- Email: [email protected]
Near-infrared (NIR) monitoring of Nylon 6 during quenching studied by projection two-dimensional (2D) correlation spectroscopy
Journal of Molecular Structure xxx (2016) 1e4
Contents lists available at ScienceDirect
Journal of Molecular Structure journal homepage: http://www.elsevier.com/locate/molstruc
Near-infrared (NIR) monitoring of Nylon 6 during quenching studied by projection two-dimensional (2D) correlation spectroscopy Hideyuki Shinzawa*, Junji Mizukado Research Institute for Sustainable Chemistry, Advanced Industrial Science and Technology (AIST), Tsukuba, Japan
a r t i c l e i n f o
a b s t r a c t
Article history: Received 11 December 2015 Received in revised form 14 January 2016 Accepted 4 February 2016 Available online xxx
Evolutionary change in supermolecular structure of Nylon 6 during its melt-quenched process was studied by Near-infrared (NIR) spectroscopy. Time-resolved NIR spectra was measured by taking the advantage of high-speed NIR monitoring based on an acousto-optic tunable filter (AOTF). Fine spectral features associated with the variation of crystalline and amorphous structure occurring in relatively short time scale were readily captured. For example, synchronous and asynchronous 2D correlation spectra reveal the initial decrease in the contribution of the NIR band at 1485 nm due to the amorphous structure, predominantly existing in the melt Nylon 6. This is then followed by the emerging contribution of the band intensity at 1535 nm associated with the crystalline structure. Consequently, the results clearly demonstrate a definite advantage of the high-speed NIR monitoring for analyzing fleeting phenomena. © 2016 Elsevier B.V. All rights reserved.
Keywords: Near-infrared Two-dimensional correlation spectroscopy Nylon 6 Projection treatment
1. Introduction This article provides an illustrative application example of highspeed near-infrared (NIR) spectroscopy for studying evolutionary change in supermolecular structure of Nylon 6 during its meltquenched process. Spectroscopic monitoring of chemical reactions and product streams is an analytical tool of practical interest [1e4]. Near-infrared (NIR) spectroscopy is particularly suited for monitoring chemical reaction systems [1,5e8]. For example, each chemical functional group in a molecule vibrates at a unique frequency, providing a specific peak in spectrum. The transient variation of components in systems results in the change in the spectral feature. By analyzing the variation of the spectral feature, it is possible to sort out the chemically meaningful information on the system altered. In addition, the utilization of the NIR region with much less absorption compared to mid infrared region enables the light to penetrate much farther into a sample. In fact, the penetration depth of NIR beam can be on the scale of centimeters and this particularly becomes useful in in probing bulk material. Recent development of high-speed NIR spectrometer based on acousto-optic tunable filter (AOTF) opened up further perspectives for process analysis [1,9e11]. An AOTF consists of a birefringent crystal made of a tellurium oxide (TeO2) and a high-frequency
* Corresponding author. E-mail address: [email protected] (H. Shinzawa).
piezoelectric transducer. By applying a specific radio frequency to TeO2 crystal, it produces acoustic vibrations that propagates through the crystal. As light goes through the crystal, interaction between the light waves and sound waves causes the crystal to act as a narrow-line band pass filter to separate a single wavelength of light from a broadband source. In other words, AOTF works as electronically tunable bandpass filter with no moving parts. The utilization of the AOTF provides a distinct advantage over the conventional grating-monochromator or interferometer based approach in terms of high speed data acquisition, especially when the variation of the system occurs in a relatively short time scale [10,11]. This is actually true to the development of supermolecular structures during polymer processing. Thus, the intrinsic properties of NIR light and AOTF brought together provide interesting opportunity to prove the even more in-depth understanding of polymer system of interest. In this article, high-speed NIR monitoring of a seemingly simple Nylon 6 polymer undergoing melt-quenching is presented as an illustrative example of an application of this characterization technique in polymer analysis. Semicrystalline Nylon 6 prepared from the melt often show a complex polymer structure consisting of folded-chain crystal lamellae embedded in a liquid-like amorphous matrix [12]. The cooling rate substantially affects the crystalline growth, which eventually influences various properties of the polymer [13]. The high-speed NIR monitoring of the evolutionary change in polymer structure of Nylon 6 during its melt-
http://dx.doi.org/10.1016/j.molstruc.2016.02.016 0022-2860/© 2016 Elsevier B.V. All rights reserved.
