-\frac{\langle x\cdot P_{1}, P_{1}\rangle_w}{\langle P_{1}, P_{1}\rangle_w}=-\frac{\langle x\cdot \left (x-\frac 12\right ), x-\frac 12\rangle_w}{\langle x-\frac 12, x-\frac 12\rangle_w}=-\frac{\int_0^1x\cdot \left (x-\frac 12\right )^2\cdot w(x)\, dx}{\int_0^1\left (x-\frac 12\right )^2\cdot w(x)\, dx}$$
That is, with a square in the numerator? (Thinking)

Ahh yes. I did the calculations again and now I get the following:
\begin{align*}&a_1=-\frac{\langle x\cdot P_0, P_0\rangle_\omega}{\langle P_0, P_0\rangle_\omega}
= -\frac{\langle x\cdot 1, 1\rangle_\omega}{\langle 1, 1\rangle_\omega}
= -\frac{\int_0^1 \omega(x)x\,dx}{\int_0^1 \omega(x)\,dx}
= -\frac{\int_0^1 x\,dx}{\int_0^1 dx}
= -\frac {\frac 12}1 \\
&P_1(x)=(a_1+x)P_0(x)+c_1P_{-1}(x)
= \left(-\frac 12+x\right)1+c_1\cdot 0 = x-\frac 12
\end{align*}

\begin{align*}&a_2=-\frac{\langle x\cdot P_{1}, P_{1}\rangle_w}{\langle P_{1}, P_{1}\rangle_w}=-\frac{\langle x\cdot \left (x-\frac 12\right ), x-\frac 12\rangle_w}{\langle x-\frac 12, x-\frac 12\rangle_w}=-\frac{\int_0^1x\cdot \left (x-\frac 12\right )^2\cdot w(x)\, dx}{\int_0^1\left (x-\frac 12\right )^2\cdot w(x)\, dx}=-\frac{\int_0^1x\cdot \left (x-\frac 12\right )^2\, dx}{\int_0^1\left (x-\frac 12\right )^2\, dx}=-\frac{\frac{1}{24}}{\frac{1}{12}}=-\frac{1}{2} \\ & c_2=-\frac{\langle P_{1}, P_{1}\rangle_w}{\langle P_{0}, P_{0}\rangle_w}=-\frac{\langle x-\frac 12, x-\frac 12\rangle_w}{\langle 1, 1\rangle_w}=-\frac{\int_0^1\left (x-\frac 12\right )^2\cdot w(x)\, dx}{\int_0^1 w(x)\, dx}=-\frac{\int_0^1\left (x-\frac 12\right )^2\, dx}{\int_0^1 1\, dx}=-\frac{\frac{1}{12}}{1}=-\frac{1}{12} \\ & P_2(x)=(a_2+x)P_{1}(x)+c_2P_{0}(x)=\left (-\frac{1}{2}+x\right )\cdot \left (x-\frac{1}{2}\right )-\frac{1}{12}\cdot 1=x^2-x+\frac{1}{6}
\end{align*}

\begin{align*}&a_3=-\frac{\langle x\cdot P_{2}, P_{2}\rangle_w}{\langle P_{2}, P_{2}\rangle_w}=-\frac{\langle x\cdot \left (x^2-x+\frac{1}{6} \right ), x^2-x+\frac{1}{6}\rangle_w}{\langle x^2-x+\frac{1}{6}, x^2-x+\frac{1}{6}\rangle_w}=-\frac{\int_0^1 x\cdot \left (x^2-x+\frac{1}{6} \right )^2\, dx}{\int_0^1 \left ( x^2-x+\frac{1}{6}\right )^2\, dx}=-\frac{\frac{1}{360}}{\frac{1}{180}}=-\frac{1}{2} \\ & c_3=-\frac{\langle P_{2}, P_{2}\rangle_w}{\langle P_{1}, P_{1}\rangle_w}=-\frac{\langle x^2-x+\frac{1}{6}, x^2-x+\frac{1}{6}\rangle_w}{\langle x-\frac 12, x-\frac 12\rangle_w}=-\frac{\int_0^1\left ( x^2-x+\frac{1}{6}\right )^2\, dx}{\int_0^1\left ( x-\frac 12\right )^2\, dx}=-\frac{\frac{1}{180}}{\frac{1}{12}}=-\frac{1}{15} \\ & P_3(x)=(a_3+x)P_{2}(x)+c_3P_{1}(x)=\left (-\frac{1}{2}+x\right )\left (x^2-x+\frac{1}{6}\right )-\frac{1}{15}\left (x-\frac 12\right )=x^3-\frac{3}{2}x^2+\frac{3}{5}x-\frac{1}{20}\end{align*}

