← all systems · see it in The System
Computational physics (computer simulation of physical systems: the practice and the simulated systems it stands for)
Computational physics is the study and implementation of numerical analysis to solve problems in physics. In the philosophers' narrow sense, a computer simulation is a program that is run on a computer and that uses step-by-step methods to explore the approximate behavior of a mathematical model.
Properties published scienceour proposal
Sourced fields, and our own readings.
| aliases | Computational physics |
| born at | 1948-1955 |
| born precision | contested; precursors back to 1791 |
| boundary | Inside: the mathematical model taken from theory, its discretisation (grid, time step, lattice, particles, sampled configurations), the algorithms and codes, the run on hardware, the numerical output with its error estimates, verification and validation, and the predictions, understanding and hypotheses it sends out. Outside: the target systems simulated (the atmosphere, black holes, protons, proteins), the theories it starts from, the experiments it is checked against, the computers it runs on, and the people who build and read it. |
| nested level | A human practice carried by computers (rung 12), a part of physics and a subset of computational science; its targets reach from quarks (lattice QCD) to the large-scale structure of the universe (N-body). |
| scale | technological |
Measures (31) published science
| Forecast improvement rate (characteristic clock of the practice) | about 4 days of lead time in 40 years (about 1 day per decade) forecast lead time at equal accuracy |
| Molecular dynamics time step (characteristic clock of the run) | a few femtoseconds |
| Time steps in a typical MD simulation | millions or billions steps |
| GPU throughput for a 50,000-atom protein system | about 1 microsecond simulated in a few days simulated time per wall-clock time |
| Operational forecast throughput target | 200-300 forecast days per wall-clock day |
| Richardson's 1922 hand forecast: computing time | at least six weeks for a six-hour forecast over two points |
| Medium-range forecast skill (ECMWF IFS, 500 hPa ACC 80%) | about 7 days |
| Hurricane track accuracy | 150 km up to 4 days ahead km at lead time |
| Predictability limit of weather forecasts | about 14 days |
| Error doubling time in weather models | about 5 days |
| Power usage effectiveness, global data-centre average | 1.67 facility power / computing power (2019) |
| Share of facility power reaching computing equipment | about 0.60 ratio (1/1.67) |
| Landauer minimum heat per irreversible operation | of the order of kT energy per irreversible function |
| GEANT4 run: emissions at 15 and 60 cores | 238 and 481 gCO2e (run time only halved by 4x cores) |
| Supercomputer emissions of Australian astronomers | 15 kt per year; 22 t per researcher CO2 per year |
| Inputs to the light hadron spectrum calculation | 3 parameters (light and strange quark masses, coupling) |
| Share of proton and neutron mass from confined quark-gluon energy | about 95 percent |
| Analysis procedures behind the systematic error (BMW) | 432 fitting procedures |
| Lattice QCD integrated autocorrelation time (BMW) | under about 10 trajectories (every tenth used) |
| Quenched spectrum determination | almost 20 years |
| Lattice QCD proton mass error | less than 2 percent |
| Wolff cluster algorithm autocorrelation time at criticality | 1-2 steps per spin (lattices up to 128 x 128) |
| MANIAC I memory | 1000 words |
| Code needed today to reproduce FPU recurrences | 15 lines (Matlab) |
| Rest mass radiated in a simulated equal-mass merger (Pretorius 2005) | on the order of 5 percent |
| Final black hole spin in the first full merger simulation | about 0.70 dimensionless spin a |
| Template waveforms in the GW150914 search | about 250,000 templates |
| Numerical relativity waveform agreement with an independent code | 99.9 percent |
| Particles in the Millennium simulation | ten billion particles |
| Cost growth of classical N-body (direct) | N squared operations per step in particle number N |
| Physics faculty surveyed on computational instruction | 1246 from 357 institutions faculty |
Open contentions (4) open contention
Where sources or readings disagree. Both stay on the map.
- C-413 Where do thoughts come from?
booked: No theory is proven, so the map lists the main ones side by side and doesn't pick one. It shows where thinking is handled in the brain, not where thoughts start.
the other reading: see human.json thought channel positions - C-1220 When did computational physics begin?
booked: The map dates it to the first electronic physics simulations, 1948-1955, and keeps the earlier by-hand precursors alongside.
the other reading: by-hand precursors: Verlet integration used by Delambre in 1791; Richardson's hand forecast 1922; Fermi's unpublished Monte Carlo for neutron diffusion in the 1930s; the encyclopedia's 'first large-scale deployment during the Manhattan Project' (a 12 hard-sphere Monte Carlo) - C-1221 Will AI replace traditional numerical weather forecasting?
