Experimental - Spectroscopy

22 Spectroscopy 26(6) June 2011 www.spectroscopyonline.com
Chemometrics in Spectroscopy
Classical Least Squares, Part VI:
Spectral Results
We continue to examine in detail the spectral behavior of three-component mixtures.
Howard Mark and Jerome Workman, Jr.
This column is the continuation of our discussion of
the classical least squares approach to calibration
(1–5). In this installment, as in the previous one (5),
we will be dealing largely with figures, and the spectra
therein. So far, we have looked at the spectra of the pure
components comprising the ternary mixtures we are
working with. We also have looked at the spectra of all
the mixtures overlaid on each other.
One important reason to perform this extensive exercise
of examining spectra is to address, and hopefully
clear up, a misconception some people have about nearinfrared
(NIR) spectra. Because of its history, NIR spectroscopy
sometimes is believed to rely on, and require,
the “magic” of chemometric analysis to obtain meaningful
scientific results. But that’s not always true. That perception
has arisen because NIR measurements most often
are made on samples that are powdered solids, part of a
dynamic mechanism, or involve other difficult conditions
that often make other types of spectral measurements
impossible.
This has created an impression that NIR exists in a
Celebrating 25 Years
The editors congratulate Howard Mark and
Jerry Workman for 25 years of statistics and
chemometrics columns in Spectroscopy.
universe of its own, disconnected from the rest of science.
Therefore, we take this opportunity to emphasize
(and maybe overemphasize) the fact that NIR is not some
“magical” technique that is different from the rest of the
universe, but in fact is the same spectroscopy we are all
used to seeing in other spectral regions. When samples
are presented to a spectrometer, they behave the same
way in the NIR as they do in the UV, visible, mid-IR, or
any other region.
In the current experiment, we have a situation where
clear liquid samples are used, and measured in a cuvette
having plane parallel windows — the near-ideal conditions
that are normally used for any spectral region. In
this way, we can demonstrate that NIR spectra, under
those conditions, do indeed behave the same way as spectra
in other spectral regions do.
Let us look at the three sets of two-component mixtures
that the experimental design includes. These are
presented in Figures 1–3. In these figures, we first present
the full spectrum, and then for each mixture we present
the spectra of the mixtures in the several subranges that
contain the useful absorbance bands.
Examining these spectra, we see some features that
were partially obscured when the spectra from all the
mixtures were plotted together. One effect that was
hidden by the overlapping spectra was the presence of
isosbestic points. Although Figure 5e from part V of this
column series (5) shows a ternary isosbestic point, that
is rare; for the most part isosbestic points are not seen