Raman scattering — light exciting molecular vibrations and shifting to lower frequency — has a standard description: the Raman gain spectrum, a frequency-domain object that predicts how much a given spectral component is amplified per unit length. This works when the light pulse is long compared to the molecular response time. The molecules reach steady state during the pulse, and the frequency-domain description is exact.
For ultrashort pulses, it fails (arXiv:2603.21734). When the pulse duration approaches or falls below the molecular vibration period, the Raman interaction is impulsive — the molecule is kicked rather than driven. A time-domain framework reveals what the steady-state description hides: temporal and spectral distortions that reduce frequency-shifting efficiency. The Raman-shifted output isn't just frequency-shifted; it's temporally reshaped, and the reshaping diverts energy away from the desired spectral channel.
The distortions can be controlled. In gas-filled hollow-core fibers, reducing the relative Raman contribution compared to the electronic (instantaneous) nonlinear response suppresses the temporal distortion. This is counterintuitive — weakening the Raman effect improves Raman-based frequency conversion. The mechanism: when the electronic response dominates, it maintains soliton integrity, and the intact soliton undergoes a clean self-frequency shift. When the Raman response dominates, the impulsive dynamics fragment the pulse.
The analytic expression for impulsive soliton self-frequency shift in gases differs significantly from the established formula for glass fibers. The glass formula assumes a fixed Raman fraction; the gas formula accounts for the tunable ratio between Raman and electronic responses.
The structural insight: the steady-state description isn't wrong — it's incomplete. It captures the frequency domain accurately but misses the temporal dynamics that become dominant at short pulse durations. The efficiency loss isn't a new physical effect; it's a temporal consequence of the same interaction that the frequency-domain picture already describes. The loss was always there; the description was too coarse to see it.