booked: The map leaves it open: numerical models remain in use, and replacement is called 'not inconceivable' but faces many open questions.
the other reading: data-driven models replace the numerical models and data assimilation outright; hybrid workflows keep both - C-1233 Are computer simulations theory, experiment, or something else?
booked: The map treats simulation as a third way between theory and experiment and keeps the other readings alongside.
the other reading: akin to theoretical physics: an extension of modelling (Frigg and Reiss, via SEP); computer experiments (Ercolessi, via the encyclopedia; Parker's materiality argument, via SEP); scientific instruments (Alvarado, via SEP); epistemically opaque results trusted like senses or testimony (Beisbart, via SEP)
Partners and parts published science
Built from / builds (11)
- Physics (the field)
- Verification and validation
- Numerical analysis (numerical differentiation and integration)
- Numerical methods
- Molecular dynamics
- Monte Carlo and molecular dynamics methods
- Data assimilation
- High-performance computing (supercomputers, GPUs) and its tuning
- Random number generator
- Force field (molecular mechanics)
- Reservoir computing against backpropagation-trained recurrent networks for forecasting chaotic dynamics
Sits inside / holds (2)
Trades with (valence) (13)
- Standard Model of particle physics via sys:sensory-environment · Computational physics (computer simulation of physical systems: the practice and the simulated systems it stands for) gain
- Quantum field theory via sys:computer · Computational physics (computer simulation of physical systems: the practice and the simulated systems it stands for) gain
- General relativity via sys:written-record · Computational physics (computer simulation of physical systems: the practice and the simulated systems it stands for) gain
- Thermal energy (heat, temperature and heat flow) via sys:thermal-energy · Computational physics (computer simulation of physical systems: the practice and the simulated systems it stands for) loss
- Human (Homo sapiens) via sys:sensory-environment · Computational physics (computer simulation of physical systems: the practice and the simulated systems it stands for) gain
- Weather via sys:atmosphere-of-earth · Computational physics (computer simulation of physical systems: the practice and the simulated systems it stands for) gain
- Computer via sys:electromagnetic-field · Computational physics (computer simulation of physical systems: the practice and the simulated systems it stands for) gain
- Algorithm (the practice of step-by-step procedures and the structure they compute) via sys:written-record · Computational physics (computer simulation of physical systems: the practice and the simulated systems it stands for) gain
- Differential equations (the practice of writing and solving laws of change, and the equations it works on) via sys:written-record · Computational physics (computer simulation of physical systems: the practice and the simulated systems it stands for) gain
- Physics (the field) via sys:sensory-environment · Computational physics (computer simulation of physical systems: the practice and the simulated systems it stands for) gain
- Experiment, measurement and uncertainty (the physics lab as a practice: designing, measuring, estimating error) via sys:sensory-environment · Computational physics (computer simulation of physical systems: the practice and the simulated systems it stands for) gain
- Fluid dynamics (the flows of liquids and gases, and the practice of describing them) via sys:computer · Computational physics (computer simulation of physical systems: the practice and the simulated systems it stands for) gain
- Chaos and nonlinear dynamics (the practice, and the chaotic behaviour it studies) via sys:computer · Computational physics (computer simulation of physical systems: the practice and the simulated systems it stands for) gain
expressed by (15)
- Templating / Standardization
- Quality Improvement/Refinement
- Interoperability/Transduction (Bridging)
- Encapsulation & Modularity (Abstraction)
- Mapping/Modeling
- Uncertainty Reduction/Learning
- Active Experimentation / Probing (Optimal Experimental Design)
- Multi-Scale Data Fusion / Triangulation
- Surprise / Change-Point Detection
- Hypothesis Portfolio / Multiple Hypothesis Tracking
- Offline Simulation & Replay-Guided Planning
- Self-Monitoring
- Validation of Effectiveness
- Redundancy & Diversity
- Quality Control & Error Correction
floor filled by (21)
- Proton
- Standard Model of particle physics
- Physics (the field)
- Black hole
- Human (Homo sapiens)
- Weather
- Computer
- Algorithm (the practice of step-by-step procedures and the structure they compute)
- Verification and validation
- Differential equations (the practice of writing and solving laws of change, and the equations it works on)
- Numerical analysis (numerical differentiation and integration)
- Experiment, measurement and uncertainty (the physics lab as a practice: designing, measuring, estimating error)
- Numerical methods
- Monte Carlo and molecular dynamics methods
- Data assimilation
- High-performance computing (supercomputers, GPUs) and its tuning
- Force field (molecular mechanics)
- Lattice QCD
- Condensed matter physics
- Fluid dynamics (the flows of liquids and gases, and the practice of describing them)
- Chaos and nonlinear dynamics (the practice, and the chaotic behaviour it studies)
represents (1)
Kinds (8)
- Numerical relativity
- Computational fluid dynamics (and computational mechanics)
- Lattice QCD
- Computational statistical physics
- Computational astrophysics (N-body simulation)
- Computational electrodynamics
- Computational chemistry
- Numerical weather prediction
wired (2)
Notes published science
- The first 'numerical experiment' (1955) found that energy in a vibrating chain came back instead of spreading out, and it was found only because someone forgot to stop the computer. The programmer, Mary Tsingou, was left off the authors for 50 years (src:dauxois-2008-fput).