Please cite this article in press as: H. Shinzawa, J. Mizukado, Near-infrared (NIR) monitoring of Nylon 6 during quenching studied by projection two-dimensional (2D) correlation spectroscopy, Journal of Molecular Structure (2016), http://dx.doi.org/10.1016/j.molstruc.2016.02.016
2
H. Shinzawa, J. Mizukado / Journal of Molecular Structure xxx (2016) 1e4
quenched process promises a great deal of practical potential in process monitoring applications. The time-resolved NIR spectra were collected at an interval of 1 s during the early stage of the quenching. While the entire feature of the NIR spectra was overwhelmed by baseline change arising from the transition from liquid to solid states, detailed changes in the spectral feature was readily elucidated with an aid of the projection two-dimensional (2D) correlation analysis [10,11,14e19]. The 2D correlation spectra of Nylon 6 clearly revealed fine details of spectral intensity variations closely related to the consumption of the amorphous structure and subsequent development of the crystalline which occur in relatively short time scale.
quenching of Nylon 6 is depicted in Fig. 1. A temperaturecontrolled cell holder was equipped with an AOTF-NIR spectrometer (Systems Engineering Inc., Tokyo). The Nylon 6 provided by SigmaeAldrich was placed in the cuvette cell with 10 mm pathlength. The sample was melted at 250 C and then quenched by setting the temperature of the cell holder at 30 C. During the quenching, a series of NIR transmittance spectra of the sample were measured for every 1 s by co-adding 16 scans. An input optical beam from broadband source is separated into wavelengths ranging 1100e1800 nm by an AOTF between crossed polarizes. The measurement of the time-resolved NIR spectra was carried out for 180 s. A background spectrum was measured with the empty cuvette cell prior to the collection.
2. Theory 4. Result and discussions 2.1. Projection treatment 4.1. Time-resolved NIR spectra of Nylon 6 Projection operation is an especially useful pretreatment method to analyze NIR spectra showing exceptional baseline change [14,15]. It aims to selectively remove the specific signal contribution from spectra by sorting out dynamic spectra into two separate sets: one which is fully aligned with a chosen projecting vector and the other which is orthogonal to the same vector. For given vector y, the projection matrix Ry is defined as
1 R y ¼ y yT y y
(1)
The m-by-m matrix Ry acts as a projector for the space spanned by y. The projected data matrix AP is obtained as
Ap ¼ R y A
(2)
The projected data Ap represents the projection of A onto the abstract space spanned by y. The portion of dynamic spectra projected onto the space spanned by such a projecting vector y will have the same trend of y. Thus, all signals contained in the projected data Ap are fully synchronized. The corresponding null-space projection is carried out as
AN ¼ I R y A ¼ A Ap
(3)
where I means m-by-m identity matrix. The null-space projected data matrix AN represents the projection of A onto the space spanned by the vectors orthogonal to y. AN is the residual after the removal of AP from A by using the information contained within y. In other words, the null-space projection selectively eliminates the portion of dynamic spectra which is synchronized with the projecting vector. There are several options to choose the source of the projector. In most cases, a single column is selected as a projector vector y from the data matrix [14,15,19]. By using a spectral intensity change at specific wavenumber where a peak is observed as the vector y, it AN becomes free from the signal contribution from the trend associated with y. Such an ability is suitable for the elimination of specific trend from spectra. In fact, baseline correction by the projection becomes especially important in the analysis of the spectra measured by the several NIR techniques where the baseline fluctuation of spectra can be caused necessarily by morphological changes of samples [10,11].
The time-resolved NIR spectra of Nylon 6 undergoing the quenching are represented in Fig. 2. Interestingly, entire feature changes from one spectrum to another, suggesting that the transient variation induced in such short time scale can be readily captured. NIR study of semicrystalline polymer samples is often complicated with the presence of overlapped contributions from coexisting crystalline and amorphous components. Fortunately, the NIR spectrum of amorphous component of Nylon 6 is approximated by the melt spectrum since the contribution of amorphous component is well represented by the initial spectrum of Nylon 6 undergoing the quenching. For example, variation of the band around 1485 nm suggests that the crystallization process is accompanied by a precipitous decrease in the intensity of the NIR band. This band is thus mostly assignable to the first overtone of the NH2 antisymmetric vibration arising from the amorphous component [12]. Another notable feature is that the decrease in the amorphous band is compensated by the rising contribution of the band around at 1530 nm due to the first overtone of the NH2 symmetric vibration associated with the crystalline component [12]. The quenching of the melt Nylon 6 results in the substantial variation in the spectral intensity. The difference in the spectral intensity change between the crystalline and amorphous bands will provide useful information to sort out the variation of the population. However, it is obvious that the change in the spectral intensity caused by the crystallization is overwhelmed by exceptionally predominant baseline changes to make the identification of the pertinent spectral intensity variations difficult. A major cause of the baseline fluctuation observed in the time-resolved NIR spectra is the change in the light scattering due to the transition from liquid to solid states. For example, the solid Nylon 6 scatters even more NIR light and the less light is then detected by the sensor. This, in turn, induces apparent increase in the light absorption by the sample to provide upward shift of the spectral baseline. Unfortunately, these changes are essentially inevitable to the NIR
3. Experimental 3.1. Time-resolved NIR spectra A schematic description of NIR monitoring of the melt-
Fig. 1. A schematic illustration of NIR monitoring based on AOFT.
