Merge pull request #22 from glotzerlab/alj-2d

ALJ 2D energy conservation test

hoomd_validation/alj_2d.py

@@ -0,0 +1,331 @@
+# Copyright (c) 2022 The Regents of the University of Michigan.
+# Part of HOOMD-blue, released under the BSD 3-Clause License.
+
+"""ALJ 2D energy conservation validation test."""
+
+from config import CONFIG
+from project_class import Project
+from flow import aggregator
+import util
+import os
+
+# Run parameters shared between simulations
+RANDOMIZE_STEPS = 5e4
+RUN_STEPS = 2e6
+WRITE_PERIOD = 4000
+LOG_PERIOD = {'trajectory': 50000, 'quantities': 2000}
+ALJ_PARAMS = {'epsilon': 1.0}
+
+# Unit area hexagon
+PARTICLE_VERTICES = [[6.20403239e-01, 0.00000000e+00, 0],
+                     [3.10201620e-01, 5.37284966e-01, 0],
+                     [-3.10201620e-01, 5.37284966e-01, 0],
+                     [-6.20403239e-01, 7.59774841e-17, 0],
+                     [-3.10201620e-01, -5.37284966e-01, 0],
+                     [3.10201620e-01, -5.37284966e-01, 0]]
+CIRCUMCIRCLE_RADIUS = 0.6204032392788702
+INCIRCLE_RADIUS = 0.5372849659264116
+
+
+def job_statepoints():
+    """list(dict): A list of statepoints for this subproject."""
+    num_particles = 42**2
+    replicate_indices = range(4)
+    params_list = [(1.0, 0.4)]
+    for kT, density in params_list:
+        for idx in replicate_indices:
+            yield ({
+                "subproject": "alj_2d",
+                "kT": kT,
+                "density": density,
+                "num_particles": num_particles,
+                "replicate_idx": idx
+            })
+
+
+def is_alj_2d(job):
+    """Test if a given job is part of the alj_2d subproject."""
+    return job.statepoint['subproject'] == 'alj_2d'
+
+
+@Project.operation(directives=dict(executable=CONFIG["executable"],
+                                   nranks=8,
+                                   walltime=1))
+@Project.pre(is_alj_2d)
+@Project.post.isfile('alj_2d_initial_state.gsd')
+def alj_2d_create_initial_state(job):
+    """Create initial system configuration."""
+    import hoomd
+    import numpy
+    import itertools
+
+    init_diameter = CIRCUMCIRCLE_RADIUS * 2 * 1.15
+
+    device = hoomd.device.CPU(msg_file=job.fn('create_initial_state.log'))
+
+    num_particles = job.statepoint['num_particles']
+    density = job.statepoint['density']
+
+    box_volume = num_particles / density
+    L = box_volume**(1 / 2.)
+
+    N = int(numpy.ceil(num_particles**(1. / 2.)))
+    x = numpy.linspace(-L / 2, L / 2, N, endpoint=False)
+
+    if x[1] - x[0] < init_diameter:
+        raise RuntimeError('density too high to initialize on square lattice')
+
+    position_2d = list(itertools.product(x, repeat=2))[:num_particles]
+
+    # create snapshot
+    snap = hoomd.Snapshot(device.communicator)
+
+    if device.communicator.rank == 0:
+        snap.particles.N = num_particles
+        snap.particles.types = ['A']
+        snap.configuration.box = [L, L, 0, 0, 0, 0]
+        snap.particles.position[:, 0:2] = position_2d
+        snap.particles.typeid[:] = [0] * num_particles
+        snap.particles.moment_inertia[:] = [0, 0, 1]
+
+    # Use hard sphere Monte-Carlo to randomize the initial configuration
+    mc = hoomd.hpmc.integrate.Sphere()
+    mc.shape['A'] = dict(diameter=init_diameter)
+
+    sim = hoomd.Simulation(device=device, seed=job.statepoint.replicate_idx)
+    sim.create_state_from_snapshot(snap)
+    sim.operations.integrator = mc
+
+    device.notice('Randomizing initial state...')
+    sim.run(RANDOMIZE_STEPS)
+    device.notice(f'Done. Move counts: {mc.translate_moves}')
+
+    hoomd.write.GSD.write(state=sim.state,
+                          filename=job.fn("alj_2d_initial_state.gsd"),
+                          mode='wb')
+
+
+def make_md_simulation(job,
+                       device,
+                       initial_state,
+                       method,
+                       sim_mode,
+                       period_multiplier=1):
+    """Make an MD simulation.
+
+    Args:
+        job (`signac.job.Job`): Signac job object.
+
+        device (`hoomd.device.Device`): hoomd device object.