- Supercomputers computed the masses of the proton and neutron from just three inputs (two quark masses and the coupling) and matched experiment; about 95% of their mass is the energy of confined quarks and gluons, not the quarks' own mass (src:durr-bmw-hadron-masses).
- Richardson's 1922 hand forecast of six hours of weather took at least six weeks; operational models now aim for 200-300 forecast days per day, yet chaos still caps accurate forecasts at about 14 days even with a flawless model (src:wp:numerical-weather-prediction, src:lannelongue-2021-green-algorithms).
Layers
The ways it takes in its world, where they meet, and how it acts back, each from L1 (where it meets the world) up.
computational physics: a physical system turned into a model, discretised, computed and checked (reader's view)
- L1: The target system and the problem: a physical system whose model gives precise predictions in principle but cannot be solved in closed form (many bodies, nonlinearity, chaos)
- L2: The mathematical model: the governing equations from theory (Newton's equations, Navier–Stokes, Einstein's field equations, the QCD Lagrangian) with their parameters (force fields, parameterizations, quark masses and coupling)
- L3: Discretisation: grid cells, time steps, a space-time lattice, particles or sampled configurations; the step bounded by stability (CFL, femtosecond MD steps)
- L4: The algorithm and code: integrators (Verlet, Runge–Kutta, leapfrog), Monte Carlo and Metropolis sampling, linear algebra and eigenvalue solvers, shared software packages
- L5: The run on the machine: time step after time step (billions in MD), on GPUs, clusters and supercomputers, with software and hardware tuned to the problem
- L6: Checking: observables with statistical and systematic errors, convergence tests, verification against the model's equations and validation against the target and experiment
- L7: Use back on the world: predictions (forecasts, waveform templates, hadron masses), understanding, and hypotheses that send out new experiments
the boardroom (shared floors)
- L4: Algorithm and code: the model, its parameters, the random stream and the idea are combined into a program of simple operations
- L5: The run: time stepping or sampling on the machine, starting from assimilated initial states
- L6: Checking: error estimates, convergence, verification and validation against experiment; the surprise that flags something new
- L7: Use: prediction, understanding and exploration sent back to the world
way in: Emotion: surprise at an output that breaks expectation
- L1: An unexpected output: energy returning to the first mode
- L5: Surprise against the expectation of thermalisation
way in: Experimental data: measured values the results are validated against
- L1: Experimental values with their widths (the light hadron spectrum)
- L5: Experiments designed to test the simulation's predictions
way in: Observations: measured states of the target fed in as starting conditions (data assimilation)
- L1: Incomplete, heterogeneous, scattered observations of the target
- L2: The instruments, with their biases and failures
- L3: Projected onto the model grid and interpolated in time
- L4: Assimilated: the forecast (first guess) corrected by the observations into an analysis (3D-Var, 4D-Var, ensemble Kalman filters)
way in: Parameters: empirical and quantum-mechanical inputs fitted into the model
- L1: Quantum-mechanical results and experimental measurements
- L2: The fitted force field or parameterization (sub-grid processes)
- L3: The set of inputs the run takes (for the light hadron spectrum, three)
way in: Randomness: the random number stream that drives Monte Carlo
- L1: A random number stream (pseudo-random, from a generator)
- L2: Random trials: pebbles thrown, moves proposed
- L3: Accepted or rejected by the Metropolis rule (detailed balance)
way in: Shared methods and codes from neighbouring fields and the community
- L1: Methods borrowed across fields (density functional theory from chemistry)
- L2: Software packages, easier to use and supporting non-experts
- L3: Validated parameter sets reused (CHARMM36m and others)
way in: Sight: the physicist watching the output (graphics, plots, trajectories)
- L1: Output drawn as graphics (first on an oscilloscope)
- L5: Too much data to look at whole: 100,000 atoms over billions of steps
way in: The machine: memory, speed and architecture the run is given
- L1: The hardware: MANIAC I with 1000 words; today GPUs and supercomputers
