Article
Artifact Correction in TemperatureDependent Attenuated Total Reflection Infrared (ATR-IR) Spectra
Applied Spectroscopy 2017, Vol. 71(8) 1868–1875 ! The Author(s) 2017 Reprints and permissions: sagepub.co.uk/journalsPermissions.nav DOI: 10.1177/0003702817690408 journals.sagepub.com/home/asp
Brian Sobieski1, Bruce Chase1, Isao Noda1,2, and John Rabolt1
Abstract A spectral processing method was developed and tested for analyzing temperature-dependent attenuated total reflection infrared (ATR-IR) spectra of aliphatic polyesters. Spectra of a bio-based, biodegradable polymer, 3.9 mol% 3HHx poly[(R)-3-hydroxybutyrate-co-(R)-3-hydroxyhexanoate] (PHBHx), were analyzed and corrected prior to analysis using two-dimensional correlation spectroscopy (2D-COS). Removal of the temperature variation of diamond absorbance, correction of the baseline, ATR correction, and appropriate normalization were key to generating more reliable data. Both the processing steps and order were important. A comparison to differential scanning calorimetry (DSC) analysis indicated that the normalization method should be chosen with caution to avoid unintentional trends and distortions of the crystalline sensitive bands.
Keywords P(3HB-co-3HHx) copolymers, thermal behavior, infrared spectroscopy, IR, two-dimensional correlation spectroscopy, 2D-COS Date received: 17 November 2016; accepted: 28 December 2016
Introduction Semi-crystalline polymers can exhibit a wide range of properties, depending on the processing methods employed. In many cases, it is the degree of crystallinity which is the molecular level property governing the macroscopic property development. The kinetics of the development of crystallinity can often be a limiting effect in further development of mechanical properties during processing. For example, rapid development of crystallinity can inhibit further development of orientation during draw processing. Since the modulus and strength of a film or fiber fundamentally arises from chain orientation, early or premature crystallization can affect the ultimate mechanical properties achieved from a draw process. A clear understanding of the crystallization process in semi-crystalline polymers is critical to the design and development of processing protocols for any material of interest, whether it be an injection molded part, a bi-axially drawn film, or a fiber. One specific subset of semi-crystalline polymers includes thermoreversible gels. In most of these systems, it is believed that the development of the gel state is driven by a crystallization process, where small crystallites form the tie points for a network which comprises the gel. The development of the mechanical properties of the gel is similarly
very dependent on the rate and extent of crystallization during the thermal cycling process.1–18 Vibrational spectroscopy is an excellent tool for following the crystallization process in polymers. Both the infrared (IR) and Raman spectra are sensitive to the conformational states. Specific conformations are necessary for the crystallization process to proceed. For example, the vibration modes for a trans conformation of the ethylene segments in poly ethylene terephthalate can be observed to increase in intensity as the polymer crystallizes.19 Similarly the carbonyl stretch can be observed at different frequencies dependent on whether the C¼O group is in a crystalline or non-crystalline domain. The use of IR spectroscopy to follow crystallization can sometimes be limited by the thickness constraints imposed by transmission absorption intensities. The use of attenuated total reflection (ATR) techniques is often utilized to overcome the limitations 1
Department of Materials Science and Engineering, University of Delaware, Newark, DE, USA 2 Danimer Scientific, Bainbridge, GA, USA Corresponding author: Bruce Chase, 201 Dupont Hall, University of Delaware, Newark, DE 19716-5600, USA. Email:
[email protected]
Chase et al. imposed by transmission measurements on polymeric samples. However, there are additional problems associated with variable temperature ATR measurements which can include temperature dependent changes in the ATR accessory, baseline drift, change in sample/crystal contact due to expansion/contraction, changes in effective penetration depth and effective path length due to changes in optical constants, and changes in intensity which accompany many of these optical effects. In order to follow changes in structure at the molecular level, these optical effects must be taken into account in processing the ATR spectra.
