Every experiment so far cut two openings in space and read the interference across the screen. In 2023, two centuries after Young, a team at Imperial College cut two openings in time: a mirror that exists twice, briefly, a few hundred femtoseconds apart. Light reflected through such a pair of "time slits" interferes with itself. But the fringes appear in its spectrum, not on a screen. Where the double slit trades position against angle, the time slit trades moments against colour.
The bench below builds the shutter function g(t) in the time domain (raised-cosine edges, adjustable rise time) and computes its spectrum as an explicit Fourier sum. The fringes are not drawn in; they fall out of the sum exactly as they fell out of the ITO mirror in the lab.
Spectral fringe spacing: measured · predicted 1000/Δt: .
Nothing new was needed to compute this page, only a change of variables. The double slit's geometry translates word for word: slit separation d becomes the time between openings Δt; slit width a becomes the opening's duration τ; position on the screen becomes frequency offset from the carrier. The classic condition d sinθ = mλ becomes δ·Δt = 2πm: colours whose phase advances by a whole cycle between the two openings add up; those half a cycle out cancel. Even the single-slit envelope comes along: one brief opening spreads the spectrum, which is energy–time uncertainty wearing engineering clothes.
Shrink Δt and the spectral fringes spread apart. It is the exact mirror of chapter 2, where moving the slits together widened the pattern. One formula, two costumes.
The translationSlide the edge rise time down and watch the fringes reach far from the carrier. Tirole's team saw fringes persisting much further than expected: their mirror was switching in about a femtosecond, faster than anyone thought ITO could.
What the lab sawSwitch to a single opening: no fringes, but the spectrum spreads. The shorter the flash, the wider its colour. A shutter is a spectrometer's enemy.
Why brevity costs purity