polymers

Article
Detection of Black Plastics in the Middle Infrared
Spectrum (MIR) Using Photon Up-Conversion
Technique for Polymer Recycling Purposes
Wolfgang Becker *, Kerstin Sachsenheimer and Melanie Klemenz
Fraunhofer Institut fuer Chemische Technologie (ICT), P.O. Box 12 40, 76327 Pfinztal, Germany;
Kerstin.Sachsenheimer@ict.fraunhofer.de (K.S.); Melanie.Klemenz@ict.fraunhofer.de (M.K.)
* Correspondence: Wolfgang.Becker@ict.fraunhofer.de; Tel.: +49-721-4640-154; Fax: +49-721-4640-111

Received: 10 July 2017; Accepted: 6 September 2017; Published: 8 September 2017

Abstract: The identification of black polymers which contain about 0.5 to 3 mass percent soot
or black master batch is still an essential problem in recycling sorting processes. Near infrared
spectroscopy (NIRS) of non-black polymers offers a reliable and fast identification, and is therefore
suitable for industrial application. NIRS is consequently widely used in polymer sorting plants.
However, this method cannot be used for black polymers because small amounts of carbon black
or soot absorb all light in the NIR spectral region. Spectroscopy in the mid infrared spectral region
(MIR) offers a possibility to identify black polymers. MIR spectral measurements carried out with
Fourier-transform infrared spectrometers (FTIR) are not fast enough to meet economic requirements
in sorting plants. By contrast, spectrometer systems based on the photon up-conversion technique
are fast and sensitive enough and can be applied to sort black polymer parts. Such a system is able to
measure several thousand spectra per second hence is suitable for industrial applications. The results
of spectral measurements of black polymers are presented.

Keywords: black plastic recycling; photon up conversion; MIR spectroscopy

1. Introduction
Black polymers are widely used in consumer products such as electronic household goods,
automobile components or IT goods. The black color is achieved by adding 0.5 to 3 mass percent soot
or black master batch in the polymer melt during the extrusion process. The recycling and reuse of these
materials is of substantial value and reliable methods for the identification of black polymers are still
under development. For industrial applications identification techniques have to be economical, i.e.,
fast enough, reliable with respect to long term stability, safe and easy to operate. Only methods fulfilling
these requirements are suitable for large scale recycling applications. For non-black plastic materials,
e.g., household goods, NIRS is the predominant spectroscopic technology. However, because of the
soot contained in black plastics, NIRS cannot be used, as all light is absorbed, rendering material
recognition and classification impossible. MIR spectroscopy is an option to measure and characterize
black plastics [1].
Black polymers represent a much wider variety of materials than do household plastic waste
since they are mostly used for technical applications with special requirements. Different additives
and filler materials, which are added in order to achieve specific properties of the plastics, complicate
the identification, since spectra of the same kind of plastic can vary dramatically if different types or
amounts of additives, e.g., flame retardants, fibers, or soot are contained in the plastic parts. Lacquer
films on the plastic part surface even prevent any spectral identification and have to be removed before
measurement. Interest in characterizing black polymers has led to a wide range of IR-techniques
e.g., attenuated total reflection (ATR) [2,3] infrared transmission [3,4] emission spectroscopy [5],

Polymers 2017, 9, 435; doi:10.3390/polym9090435 www.mdpi.com/journal/polymers

Polymers 2017, 9, 435 2 of 9

and photoacoustic spectroscopy [6]. In several publications the use of reflectance measurements was
demonstrated for characterizing soot filled polymers [2,7–10].
Photon up-conversion is a possibility to combine the infrared and the NIR spectral range.
Photons are transformed from the IR spectral range into the near infrared spectral range, in which fast
silicon detectors can be used. Up-conversion is carried out by adding the infrared photon energy to
that of the photons of a Neodym-doped Yttrium–Vanadat laser to produce the sum frequency, resulting
in a higher energy than the band gap of silicon material.
This concept offers the possibility to measure plastic parts in the infrared range and
the up-converted light can be evaluated with conventional silicon detector technology [11–13].
In the present paper, the reflection spectra of some technical plastics are shown and compared
with conventional Fourier transform spectrometer measurements to demonstrate the potential of
up-conversion spectroscopy for plastic sorting applications.
The system was tested and measurements were performed at the Fraunhofer-Institut fuer
Chemische Technologie (Fh-ICT). The spectrometer system based on photon up-conversion was
developed and constructed at the Fraunhofer-Institut fuer Physikalische Messtechnik (Fh-IPM).

