<html>
  <head>
    <meta content="text/html; charset=utf-8" http-equiv="Content-Type">
  </head>
  <body text="#000000" bgcolor="#FFFFFF">
    Hi Dongwook,<br>
    <br>
    I'm using the HLLC Riemann solver. I've attached a recent log file
    from a short run.<br>
    <br>
    Thanks!<br>
    <br>
    Best,<br>
    Rukmani<br>
    <br>
    <br>
    <div class="moz-cite-prefix">On 02/18/2016 10:31 PM, Dongwook Lee
      wrote:<br>
    </div>
    <blockquote
      cite="mid:7690B0E9-99A8-4F91-B1E5-59A5F5849056@gmail.com"
      type="cite">
      <meta http-equiv="content-type" content="text/html; charset=utf-8">
      <div>Dear Rukmani,</div>
      <div><br>
      </div>
      <div>What kind of Riemann solver are you using?</div>
      <div>Can you send me a log file or flash.par?</div>
      <div><br>
      </div>
      <div>Thanks,</div>
      <div>Dongwook</div>
      <div><br>
        On Feb 18, 2016, at 6:59 PM, Rukmani Vijayaraghavan <<a
          moz-do-not-send="true" href="mailto:rukmani@virginia.edu"><a class="moz-txt-link-abbreviated" href="mailto:rukmani@virginia.edu">rukmani@virginia.edu</a></a>>
        wrote:<br>
        <br>
      </div>
      <blockquote type="cite">
        <div>
          <meta content="text/html; charset=utf-8"
            http-equiv="Content-Type">
          Hi Jason, Klaus,<br>
          <br>
          This block-by-block variation is correlated with similar
          variation in other fluid variables (density, pressure), and
          this persists even where there is a zero velocity inflow, as
          well as with a uniform grid, and with both USM and PPM (pure
          hydro) solvers. Modifying the gravity solver from Multigrid to
          Multipole doesn't make a difference either. I'm using the
          FLASH Gamma EOS unit. As far as I've seen, there is no
          variation in B-field across grid cells adjacent to block /
          refinement boundaries, this only happens at the inflow edge. <br>
          <br>
          I also update the magnetic field face variables (MAG_FACE_VAR
          and/or MAGI_FACE_VAR), with no effect. Div(B) still seems to
          be 0. <br>
          <br>
          Any other suggestions would be great!<br>
          <br>
          Thanks,<br>
          Rukmani<br>
          <br>
          <br>
          <br>
          <div class="moz-cite-prefix">On 02/18/2016 01:30 PM, Jason
            Galyardt wrote:<br>
          </div>
          <blockquote
cite="mid:CABdF9uevM2PyrdYd_1Lz1Ofx11u0mTKEjCRHxboEA9aB6_YFKA@mail.gmail.com"
            type="cite">
            <div dir="ltr">
              <div>
                <div>
                  <div>
                    <div>
                      <div>Hi Rukmani,<br>
                        <br>
                      </div>
                      I used a spatially varying wind; the velocity of
                      the wind varies along the boundary, but it has a
                      well-defined, time-independent form. I've also
                      seen problems with more realistic B-field
                      geometries which (to my horror) included step
                      functions in the domain interior. I had to smooth
                      these out to avoid unphysical evolution in those
                      regions. <br>
                      <br>
                      I've also seen some modest increase in B-field
                      magnitude for the cells adjacent to a refinement
                      boundary. I haven't reported the latter previously
                      because I haven't had time to figure out what's
                      going on there. You might try setting lrefine_min
                      = lrefine_max to get uniform refinement and see
                      whether that helps (some of our group's
                      simulations do this).<br>
                      <br>
                    </div>
                    The block by block variation does seem strange. I
                    would expect this kind of variation to be correlated
                    with variation in another variable. How do the other
                    variables look in the problem region?<br>
                    <br>
                  </div>
                  Another idea: could this variation be tied to the
                  equation of state? If you're using one of the
                  supported FLASH EOS units, you're probably fine.<br>
                  <br>
                </div>
                Regards,<br>
              </div>
              Jason<br>
              <br>
            </div>
            <div class="gmail_extra"><br>
              <div class="gmail_quote">On Thu, Feb 18, 2016 at 11:07 AM,
                Rukmani Vijayaraghavan <span dir="ltr"><<a
                    moz-do-not-send="true"
                    class="moz-txt-link-abbreviated"
                    href="mailto:rukmani@virginia.edu"><a class="moz-txt-link-abbreviated" href="mailto:rukmani@virginia.edu">rukmani@virginia.edu</a></a>></span>
                wrote:<br>
                <blockquote class="gmail_quote" style="margin:0 0 0
                  .8ex;border-left:1px #ccc solid;padding-left:1ex">
                  <div text="#000000" bgcolor="#FFFFFF"> Hi Jason,<br>
                    <br>
                    Thanks! I'm using FLASH 4.2, I'll try using 4.3 to
                    see if that makes a difference. I haven't tried
                    refining on the magnetic variables yet.<br>
                    <br>
                    For the different runtime parameters -- <br>
                    <br>
                    1. I've tried cfl = 0.5 and 0.8, but nothing lower
                    yet. I'll check to see if that works.<br>
                    <br>
                    2. For the Riemann Solver, I've found HLLC to be a
                    bit more dissipative than HLLD, and therefore
                    marginally better at smoothing out the magnetic
                    field at the edges. Ditto with second order
                    MUSCL-Hancock over third order PPM.<br>
                    <br>
                    3. All the other runtime parameters are mostly the
                    same. I don't refine on the magnetic variables, but
                    I tried higher overall lrefine_min (to make sure the
                    outer edges get further refined) and it didn't help
