(39 intermediate revisions by 9 users not shown)
Line 1: Line 1:
 
== Homework 9 Collaboration Area ==
 
== Homework 9 Collaboration Area ==
  
−
Here are some hints about the problems on Legendre Polynomials.
+
Click on
  
−
See p. 180 for a list of the first few Legendre Polynomials.
+
[[LegendreMA527Fall2010 |Legendre hints from Bell]]
  
−
The even numbered Legendre Polynomials only involve even powers of  x,
+
for some hints about Legendre Polynomials.
−
so they are even functions.
+
  
−
The odd numbered Legendre Polynomials only involve odd powers of  x,
+
Question Page 597, Problem 5:
−
so they are odd functions.
+
  
−
The Legendre Polynomials are orthogonal on the interval [-1,1].
+
What do we do with the x in the first term of this problem?
  
−
p. 209, 5. asks you to show that
+
Answer: When you do a Laplace transform wrt t, the x floats along like when you do d/dt(x*t).
  
−
<math>P_n(\cos\theta)</math>
+
Additional question:  I get to this point after taking the Laplace Transform:
  
−
are orthogonal on [0,pi] with respect to the weight function
+
d/dx*W(x,s) + s/x*W(x,s) = 1/s^2.
  
−
<math>\sin\theta,</math>
+
Is this correct?
  
−
i.e., to show that
+
Answer: Yes, that's what I got.  Check out lecture 23 around the 28 minute mark to see what to do next.
  
−
<math>\int_0^\pi P_n(\cos\theta)P_m(\cos\theta)\sin\theta\ d\theta=0
+
Message from Bell:  When you solve that ODE as a first order linear ODE in the x variable, you should get an integrating factor of  x^s.  Remember that the arbitrary constant that you get in the solution might depend on  s.  So I like
−
\qquad\text{if }n\ne m.</math>
+
to write  C(s) for the arbitrary constant.  You will need to show that  C(s) must be zero by showing that
  
−
The key here is to use the change of variables
+
W(0,s)=0.
  
−
<math>x=\cos\theta</math>
+
This follows from one of the boundary conditions given in the problem.
  
−
and convert the integral to one in  x  over the interval [-1,1], where you can use the orthogonality of the Legendre Polynomials.
+
Page 209 Problem 5:
 +
I've done this problem 5 times over, (yes, again after reading Prof. Bell's notes) and I keep getting 2/3 out of the integral rather than the necessary 0. Whether it's cos^3 or x^3, I don't see any way to get anything other than Pm*Pn/3*(-1 - 1) = 2/3 * PnPm rather than the 0 we're looking for. Or do I just not understand orthogonality??
 +
 +
New Question: How is everyone else "showing orthogonality?" Are you evaluating all the available (11) polynomials, doing just a few and making a statement, or have you figured out how to integrate (11) itself?
  
−
p. 216, problems 1 and 3 ask you to expand a given function in terms of Legendre Polynomials.  Here, you will use the fact that if  Q(x)  is a polynomial of degree  N, then
+
From Bell:  My hint about p. 209, problem 5 was to make a change of variables,
  
−
<math>Q(x)=\sum_{n=0}^N c_nP_n(x)</math>
+
x = cos theta.
  
−
where the coefficients c_n  are computed via orthogonality:
+
After that, you'll reduce the orthogonality of P_n(cos theta) with respect
 +
to the weight function Sin(theta) on [0,pi] to the orthogonality of the
 +
Legendre Polynomials on [-1,1]. The orthogonality of the Legendre Polynomials on the interval [-1,1] is a given fact, so you shouldn't be computing actual integrals to verify the orthogonality for the first few Legendre Polynomials listed on p. 180.
  
−
<math>\int_{-1}^1 Q(x)P_m(x)\ dx=c_m\int_{-1}^1 P_m(x)P_m(x)\ dx.
+
Question Page 209 Problem 7:
−
</math>
+
  
−
You will need to use the fact given on page 212 that
+
Where does the <math>\pi</math> come from in this solution?
  
−
<math>\int_{-1}^1 P_m(x)^2\ dx=\frac{2}{2m+1}.</math>
+
Answer: When you do the positive lambda case, you get A = 0 and let B = 1 => Sin(5*mu) = 0. If mu = m*pi/5, this equation is true. I let B=1 because we cannot have both A and B = 0.
  
−
p. 216, 5. asks you to show that if f(x)  is even, then all the odd
+
Question: Page 209, Problem 17:
−
coefficients in its Legendre expansion must vanish, i.e., that
+
For the given equation, shouldn't p=1, q=16, r=1? These values differ from the textbook's values.
  
