A thermoelectric cloud chamber that makes ionizing radiation visible using a dual Peltier cooling system, paired with a computer vision pipeline that detects particle tracks, classifies them by type, and uses the underlying randomness to generate cryptographic keys.
A cloud chamber is a particle detector that makes invisible radiation visible.
When a supersaturated vapor cools rapidly, charged particles passing through it leave ionization trails. These trails seed condensation, and droplets form along the particle's path, appearing as visible streaks.
Traditional cloud chambers use dry ice for cooling. This one uses a thermoelectric (Peltier) cooling system instead, trading the simplicity of dry ice for a design that runs continuously without consumables.
 |
 |
 |
| The Setup | ||
- 99.9% isopropyl alcohol
- Aluminium base plate
- TEC1-12706 Peltier modules (x2)
- 12V 20A SMPS
- XL4016 DC-DC buck converter
- Insulated chamber enclosure
- Heat sink assembly
Two Peltier stages work together:
- Bottom Peltier (12V): the main cooling stage, pulling heat from the aluminium plate.
- Top Peltier (5V, via XL4016): stabilizes the gradient to prevent thermal runaway and improve condensation.
The aluminium plate sits above the cold junction and is where alcohol vapor condenses into visible tracks.
The first prototype used an LM2596 buck converter for the top Peltier stage. Under real load, current draw exceeded its safe limit and it burned out. It was replaced with an XL4016, which handles higher current with better thermal stability. The fix wasn't a redesign, just picking a converter rated for what the circuit actually draws under load, not just its nameplate spec.
- The aluminium plate is cooled via the Peltier stack.
- Isopropyl alcohol is introduced and evaporates, saturating the chamber.
- A steep temperature gradient forms just above the plate.
- Charged particles (cosmic ray muons and background radiation) ionize the vapor along their path.
- Condensation forms visible streaks along the ionized trail.
230V AC (Mains)
│
▼
12V 20A SMPS (DC Supply)
│
┌───────────┴───────────┐
│ │
▼ ▼
Bottom Peltier (TEC1-12706) XL4016 Buck Converter
│ │
│ ▼
│ Adjustable Output (5V)
│ │
│ ▼
│ Top Peltier (TEC1-12706)
│
└───────────┬───────────┘
│
▼
Aluminium Cold Plate
A camera mounted above the chamber feeds a Python vision pipeline (OpenCV) that watches for tracks, classifies them, and turns the underlying randomness into usable cryptographic material.
Computer Vision. Background subtraction isolates tracks from the static chamber background. Each detected track is measured for length, width, and straightness, the same geometric cues a physicist would use to read a chamber by eye.
Classification. Track shape maps onto known particle signatures: alpha particles are short, thick, and straight; muons are long, thin, and straight; betas are long and thin but visibly wander, since they scatter more easily. The classifier is a starting point, not a substitute for calibrating against a known source like an Am-241 smoke detector pellet.
Randomness. This is the same principle behind Cloudflare's lava lamp wall: a physical process too chaotic to predict is used to seed a random number generator. Each new track feeds a running hash pool along with the timing gap since the last event. Random bytes and keys are then derived from this pool using standard cryptographic key-derivation methods.
Cloud Chamber
│
Camera (60 FPS)
│
OpenCV Pipeline
│
Track Detection
│
Feature Extraction
│
Particle Classification
├──────────────┐
│ │
Radiation Log Entropy Pool
│
HKDF
│
Cryptographic Keys
Hardware prototype complete and running. Software pipeline complete: real-time track detection, particle classification, and entropy-based key generation.
Next: calibrating the classifier against a known radioactive source, and formally measuring the true randomness output using standard entropy-estimation methods.