Application of Generalized Two-Dimensional Infrared Correlation Spectroscopy to the Study of a Hydrogen-Bonded Blend

1074 Volume 58, Number 9, 2004 APPLIED SPECTROSCOPY0003-7028 / 04 / 5809-1074$2.00 / 0q 2004 Society for Applied Spectroscopy

Application of Generalized Two-Dimensional InfraredCorrelation Spectroscopy to the Study of aHydrogen-Bonded Blend

HE HUANG, SERGHEI MALKOV, MICHAEL M. COLEMAN, andPAUL C. PAINTER*Department of Materials Science and Engineering, Penn State University, University Park, Pennsylvania 16802

FIG. 1. Scale-expanded FT-R spectra in the region of 1800–1650 cm21

recorded at room temperature of PVPh/PMMA blends cast from MIBK.(1) 80/20, (2) 60/40, (3) 40/60, and (4) 20/80.

FIG. 2. Schematic of curve-fitting for PVPh/PMMA blends at 200 8C.(A) PVPh/PMMA 5 80/20, (B) PVPh/PMMA 5 40/60.

In the preceding studies in this series, generalized two-dimensional(2D) infrared correlation spectroscopy has been applied to the studyof polymer blends with relatively weak intermolecular interactions.In this paper, a miscible system with strong intermolecular inter-actions, hydrogen-bonded blends of poly(4-vinyl phenol) (PVPh)and poly(methyl methacrylate) (PMMA), is considered. It has beenfound that band positions in 2D plots are dependent on the datasets used, due to large peak shifts and/or bandwidth changes. Thisobservation complements our preceding studies, in which it wasfound that new features correspond to maxima, minima, or pointsof inflection in the difference spectra used to generate the 2D plotsand are not normal modes of vibration of specific functional groups.Great care needs to be taken in order not to interpret artifacts ofthe procedure in terms of new spectroscopic features.

Index Headings: Generalized two-dimensional (2D) infrared corre-lation spectroscopy; Hydrogen-bonded blends; Poly(4-vinyl phenol);PVPh; Poly(methyl methacrylate); PMMA; Peak shifts; Bandwidthchanges.

INTRODUCTION

In the first two papers of this series,1,2 the results ofsystematically applying generalized infrared correlationspectroscopy to the characterization of polymer blendswere presented. First, two-dimensional (2D) correlationinfrared spectroscopy was applied to a study of immis-cible blends of polystyrene (PS) and poly(methyl meth-acrylate) (PMMA).1 Asynchronous spectra should not beobtained from such mixtures, but usually are. We dem-onstrated that these spectra were largely due to small dif-ferences in bandwidths that are a result of sample prep-

Received 6 December 2003; accepted 4 May 2004.* Author to whom correspondence should be sent.

aration problems. The bandwidth changes result in theappearance of a characteristic pattern of bands in asyn-chronous spectra, which appear to be very sensitive tothese effects. As for miscible blends with relatively weakintermolecular interactions,2 such as blends of polysty-rene (PS) with poly(vinyl methyl ether) (PVME) andpoly(2,6-dimethyl-1,4-phenylene oxide) (PPO), a moreimportant role is played by peak shifts, which also giverise to a characteristic pattern of bands in asynchronousspectra. These new features have previously been inter-preted in terms of the detection of hidden bands, specificinteractions, and conformational changes,1,2 but they are

APPLIED SPECTROSCOPY 1075

TABLE I. Polymer samples used in the PVPh/PMMA blends.

Name Abbreviation M.W.a Source

Poly(4-vinyl phenol)b

Poly(methyl methacrylate)bPVPhPMMA

20 00015 000

Aldrich Chemical Corp., Inc.Scientific Polymer Production, Inc.

a Weight average.b Both polymers are atactic.

FIG. 3. (A) Synchronous and (B) asynchronous contour plots in theregion 1800 to 1650 cm21 of the data obtained from four blends ofcomposition (by weight): 90/10, 80/20, 70/30, and 60/40, PVPh/PMMA(mean normalization was not performed on the spectra before 2D cor-relation analysis).

