ACM 100b
Boundary value problems and the Sturm-Liouville ODE - part 2
Dan Meiron
Caltech
February 6, 2013
D. Meiron (Caltech)
ACM 100b - Methods of Applied Mathematics
February 6, 2013
1 / 25
Recap - boundary value problems
Last lecture we introduced the concept of boundary value problems We illustrated some of the issues by solving the heat equation by separation of variables 2 =D 2 0x 1 t x with homogenous boundary conditions (0, t ) = 0 (1, t ) = 0 t 0
and an initial condition (x , 0) = 0 (x , 0) This led us to a homogeneous boundary value problem of the form d 2X + 2 X = 0 0x 1 dx 2 which had a countably innite number of solutions for special values of : Xn (x ) = A sin(n x ) n = n n = 1, 2, . . .
D. Meiron (Caltech) ACM 100b - Methods of Applied Mathematics February 6, 2013 2 / 25
Recap - boundary value problems
The general solution to the heat problem was
(x , t ) =
n=1
Bn exp(n2 2 t ) sin(n x ).
which satises the boundary conditions, because the sines vanish at x = 0, 1 But there is also an initial condition to satisfy. At t = 0 we have some starting distribution of heat in the rod: (x , 0) = 0 (x ). In order to satisfy this condition we substitute t = 0 into
(x , t ) =
n=1
Bn exp(n2 2 t ) sin(n x ) to get
0 (x ) =
n=1
D. Meiron (Caltech)
Bn sin(n x )
February 6, 2013 3 / 25
ACM 100b - Methods of Applied Mathematics
Recap - boundary value problems
So we would have a solution that satises all the conditions if we could gure out the coefcients Bn in the expression
0 (x ) =
n=1
Bn sin(n x )
As promising as this looks, there are some unanswered questions: How does one determine Bn ? If you can determine Bn is there only one choice that works? Even if there is a unique choice of Bn can you show the series converges to 0 (x ) as n ? If it converges at t = 0 does it converge for t > 0?
D. Meiron (Caltech)
ACM 100b - Methods of Applied Mathematics
February 6, 2013
4 / 25
The Sturm-Liouville ODE
We will answer all these questions shortly. But what we want to emphasize right now is that the type of ODE problem we just solved is actually quite common. It turns out that the ODE we solved above in the x -direction is a special case of a second order ODE boundary value problem called the Sturm-Liouville problem. The Sturm-Liouville ODE is given by d dx p (x ) dy dx q (x )y (x ) + r (x )y (x ) = 0, a < x < b,
You will also see this ODE written as d dx p(x ) dy dx + q (x )y (x ) = r (x )y (x ), a < x < b,
This is called the positive denite form.
D. Meiron (Caltech) ACM 100b - Methods of Applied Mathematics February 6, 2013 5 / 25
Boundary conditions for the Sturm-Liouville ODE
The boundary conditions are homogeneous c1 y (a) + c2 y (a) = 0, d1 y (b) + d2 y (b) = 0 For the boundary conditions c1 y (a) + c2 y (a) = 0, d1 y (b) + d2 y (b) = 0 the coefcients c1 , c2 , d1 , d2 , are assumed to be real constants. The boundary conditions above are said to be separated because they provide conditions on only one end point at a time. Later on we will relax this a bit.
D. Meiron (Caltech) ACM 100b - Methods of Applied Mathematics February 6, 2013 6 / 25
The heat equation problem led to a Sturm-Liouville ODE
If we recall the boundary value problem we solved earlier for the heat equation d 2 X (x ) + 2X = 0 dx 2 X (0) = 0 X (1) = 0.
we see this problem is indeed an example of the Sturm-Liouville ODE: d dx p (x ) dy dx q (x )y (x ) + r (x )y = 0, a < x < b,
We have p = 1, q = 0 r (x ) = 1 and = 2 Well see some other examples shortly where the coefcients p, q , r are not just constants.
D. Meiron (Caltech) ACM 100b - Methods of Applied Mathematics February 6, 2013 7 / 25
The Sturm-Liouville ODE
We will assume in the ODE d dx p (x ) dy dx q (x )y (x ) + r (x )y = 0, a < x < b,
that the coefcient functions p(x ), q (x ) and r (x ) are all continuous in the interval a x b. We also assume that p (x ) is also continuous in this interval. Most importantly, we will assume that p(x ) and r (x ) are strictly positive over the interval a x b. And, as usual, there is no loss of generality if we restrict our attention to a specic interval so we will assume in what follows that a = 0 and b = 1.
