Open AccessAnalysis of complex multidimensional optical spectra by linear prediction
Author(s) -
Ethan Swagel,
Jeffrey Paul,
Alan D. Bristow,
J. K. Wahlstrand
Publication year - 2021
Publication title -
optics express
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.394
H-Index - 271
ISSN - 1094-4087
DOI - 10.1364/oe.442532
Subject(s) - spectral line , fourier transform , ideal (ethics) , singular value decomposition , transformation (genetics) , optics , time domain , nonlinear system , physics , noise (video) , frequency domain , discrete fourier transform (general) , algorithm , statistical physics , computational physics , mathematical analysis , computer science , mathematics , fourier analysis , quantum mechanics , fractional fourier transform , image (mathematics) , chemistry , philosophy , biochemistry , epistemology , artificial intelligence , computer vision , gene
We apply Linear Prediction from Singular Value Decomposition (LPSVD) to two-dimensional complex optical data in the time-domain to generate spectra with advantages over discrete Fourier transformation (DFT). LPSVD is a non-iterative procedure that fits time-domain complex data to the sum of damped sinusoids, or Lorentzian peaks in the spectral domain. Because the fitting is linear, it is not necessary to give initial guess parameters as in nonlinear fits. Although LPSVD is a one-dimensional algorithm, it can be performed column-wise on two-dimensional data. The method has been extensively used in 2D NMR spectroscopy, where spectral peaks are typically nearly ideal Lorentzians, but to our knowledge has not been applied in the analogous optical technique, where peaks can be far from Lorentzian. We apply LPSVD to the analysis of zero, one, and two quantum electronic two-dimensional spectra from a semiconductor microcavity. The spectra consist of non-ideal, often overlapping peaks. We find that LPSVD achieves a very good fit even on non-ideal data. It reduces noise and eliminates discrete distortions inherent in the DFT. We also use it to isolate and analyze weak features of interest.