1.1. Spectral Identification of Plastic Parts

The main spectral method for industrial plastic identification is near infrared spectroscopy.
Rugged detector systems based on indium gallium arsenide (InGaAs) material can be applied
which are sensitive and fast enough for sorting applications. The dispersive spectrometer systems
possess no moveable parts and are therefore insensitive to mechanical vibrations and can be protected
against dust exposure. Sophisticated systems reach a measurement rate up to several ten thousand
spectra per second which is needed for the economic running of a sorting facility.
The NIR spectrum expands the wavelength range from 780 to 2500 nm which corresponds to
a wavenumber range from about 12,800 to 4000 cm−1 . The NIR range is often subdivided into two
parts because different detectors are used. In the short near infrared (SNIR) region, with wavelengths
from 780 to about 1050 nm, silicon based sensors are mainly used, while in the “classical” near
infrared range, from 1050 to 2500 nm wavelength, InGaAs sensors predominate. The NIR range
is dominated by overtone and combination bands of C–H, N–H, O–H, and C–O functional groups.
The identification of plastics is mainly based on the stretching vibration modes of CH, CH2 and
CH3 groups between about 1.1 and 1.25 µm, which correspond to the second overtone, the first
overtone ranging from 1.65 and 1.7 µm, and the combination bands, respectively [14]. The main
limitation of plastic identification with NIR spectroscopy is that black plastics cannot be detected.
The soot they contain absorbs all light in the NIR range and no detectable signal can be evaluated for
material separation.
In the MIR spectral region from 2.5 to about 16 µm wavelength, which corresponds a wavenumber
range from 4000 to about 600 cm−1 , the different kinds of plastic material show additional vibrational
modes like deformation, rocking, and twisting modes due to their molecular structure. Beside the C–H
group, other molecules like O–H, N–H and O–C also contribute with their fundamental vibrations to
the spectral features. The various molecular groups with their different vibrational modes generate
a unique spectrum of each polymer in the spectral range between 2500 and 600 cm−1 , which allow
a definite identification. This spectral range is therefore called the fingerprint region, and MIR
spectroscopy is the predominant analytical method for polymer investigation and characterization.
Another important advantage of the MIR spectral range is that black polymers can be measured.
Reflectance spectra can be measured and allow identification of the black polymers. This is shown in
Figure 1 where the reflectance spectra of a black polypropylene (PP) polymer part and a non-black
part are compared. The spectra show the main spectral characteristics, and the filler material has no
obvious influence on the optical properties of the polymers. MIR is therefore a suitable analytical
method for the identification of black polymers.
 Polymers 2017, 9, 435 3 of 9
Polymers 2017, 9, 435 3 of 9

0.18

0.16

0.14

0.12
reflectance
0.10
PP black
0.08

0.06
PP non-black
0.04

0.02

0.00
4000 3500 3000 2500 2000 1500 1000 500
-1
wavenumber / cm

Figure 1. Comparison of IR spectra of the surface reflectance of a black and a non-black polypropylene
Figure 1. Comparison of IR spectra of the surface reflectance of a black and a non-black polypropylene
(PP) plastic part.
(PP) plastic part.

1.2. Non-Linear Photon Up Conversion

1.2. Non-Linear Photon Up Conversion
In the photon up conversion process infrared photons are converted to higher energy photons in
In the
the NIRphoton upinconversion
range i.e., processrange
the silicon spectral infrared photons
by sum are converted
frequency. to higher
An MIR photon energyωphotons
with energy MIR
in theinteracts
NIR range
withi.e., in the silicon
a pumping spectral
laser photon range
ωLS in a by sum frequency.
non-linear An
medium to MIR photon
generate a thirdwith energy
photon with ωMIR
interacts with
energy ωSFa. pumping laser photon ωLS in a non-linear medium to generate a third photon with
energy ωSF. ω MIR + ω LS = ωSF (1)

ω
Using this mechanism the high conversion + ωof signals
= ω has long been known [15]. By using (1)
periodically polarized non-linear crystals made of lithium niobate oxide (PPLN), and by the proposal
Using this mechanism the high conversion of signals has long been known [15]. By using
to carry out the high conversion directly in the laser resonator, whereby the laser power is amplified,
periodically polarized
a conversion rate of non-linear crystals
more than 10% can bemade of lithium
achieved. niobate
In addition oxide
to the (PPLN),
energy and byaccording
conservation the proposal
to carry out the high conversion directly in the laser resonator, whereby the laser power
to Equation (1), a phase matching of the light wave in the crystal must also be present. This is can
amplified,
be
a conversion
formulatedrate of more
in the than
k-space 10%
with can beand
k-vectors achieved.
in the oneIndimensional
addition tocase
the itenergy conservation
corresponds according
to Equation (2).
to Equation (1), a phase matching of the light wave in the crystal must also be present. This can be
1 n LS n MIR n
formulated in the k-space with k-vectors and
+ in the
+ one = SF
dimensional case it corresponds to Equation
(2)
ΛPP λ LS λ MIR λSF
(2).
ΛPP is the polarization period of the conversion crystal and the different λi are the wavelengths
1
+
of the interacting waves. The ni are the refractive +indexes=of the media at the respective wavelength. (2)
Λ λ λ λ
The refractive index of the crystal is temperature dependent. By choosing a temperature and the
ΛPP is the polarization
polarization period of theperiod
crystal of
thethe conversion
wavelength crystal
range of theand the different
converter λi are the
is determined. wavelengths
More details
and applications of photon up conversion can be found in [16].
of the interacting waves. The ni are the refractive indexes of the media at the respective wavelength.
The refractive index of the crystal is temperature dependent. By choosing a temperature and the
2. Experiments
polarization period of the crystal the wavelength range of the converter is determined. More details
The principal
and applications system
of photon setup (Fraunhofer
up conversion Institut
can be found für Physikalische Messtechnik IPM,
in [16].
Freiburg, Germany) which was used to measure the plastic samples is shown in Figure 2. The IR light
source was a black body radiator with a temperature of 900 ◦ C. The sample parts were plastic plates
2. Experiments
with a thickness of 5 mm. The samples were not electrically conductive. This distinction is important
The principal
because system
black plastic setupwhich
samples (Fraunhofer Institut
are electrically für Physikalische
conductive also absorbMesstechnik IPM, in
the entire radiation Freiburg,
the
Germany) whichrange
MIR spectral was and
used to measure
could the plastic The
not be characterized. samples iswas
sample shown in Figure
irradiated with 2.
theThe IRbody
black lightand
source
the reflected light passed the converter module with the PPLN crystal. The generated NIR
was a black body radiator with a temperature of 900 °C. The sample parts were plastic plates with a signal was
thickness of 5 mm. The samples were not electrically conductive. This distinction is important
because black plastic samples which are electrically conductive also absorb the entire radiation in the
MIR spectral range and could not be characterized. The sample was irradiated with the black body
and the reflected light passed the converter module with the PPLN crystal. The generated NIR signal
 Polymers 2017, 9, 435 4 of 9