D. Meiron (Caltech)
ACM 100b - Methods of Applied Mathematics
February 6, 2013
8 / 25
Regular vs. singular Sturm-Liouville ODE
Note we have already insisted in the S-L ODE d dx p (x ) dy dx q (x )y (x ) + r (x )y = 0, a < x < b,
that p(x ) and r (x ) be strictly positive There are a few more restrictions that we will impose and then relax later. First well insist the boundary conditions be of the separable form c1 y (a) + c2 y (a) = 0, d1 y (b) + d2 y (b) = 0 Second, we insist that the domain a x b be nite You can see that if p(x ) vanishes our ODE becomes singular But the ODE will also be singular if the domain is made innite A S-L problem on a nite domain with separable boundary conditions and p(x ) > 0 and w (x ) > 0 is called a regular Sturm-Liouville problem
D. Meiron (Caltech) ACM 100b - Methods of Applied Mathematics February 6, 2013 9 / 25
Another example of the S-L ODE - the Bessel equation
We will examine a Sturm-Liouville ODE that does not have simple solutions like sines and cosines This is the Bessel equation It comes up when we solve the heat equation in cylindrical coordinates as we will show later The Bessel ODE is given by d 2 y (x ) 1 dy (x ) m2 2 + + x dx dx 2 x2 y (x ) = 0 x1 x x2 .
Note as written its not in the typical S-L form which is d dx p(x ) d y (x ) + q (x )y (x ) = r (x )y (x ) dx
D. Meiron (Caltech)
ACM 100b - Methods of Applied Mathematics
February 6, 2013
10 / 25
The Bessel equation in S-L form
But we can rewrite the ODE so that it is in S-L form Multiply each term of the ODE d 2 y (x ) 1 dy (x ) m2 2 + + x dx dx 2 x2 by x : x m2 d 2 y (x ) dy (x ) 2 + + x dx x dx 2 y (x ) = 0 x1 x x2 . y (x ) = 0 x1 x x2 .
We can write this in the form d dy m2 x y (x ) + 2 xy (x ) = 0 dx dx x Now we see this is in S-L form with m2 r (x ) = x x Note we replaced by 2 but this is just for convenience p(x ) = x q (x ) =
D. Meiron (Caltech) ACM 100b - Methods of Applied Mathematics February 6, 2013 11 / 25
Boundary conditions for the Bessel equation
Because m2 r (x ) = x x we see that p(x ) > 0 and r (x ) > 0 except where x = 0 which is a singular point So for now well take our domain x1 x x2 to be 1 x 2 This is safely away from x = 0 so the ODE is nonsingular For boundary conditions well take p(x ) = x q (x ) = y (1) = 0 y (2) = 0.
which are of separable type. This problem is now a regular S-L ODE problem Well take m = 0 for simplicity d dx x dy dx + 2 xy (x ) = 0. 1 x 2 y (1) = 0 y (2) = 0
February 6, 2013 12 / 25
D. Meiron (Caltech)
ACM 100b - Methods of Applied Mathematics
The solutions of the Bessel equation
The solutions of this ODE are new functions called Bessel functions They are not simple functions likes sines and cosines but we will see shortly that they share some important properties with sines and cosines. The ODE is second order so has two linearly independent solutions It has a regular singular point at x = 0. The Frobenius theory tells us there is one singular solution that blows up at x = 0 and one regular solution that does not blow up at the origin The regular solution is labeled J0 (x ) The singular solution is labeled Y0 (x ) Note that for our purposes since our domain is 1 x 2 both of these linearly independent solutions are smooth over this interval.
D. Meiron (Caltech) ACM 100b - Methods of Applied Mathematics February 6, 2013 13 / 25
Solution of the Bessel S-L problem
For our purposes all we need to know is that J0 and Y0 are linearly independent The general solution to d dx is y (x ) = c1 J0 (x ) + c2 Y0 (x ). Now in order to satisfy the boundary conditions we must have C1 J0 (x1 ) + C2 Y0 (x1 ) = 0, C1 J0 (x2 ) + C2 Y0 (x2 ) = 0. or using that x1 = 1 and x2 = 2 we have C1 J0 () + C2 Y0 () = 0, C1 J0 (2) + C2 Y0 (2) = 0.
