Envelope Probes

[Z10Frank]

A small case of 1D gaussian envelope propagation in vacuum: Laser_pulse_1D_propagation_envelope.txt

At t=0 the Probes and Fields diags for Env_E_abs show a field identically 0 (maybe it is not computed at t=0). The Field Env_A_abs seems normal at t=0. For t>0, the Fields Env_E_abs and Env_A_abs are correct, but Probes appear fragmented and with ripples at the patch boundaries: S.Probe.Probe0("Env_A_abs").slide()

[Z10Frank]

Same happens with 1 OpenMP thread and 1 MPI.

[mccoys]

Is this new? Only in 1d? Only envelope fields?

[Z10Frank]

I don't know if it is new. The equivalent simulation without envelope seems to work perfectly: Laser_pulse_1D_propagation_without_envelope.txt

S.Probe.Probe0("Ey").slide()

[Z10Frank]

In LaserEnvelope1D::initEnvelope the Env_E_abs field seems to be computed at the envelope initialisation: ( *Env_Eabs1D )( i )= std::abs( ( ( *A1D )( i )-( *A01D )( i ) )/timestep - i1*omega*( *A1D )( i ) );, maybe it's a problem in the diags.

[mccoys]

But does the probe output correctly Ex

[Z10Frank]

In the envelope simulation it will always be zero, it's just a laser pulse in vacuum, so the only non zero quantities will be Env_E_abs and Env_A_abs.

[Z10Frank]

I have tried with the 2D and AM benchmarks, everything seems correct. With the 1D benchmark the bug appears, so it seems that the bug is limited to 1D.

[Z10Frank]

Removing the moving window and using only 1 patch the bug is the same. When the laser moves to the right the discontinuity in Env_E_abs seems to progressively disappear.

[Z10Frank]

Moving the laser initial position at the center allowed to study the phenomenon. Left panels: Fields (correct); right panels: Probes (with bug) At the start the field in the Probes seems correct (maybe a little shifted) on the left hand side of the window and damped on the right hand side:

Afterwards, the laser moves on the right, the discontinuity disappears but the field remains damped:

I tried substituting the same lines of code that work for other interpolators, and the result is the same. Maybe there is a problem in the Probes definition in 1D for the envelope fields, or at the moment of the output dump?