Polymers 2017, 9,
measured 435 a single photon avalanche diode area sensor. The sensor used was also a development
with 4 of 9
within the framework of the research project LITRAN within the Fraunhofer Gesellschaft, carried out
Gesellschaft, carried out by
by the Fraunhofer-Institut für the Fraunhofer-Institut
intergrierte für intergrierte
Schaltungen (Fh-IIS). Schaltungen
The measurement (Fh-IIS).
time was 0.3–0.5The
s
measurement time was 0.3–0.5 s for one sample. The reference measurement I was carried
for one sample. The reference measurement IR was carried out with an uncoated aluminum mirror.
R out with
anThe
uncoated aluminum
reflectance mirror. The
was calculated reflectance
according was calculated
to Equation (3). according to Equation (3).

RR==
IS − IN (3)
(3)
IR − IN

Figure
Figure2.2.Schematic drawingofof
Schematic drawing thethe measuring
measuring setupsetup
for thefor the measurement
measurement of black
of black plastics plastics in
in reflection
reflection
with thewith theup
photon photon up conversion
conversion set-up. set-up.

IS is
IS the sample
is the sampleintensity
intensityspectrum
spectrumand
andIIRRisisintensity
intensityofofthe
the reference.
reference.
IN IisN the wavelength dependent noise and the dark current
is the wavelength dependent noise and the dark current spectrum. spectrum.
The Thepoling
poling period
period ΛPPΛof the PPLN crystal was ΛPP = 21.34 μm and the temperature of the PPLN
PP of the PPLN crystal was ΛPP = 21.34 µm and the temperature of the
crystal
PPLNwas 50.02
crystal was°C
50.02 ◦
for Callformeasurements.
all measurements.TheThe measured
measured NIRNIRspectra
spectra were then
thenassigned
assigned to to
wavelengths in the
wavelengths in the MIR. MIR.
In Inthethefollowing
followingsection,
section, the
the measurements
measurements with withthe
thephoton
photon up-conversion
up-conversion system
system are referred
are referred to
to as LITRAN (from Light Transformation). This will simplify the presentation.
as LITRAN (from Light Transformation). This will simplify the presentation.

3. Results
3. Results
AA polypropylene (PP) black plastic sample was measured with the photon up conversion system
polypropylene (PP) black plastic sample was measured with the photon up conversion system
(Fraunhofer Institut für Physikalische Messtechnik IPM,) as shown in Figure 2. The temperature of the
(Fraunhofer Institut für Physikalische◦ Messtechnik IPM,) as shown in Figure 2. The temperature of
conversion crystal T was 50.02 C and the poling period Λ was 21.34 µm. Thus, the spectral
the conversion crystal PPLN
TPPLN was 50.02 °C and the poling period PPΛPP was 21.34 μm. Thus, the spectral
range was determined from 3.2 to 4.15 µm wavelength. Only one measurement was taken with
range was determined from 3.2 to 4.15 μm wavelength. Only one measurement was taken with no
no accumulations. To have a comparison and to evaluate the quality of the measured spectra of
accumulations. To have a comparison and to evaluate the quality of the measured spectra of the
the LITRAN system, the plastic samples were also measured with a commercial FTIR spectrometer
LITRAN system, the plastic samples were also measured with a commercial FTIR spectrometer
(Fraunhofer Institut für Physikalische Messtechnik IPM). The results of the spectral measurements and
(Fraunhofer Institut
the comparison are für Physikalische
shown in Figure 3. Messtechnik IPM). The results of the spectral measurements
and the comparison are shown in Figure 3.
Polymers 2017, 9, 435 5 of 9
Polymers
Polymers2017,
2017,9,9,435
435 5 5ofof9 9

