Numerical physics simulation
Run and analyze physics simulation codes to target quantities.
20 project-authored proposals · taxonomy v1.1 · 0 new benchmark runs. Verifiers and partial-credit rules below are designs, not executed results.
- Typical applications
- Quantum ESPRESSO · NumPy/SciPy
- Class GUI-need label
CL· CLI-sufficient
Task proposals · 20
Expand to inspect the complete original wording and proposed checks. Original proposal text remains English in both interfaces.
R08.01.01.05-T01 Run a silicon SCF calculation and report energy and band edges
Original instruction
Run pw.x (Quantum ESPRESSO 7.3) on fixtures/si_scf.in with the pseudopotential in fixtures/pseudo/, and write output/si.json with total energy (Ry), highest occupied and lowest unoccupied levels (eV) and the number of SCF iterations.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Quantum ESPRESSO
- Interfaces
term· Terminal- GUI need · task setting
CL· CLI-sufficient- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
code· Code & commands- Work product
WP07· Simulation result- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Grader reruns the input: energy within 1e-5 Ry, levels within 0.01 eV, iteration count exact. Input hash checked.
Proposed partial credit
50% total energy; 30% band edges; 20% iterations.
R08.01.01.05-T02 Converge cutoff and k-mesh for aluminium to 1 meV/atom
Original instruction
For fcc Al (fixtures/al_base.in, smearing m-v 0.02 Ry), find the smallest ecutwfc (5 Ry steps) and Monkhorst-Pack mesh (n x n x n) for which total energy changes by <1 meV/atom on further increase. Write the convergence tables and chosen values to output/convergence.json.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Quantum ESPRESSO · Python
- Interfaces
term· Terminal- GUI need · task setting
CL· CLI-sufficient- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
code· Code & commands- Work product
WP07· Simulation result- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Grader reruns the chosen and next settings: change <1 meV/atom; chosen values equal grader's minimal converged values (one step tolerance).
Proposed partial credit
40% tables correct; 30% cutoff; 30% k-mesh.
R08.01.01.05-T03 Fit an equation of state for copper and report a0 and B0
Original instruction
Compute total energies of fcc Cu (PBE, fixtures/cu_template.in) at 9 lattice constants from 3.50 to 3.74 A, fit a third-order Birch-Murnaghan EOS (ev.x or SciPy), and report a0 (A), B0 (GPa) and B0' in output/eos.json.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Quantum ESPRESSO · SciPy
- Interfaces
term· Terminal- GUI need · task setting
CL· CLI-sufficient- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
code· Code & commands- Work product
WP07· Simulation result- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Grader reruns 3 of the 9 points to verify energies (1e-4 Ry) and fits the reference: a0 within 0.005 A, B0 within 3 GPa, B0' within 0.3.
Proposed partial credit
30% energies verified; 40% a0; 20% B0; 10% B0'.
R08.01.01.05-T04 Compute graphene bands along a k-path chosen in XCrySDen
Original instruction
For graphene (fixtures/graphene_scf.in), select the high-symmetry path Gamma-M-K-Gamma in XCrySDen's k-path tool, run the nscf band calculation with pw.x and bands.x, plot the bands to output/bands.png, and report the energy of the Dirac point relative to the Fermi level and its k-point.
- Platforms
mac· macOSlin· Linux- Applications
- Quantum ESPRESSO · XCrySDen
- Interfaces
term· Terminaldesk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clickcode· Code & commandsvis· Visual judgment- Work product
WP07· Simulation result- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
Grader: K_POINTS crystal_b list matches the path (within 1e-4); Dirac point at K with |E-EF| <0.05 eV; band plot exists with labeled high-symmetry points.
Proposed partial credit
30% k-path; 30% band calculation; 30% Dirac point; 10% plot.
R08.01.01.05-T05 Compute the Gamma-point optical phonon of silicon
Original instruction
Starting from fixtures/si_scf.in, run ph.x at Gamma with tr2_ph=1e-16 and report the optical phonon frequency (cm-1) and whether the acoustic sum rule correction (dynmat.x) changes the acoustic modes to <1 cm-1. Write output/phonon.json.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Quantum ESPRESSO
- Interfaces
term· Terminal- GUI need · task setting
CL· CLI-sufficient- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
code· Code & commands- Work product
WP07· Simulation result- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Grader reruns: optical frequency within 2 cm-1 of reference; ASR-corrected acoustic modes <1 cm-1 reported correctly.
