A fit is a smooth curve that follows your wiggly measurements as closely as it can. The numbers that make the curve fit best tell you something real, like how strong a magnet is or how long a qubit remembers.

Levenberg–Marquardt least squares: the solver takes steps that blend gradient descent (safe when far away) with Gauss–Newton (fast when close). The uncertainties come from the curvature of the error surface at the minimum. The animation shows every accepted step.

Uncertainties are 1σ from (JᵀJ)⁻¹ scaled by the residual variance, assuming independent Gaussian noise; correlated noise (slow drift) makes them optimistic, so check the residual plot. ODMR lines are found on a lightly smoothed copy and pruned below 2.5σ significance. The Arrhenius fit corrects for the T^1.5 density of states and T^−2.2 lattice mobility before extracting E_A.