1. PF Contest - Win "Conquering the Physics GRE" book! Click Here to Enter
    Dismiss Notice
Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Separable Partial ODE

  1. Nov 4, 2008 #1
    Find the general solution of:

    x.du/dx - (1/2).y.du/dy=0

    I know that for it to be separable u(x,y)=X(x)Y(y)


    x.Y(y)dX/dx + (1/2).X(x)dY/dy = 0

    which cancels to:

    x/X(x).(dX/dx) = - y/2Y(y).dY/dy


    X(x) = -c. Y(y) c is some constant


    x/X(x).dX/dx = c

    from here i am stuck. Any help appreciated
  2. jcsd
  3. Nov 4, 2008 #2


    User Avatar
    Staff Emeritus
    Science Advisor
    Gold Member

    You lost a minus sign in your second equation.

    Also, separation of variables is your friend. Just because you learned a method of solving differential equations a while ago doesn't mean you should abandon it

    [tex]\frac{x}{X(x)}\frac{dX}{dx} = c[/tex]


    [tex]\frac{dX}{X} = c \frac{dx}{x}[/tex]
  4. Nov 4, 2008 #3
    Thanks I get
    c.ln(x) + C=ln(X) => X=Ax^C
    c.ln(y) +C= 1/2 ln(Y) => Ay^2C

    so u(x,y)=X.Y=A^2(y^2C)(x^C)

    ...the contants for integration are required right? Just wondering because the second part of the question would be easier if they weren't there:

    determine u(x,y) which satisfies the boundary conditon u(1,y)= 1 +sin y

    I would go about this by making 1+siny = A^2((1)^C(Y^2C))

    => A^2=[1 - Sin(y)]/[y^2C]

    so u(x,y) satisfied by the boundary condtion is:

    u(x,y)=[1 - Sin(y)]/[y^2C].x^C.y^2C

    y^2C cancel so:

    u(x,y)=[1 - Sin(y)].x^C

    BUT how do you get rid of that C or am i being stupid and iv made a previous mistake/mistakes

    EDIT :OK Long winded but from finding out X and Y I can find the original c and hence A by finding the ratio between X an Y which is x/y^2 so A is known. Then the boundary condtion is simply finding the constant of integration c with known A? Il leave the previous mess on the post just to check your opinion.
    Last edited: Nov 4, 2008
  5. Nov 4, 2008 #4


    User Avatar
    Staff Emeritus
    Science Advisor
    Gold Member

    You have some sloppy coefficient carrying there, but got lucky and ended up with an equivalent formula anyway

    \frac{dX}{X} = c \frac{dx}{x}


    [tex] ln(X) = cln(x) + C_x[/tex]
    which gives us

    [tex]X = x^c*A_x[/tex] where [tex] A_x = e^{C_x}[/tex]. This is important because the constant of integration obtained in the formula for Y need not be the same (it happens not to matter in this case, but you could shoot yourself in the foot). Similarly

    [tex]\frac{dY}{Y} = 2c\frac{dy}{y}[/tex] which gives us

    [tex]ln(Y) = 2cln(y) + C_y[/tex]

    and hence
    [tex]Y = A_yy^{2c}[/tex] where [tex]A_y = e^{C_y}[/tex].


    [tex]u(x,y) = A_xA_yx^cy^{2c}[/tex] we combine the multiplicative constants to get

    [tex]u(x,y) = Ax^cy^{2c}[/tex]

    But hark! This gives not the final solution. Because any linear combination of these is also a solution. In particular any infinite summation that happens to uniformly converge is also a solution. You know an infinite summation of 1+siny in terms of y, so try finding A's and c's such that when you sum up all the terms and set x=1 you get the power series of 1+siny
Know someone interested in this topic? Share this thread via Reddit, Google+, Twitter, or Facebook