                    -- the same block-based discontinuity persists.<br>
                    <br>
                    4. I'm using a constant wind inflow for this
                    particular run. One thing I checked to see was if
                    there was a round off error in reading my input
                    variables into double precision arrays, and this
                    tiny "seed" instability might grow, but it doesn't
                    seem to be an issue. What is strange is that the
                    value (and sign) of the initial instability varies
                    block-by-block. In your simulations, did you use a
                    constant wind?<br>
                    <br>
                    Thanks,<br>
                    Rukmani
                    <div>
                      <div class="h5"><br>
                        <br>
                        <div>On 02/18/2016 09:38 AM, Jason Galyardt
                          wrote:<br>
                        </div>
                        <blockquote type="cite">
                          <div dir="ltr">
                            <div>
                              <div>
                                <div>
                                  <div>Hi Rukmani,<br>
                                    <br>
                                  </div>
                                  I've had some similar issues with MHD
                                  runs. You didn't mention which version
                                  of FLASH you're using, but I've found
                                  the latest (v4.3) to be a bit more
                                  stable than v4.2 or v2.5. As for
                                  runtime parameters, found the
                                  following combination to be helpful:<br>
                                  <br>
                                  #~~~~<br>
                                </div>
                                <div># Refine on the magnetic variables:<br>
                                </div>
                                <div>refine_var_1 = "dens"<br>
                                  refine_var_2    = "magp"<br>
                                </div>
                                <div># -OR-<br>
                                </div>
                                <div># refine_var_2 = "magx"<br>
                                </div>
                                <div># refine_var_3 = "magy"<br>
                                </div>
                                <div># refine_var_4 = "magz"<br>
                                </div>
                                <div># prefer higher refinement,
                                  according to magp (default
                                  refine_cutoff_X = 0.8)<br>
                                  refine_cutoff_2 = 0.7<br>
                                  # refine_cutoff_3 = 0.7<br>
                                  # refine_cutoff_4 = 0.7<br>
                                  <br>
                                </div>
                                <div># Lower CFL: between 0.25 and 0.5<br>
                                </div>
                                <div>cfl = 0.5<br>
                                  <br>
                                </div>
                                <div># Use second order MUSCL-Hancock
                                  reconstruction scheme<br>
                                </div>
                                <div>order = 2<br>
                                </div>
                                <div><br>
                                  # I've mostly used the "hybrid" slope
                                  limiter, but occasionally I've found
                                  the "minmod" useful in particularly
                                  difficult situations <br>
                                </div>
                                <div>slopeLimiter    = "hybrid"<br>
                                  <br>
                                  # use flattening (dissipative)
                                  (originally for PPM)<br>
                                  use_flattening    = .true. <br>
                                  <br>
                                  # Use high order algorithm for E-field
                                  construction<br>
                                  E_modification  = .true.<br>
                                  <br>
                                  # Update magnetic energy using
                                  staggered B-fields<br>
                                  energyFix       = .true.<br>
                                  <br>
                                  # Prolongation method (injecton_prol,
                                  balsara_prol) -- Using Balsara's
                                  method is particularly critical, in my
                                  experience.<br>
                                  prolMethod      = "BALSARA_PROL"<br>
                                  <br>
                                </div>
                                <div># For the Riemann solver, I use
                                  HLLD for MHD runs, and HLLC for pure
                                  hydro runs.<br>
                                </div>
                                <div>RiemannSolver    = "HLLD"<br>
                                </div>
                                <div>#~~~~<br>
                                </div>
                                <br>
                                <div>What sort of inflow conditions have
                                  you implemented? Small non-linearities
                                  in the inflow can grow into large
                                  unphysical features over time (I've
                                  seen this happen in my own
                                  simulations). So, it's worth checking
                                  your boundary condition code for
                                  undesirable features. In any case, I
                                  hope this helps.<br>
                                </div>
                                <br>
                                Sean: is the E_upwind option available
                                for the unsplit MHD solver in FLASH 4.3?