Please cite this article in press as: H. Shinzawa, J. Mizukado, Near-infrared (NIR) monitoring of Nylon 6 during quenching studied by projection two-dimensional (2D) correlation spectroscopy, Journal of Molecular Structure (2016), http://dx.doi.org/10.1016/j.molstruc.2016.02.016
H. Shinzawa, J. Mizukado / Journal of Molecular Structure xxx (2016) 1e4
Fig. 2. Time-resolved NIR spectra of Nylon 6 during quenching.
transmittance spectra based on the penetration of the light. The baseline correction of the NIR spectra, thus, becomes especially important task in elucidating the truly meaningful intensity variation of the NIR bands.
4.2. Projection 2D correlation analysis The baseline correction of the NIR spectra involves two steps. The first step is simple subtraction, i.e. offset-correction, to remove additive scattering factor from the original NIR spectra [10,11]. It is then followed by projection treatment to remove multiplicative scatter factor from the offset-corrected spectra [10,11]. For example, Fig. 3 illustrates offset-corrected spectra derived from the raw NIR spectra of the quenched Nylon 6 sample shown in Fig. 2. Note that the baseline offset is carried out by subtracting the spectral intensity at 1300 nm to make the intensities zero. This simple subtraction aims to remove the additive scatter factor providing upward shift to the spectrum. In Fig. 3, it is noted that the spectra still show fluctuations of the spectral intensity even in the region where no chemically meaningful peaks exist. Fig. 4 represents spectral intensity variations at 1485 (amorphous component), 1530 (crystalline component) and 1600 nm where no obvious peaks are observed. Note that the intensity variation at 1600 nm mainly reflects the baseline increase arising from the multiplicative scatter factor. The intensities of the bands associated with the amorphous component show an initial increase at the very onset of the quenching followed by the subsequent decrease. Importantly, it
3
Fig. 4. Spectral intensity variations at 1485 (amorphous component), 1530 (crystalline component) and 1600 nm derived from offset-corrected spectra.
should be noted that the change in the intensity of the crystalline peak results in somewhat similar feature, while the quantity of the crystalline component is supposed to essentially increase during the melt-quenching process. Such trend is also true for the variation at 1600 nm where no meaningful peaks exist. It is thus likely that the intensity variations in the whole spectral region are still overwhelmed with the contribution from the baseline change associated with multiplicative scatter factor. It should be pointed out here that the intensity variation at 1600 nm in the offset spectra purely reflects the signal contribution from the multiplicative scatter factor. In other words, the attenuation of the portion, which is synchronized with this trend, leads to the highly selective removal of the contribution from the multiplicative scatter factor in the whole spectral region. Fig. 5 represents the null-space projected spectra derived from the offset-corrected spectra by using the spectral intensity change at 1600 nm as the vector y. One can find that the spectra are now free from the additive and multiplicative scatter effects after the application of the baseline correction treatments. While the spectral intensities can be both positive and negative after the projection treatment, certain information such as the relative direction of signal changes and the sequential order of intensity variations, is thus fully preserved. The intensity of the amorphous band at 1485 nm shows obvious decrease, mostly indicating the reduction of the amorphous content. It is also noted that the crystalline band
Fig. 5. Projection-corrected spectra obtained by projecting offset-collected spectra onto null-space of their signals at 1600 nm. Fig. 3. Offset-corrected spectra derived from original NIR spectra.
Please cite this article in press as: H. Shinzawa, J. Mizukado, Near-infrared (NIR) monitoring of Nylon 6 during quenching studied by projection two-dimensional (2D) correlation spectroscopy, Journal of Molecular Structure (2016), http://dx.doi.org/10.1016/j.molstruc.2016.02.016
4
H. Shinzawa, J. Mizukado / Journal of Molecular Structure xxx (2016) 1e4
Fig. 6. (A) Synchronous and (B) asynchronous correlation spectra calculated from the projection-corrected spectra shown in Fig. 5.
