Sd. Yoo et De. Aspnes, Elimination of endpoint-discontinuity artifacts in the analysis of spectrain reciprocal space, J APPL PHYS, 89(12), 2001, pp. 8183-8192
Reciprocal-space analysis offers several advantages for determining critica
l point parameters in optical and other spectra, for example the separation
of baseline effects, information, and noise in low-, medium-, and high-ind
ex Fourier coefficients, respectively. However, endpoint-discontinuity arti
facts can obscure much of the information when segments are isolated for an
alysis. We developed a procedure for eliminating these artifacts and recove
ring buried information by minimizing in the white-noise region the mean-sq
uare deviation between the Fourier coefficients of the data and those of lo
w-order polynomials, then subtracting the resulting coefficients from the d
ata over the entire range. We find that spectral analysis is optimized if n
o false data are used, i.e., when the number of points transformed equals t
he number of actual data points in the segment. Using fractional differenti
ation we develop a simple derivation of the variation of the reciprocal-spa
ce coefficients with index n for Lorentzian and Gaussian line shapes in dir
ect space. More generally, we show that the definition of critical point en
ergies in terms of phase coherence of the Fourier coefficients allows these
energies to be determined for a broad class of line shapes even if the dir
ect-space line shapes themselves are not known. Limitations for undersample
d or highly broadened spectra are discussed, along with extensions to two-
or higher-dimensional arrays of data. (C) 2001 American Institute of Physic
s.