D. Meiron (Caltech) ACM 100b - Methods of Applied Mathematics February 6, 2013 14 / 25
dy dx
+ 2 xy (x ) = 0.
Homogeneous solutions of the Bessel S-L problem
Note this system of equations is homogeneous So in general there is only the trivial solution C1 = C2 = 0 But if for some special values of we could get the determinant J0 () Y0 () J0 (2) Y0 (2) to vanish, we could get nontrivial special solutions. But because the system is homogeneous they would not be unique. So the question is - does this determinant () = J0 () Y0 () = J0 ()Y0 (2) Y0 ()J0 (2) J0 (2) Y0 (2)
vanish for some values of ?
D. Meiron (Caltech) ACM 100b - Methods of Applied Mathematics February 6, 2013 15 / 25
The determinant for our S-L problem
It is not easy to see where this determinant () = J0 ()Y0 (2) Y0 ()J0 (2) might vanish The Bessel functions are not simple things like sines and cosines. So we resort to computing it numerically and evaluating the determinant for various values of .
D. Meiron (Caltech)
ACM 100b - Methods of Applied Mathematics
February 6, 2013
16 / 25
Plotting the determinant
The determinant () is plotted below.
D. Meiron (Caltech)
ACM 100b - Methods of Applied Mathematics
February 6, 2013
17 / 25
Locations of the roots of the determinant
As can be seen, the determinant oscillates and crosses zero at various values of . At these values of , we expect we can get nontrivial solutions that satisfy the boundary conditions. While these are not simple values, like = n, n = 1, 2, 3 . . ., we see that there is some structure that can be seen in the locations where the determinant vanishes. Below we calculate the location of the rst 20 zeros of the determinant:
D. Meiron (Caltech)
ACM 100b - Methods of Applied Mathematics
February 6, 2013
18 / 25
The zeros take on a simple pattern as n
If we calculate the crossings of adjacent zeroes and see how far apart they are, we can see that the spacing of adjacent zeroes approaches a constant. That constant seems to be getting close to . In fact the numbers themselves seem to be approaching = n for n large This is similar to the values of we calculated when we solved the heat equation. It will turn out this is not an accident.
D. Meiron (Caltech)
ACM 100b - Methods of Applied Mathematics
February 6, 2013
19 / 25
The solutions look sort of like sines
Additional similarities with sines and cosines can be seen when we look at the actual solutions. When the determinant crosses zero we can get a solution (unique up to some multiplicative constant). The values of y (x ) are plotted in the gures below for increasing values of .
D. Meiron (Caltech)
ACM 100b - Methods of Applied Mathematics
February 6, 2013
20 / 25
The solution for the lowest value of
Figure: The solution corresponding to the rst value of for which the determinant vanishes
D. Meiron (Caltech) ACM 100b - Methods of Applied Mathematics February 6, 2013 21 / 25
The solution for the 6th root
Figure: The solution corresponding to the sixth value of for which the determinant vanishes
D. Meiron (Caltech) ACM 100b - Methods of Applied Mathematics February 6, 2013 22 / 25
The solution for the 11th root
Figure: The solution corresponding to the 11th value of for which the determinant vanishes
D. Meiron (Caltech) ACM 100b - Methods of Applied Mathematics February 6, 2013 23 / 25
The solution for the 16th root
Figure: The solution corresponding to the 16th value of for which the determinant vanishes
D. Meiron (Caltech) ACM 100b - Methods of Applied Mathematics February 6, 2013 24 / 25
Overview of the solutions
We can see that these solutions look increasingly sinusoidal Later in a quantitative sense which we will make precise, they do get increasingly close to sin m x (times an envelope function which modulates the amplitude) Here m describes the m + 1th value of for which the determinant vanishes. This too is not an accident Its a general feature of the S-L problem which we will demonstrate. In order to see why this ODE is so special, we rst have to derive an important identity which allows us to understand the properties of the solutions of this boundary value problem.
D. Meiron (Caltech)
ACM 100b - Methods of Applied Mathematics
February 6, 2013
25 / 25
