Implement injection of particles from the embedded boundary #5208

RemiLehe · 2024-09-03T19:54:37Z

Overview

This PR implements flux injection of particles from the embedded boundary.

It also adds a test that emits particles from a sphere in 3D as represented here:

as well as RZ and 2D versions of this test. (In 2D, the particles are emitted from a cylinder.)

As can be seen in the above movie, particles are emitted from a single point within each cell (the centroid of the EB), instead of being emitted uniformly on the surface of the EB within the cell. This could be improved in a future PR.

The implementation as well as the user interface largely re-use the infrastructure for the flux injection from a plane. However, as a result, the user interface is perhaps not very intuitive. In particular, when specify the velocity distribution, uz represents the direction normal to the EB while ux, uy represent the tangential directions. This again will be improved in follow-up PR.

Follow-up PRs

Change the interface of gaussianflux so as to specify the tangential and normal distribution. In other words, instead of:

electron.momentum_distribution_type = gaussianflux
electron.ux_th = 0.01
electron.uy_th = 0.01
electron.uz_th = 0.1
electron.uz_m = 0.07

we would do:

electron.momentum_distribution_type = gaussianflux
electron.u_tangential_th = 0.01  # Tangential to the emitting surface
electron.u_normal_th = 0.1   # Normal to the emitting surface
electron.u_normal_m = 0.07

Change the interface so that the user does not need to specify the number of macroparticles per cell (which is problematic for EB, since difference cell contain different EB surface, and should in general emit different numbers of macroparticles). Instead, we would specify the weight of macroparticles, i.e. instead of

electron.injection_style = NFluxPerCell
electron.num_particles_per_cell = 100   
electron.flux_function(x,y,z,t) = "1.”

we would do

electron.injection_style = NFluxPerCell
electron.flux_macroweight = 200   # Number of physical particles per macroparticle
electron.flux_function(x,y,z,t) = "4e12” # Number of physical particles emitted per unit time and surface

Add a way for the user to specify the total flux across the whole emitting surface
Example:

electron.flux_function(x,y,z,t) = "(x>-1)*(x<1)"
electron.total_flux = 4e12 # physical particle / second (not per unit area)

(In that case, flux_function would be rescaled internally by WarpX so as to emit the right number of particles.)

Add PICMI interface
Emit the particles uniformly from the surface of the EB within one cell

This reverts commit 20d1fba.

Examples/Tests/flux_injection/inputs_test_3d_flux_injection_from_eb

…om_eb Make sphere larger

…jection_from_eb

RemiLehe · 2024-10-10T12:53:18Z

Source/Particles/AddPlasmaUtilities.H

@@ -22,6 +22,29 @@
 #include <AMReX_REAL.H>
 #include <AMReX_RealBox.H>

+struct PDim3 {


I had to move this object from PhysicalParticleContainer.cpp to AddPlasmaUtilities.H as part of this PR.

RemiLehe · 2024-10-10T13:02:07Z

Source/Particles/PhysicalParticleContainer.cpp

-        AMREX_GPU_HOST_DEVICE
-        PDim3& operator=(PDim3&&)       = default;
-    };
-


I had to move this object from PhysicalParticleContainer.cpp to AddPlasmaUtilities.H as part of this PR.

dpgrote · 2024-10-10T16:22:55Z

Source/Particles/AddPlasmaUtilities.H

+
+#ifdef AMREX_USE_EB
+/*
+  This computes the scale_fac (used for setting the particle weights) on a on area basis


This should use the standard documentation format, like

/* * \brief This computes... * * \param[in] dx * /

Though I do notice that no other docs in this file match this format.

dpgrote · 2024-10-10T16:26:24Z

Source/Particles/AddPlasmaUtilities.H

+
+    // Rotate the momentum along the theta direction (in the z-x plane)
+    amrex::Real const cos_theta = nz;
+    amrex::Real const sin_theta = std::sqrt(1-nz*nz);


1._rt

To simplify this, could the theta and phi rotations be combined into one set of expressions, a 3D rotation?

dpgrote · 2024-10-10T16:31:57Z

Source/Particles/PhysicalParticleContainer.cpp

@@ -1371,6 +1348,22 @@ PhysicalParticleContainer::AddPlasmaFlux (PlasmaInjector const& plasma_injector,
    const auto dx = geom.CellSizeArray();
    const auto problo = geom.ProbLoArray();