Experimental Materials and Sample Preparation Poly[(R)-3-hydroxybutyrate-co-(R)-3-hydroxyhexanoate] (PHBHx) 3.9 mol % 3-hydroxyhexanoate (3HHx) was supplied by the Procter and Gamble Company with no further purification. The weight average molecular weight of the polymer was 843 kg/mol. To erase the thermal history of the polymer and to establish a known thermal history, the sample was held above the melt temperature in the differential scanning calorimeter (DSC) before transferring to the ATR.
Differential Scanning Calorimetry The PHBHx sample was sealed in a TA Instruments Tzeropan (R) with a Tzero (R) hermetic lid for DSC analysis using a TA Instruments Discovery DSC with an empty sealed pan as the reference. An initial heating ramp was carried out to 180 C to erase the thermal history, after which the sample was cooled to 20 C. This process was repeated once to obtain thermal information on the sample with the known thermal history for comparison with the IR analysis. The heating and cooling rates were 2 C/min for all ramps.
Fourier Transform Infrared Spectroscopy Fourier transform infrared spectroscopy (FT-IR) spectra were recorded using a Thermo Nicolet 670 Nexus FT-IR spectrometer with a Specac Golden Gate heated diamond ATR attachment, DTGS KBr detector, and a KBr beamsplitter. Sixteen scans at 4 cm1 resolution were co-added and examined between 4000 and 600 cm1. A single background was taken before the sample was placed on the ATR attachment at ambient temperature. One hundred percent line spectra relative to the initial background were collected on the empty cell during a heating ramp at 2 C/min from ambient to 180 C with a 10 min hold at 180 C. Temperature data were manually recorded for each spectrum to the nearest 1 C. Spectra of PHBHx were collected in a similar fashion.
1869
Spectra Processing All spectral processing was conducted using Essential FT-IR v3.50.032 (Madison, WI, USA) with the aid of home build code in Matlab (The Mathworks, Inc.). Subtraction of the ATR 100% line was performed with code written in Matlab. Baseline correction and ATR correction were performed in Essential FT-IR. Baseline correction utilized manually selected multiple points fit using cubic spline. The ATR correction in Essential FT-IR uses an initial estimate of the refractive index in a non-absorbing region. The real and imaginary refractive indices are then compared in an iterative calculation which attempts to converge to the best fit to the observed spectrum.21,22 The ATR correction was performed with a refractive index for the crystal of 2.4 and for the sample of 1.5, a tolerance of 1E-5, an angle of incidence of 45 , a single reflection, and a maximum of 100 iterations to converge.
Two-Dimensional Correlation Spectroscopy Two-dimensional correlation spectroscopy (2D-COS) synchronous and asynchronous spectra were generated using homebuilt code in Matlab based on the procedure developed by Noda.23–29 Analysis of the spectra was performed using the so-called Noda’s Rule, which states that if the sign of the synchronous and asynchronous spectra for a pair of independent wavenumbers, n1 and n2, are the same, then the spectral intensity change of n1 occurs before that at n2. Additionally, if the signs are opposite then the spectral intensity change of n2 occurs before that at n1. All 2DCOS spectra were plotted in Origin combined with their respective initial and final spectra from the corresponding data set. For all 2D-COS figures, blue denotes a negative and red a positive correlation.
Results and Discussion The spectra at several temperatures, as measured, and after corrections are shown in Figures 1 and 2. Figure 3 shows the scale expanded carbonyl-stretching region for the same spectra. The variable temperature ATR spectra have several effects that can give rise to distortion of the spectral features. As the ATR stage is heated, the absorptions arising from the accessory itself can change with temperature. This is evident in Figure 1a, where the features around the diamond absorption bands near 2000 cm1 change with temperature. While this region by itself does not severely interfere with the spectral regions of interest, there is clearly a trailing edge of the diamond absorption feature that extends into the region of the carbonyl-stretching mode at 1720 cm1. Similar effects can also be present in other regions of the spectrum. To correct for this temperature dependence, a set of 100% line spectra were taken of the empty ATR accessory using a background at room temperature under the same heating rate as the sample was
1870
Applied Spectroscopy 71(8)
Figure 1. Neat 3.9 mol % 3HHx PHBHx IR (a) raw and (b) fully corrected spectra obtained during a heating cycle.