at 1535 nm is still overwhelmed with the neighboring amorphous band and the intensity seemingly decreases. The gradual increase in the intensity observed over the 1350e1450 nm region may be associated with the contributions from the developments of the crystalline structure. Thus, the application of 2D correlation analysis to the projection corrected spectra becomes important to sort out the pertinent changes of the rather broad and highly overlapped bands corresponding to the contributions from crystalline and amorphous components. (A) Synchronous and (B) asynchronous correlation spectra derived from the projection-treated NIR spectra of Nylon 6 are represented in Fig. 6, respectively. The plot of the reference spectrum is placed at the top and side of the contour map. Negative correlation intensity areas of the contour map are represented by the shading. The synchronous spectrum develops one broad autopeak near 1485 nm, reflecting the gradual increase in the spectral intensity in this NIR region. This is mostly because of the obvious contribution from the amorphous band. A broad negative correlation peak around at (1485, 1400) suggests that spectral intensities at these wavelengths change in the direction opposite each other during the quenching process, probably indicating the presence of the crystalline bands over the 1350e1450 nm region. The corresponding asynchronous spectrum provides an even clearer picture of the time-dependent changes in the NIR intensities of component bands. For example, it develops a positive cross peak at the coordinate (1485, 1535). The appearance of the asynchronous correlation peak suggests that the decrease in the spectral intensity at 1485 nm predominantly occurs before the decrease that at 1535 nm. This can be interpreted to mean that the development of the crystalline component provides additional contribution delaying the intensity variation at 1535 nm. Another notable feature can also be observed along the 1485 nm coordinate of the asynchronous correlation spectrum. The correlation peaks observed over the 1350e1450 nm region along the 1485 nm coordinate reveal the presence of the component developing after the predominant consumption of the amorphous. In other words, the development of the correlation peaks also suggests the multiple crystalline bands appeared in this NIR region. It is thus likely that the, in the earlier stage of the quenching, the disappearance of amorphous component occurs first. The consumption of the amorphous structure is then followed by the development of the crystalline structure. Such observation is actually consistent with the expected population changes of crystalline and amorphous components during the quenching of the molten semicrystalline polymer. Consequently, determination of the sequential order as well as clear differentiation between bands located very close to each other, such as crystalline and amorphous components, is demonstrated.
5. Conclusion 2D projection correlation analysis, coupled with high-speed NIR measurements yields an interesting result on the structure variation of Nylon 6 undergoing the melt-quenching. Although the transient variation of the Nylon 6 sample was readily captured as timeresolved NIR spectra, the main feature of the NIR spectra was overwhelmed by the contribution from the baseline change. Projection treatment was applied to selectively separate the signal contribution of interest from the baseline fluctuation. The 2D correlation spectra derived from the projection-corrected spectra suggest the presence of polymer structures undergoing different variations during the quenching. For example, the synchronous and asynchronous 2D correlation spectra reveal the initial decrease in the contribution of the NIR band at 1485 nm due to the amorphous structure, predominantly existing in the melt Nylon 6. This is then followed by the emerging contribution of the band intensity at 1535 nm associated with the crystalline structure. These results reveal the transient variation of the polymer structures occurring in relatively short time scale, demonstrating a definite advantage of the high-speed NIR monitoring for analyzing fleeting phenomena we often encounter in the real world.
References [1] K. Bakev, Process Analytical Technology. Spectroscopic Tools and Implementation Strategies for the Chemical and Pharmaceutical Industries, Blackwell Publishing Ltd., Oxford, 2005. [2] L. Helmdach, M.P. Feth, C. Minnich, J. Ulrich, Chem. Eng. Process. Process Intensif. 70 (2013) 184. [3] S.S. Kadam, A. Mesbah, E. van der Windt, H.J.M. Kramer, Chem. Eng. Res. Des. 89 (2011) 995. [4] I. Noda, W.M. Allen, S.E. Lindberg, Appl. Spectrosc. 63 (2009) 224. [5] M. Blanco, I. Villarroya, Trends Anal. Chem. 21 (2002) 240. [6] J.G. Rosas, H. de Waard, T. De Beer, C. Vervaet, J.P. Remon, W.L.J. Hinrichs, H.W. Frijlink, M. Blanco, J. Pharm. Biomed. Anal. 97 (2014) 39. [7] G. Mirschel, O. Daikos, K. Heymann, T. Scherzer, B. Genest, C. Sommerer, C. Steckert, Progr. Org. Coat. 77 (2014) 719. [8] J.P. Reed, D. Devlin, S.R.R. Esteves, R. Dinsdale, A.J. Guwy, Biores. Tech. 133 (2013) 398. [9] B.L. Heffner, D.A. Smith, J.E. Baran, K.W. Cheung, Electron. Lett. 24 (1988) 1562. [10] H. Shinzawa, W. Kanematsu, I. Noda, Vib. Spectrosc. 70 (2014) 53. [11] H. Shinzawa, W. Kanematsu, I. Noda, Vib. Spectrosc. 78 (2015) 34. [12] Y. Xu, P. Wu, J. Mol. Struct. 833 (2007) 145. [13] L.W. McKeen, The Effect of Creep and Other Time Related Factors on Plastics and Elastomers (Plastics Design Library), William Andrew, Norwich, 1992. [14] I. Noda, J. Mol. Struct. 974 (2010) 116. [15] I. Noda, Vib. Spectrosc. 60 (2012) 146. [16] I. Noda, Appl. Spectrosc. 47 (1993) 1329. [17] I. Noda, Appl. Spectrosc. 54 (2000) 994. [18] I. Noda, Y. Ozaki, Two-dimensional Correlation Spectroscopy, Willey & Sons, Chichester, West Sussex, 2004. [19] M.K. Kim, S.R. Ryu, I. Noda, Y.M. Jung, Vib. Spectrosc. 60 (2012) 163.