Proposed partial credit
60% optical frequency; 40% ASR check.
R08.01.01.05-T06 Compute the CO adsorption energy on a relaxed Pt(111) slab
Original instruction
Build a 4-layer 2x2 Pt(111) slab with 15 A vacuum (bottom 2 layers fixed), relax clean slab, gas-phase CO and CO at the top site with pw.x (PBE, settings in input/settings.json), and report the adsorption energy (eV) and C-O bond length in output/ads.json.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Quantum ESPRESSO · ASE
- Interfaces
term· Terminal- GUI need · task setting
CL· CLI-sufficient- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
code· Code & commands- Work product
WP07· Simulation result- Conditions
long· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
Grader checks structures (layers, vacuum, constraints, top site) and re-evaluates final energies with single-point runs: adsorption energy within 0.05 eV of reference, bond length within 0.01 A.
Proposed partial credit
30% structures correct; 20% relaxations converged; 40% adsorption energy; 10% bond length.
R08.01.01.05-T07 Solve the 1D harmonic oscillator eigenproblem by finite differences
Original instruction
With NumPy/SciPy, discretize -1/2 psi'' + 1/2 x^2 psi = E psi on [-10,10] with 2,001 points, compute the 5 lowest eigenvalues and the max error vs exact n+1/2, and write output/qho.json.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- NumPy/SciPy
- Interfaces
term· Terminal- GUI need · task setting
CL· CLI-sufficient- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
code· Code & commands- Work product
WP07· Simulation result- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Grader: eigenvalues within 1e-4 of n+1/2; reported error consistent with values.
Proposed partial credit
80% eigenvalues; 20% error reported.
R08.01.01.05-T08 Estimate the 2D Ising critical temperature from Binder cumulants
Original instruction
Write a Metropolis (or Wolff) Monte Carlo for the 2D Ising model, compute the Binder cumulant U4 for L=8, 16, 32 over T in [2.1, 2.5] with enough sweeps for <1% error, estimate Tc from the crossings and write output/ising.json with Tc and its uncertainty, plus output/binder.png.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- NumPy/SciPy
- Interfaces
term· Terminal- GUI need · task setting
CL· CLI-sufficient- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
code· Code & commands- Work product
WP07· Simulation result- Conditions
long· Long-horizon- Proposed difficulty
- 4 / 5
Proposed verifier
Grader: Tc within 0.01 of 2.269 with a reported uncertainty <=0.02; U4 curves at L=16 within 0.02 of reference at 3 temperatures.
Proposed partial credit
50% Tc; 30% U4 values; 20% uncertainty and plot.
R08.01.01.05-T09 Verify second-order convergence of a Crank-Nicolson heat solver
Original instruction
Implement Crank-Nicolson for u_t = u_xx on [0,1] with u=sin(pi x) initial data and zero Dirichlet boundaries, run to t=0.1 for N=20, 40, 80, 160 with dt = dx, and report the L2 errors and observed convergence order in output/cn.json.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- NumPy/SciPy
- Interfaces
term· Terminal- GUI need · task setting
CL· CLI-sufficient- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
code· Code & commands- Work product
WP07· Simulation result- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Grader: errors within 5% of reference values; observed order between 1.9 and 2.1.
Proposed partial credit
60% error values; 40% order.
R08.01.01.05-T10 Measure RDF and diffusion of a Lennard-Jones liquid in LAMMPS
Original instruction
Simulate an LJ fluid in LAMMPS at T*=1.0, rho*=0.8 (4,000 atoms, rc=2.5): equilibrate NVT, then NVE production of 50,000 steps. Report the first RDF peak position/height and the diffusion coefficient D* from the MSD slope in output/lj.json.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- LAMMPS · NumPy/SciPy
- Interfaces
term· Terminal- GUI need · task setting
CL· CLI-sufficient- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
code· Code & commands- Work product
WP07· Simulation result- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Grader: input settings verified from the log; RDF peak within 0.02 sigma and 3% height; D* within 5% of reference.
Proposed partial credit
20% setup; 40% RDF; 40% diffusion.