FIG. 4. (A) Synchronous and (B) asynchronous contour plots in theregion 1800 to 1650 cm21 of the data obtained from four blends ofcomposition (by weight): 40/60, 30/70, 20/80, and 10/90 PVPh/PMMA(mean normalization was not performed on the spectra before 2D cor-relation analysis).

1076 Volume 58, Number 9, 2004

FIG. 5. FT-IR spectra of PVPh/PMMA blends (by weight) in the region of 1800–1650 cm21. (1) 90/10, (2) 80/20, (3) 70/30, and (4) 60/40.

actually artifacts that correspond to lobes or points ofinflection in the difference spectra used to generate the2D plots. Similar considerations apply to 2D-IR studiesof materials, such as N-methylacetamide, that have beenstudied as a function of temperature.3

In this paper, the application of generalized 2D corre-lation IR spectroscopy will be extended to a misciblesystem with strong intermolecular interactions, hydrogen-bonded blends of poly(4-vinyl phenol) (PVPh) andpoly(methyl methacrylate) (PMMA). Hydrogen-bondedblends present similar but also separate and interestingissues. This is in part due to the fact that peak shifts andbandwidth changes are much larger in hydrogen-bondedblends than those in immiscible blends or miscible blendswith relatively weak intermolecular interactions.

Hydrogen-bonded PVPh/PMMA blends have beencharacterized extensively by 2D Raman and near-infrared(NIR) spectroscopy,4 traditional ‘‘linear’’ infrared spec-troscopy,5,12,16,18–20 nuclear magnetic resonance (NMR),13

differential scanning calorimetry (DSC),7,11,14,15 dynamicmechanical thermal analysis (DMTA),16 and small angleX-ray scattering (SAXS),16 etc. It is the first work citedthat is relevant here. Ren et al.4 studied what they called‘‘partially miscible’’ PVPh/PMMA blends, where theconcentration of PVPh varied systematically between 1%and 10%, using 2D Raman and NIR correlation spec-troscopy. These blends were actually phase-separated asa result of the Dx effect and became single phase uponannealing.5 Using 2D Raman correlation analysis, Ren etal.4 identified a previously undetected band at 1706 cm21

for the hydrogen-bonded C5O of PMMA, and two bandsat 599 and 549 cm21 for the ester group. The Raman NIRhetero-correlation analysis detects two bands at 6764 and4637 cm21 for the hydrogen-bonded hydroxyl group ofPVPh, and one band at 5052 cm21 for the hydrogen-bonded C5O group of PMMA, modes which they reportare not readily identifiable from the one-dimensionalspectra.

Results such as these have led many groups to believethat 2D-IR correlation spectroscopy could be a powerfultool for band resolution in many systems. Unfortunately,

our initial studies suggest that many new features are ar-tifacts. Here we extend our work to an infrared study ofPVPh/PMMA blends over the whole composition range,instead of at low concentrations of PVPh. We will showthat, as before, new features generated in asynchronousplots are simply a result of frequency shifts and/or band-width changes. It will also be shown that the new bandsthat are revealed are dependent on the data set used, andtherefore the difference spectra used for 2D correlationanalysis, confirming that these ‘‘new’’ bands are not thefundamental normal modes of vibrations of functionalgroups. This is an important complement to the findingsof the preceding studies,1–3 that new features correspondto lobes or points of inflection in the difference spectraused to generate the 2D plots.

EXPERIMENTAL

The polymer samples used in this study are listed inTable I. Sample preparation techniques and calculationmethodologies are as described previously.1 The solventused to make polymer blend solutions was methyl iso-butanone (MIBK). These blends do not phase separatewhen films are cast from this solvent (i.e., there is no Dxeffect).8,9 All the spectra employed in the 2D correlationanalysis were mean normalized following our precedingstudies,1 except as stated in the text. Mean normalizationis a process in which all points of a spectrum are dividedby its mean value.21

RESULTS AND DISCUSSION

C5O Stretching Modes in the Region of 1800–1650cm21. The miscibility of PVPh/PMMA blends has beena somewhat controversial issue in the past, but it is gen-erally accepted that PVPh will form miscible blends withPMMA if the right solvents and sample pretreatment con-ditions are used.5,8,9 The main driving force for miscibilityis the hydrogen bonds formed between the C5O groupof PMMA and the –OH group of PVPh. This results inthe appearance of a new band around 1707 cm21, whichis assigned to the hydrogen-bonded C5O band. Figure 1


FIG. 6. (A) Synchronous and (B) asynchronous contour plots in theregion 1770 to 1670 cm21 of the data obtained from four blends ofcomposition (by weight): 40/60, 30/70, 20/80, and 10/90, PVPh/PMMA.

