Simple ODE Question kinda.

  • Thread starter STEMucator
  • Start date
  • #1
STEMucator
Homework Helper
2,075
140

Homework Statement



So yeah, my first time playing with ODEs, how exciting. So my prof gave us a few suggested exercises and I want to know whether I'm actually doing these properly or not. The question and all relevant things will be included in the picture below :

http://gyazo.com/b3be5b5d56ce2201cf88bce5ea8d5838

Homework Equations





The Attempt at a Solution



(1a) So! I figured I should probably break this up into cases since I'm dealing with a piecewise function.

Case #1 : Suppose x>c, then y(x) = (x-c)^2 and y' = 2(x-c). Now we check that this satisfies our equation!

y' = 2[itex]\sqrt{y}[/itex]
2(x-c) = 2[itex]\sqrt{(x-c)^2}[/itex]
2(x-c) = 2(x-c)
1 = 1

Case #2 : Suppose x ≤ c, then y(x) = 0 and y' = 0. Now we check that this also satisfies our equation.

y' = 2[itex]\sqrt{y}[/itex]
y' = 2[itex]\sqrt{0}[/itex]
0 = 0

Check and check. ∴ y(x) satisfies our ODE for all x as desired.

Now for the figure, I've never drawn a direction field before, so I got the computer to do it for me for the first time. I know of course that the equilibrium point occurs at y' = 0 ( In this case it turns out after solving we get y = 0 ). Here's a picture of the direction field when we have the IVP y(0) = 0 :

http://gyazo.com/443b52dec28b53e61b3c7ffb3c0b4369

I'm not sure how this demonstrates that we have infinitely many solutions, perhaps someone could elaborate for me?

I'll save my attempt at (1b) for after I know what I'm doing with part (1a).
 

Answers and Replies

  • #2
jbunniii
Science Advisor
Homework Helper
Insights Author
Gold Member
3,447
227
Your proof for case #1 is fine. Your proof for case #2 is really only valid for x < c. This is because x = c is an endpoint of the intervals where the two pieces of the function are joined. To see what can go wrong, consider f(x) = |x|, which has derivative equal to -1 for negative x, +1 for positive x, but is not differentiable at x = 0.

Therefore you need to provide a proof that y is differentiable at c, and that y'(c) = 0.
 
  • #3
jbunniii
Science Advisor
Homework Helper
Insights Author
Gold Member
3,447
227
As for the figure, I'm not much into differential equations, so I'm not sure what it is you have plotted, nor how (or if) it illustrates that the differential equation has infinitely many solutions. Can you explain in words what the plot is depicting?

I'm also not sure why you need a plot in order to conclude that there are infinitely many solutions. If you show that your y(x) is a solution:
[tex]y(x) = \begin{cases}
0 & \text{for }x \leq c \\
(x - c)^2 & \text{for }x > c
\end{cases}[/tex]
then aren't there infinitely many values of [itex]c[/itex] that satisfy [itex]y' = 2\sqrt{y}[/itex] and [itex]y(0) = 0[/itex]?
 
Last edited:
  • #4
STEMucator
Homework Helper
2,075
140
Your proof for case #1 is fine. Your proof for case #2 is really only valid for x < c. This is because x = c is an endpoint of the intervals where the two pieces of the function are joined. To see what can go wrong, consider f(x) = |x|, which has derivative equal to -1 for negative x, +1 for positive x, but is not differentiable at x = 0.

Therefore you need to provide a proof that y is differentiable at c, and that y'(c) = 0.
Ah, so my case #2 is really a case #2 and #3. So let me correct that :

Case #2 : Suppose x < c, then y(x) = 0 and y' = 0. Now we check that this also satisfies our equation.

y' = 2sqrt(y)
y' = 2sqrt(0)
0 = 0

Case #3 : Suppose x = c, then we must show that y is differentiable at c.

y'(c) = [itex]lim_{h→0} \frac{y(c+h) - y(c)}{h} = lim_{h→0} \frac{h^2 - 0}{h} = 0[/itex]

Now would I be able to conclude the exact same thing as in Case 2 or am I misinterpreting what you said?

