MD in a Box
Interactive 2D molecular dynamics simulation running on a Raspberry Pi Pico 2. Based on the browser-based MD simulation by Ulf R. Pedersen: https://urp.dk/md/
| Liquid phase | Crystal phase |
|---|---|
 |
 |
Physics
Atoms are modelled as classical particles interacting via pair forces, following Newton’s laws of motion. A Langevin thermostat maintains the temperature.
Hardware
| Component | Details |
|---|---|
| Raspberry Pi Pico 2 | RP2350, dual-core ARM Cortex-M33 @ 150 MHz, 520 KB RAM |
| Pico Display Pack 2.0 | 320×240 pixel display (Pimoroni) |
| BNO055 9-axis IMU | Gravity vector + linear acceleration (DFRobot breakout) |
Total hardware cost: ~€65
Architecture
The simulation uses both CPU cores:
- Core 0 — Display rendering (PicoGraphics) and button input
- Core 1 — Physics simulation, running C-accelerated MD steps continuously
Particle positions are stored in shared memory arrays read by Core 0 for drawing. The accelerometer feeds tilt (gravity direction) and movement (box linear acceleration) into the simulation.
Performance
The MD force loop and Leap Frog integrator are implemented as a custom C
extension to ulab (user.c), called from MicroPython. This gives a large
speedup over pure Python/ulab:
| N atoms | ulab (MDPS) | ulab + C (MDPS) | Speedup |
|---|---|---|---|
| 16 | 78 | 12,674 | ~160× |
| 25 | 35 | 6,816 | ~195× |
| 36 | 18 | 3,960 | ~220× |
| 64 | 5 | 1,553 | ~310× |
| 100 | — | 748 | — |
*Measured on Pico 2 as pure MD run, with no graphics or physical input.
In the full demo (display + accelerometer + 100 atoms): ~500 MD steps/second.
Controls
| Button | Action |
|---|---|
| A | Cycle options: Temperature → Gravity → Pause → Coupling → Box Vel scale |
| X | Increase selected parameter |
| Y | Decrease selected parameter |
| X + Y | Save screenshot (PPM file) |
| Hold B (1 sec) | Shutdown |
Physical interaction via the accelerometer:
- Tilt the device → gravity follows orientation
- Shake it → box walls move, moving the atoms, leading to heating
Files
Simulation cores
| File | Description |
|---|---|
md_sim_ulab_numpy.py |
Pure Python/ulab simulation class (no C), runs on MicroPython with ulab and standard Python with numpy |
md_sim_ulab_user_c_func.py |
Simulation class using C-accelerated force loop integrated with ulab |
user.c |
Custom ulab C extension for ulab: Leap Frog integrator + LJ forces |
Pico applications
| File | Description |
|---|---|
run_show_sim_ulab_user_c_func_dual_core_pico_dislay.py |
Main demo: dual-core display + physics |
run_test_ulab_user_c_func.py |
Benchmark (no display) — C version |
run_test_ulab_user_c_func_pico_display.py |
Benchmark with display output — C version |
run_test_ulab_numpy.py |
Benchmark (no display) — pure ulab / numpy version |
bno055_handler.py |
BNO055 accelerometer wrapper |
Desktop
| File | Description |
|---|---|
run_show_sim_numpy_pc.py |
Desktop visualization with optional matplotlib plot |
Building the Firmware
MicroPython / ulab / display
The C extension requires a custom MicroPython + ulab build.
user.c should be placed in the ulab folder:
ulab/code/user/user.c
and enabled it by setting ULAB_HAS_USER_MODULE in ulab.h.
For the full integration with pico display 2, pimoroni modules are also needed.
Easiest build strategy is to use a modified version of the pimoroni-pico-rp2350 build.
Pimoroni provides a CI setup at:
https://github.com/pimoroni/pimoroni-pico-rp2350.git
The script
setup_and_build_firmware.sh
utilizes this CI setup to perform a build (as of February 2026), might be broken in the future.
BNO055 accelerometer support
For supporting the accelerometer the micropython-bno055 library must be added to the device.
The library is found at:
https://github.com/micropython-IMU/micropython-bno055
Quick Start
# Desktop test (numpy / matplotlib version )
python3 run_show_sim_numpy_pc.py
# Upload to Pico
mpremote fs cp md_sim_ulab_user_c_func.py :
mpremote fs cp bno055_handler.py :
mpremote fs cp run_show_sim_ulab_user_c_func_dual_core_pico_dislay.py :
# Run on Pico
mpremote run run_show_sim_ulab_user_c_func_dual_core_pico_dislay.py
References
- Ulf R. Pedersen’s browser MD simulation: https://urp.dk/md/
- ulab (numpy for MicroPython): https://github.com/v923z/micropython-ulab
- Pimoroni Pico libraries: https://github.com/pimoroni/pimoroni-pico
- MicroPython-bno055: https://github.com/micropython-IMU/micropython-bno055
License
MIT — see LICENSE.
Authors
Bo Jakobsen, IMFUFA, Roskilde University (boj@ruc.dk) Based on the browser-based MD simulation by Ulf R. Pedersen (urp@ruc.dk)