+
+        initial_state (str): Path to the gsd file to be used as an initial state
+            for the simulation.
+
+        method (`hoomd.md.methods.Method`): hoomd integration method.
+
+        sim_mode (str): String identifying the simulation mode.
+
+        ThermodynamicQuantities is added by default, any more quantities should
+            be in this list.
+
+        period_multiplier (int): Factor to multiply the GSD file periods by.
+    """
+    import hoomd
+    from hoomd import md
+
+    incircle_d = INCIRCLE_RADIUS * 2
+    circumcircle_d = CIRCUMCIRCLE_RADIUS * 2
+    r_cut = max(2**(1 / 6) * incircle_d,
+                circumcircle_d + 2**(1 / 6) * 0.15 * incircle_d)
+
+    # pair force
+    nlist = md.nlist.Cell(buffer=0.4)
+    alj = md.pair.aniso.ALJ(default_r_cut=r_cut, nlist=nlist)
+    alj.shape['A'] = {
+        "vertices": PARTICLE_VERTICES,
+        "faces": [],
+        "rounding_radii": 0
+    }
+    alj.params[("A", "A")] = {
+        "epsilon": ALJ_PARAMS['epsilon'],
+        "sigma_i": incircle_d,
+        "sigma_j": incircle_d,
+        "alpha": 0
+    }
+
+    # integrator
+    integrator = md.Integrator(dt=0.0001,
+                               methods=[method],
+                               forces=[alj],
+                               integrate_rotational_dof=True)
+
+    # compute thermo
+    thermo = md.compute.ThermodynamicQuantities(hoomd.filter.All())
+
+    # add gsd log quantities
+    logger = hoomd.logging.Logger(categories=['scalar', 'sequence'])
+    logger.add(thermo,
+               quantities=[
+                   'pressure',
+                   'potential_energy',
+                   'kinetic_temperature',
+                   'kinetic_energy',
+                   'translational_kinetic_energy',
+                   'rotational_kinetic_energy',
+               ])
+    logger.add(integrator, quantities=['linear_momentum'])
+
+    # simulation
+    sim = util.make_simulation(job, device, initial_state, integrator, sim_mode,
+                               logger, WRITE_PERIOD,
+                               LOG_PERIOD['trajectory'] * period_multiplier,
+                               LOG_PERIOD['quantities'] * period_multiplier)
+    sim.operations.add(thermo)
+
+    # thermalize momenta
+    sim.state.thermalize_particle_momenta(hoomd.filter.All(), job.sp.kT)
+
+    return sim
+
+
+def run_nve_md_sim(job, device, run_length):
+    """Run the MD simulation in NVE."""
+    import hoomd
+
+    initial_state = job.fn('alj_2d_initial_state.gsd')
+    nvt = hoomd.md.methods.NVE(hoomd.filter.All())
+    sim_mode = 'nve_md'
+
+    sim = make_md_simulation(job,
+                             device,
+                             initial_state,
+                             nvt,
+                             sim_mode,
+                             period_multiplier=300)
+
+    # Run for a long time to look for energy and momentum drift
+    device.notice('Running...')
+    sim.run(run_length)
+    device.notice('Done.')
+
+
+@Project.operation(directives=dict(walltime=48,
+                                   executable=CONFIG["executable"],
+                                   nranks=8))
+@Project.pre.after(alj_2d_create_initial_state)
+@Project.post.true('alj_2d_nve_md_cpu_complete')
+def alj_2d_nve_md_cpu(job):
+    """Run NVE MD on the CPU."""
+    import hoomd
+    device = hoomd.device.CPU(msg_file=job.fn('run_nve_md_cpu.log'))
+    run_nve_md_sim(job, device, run_length=600e6)
+
+    if device.communicator.rank == 0:
+        job.document['alj_2d_nve_md_cpu_complete'] = True
+
+
+@Project.operation(directives=dict(walltime=48,
+                                   executable=CONFIG["executable"],
+                                   nranks=1,
+                                   ngpu=1))
+@Project.pre.after(alj_2d_create_initial_state)
+@Project.post.true('alj_2d_nve_md_gpu_complete')
+def alj_2d_nve_md_gpu(job):
+    """Run NVE MD on the GPU."""
+    import hoomd
+    device = hoomd.device.GPU(msg_file=job.fn('run_nve_md_gpu.log'))
+    run_nve_md_sim(job, device, run_length=600e6)
+