−
<math>\int_{-1}^1 f(x)P_n(x)\ dx=0</math>
+
Answer:  If that were the case, then the equation would be
  
−
if  n  is odd.  Recall that if n is odd,  P_n  is odd.  An even times an odd is a ... etc.
+
[py']' + (q+ lambda r) y =
  
−
Page 209, Question 17:
+
[1 y']' + (16 + lambda) y =
−
For the given equation, shouldn't p=1, q=16, r=1? These values differ from the textbook's values.
+
 
 +
y" + (16 + lambda) y = 0
 +
 
 +
and it ain't.  You need to use problem  6  in the same section to get p,q, and r.
 +
 
 +
Question: Why isn't q=pg=16*exp(8x)?
 +
 
 +
Answer:  Here is the idea of problem 6.  We have the equation
 +
 
 +
<math>y'' + 8 y' + (\lambda + 16)y=0.</math>
 +
 
 +
Multiply that equation by p(x).  You get
 +
 
 +
<math>py'' + 8p y' + (\lambda p+ 16p)y=0.</math>
 +
 
 +
If this were in Sturm-Liouville form, it would look like
 +
 
 +
<math>[py']'+ (q + \lambda r) y = </math>
 +
 
 +
<math>py'' + p'y' + (q+ \lambda r) y = 0.</math>
 +
 
 +
By comparing those two, we see that we need
 +
 
 +
<math>p'=8p</math>
 +
 
 +
and q=16p  and r=p.  Solving the ODE for p yields
 +
 
 +
<math>p(x)=e^{8x}.</math>
 +
 
 +
(We can take the arbitrary constant in the solution
 +
to be a convenient value because we just want one
 +
p(x) that has this property.)
 +
 
 +
Finally, we get
 +
 
 +
<math>p(x)=e^{8x},\quad q(x)=16e^{8x},\quad\text{and }r(x)=e^{8x}.</math>
 +
 
 +
Hmmm. I see what you mean.  I think the answer in the back of the book is wrong.--[[User:Bell|Steve Bell]] 12:07, 23 October 2010 (UTC)
 +
 
 +
p. 216, #s 1 and 3:
 +
 
 +
I am using the hints given, but I'm still not sure I'm doing this correctly. For example, for #1, I've calculated c4 as c4 = [((2*4)+1)/2] * integral from -1 to 1 of (7x^4-6x^2)(P_4(x)) dx, where P_4= (1/8)(35x^4-30x^2+3), and then I would do something similar for C3, C2, C1, and C0. But if I find C4 this way, I'll get an answer where there's an x^9 term, and I don't see how to get an answer in terms for P_4 and P_1, like there is in the back of the book.
 +
 
 +
Answer: These are definite integrals carried out from -1 to 1, so you should get a numerical value for the solution, which will be the coefficient cm of the mth legendre polynomial. For instance, for m=0, you'll have c_0 =(2*0+1)/2 * integral from -1 to 1 of (7x^4-6x^2)*P_0(x)dx, where P_0(x)=1. The solution of that integral is -3/5, which matches the value for the coefficient of P_0(x) in the back of the book. --[[User:Jkusnick|Jkusnick]] 08:47, 25 October 2010 (UTC)
 +
 
 +
p. 209 #7 & #17:
 +
 
 +
To verify orthogonality, do you just use the Theorem, or do you have to do the integral?
 +
 
 +
Answer:  The theorem says the eigenfunctions are orthogonal.  However, to VERIFY that, you'll have to compute the integrals.
 +
 
 +
p. 209, #17
 +
 
 +
Answer:
 +
 +
You can solve for lambda in a way similar to the way Prof. Bell did it at the beginning of class on 10/20/10.  From the original problem you get:  r^2+8r+(16+lamda)=0.  So r = sqrt(-lambda)-4.  As usual, the lambda >0 case gives you non-zero solutions.  Since our root is complex of the form -4 plus/minus i*sqrt(lambda), where mu = sqrt(lambda).  We can find lambda that allow non-zero solutions like in 7 by using the general form of:
 +
 
 +
y = exp(-4*x)*(A*cos(mu*x)+B*sin(mu*x)).  Plug in boundary conditions and Ta-Da.
  
 
[[2010 MA 527 Bell|Back to the MA 527 start page]]  
 
[[2010 MA 527 Bell|Back to the MA 527 start page]]  

Latest revision as of 10:00, 27 October 2010

Homework 9 Collaboration Area

Click on

Legendre hints from Bell

for some hints about Legendre Polynomials.

Question Page 597, Problem 5:

What do we do with the x in the first term of this problem?

Answer: When you do a Laplace transform wrt t, the x floats along like when you do d/dt(x*t).

Additional question: I get to this point after taking the Laplace Transform:

d/dx*W(x,s) + s/x*W(x,s) = 1/s^2.

Is this correct?

Answer: Yes, that's what I got. Check out lecture 23 around the 28 minute mark to see what to do next.

Message from Bell: When you solve that ODE as a first order linear ODE in the x variable, you should get an integrating factor of x^s. Remember that the arbitrary constant that you get in the solution might depend on s. So I like to write C(s) for the arbitrary constant. You will need to show that C(s) must be zero by showing that

W(0,s)=0.

This follows from one of the boundary conditions given in the problem.

