<html>
  <head>
    <meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    On 02/22/2012 05:13 PM, Jed Brown wrote:
    <blockquote
cite="mid:CAM9tzSkqM7=TS8UYnFCdJcB7zdxjUBKgSLv_HCCnSb2+HvMbrQ@mail.gmail.com"
      type="cite">
      <div class="gmail_quote">On Wed, Feb 22, 2012 at 18:05, Patrick
        Alken <span dir="ltr">&lt;<a moz-do-not-send="true"
            href="mailto:patrick.alken@colorado.edu">patrick.alken@colorado.edu</a>&gt;</span>
        wrote:<br>
        <blockquote class="gmail_quote" style="margin:0 0 0
          .8ex;border-left:1px #ccc solid;padding-left:1ex">
          Hi all,<br>
          <br>
           I have been trying to track down a problem for a few days
          with solving a linear system arising from a finite differenced
          PDE in spherical coordinates. I found that PETSc managed to
          converge to a nice solution for my matrix at small grid sizes
          and everything looks pretty good.<br>
          <br>
           But when I try larger more realistic grid sizes, PETSc fails
          to converge. After trying with another direct solver library,
          I found that the direct solver found a solution which exactly
          solves the matrix equation,</blockquote>
        <div><br>
        </div>
        <div>This never happens, so what do you mean? You compute the
          residual and it's similar to what you expect the rounding
          error to be?</div>
      </div>
    </blockquote>
    <br>
    Yes I mean the direct solver residual is around 10e-15. The PETSc
    residual is 4e00<br>
    <br>
    <blockquote
cite="mid:CAM9tzSkqM7=TS8UYnFCdJcB7zdxjUBKgSLv_HCCnSb2+HvMbrQ@mail.gmail.com"
      type="cite">
      <div class="gmail_quote">
        <div> </div>
        <blockquote class="gmail_quote" style="margin:0 0 0
          .8ex;border-left:1px #ccc solid;padding-left:1ex"> but when
          plotting the solution, I see that it oscillates rapidly
          between the grid points and therefore isn't a satisfactory
          solution. (At smaller grids the solution is nice and smooth)<br>
        </blockquote>
        <div><br>
        </div>
        <div>What sort of PDE are you solving?</div>
      </div>
    </blockquote>
    <br>
    The PDE is:<br>
    <br>
    grad(f) . B = g<br>
    <br>
    where B is a known vector field, g is a known scalar function, and f
    is the unknown scalar function to be determined (I am discretizing
    this equation for f in spherical coords)<br>
    <br>
    <blockquote
cite="mid:CAM9tzSkqM7=TS8UYnFCdJcB7zdxjUBKgSLv_HCCnSb2+HvMbrQ@mail.gmail.com"
      type="cite">
      <div class="gmail_quote">
        <div> </div>
        <blockquote class="gmail_quote" style="margin:0 0 0
          .8ex;border-left:1px #ccc solid;padding-left:1ex">
          <br>
           I was wondering if this phenomenon is common in PDEs? and if
          there is any way to correct for it?<br>
          <br>
           I am currently using 2nd order centered differences for
          interior grid points, and 1st order forward/backward
          differences for edge points. Would it be worthwhile to try
          moving to 4th order differences instead? Or would that make
          the problem worse?<br>
          <br>
           I've even tried smoothing the parameters which go into the
          matrix entries using moving averages...which doesn't seem to
          help too much.<br>
          <br>
           Any advice from those who have experience with this
          phenomenon would be greatly appreciated!<br>
          <br>
          Thanks,<br>
          <font color="#888888">
            Patrick<br>
          </font></blockquote>
      </div>
      <br>
    </blockquote>
    <br>
  </body>
</html>