Also the graph I posted represents the direction field associated with the ODE as well as its equilibrium point at y = 0.
 
  • #6
jbunniii
Science Advisor
Homework Helper
Insights Author
Gold Member
3,447
227
Case #3 : Suppose x = c, then we must show that y is differentiable at c.

y'(c) = [itex]lim_{h→0} \frac{y(c+h) - y(c)}{h} = lim_{h→0} \frac{h^2 - 0}{h} = 0[/itex]
You have to be slightly careful here. The expression [itex](h^2 - 0)/h[/itex] is only correct for positive [itex]h[/itex]. So what you have shown is
[tex]\lim_{h \rightarrow 0^+}\frac{y(c+h) - y(c)}{h} = 0[/tex]
You also need to show that
[tex]\lim_{h \rightarrow 0^-}\frac{y(c+h) - y(c)}{h} = 0[/tex]
in order to conclude that
[tex]\lim_{h \rightarrow 0}\frac{y(c+h) - y(c)}{h} = 0[/tex]
Of course, this is obviously true for this function, but you definitely need to check the limits from both sides because for some functions they won't match, so the function won't be differentiable.
 
  • #7
jbunniii
Science Advisor
Homework Helper
Insights Author
Gold Member
3,447
227
Also, here's a much better image of whats going on at the initial point y(0) = 0

http://gyazo.com/fc97b2c612ad68b1565d1749812aeeb4
So what is this plot telling me? How does it show that there are infinitely many solutions? I am not being pedantic here, I don't know what I am looking at. Explain in words what a direction field is, and why it demonstrates what you are trying to show.
 
  • #8
STEMucator
Homework Helper
2,075
140
You have to be slightly careful here. The expression [itex](h^2 - 0)/h[/itex] is only correct for positive [itex]h[/itex]. So what you have shown is
[tex]\lim_{h \rightarrow 0^+}\frac{y(c+h) - y(c)}{h} = 0[/tex]
You also need to show that
[tex]\lim_{h \rightarrow 0^-}\frac{y(c+h) - y(c)}{h} = 0[/tex]
in order to conclude that
[tex]\lim_{h \rightarrow 0}\frac{y(c+h) - y(c)}{h} = 0[/tex]
Of course, this is obviously true for this function, but you definitely need to check the limits from both sides because for some functions they won't match, so the function won't be differentiable.
Yes of course, fair enough. So for argument sake saving some latex, since the limit would match from both sides, we could conclude that y(c) = 0 = y'(c). Then it would be literally a copy and pasta of the case 2 proof.

So onto the more important matter of the model.

The direction field represents the solutions of the differential equation which vary according to the constant c. So, depending on c, the graph will fit certain portions of the direction field better.
 
  • #9
jbunniii
Science Advisor
Homework Helper
Insights Author
Gold Member
3,447
227
Yes of course, fair enough. So for argument sake saving some latex, since the limit would match from both sides, we could conclude that y(c) = 0 = y'(c). Then it would be literally a copy and pasta of the case 2 proof.
Yes.

The direction field represents the solutions of the differential equation which vary according to the constant c. So, depending on c, the graph will fit certain portions of the direction field better.
OK, I'm an analysis guy and not a differential equations guy. I like my proofs to be proofs instead of pictures. When I hear the word "illustrate" I think of it as a synonym for "demonstrate" or "prove", whereas to me, your plot merely SUGGESTS that there may be infinitely many solutions. However, you are in a better position than I am to know what your instructor expects for this question.

If someone asked me to demonstrate that there are infinitely many solutions satisfying [itex]y' = 2\sqrt{y}[/itex] and [itex]y(0) = 0[/itex], I would say: well, we just showed that
[tex]
y(x) = \begin{cases}
0 & \text{for }x \leq c \\
(x - c)^2 & \text{for }x > c
\end{cases}
[/tex]
satisfies [itex]y' = 2\sqrt{y}[/itex] for any [itex]c[/itex]. Furthermore, it also satisfies [itex]y(0) = 0[/itex] for any [itex]c \geq 0[/itex]. This gives us infinitely many solutions.