Figure 2. Neat 3.9 mol % 3HHx PHBHx scale expanded IR (a) raw and (b) fully corrected spectra obtained during a heating cycle.
measured. Spectral subtraction of the 100% line from the sample spectra was performed using spectra of the heated ATR element at corresponding temperatures. Subtraction of the 100% line not only flattens the baseline but also removes some of the distortion of the carbonyl band. One of the effects of the correction for the temperature-dependent diamond spectrum is the introduction of a baseline offset. Once the 100% line spectra have been subtracted from the sample spectra, baseline and ATR correction are performed. It is important to carry out these two corrections in the proper order. If the ATR correction is performed before the baseline correction, the ATR correction is more likely to fail to converge due to the offset of the baseline. The second step in the correction process is to remove the baseline offset using standard multipoint baseline correction. In this case a cubic spline fit to the baseline was employed. Once the baseline is flat and the spectrum has no offset, the ATR correction can be applied. Finally, a normalization factor must be applied. This comes about due to the change in sample/crystal contact that occurs during heating or cooling. Normalization of the spectra was performed to account for any changes in the
intensities of the peaks due to sample contact with the ATR crystal and other fluctuations. Because temperature is a measure of the kinetic energy of the sample, all the vibrational frequencies should vary with temperature, resulting in apparent minor band shifts and intensity variations caused by changes in the full width half-height (FWHH) of the bands. For this reason, it is inappropriate to normalize to a single spectral band intensity, even if it is a band which is independent of crystallinity. It was found that an approach commonly used in statistical and chemometric analyses generated the best results, specifically standard normal variate (SNV) normalization: xi,norm ¼ xi w1 i Pn x i ¼
wi ¼
j¼1
xi,j
n sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi Pn x x i,j i j¼1 n1
ð1Þ
ð2Þ
ð3Þ
Chase et al. where n is the total number of spectral variables per spectrum, in this case wavenumbers, i is a variable denoting the current spectrum, j is the value for the wavenumber, and xi,j is the spectral intensity, in this case IR absorbance, of the ith spectrum and the jth wavenumber. Furthermore, wi is the
Figure 3. Neat 3.9 mol % 3HHx PHBHx scale expanded IR (a) raw, (b) corrected using 1054 cm1 band as normalization, and (c) corrected using SNV normalization spectra obtained during a heating cycle.
1871 normalization factor for the ith spectrum calculated from the standard normal variate of the spectrum, x i is the mean value of the ith spectrum, and xi,norm is the normalized spectral intensity vector of the ith spectrum. The effectiveness of the corrections is shown in Figures 1 and 2. Note the appearance of multiple isosbestic points for the corrected spectra. This result suggests that the model of a two phase, crystalline/non-crystalline composition may be, to first order, correct. As was previously mentioned, the normalization factor must be chosen carefully. For instance, though the IR band centered at 1054 cm1 is not directly associated with the crystalline phase of the polymer, normalizing to the spectral intensity at 1054 cm1 causes artificial changes in the spectral intensities of some bands. Figure 3b illustrates this. Compared to the uncorrected and the fully corrected SNV normalized spectra, the spectra normalized to 1054 cm1 now exhibit an increase in the 1720 cm1 crystalline band going from 41 C to 131 C. This increase would suggest that recrystallization occurs in the system on heating. Recrystallization is not unusual for semi-crystalline polymers; however, a comparison to the thermal analysis results from DSC in Figure 4 reveals no exotherm at 41 C indicating that no recrystallization occurred in this sample upon heating. Additionally, the DSC thermal curve indicates the existence of two distinct endothermic melting transitions for this polymer sample with peak temperatures of 145 C and 155 C and comparable transitional enthalpies. Multiple melting points are not uncommon in semi-crystalline polymers and are typically attributable to separate crystalline phases, distinct crystal size distribution, and/or recrystallization. While this is an interesting phenomenon that requires explanation, it is outside of the scope of this paper and will be revisited in future publications by this group. However, preliminary results indicate that the dual
Figure 4. Differential scanning calorimetry heating trace of neat 3.9 mol % 3HHx PHBHx at 2 C/min after cooling from melt at 2 C/min.