- L2: Throughput: one GPU simulates about a microsecond of a 50,000-atom system in a few days
- L3: Software and hardware tuned to the problem
- L4: The run spread over many cores, with a best number for time and emissions
way in: Theory: the governing equations handed over by theoretical physics
- L1: Theories based on mathematical models, precise in principle
- L2: The model chosen for the target: the Navier–Stokes system for the atmosphere; the QCD Lagrangian on a lattice
- L3: Discretised: the Lagrangian put on a hypercubic lattice with spacing a, extrapolated back to the continuum
way in: Thought: the physicist's idea, intuition and mental simulation
- L1: Contested: where thoughts come from is not settled (human walk C-413); here, the problem held in mind
- L2: An intuitive physics engine: approximate, probabilistic mental simulation
- L3: The idea of testing a physical idea by computing it (Fermi's numerical experiment)
way out: Forecasts issued to the public
- L1: Weather users: high-impact weather, the renewable energy sector
- L6: Uncertainty attached from ensembles
- L7: Live predictions to the public on a fixed schedule
way out: Heat and emissions: the run's electricity turned to heat
- L1: Greenhouse emissions per run
- L2: The facility draws more than the computers use (cooling, power delivery)
- L5: Each irreversible operation generates heat of order kT at least
way out: Predictions and templates handed to experiments
- L1: Detector data searched with them
- L4: Template waveforms built ahead of the search
- L7: Hypotheses that lead to new experimental work
way out: Results, codes and methods published and taught
- L1: Taught on homework and projects, rarely in class or exams
- L7: Results reported (written up for others)
Sources (30) published science
Data assimilation (Wikipedia, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Hollingsworth, S. A. & Dror, R. O. (2018) Molecular dynamics simulation for all. Neuron 99, 1129, doi:10.1016/j.neuron.2018.08.011 (PMC6209097) (behind its links)
Krauth W. Introduction to Monte Carlo algorithms (lecture notes, Advances in Computer Simulation, Springer 1998). arXiv:cond-mat/9612186 (behind its links)
Numerical weather prediction (Wikipedia, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Pretorius, F. (2005) Evolution of Binary Black Hole Spacetimes, Phys. Rev. Lett. 95:121101, doi:10.1103/PhysRevLett.95.121101, arXiv:gr-qc/0507014 (behind its links)
Dauxois T. Fermi, Pasta, Ulam and a mysterious lady. arXiv:0801.1590 (Physics Today 2008) (behind its links)
Lannelongue L, Grealey J, Inouye M. Green Algorithms: quantifying the carbon footprint of computation. Adv Sci 8:2100707 (2021). PMC8224424 (behind its links)
Schultz MG et al. Can deep learning beat numerical weather prediction? Phil Trans R Soc A 379:20200097 (2021). PMC7898133, CC BY (behind its links)
Landauer R. Irreversibility and heat generation in the computing process. IBM J Res Dev 5:183 (1961) (abstract via OpenAlex) (behind its links)
Dürr S, Fodor Z et al. (BMW collaboration). Ab-initio determination of light hadron masses. arXiv:0906.3599 (behind its links)
Lattice QCD (Wikipedia, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Wolff U. Collective Monte Carlo updating for spin systems. Phys Rev Lett 62:361 (1989) (abstract via OpenAlex) (behind its links)
N-body simulation (Wikipedia, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Molecular dynamics (Wikipedia, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Monte Carlo method (Wikipedia, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Ferrenberg AM, Landau DP, Wong YJ. Monte Carlo simulations: hidden errors from 'good' random number generators. Phys Rev Lett 69:3382 (1992) (abstract via OpenAlex) (behind its links)
Computer Simulations in Science. Stanford Encyclopedia of Philosophy (current revision) (behind its links)
Dürr, S. et al. (2008). Ab-initio Determination of Light Hadron Masses. Science 322, 1224. doi:10.1126/science.1163233 (arXiv:0906.3599) (behind its links)
Duerr S. et al. (2008) Ab initio determination of light hadron masses. Science 322:1224; arXiv:0906.3599 (behind its links)
Numerical relativity (Wikipedia, rev 1375775298, 2026-09-20; via Wikimedia Enterprise cache) (behind its links)
Slotnick J et al. (2014) CFD Vision 2030 Study: A Path to Revolutionary Computational Aerosciences. NASA/CR-2014-218178 (NTRS 20140003093) (behind its links)
Batygin K, Laughlin G. On the dynamical stability of the Solar System. ApJ 683:1207 (2008), arXiv:0804.1946 (behind its links)