PP black
PP black

1.0
1.0
FTIR
FTIR
LITRAN
LITRAN
0.9
0.9

reflectance
as
CH22 as
0.8
0.8

0.7
0.7
CH33
0.6
0.6

CH3
0.5
0.5
3.2
3.2 3.3
3.3 3.4
3.4 3.5
3.5 3.6
3.6 3.7
3.7 3.8
3.8 3.9
3.9 4.0
4.0 4.1
4.1
wavelength
wavelength // µm
µm

Figure 3. Fourier-transform infrared spectrometers (FTIR) and the measurements with the photon
Figure3.3.Fourier-transform
Figure Fourier-transform infrared
infrared spectrometers
spectrometers (FTIR)
(FTIR)and
andthe
themeasurements
measurementswithwiththe
thephoton
photonup-
up-
up-conversion system are referred to as LITRAN (from Light Transformation) of the reflectance
conversion system are referred to as LITRAN (from Light Transformation) of the reflectance spectrum
conversion system are referred to as LITRAN (from Light Transformation) of the reflectance spectrum
spectrum of PP black polymer. LITRAN is the spectrum measured with the up conversion system.
ofPP
of PPblack
blackpolymer.
polymer. LITRAN
LITRAN is is the
the spectrum
spectrum measured
measured with
withthe
theup
upconversion
conversionsystem.
system.

The
TheFTIR
FTIRspectra
spectrawereweremeasured
measuredwith −
−11 Functionalgroups
The FTIR spectra were measured withaaaspectral
with spectralresolution
spectral resolutionof
resolution of444cm
of cm−1...Functional
cm Functional groups
groupsofof the
ofthe
the
PPpolymer
PP
PP polymerare
polymer areassigned
are assignedtoto
assigned tothe
the
the corresponding
corresponding
corresponding reflection
reflection
reflection peaks.
peaks. ByBy
peaks. comparing
comparing
By comparing thethe
twotwo
the spectra
spectra
two oneone
spectra can
one
canthat
see
can seethe
see that
that the spectral
spectral
the spectral
resolution resolution of
of the
of the LITRAN
resolution LITRAN
the system
LITRAN system
is lower
system isis lower
compared lower tocompared
the toto the
the FTIR
FTIR measurements.
compared FTIR
measurements. The reflection peaks are wider but present and can be used for the recognitionand
measurements.
The reflection The
peaks reflection
are wider peaks
but are
present wider
and but
can present
be used and
for can
the be used
recognition for the
and recognition
differentiation of
and
differentiation
other of
types of plastic.
differentiation other types of plastic.
of other types of plastic.
Figure
Figure 4 shows
Figure44shows shows the
thethe
FTIRFTIR and
and and
FTIR LITRAN
LITRAN
LITRAN reflectance
reflectance spectra
reflectance spectra
of black
spectra of
ofPEblack PE
PE (polyethylene).
(polyethylene).
black The
The reflection
(polyethylene). The
reflection peaks are wider as in the case of the LITRAN PP spectrum. In addition, the PE
reflection
peaks are peaksasare
wider in wider
the case as
of in
the the case of
LITRAN PP the LITRAN
spectrum. In PP spectrum.
addition, the PEInmeasurements
addition, the also
PE
measurements
show a wavelength alsoshift
showcompared
a wavelengthto theshift
FTIR compared
spectrum toofthe
the FTIR
PE spectrum
sample and ofthere
the PE
is ansample and
additional
measurements also show a wavelength shift compared to the FTIR spectrum of the PE sample and
there is an additional
shoulder shoulder in the spectrum atcause
around for3.4 μm. The cause forthe
this might beanglethat the
there is aninadditional
the spectrum at around
shoulder in the3.4 µm. The
spectrum at around this
3.4 might
μm. The be that
cause for thisentrance
might be thatand the
entrance
phase angle
matching and phase matching in the PPLN crystal varied. This means the reason for these
entrance angle in andthephase
PPLNmatching
crystal varied.
in theThis
PPLNmeans the reason
crystal varied.for these
This features
means the are caused
reason forby the
these
features are caused
measurement set-upby by
andthe measurement set-up
withand hasmaterial
nothing to do with some material properties.
features are caused thehas nothing to do
measurement set-up some
and has nothingproperties.
to do with some material properties.

