Interactive lab › Start from the answer and find the Hamiltonian
Interactive lab · 26/66 · Hamiltonian engineering · inverse problems
Start from the answer and find the Hamiltonian
Hamiltonian engineering is the one discipline behind all of RCQT's work: a sensor, a simulator and a computer each need a Hamiltonian that does a particular job. Physics usually runs forward, from H to dynamics; engineering runs it backwards. Inverse engineering chooses the trajectory and solves for the control; Floquet design averages a native interaction into a different one with pulse sequences; Hamiltonian learning infers H from measured data. Optimal control, the fourth approach, has its own lab.
What you are seeing
- Inverse engineering: the dashed path is what you ask for; the control W = r × ṙ follows from it directly. If only Ω can be used, the path cannot cross the equator while turning: the readout says so.
- Faster costs power: the peak Rabi frequency grows as 1/T, a quantum speed limit.
- Floquet design: three spins with an XY, dipolar or Ising coupling. Global π/2 pulses permute the axes; the time spent in each frame sets the effective couplings. The reachable targets form a triangle; outside it no positive times exist.
- Hamiltonian learning: noisy Rabi data are fitted for Ω and Δ. The χ² map has two minima, because these data cannot tell the sign of Δ; more shots sharpen the answer as 1/√shots, down to the Cramér–Rao bound.
The numbers in the model
| Inverse control | W = r × ṙ + λ r |
|---|---|
| Path schedule | s = 10τ³ − 15τ⁴ + 6τ⁵ |
| Effective coupling | J_eff = Σ t_k P_k(a, b, c) |
| Learning | grid search + Gauss–Newton, 40 time points |
Go further
- One discipline: Hamiltonian engineering
- Lab: optimal control of one qubit
- Glossary: Floquet engineering
- Glossary: average Hamiltonian
- RCQT × CEQT: who does what