Fourier Transform Infrared (FT-IR) Spectroscopic ... - OSA Publishing

3 downloads 0 Views 696KB Size Report
Fourier Transform Infrared (FT-IR). Spectroscopic Imaging Analysis of Partially. Miscible PMMA–PEG Blends Using. Two-Dimensional Disrelation Mapping.
Article

Fourier Transform Infrared (FT-IR) Spectroscopic Imaging Analysis of Partially Miscible PMMA–PEG Blends Using Two-Dimensional Disrelation Mapping

Applied Spectroscopy 2017, Vol. 71(6) 1189–1197 ! The Author(s) 2016 Reprints and permissions: sagepub.co.uk/journalsPermissions.nav DOI: 10.1177/0003702816670917 journals.sagepub.com/home/asp

Hideyuki Shinzawa1,2, Junji Mizukado1, and Sergei G. Kazarian2

Abstract A novel technique called disrelation spectroscopic imaging describes the process of identifying an area where a coordinated or out-of-phase change in pattern of spectral absorbance occurs. Disrelation mapping can be viewed as a spatial filter based on the well-established two-dimensional (2D) correlation function to highlight specific areas where disrelated variation occurs between n1 and n2. Disrelation intensity develops only if the spectral absorbance measured at n1 and n2 vary out of phase with each other within a specific spatial area. The disrelation mapping locates regions where absorbance varies in a dissimilar manner because of the contribution from species of different physical or chemical origins. Consequently, it becomes possible to probe onset of molecular interactions or presence of intermediate forms between components, which is not fully detected by the conventional visualizations based on a single wavenumber. Data analysis using disrelation mapping applied to Fourier transform infrared (FT-IR) spectroscopic images is presented in this study. Data sets of FT-IR spectroscopic images of blends of poly(methyl methacrylate) (PMMA) and polyethylene glycol (PEG) were subjected to the disrelation mapping. It was found that the disrelation intensity between 1730 and 1714 cm–1 becomes especially acute around the spatial boundary between PMMA and PEG domains within the studied blend sample. Thus the band at 1730 cm–1 most likely represents the C¼O stretching mode of the C¼OH–O species due to the intermolecular hydrogen bonding between PMMA and PEG. The appearance of such disrelation is more noticeable in the PEG-rich region, for the PEG with low molecular weight. Consequently, it suggests that the blends of PMMA and PEG are partially miscible at the molecular level and these intermolecular interactions are affected by the quantity of the terminal –OH groups of the PEG.

Keywords Fourier transform infrared imaging, FT-IR, multiple perturbation two-dimensional correlation spectroscopy, 2D, ATR, poly(methyl methacrylate), PMMA, polyethylene glycol, PEG Date received: 25 July 2016; accepted: 23 August 2016

Introduction Spectroscopic imaging is a technique to measure spectra for every pixel dividing an area into thousands of spatial parts.1–3 For example, Fourier transform infrared (FT-IR) spectroscopic imaging is based on use of focal plane array IR detectors, where each pixel of such a detector simultaneously measures a full IR spectrum, and each image generated is a distribution of absorbance of a spectral band at a particular wavenumber in the measured area as a function of all pixels.4–6 The visual inspection of the FT-IR image allows examination of the distribution of specific chemical components within the observed area.7,8 In this study, a technique called disrelation mapping is used to describe and identify areas where coordinated or

out-of-phase changes in the pattern of spectral absorbance occurs. A schematic illustration of disrelation mapping is shown in Figure 1. In measuring a set of FT-IR spectra 1 National Institute of Advanced Industrial Science and Technology (AIST), Japan 2 Department of Chemical Engineering, Imperial College London, London, UK

Corresponding authors: Hideyuki Shinzawa, National Institute of Advanced Industrial Science and Technology (AIST), Japan. Email: [email protected] Sergei G. Kazarian, Department of Chemical Engineering, Imperial College London, London, UK. Email: [email protected]

1190

Applied Spectroscopy 71(6)

Figure 1. A schematic illustration of disrelation mapping based on 2D correlation convolution filter.

