2

Question: What form must $g(x)$ have in order that the following problem have a solution? $u_x+3u_y-u=1,u(x,3x)=g(x)$. If $g(x)$ has the required form, will there be more than one solution?

My attempt:

$u_x+3u_y-u=1$

First we need the slope and the characteristic lines.

The characteristic line is in the form of $bx-ay =d$ for the equation $au_x+bu_y+cu=f(x,y)$

Since $ b = 3$ and $ a = 1$, our slope is $\frac{3}{1} \rightarrow 3$

Our characteristic lines are

$3x-y = w$ and $y = z$

and using the change of variables, we have

$ x = \frac{1}{3}w+\frac{1}{3}z$ and $ z =y$

Using the chain rule, we have

$V_wW_x+V_zZ_x + 3(V_wW_y+V_zZ_y ) - v= 1$

and taking the partial derivatives, we have

$ W_x = 3, W_y = -1, Z_x = 0$ and $Z_y =1$

Our equation becomes

$V_z - v= 1$

$\frac{dv}{dz}-v=1$

$l(z) = e^{-\frac{1}{3}z}$

so multiplying the integrating factor we have

$e^{-\frac{1}{3}z}V_z - e^{-\frac{1}{3}z}v= e^{-\frac{1}{3}z}$

Taking the reverse product rule, we have

$v(e^{\frac{-1}{3}z})= \int \frac{1}{3}e^{\frac{-1}{3}z} $

Integrating with respect to z, we have

$v(e^{\frac{-1}{3}z})= -e^{\frac{-1}{3}z} +C(w)$

Dividing $(e^{\frac{-1}{3}z})$, we have

$v= -1 +C(w)e^{\frac{1}{3}z}$

Since $ z = y$

$v= -1 +C(w)e^{\frac{1}{3}y}$

and $y = 3x$

$v= -1 +C(w)e^{\frac{1}{3}3x}$

$v= -1 +C(w)e^{x}$

So I am stuck at the $w$ part. The only thing that I can think of is to substitute $ w =3x-y$ and I am not sure what to do. The answer to do is $ g(x) = -1+2e^x$ and I almost have it except I don't know where the 2 comes from.

$v= -1 +C(3x-y)e^{x}$

and from the condition $u(x,3x) = g(x)$

$ v = -1+C(3x-3x)e^{x}$

$v= =1+C(0)e^{x}$

what the!!

Edit: I was stressing out for no reason. It turns out that as long as $C(0) =k$ or we are in $C^1$ the solution of $g(x)$ has to be $ 1 +ke^x$. Now I feel like a dummy.

usukidoll
  • 2,074

1 Answers1

1

I would approach this differently. First, $u\equiv -1$ is a particular solution of the inhomogeneous equation, so we only need the general solution for homogeneous one, $u_x +3u_y-u = 0$. Take exponential ansatz $u=C e^v$ and simplify to $v_x+3v_y - 1=0$. Again, a particular solution is easy to find: $v(x,y)=x$. The general solution of $v_x+3v_y=0$ is any function that is constant on lines of slope $3$, which is $f(y-3x)$.

Collect together: $x+f(y-3x)$ solves $v_x+3v_y - 1=0$, hence $u(x,y)=C e^{x+f(y-3x)} -1$ solves the original equation. Tidy up: $$u(x,y) = e^x h(y-3x)-1,\quad h\text{ arbitrary}\tag{1}$$ which is the same what you found.

Since $u(x,3x) = e^x h(0)-1$, you need some pretty special values on that line to have a solution. But with these special values you have lots of solutions, since $h(t)$ for $t\ne 0$ can be anything.


In the special case $f(x,3x)=-1+2e^x$ we see that $h(0)=2$. This is the only constraint on $h$. You can take $h(t)\equiv 2$, or $h(t)=2+t$, or $h(t)=2e^{t}$... all these give different solutions of the PDE according to (1).

  • there are a ton of solutions. I'm supposed to find two of them when $g(x) = -1+2e^x$. That would mean that I need to find $g(x) = -1+2ke^x$ which was the constant k that I nabbed earlier. But the problem is ... how would I know if it's right since I can't be just picking at random... Oh I see... the 2 solutions that I need have to satisfy the conditions of $u(x,3x) = 1+2e^x$

    So do I have to solve the entire equation again with the condition of $u(x,3x) = 1+2e^x$?