FIG. 7. (A) Mean-normalized FT-IR spectra and (B) the differencespectra of low PVPh (40–10%) content PVPh/PMMA blends.

shows the ‘‘one-dimensional’’ IR spectra of PVPh/PMMA blends as a function of composition in the 1800to 1600 cm21 region of the spectrum. Each spectrum hasbeen normalized to the intensity of the ‘‘free’’ C5O bandin this plot. Figure 2 shows the result of curve fitting. Itcan be seen that two bands give an exceptionally goodfit: the band around 1730 cm21 is the ‘‘free’’ C5O bandand the band around 1707 cm21 is the hydrogen-bondedC5O band. This is a well-known result.22

Two-dimensional correlation analysis of these blends,using (separately) the spectra obtained from high (90–60%) and low (40–10%) PVPh content blends are con-sistent with this conclusion, as shown in Figs. 3 and 4.(Mean normalization was not performed on these data

before 2D analysis.) Only two bands are revealed in bothsynchronous and asynchronous plots. In the synchronousspectra (Figs. 3A and 4A), the two cross-peaks at 1732and 1707 cm21 are positive, implying that they are chang-ing in the same ‘‘direction’’ as the PVPh content. At firstglance, this is counter-intuitive, because it has been firmlyestablished5,23 that the hydrogen-bonded C5O band in-creases in intensity at the expense of the ‘‘free’’ C5Oband as the PVPh content of the blend is increased. Inother words, the ‘‘free’’ and hydrogen-bonded C5Obands should change in different directions. This is not acontradiction, however. If the ‘‘unnormalized’’ spectra ofhigh PVPh content blends in Fig. 5 are carefully exam-ined, it can be seen that the relative intensity of ‘‘free’’and hydrogen-bonded C5O bands are indeed changinginversely as the PVPh content increases from 60% to90%, but the absolute values of the carbonyl band inten-sities of the four PVPh/PMMA blends decrease with in-creasing PVPh content. This gives rise to the two positivecross-peaks at 1732 and 1707 cm21 in Figs. 3A and 4A,showing that they are changing in the same direction inabsolute terms.

It is interesting to notice that if mean normalization21

is performed on the spectra of low PVPh content PVPh/PMMA blends before 2D correlation analysis, the twocross-peaks in Fig. 4A change signs from positive to neg-ative, as shown in Fig. 6A. The two negative cross-peaks,of course, indicate that they are changing in different di-rections, one increasing and the other one decreasing in


FIG. 8. (A) Mean-normalized FT-IR spectra and (B) the differencespectra of high PVPh (90–60%) content PVPh/PMMA blends.

FIG. 9. Scale-expanded FT-IR spectra of PVPh/PMMA blends in the region of 3600–3100 cm21. (1) Pure PVPh, (2) PVPh/PMMA 5 80/20, (3)PVPh/PMMA 5 60/40, (4) PVPh/PMMA 5 40/60, and (5) PVPh/PMMA 5 20/80.

intensity. This can be seen directly in the mean-normal-ized Fourier transform infrared (FT-IR) spectra and thedifference spectra obtained by so-called mean centering(subtracting the mean spectrum from all others in the dataset), as shown in Fig. 7. It should also be noted that after

mean normalization the asynchronous plot of the blendswith 40–10% PVPh content has changed dramatically(Fig. 6B). The ‘‘free’’ C5O band at 1730 cm21 is nowapparently split into two bands at 1736 and 1725 cm21,respectively. This pattern is very similar to what is ob-served as a result of a peak shift.1,2 It is easy to show, asin preceding studies in this series,1,2 that these two newbands correspond to the lobes or points of inflection inthe difference spectra, which can be clearly seen in Fig.7B. In other words, they are not newly revealed funda-mental modes of vibration.

