case_fdtd_pulse_longrun

测试目标

case_fdtd_pulse_longrun 先用一个短时、局域、光滑脉冲激励场量，再关闭源项长时间传播。 它比纯无源稳定性测试更接近“有扰动进入系统后是否慢增长”的场景。

覆盖子程序

覆盖 E01 2D RZ TEz/TMz、E02 3D cylindrical m=0/m=1 和 E03 3D Cartesian 的 FDTD 更新核。

文件说明

文件

作用

test_pulse_*.f90

五个带短脉冲的长跑主程序。

pulse_common.f90

能量、轴线带、第一环和增长率监控。

pulse_longrun_summary.csv

case_name,CFL,Npulse,Ntotal,final_energy_ratio,post_pulse_growth_rate,result。

logs/*.log

每个 case 的时间序列监控。

运行方式

source ~/.venv/bin/activate
cd tests/005_maxwell/case_fdtd_pulse_longrun
bash run.sh

可选：bash run.sh 50000 200 80 5e-5，参数为 Ntotal monitor_every Npulse pulse_amp。

主流程

  1. 初始化零场或小幅背景场。

  2. 在前 Npulse 步加入 sin^2 时间包络和 Gaussian 空间包络源项。

  3. 每步执行 H 更新、边界填充、E 更新。

  4. Npulse 后完全关闭源项。

  5. 记录脉冲结束时能量 E_at_pulse_end。

  6. 长时间传播后计算 final_energy_ratio 和 post_pulse_growth_rate。

典型例子：3D cylindrical m=1 脉冲长跑

test_pulse_3d_cyl_m1.f90 是最敏感的一个子例子，因为它同时包含 phi 周期缝、 轴线附近处理和 cylindrical metric。默认设置为 nr=40,nphi=32,nz=64，总步数 20000，前 60 步加入幅值 1e-4 的短脉冲，之后完全无源传播。

integer, parameter :: nsteps_default = 20000
integer, parameter :: npulse_default = 60
real, parameter :: pulse_amp_default = 1.0e-4

call parse_int_arg(1, nsteps_default, nsteps)
call parse_int_arg(3, npulse_default, npulse)
call parse_real_arg(4, pulse_amp_default, pulse_amp)

主循环和稳定性测试类似，但在 E 更新之后、边界填充之前给 Ez 加入局域脉冲源。 当 n == npulse 时，测试记录 energy_pulse_end，后续所有增长率都相对于这个“源项关闭时刻”来解释。

call sub_E02_fdtd_3d_cylindrical_H(..., Er,Ephi,Ez,Hr,Hphi,Hz,dt,dr,dphi,dz,mu)
call fill_h_boundaries(nr,nphi,nz,Hr,Hphi,Hz)

call sub_E02_fdtd_3d_cylindrical_E(..., Er,Ephi,Ez,Hr,Hphi,Hz,dt,dr,dphi,dz,ep)
if (n <= npulse) call add_pulse_source(n,npulse,pulse_amp,nr,nphi,nz,dr,dphi,dz,Ez)
call fill_e_boundaries(nr,nphi,nz,Er,Ephi,Ez)

if (n == npulse) energy_pulse_end = total_energy

重点调用方式

call sub_*_H(...)
call fill_h_boundaries(...)
call sub_*_E(...)
if (n <= Npulse) call add_pulse_source(...)
call monitor_post_pulse_growth(...)

结果判断

final_energy_ratio = E_final / E_at_pulse_end，post_pulse_growth_rate 是脉冲结束后的对数增长率。 3D cylindrical m=1 的当前参考结果：

SUMMARY_CSV,3D_CYL_M1,0.800,60,20000,1.0088E+00,2.2835E-04,stable

常见误读

脉冲后能量不一定单调下降，因为这里没有吸收边界，周期方向会发生回绕。判据关注的是脉冲关闭后是否出现持续增长或发散。

贡献者

刘哲，赵隐剑 · 哈尔滨工业大学 · 2026/04/09

Test Goal

case_fdtd_pulse_longrun injects a short localized smooth pulse, then turns the source off and propagates for a long time. It is closer than a pure no-source test to the question: does the scheme slowly grow after a finite disturbance enters the system?

