kUPS MD Tutorials

An executable introduction for readers who know Python and elementary calculus but may be new to molecular dynamics. Each lesson connects physical intuition, equations, a compact JAX implementation, the corresponding kUPS interface, and an interpreted simulation result.

1. What Does a Molecular-Dynamics Program Actually Store and Update?

   Build the mental model behind MD: atomic state, potential energy, JAX forces, discrete trajectories, periodic cells, ensembles, and the corresponding kUPS objects.

   Start here; Sungsoo Ahn; August 04, 2026; 12 min read; lesson 1 of 13
2. What Distribution Does an MD Simulation Start From?

   Build an atomic state, derive a Maxwell–Boltzmann momentum draw in JAX, remove box translation, and inspect the state returned by kUPS.

   Tutorial; Sungsoo Ahn; July 14, 2026; 11 min read; lesson 2 of 13
3. What Does an MD Integrator Actually Approximate?

   Translate velocity Verlet from Newton's equations into JAX, map its kick–drift–force–kick structure to kUPS, and inspect a real atomic trajectory.

   Tutorial; Sungsoo Ahn; July 14, 2026; 14 min read; lesson 3 of 13
4. Where Does Error Enter an MD Trajectory?

   Separate timestep, arithmetic, initial-condition, and force-model error with transparent JAX controls and matched kUPS trajectories.

   Tutorial; Sungsoo Ahn; July 14, 2026; 15 min read; lesson 4 of 13
5. What Does a Thermostat Really Do?

   Derive BAOAB Langevin dynamics, implement it transparently in JAX, and compare real kUPS BAOAB and CSVR trajectories.

   Tutorial; Sungsoo Ahn; July 14, 2026; 15 min read; lesson 5 of 13
6. How Does a Barostat Move the Simulation Cell?

   Derive stochastic cell rescaling in JAX, then interpret isotropic and flexible-cell NPT trajectories run through kUPS.

   Tutorial; Sungsoo Ahn; July 14, 2026; 16 min read; lesson 6 of 13
7. When Is a Trajectory Long Enough to Trust?

   Derive autocorrelation and effective sample size in JAX, then apply them to independent kUPS trajectories and actual atom motion.

   Tutorial; Sungsoo Ahn; July 14, 2026; 16 min read; lesson 7 of 13
8. How Do Trajectories Become Physical Observables?

   Implement a periodic RDF and coordination estimator in JAX, then measure structure and velocity memory from real kUPS trajectories.

   Tutorial; Sungsoo Ahn; July 14, 2026; 16 min read; lesson 8 of 13
9. How Do Equilibrium Samples Become Free Energies?

   Implement probability-to-free-energy conversion and bias reweighting in JAX, then transform a real kUPS RDF into a supported pair PMF.

   Tutorial; Sungsoo Ahn; July 14, 2026; 16 min read; lesson 9 of 13
10. When Can You Trust a Free-Energy Estimate?

    Derive FEP and BAR, implement both estimators in JAX, and diagnose whether sampled configurations actually support a free-energy difference.

    Tutorial; Sungsoo Ahn; July 14, 2026; 17 min read; lesson 10 of 13
11. What Does Umbrella Sampling Actually Sample?

    Apply a harmonic bias to atomic coordinates in JAX, reconstruct connected windows with WHAM, and interpret real kUPS Ar-pair trajectories.

    Tutorial; Sungsoo Ahn; July 14, 2026; 21 min read; lesson 11 of 13
12. Bias the Landscape or Drive the System?

    Separate adaptive metadynamics bias from nonequilibrium steering, implement both central updates in JAX, and interpret real kUPS Ar-pair paths.

    Tutorial; Sungsoo Ahn; July 14, 2026; 20 min read; lesson 12 of 13
13. From Atomic Graph to Learned-Force MD

    Build a differentiable atomic graph in JAX, map it to a pinned Tojax MACE potential in kUPS, and separate numerical stability from model accuracy.

    Tutorial; Sungsoo Ahn; July 14, 2026; 24 min read; lesson 13 of 13