under varying spatial positions, a set of FT-IR spectroscopic images is obtained. Analysis of the spectroscopic imaging data can be done by slicing the cube at a fixed wavenumber to visualize the variation of the spectral absorbance.3,6 On the other hand, the disrelation map is obtained by carrying out a so-called convolution filter based on cross-correlation analysis of FT-IR images measured at two different wavenumbers. For example, a set of spectra falling inside of a convolution window defined by neighboring pixels are subjected to two-dimensional (2D) correlation analysis to derive disrelation intensity within the local region.9–11 Disrelation intensity develops only if spectral absorbance measured at two different wavenumbers vary out of phase with each other. In other words, it works as an index representing asynchronous or dissimilar changes of the patterns between the images constructed with the two different wavenumbers. Thus, one can expect to see the appearance of a substantial level of correlation intensity only in the regions where chemically or physically meaningful variation occurs. The conventional visual inspection of the IR image highlights the spatial regions having the significant spectral absorbance, which, in turn, allows probing the areas where specific component dominantly exists. On the other hand, the disrelation map emphasizes the regions where spectral absorbance varies in a dissimilar manner. It therefore becomes possible to probe the onset of molecular interaction or presence of intermediate form between components, which is not fully detected by the conventional visualization based on single wavenumber. An illustrative example is provided with IR imaging analysis of partially miscible polymer blends of poly(methyl methacrylate) (PMMA) and polyethylene glycol (PEG), each having a different molecular weight of PEG, to show how this techniques can be utilized. A series of analyses based on 2D correlation analysis is applied to the IR imaging data to demonstrate the ability to sort

out the key information, which eventually assists the unambiguous interpretation of the system. In this work, the spectroscopic evidence clearly shows PMMA and PEG substantially interact at the molecular level and this interaction is also affected by the quantity of the terminal –OH group of the PEG, which is an important requirement to develop the hydrogen bonding with PMMA. Consequently, the partial molecular interaction allows the blends to respond to the spatial perturbations in different manner.

Methods Doubly Two-Dimensional Correlation Spectra The main idea behind disrelation map is 2D correlation analysis with two spatial perturbation variables, which is often referred to as multiple perturbation 2D correlation analysis.12–17 We assume that a set of spectra A(n,p,q) of a system under multiple perturbations are collected as a function of an appropriate spectral variable, like wavenumber n. The two additional variables, p ¼ 1,2,. . .,P and q ¼ 1,2,. . .,Q, represent different perturbation variables, e.g., x- and y-axis coordinates, respectively. Collective 2D correlation spectra are derived directly from the three-way data matrix A(n,p,q) by lumping two perturbation variables p and q together.16,17 Reference and dynamic spectra over p and q are described as P X 1 X  AðnÞ ¼ Aðn, p, qÞ PQ p¼1 q¼1

ð1Þ

~ p, qÞ ¼ Aðn, p, qÞ  AðnÞ  Aðn,

ð2Þ

Q

Shinzawa et al.

1191

Doubly synchronous and asynchronous correlation spectra for p and q planes can be described as pq ðn1 , n2 Þ ¼

Q P X 1 X ~ 1 , p, qÞ  Aðn ~ 2 , p, qÞ Aðn PQ  1 p¼1 q¼1

ð3Þ pq ðn1 , n2 Þ ¼

Q P X 1 X ~ 1 , p, qÞ  A~ 6¼ ðn2 , p, qÞ Aðn pq PQ  1 p¼1 q¼1

ð4Þ where A~ 6¼ represents pq ðn2 , p, qÞ transformation defined by A~ 6¼ pq ðn2 , p, qÞ ¼

P X

Npr

r¼1

 Npr ¼

0 1=ðr  pÞp

Q X

the

Hilbert–Noda

~ 2 , r, sÞ Nqs Aðn

ð5Þ

s¼1

for p ¼ r otherwise

ð6Þ

In the spatial 2D correlation analysis, the intensity of a synchronous 2D correlation spectrum pq ðn1 , n2 Þ represents the simultaneous or coincidental changes of spectral absorbance at n1 and n2 caused by the increase in distance from a reference point on image.14,15 On the other hand, the intensity of an asynchronous spectrum pq ðn1 , n2 Þ indicates the sequential or successive changes of spectral absorbance at n1 and n2, caused by the change of the spatial coordinate.16,17 In other words, the doubly correlated 2D correlation spectra reveal the difference in the change of pattern on the images constructed with n1 and n2. It is important to point out that the determination of the sequential order is generally less significant in the actual practice of the spatial 2D correlation analysis. For example, if the reference point is switched to the point on the opposite side of the image or we rotate the image 180 , the sign of asynchronous correlation becomes opposite providing the different sequential order of the events. It is thus convenient to estimate the disrelation spectrum as an approximation for the doubly asynchronous spectrum by circumventing the Hilbert–Noda transformation, which is computationally somewhat demanding. 2pq ðn1 , n2 Þ ¼ pq ðn1 , n1 Þpq ðn2 , n2 Þ  2pq ðn1 , n2 Þ