Please cite this article in press as: H. Shinzawa, J. Mizukado, Near-infrared (NIR) monitoring of Nylon 6 during quenching studied by projection two-dimensional (2D) correlation spectroscopy, Journal of Molecular Structure (2016), http://dx.doi.org/10.1016/j.molstruc.2016.02.016
Contents lists available at ScienceDirect
Journal of Molecular Structure journal homepage: http://www.elsevier.com/locate/molstruc
Near-infrared (NIR) monitoring of Nylon 6 during quenching studied by projection two-dimensional (2D) correlation spectroscopy Hideyuki Shinzawa*, Junji Mizukado Research Institute for Sustainable Chemistry, Advanced Industrial Science and Technology (AIST), Tsukuba, Japan
a r t i c l e i n f o
a b s t r a c t
Article history: Received 11 December 2015 Received in revised form 14 January 2016 Accepted 4 February 2016 Available online xxx
Evolutionary change in supermolecular structure of Nylon 6 during its melt-quenched process was studied by Near-infrared (NIR) spectroscopy. Time-resolved NIR spectra was measured by taking the advantage of high-speed NIR monitoring based on an acousto-optic tunable filter (AOTF). Fine spectral features associated with the variation of crystalline and amorphous structure occurring in relatively short time scale were readily captured. For example, synchronous and asynchronous 2D correlation spectra reveal the initial decrease in the contribution of the NIR band at 1485 nm due to the amorphous structure, predominantly existing in the melt Nylon 6. This is then followed by the emerging contribution of the band intensity at 1535 nm associated with the crystalline structure. Consequently, the results clearly demonstrate a definite advantage of the high-speed NIR monitoring for analyzing fleeting phenomena. © 2016 Elsevier B.V. All rights reserved.
Keywords: Near-infrared Two-dimensional correlation spectroscopy Nylon 6 Projection treatment
1. Introduction This article provides an illustrative application example of highspeed near-infrared (NIR) spectroscopy for studying evolutionary change in supermolecular structure of Nylon 6 during its meltquenched process. Spectroscopic monitoring of chemical reactions and product streams is an analytical tool of practical interest [1e4]. Near-infrared (NIR) spectroscopy is particularly suited for monitoring chemical reaction systems [1,5e8]. For example, each chemical functional group in a molecule vibrates at a unique frequency, providing a specific peak in spectrum. The transient variation of components in systems results in the change in the spectral feature. By analyzing the variation of the spectral feature, it is possible to sort out the chemically meaningful information on the system altered. In addition, the utilization of the NIR region with much less absorption compared to mid infrared region enables the light to penetrate much farther into a sample. In fact, the penetration depth of NIR beam can be on the scale of centimeters and this particularly becomes useful in in probing bulk material. Recent development of high-speed NIR spectrometer based on acousto-optic tunable filter (AOTF) opened up further perspectives for process analysis [1,9e11]. An AOTF consists of a birefringent crystal made of a tellurium oxide (TeO2) and a high-frequency
* Corresponding author. E-mail address: [email protected] (H. Shinzawa).
piezoelectric transducer. By applying a specific radio frequency to TeO2 crystal, it produces acoustic vibrations that propagates through the crystal. As light goes through the crystal, interaction between the light waves and sound waves causes the crystal to act as a narrow-line band pass filter to separate a single wavelength of light from a broadband source. In other words, AOTF works as electronically tunable bandpass filter with no moving parts. The utilization of the AOTF provides a distinct advantage over the conventional grating-monochromator or interferometer based approach in terms of high speed data acquisition, especially when the variation of the system occurs in a relatively short time scale [10,11]. This is actually true to the development of supermolecular structures during polymer processing. Thus, the intrinsic properties of NIR light and AOTF brought together provide interesting opportunity to prove the even more in-depth understanding of polymer system of interest. In this article, high-speed NIR monitoring of a seemingly simple Nylon 6 polymer undergoing melt-quenching is presented as an illustrative example of an application of this characterization technique in polymer analysis. Semicrystalline Nylon 6 prepared from the melt often show a complex polymer structure consisting of folded-chain crystal lamellae embedded in a liquid-like amorphous matrix [12]. The cooling rate substantially affects the crystalline growth, which eventually influences various properties of the polymer [13]. The high-speed NIR monitoring of the evolutionary change in polymer structure of Nylon 6 during its melt-
http://dx.doi.org/10.1016/j.molstruc.2016.02.016 0022-2860/© 2016 Elsevier B.V. All rights reserved.