+#ifdef AMREX_USE_EB
+    bool inject_from_eb = plasma_injector.m_inject_from_eb; // whether to inject from EB or from a plane


bool const

dpgrote · 2024-10-10T16:44:45Z

Source/Particles/PhysicalParticleContainer.cpp

+                // Skip cells that do not overlap with the plane
+                if (!flux_pos->overlapsWith(lo, hi)) { return; }
+            }
+            const int num_ppc_int = static_cast<int>(num_ppc_real_in_this_cell + amrex::Random(engine));


This looses the change made in PR #5272. The calculation of num_ppc_int should happen after the r factor is included.

dpgrote · 2024-10-10T16:50:36Z

Source/Particles/PhysicalParticleContainer.cpp

@@ -1659,6 +1736,7 @@ PhysicalParticleContainer::AddPlasmaFlux (PlasmaInjector const& plasma_injector,
                // For cylindrical emission (flux_normal_axis==0
                // or flux_normal_axis==2), the emission surface depends on
                // the radius ; thus, the calculation is finalized here
+


Can this blank line be delete? It's an unrelated format change.

dpgrote · 2024-10-10T17:28:29Z

Source/Particles/PhysicalParticleContainer.cpp

+#ifdef AMREX_USE_EB
+                if (inject_from_eb) {
+#if defined(WARPX_DIM_3D)
+                    pos.x = overlap_corner[0] + (iv[0] + 0.5_rt + eb_bnd_cent_arr(i,j,k,0))*dx[0];


Is this putting all of the particles at the same location, the EB boundary center? Is there a way to distribute the particles around on the EB surface?

Yes, that's absolutely correct: all the particles are emitted from a single point. (See also the PR description.)
We should indeed distribute the particles over the EB surface. This should be doable but will take a bit more time. This will be added in a follow-up PR.

I should have read the whole PR description :)

dpgrote

Looks good! Thanks!

…X#5208) # Overview This PR implements flux injection of particles from the embedded boundary. It also adds a test that emits particles from a sphere in 3D as represented here: ![movie](https://github.com/user-attachments/assets/1e76cf87-fd7d-4fa3-8c83-363956226a42) as well as RZ and 2D versions of this test. (In 2D, the particles are emitted from a cylinder.) As can be seen in the above movie, particles are emitted from a single point within each cell (the centroid of the EB), instead of being emitted uniformly on the surface of the EB within the cell. This could be improved in a future PR. The implementation as well as the user interface largely re-use the infrastructure for the flux injection from a plane. However, as a result, the user interface is perhaps not very intuitive. In particular, when specify the velocity distribution, `uz` represents the direction normal to the EB while `ux`, `uy` represent the tangential directions. This again will be improved in follow-up PR. # Follow-up PRs - [ ] Change the interface of `gaussianflux` so as to specify the tangential and normal distribution. In other words, instead of: ``` electron.momentum_distribution_type = gaussianflux electron.ux_th = 0.01 electron.uy_th = 0.01 electron.uz_th = 0.1 electron.uz_m = 0.07 ``` we would do: ``` electron.momentum_distribution_type = gaussianflux electron.u_tangential_th = 0.01 # Tangential to the emitting surface electron.u_normal_th = 0.1 # Normal to the emitting surface electron.u_normal_m = 0.07 ``` - [ ] Change the interface so that the user does not need to specify the number of macroparticles per cell (which is problematic for EB, since difference cell contain different EB surface, and should in general emit different numbers of macroparticles). Instead, we would specify the weight of macroparticles, i.e. instead of ``` electron.injection_style = NFluxPerCell electron.num_particles_per_cell = 100 electron.flux_function(x,y,z,t) = "1.” ``` we would do ``` electron.injection_style = NFluxPerCell electron.flux_macroweight = 200 # Number of physical particles per macroparticle electron.flux_function(x,y,z,t) = "4e12” # Number of physical particles emitted per unit time and surface ``` - [ ] Add a way for the user to specify the total flux across the whole emitting surface Example: ``` electron.flux_function(x,y,z,t) = "(x>-1)*(x<1)" electron.total_flux = 4e12 # physical particle / second (not per unit area) ``` (In that case, `flux_function` would be rescaled internally by WarpX so as to emit the right number of particles.) - [ ] Add PICMI interface - [ ] Emit the particles uniformly from the surface of the EB within one cell

RemiLehe force-pushed the injection_from_eb branch from cc5a0e9 to 6548719 Compare September 3, 2024 19:58

Implement changes by Maya Zeldin in a separate commit

20d1fba

RemiLehe force-pushed the injection_from_eb branch from 6548719 to 20d1fba Compare September 3, 2024 21:18

Add a few TODOs

add9df0

RemiLehe added the component: boundary PML, embedded boundaries, et al. label Sep 3, 2024

RemiLehe added 16 commits September 4, 2024 09:55

Add test example

e56d748

Revert "Implement changes by Maya Zeldin in a separate commit"

1f24d50

This reverts commit 20d1fba.