PE black
0.65
PE black
0.65
FTIR
0.60 FTIR
LITRAN
0.60 LITRAN
0.55
0.55
reflectance

0.50
reflectance

0.50
0.45
0.45
0.40 CH2
0.40 CH2
0.35
0.35
0.30
0.30
CH2
0.25 CH2
0.25 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 4.0 4.1
3.2 3.3 3.4 3.5wavelength
3.6 3.7/ µm3.8 3.9 4.0 4.1
wavelength / µm
Figure4.4. FTIR
Figure and LITRAN
FTIR and LITRANreflectance
reflectancespectrum
spectrumofof polyethylene
polyethylene (PE)
(PE) black
black polymer.
polymer. LITRAN
LITRAN is theis
Figure
the 4. FTIR
spectrum
spectrum and LITRAN
measured
measured thereflectance
with
with the
up up spectrum
conversion
conversion of polyethylene (PE) black polymer. LITRAN is the
system.
system.
spectrum measured with the up conversion system.
Polymers 2017, 9, 435 6 of 9
Polymers 2017, 9, 435 6 of 9

Figure
Figure 55 shows
shows the
the FTIR
FTIR and
and LITRAN
LITRAN spectrum
spectrum of
of PA66
PA66 (polyamide
(polyamide 66). AsAs in
in all
all measurements
measurements
of
of the
the plastic
plastic samples,
samples, the
the spectra
spectra in
in the
the wavelength
wavelength range
range from
from 3.6
3.6 μm
µm to
to approximately
approximately 44 μm
µm show
show
aa very
very different curvature. In
In this
this spectral
spectral range
range the
the sensitivity
sensitivity and
and accuracy
accuracy of
of the
the LITRAN
LITRAN system
system
decreases,
decreases, leading to this spectral behavior.

PA66 black
0.36
FTIR
LITRAN
0.34

0.32
reflectance

0.30
CH2
0.28

0.26

0.24 CH2

0.22
3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 4.0 4.1
wavelength / µm

Figure
Figure 5.
5. FTIR
FTIR and
andLITRAN
LITRANreflectance
reflectance spectrum
spectrum of
of polyamide
polyamide 66
66 (PA
(PA 66)
66) black
black polymer.
polymer. LITRAN
LITRAN is
is
the
the spectrum
spectrum measured
measured with the up conversion system.

3.1.
3.1. Principal
Principal Component
Component Analysis
Analysis (PCA)
(PCA)
PCA
PCA isis an
an unsupervised
unsupervised learning
learning method.
method. There
There are
are nono targets,
targets, but
but attempts
attempts areare made
made to to extract
extract
similarities
similarities and
and differences
differences from
from the
the measured
measured data—the
data—the so-called
so-called characteristics.
characteristics. InIn the
the case
case ofof the
the
plastic samples, these characteristics are the measured spectra. The measured
plastic samples, these characteristics are the measured spectra. The measured plastic spectra are plastic spectra are first
combined into ainto
first combined data matrix
a data M. InM.
matrix theInPCA, the original
the PCA, data data
the original matrix X is now
matrix approached
X is now approachedby two by
smaller matrices
two smaller T andT L,
matrices which
and are referred
L, which to as to
are referred theasscore matrix
the score T andT the
matrix andloading
the loadingmatrix L. The
matrix L.
columns in the loading matrix are also called principle components PCs and they
The columns in the loading matrix are also called principle components PCs and they are determined are determined to
be mutually
to be mutually orthogonal.
orthogonal. Every
Every loading
loading ororprinciple
principlecomponent
componentdescribes
describesaaspecific
specificamount
amountof of the
the
contained variance of the data matrix M and each main component describes
contained variance of the data matrix M and each main component describes a maximum of variance a maximum of variance
that
that has
has not
not yet
yet been
been achieved
achieved byby the
the previous
previous ones.
ones. There
There areare different
different methods
methods of of decomposing
decomposing the the
data
data matrix
matrix M M into
into the
the submatrices
submatrices TT andand L.L. One
One of
ofthese
theseisisthethesingular
singularvalue
valuedecomposition
decomposition(SVN), (SVN),
which
which uses
uses the
the maximum
maximum variance
variance inin the
the measured
measured datadata asas aa criterion.
criterion. The
The score
score indicates
indicates how
how much
much
of a specific loading is contained in the measured data. The PCA can
of a specific loading is contained in the measured data. The PCA can be formulated as: be formulated as:
= . (4)
M = T Lt (4)
For the numerical determination of the matrices T and L, the NIPALS algorithm is often used.
For the
The PCA numerical
evaluation determination
of the of the
available data carried T
wasmatrices and
out L, the
with the NIPALS algorithm
Unscrambler is often
® software used.
package
The PCA evaluation of the(Oslo,
available data was carried out with the PCA
Unscrambler ® software package
from CAMO version 10.X Norway). More information about can be found in [17,18]. The
from CAMO
calculation version
of the PCA 10.X (Oslo, took
evaluation Norway). More
only a few information
seconds. about PCA
For processing can beseveral
purposes found thousand
in [17,18].
The calculation
evaluations of the PCA
per second can beevaluation
done. took only a few seconds. For processing purposes several
thousand evaluations per second can be done.
3.2. Distinguishing Black Plastic
3.2. Distinguishing Black Plastic
In order to examine whether the LITRAN system can be used to distinguish between different
In materials,
plastic order to examine
severalwhether the LITRAN
commercial plastics system can be used
were measured to distinguish
with the LITRAN between
systemdifferent
in the
plastic materials, several commercial plastics were measured with the LITRAN system
wavelength range of 3.2 to 3.6 μm, which is determined by the poling period and crystal temperature in the
and subjected to a principle component analysis (PCA). The score plot of the measurements is shown
in Figure 6.
Polymers 2017, 9, 435 7 of 9

wavelength range of 3.2 to 3.6 µm, which is determined by the poling period and crystal temperature
and subjected to a principle component analysis (PCA). The score plot of the measurements is shown
Polymers 2017, 9, 435 7 of 9
in Figure 6.