ð7Þ

The disrelation can be seen as a portion of the total joint variance of signal fluctuations measured at n1 and n2, that is not accounted for by covariance 2pq ðn1 , n2 Þ.9,12,16,17 The intensity of pq ðn1 , n2 Þ can be used as an index to estimate dissimilar change of the patterns on the images constructed with n1 and n2, while it provides only positive values. The use of the disrelation spectrum becomes especially

attractive if the main purpose of 2D analysis is simply to identify pertinent peaks or differentiate overlapped peaks by taking advantage of the high-resolution feature of 2D spectroscopy. For example, one can expect to see obvious appearance of the correlation peak, when patterns of the distribution at n1 and n2 vary out of phase with each other (that is, delayed or accelerated) due to the development of peaks associated with molecular interaction or complex.17

Disrelation Map Thus far, multiple 2D correlation analysis of the spectroscopic imaging data was mainly used to identify the presence of a specific pair of spectral bands, showing dissimilar change in absorbance from each other on the image plane.16,17 In this study, this technique is extended to identify specific spatial area where the dissimilar variation predominantly occurs. Disrelation mapping is based on doubly correlated analysis applied to each element of the image with its local neighbors. Figure 2 illustrates a schematic illustration to construct the image based on the disrelation intensity. Now we assume a set of IR images measured at between n1 and n2 under varying spatial coordinate. Spectra in a spatial region falling inside of a convolution window defined by neighboring pixels are subjected to multiple 2D correlation analysis described above. A new value for the center pixel in the window is obtained as disrelation intensity between v1 and v2. The window is moved across the entire plane of the image and it eventually generates a new image based on the disrelation intensity pq ðn1 , n2 Þ. It is interesting to point out that this scheme is somewhat analogous to so-called convolution filter which is widely used in the field of image processing.18 Namely, this technique can be seen as a kind of spatial filter based on the 2D correlation function to highlight specific area where disrelated variation between n1 and n2 predominantly occurs. pq ðn1 , n2 Þ develops only if spectral absorbance at n1 and n2 within the local spatial area varies out of phase with each other and not simultaneously. In other words, no obvious generation of pq ðn1 , n2 Þ can be observed if the changes of spectral absorbance at n1 and n2 in the neighboring pixels complements each other. One thus can expect to see appearance of the substantial level of correlation intensity only in the areas where chemically or physically meaningful variation occurs. This feature is well suited for the identification of the dissimilar variation of spectral absorbance by the development of peaks associated with molecular interactions or complexes, predominantly appearing at the boundary between the components. Since the correlation intensity is determined from dynamic spectral intensities observed at two arbitrarily chosen spectral variables v1 and v2, there are many options in the selection of a specific pair of n1 and n2 to construct the image based on pq ðn1 , n2 Þ. One option to choose a specific pair of spectral variables is to carry out doubly 2D

1192

Applied Spectroscopy 71(6)

Figure 2. A convolution filter to highlight disrelation within local spatial area.

correlation analysis first with whole data to identify the development of correlation peaks at specific coordinate (n1,n2) on a 2D correlation map. Only the pair of spectral image planes defined at n1 and n2 is then subjected to a series of manipulation process based on the 2D correlation convolution filter. This reduces computational steps by circumventing the need to calculate correlation intensities of all the possible combinations spectral variables n1 and n2.

