Measuring the phase of a light pulse normally requires a reference — another pulse with known phase, or a delayed copy of the same pulse. The interference between the unknown and the reference encodes the phase difference in the fringe pattern. Without a reference, you have only the intensity spectrum, which discards the phase entirely. This is the phase problem: spectral intensity alone doesn't tell you when the different frequency components arrive relative to each other.
Cheeran, Müftüoğlu, Saeed, Fischer, and Chemnitz (arXiv 2602.22845, February 2026) demonstrate phase retrieval without a reference, without computation, and without an interferometer. The pulse provides its own reference through the Kerr effect.
The mechanism is self-interference. When a short optical pulse propagates through a Kerr medium (any material with an intensity-dependent refractive index), the high-intensity peak of the pulse experiences a different refractive index than the low-intensity wings. The peak acquires extra phase from the nonlinear index; the wings travel at the linear phase velocity. If the pulse has an abrupt phase feature — a step, a jump, a transition — the Kerr effect converts the phase feature into a spectral interference pattern. The high-intensity portion and the residual linear-wave component interfere in the frequency domain, producing fringes whose spacing and contrast encode the original phase transition.
The key insight: the pulse doesn't need an external reference because it carries two components with it — the nonlinear peak and the linear wings — that naturally interfere after Kerr propagation. The Kerr medium creates the reference from the pulse itself.
The phase sensitivity is π/385 — the ability to detect phase shifts almost 400 times smaller than a half-wave. This is achieved at 80 MHz repetition rate with 50 picojoule pulses, requiring no iterative algorithms, no numerical phase retrieval, no computational post-processing. The phase information is directly readable from the spectral intensity pattern.
The elimination of the computational step is significant. Conventional reference-free phase retrieval methods (FROG, d-scan, SPIDER) measure spectral or temporal data and then solve an inverse problem to recover the phase. The inverse problem is computationally intensive and can have ambiguities. Here, the Kerr medium performs the inversion physically: the spectral fringes are the answer, not the input to a computation.
The approach works for abrupt phase transitions — steps, boundaries, switching events — where the Kerr-induced spectral interference produces a clean, unambiguous signature. For smoothly varying phases, the fringe pattern becomes complex and the direct readability may be lost. The method has a natural domain: pulse shaping, optical switching, and phase-encoded communications where the phase features are discrete.
The pulse measures itself.