1872 melting behavior is due to the morphology of the polymer. The smaller and imperfect alpha crystals melt during the first endotherm because they are less stable than the larger and more perfectly formed crystals, which melt during the second endotherm. Because both endothermic transitions are caused by melting of the alpha crystals, the overall trend in the IR spectra does not indicate two separate transitions. Standard normal variate normalization showed a monotonic decrease in intensity at 1720 cm1 with increasing temperature. This eliminates the suggestion of recrystallization behavior found when normalizing to the 1054 cm1 band. Close examination of the 1054 cm1 band intensity in Figure 2b revealed that the 1054 cm1 band decreased with temperature. This decrease in 1054 cm1 was the cause of the apparent increase in the crystal band intensities. Standard normal variate normalization removed the initial increase in the 1720 cm1 vibrational frequency and revealed that the 1054 cm1 band decreases with temperature, while the 1720 cm1 remains constant before decreasing upon melting. It was the decrease in 1054 cm1 spectral intensity that caused the apparent increase in the carbonyl peak, because the constant value at 1720 cm1 was divided by a decreasing value. Therefore, one must be cautious when selecting the normalization method to avoid artifacts in the resulting spectra. Standard normal variate normalization is a good general normalization method, for it removes any user bias, but may not be appropriate in every situation, such as a variation in sample concentration between spectra. It is always advisable to consider several normalization approaches and to look for variations which may indicate artifacts resulting from normalization. A comparison between the raw and corrected spectra is shown in Figure 5. By correctly processing the spectra, the diamond absorbance distortion has been minimized, the baseline returned to zero and flattened, and the relative intensities of the bands corrected. Without this spectral processing method, the trends observed in the data set would be distorted and the spectra analyzed using 2DCOS might be misleading. In addition, a direct comparison of band intensities between separate spectra would be unreliable, if not impossible. The 2D-COS synchronous and asynchronous spectra were generated for the full temperature perturbation from 22 C to 180 C using the uncorrected and fully corrected spectra to determine the necessity of the proposed spectral processing. The 2D-COS spectra combined with the initial and final spectra in the data set of the uncorrected spectra of neat 3.9 mol % 3HHx PHBHx during a heating cycle are shown in Figure 6. Immediately one notices several issues with the results for the uncorrected spectra. The one-dimensional (1D) spectra illustrate the shift of the baseline with increasing temperature due to the increase in the IR absorbance of the ATR crystal. This baseline shift indicates that some of
Applied Spectroscopy 71(8)
Figure 5. Infrared spectra of neat 3.9 mol % 3HHx PHBHx at 180 C (a) no data correction and (b) after data correction.
the specific correlation peaks may be untrustworthy, since they could be attributable to this variation. Second, the synchronous spectrum contains auto peaks centered at 1718 and 1732 cm1, respectively, for the crystalline band and the amorphous band. While frequency shifting is a known issue of ATR-IR spectra, nevertheless it makes comparison with the literature difficult. The asynchronous spectrum depicts the most obvious issue with the uncorrected spectra, namely the intense cross-peak extending along 1720 cm1 beyond 1800 cm1. This cross-peak is approximately onethird the intensity of the maximum correlation and is caused simply by the baseline drift with increasing temperature. Such an intense correlation could cover up additional information in the 2D-COS spectra and also calls any other crosspeaks of that intensity into question, for they could originate from the instrument and not the sample. Questioning of the source of the correlation peak is especially true for the spectra in Figure 6, because the main cross-peak feature is in the same direction, negative, as the baseline drift. By utilizing Noda’s Rule for 2D-COS analysis, it was found that the order of band change was: decrease in 1687 cm1,
Chase et al.