Page 209 Problem 5: I've done this problem 5 times over, (yes, again after reading Prof. Bell's notes) and I keep getting 2/3 out of the integral rather than the necessary 0. Whether it's cos^3 or x^3, I don't see any way to get anything other than Pm*Pn/3*(-1 - 1) = 2/3 * PnPm rather than the 0 we're looking for. Or do I just not understand orthogonality??

New Question: How is everyone else "showing orthogonality?" Are you evaluating all the available (11) polynomials, doing just a few and making a statement, or have you figured out how to integrate (11) itself?

From Bell: My hint about p. 209, problem 5 was to make a change of variables,

x = cos theta.

After that, you'll reduce the orthogonality of P_n(cos theta) with respect to the weight function Sin(theta) on [0,pi] to the orthogonality of the Legendre Polynomials on [-1,1]. The orthogonality of the Legendre Polynomials on the interval [-1,1] is a given fact, so you shouldn't be computing actual integrals to verify the orthogonality for the first few Legendre Polynomials listed on p. 180.

Question Page 209 Problem 7:

Where does the $ \pi $ come from in this solution?

Answer: When you do the positive lambda case, you get A = 0 and let B = 1 => Sin(5*mu) = 0. If mu = m*pi/5, this equation is true. I let B=1 because we cannot have both A and B = 0.

Question: Page 209, Problem 17: For the given equation, shouldn't p=1, q=16, r=1? These values differ from the textbook's values.

Answer: If that were the case, then the equation would be

[py']' + (q+ lambda r) y =

[1 y']' + (16 + lambda) y =

y" + (16 + lambda) y = 0

and it ain't. You need to use problem 6 in the same section to get p,q, and r.

Question: Why isn't q=pg=16*exp(8x)?

Answer: Here is the idea of problem 6. We have the equation

$ y'' + 8 y' + (\lambda + 16)y=0. $

Multiply that equation by p(x). You get

$ py'' + 8p y' + (\lambda p+ 16p)y=0. $

If this were in Sturm-Liouville form, it would look like

$ [py']'+ (q + \lambda r) y = $

$ py'' + p'y' + (q+ \lambda r) y = 0. $

By comparing those two, we see that we need

$ p'=8p $

and q=16p and r=p. Solving the ODE for p yields

$ p(x)=e^{8x}. $

(We can take the arbitrary constant in the solution to be a convenient value because we just want one p(x) that has this property.)

Finally, we get

$ p(x)=e^{8x},\quad q(x)=16e^{8x},\quad\text{and }r(x)=e^{8x}. $

Hmmm. I see what you mean. I think the answer in the back of the book is wrong.--Steve Bell 12:07, 23 October 2010 (UTC)

p. 216, #s 1 and 3:

I am using the hints given, but I'm still not sure I'm doing this correctly. For example, for #1, I've calculated c4 as c4 = [((2*4)+1)/2] * integral from -1 to 1 of (7x^4-6x^2)(P_4(x)) dx, where P_4= (1/8)(35x^4-30x^2+3), and then I would do something similar for C3, C2, C1, and C0. But if I find C4 this way, I'll get an answer where there's an x^9 term, and I don't see how to get an answer in terms for P_4 and P_1, like there is in the back of the book.

Answer: These are definite integrals carried out from -1 to 1, so you should get a numerical value for the solution, which will be the coefficient cm of the mth legendre polynomial. For instance, for m=0, you'll have c_0 =(2*0+1)/2 * integral from -1 to 1 of (7x^4-6x^2)*P_0(x)dx, where P_0(x)=1. The solution of that integral is -3/5, which matches the value for the coefficient of P_0(x) in the back of the book. --Jkusnick 08:47, 25 October 2010 (UTC)

p. 209 #7 & #17:

To verify orthogonality, do you just use the Theorem, or do you have to do the integral?

Answer: The theorem says the eigenfunctions are orthogonal. However, to VERIFY that, you'll have to compute the integrals.

p. 209, #17

Answer:

You can solve for lambda in a way similar to the way Prof. Bell did it at the beginning of class on 10/20/10. From the original problem you get: r^2+8r+(16+lamda)=0. So r = sqrt(-lambda)-4. As usual, the lambda >0 case gives you non-zero solutions. Since our root is complex of the form -4 plus/minus i*sqrt(lambda), where mu = sqrt(lambda). We can find lambda that allow non-zero solutions like in 7 by using the general form of:

y = exp(-4*x)*(A*cos(mu*x)+B*sin(mu*x)). Plug in boundary conditions and Ta-Da.

Back to the MA 527 start page

To Rhea Course List

Retrieved from "https://www.projectrhea.org/rhea/index.php?title=HWK9MA527Fall2010&oldid=41382"
Shortcuts
Help Main Wiki Page Random Page Special Pages Log in
Search

Alumni Liaison

Ph.D. 2007, working on developing cool imaging technologies for digital cameras, camera phones, and video surveillance cameras.

Buyue Zhang