Experimental Procedure Material Polyethylene glycols (PEGs) with average molecular weights of 2000 and 2 000 000 g/mol and PMMA were supplied by Wako Chemicals Company. They were thoroughly dried in vacuum at 70  C overnight before the manipulation. A total of 1.5 g of PMMA and PEG were melt mixed with a Laboplastmill micro KF-6 V extruder (Toyo Seiki Seisakusho, Ltd., Japan) at 10 rpm and 190  C for 1 min. One gram blend samples were hot-pressed at 190  C for 10 min. The samples were press-molded with a Teflon spacer of 1 mm thickness by applying 5 MPa pressure and then quenched in the ice water.

Fourier Transform Infrared Spectroscopic Imaging A Golden Gate diamond attenuated total reflection (ATR) accessory with pressure control (Specac Ltd., UK) was aligned in the external sample compartment of Tensor 27 spectrometer (Bruker, Germany) coupled with a macrochamber containing a 64  64 FPA detector for macro ATR FT-IR spectroscopic imaging measurements. Samples were placed on the measuring surface of a diamond prism of an ATR accessory. Constant pressure was applied by a torque wrench to provide a close contact between the sample and prism. Sets of FT-IR spectra were measured by co-adding 256 scans over a region of 3100–1600 cm–1

with spectral resolution of 4 cm–1 and spatial resolution of approximately 20 mm.

Results and Discussion Fourier Transform Infrared Spectra of PMMA/PEG Blends PMMA is partially miscible with specific polymers, such as poly(4-vinylphenol) (PVPh), polyethylene oxide (PEO), and PEG, by hydrogen bonding of the C¼O of PMMA and the OH of them.19–21 For example, the C¼O group of PMMA interacts with the terminal –OH group of PEG to develop C¼OH–O species. The analysis of the IR imaging data thus offers an interesting opportunity to identify the generation of the molecular-level interaction caused by the coexistence of the PMMA and PEG. Figure 3 shows FT-IR spectra of the blend of PMMA and PEG-2000000 over the (a) 3100–2700 cm–1 and (b) 1800–1600 cm–1 regions, which are extracted from interfacial region of PMMA/PEG2000000 blend. Note that actual spatial position of the interfacial region can be found in Figure 4, which will be discussed below. In Figure 3a, several peaks arising from PMMA as well as PEG can be found in this region. For example, a peak observed at 2882 cm–1 is assignable to asymmetric stretching mode of CH2 group associated with PEG. A shoulder peak, on the other hand, is observed at 2953 cm–1, which can be assigned to asymmetric stretching mode of CH2 group associated with PMMA. Such a spectral feature obviously suggests the coexistence of multiple peaks arising from PMMA and PEG and it eventually causes overlapping of the peaks in this region. In Figure 3b, peaks associated with PMMA can be observed in the IR region. For example, a peak observed around at 1728 cm– 1 is assignable to the C¼O stretching mode of PMMA.22,23 The absorbance of the bands at 2882 and 1728 cm–1 essentially represent the quantity of the PEG and PMMA in the sample. By plotting the absorbance of the bands, it is

Shinzawa et al.

1193

Figure 3. FT-IR spectra of PMMA/PEG-2000000 blend extracted from interfacial region of PMMA and PEG blend, indicated by gray bars in Figure 4.

Figure 4. FT-IR images of PMMA/PEG-2000000 blend constructed with the distribution of absorbance of spectral bands at a 2882 cm–1 (PEG band), (b) 1728 cm–1 (PMMA band), (c) 2882 cm–1 (PEG band), and (d) 1728 cm–1 (PMMA band). Gray bars in the images indicate the interfacial area where the spectra were extracted for Figure 3.

possible to construct an image revealing the distribution of PEG or PMMA over the observed area of the sample. Figure 4a and b illustrates FT-IR images of the polymer blend of PMMA/PEG-2000000 generated with the spectral absorbance at 2882 and 1728 cm–1, respectively. Gray bars in the figures indicate the area where the spectra were extracted to construct Figure 3. Clearly distinguishable lumps with significant absorbance can be seen in Figure 4a and b, revealing the regions where the PEG or PMMA predominantly exists. It is also interesting to point out that the demarcation between the PMMA and PEG in Figure 4a is somewhat unclear compared to that in Figure 3b. This is mostly due to the overlapping of the PMMA and PEG bands in the IR region, which eventually

provides additional signal contribution to the PEG band at 2882 cm–1. Fourier transform IR images of the polymer blend of PMMA/PEG-2000 constructed with the spectral absorbance at (c) 2882 and (d) 1728 cm–1 are illustrated in Figure 4, respectively. Again, highly coagulated PMMA can be found around the center of the image in Figure 4d, while the demarcation between the PMMA and PEG is not so clear in Figure 4c due to the overlapping of the PMMA and PEG bands. The practical utility of this type of visualization will be found in the area of identifying distribution of the components by virtually exploring the image. However, this analysis is not so useful in probing the possible interaction between different two components. For example, the