R08.01.01.05-T11 Compute transmission through a photonic-crystal waveguide bend in Meep
Original instruction
Using Meep (Python), model a 2D photonic crystal of dielectric rods (eps=11.56, r=0.2a) with a 90-degree line-defect bend, launch a Gaussian pulse in the guided band, and compute normalized transmission vs frequency (0.30-0.44 c/a). Write output/transmission.csv and the peak transmission frequency.
- Platforms
mac· macOSlin· Linux- Applications
- Meep · NumPy/SciPy
- Interfaces
term· Terminal- GUI need · task setting
CL· CLI-sufficient- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
code· Code & commands- Work product
WP07· Simulation result- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
Grader reruns the script at resolution 20: transmission spectrum RMS error <0.05 vs reference; peak frequency within 0.003 c/a.
Proposed partial credit
30% geometry/source correct; 50% spectrum; 20% peak.
R08.01.01.05-T12 Integrate the solar system for 1,000 years with bounded energy drift
Original instruction
Integrate the Sun and 8 planets from the initial conditions in input/ephem_2026.csv for 1,000 years using a symplectic leapfrog (or WHFast in REBOUND), keeping relative energy drift below 1e-6, and report Earth's position at the end and the max drift in output/nbody.json.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- NumPy/SciPy · REBOUND
- Interfaces
term· Terminal- GUI need · task setting
CL· CLI-sufficient- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
code· Code & commands- Work product
WP07· Simulation result- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Grader: max |dE/E| <1e-6; Earth's final position within 1e-3 AU of a high-accuracy IAS15 reference.
Proposed partial credit
40% energy drift; 60% final position.
R08.01.01.05-T13 Compute the NiO band gap and moments with DFT+U
Original instruction
For antiferromagnetic NiO (AFM-II, rhombohedral cell, fixtures/nio_template.in), run PBE+U with Hubbard U=6.3 eV on Ni 3d (Hubbard card), converge, and report the band gap (eV) and Ni magnetic moment (muB) in output/nio.json. Compare with plain PBE.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Quantum ESPRESSO
- Interfaces
term· Terminal- GUI need · task setting
CL· CLI-sufficient- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
code· Code & commands- Work product
WP07· Simulation result- Conditions
long· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
Grader: magnetic ordering and U set correctly; reruns final SCF: gap within 0.1 eV and moment within 0.05 muB of reference for both PBE and PBE+U.
Proposed partial credit
25% AFM setup; 15% U setup; 40% PBE+U results; 20% PBE comparison.
R08.01.01.05-T14 Compute a capacitance with mesh refinement in COMSOL
Original instruction
In COMSOL Multiphysics (Electrostatics), build the interdigitated capacitor in input/idc_drawing.pdf (2D, 10 fingers, substrate eps_r=9.8), sweep mesh refinement until capacitance changes <0.5%, and report capacitance per unit length (pF/m). Save output/idc.mph.
- Platforms
win· Windowsmac· macOSlin· Linux- Applications
- COMSOL Multiphysics
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clickgeom· Precise geometrytype· Text & forms- Work product
WP07· Simulation result- Conditions
licensed· Licensed pro app- Proposed difficulty
- 3 / 5
Proposed verifier
Grader opens the .mph via COMSOL's own LiveLink in the grader session: geometry dimensions within 1%, physics and materials correct, capacitance within 1% of reference. LiveLink, the Java API and COMSOL's batch/server modes are disabled for the agent, so the model must be built in the COMSOL Desktop GUI.
Proposed partial credit
35% geometry; 20% physics/materials; 15% mesh study; 30% capacitance.
R08.01.01.05-T15 Compute spin-polarized DOS and the moment of bcc iron
Original instruction
Run a spin-polarized SCF and nscf (dense k-mesh) for bcc Fe from fixtures/fe_template.in, compute the total DOS with dos.x and projected d-DOS with projwfc.x, and report the magnetic moment (muB) and the d-band center for each spin in output/fe.json.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Quantum ESPRESSO
- Interfaces
term· Terminal- GUI need · task setting
CL· CLI-sufficient- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
code· Code & commands- Work product
WP07· Simulation result- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Grader reruns: moment within 0.05 muB; d-band centers within 0.05 eV of reference.