Merge branch 'development' into injection_from_eb

8897ccd

Move test

d9a1c86

Find boxes that overlap with the EB

1ef9bc4

Scale the number of particles per cell

00328ce

Cleanup

3888459

Minor refactoring of refined injection for AddPlasmaFlux

89c684f

Merge branch 'simplify_fine_injectionwq' into injection_from_eb

2e0779d

Some code refactoring

09e1174

Scale emitting flux by the area of the boundary

f7b3e92

Additional simplification

3c00605

Minor fixes

f16b01c

Merge branch 'simplify_fine_injectionwq' into injection_from_eb

b6310ae

Implement rotation of momentum

6d06352

Fix initialization

2c05fe6

RemiLehe commented Sep 16, 2024

View reviewed changes

Examples/Tests/flux_injection/inputs_test_3d_flux_injection_from_eb Outdated Show resolved Hide resolved

RemiLehe added 8 commits September 16, 2024 11:45

Update Examples/Tests/flux_injection/inputs_test_3d_flux_injection_fr…

6fe4b13

…om_eb Make sphere larger

Update test

9ecd54a

Merge branch 'injection_from_eb' of github.com:RemiLehe/WarpX into in…

1d0722c

…jection_from_eb

Fix parallelization issue

c123e14

Temporarily deactivate checksum

04d5a5e

Check velocity distribution

1a57716

Add one more test for perpendicular direction

a408735

Add code for the 3rd direction

b203a53

RemiLehe added 6 commits October 8, 2024 18:02

Add checksums

fdc4d1b

Reactivate checksum test

c0368af

Fix compilation tests

f3bc86c

Fix clang-tidy error

57e29d9

Update checksum

b56c3f9

Merge branch 'development' into injection_from_eb

3b68f69

RemiLehe changed the title ~~[WIP] Implement injection of particles from the embedded boundary~~ Implement injection of particles from the embedded boundary Oct 10, 2024

RemiLehe added 3 commits October 10, 2024 05:22

Merge branch 'development' into injection_from_eb

5a7b468

Fix const-ness

1d723c0

Add documentation

de16bfa

RemiLehe commented Oct 10, 2024

View reviewed changes

Update comment

b7f6f50

RemiLehe commented Oct 10, 2024

View reviewed changes

Update comment

404f262

RemiLehe requested a review from dpgrote October 10, 2024 16:18

RemiLehe assigned dpgrote and RemiLehe and unassigned RemiLehe Oct 10, 2024

dpgrote reviewed Oct 10, 2024

View reviewed changes

Correct const-ness

09703f5

dpgrote reviewed Oct 10, 2024

View reviewed changes

Incorporate review comments

249be0b

RemiLehe requested a review from dpgrote October 10, 2024 20:31

dpgrote approved these changes Oct 10, 2024

View reviewed changes

RemiLehe merged commit 1794629 into ECP-WarpX:development Oct 10, 2024
32 of 37 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Implement injection of particles from the embedded boundary #5208

Implement injection of particles from the embedded boundary #5208

RemiLehe commented Sep 3, 2024 •

edited

Loading

RemiLehe Oct 10, 2024

RemiLehe Oct 10, 2024

dpgrote Oct 10, 2024

dpgrote Oct 10, 2024

dpgrote Oct 10, 2024

dpgrote Oct 10, 2024

dpgrote Oct 10, 2024

dpgrote Oct 10, 2024

RemiLehe Oct 10, 2024

dpgrote Oct 10, 2024

dpgrote left a comment

Implement injection of particles from the embedded boundary #5208

Implement injection of particles from the embedded boundary #5208

Conversation

RemiLehe commented Sep 3, 2024 • edited Loading

Overview

Follow-up PRs

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

dpgrote left a comment

Choose a reason for hiding this comment

RemiLehe commented Sep 3, 2024 •

edited

Loading