1194

Applied Spectroscopy 71(6)

Figure 5. (a) Synchronous and (b) disrelation spectra derived from IR spectra of PMMA/PEG-2000000 blend.

change of the color observed in this figure is obviously dominated by the change in the PMMA content but it provides no clues on the onset of the interaction between PMMA and PEG which one expects to observe at the interface of them. It is rather difficult to directly extract more detailed evolution of the spectral features associated with the molecular interaction from the images. Disrelation mapping is a powerful tool to effectively achieve such a task.

Two-Dimensional Correlation Analysis The basic idea behind the disrelation mapping to identify the interaction between PMMA and PEG is straightforward. If a certain PMMA band is influenced by the presence of PEG, the variation of its spectral absorbance may indeed become different from other PMMA bands. In other words, by looking at the difference in the variation between two wavenumbers, it becomes possible to elucidate the presence of chemically meaningful interactions present in the analyzed samples. Doubly synchronous and disrelation spectra derived from the FT-IR spectra of the polymer blend of PEG and PEG2000000 are illustrated in Figure 5a and b, respectively. The plot of the reference spectrum is placed at the top and side of the contour map. The top left corners in Figure 4 were used as reference points for the calculation. Doubly correlated 2D correlation spectra represent similar or dissimilar spectral intensity variation induced by an increase in geometric distance from the top left corner on the image. The entire spectral plane of the synchronous correlation spectrum is covered with a broad auto-correlation peak observed at 1725 cm–1, reflecting that the bands associated with PMMA are all synchronized, i.e., they are all varying together. Such coincidental changes of PMMA bands mostly suggest the variation of complementary change in the PMMA and PEG contents observed over the entire plane of the image. Disrelated correlation appears only if the patterns of the changes observed at different spectral variables are not identical. This feature becomes especially useful in sorting

out subtle and complex changes in spectral feature and establishing unambiguous assignments of bands. For example, a cross-peak at the coordinate (1730 cm–1, 1714 cm–1) in Figure 5b suggests that different trends in the change in the spectral absorbance at these spectral variables. In other words, when we move from the reference point to the other side of the image, the spectral absorbance at 1730 cm–1 and 1714 cm–1 varies in different manners. The development of cross-peaks between the PMMA bands becomes especially important. For example, if the decrease in the PEG is completely compensated with the development of the PMMA, one should expect to see only the development of a synchronous correlation peak but no asynchronous correlation. Thus, the appearance of the asynchronous correlation peak, in turn, indicates the presence of additional contribution accelerating or delaying the variation of the PMMA band at 1714 cm–1. Such different trends in the change of pattern should reveal the generation of species originating from different spatial dependences, suggesting the presence of different PMMA species. While the origin of this band is not clear at this moment, the disrelation mapping provides key information to elucidate it later. Spectroscopic images of PMMA/PEG-2000000 blend obtained with plotting the distribution of absorbance at (a) 1730 and (b) 1714 cm–1 are presented in Fig. S1 (see Supplemental Material). The entire features of Fig. S1 end up with more or less similar patterns, although it is obvious from the disrelation correlation spectra that the spectral absorbance at 1730 and 1714 cm–1 varies in the different manner. The difficulty in identifying the relevant differences between the distribution patterns arises from the fact that the main features of the images are characterized by the presence of the spatial region having a strong absorbance. Fortunately, the difference in the distribution can often be accentuated by using the ratio of the absorbance collected at two different wavenumbers. For instance, Fig. S2 (see Supplemental Material) illustrates images of the (a) PMMA/PEG-2000000 and (b) PMMA/PEG-2000 blends,

Shinzawa et al.