1873
Figure 6. Two-dimensional correlation spectroscopy synchronous and asynchronous spectra of uncorrected neat 3.9 mol % 3HHx PHBHx temperature perturbation from 22 to 180 C during heating cycle.
Figure 7. Two-dimensional correlation spectroscopy synchronous and asynchronous spectra of fully corrected neat 3.9 mol% 3HHx PHBHx temperature perturbation from 22 C to 180 C during heating cycle.
increase in 1740 cm1, decrease in 1706 cm1, decrease in 1712 cm1, increase in 1737 cm1, decrease in 1715 cm1, increase in 1732 cm1, increase in 1730 cm1, decrease in 1720 cm1, and decrease in 1718 cm1. The 2D-COS spectra combined with the initial and final spectra in the data set of the fully corrected spectra of neat 3.9 mol % 3HHx PHBHx during a heating cycle are shown in Figure 7. With the fully processed spectra, 2D-COS generated a synchronous spectrum with auto peaks at 1721 and 1740 cm1, respectively, for the crystalline and amorphous bands. As with the 1D spectra, these vibrational
frequencies are in agreement with the literature and allow direct comparison with previous results. By utilizing Noda’s Rule for 2D-COS analysis, it was found that the order of band change was: decrease in 1699 cm1, decrease in 1686 cm1, increase in 1754 cm1, increase in 1740 cm1, decrease in 1726 cm1, and decrease in 1721 cm1. The order and band distinction is in close agreement with previous work.30,31 Relating back to the DSC analysis, the 2D-COS analysis reveals an indication of the dual melting phenomenon observed in the DSC. There is asynchronous behavior between the alpha crystal
1874 bands at 1721 and 1726 cm1, the first indication that the two endotherms are both due to the melt of alpha crystals. It is interesting to note that the overall pertinent results from the 2D-COS analysis of the temperaturedependent spectra are actually retained even in the uncorrected spectra. It shows that 2D-COS as an analysis tool is quite robust and can generate reliable results even when the spectra are heavily distorted. However, if specific details or detection of smaller transitions are required, the processing steps outlined in this manuscript are necessary. For instances, while the raw spectra were still able to generate the correct overall order of change in the spectra, the 2D-COS analysis generated an additional four vibrational bands, confusing the order of spectral variation. Additionally, though it is difficult to compare due to the ATR correction, it appears that the feature at 1699 cm1, which is easily distinguished in the processed asynchronous spectrum, is completely lost in the raw spectra. Similarly, what has been called a possible intermediate around 1730 cm1,30,31 in this case 1726 cm1, is difficult to observe in the raw spectra. It may be that the feature at 1720 cm1 in the uncorrected spectra is this intermediate, but its close position to the main crystalline band at 1718 cm1 causes such a distinction to be questionable at best.
Conclusion A spectral processing method was developed and tested for analyzing temperature-dependent IR spectra of a bio-based, biodegradable polymer. It was found that removal of the temperature variation of ATR diamond absorbance, correction of the baseline, ATR correction, and normalization were key to generating useable data. In addition, the order of processing and the normalization method used were just as important to avoid any distortions and anomalies. Two-dimensional correlation spectroscopy was performed to determine the importance of the spectral processing by comparing raw spectra to those fully corrected. It was found that, due to the robust nature of 2D-COS, the overall order of vibrational band intensity changes was similar between the data sets, but that fine details were difficult to distinguish. Furthermore, the distortions in the raw spectra raise concerns pertaining to the source of the correlation peaks, making results from the uncorrected spectra questionable at best. Conflict of Interest The authors report there are no conflicts of interest.