    – usukidoll Sep 14 '14 at 06:02
  • @chinny84

    suppose I want to find two solutions that satisfy the PDE . I was looking over my notes and found a problem similar to what I am required to do. It had two equations with conditions and the g(x), but for this case I have

    $u(x,3x)=-1+ke^x$ and $u(x,3x) = 1+2e^x$. If I take one of the equations I will have $1+2e^x = -1+ke^x$ which would be $1+1+2e^x-ke^x = 0$ and results in $2e^x-ke^x = 0$. Factoring the $e^x$ I have $e^x(2-k) = 0$ so now what do I do?

    – usukidoll Sep 14 '14 at 08:22
  • @chinny84 sorry I meant $2+2e^x-ke^x =0 \rightarrow 2+e^x(2-k)=0$ – usukidoll Sep 14 '14 at 08:29
  • can you help me out with what I have mentioned in the comment? We need to find 2 solutions that would satisfy the equation, but all I got was to set each equation to each other and try to solve for k. That would be $2e^x-e^{x}k=-2 \rightarrow -e^{x}k=-2-2e^x $ ..... @Thursday – usukidoll Sep 14 '14 at 10:38
  • $\rightarrow k =\frac{-2}{-e^{x}} -\frac{-2e^x}{-e^{x}} \rightarrow k = \frac{2}{e^{x}}-2 \rightarrow k =2e^{-x}-2$. If I plug in the condition, it would still be x because there aren't any y variables whatsoever. @Thursday – usukidoll Sep 14 '14 at 10:40
  • so basically as long as we have something that has a 2 in it due to the fact that it is the only restriction we can take $2cos(t)$, $2e^t$... so how do I know that it satisfies the PDE? Do I just take the derivatives or resolve the whole problem? @Thursday – usukidoll Sep 15 '14 at 03:50
  • The formula (1) in my answer isn't just some random formula, is it? It was obtained from the PDE. You can check that any function it gives satisfies the PDE. Or, you can first plug in concrete $h$ and then check. –  Sep 15 '14 at 03:51
  • so I just choose a solution in the form of $-1+2e^x$ and then just take the derivatives of it? I'm a bit lost. I know to check I need to take partial derivatives depending on the order . Can you give me an example? @Thursday – usukidoll Sep 15 '14 at 03:54
  • Lost? You don't know how to find partial derivatives of $2e^x-1$? –  Sep 15 '14 at 03:56
  • of course I do... $u_y$ would be 0 due to the fact that there is no y variable. for $u_x = 2e^x$. which leaves me with $2e^x -(-1+2e^x) = 1$

    so... yup the $2e^x$ cancels out and the negative distributes all the way making it $1=1$

    – usukidoll Sep 15 '14 at 03:58
  • And so, $u_x +3u_y-u $ simplify to... $1$, as desired. –  Sep 15 '14 at 03:59
  • yes, but now do we need some other solution like $2+x$ then $u_x$ = 1 and so $1-(2e^x-1) = 1$... that's not going to get rid of the 2e^x if I do that. – usukidoll Sep 15 '14 at 04:00
  • No, $2+x$ is not a solution. Read the rest of my answer again. You can choose function $h$ to be $h(t)=2+t$; note I'm using another letter here, not to confuse the argument of $h$ with $x$ or $y$. With this choice, formula (1) produces $u(x,y)=e^x(y-3x+2)-1$. Go ahead and check that this is also a solution. –  Sep 15 '14 at 04:02
  • $u_x = e^x(y-3x+2)(-3) +(y-3x+2)e^x$...$u_y = e^x(y-3x+2)$ ..................$-2e^x(y-3x+2)+3e^x(y-3x+2)-e^x(y-3x+2)+1 = 1$...yes I do see that $ 1=1$ I must have gotten confused with your notation since my book does it differently. I have one more question that I'm going to post. I was wondering if you could check it out... I am stuck at taking the integration wrt z – usukidoll Sep 15 '14 at 04:14
  • wait I think it's best if I do the whole problem again. I had my z handwriting confused with the 2 oy! but thanks for your help. I'll tag you again if I get stuck. – usukidoll Sep 15 '14 at 04:21
  • http://math.stackexchange.com/questions/931921/find-the-particular-solution-of-u-x2u-y-4u-exy-satisfying-the-following-s @Thursday here is the question. – usukidoll Sep 15 '14 at 05:39