Please cite this article in press as: H. Shinzawa, J. Mizukado, Near-infrared (NIR) monitoring of Nylon 6 during quenching studied by projection two-dimensional (2D) correlation spectroscopy, Journal of Molecular Structure (2016), http://dx.doi.org/10.1016/j.molstruc.2016.02.016
2
H. Shinzawa, J. Mizukado / Journal of Molecular Structure xxx (2016) 1e4
quenched process promises a great deal of practical potential in process monitoring applications. The time-resolved NIR spectra were collected at an interval of 1 s during the early stage of the quenching. While the entire feature of the NIR spectra was overwhelmed by baseline change arising from the transition from liquid to solid states, detailed changes in the spectral feature was readily elucidated with an aid of the projection two-dimensional (2D) correlation analysis [10,11,14e19]. The 2D correlation spectra of Nylon 6 clearly revealed fine details of spectral intensity variations closely related to the consumption of the amorphous structure and subsequent development of the crystalline which occur in relatively short time scale.
quenching of Nylon 6 is depicted in Fig. 1. A temperaturecontrolled cell holder was equipped with an AOTF-NIR spectrometer (Systems Engineering Inc., Tokyo). The Nylon 6 provided by SigmaeAldrich was placed in the cuvette cell with 10 mm pathlength. The sample was melted at 250 C and then quenched by setting the temperature of the cell holder at 30 C. During the quenching, a series of NIR transmittance spectra of the sample were measured for every 1 s by co-adding 16 scans. An input optical beam from broadband source is separated into wavelengths ranging 1100e1800 nm by an AOTF between crossed polarizes. The measurement of the time-resolved NIR spectra was carried out for 180 s. A background spectrum was measured with the empty cuvette cell prior to the collection.
2. Theory 4. Result and discussions 2.1. Projection treatment 4.1. Time-resolved NIR spectra of Nylon 6 Projection operation is an especially useful pretreatment method to analyze NIR spectra showing exceptional baseline change [14,15]. It aims to selectively remove the specific signal contribution from spectra by sorting out dynamic spectra into two separate sets: one which is fully aligned with a chosen projecting vector and the other which is orthogonal to the same vector. For given vector y, the projection matrix Ry is defined as
1 R y ¼ y yT y y
(1)
The m-by-m matrix Ry acts as a projector for the space spanned by y. The projected data matrix AP is obtained as
Ap ¼ R y A
(2)
The projected data Ap represents the projection of A onto the abstract space spanned by y. The portion of dynamic spectra projected onto the space spanned by such a projecting vector y will have the same trend of y. Thus, all signals contained in the projected data Ap are fully synchronized. The corresponding null-space projection is carried out as
AN ¼ I R y A ¼ A Ap
(3)
where I means m-by-m identity matrix. The null-space projected data matrix AN represents the projection of A onto the space spanned by the vectors orthogonal to y. AN is the residual after the removal of AP from A by using the information contained within y. In other words, the null-space projection selectively eliminates the portion of dynamic spectra which is synchronized with the projecting vector. There are several options to choose the source of the projector. In most cases, a single column is selected as a projector vector y from the data matrix [14,15,19]. By using a spectral intensity change at specific wavenumber where a peak is observed as the vector y, it AN becomes free from the signal contribution from the trend associated with y. Such an ability is suitable for the elimination of specific trend from spectra. In fact, baseline correction by the projection becomes especially important in the analysis of the spectra measured by the several NIR techniques where the baseline fluctuation of spectra can be caused necessarily by morphological changes of samples [10,11].
The time-resolved NIR spectra of Nylon 6 undergoing the quenching are represented in Fig. 2. Interestingly, entire feature changes from one spectrum to another, suggesting that the transient variation induced in such short time scale can be readily captured. NIR study of semicrystalline polymer samples is often complicated with the presence of overlapped contributions from coexisting crystalline and amorphous components. Fortunately, the NIR spectrum of amorphous component of Nylon 6 is approximated by the melt spectrum since the contribution of amorphous component is well represented by the initial spectrum of Nylon 6 undergoing the quenching. For example, variation of the band around 1485 nm suggests that the crystallization process is accompanied by a precipitous decrease in the intensity of the NIR band. This band is thus mostly assignable to the first overtone of the NH2 antisymmetric vibration arising from the amorphous component [12]. Another notable feature is that the decrease in the amorphous band is compensated by the rising contribution of the band around at 1530 nm due to the first overtone of the NH2 symmetric vibration associated with the crystalline component [12]. The quenching of the melt Nylon 6 results in the substantial variation in the spectral intensity. The difference in the spectral intensity change between the crystalline and amorphous bands will provide useful information to sort out the variation of the population. However, it is obvious that the change in the spectral intensity caused by the crystallization is overwhelmed by exceptionally predominant baseline changes to make the identification of the pertinent spectral intensity variations difficult. A major cause of the baseline fluctuation observed in the time-resolved NIR spectra is the change in the light scattering due to the transition from liquid to solid states. For example, the solid Nylon 6 scatters even more NIR light and the less light is then detected by the sensor. This, in turn, induces apparent increase in the light absorption by the sample to provide upward shift of the spectral baseline. Unfortunately, these changes are essentially inevitable to the NIR
3. Experimental 3.1. Time-resolved NIR spectra A schematic description of NIR monitoring of the melt-
Fig. 1. A schematic illustration of NIR monitoring based on AOFT.