1195

Figure 6. Disrelation maps of a PMMA/PEG-2000000 and (b) PMMA/PEG-2000 blends calculated with disrelation intensity between 1730 and 1714 cm–1.

generated by plotting the distribution of –the ratio of the spectral absorbance at 1714 and 1730 cm–1. As expected, one can see a significant development in the value of the ratio that can be seen around the boundary between PMMA and PEG in Fig. S2a. The region expressing a brighter color indicates an increased contribution arising from the relative absorbance at 1714 cm–1. On the other hand, the region on the right half of the image is covered with meaningless artifacts. Generation of the unwanted pattern is probably related to the weak spectral absorbance in this region. Such effect becomes even more acute for the PMMA/PEG-2000 blends where no significant difference in the ratio is observed in Fig. S2b. Thus, accentuating the difference in the change of absorbance around each image pixel becomes important.

Disrelation Map The spatial filter based on the disrelation becomes particularly useful to locate the areas where the interaction of PMMA and PEG occurs. Figure 6a represents a disrelation map derived from the absorbance of the spectral bands at 1730 and 1714 cm–1 of PEG/PEG-2000000. The selection of the size of the convolution window is somewhat arbitrary, while the smallest window size is 3  3 pixels because of the fundamental requirement of correlation analysis. For instance, by increasing the window size (e.g., 5  5, 7  7, and 9  9 pixels), one can attenuate the contribution from the coincidental correlation intensity or artifact in the local spatial area. The use of an excessive window size, on the other hand, results in the exclusion of some pixels around the end of the image. This becomes especially problematic when the meaningful disrelation pattern is observed around at the corner of the image. Thus, care must be taken in the selection of the window size. In our case, the use of 5  5 pixels turned out to be useful to enhance the correlation features on the disrelation map by the side-by-side comparison of several window sizes. It is noted that, in Figure 6a, disrelation intensity between 1730 and 1714 cm–1 becomes especially strong around the boundary

between PMMA and PEG. The spectral absorbance at 1730 and 1714 cm–1 varies out of phase with each other within the local spatial area. In other words, the absorbance at 1730 cm–1 changes with the increase or decrease of the amount of PMMA within the local area, but the absorbance at 1714 cm–1 does not follow this trend. Another notable observation here is absence of significant disrelation intensity in the region where only PEG or PMMA dominantly exists. The observations above suggest that the disrelation intensity between 1730 and 1714 cm–1 is due to the presence of additional factor providing the signal contribution, which generates only when PMMA and PEG are brought together. Specifically, the dissimilar variations between the two spectral bands of PMMA are caused by the presence of PEG. It is hence reasonable to draw a conclusion that there are at least two distinct types of molecular arrangements in the PMMA, each represented by a slightly different characteristic wavenumber for corresponding vibrational modes. The lower wavenumber band at 1714 cm–1 most likely represents the C¼O stretching mode of the C¼OH–O species due to the intermolecular hydrogen bonding between PMMA and PEG, generating when PMMA and PEG are brought together. In fact, such development of hydrogenbonded species agrees with our previous study based on Raman spectroscopy.17 We now have found that the disrelation between 1730 and 1714 cm–1 can be viewed as an index revealing the development of the C¼OH–O species. It is interesting to see what happens when a similar analytical approach is used on the sample including the lower molecular weight PEG, in other words, more terminal –OH groups per unit volume. A disrelation map derived from the FT-IR spectra of PMMA/PEG-2000 is illustrated in Figure 6b. As expected, one can find significant decrease in the disrelation intensity in the middle of the FT-IR image because of the dominant presence of the PMMA component and subsequent exclusion of the PEG from this region. On the other hand, careful inspection of Figure 6b reveals the development of a significant level of disrelation intensity can be observed even in the PEG-rich region despite the relatively low

1196 concentration of the PMMA. Such developments of the disrelation intensity can be interpreted to mean that the interactions between the PMMA and PEG are more noticeable in those regions, indicating that the increased –OH groups of the PEG provide more opportunity for the PEG to develop the hydrogen bonding. The PMMA and PEG are partially miscible and this is closely related to the presence of the terminal –OH group of the PEG, which is a requirement to develop the intermolecular hydrogen bonding between the PMMA and PEG. It is also interesting to point out that identifying the existence of such interactions is indeed rather difficult from the images shown in Figure 4 and Figs. S1 and S2. Consequently, the spectroscopic evidence indicating that the PMMA and PEG in the blends are partially miscible at the molecular level is readily derived from the disrelation mapping in the intuitively understandable manner. It also revealed that the presence of interaction depends on the quantity of the terminal –OH group of the PEG, which forms hydrogen bonding with the PMMA.