However, I would probably lose points because there is no picture in my above proof...
 
  • #10
STEMucator
Homework Helper
2,075
140
Yes.


OK, I'm an analysis guy and not a differential equations guy. I like my proofs to be proofs instead of pictures. When I hear the word "illustrate" I think of it as a synonym for "demonstrate" or "prove", whereas to me, your plot merely SUGGESTS that there may be infinitely many solutions. However, you are in a better position than I am to know what your instructor expects for this question.

If someone asked me to demonstrate that there are infinitely many solutions satisfying [itex]y' = 2\sqrt{y}[/itex] and [itex]y(0) = 0[/itex], I would say: well, we just showed that
[tex]
y(x) = \begin{cases}
0 & \text{for }x \leq c \\
(x - c)^2 & \text{for }x > c
\end{cases}
[/tex]
satisfies [itex]y' = 2\sqrt{y}[/itex] for any [itex]c[/itex]. Furthermore, it also satisfies [itex]y(0) = 0[/itex] for any [itex]c \geq 0[/itex]. This gives us infinitely many solutions.

However, I would probably lose points because there is no picture in my above proof...
I also prefer the rigor associated with analysis, this is a very new and... not so well defined topic for me.

I see what you're saying, and I see how it associates itself with the model, but I suppose that asking my professor about this would be a better idea.

Part b seems pretty obvious now though.

(1b) So we are given y' = 2sqrt(y), y(0) = b and obviously y'(0) = 1.

y' = 2sqrt(y)
1 = 2sqrt(b)
1/2 = sqrt(b)
1/4 = b

So for (i), we have no solutions if b < 0 and for (ii) If b = 1/4 we have a unique solution.

I think that's it.
 
  • #11
jbunniii
Science Advisor
Homework Helper
Insights Author
Gold Member
3,447
227
(1b) So we are given y' = 2sqrt(y), y(0) = b and obviously y'(0) = 1.
Why is y'(0) obviously 1?
 
  • #12
STEMucator
Homework Helper
2,075
140
Why is y'(0) obviously 1?
Nice and early in the morning, best time to figure this out.

If y(0) = b, then consider :

y'(0) = [itex]lim_{h→0^+} \frac{y(0+h) - y(0)}{h}[/itex]

and

y'(0) = [itex]lim_{h→0^-} \frac{y(0+h) - y(0)}{h}[/itex]

Is this what you meant?
 
  • #13
jbunniii
Science Advisor
Homework Helper
Insights Author
Gold Member
3,447
227
Nice and early in the morning, best time to figure this out.

If y(0) = b, then consider :

y'(0) = [itex]lim_{h→0^+} \frac{y(0+h) - y(0)}{h}[/itex]

and

y'(0) = [itex]lim_{h→0^-} \frac{y(0+h) - y(0)}{h}[/itex]

Is this what you meant?
No, you said we are given y' = 2sqrt(y), and y(0) = b. But why would this imply that y'(0) = 1? I don't think it does.
 
  • #14
STEMucator
Homework Helper
2,075
140
No, you said we are given y' = 2sqrt(y), and y(0) = b. But why would this imply that y'(0) = 1? I don't think it does.
Ohh wait a second here, I don't even need to know y'(0) because I have y(x) come to my rescue I believe.
 
  • #15
STEMucator
Homework Helper
2,075
140
My only question is about the algebra then.

So I'm given y' = 2sqrt(y), y(0) = b and I know from part (1a) that y(x) is a general solution to this equation.

So :

For x > c :
b = (0-c)^2
sqrt(b) = c

So we only have a solution if b ≥ 0

For x ≤ c :
b = 0

Thus we would have no solution if b < 0 and a unique solution only if b = 0.
 
  • #16
jbunniii
Science Advisor
Homework Helper
Insights Author
Gold Member
3,447
227
Thus we would have no solution if b < 0 and a unique solution only if b = 0.
But didn't you just show in part (a) that there are infinitely many solutions when b = 0?

I agree that there are no solutions if y(0) = b < 0, for in that case y'(0) = 2sqrt(y(0)) = 2sqrt(b) is undefined. (Assuming we're working with real-valued functions.)
 