Figure 6. PC1–PC2
PC1–PC2 plot
plot of
of the
the scores
scores from
from principle
principle component
component analysis (PCA) evaluation of the black
plastic samples. Each dot in the plot in Figure 6 represents
represents aa complete
complete spectrum.
spectrum.

The
The plastic
plastic samples
samplesare arebulkbulkplastics
plasticssuch
such as as
PPPP andand PE PEandand
technical plastics
technical such such
plastics as ABS as
(acrylonitrile–butadiene–styrene) and PBT (polybutylenterephthalate),
ABS (acrylonitrile–butadiene–styrene) and PBT (polybutylenterephthalate), which are used in which are used in the
automotive
the automotive and andelectrical industries.
electrical ThereThere
industries. are also somesome
are also blends withwith
blends PC (polycarbonate).
PC (polycarbonate). All
measured
All measured blackblack
plastic samples
plastic are commercially
samples are commercially available and are
available andused for industrial
are used goods.
for industrial The
goods.
commercial names are each indicated in the framing of the so called clusters.
The commercial names are each indicated in the framing of the so called clusters. At each cluster At each cluster the type
of
theplastic
type ofisplastic
shown. The twoThe
is shown. principle components
two principle PC1 (98%)
components and PC2
PC1 (98%) and (2%) contain
PC2 (2%) 100%100%
contain of theof
spectral data.
the spectral data.
The
The twotwodimensional
dimensionalscore scoreplotplot shows
shows thethe different
different clusters
clusters representing
representing the spectra
the spectra of the
of the plastic
plastic samples. PE (Hostalen) and PP (Hifax) are all in the positive part
samples. PE (Hostalen) and PP (Hifax) are all in the positive part of the axis of component PC2. of the axis of component
PC2.
Both Both plastics
plastics are thermoplastics.
are thermoplastics. Granex
Granex is aismixture
a mixture ofof bothtypes
both typesofofrecycled
recycledPP PPand
andPE,PE, soso the
the
Granex cluster is positioned near the PE and PP cluster. Since the Granex cluster
Granex cluster is positioned near the PE and PP cluster. Since the Granex cluster is closer to the PP is closer to the PP
cluster, one can
cluster, one can also
also conclude
conclude thatthat the
the Granex
Granex mixture
mixture hashas aa larger
larger share
shareofofPP
PPpolymer.
polymer.
Cycolac and Lustran are ABS plastics from two
Cycolac and Lustran are ABS plastics from two different manufacturers. different manufacturers. As aAsresult, the
a result,
composition of this blend also varies and there are spectral differences.
the composition of this blend also varies and there are spectral differences. These two clusters These two clusters are
consequently
are consequently far apart in theinPCA
far apart the plot.
PCA This
plot.makes
This it possible
makes to identify
it possible ABS plastics
to identify ABSfrom different
plastics from
manufacturers and sort them accordingly. This can be decisive for the regranulation
different manufacturers and sort them accordingly. This can be decisive for the regranulation of of these plastics
these
becausebecause
plastics the mixing of ABSoffrom
the mixing ABS different sources
from different and and
sources thusthusdifferent compositions
different compositions cancanleadleadto to
a
worsening
a worseningofofthe thequality
qualityofofthetheregranulated
regranulatedmaterial.
material.
The same is valid for
The same is valid for the PC/ABS the PC/ABS and PC/PBT
and PC/PBT blends.
blends. These Thesehave
samples samples
differenthave different
compositions
compositions
compared, for compared,
example, to for ABSexample, to ABS
or pure PBT, or pure PBT,possess
and consequently and consequently possess
different spectral different
properties in
spectral properties in the MIR. The t-score and the cluster are well separated from
the MIR. The t-score and the cluster are well separated from the other samples, indicating that the black the other samples,
indicating
plastic that the
samples canblack plastic samples
be separated can be separated
and characterized and characterized
by the LITRAN system orby the LITRAN
photon system
up conversion.
or photon up conversion.
The principle components PC1 and PC2 of the PCA for the black plastic samples are shown in
The principle
Figure 7a,b respectively.components PC1 and
The principle PC2 of theindicate
components PCA forthe themain
blackspectral
plastic information
samples are contained
shown in
Figure 7a,b respectively. The principle components indicate the main spectral information contained
in each spectrum. The first PC contains most of the spectral variances and the second PC indicates
the second highest percentage of variance and so on. The scores of the samples in the score plot again
indicate how much of the spectral information is contained in the measured spectra. This makes it
possible to distinguish identical or similar plastic samples or highly differentiated samples from one
another, and thus to classify them into plastic varieties. Looking at the measurements of the black
Polymers 2017, 9, 435 8 of 9