                                My recollection is that it caused some
                                problems in previous versions....<br>
                                <br>
                              </div>
                              Regards,<br>
                            </div>
                            Jason<br>
                            <br>
                          </div>
                          <div class="gmail_extra"><br>
                            <div class="gmail_quote">On Wed, Feb 17,
                              2016 at 9:22 PM, Rukmani Vijayaraghavan <span
                                dir="ltr"><<a moz-do-not-send="true"
                                  class="moz-txt-link-abbreviated"
                                  href="mailto:rukmani@virginia.edu">rukmani@virginia.edu</a>></span>
                              wrote:<br>
                              <blockquote class="gmail_quote"
                                style="margin:0 0 0 .8ex;border-left:1px
                                #ccc solid;padding-left:1ex">Hi
                                everyone,<br>
                                <br>
                                I've come across an error when updating
                                fluid variables at the inflow edge of a
                                wind tunnel simulation. I'm running a
                                simulation of a galaxy (with active dark
                                matter particles, gas, and passive
                                particles) in a box, whose fluid is
                                initialized to be identical to the
                                incoming wind (with vx, vy, vz = 600
                                km/s, 0, 0). There is a small error (on
                                the order of 1%) when updating grid
                                cells near the inflow boundary (with
                                both USM and PPM solvers), and this
                                error is spatially correlated with 
                                block boundaries. While this error
                                itself is tolerable as far as the
                                density and pressure go, this has bad
                                consequences for the magnetic field
                                which grows as the wind propagates
                                through the box (see attached figure,
                                xl_boundary). This figure shows slices
                                of Bx at two timesteps (annotated with
                                block boundaries and magnetic field
                                vectors). The dynamic range of Bx in
                                this image has been reduced to highlight
                                these discontinuities. At the timesteps
                                shown in the attached image, the
                                fluctuations in Bx are ~1%, but grow
                                with time up to order unity. I've tried
                                a variety of Riemann solvers (HLLC,
                                HLLD, Roe, Hybrid), slope limiters (mc,
                                minmod, etc.), interpolation orders,
                                prolongation methods, turning on and off
                                specific USM switches, but nothing seems
                                to solve this issue so far. Has anybody
                                else dealt with and/or successfully
                                solved this issue?<br>
                                <br>
                                Thanks,<br>
                                Rukmani<span><font color="#888888"><br>
                                    <br>
                                    -- <br>
                                    Rukmani Vijayaraghavan<br>
                                    NSF Astronomy & Astrophysics
                                    Postdoctoral Fellow<br>
                                    University of Virginia<br>
                                    <a moz-do-not-send="true"
                                      href="mailto:rukmani@virginia.edu"
                                      target="_blank">rukmani@virginia.edu</a><br>
                                    <br>
                                  </font></span></blockquote>
                            </div>
                            <br>
                          </div>
                        </blockquote>
                        <br>
                        <pre cols="72">-- 
Rukmani Vijayaraghavan
NSF Astronomy & Astrophysics Postdoctoral Fellow
University of Virginia
<a moz-do-not-send="true" href="mailto:rukmani@virginia.edu" target="_blank">rukmani@virginia.edu</a></pre>
                      </div>
                    </div>
                  </div>
                </blockquote>
              </div>
              <br>
            </div>
          </blockquote>
          <br>
          <pre class="moz-signature" cols="72">-- 
Rukmani Vijayaraghavan
NSF Astronomy & Astrophysics Postdoctoral Fellow
University of Virginia
<a moz-do-not-send="true" class="moz-txt-link-abbreviated" href="mailto:rukmani@virginia.edu">rukmani@virginia.edu</a></pre>
        </div>
      </blockquote>
    </blockquote>
    <br>
    <pre class="moz-signature" cols="72">-- 
Rukmani Vijayaraghavan
NSF Astronomy & Astrophysics Postdoctoral Fellow
University of Virginia
<a class="moz-txt-link-abbreviated" href="mailto:rukmani@virginia.edu">rukmani@virginia.edu</a></pre>
  </body>
</html>