Determination of cobalt content,
solids content, specific gravity and viscosity
in paint driers by Vis-NIR spectroscopy
Introduction to paint drier industry
Paint is generally considered as a mixture of pigment, binder, solvent, driers and other
additives. Though the basic function of driers is to decrease the drying time after appli-
cation, it significantly affects the gloss and clarity of paint coatings. There are many end
markets for paints and coatings, ranging from home builders to original equipment
manufacturers, many of which are subject to regulations that require testing for comp-
liance. Therefore, producers as well as the end user of paints and additives like paint
driers need to control the quality regarding the specification of multiple parameters.
The test procedures are specified by ASTM procedures (see addi-
tional poster). The determination of all parameters of interest re -
quires four primary methods – balance & oven, titration equip-
ment, a hydrometer and a viscometer. These four independent
analysis methods can be replaced by a single measurement using
Vis-NIR spectroscopy: the visible range is used for a direct deter-
mination of metal-complex and the near-infrared region for the
simultaneous quantification of physical and chemical parameters.
Experimental setup – Cobalt content
5 samples of cobalt octoate with different cobalt content (4%, 6%,
8%, 10%, and 12%) were provided by a producer of paint driers.
10 dilutions were prepared by mixing the 5 initial concentrations by
different ratios. Thus, the sample amount was increased from 5 to
15 samples. Each sample was acquired in transmission over the full
wavelength range (400–2500 nm) using the NIRS XDS RapidLiquid
Analyzer (2.921.1410). The samples were placed in quartz glass cuvettes
of 1 mm path length (6.7401.200).
The reference values necessary for the determination of a quantitative model for the
pre diction of the cobalt content were derived by titration using the Cu ISE (AW TI CH1-
1160-012014).
Vision, with its Partial Least Squares (PLS) algorithm, was used to develop quantitative
prediction models for cobalt content in paint driers. Therefore, absorption bands of the
Vis-range (400–780 nm) were chosen. The spectral data were pre-treated using Standard
Normal Variate (SNV) to get rid of light scattering effects. Internal cross validation
(leave-one-out method) was applied to verify the performance of the derived quantita-
tive model.
Determination of cobalt content
The spectral data, pre-treated with SNV, displays a good correlation between the cobalt
content and the absolute absorbance value over the full wavelength range (400–2500 nm).
The PLS model with 3 factors based on spectral information of the visible range
(400–780 nm) leads to excellent results, see Fig. 1.
Comparison of results derived
from Vis and NIR region
Cobalt octoate is a complex formed by two ethylhexanoate ligands associated to one
Co2+ ion. In the NIR region, characteristic combination bands can be found of the ligand
(NIR active) whereas the bands in the visible range arise due to complexation (Co is NIR
inactive; the complex is blue). To verify that the visible range contains the information
of interest, the statistical values for three models (Vis, Vis + NIR and NIR) were compared,
see Tab.1.
Determination of physical and
chemical properties
The NIR wavelength range can be used for the determination of solid content, specific
gravity and viscosity of cobalt octoate. Fig. 2–4 shows the correlation of calculated
values to reference values from the developed quantitative methods using the NIR region
(800–2500 nm).
Conclusion
Paint industry is growing. Vis-NIR spectroscopy has a very high potential in this market
segment due to its ability to determine multiple parameters with the same analyzer and
on the same time. This analyzer serves the producer and the customer of paints to the
same extent for quality control of the raw materials and the final product, respectively.
The visible range (400–780 nm) provides direct prediction results for the cobalt content
and can be used to replace the time-consuming and waste-producing wet-chemistry.
The NIR region (780–2500 nm) shows excellent results for the simultaneous prediction
of three properties (solids content, specific gravity and viscosity). Therefore, Vis-NIR
spectroscopy, compared to only NIR spectroscopy, benefits from the extended wave-
length range from 400–2500 nm to get all results with higher accuracy and precision.
Experimental setup – Physical and
chemical properties
4 out of the 5 initial samples were provided with their certificate of specifications for
solids content, specific gravity and viscosity. Spectral information of the NIR wavelength
region (800–2500 nm) and the provided reference values were correlated to develop
three quantitative model. The data was pre-treated using a baseline correction at 800 nm.
Tab. 1: Comparison of results for three quantitative models based on three different
wavelength regions
The best statistical values, meaning highest correlation (R2) and lowest error for calibra-
tion (SEC) and validation (SECV), were achieved using the visible range from 400–780 nm.
Predicting the cobalt content using the NIR range would be an indirect determination
via the ligand and can yield strongly diverging results.
Fig. 1: a) Absorption spectrum of the visible range (400–780 nm) of cobalt octoate with
five different cobalt concentrations. b) Correlation of calculated values to reference
values as a result of quantitative method development of the cobalt content.
Fig. 2: The provided values for the solids
content ranged from 33–67.5%. A PLS
model using 3 factors shows a high corre-
lation (R2 = 0.999, SEC = 0.240%, SECV =
0.285%) bet ween the provided reference
values and the calculated values.
Fig. 3: The provided values for the speci fic
gravity ranged from 0.89–1.00. A PLS
model using 3 factors shows a high corre-
lation (R2 = 0.997, SEC = 0.003, SECV =
0.003) between the provided reference
values and the calculated values.
Fig. 4: The provided values for the viscos-
ity ranged from 80–800 mPas. A PLS model
using 4 factors shows a high correlation
(R2 = 0.999, SEC = 9.3 mPas, SECV =
10.9 mPas) between the provided refer-
ence values and the calculated values.
400–780 nm
780–2500 nm
400–2500 nm
R2
0.999
0.998
0.999
SEC
0.084
0.126
0.096
SECV
0.088
0.132
0.100