Conclusion In this work, we presented the application of disrelation mapping, analogous of 2D correlation approach, to FT-IR spectroscopic imaging data for the first time, as demonstrated on an example system of polymer blends. Disrelation mapping can be viewed as a spatial filter based on the 2D correlation function to highlight specific areas where disrelated variation between n1 and n2 predominantly occurs. Disrelation intensity develops only if spectral absorbance at n1 and n2 within the local spatial area varies out of phase with each other. On the other hand, no obvious generation of the correlation intensity can be observed if the changes of absorbance for spectral bands at n1 and n2 in the neighboring pixels complement each other. One thus can expect to see the appearance of a substantial level of correlation intensity only in the area where chemically or physically meaningful variation occurs. The application of disrelation mapping to FT-IR spectroscopic imaging data is present in this study where FT-IR imaging data of blends of PMMA and PEG were measured and subjected to disrelation analysis. The 2D correlation analysis of the FT-IR imaging data revealed the disrelation peaks at (1730 cm–1, 1714 cm–1), showing the presence of a specific PMMA band around at 1714 cm–1 whose variation of absorbance did not follow the trend of the other PMMA bands. Disrelation maps were then constructed with the correlation peak intensity observed at the coordinate (1730 cm–1, 1714 cm–1) in the 2D correlation spectrum. Visual inspection of the disrelation map revealed that the disrelation intensity between 1730 and 1714 cm–1 becomes especially strong around the boundary between the PMMA and PEG-2000000 in a blend sample. It was thus suggested that the band at 1714 cm–1 most likely represents the C¼O

Applied Spectroscopy 71(6) stretching mode of the C¼OH–O species due to the intermolecular hydrogen bonding between the PMMA and PEG, which appears only when the PMMA and PEG are brought together. The development of such disrelation intensity was occurring more frequently in the PEG-rich region of the PMMA/PEG-2000 blend. Consequently, it suggests that the blends of the PMMA and PEG are partially miscible at the molecular level and this molecular interaction is essentially affected by the quantity of the terminal –OH group of the PEG. Conflict of Interest The authors report there are no conflicts of interest.

Funding This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Supplemental Material All supplemental material mentioned in the text, consisting of two figures, is available in the online version of the journal.

References 1. S. Sˇasˇic´, Y. Ozaki. Raman, Infrared and Near-Infrared Chemical Imaging. Hoboken, NJ: John Wiley and Sons Inc., 2010. 2. R. Salzer, H.W. Siesler. Infrared and Raman Spectroscopic Imaging. Weinheim, Germany: Wiley-VC, 2009. 3. H. Shinzawa, K. Awa, W. Kanematsu, Y. Ozaki. ‘‘Multivariate Data Analysis for Raman Spectroscopic Imaging’’. J. Raman Spectrosc. 2009. 40(12): 1720–1725. 4. S.G. Kazarian, K.L.A. Chan. ‘‘ATR-FTIR Spectroscopic Imaging: Recent Advances and Applications to Biological Systems’’. Analyst. 2013. 138(7): 1940–1951. 5. S.G. Kazarian, A.V. Ewing. ‘‘Applications of Fourier Transform Infrared Spectroscopic Imaging to Tablet Dissolution and Drug Release’’. Expert Opin. Drug Deliv. 2013. 10(9): 1207–1221. 6. C. Gendrina, Y. Roggo, C. Collet. ‘‘Pharmaceutical Applications of Vibrational Chemical Imaging and Chemometrics: A Review’’. J. Pharm. Biomed. Anal. 2008. 48(1): 533–553. 7. S.G. Kazarian, K.L.A. Chan. ‘‘FTIR Imaging of Polymeric Materials under High-pressure Carbon Dioxide’’. Macromolecules. 2004. 37(2): 579–584. 8. A. Gupper, P. Wilhelm, M. Schmied, S.G. Kazarian, K.L.A. Chan, J. Reussner. ‘‘Combined Application of Imaging Methods for the Characterisation of A Polymer Blend’’. Appl. Spectrosc. 2002. 56(12): 1515–1523. 9. I. Noda, Y. Ozaki. Two-Dimensional Correlation Spectroscopy. Chichester, UK: John Wiley & Sons, 2004. 10. I. Noda. ‘‘Generalized Two-dimensional Correlation Method Applicable to Infrared, Raman, and Other Types of Spectroscopy’’. Appl. Spectrosc. 1993. 47(9): 1329–1336. 11. I. Noda. ‘‘Determination of Two-dimensional Correlation Spectra Using the Hilbert Transform’’. Appl. Spectrosc. 2000. 54(7): 994–999. 12. H. Shinzawa, S. Morita, K. Awa, M. Okada, I. Noda, Y. Ozaki, H. Sato. ‘‘Multiple Perturbation Two-dimensional Correlation Analysis of Cellulose by Attenuated Total Reflection Infrared Spectroscopy’’. Appl. Spectrosc. 2009. 63(5): 501–506. 13. H. Shinzawa, K. Awa, I. Noda, Y. Ozaki. ‘‘Multiple-perturbation Twodimensional Near-infrared Correlation Study of Time-dependent