Funding The authors acknowledge the support from Delaware NSF EPSCoR Grant no. 1301765 and the NSF DMR Polymers Program Grant no. 1407255.
Applied Spectroscopy 71(8) References 1. D.R. Picout, S.B. Ross-Murphy. (2002). Thermoreversible and Irreversible Physical Gels from Biopolymers. In: Y. Osada, A.R. Khokhlov, editors. Polymer Gels and Networks. New York: Marcel Dekker. Pp. 27–46. 2. C. Daniel, C. Dammer, J.M. Guenet. ‘‘On the Definition of Thermoreversible Gels: The Case of Syndiotactic Polystyrene’’. Polymer. 1994. 35(19): 4243–4246. 3. A. Prasad, H. Marand, L. Mandelkern. ‘‘Supermolecular Morphology of Thermoreversible Gels Formed from Homogeneous and Heterogeneous Solutions’’. J. Polym. Sci., Part B.: Polym. Phys. 1993. 31(12): 1819–1835. 4. C. Daniel, M.D. Deluca, J.M. Guenet, A. Bruˆlet, A. Menelle. ‘‘Thermoreversible Gelation of Syndiotactic Polystyrene in Benzene’’. Polymer. 1996. 37(7): 1273–1280. 5. C. Daniel, A. Menelle, A. Brulet, J.M. Guenet. ‘‘Thermoreversible Gelation of Syndiotactic Polystyrene in Toluene and Chloroform’’. Polymer. 1997. 38(16): 4193–4199. 6. M. Tazaki, R. Wada, M.O. Abe, T. Homma. ‘‘Crystallization and Gelation of Poly(vinylidene fluoride) in Organic Solvents’’. J. Appl. Polym. Sci. 1997. 65(8): 1517–1524. 7. P.J. Yadav. ‘‘Solvent Retention, Thermodynamics, Rheology and Small Angle X-ray scattering Studies on Thermoreversible Poly(vinylidene fluoride) Gels’’. J. Phys. Chem. B. 2010. 114(35): 11420–11429. 8. Y. Peng, B. Sun, P. Wu. ‘‘Two-Dimensional Correlation Infrared Spectroscopic Study on the Crystallization and Gelatio of Poly(vinylidene fluoride) in Cyclohexanone’’. Appl. Spectrosc. 2008. 62(3): 295–301. 9. M. Okabe, R. Wada, M. Tazaki, T. Homma. ‘‘The Flory-Huggins Interaction Parameter and Thermoreversible Gelation of Poly(vinylidene fluoride) in Organic Solvents’’. Polym. J. 2003. 35(10): 798–803. 10. M. Dahmani. ‘‘Poly(vinyl chloride) Thermoreversible Gels: A Fractal Approach to the Relation between Structure and Mechanical Properties’’. Macromolecules. 1997. 30(5): 1463–1468. 11. K. Kawanishi. ‘‘Thermodynamic Consideration of the Sol-Gel Transition in Polymer Solutions’’. Polymer. 1987. 28(6): 980–984. 12. H. Reinecke. ‘‘Complex Formation in Poly(vinyl chloride) Thermoreversible Gels’’. Macromolecular Symposia. 1997. 114(1): 309–314. 13. N. Fazel. ‘‘Molecular Structure and Thermal Behavior of Poly(methyl methacrylate) Thermoreversible Gels and Aggregates’’. Macromolecules. 1994. 27(14): 3836–3842. 14. A. Saiani. ‘‘On the Helical Form in Syndiotactic Poly(methyl methacrylate) Thermoreversible Gels As Revealed by Small-Angle Neutron Scattering’’. Macromolecules. 1997. 30(4): 966–972. 15. A. Saiani. ‘‘Nanostructure and Helicity in Syndiotactic Poly(methyl methacrylate) Thermoreversible Gels’’. Macromolecules. 