FieldProbe Bug

[zhazhajust]

Description

The question is that 2 case: (1) I set a field probe line with starting locate at out of the simulation windows, and setting a moving window with "do_moving_window_FP" to be OFF, an the time that field probe line got in the simulation window, it will cause segment fault. (2) Set the "do_moving_window_FP" ON so that the field probe line will follow the moving window, but when the timestep coming into the end_moiving_step, no longer fieldprobe to be compute.

After long time debug, I find out that both the two case will cause "MyParIter" have no valide particle, so the field computation for each particles will be skip because don't have any valid particles in the iterator. But I don't know how to resolve it, because the particle "x, y, z" position work well in the init function of "FieldProbe".

//FieldProbe.cpp
    using MyParIter = FieldProbeParticleContainer::iterator;
    for (MyParIter pti(m_probe, lev); pti.isValid(); ++pti)

Input file

#algo
algo.maxwell_solver = Yee #psatd
algo.particle_shape = 3

#amr
amr.max_grid_size = 32
amr.max_level = 0
amr.n_cell = 360 80

#boundary
boundary.field_hi = pml pml
boundary.field_lo = pml pml
boundary.particle_hi = absorbing absorbing
boundary.particle_lo = absorbing absorbing

#diagnostics
diagnostics.diags_names = species field

#electrons
electrons.charge = -q_e
electrons.density_function(x,y,z) = "(x-x1>=0)(x-x2<0)(x - x1)(n0 - 0)/(x2 - x1) + (x-x2>=0)(x-x3<0)n0 + (x-x3>=0)(x-x4<0)(x - x4)(0 - n0)/(x4 - x3)"
electrons.do_continuous_injection = 1
electrons.initialize_self_fields = 0
electrons.injection_style = nuniformpercell
electrons.mass = m_e
electrons.momentum_distribution_type = constant
electrons.num_particles_per_cell_each_dim = 1 1 1
electrons.profile = parse_density_function
electrons.ux = 0.0
electrons.uy = 0.0
electrons.uz = 0.0
electrons.xmax = 0.0005
electrons.xmin = 0.0
electrons.ymax = 0.00015
electrons.ymin = -0.00015
electrons.zmax = 0.00015
electrons.zmin = -0.00015

#field
field.diag_type = Full
field.fields_to_plot = Bx By Bz Ex Ey Ez jx jy jz rho
field.file_prefix = ./FieldDiagnostic/
field.format = openpmd
field.intervals = 100
field.openpmd_backend = h5
field.write_species = 0

#geometry
geometry.dims = 2
geometry.prob_hi = 0.0 0.00001
geometry.prob_lo = -0.000060 -0.00001

#laser1
laser1.direction = 1 0 0
laser1.do_continuous_injection = 0
laser1.e_max = 3210000000000.0
laser1.polarization = 0 1 0
laser1.position = -1e-05 0.0 0.0
laser1.profile = Gaussian
laser1.profile_duration = 3.2e-14
laser1.profile_focal_distance = 1e-05
laser1.profile_t_peak = 6.4e-14
laser1.profile_waist = 1e-05
laser1.wavelength = 8e-07

#lasers
lasers.names = laser1

#max_step
max_step = 1000

#my_constants
my_constants.n0 = 1.74e+24
my_constants.x1 = 0.0
my_constants.x2 = 2e-05
my_constants.x3 = 0.0005
my_constants.x4 = 0.0006

#species
species.diag_type = Full
species.electrons.plot_filter_function(t,x,y,z,ux,uy,uz) = sqrt(uzuz+uxux+uy*uy) > 1.2
species.electrons.variables = ux uy uz w
species.fields_to_plot = none
species.file_prefix = ./SpecDiagnostic/
species.format = openpmd
species.intervals = 200
species.openpmd_backend = h5
species.species = electrons
species.write_species = 1

#warpx
warpx.cfl = 0.95
warpx.do_dive_cleaning = 0
warpx.do_moving_window = 1
warpx.end_moving_window_step = 100
warpx.moving_window_dir = x
warpx.moving_window_v = 1.0
warpx.serialize_initial_conditions = 1
warpx.start_moving_window_step = 0
warpx.use_filter = 1
warpx.verbose = 1

#Diag
warpx.reduced_diags_names = FieldProbe_ScatLine
FieldProbe_ScatLine.type = FieldProbe
FieldProbe_ScatLine.intervals = 1
FieldProbe_ScatLine.probe_geometry = Line
FieldProbe_ScatLine.x_probe = -0.05e-6
FieldProbe_ScatLine.z_probe = -8e-6
FieldProbe_ScatLine.x1_probe = -0.05e-6
FieldProbe_ScatLine.z1_probe = 8e-6
FieldProbe_ScatLine.resolution = 80
FieldProbe_ScatLine.do_moving_window_FP = 1

[zhazhajust]

And when the refinement mesh open, only the level 1 field will be probe but ignore the level 0 field.

[ax3l]

Thank you for the report!

Crash for Probes that run out of the sim domain

Urgh, that is a bug - let's fix it. Currently we only check probes are in the sim domain during construction phase.

We could approach this by masking probes that run out of the domain (or removing them forever once it happens).

MR: Probes Field 1

refinement mesh [...]: only the level 1 field will be probe but ignore the level 0 field.

We want to use the physical field in aux, which it seems to do:

• https://github.com/ECP-WarpX/WarpX/blob/23.11/Source/WarpX.H#L485-L486
• https://github.com/ECP-WarpX/WarpX/blob/23.11/Source/Diagnostics/ReducedDiags/FieldProbe.cpp#L393-L399

What are you comparing to?

[zhazhajust]

Thanks for help!
The question is about the FieldProbe. And i have addressed the question (2).
I findout that the in the
https://github.com/ECP-WarpX/WarpX/blob/23.11/Source/Diagnostics/ReducedDiags/FieldProbe.cpp#L393-L399

                const auto prob_lo = WarpX::GetInstance().Geom(0).ProbLo()[WarpX::moving_window_dir];
                amrex::Real probe_move = prob_lo - prob_lo_prev;
                prob_lo_prev = prob_lo;
                
                // move_dist = probe_move;

                const auto temp_warpx_moving_window = WarpX::moving_window_dir;
                if (temp_warpx_moving_window == 0)
                {
                    x_probe += probe_move;
                }

The x_probe will move when the moving window working and the flag on.

And then i rewritten the file to make it could recognize the both level, and then rewritten the new sort method because the id number will be discontinues.

        // push back idx
        std::vector<long int> idx_vec(m_valid_particles_level[j]);
        std::iota(idx_vec.begin(), idx_vec.end(), 0);
        // sort idx as id number(m_data_out_level[j][i1*noutputs]) order
        std::sort (idx_vec.begin(), idx_vec.end(), [&](int i1,int i2){
            return static_cast<long int>(m_data_out_level[j][i1*noutputs]) < 
                static_cast<long int>(m_data_out_level[j][i2*noutputs]);
        });
        // push back data
        for (long int i = 0; i < m_valid_particles_level[j]; i++)
        {
            for (long int k = 0; k < noutputs; k++)
            {
                // sorted_data[idx * noutputs + k] = m_data_out_level[j][i * noutputs + k];
                sorted_data[i * noutputs + k] = m_data_out_level[j][idx_vec[i] * noutputs + k];
            }
        }

Finally, i had add a new pull request above modify at "Fix the bug for diag and fieldprobe #4454".

[ax3l]

X-ref #4457