Covered Routines

The tests cover E01 2D RZ TEz/TMz, E02 3D cylindrical m=0/m=1, and E03 3D Cartesian FDTD kernels.

Files

File

Role

test_pulse_*.f90

Five long-run drivers with a short pulse.

pulse_common.f90

Energy, axis-band, first-ring, and growth-rate monitors.

pulse_longrun_summary.csv

Summary columns for case, CFL, pulse length, total steps, energy ratio, growth rate, and result.

logs/*.log

Per-case time-history monitors.

Run Command

source ~/.venv/bin/activate
cd tests/005_maxwell/case_fdtd_pulse_longrun
bash run.sh

Optional: bash run.sh 50000 200 80 5e-5 sets Ntotal monitor_every Npulse pulse_amp.

Main Flow

  1. Initialize zero or small background fields.

  2. For the first Npulse steps, add a sin^2 temporal envelope and Gaussian spatial pulse.

  3. Advance H, fill boundaries, and advance E each step.

  4. Disable the source completely after Npulse.

  5. Record energy at pulse end.

  6. Compute final_energy_ratio and post_pulse_growth_rate after the long run.

Example: 3D Cylindrical m=1 Pulse Long Run

test_pulse_3d_cyl_m1.f90 is the most sensitive subcase because it includes the phi periodic seam, near-axis handling, and cylindrical metric terms. The default grid is nr=40,nphi=32,nz=64. It runs 20000 steps, injects a 1e-4 pulse for the first 60 steps, and then propagates with no source.

integer, parameter :: nsteps_default = 20000
integer, parameter :: npulse_default = 60
real, parameter :: pulse_amp_default = 1.0e-4

call parse_int_arg(1, nsteps_default, nsteps)
call parse_int_arg(3, npulse_default, npulse)
call parse_real_arg(4, pulse_amp_default, pulse_amp)

The main loop resembles the stability test, but it adds a localized pulse to Ez after the electric update and before the electric boundary fill. When n == npulse, the test records energy_pulse_end; all post-pulse growth metrics are interpreted relative to the source-off moment.

call sub_E02_fdtd_3d_cylindrical_H(..., Er,Ephi,Ez,Hr,Hphi,Hz,dt,dr,dphi,dz,mu)
call fill_h_boundaries(nr,nphi,nz,Hr,Hphi,Hz)

call sub_E02_fdtd_3d_cylindrical_E(..., Er,Ephi,Ez,Hr,Hphi,Hz,dt,dr,dphi,dz,ep)
if (n <= npulse) call add_pulse_source(n,npulse,pulse_amp,nr,nphi,nz,dr,dphi,dz,Ez)
call fill_e_boundaries(nr,nphi,nz,Er,Ephi,Ez)

if (n == npulse) energy_pulse_end = total_energy

Core Call Pattern

call sub_*_H(...)
call fill_h_boundaries(...)
call sub_*_E(...)
if (n <= Npulse) call add_pulse_source(...)
call monitor_post_pulse_growth(...)

Result Interpretation

final_energy_ratio = E_final / E_at_pulse_end. The growth rate is the post-pulse logarithmic energy growth rate. Current 3D cylindrical m=1 reference output:

SUMMARY_CSV,3D_CYL_M1,0.800,60,20000,1.0088E+00,2.2835E-04,stable

Common Pitfall

Post-pulse energy need not decrease monotonically because these runs do not use absorbing boundaries; periodic directions can wrap. The criterion is persistent post-source growth or divergence.

Contributors

Zhe LIU, Yinjian ZHAO · Harbin Institute of Technology · 2026/04/09