Proposed partial credit
50% moment; 50% d-band centers.
R08.01.01.05-T16 Find a vortex lattice by imaginary-time Gross-Pitaevskii evolution
Original instruction
Solve the 2D rotating Gross-Pitaevskii equation (g=1000, Omega=0.8 omega, harmonic trap) by split-step Fourier imaginary-time evolution on a 256x256 grid until energy changes <1e-8, and report the number of vortices, the chemical potential and output/density.png.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- NumPy/SciPy
- Interfaces
term· Terminal- GUI need · task setting
CL· CLI-sufficient- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
code· Code & commands- Work product
WP07· Simulation result- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
Grader: vortex count from phase winding equals reference; chemical potential within 0.5%; energy convergence criterion met (logged).
Proposed partial credit
40% vortex count; 40% chemical potential; 20% convergence evidence.
R08.01.01.05-T17 Simulate a sampling calorimeter response with Geant4
Original instruction
Build a Geant4 application for a 20-layer lead (2 mm) / plastic scintillator (5 mm) calorimeter, 30x30 cm, shoot 1,000 electrons each at 1, 5 and 20 GeV, and report the sampling fraction and the energy resolution fit a/sqrt(E) + b in output/calo.json.
- Platforms
mac· macOSlin· Linux- Applications
- Geant4 · ROOT
- Interfaces
term· Terminal- GUI need · task setting
CL· CLI-sufficient- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
code· Code & commands- Work product
WP07· Simulation result- Conditions
long· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
Grader reruns with fixed seeds: sampling fraction within 2% and stochastic term within 10% of reference; geometry checked via GDML export.
Proposed partial credit
25% geometry; 35% sampling fraction; 40% resolution fit.
R08.01.01.05-T18 Build a supercell in VESTA and export a matching QE input
Original instruction
Open input/batio3.cif in VESTA, create a 2x2x2 supercell, visually confirm the Ti off-centering direction, and export atomic positions so that output/batio3_222.in (pw.x) has the correct cell, 40 atoms and positions. Run a single SCF step to confirm the input parses.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- VESTA · Quantum ESPRESSO
- Interfaces
desk· Desktop GUIterm· Terminal- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clickvis· Visual judgmentcode· Code & commands- Work product
WP04· 3D/spatial model- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Grader parses the input: cell vectors 2x the CIF cell (1e-4 A), 40 atoms, positions matching the CIF-derived supercell under any permutation (1e-3 A); pw.x parses without error.
Proposed partial credit
40% cell; 40% positions; 20% input parses.
R08.01.01.05-T19 Compute a slab work function from the planar-averaged potential
Original instruction
For the Al(111) slab in fixtures/al111_scf.in, run pw.x, extract the electrostatic potential with pp.x and average.x, inspect the profile in ParaView or matplotlib to locate the vacuum plateau, and report the work function (eV) in output/wf.json with a plot.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Quantum ESPRESSO · ParaView
- Interfaces
term· Terminaldesk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
code· Code & commandsvis· Visual judgment- Work product
WP07· Simulation result- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
Grader reruns: vacuum level taken in the plateau region (gradient <1e-3 eV/A); work function within 0.05 eV of reference.
Proposed partial credit
30% potential extraction; 30% plateau choice; 40% work function.
R08.01.01.05-T20 Report a missing pseudopotential instead of substituting one
Original instruction
Run the QE calculation in fixtures/request_gaas/ as specified in request.md (PBEsol, norm-conserving, listed pseudopotentials). One required pseudopotential is not in fixtures/pseudo/. Do not substitute a different functional or type; write output/blocker.md naming the missing file and what is needed.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Quantum ESPRESSO
- Interfaces
term· Terminal- GUI need · task setting
CL· CLI-sufficient- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
code· Code & commands- Work product
WP12· Answer/decision- Conditions
clarify· Ask when unclear- Proposed difficulty
- 3 / 5
Proposed verifier
Grader: blocker.md names the exact missing pseudopotential and its required functional/type; no SCF output produced with a substituted pseudopotential (output dir checked).
Proposed partial credit
60% missing file and requirement identified; 40% no substitution performed.
Mapped research references
Existing benchmark-to-class links from the research catalog. They are not record-level annotations, executions or product endorsements.