← SCPN-Quantum-Control

One mixing angle, two drives

The transfer is governed by a single mixing angle $\theta(t)$ that rotates the dark state from the initial level to the target. The two Rabi drives are its sine and cosine, so their powers always add to the same total — the pump grows exactly as the Stokes shrinks. ICI splits the schedule into three: a fast ramp in, a slow adiabatic sweep across the middle, and a fast ramp out, the shape a time-optimal control argument (Pontryagin's principle) singles out.

$$\Omega_P(t)=\Omega_0\sin\theta(t),\qquad \Omega_S(t)=\Omega_0\cos\theta(t),\qquad \Omega_P^2+\Omega_S^2=\Omega_0^2$$

The leak sets the cost: while $\theta$ is mid-sweep the state has some excited-level character $\sin^2\theta$, and the estimated loss is the decay rate times that population, time-averaged. Sharper end-ramps (a larger jump angle) spend less time exposed — the trade the panel lets you feel.

Live — ICI pump & Stokes envelopes from the library's formulas
pump Ωₚ = Ω₀ sinθ Stokes Ωₛ = Ω₀ cosθ
estimated fidelity: excited population ⟨sin²θ⟩: total power Ω₀:

The left panel is the pump and Stokes envelopes in time: Stokes (green) starts high and the pump (blue) rises to meet it — the counterintuitive ordering that keeps the dark state populated. The right panel plots the same drives against each other; because their powers sum to a constant, the trajectory rides a quarter of a circle, and the mixing angle is just the angle along it. Widen the jump angle and the path snaps more sharply to the ends, cutting the exposed excited population and lifting the estimated fidelity — up to the point where the sweep is no longer adiabatic.

Deeper: what the estimate does and does not claim
These are the library's documented ICI envelopes: the three-segment mixing angle (a 5% ramp in, an adiabatic sweep, a 5% ramp out) and the power-constraint drives $\Omega_0\sin\theta,\ \Omega_0\cos\theta$. The fidelity shown is the library's fast estimate — one minus the decay rate times the time-averaged excited population over the peak Rabi frequency — not a full master-equation solve; it captures the loss channel, not every coherent error. The envelopes and the estimate were reproduced here and checked against a reference computation of the same formulas to better than $10^{-14}$. A full three-level Lindblad evolution is a separate step in the library.

Where it fits

Pulse shaping is the layer between an abstract gate and the analog drive a device actually applies. ICI is one shape in a family — alongside the hypergeometric envelopes that subsume Allen–Eberly and STIRAP — and the platform picks among them to hit a transfer quickly while staying inside the hardware's power and coherence budget. It is the same concern as Trotter depth and error mitigation, seen from the waveform end.

ShapeIdeaGood for
ICI (this panel)bang–adiabatic–bang mixing angletime-optimal transfer past a leak
Hypergeometric (α,β)${}_2F_1$ envelope familysubsumes Allen–Eberly, STIRAP
Plain adiabaticslow smooth sweeprobustness when time is cheap

Evidence boundary: the ICI envelopes are the library's exact documented formulas, rebuilt live in your browser and checked against a reference to better than $10^{-14}$. The fidelity is the library's fast loss estimate, not a full master-equation result, and the panel is an illustration of the pulse shape, not a hardware calibration.