Applied_Spectroscopy_Practica_01_01 - 27

Kara Ali et al.
153
step. Meanwhile, it is also important to minimize the amount
of H used to minimize the cost of the process, both from an
economic and environmental standpoint. Quantifying the
residual solvents in the recycled pellets is also essential
because it is a direct measure of the devolatilization process
efficiency. Therefore, it is necessary to implement a rapid,
direct, and easy quantification technique for residual solvents
throughout the process.
While mass spectrometry provides outstanding levels of
sensitivity and selectivity, it is expensive and cannot readily
be implemented for routine analysis and process optimization
at a production plant. Gas chromatography is a more
affordable alternative, but a long elution time makes it less
practical for the quality control aspect of the intended
application.
Thermogravimetric analysis was also considered but the
evaporation of H and Cy was poorly separated both at
slow and fast scanning rates. By comparison with these various
techniques, infrared spectroscopy (IR) offers several
advantages such as rapid measurements and data analysis,
low cost, robustness, and easy operation by nonqualified
operators at the plant level.18,19 In this paper, we develop
and validate calibration models allowing the rapid and accurate
quantification of Cy and H in PS by IR spectroscopy.
We also apply the model at the pilot plant level to representative
pastes and pellets produced during the dissolution/
precipitation recycling of PS waste material.
Experimental
Materials and Methods
High-performance liquid chromatography grade H (≥96%),
Cy (99%), 1,2-dichloroethane (DCE; anhydrous, 99.8%), and
PS (Mw∼ 350,000, Mn ∼170 000) were purchased from
Sigma-Aldrich and used as received. Calibration standards
containing either Cy, H, or both, were prepared according
to the following procedures. In all cases, approximately
600 mg of PS was first solubilized in an appropriate mass of
DCE in order to obtain solutions of 30 wt% PS. For the
Cy-only model, 26 calibration standards were prepared by
adding weighted amounts of Cy to the PS solution to reach
Cy concentrations ranging from 0 wt% to 16 wt% relative
to the total mass (excluding DCE). Similarly, for the H-only
model, 26 calibration standards containing from 0 wt% to
40 wt% of H relative to the total mass (excluding DCE)
were prepared. Finally, for the mixed Cy + H model,
selected masses of both Cy and H were added to the 30
wt% solutions of PS in DCE. A total of 67 solutions were prepared
and used either as calibration or validation standards.
Partial least squares (PLS) regression models were built
using approximately 75% of the samples as the calibration
set and the remaining 25% of the samples as an independent
validation set to evaluate the model performance.
To test the applicability of the models in real-life conditions,
PS samples were extracted at different stages of the
recycling process in a pilot-scale plant. This plant can produce
about 5 kg of recycled resin per hour. It consists of several
modules of dissolution, purification, precipitation, washing,
and extrusion. A weighted amount of each sample was dissolved
in an appropriate mass of DCE to approximate the
PS concentration used in the calibration standards. Three
replicates were used for each sample.
For convenience, the models were built using either a
spectrometer located in the university laboratory or one
located at the pilot plant. Efforts were made to make the
sampling and measurement parameters as similar as possible
between the two instruments, but no attempt was made to
transfer calibration models from one instrument to the
other. Rather, samples were systematically quantified by measuring
their spectrum with the same instrument that was
used to create the calibration model. More specifically, the
single-solvent (Cy-only and H-only) models were created
using a Vertex 70 Fourier transform infrared (FT-IR) spectrometer
(Bruker Optics) equipped with a deuterated
L-alanine doped tri-glycine sulfate detector and a single
reflection ZnSe attenuated total reflection (ATR) accessory
(MIRacle, Pike Technologies), while the mixed-solvent (Cy
+ H) model was created using a Thermo Scientific Nicolet
iS10 FT-IR spectrometer equipped with a deuterated triglycine
sulfate detector and a single reflection diamond ATR
accessory. To minimize the risk of solvent evaporation during
the analysis, a large (at least 1 mm thick) droplet of the highly
viscous solution was deposited on the ATR element. Spectra
were recorded with a 4 cm−1 resolution by averaging scans
for 1 min for the samples and 2 min for the backgrounds.
This acquisition time was selected as a compromise between
the need for a good signal-to-noise ratio and the avoidance of
a bias due to problematic solvent evaporation.
ATQ Analyst 9 (Thermo Nicolet Corporation) was used
for spectral processing and multivariate analysis. The performance
of the multivariate models was evaluated in terms of
root mean square error of calibration (RMSEC), root mean
square error of prediction (RMSEP), root mean square
error of cross-validation (RMSECV), and determination coefficient
Applied_Spectroscopy_Practica_01_01

Table of Contents for the Digital Edition of Applied_Spectroscopy_Practica_01_01