+    if device.communicator.rank == 0:
+        job.document['alj_2d_nve_md_gpu_complete'] = True
+
+
+agg = aggregator.groupby(key=['kT', 'density', 'num_particles'],
+                         sort_by='replicate_idx',
+                         select=is_alj_2d)
+
+
+@agg
+@Project.operation(directives=dict(walltime=1, executable=CONFIG["executable"]))
+@Project.pre(
+    lambda *jobs: util.true_all(*jobs, key='alj_2d_nve_md_cpu_complete'))
+@Project.pre(
+    lambda *jobs: util.true_all(*jobs, key='alj_2d_nve_md_gpu_complete'))
+@Project.post(lambda *jobs: util.true_all(
+    *jobs, key='alj_2d_conservation_analysis_complete'))
+def alj_2d_conservation_analyze(*jobs):
+    """Analyze the output of NVE simulations and inspect conservation."""
+    import gsd.hoomd
+    import numpy
+    import math
+    import matplotlib
+    import matplotlib.style
+    import matplotlib.figure
+    matplotlib.style.use('ggplot')
+    from util import read_gsd_log_trajectory, get_log_quantity
+
+    sim_modes = ['nve_md_cpu', 'nve_md_gpu']
+
+    energies = []
+    linear_momenta = []
+
+    for job in jobs:
+        job_energies = {}
+        job_linear_momentum = {}
+
+        for sim_mode in sim_modes:
+            with gsd.hoomd.open(job.fn(sim_mode
+                                       + '_quantities.gsd')) as gsd_traj:
+                # read GSD file once
+                traj = read_gsd_log_trajectory(gsd_traj)
+
+            job_energies[sim_mode] = (
+                get_log_quantity(
+                    traj, 'md/compute/ThermodynamicQuantities/potential_energy')
+                + get_log_quantity(
+                    traj, 'md/compute/ThermodynamicQuantities/kinetic_energy'))
+            job_energies[sim_mode] = (
+                job_energies[sim_mode]
+                - job_energies[sim_mode][0]) / job.statepoint["num_particles"]
+
+            momentum_vector = get_log_quantity(traj,
+                                               'md/Integrator/linear_momentum')
+            job_linear_momentum[sim_mode] = [
+                math.sqrt(v[0]**2 + v[1]**2 + v[2]**2) for v in momentum_vector
+            ]
+
+        energies.append(job_energies)
+
+        linear_momenta.append(job_linear_momentum)
+
+    # Plot results
+    def plot(*, ax, data, quantity_name, legend=False):
+        for i, job in enumerate(jobs):
+            for mode in sim_modes:
+                ax.plot(numpy.asarray(data[i][mode]),
+                        label=f'{mode}_{job.statepoint.replicate_idx}')
+        ax.set_xlabel("frame")
+        ax.set_ylabel(quantity_name)
+
+        if legend:
+            ax.legend()
+
+    fig = matplotlib.figure.Figure(figsize=(10, 10 / 1.68 * 2), layout='tight')
+    ax = fig.add_subplot(2, 1, 1)
+    plot(ax=ax, data=energies, quantity_name=r"$E / N$", legend=True)
+
+    ax = fig.add_subplot(2, 1, 2)
+    plot(ax=ax, data=linear_momenta, quantity_name=r"$p$")
+
+    fig.suptitle("ALJ 2D conservation tests: "
+                 f"$kT={job.statepoint.kT}$, $\\rho={job.statepoint.density}$, "
+                 f"$N={job.statepoint.num_particles}$")
+    filename = f'alj_2d_conservation_kT{job.statepoint.kT}_' \
+               f'density{round(job.statepoint.density, 2)}_' \
+               f'N{job.statepoint.num_particles}.svg'
+
+    fig.savefig(os.path.join(jobs[0]._project.path, filename),
+                bbox_inches='tight')
+
+    for job in jobs:
+        job.document['alj_2d_conservation_analysis_complete'] = True

hoomd_validation/init.py

@@ -8,11 +8,12 @@
 import config
 
 # import subprojects
+import alj_2d
 import lj_fluid
 import hard_disk
 import hard_sphere
 
-subprojects = [lj_fluid, hard_disk, hard_sphere]
+subprojects = [alj_2d, lj_fluid, hard_disk, hard_sphere]
 
 project = signac.init_project(name="hoomd-validation", root=config.project_root)
 

hoomd_validation/project.py

@@ -7,11 +7,13 @@
 from project_class import Project
 
 # Define subproject flow operations
+import alj_2d
 import lj_fluid
 import hard_disk
 import hard_sphere
 
 __all__ = [
+    "alj_2d",
     "lj_fluid",
     "hard_disk",
     "hard_sphere",