Shinzawa et al.

14.

15.

16.

17.

18.

Water Absorption Behavior of Cellulose Affected by Pressure’’. Appl. Spectrosc. 2013. 67(2): 163–170. H. Shinzawa, M. Nishida, W. Kanematsu, T. Tanaka, K. Suzuki, I. Noda. ‘‘Parallel Factor (PARAFAC) Kernel Analysis of Temperature- and Composition-Dependent NMR Spectra of Poly(lactic acid) Nanocomposites’’. Analyst. 2012. 137(8): 1913–1921. H. Shinzawa, T. Genkawa, W. Kanematsu. ‘‘Pressure-Induced Association of Oleic Acid (OA) under Varying Temperature Studied by Multiple Perturbation Two-dimensional (2D) IR Correlation Spectroscopy’’. Vib. Spectrosc. 2012. 1028: 164–169. H. Shinzawa, T.N. Murakami, M. Nishida, W. Kanematsu, I. Noda. ‘‘Near-Infrared (NIR) Imaging Analysis of Polylactic Acid (PLA) Nanocomposite by Multiple-Perturbation Two-dimensional (2D) Correlation Spectroscopy’’. J. Mol. Str. 2014. 1069: 171–175. H. Shinzawa, K. Hashimoto, H. Sato, W. Kanematsu, I. Noda. ‘‘Multiple-Perturbation Two-Dimensional (2D) Correlation Analysis for Spectroscopic Imaging Data’’. J. Mol. Str. 2014. 1069: 176–182. W.K. Pratt. Introduction to Digital Image Processing. Boca Raton, FL: CRC Press, 2013.

1197 19. J. Dong, Y. Ozaki. ‘‘FTIR and FT-Raman Studies of Partially Miscible Poly(methyl methacrylate)/poly(4-vinylphenol) Blends in Solid States’’. Macromolecules. 1997. 30(2): 286–292. 20. Y. Ward, Y. Mi. ‘‘The Study of Miscibility and Phase Behaviour of Phenoxy Blends Using Raman Spectroscopy’’. Polymer. 1999. 40(9): 2465–2468. 21. Y. Ren, A. Matsushita, K. Matsukawa, H. Inoue, Y. Minami, I. Noda, Y. Ozaki. ‘‘Two-Dimensional Fourier-Transform-Raman and Nearinfrared Correlation Spectroscopy Studies of Poly(methyl methacrylate) Blends: 2. Partially Miscible Blends of Poly(methyl methacrylate) and Poly(4-vinylphenol)’’. Vib. Spectrosc. 2000. 23(2): 207–218. 22. L. Pan, X. Pei, R. He, Q. Wan, J. Wang. ‘‘Multiwall Carbon Nanotubes/ Polycaprolactone Composites for Bone Tissue Engineering Application’’. Colloids Surf. B Biointerfaces. 2012. 93(5): 226–234. 23. S. Rajendran, T. Uma. ‘‘Conductivity Studies on PVC/PMMA Polymer Blend Electrolyte’’. Mater. Lett. 2000. 44(3–4): 242–247.