Quantum double slit
Quantum Double Slit, published by @hyle, evolves one nonrelativistic electron wave packet in a 56 × 40 × 22 volume. Its complex amplitude propagates along X, Y and Z. Two finite-height openings cause both sideways and vertical diffraction. The source and initial world recipe are ordinary editable Hyle files.
Watch the experiment
Section titled “Watch the experiment”Open the simulation from Atlas in Studio. Use CPU-64 as the numerical reference, or GPU-32 on a supported device. The packet reaches the wall within the first 40 ticks; run around 240 ticks to collect its detector exposure. CPU execution can take several minutes. Recompile to restore the initial normalized packet.
- Translucent cyan: the electron’s probability density,
psi.real² + psi.imag². Opacity uses a nonlinear scale so low-density regions remain visible. - Grey wall: an ideal reflecting barrier with two rectangular apertures.
- Amber: probability absorbed inside the detector volume, accumulated over time.
Rotate the camera to see the wave’s depth. A Z slice at 0 reveals the horizontal
interference pattern; a Y slice near 5 cuts through one opening and reveals
vertical diffraction. View along X toward the detector to see its YZ exposure
pattern. Select a cell to inspect psi.real, psi.imag, received and sink.
These cells discretize an electron’s wavefunction, not a material filling the chamber or little pieces of an electron. The detector shows expected probability from repeated identically prepared experiments, not individually sampled electron hits. Camera movement and slicing have no physical effect.
What Hyle is calculating
Section titled “What Hyle is calculating”Neighbor reads and thirteen ordered phases approximate the three-dimensional time-dependent Schrödinger equation. Each kinetic phase couples disjoint pairs through unitary complex rotations. Six axis/parity operators run in a symmetric Suzuki–Trotter sequence between two absorbing half steps. Interference emerges from amplitude propagation; no fringe pattern is prescribed. See the solver method.
Units set ℏ = m = Δx = 1, with Δt = 0.4. The initial Gaussian has mean momentum 0.8 along X and is normalized to probability one. The wall is two cells thick. Each slit is four cells wide and eight high, with centers twelve cells apart. An absorbing detector begins eighteen cells after the wall’s front face; its absorption increases smoothly across twenty cells. Other outer boundaries also absorb to reduce reflected waves.
received records detector absorption; sink records other boundary losses.
The sum of psi.real² + psi.imag² + received + sink over the lattice should remain
approximately one. The constant kinetic diagonal is omitted because it changes
only global phase. The ±2 amplitude guards remain inactive for normalized states.
Accuracy and limits
Section titled “Accuracy and limits”Checks cover propagation along all three axes, impermeable walls, probability accounting, single-slit controls, coherent interference and exact evolution of a small three-dimensional Hamiltonian. Halving the timestep reduces that exact-matrix error by about four, as expected for the second-order split.
At tick 240, the reference records about 12.75% of the initial probability in the detector. Most of the remainder has reflected from the wall or reached other absorbing boundaries. Halving the timestep changes detector exposure by about 0.0013 in absolute L1 probability distance; halving lattice spacing changes it by about 0.0232. This is a deliberately coarse demonstration, not a calibrated laboratory prediction. Smaller cells improve spatial accuracy but increase cost.
The model includes spatial electron interference, but omits spin dynamics, electron–electron interactions, material-scale scattering, detector electronics and which-path measurement. For this field-free single-electron setup, a scalar spatial wavefunction suffices to demonstrate interference. GPU-32 introduces roundoff relative to CPU-64; equal precision never guarantees identical results across devices.
Edit the world recipe for controlled changes to slit geometry or the incoming packet. Painting default Wave cells can invalidate parity or normalization, so the validation applies to the published initial state. Long sessions can also reach Studio’s timeline storage budget; download useful bakes before discarding them.