in each spectrum. The first PC contains most of the spectral variances and the second PC indicates
the second highest percentage of variance and so on. The scores of the samples in the score plot again
indicate how much of the spectral information is contained in the measured spectra. This makes
it possible
Polymers 2017,to distinguish identical or similar plastic samples or highly differentiated samples from
9, 435 8 of 9
one another, and thus to classify them into plastic varieties. Looking at the measurements of the
plastic samples
black plastic in the score
samples in theplot in Figure
score plot in6,Figure
one can6,seeonethat
canthe
seesamples
that thecontaining
samples ABS, PC, or ABS,
containing PBT
are arranged along the PC1. From this it can be concluded that the PC1 represents
PC, or PBT are arranged along the PC1. From this it can be concluded that the PC1 represents the the characteristics
of these plastic of
characteristics grades.
these It is noteworthy
plastic grades. Itthat the ABS sample
is noteworthy withABS
that the the sample
mark namewithCycolac
the mark and the
name
ABS sample Lustran are far apart. The reason for this is that these plastics
Cycolac and the ABS sample Lustran are far apart. The reason for this is that these plastics come come from different
manufacturers and therefore partly
from different manufacturers differentpartly
and therefore additives haveadditives
different been addedhaveorbeen
the added
blend composition
or the blend
varies. PE and PP samples have therefore a sore value at PC1 near zero.
composition varies. PE and PP samples have therefore a sore value at PC1 near zero. Looking Looking at the position of
at the
the scores of the samples with respect to PC2, it can be seen that the samples PE and
position of the scores of the samples with respect to PC2, it can be seen that the samples PE and PP PP have positive
values and the
have positive scores
values of the
and other samples
scores aresamples
of other negative. areThis meansThis
negative. thatmeans
with the
thathelp
withofthethe PC2
help of the
the
thermoplastic samples can be distinguished from the non-thermoplastic
PC2 the thermoplastic samples can be distinguished from the non-thermoplastic ones. ones.

3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 4.0 4.1
3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 4.0 4.1
wavelength / µm wavelength / µm

(a) (b)

Figure
Figure 7.
7. The
The PCA
PCA evaluation
evaluation of
of black
black plastic
plastic samples
samples (a)
(a) PC1;
PC1; (b)
(b) PC2.
PC2.

4. Conclusions
4. Conclusions
As shown, MIR spectroscopy is an efficient method for spectrally characterizing black plastic
As shown, MIR spectroscopy is an efficient method for spectrally characterizing black plastic
parts and assigning them to the corresponding polymer types. Based on measurements of the
parts and assigning them to the corresponding polymer types. Based on measurements of the
reflectivity in the spectral range from 3.2 to about 3.6 μm with the LITRAN system, in addition to
reflectivity in the spectral range from 3.2 to about 3.6 µm with the LITRAN system, in addition
pure plastic grades, blends can also be detected and even assigned to different manufacturers. In the
to pure plastic grades, blends can also be detected and even assigned to different manufacturers.
case of re-granulates, for example, the proportions of the base material can be determined in principle
In the case of re-granulates, for example, the proportions of the base material can be determined in
according to corresponding calibration models. These are fundamental prerequisites for assuring
principle according to corresponding calibration models. These are fundamental prerequisites for
accurate sorting and for achieving a high grade of recycled black plastic parts. The LITRAN system
assuring accurate sorting and for achieving a high grade of recycled black plastic parts. The LITRAN
provides the prerequisite for the economical sorting of black plastics, since measuring rates of several
system provides the prerequisite for the economical sorting of black plastics, since measuring rates
thousand measurements per second can be achieved. As a result, sorting systems with a sufficient
of several thousand measurements per second can be achieved. As a result, sorting systems with
mass throughput can be implemented. In the case of engineering plastics, the determination of
a sufficient mass throughput can be implemented. In the case of engineering plastics, the determination
possible polymer additives and polymer mixtures is also important. Many engineering plastics
of possible polymer additives and polymer mixtures is also important. Many engineering plastics
contain, for example, flame retardants, some of which are prohibited and have to be sorted out before
contain, for example, flame retardants, some of which are prohibited and have to be sorted out before
the re-granulation. The spectral range of the LITRAN system can be adapted to detect these
the re-granulation. The spectral range of the LITRAN system can be adapted to detect these substances.
substances.
The LITRAN system also offers the possibility to expand existing sorting systems with NIR sorting
The LITRAN system also offers the possibility to expand existing sorting systems with NIR
into the MIR range.
sorting into the MIR range.
Because the absorption bands or signals are 10 to 100 times stronger in the MIR range than in
Because the absorption bands or signals are 10 to 100 times stronger in the MIR range than in
the NIR range, hence the sorting of normal non-black household plastics can also be carried out with
the NIR range, hence the sorting of normal non-black household plastics can also be carried out with
the LITRAN system. However, as a prerequisite, the measuring rate of the LITRAN system would have
the LITRAN system. However, as a prerequisite, the measuring rate of the LITRAN system would
to be able to perform several 10,000 to 100,000 measurements per second. Optical components and the
have to be able to perform several 10,000 to 100,000 measurements per second. Optical components
and the measurement setup of the LITRAN system are in principle capable of extending the
application range of photon up conversion to other industrial applications.