1999. 32(3): 657–663. 16. C. Daniel. ‘‘Study of Ethylene-Propylene Copolymer Thermoreversible Gels’’. Macromolecules. 1999. 32(26): 8938–8944. 17. H.M. Tan. ‘‘Relationship between Crystallinity and Thermoreversible Gelation’’. Eur. Polym. J. 1983. 19(10): 1021–1025. 18. A. Turchetto. ‘‘Gel-Sol Phase Transition of Poly(3-hydroxybutyrate) in Dimethylformamide’’. Thermochimica Acta. 1995. 269–270: 307–317. 19. X.T. Li, Y. Zhang, G.Q. Chen. ‘‘Nanofibrous Polyhydroxyalkanoate Matrices as Cell Growth Supporting Materials’’. Biomaterials. 2008. 29(27): 3720–3728. 20. S. Frisk, R.M. Ikeda, D.B. Chase, J.F. Rabolt. ‘‘Determination of the Molecular Orientation of Poly(propylene terephthalate) Fibers Using Polarized Raman Spectroscopy: A Comparison of Methods’’. Appl. Spectrosc. 2004. 58(3): 279–286. 21. J.E. Bertie, H.H. Eysel. ‘‘Infrared Intensities of Liquids I: Determination of Infrared Optical and Dielectric-Constants by FT-IR Using the Circle ATR Cell’’. Appl. Spectrosc. 1985. 39(3): 392–401.
Chase et al. 22. J.E. Bertie. ‘‘Optical Constants’’. In: J.M. Chalmers, P.R. Griffiths, editors. Handbook of Vibrational Spectroscopy. Chichester: John Wiley and Sons, 2002. Pp. 88–100. 23. I. Noda. ‘‘Two-Dimensional Infrared (2D IR) Spectroscopy: Theory and Applications’’. Appl. Spectrosc. 1990. 44(4): 550–561. 24. I. Noda. ‘‘Generalized Two-Dimensional Correlation Method Applicable to Infrared, Raman, and Other Types of Spectroscopy’’. Appl. Spectrosc. 1993. 47(9): 1329–1336. 25. I. Noda, A.E. Dowrey, C. Marcott, G.M. Story, Y. Ozaki. ‘‘Generalized Two-Dimensional Correlation Spectroscopy’’. Appl. Spectrosc. 2000. 54(7): 236A–248A. 26. I. Noda. ‘‘Determination of Two-Dimensional Correlation Spectra Using the Hilbert Transform’’. Appl. Spectrosc. 2000. 54(7): 994–999. 27. I. Noda. ‘‘General Theory of Two-Dimensional (2D) Analysis’’. In: J.M. Chalmers, P.R. Griffiths, editors. Handbook of Vibrational
1875
28. 29.
30.
31.
Spectroscopy V.3. Chichester, UK: John Wiley and Sons, 2002. Pp. 2123–2134. I. Noda. ‘‘Close-up View on the Inner Workings of Two-Dimensional Correlation Spectroscopy’’. Vib. Spectrosc. 2012. 146–153. I. Noda. ‘‘Techniques Useful in Two-Dimensional Correlation and Codistribution Spectroscopy (2DCOS and 2DCDS) Analyses’’. J. Mol. Struct. 2016. 1124: 29–41. H. Sato, R. Murakami, A. Padermshoke, F. Hirose, K. Senda, I. Noda, Y. Ozaki. ‘‘Infrared Spectroscopy Studies of CHO Hydrogen Bondings and Thermal Behavior of Biodegradable Poly(hydroxyalkanoate)’’. Macromolecules. 2004. 37(19): 7203–7213. A. Padermshoke, H. Sato, Y. Katsumoto, S. Ekgasit, I. Noda, Y. Ozaki. ‘‘Melting Behavior of Poly(3-hydroxybutyrate) Investigated by TwoDimensional Infrared Correlation Spectroscopy’’. Vib. Spectrosc. 2004. 64(4): 550.