  • #17
STEMucator
Homework Helper
2,075
140
But didn't you just show in part (a) that there are infinitely many solutions when b = 0?

I agree that there are no solutions if y(0) = b < 0, for in that case y'(0) = 2sqrt(y(0)) = 2sqrt(b) is undefined. (Assuming we're working with real-valued functions.)
Okay, so there are no solutions if b < 0 since we need the sqrt to be defined. There are infinitely many solutions when b = 0 because y(0) = 0 = b from part (1a) showed that there are infinite solutions.

So (i) is answered and that leaves only one case left, that is when b>0. Though there could be far too many possibilities considering when b>0 for there to be a unique solution?
 
  • #18
jbunniii
Science Advisor
Homework Helper
Insights Author
Gold Member
3,447
227
So (i) is answered and that leaves only one case left, that is when b>0. Though there could be far too many possibilities considering when b>0 for there to be a unique solution?
Do you have any theorems that can tell you what conditions will guarantee a unique solution?
 
  • #21
STEMucator
Homework Helper
2,075
140
Unfortunately that doesn't apply because your differential equation is not of that form.

Do you have this theorem of Picard:

http://student.fizika.org/~aficnar/Korisno/Picard.pdf
Sadly I do not have this theorem accessible to me at the moment.

What about if :

y' = 2sqrt(y)

So

y'(0) = 2sqrt(b)

Thus

y(0) = 2sqrt(b)x + b

Would this not be unique ^ Since from part a

y'(0) = 0
y(0) = 0

So for b < 0, we have no solutions. For b = 0 we have infinitely many, and then the only unique solution according to the IVP for part b would be y(0) = 2sqrt(b)x + b?
 
  • #22
jbunniii
Science Advisor
Homework Helper
Insights Author
Gold Member
3,447
227
Sadly I do not have this theorem accessible to me at the moment.

What about if :

y' = 2sqrt(y)

So

y'(0) = 2sqrt(b)

Thus

y(0) = 2sqrt(b)x + b
How did you get that? It doesn't even make any sense - the right side depends on x, but the left side does not.

I'm not sure if you will be able to prove uniqueness without some kind of nontrivial theorem. Perhaps the question wants you to look at the plot and argue that it "suggests" uniqueness when b > 0? I recommend checking with your instructor to see what is intended here.

What you should be able to do, however, is recognize why only one function of the form
[tex]y(x) = \begin{cases}
0 & \text{for }x \leq c \\
(x - c)^2 & \text{for }x > c
\end{cases}[/tex]
can be a solution if y(0) = b > 0, versus infinitely many solutions of this form if y(0) = 0.

But this doesn't prove overall uniqueness because you haven't ruled out that there could be solutions that are not of the form indicated above.
 
Last edited:
  • #23
STEMucator
Homework Helper
2,075
140
How did you get that? It doesn't even make any sense - the right side depends on x, but the left side does not.

I'm not sure if you will be able to prove uniqueness without some kind of nontrivial theorem. Perhaps the question wants you to look at the plot and argue that it "suggests" uniqueness when b > 0? I recommend checking with your instructor to see what is intended here.
I got a bit desperate trying to wrap my head around this, but I suppose asking my professor what is best would be a good idea in this case.
 
  • #24
jbunniii
Science Advisor
Homework Helper
Insights Author
Gold Member
3,447
227
I got a bit desperate trying to wrap my head around this, but I suppose asking my professor what is best would be a good idea in this case.
Yeah, that's what I would do.

I edited my post above, adding a partial step you can take, but I'm not sure if you have the machinery for a full proof for this problem.
 

Related Threads on Simple ODE Question kinda.

  • Last Post
Replies
2
Views
993
  • Last Post
Replies
0
Views
837
  • Last Post
Replies
11
Views
2K
  • Last Post
Replies
4
Views
986
Replies
13
Views
1K
  • Last Post
Replies
2
Views
2K
  • Last Post
Replies
6
Views
949
  • Last Post
Replies
3
Views
1K
  • Last Post
Replies
4
Views
1K
  • Last Post
Replies
2
Views
1K
Top