Please cite this article in press as: H. Shinzawa, J. Mizukado, Near-infrared (NIR) monitoring of Nylon 6 during quenching studied by projection two-dimensional (2D) correlation spectroscopy, Journal of Molecular Structure (2016), http://dx.doi.org/10.1016/j.molstruc.2016.02.016
H. Shinzawa, J. Mizukado / Journal of Molecular Structure xxx (2016) 1e4
Fig. 2. Time-resolved NIR spectra of Nylon 6 during quenching.
transmittance spectra based on the penetration of the light. The baseline correction of the NIR spectra, thus, becomes especially important task in elucidating the truly meaningful intensity variation of the NIR bands.
4.2. Projection 2D correlation analysis The baseline correction of the NIR spectra involves two steps. The first step is simple subtraction, i.e. offset-correction, to remove additive scattering factor from the original NIR spectra [10,11]. It is then followed by projection treatment to remove multiplicative scatter factor from the offset-corrected spectra [10,11]. For example, Fig. 3 illustrates offset-corrected spectra derived from the raw NIR spectra of the quenched Nylon 6 sample shown in Fig. 2. Note that the baseline offset is carried out by subtracting the spectral intensity at 1300 nm to make the intensities zero. This simple subtraction aims to remove the additive scatter factor providing upward shift to the spectrum. In Fig. 3, it is noted that the spectra still show fluctuations of the spectral intensity even in the region where no chemically meaningful peaks exist. Fig. 4 represents spectral intensity variations at 1485 (amorphous component), 1530 (crystalline component) and 1600 nm where no obvious peaks are observed. Note that the intensity variation at 1600 nm mainly reflects the baseline increase arising from the multiplicative scatter factor. The intensities of the bands associated with the amorphous component show an initial increase at the very onset of the quenching followed by the subsequent decrease. Importantly, it
3
Fig. 4. Spectral intensity variations at 1485 (amorphous component), 1530 (crystalline component) and 1600 nm derived from offset-corrected spectra.
should be noted that the change in the intensity of the crystalline peak results in somewhat similar feature, while the quantity of the crystalline component is supposed to essentially increase during the melt-quenching process. Such trend is also true for the variation at 1600 nm where no meaningful peaks exist. It is thus likely that the intensity variations in the whole spectral region are still overwhelmed with the contribution from the baseline change associated with multiplicative scatter factor. It should be pointed out here that the intensity variation at 1600 nm in the offset spectra purely reflects the signal contribution from the multiplicative scatter factor. In other words, the attenuation of the portion, which is synchronized with this trend, leads to the highly selective removal of the contribution from the multiplicative scatter factor in the whole spectral region. Fig. 5 represents the null-space projected spectra derived from the offset-corrected spectra by using the spectral intensity change at 1600 nm as the vector y. One can find that the spectra are now free from the additive and multiplicative scatter effects after the application of the baseline correction treatments. While the spectral intensities can be both positive and negative after the projection treatment, certain information such as the relative direction of signal changes and the sequential order of intensity variations, is thus fully preserved. The intensity of the amorphous band at 1485 nm shows obvious decrease, mostly indicating the reduction of the amorphous content. It is also noted that the crystalline band
Fig. 5. Projection-corrected spectra obtained by projecting offset-collected spectra onto null-space of their signals at 1600 nm. Fig. 3. Offset-corrected spectra derived from original NIR spectra.
Please cite this article in press as: H. Shinzawa, J. Mizukado, Near-infrared (NIR) monitoring of Nylon 6 during quenching studied by projection two-dimensional (2D) correlation spectroscopy, Journal of Molecular Structure (2016), http://dx.doi.org/10.1016/j.molstruc.2016.02.016
4
H. Shinzawa, J. Mizukado / Journal of Molecular Structure xxx (2016) 1e4
Fig. 6. (A) Synchronous and (B) asynchronous correlation spectra calculated from the projection-corrected spectra shown in Fig. 5.
at 1535 nm is still overwhelmed with the neighboring amorphous band and the intensity seemingly decreases. The gradual increase in the intensity observed over the 1350e1450 nm region may be associated with the contributions from the developments of the crystalline structure. Thus, the application of 2D correlation analysis to the projection corrected spectra becomes important to sort out the pertinent changes of the rather broad and highly overlapped bands corresponding to the contributions from crystalline and amorphous components. (A) Synchronous and (B) asynchronous correlation spectra derived from the projection-treated NIR spectra of Nylon 6 are represented in Fig. 6, respectively. The plot of the reference spectrum is placed at the top and side of the contour map. Negative correlation intensity areas of the contour map are represented by the shading. The synchronous spectrum develops one broad autopeak near 1485 nm, reflecting the gradual increase in the spectral intensity in this NIR region. This is mostly because of the obvious contribution from the amorphous band. A broad negative correlation peak around at (1485, 1400) suggests that spectral intensities at these wavelengths change in the direction opposite each other during the quenching process, probably indicating the presence of the crystalline bands over the 1350e1450 nm region. The corresponding asynchronous spectrum provides an even clearer picture of the time-dependent changes in the NIR intensities of component bands. For example, it develops a positive cross peak at the coordinate (1485, 1535). The appearance of the asynchronous correlation peak suggests that the decrease in the spectral intensity at 1485 nm predominantly occurs before the decrease that at 1535 nm. This can be interpreted to mean that the development of the crystalline component provides additional contribution delaying the intensity variation at 1535 nm. Another notable feature can also be observed along the 1485 nm coordinate of the asynchronous correlation spectrum. The correlation peaks observed over the 1350e1450 nm region along the 1485 nm coordinate reveal the presence of the component developing after the predominant consumption of the amorphous. In other words, the development of the correlation peaks also suggests the multiple crystalline bands appeared in this NIR region. It is thus likely that the, in the earlier stage of the quenching, the disappearance of amorphous component occurs first. The consumption of the amorphous structure is then followed by the development of the crystalline structure. Such observation is actually consistent with the expected population changes of crystalline and amorphous components during the quenching of the molten semicrystalline polymer. Consequently, determination of the sequential order as well as clear differentiation between bands located very close to each other, such as crystalline and amorphous components, is demonstrated.