Author Contributions: Kerstin Sachsenheimer and Melanie Klemenz performed and optimized the experiments
and made the measurements. Wolfgang Becker evaluated the measurements and wrote the paper.
Polymers 2017, 9, 435 9 of 9

measurement setup of the LITRAN system are in principle capable of extending the application range
of photon up conversion to other industrial applications.

Author Contributions: Kerstin Sachsenheimer and Melanie Klemenz performed and optimized the experiments
and made the measurements. Wolfgang Becker evaluated the measurements and wrote the paper.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Becker, W.; Eisenreich, N. Identification of black polymers in the mid-infrared spectral region with
an FTIR-spectrometer. CARE Innov. 1998, 98, 88–95.
2. Graf, R.T.; Koenig, J.L.; Ishida, H. Comparison of FT-IR Transmission, specular reflectance and attenuated
reflectance spectra of polymers. In Fourier Transform Infrared Characterization of Polymers; Ishida, H., Ed.;
Springer: New York, NJ, USA, 1987; pp. 385–395.
3. Hart, W.W.; Painter, P.C.; Koenig, J.L.; Coleman, M.M. A fourier transform infrared method of studying
elastomers filled with carbon black. Appl. Spectrosc. 1977, 31, 220–224. [CrossRef]
4. Devlin, E.F. The fourier transform infrared spectrum of cured, black-reinforced SBR. J. Polym. Sci. C 1981.
[CrossRef]
5. Chalmers, J.M.; Mackenzie, M.W.; Sharp, J.L. Charaterization of thin layer chromatographically separated
fractions by fourier transform infrared diffuse reflectance spectrometry. Anal. Chem. 1987, 59, 415–418.
[CrossRef]
6. Chalmers, J.M.; Mackenzie, M.W.; Poole, N. Some observations on FTIR emission spectroscopy of black solid
samples. Mikrochim. Acta 1988, 95, 249–253. [CrossRef]
7. Claybourn, M.; Colombel, P.; Chalmers, J. Characterization of carbon-filled polymers by specular reflection.
Appl. Spectrosc. 1991, 45, 279–286. [CrossRef]
8. Becker, W. Using Spectroscopic Methods for Identification of Technical Plastics from E & E and Elv Sector.
In Proceedings of the Going Green—Care Innovation Conference, Vienna, Austria, 13–16 November 2006.
9. Becker, W.; Eisenreich, N. Identification of black polymers in the mid-infrared spectral region with a FTIR
Spectrometer. In Proceedings of the CARE Innovation Eco-Efficient Concepts for the Electronic Industry
towards Sustainability, Vienna, Austria, 16–19 November 1998.
10. Becker, W. Prozessanalytik in der Kunststoffindustrie. In Prozessanalytik: Strategien und Fallbeispiele aus der
Industriellen Praxis; Kessler, R.W., Ed.; Wiley-VCH: Weinheim, Germany, 2006.
11. Dam, J.S.; Tidemand-Lichtenberg, P.; Pedersen, C. Room-temperature mid-infrared single-photon spectral
imaging. Nat. Photonics 2012, 6, 788–793. [CrossRef]
12. Hu, Q.; Seidelin Dam, J.; Pedersen, C.; Tidemand-Lichtenberg, P. High-resolution mid-IR spectrometer based
on frequency upconversion. Opt. Lett. 2012, 37, 5232–5234 [CrossRef] [PubMed]
13. Gurski, T.R.; Epps, H.W.; Maran, S.P. Upconversion of broadband infrared spetra. Appl. Opt. 1978, 17,
1238–1242. [CrossRef] [PubMed]
14. Workman, J.; Weyer, L. Practical Guide and Spectral Atlas for Interpretive Near-Infrared Spectroscopy, 2nd ed.;
CRC Press: Boca Raton, FL, USA, 2012.
15. Midwinter, J.E. Image conversion from 1.6 µ to the visible in lithium niobate. Appl. Phys. Lett. 1968, 12, 68–70.
[CrossRef]
16. Lambrecht, A.; Bolwien, C.; Herbst, J.; Kühnemann, F.; Sandfort, V.; Wolf, S. Neue Methoden der
laserbasierten Gasanalytik. Chem. Ing. Tech. 2016, 88, 746–755. [CrossRef]
17. Næs, T. A User-Friendly Guide to Multivariate Calibration and Classification; NIR Publications: Chichester,
UK, 2002.
18. Kessler, W. Multivariate Datenanalyse für die Pharma-, Bio- und Prozessanalytik; Wiley VCH GmbH&Co. KgaA:
Weinheim, Germany, 2007.

© 2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).