Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

2nd order PDE (I think?)

  1. Aug 18, 2010 #1
    Hi, I've spent days trying to solve some equations in a paper (referenced below) that describes it as a "straightforward, albeit lengthy integration," but I can't work out the "straightforward" bit. The notation is also odd, which doesn't seem to help my problem. Perhaps someone could help?

    I am trying to solve for the function [tex]\beta^{(0)}[/tex], which is a function of T and P (though P is held constant). Ultimately, I need a value of [tex]\beta^{(0)}[/tex] at a specific temperature T.

    I am given in the paper:
    [tex]\beta^{(0)L}=(\frac{\partial\beta^{(0)}}{\partial T})_{P}[/tex]
    [tex]\beta^{(0)J}=(\frac{\partial\beta^{(0)L}}{\partial T})_{P}+(2/T)\beta^{(0)L}[/tex]

    Also given,
    [tex]\beta^{(0)J}=6U_{5}+\frac{2U_{6}}{T}+\frac{U_{7}}{T^{2}}+\frac{526U_{8}}{T(T-263)^{3}}[/tex]
    where the U values are empirical constants.

    I (think) I managed to solve the following expression for [tex]\beta^{(0)}[/tex]:
    [tex]\beta^{(0)}=\int\beta^{(0)J}\frac{T}{1+2ln(T)}\partial T[/tex]
    but I cannot figure out how to integrate that expression.

    Thus, I also tried to express the functions as a PDE:
    [tex]\beta^{(0)J}=(\frac{\partial^{2}\beta^{(0)}}{\partial T^{2}})_{P}+\frac{2}{T}(\frac{\partial\beta^{(0)}}{\partial T})_{P}[/tex]
    and then substitute the U-series empirical expression of [tex]\beta^{(0)J}[/tex] and subtract it from each side. However, I am as at much of a loss to solve that expression for [tex]\beta^{(0)}[/tex] as the integral expression above.

    Can anyone help me out at all? These equations are important to work out some geochemical thermodynamics I need to set up an experiment.

    Reference:
    Rogers, P. S. Z. and Pitzer, K. S., High-Temperature Thermodynamic Properties of Aqueous Sodium-Sulfate Solutions. Journal of Physical Chemistry 85 (20), 2886 (1981).
     
  2. jcsd
  3. Aug 19, 2010 #2
    In fact, the method that comes into my mind is VERY lenghty:

    1) First, decompose the T(T-263)^3 term with Hermite theorem:

    [tex]\frac{526U_8}{T(T-263)^3}=\frac{a_1}{T}+\frac{a_2}{(T-263)^3}+\frac{a_3}{(T-263)^2}+\frac{a_4}{(T-263)}[/tex]

    To find the coefficients a_1 match the two sides, you should find (I made the calculations very quickly, so I don't guarantee that's correct)

    [tex]a_1=-\frac{526U_8}{(263)^3}[/tex]

    [tex]a_2=\frac{526U_8}{263}[/tex]

    [tex]a_3=-\frac{526U_8}{(263)^2}[/tex]

    [tex]a_4=\frac{2\cdot 526U_8}{(263)^3}[/tex]

    2) Now try the ansatz:

    [tex]\beta^{(0)L}=b_1T+b_2+b_3\log T+b_4/T+b_5/T^2+b_6/(T-263)+b_7/(T-263)^2+b_8/(T-263)^3[/tex]

    Calculate the derivative of this and substitute into the differential equation for [tex]\beta^{(0)L}[/tex], finding in this way the coefficients b_i.

    3) Finally, to find [tex]\beta^{(0)J}[/tex], simply integrate, which should be simple because [tex]\beta^{(0)L}[/tex] is a sum of things easy to integrate.

    4) I'm afraid this procedure will cost you half a day...!
     
  4. Aug 19, 2010 #3

    Mute

    User Avatar
    Homework Helper

    Multiply the second equation by an integrating factor [itex]\mu(T)[/itex], demanding that the right hand side is of the product rule form u v' + u' v. This gives an easy DE for mu, but you can also just check that [itex]\mu = T^2[/itex] will work. You thus have

    [tex]T^2 \beta^{(0)J} = \frac{\partial}{\partial T}(T^2\beta^{(0)L}).[/tex]

    Integrating,

    [tex]\beta^{(0)L}(T,P) = \left(\frac{T_0}{T}\right)^2\beta^{(0)L}(T_0,P) + \frac{1}{T^2} \int_{T_0}^T dt~t^2 \beta^{(0)J}(t)[/tex]
    where the pressure dependence enters as the initial condition on beta^L.

    Since [itex]\beta^{(0)L}[/itex] is just a derivative of [itex]\beta^{(0)}[/itex], you can just integrate again,

    [tex]\beta^{(0)}(T,P) = \beta^{(0)}(T_0,P) + T_0^2\beta^{(0)L}(T_0,P)\left(\frac{1}{T_0}-\frac{1}{T}\right) + \int_{T_0}^T \frac{d\tau}{\tau^2}~\int_{T_0}^\tau dt~t^2 \beta^{(0)J}(t)[/tex]

    This might not be the most pleasant set of integrals to do, but it looks like they should be doable in closed form.
     
    Last edited: Aug 19, 2010
Know someone interested in this topic? Share this thread via Reddit, Google+, Twitter, or Facebook




Similar Discussions: 2nd order PDE (I think?)
Loading...