\begin{align*}&a_4=-\frac{\langle x\cdot P_{3}, P_{3}\rangle_w}{\langle P_{3}, P_{3}\rangle_w}=-\frac{\int_0^1x\cdot \left (x^3-\frac{3}{2}x^2+\frac{3}{5}x-\frac{1}{20}\right )^2\, dx}{\int_0^1\left (x^3-\frac{3}{2}x^2+\frac{3}{5}x-\frac{1}{20}\right )^2\, dx}=-\frac{\frac{1}{5600}}{\frac{1}{2800}}=-\frac{1}{2} \\ &c_4=-\frac{\langle P_{3}, P_{3}\rangle_w}{\langle P_{2}, P_{2}\rangle_w}=-\frac{\int_0^1 \left (x^3-\frac{3}{2}x^2+\frac{3}{5}x-\frac{1}{20} \right )^2\, dx}{\int_0^1\left (x^2-x+\frac{1}{6}\right )^2\, dx}=-\frac{\frac{1}{2800}}{\frac{1}{180}}=-\frac{9}{140} \\ &P_4= (a_4+x)P_{3}(x)+c_4P_{2}(x)=\left (-\frac{1}{2}+x\right )\left (x^3-\frac{3}{2}x^2+\frac{3}{5}x-\frac{1}{20}\right )-\frac{9}{140}\left (x^2-x+\frac{1}{6}\right )=x^4-2x^3+\frac{9}{7}x^2-\frac{2}{7}x+\frac{1}{70}\end{align*}

Is there a way s that we can verify if we got the correct polynomial? (Wondering)
Klaas van Aarsen said:
We wouldn't get exactly the same roots, since we have scaled [-1,1] to [0,1].
But we should get more or less the same distribution yes.
And I would also expect the roots to be in the interval [0,1], because otherwise we cannot use them to calculate $f(x_i)$. (Worried)

Now we get these roots: Wolfram, which are in the interval $[0,1]$.

(Wondering)
 
Last edited by a moderator:
  • #15
mathmari said:
Is there a way that we can verify if we got the correct polynomial?

Yes. We can verify if $\langle P_i,P_j\rangle_w=\delta_{ij}$. (Thinking)

mathmari said:
Now we get these roots: Wolfram, which are in the interval $[0,1]$.

That looks about right. Nice! (Happy)
 
  • #16
Klaas van Aarsen said:
Yes. We can verify if $\langle P_i,P_j\rangle_w=\delta_{ij}$. (Thinking)

But it doesn't hold that $\langle P_i,P_i\rangle_w=1$, does it? (Wondering)

If we took the other weight function, $w(x)=\frac{1}{x+3}$ instead of $w(x)=1$, would we get other polynomials? (Wondering)
 
Last edited by a moderator:
  • #17
mathmari said:
But it doesn't hold that $\langle P_i,P_i\rangle_w=1$, does it?

True. The polynomials are orthogonal, but they have not been normalized to $1$. (Blush)
So we should have for each $i\ne j$ that $\langle P_i,P_j\rangle_w=0$.

mathmari said:
If we took the other weight function, $w(x)=\frac{1}{x+3}$ instead of $w(x)=1$, would we get other polynomials?

I think so yes. (Thinking)
 
  • #18
Klaas van Aarsen said:
True. The polynomials are orthogonal, but they have not been normalized to $1$. (Blush)
So we should have for each $i\ne j$ that $\langle P_i,P_j\rangle_w=0$.

Ah ok! (Malthe)
Klaas van Aarsen said:
I think so yes. (Thinking)

But both are corrects, aren't they? (Wondering)
 
  • #19
mathmari said:
Ah ok!

But both are corrects, aren't they?

Yes. (Nod)
 
Back
Top