5. Conclusion 2D projection correlation analysis, coupled with high-speed NIR measurements yields an interesting result on the structure variation of Nylon 6 undergoing the melt-quenching. Although the transient variation of the Nylon 6 sample was readily captured as timeresolved NIR spectra, the main feature of the NIR spectra was overwhelmed by the contribution from the baseline change. Projection treatment was applied to selectively separate the signal contribution of interest from the baseline fluctuation. The 2D correlation spectra derived from the projection-corrected spectra suggest the presence of polymer structures undergoing different variations during the quenching. For example, the synchronous and asynchronous 2D correlation spectra reveal the initial decrease in the contribution of the NIR band at 1485 nm due to the amorphous structure, predominantly existing in the melt Nylon 6. This is then followed by the emerging contribution of the band intensity at 1535 nm associated with the crystalline structure. These results reveal the transient variation of the polymer structures occurring in relatively short time scale, demonstrating a definite advantage of the high-speed NIR monitoring for analyzing fleeting phenomena we often encounter in the real world.
References [1] K. Bakev, Process Analytical Technology. Spectroscopic Tools and Implementation Strategies for the Chemical and Pharmaceutical Industries, Blackwell Publishing Ltd., Oxford, 2005. [2] L. Helmdach, M.P. Feth, C. Minnich, J. Ulrich, Chem. Eng. Process. Process Intensif. 70 (2013) 184. [3] S.S. Kadam, A. Mesbah, E. van der Windt, H.J.M. Kramer, Chem. Eng. Res. Des. 89 (2011) 995. [4] I. Noda, W.M. Allen, S.E. Lindberg, Appl. Spectrosc. 63 (2009) 224. [5] M. Blanco, I. Villarroya, Trends Anal. Chem. 21 (2002) 240. [6] J.G. Rosas, H. de Waard, T. De Beer, C. Vervaet, J.P. Remon, W.L.J. Hinrichs, H.W. Frijlink, M. Blanco, J. Pharm. Biomed. Anal. 97 (2014) 39. [7] G. Mirschel, O. Daikos, K. Heymann, T. Scherzer, B. Genest, C. Sommerer, C. Steckert, Progr. Org. Coat. 77 (2014) 719. [8] J.P. Reed, D. Devlin, S.R.R. Esteves, R. Dinsdale, A.J. Guwy, Biores. Tech. 133 (2013) 398. [9] B.L. Heffner, D.A. Smith, J.E. Baran, K.W. Cheung, Electron. Lett. 24 (1988) 1562. [10] H. Shinzawa, W. Kanematsu, I. Noda, Vib. Spectrosc. 70 (2014) 53. [11] H. Shinzawa, W. Kanematsu, I. Noda, Vib. Spectrosc. 78 (2015) 34. [12] Y. Xu, P. Wu, J. Mol. Struct. 833 (2007) 145. [13] L.W. McKeen, The Effect of Creep and Other Time Related Factors on Plastics and Elastomers (Plastics Design Library), William Andrew, Norwich, 1992. [14] I. Noda, J. Mol. Struct. 974 (2010) 116. [15] I. Noda, Vib. Spectrosc. 60 (2012) 146. [16] I. Noda, Appl. Spectrosc. 47 (1993) 1329. [17] I. Noda, Appl. Spectrosc. 54 (2000) 994. [18] I. Noda, Y. Ozaki, Two-dimensional Correlation Spectroscopy, Willey & Sons, Chichester, West Sussex, 2004. [19] M.K. Kim, S.R. Ryu, I. Noda, Y.M. Jung, Vib. Spectrosc. 60 (2012) 163.
Please cite this article in press as: H. Shinzawa, J. Mizukado, Near-infrared (NIR) monitoring of Nylon 6 during quenching studied by projection two-dimensional (2D) correlation spectroscopy, Journal of Molecular Structure (2016), http://dx.doi.org/10.1016/j.molstruc.2016.02.016