Mean normalization of the blends with a high PVPhcontent did not reveal the same features in the synchro-nous and asynchronous plots, which appeared very sim-ilar to those shown in Fig. 3. This is because unlike thelow PVPh content blends, mean normalization does notchange the overall trend of intensity changes in the freeand bonded bands, as shown in Fig. 8.

Finally in this section, we would like to point out thatif the intensities of the hydrogen-bonded and free bandsvaried in a linear fashion with concentration, we wouldnot expect to see an asynchronous spectrum at all, assum-ing there are no issues with bandwidth changes or fre-quency shifts as a result of sample preparation problems.1

This is because the type of correlation analysis used in2D-IR determines the degree of linear dependence be-tween variables. An asynchronous plot should only beproduced if the relative intensities of the two bands areunrelated, or if they vary in a nonlinear fashion with con-centration. The relative intensities of the free and hydro-gen-bonded bands do indeed vary in a nonlinear mannerwith one another as a function of concentration,1 so inthat sense the asynchronous spectra are ‘‘real’’ and notdue to the types of sample preparation problems that gaverise to these spectra in studies of miscible blends.1 Nev-ertheless, artifacts can still be produced as a result offrequency shifts, etc., as shown above.

Hydroxyl Stretching Modes of PVPh in the 3600 to3100 cm21 Region. The C5O stretching region of the


FIG. 10. (A) Synchronous and (B) asynchronous contour plots in theregion 3600 to 3100 cm21 of the data obtained from four blends ofcomposition (by weight): 90/10, 80/20, 70/30, and 60/40, PVPh/PMMA(no mean normalization to the spectra before 2D correlation analysis).

FIG. 11. Difference spectra of PVPh/PMMA blends with high PVPhconcentrations.

spectrum is relatively simple. Now, we turn our attentionto the hydroxyl stretching vibrations of PVPh, which ab-sorb in the 3600 to 3100 cm21 region of the spectrum,where there are various broad, overlapping bands. Figure9 shows the scale-expanded IR spectra of PVPh andPVPh/PMMA blends in this region. It can be seen thatin the spectrum of PVPh, there are very broad overlap-ping bands, with the overall band profile being centerednear 3385 cm21. This can be attributed to the wide dis-tribution of hydrogen-bonded hydroxyl groups found inthis polymer. There is also a much narrower band ob-

served around 3530 cm21, which has been assigned tofree (non-hydrogen-bonded) hydroxyl groups.23

After mixing with PMMA, there should be at leastthree components in this broad envelope.5,23 The first isdue to free hydroxyls, the second to self-associated hy-droxyls (hydroxyl groups between PVPh molecules), andthe third to inter-associated hydroxyls (hydroxyl groupshydrogen bonded to the carbonyl group of PMMA). Be-cause the average overall strength of the hydrogen bondbetween PVPh and PMMA is less than that of the self-associated pure PVPh, it is expected that the peak posi-tion of inter-associated hydroxyls would lie between theself-associated hydroxyls band and the free hydroxylband.5,23 The IR spectra shown in Fig. 9 actually indicatethat the free –OH band becomes undetectable at highPMMA concentrations, and the intensity of the self-as-sociated hydroxyl band decreases at the expense of theinter-association band. As a result, the overall peak max-imum shifts to higher wavenumber, and the bandwidthdecreases as the PMMA content increases in the blends,as shown in Fig. 9. This profile does not show the (atleast) three components one would expect and it is inter-esting to see if 2D correlation analysis is capable of re-vealing these.

Figure 10 shows the synchronous and asynchronousplots of PVPh/PMMA blends with PVPh content between90% and 60% (the spectra were not mean normalized).In the synchronous plot, three bands are clearly resolvedat 3550, 3450, and 3250 cm21. However, the asynchro-nous spectrum in Fig. 10B suggests that there are fourbands, located at 3550, 3450, 3400, and 3325 cm21, re-spectively. The first two bands are identical to thosefound in the synchronous plot, but the last two bands(3400 and 3325 cm21) are not. The ‘‘standard’’ interpre-tation of these results would be in terms of the presenceof five different, difficult to resolve bands. However,these bands have their origin in the difference spectraused to generate these plots, as shown in Fig. 11. Theoverall width of the band profiles changes systematicallywith PMMA content, and the position of maximum in-tensity also shifts systematically with concentration. Sub-tracting a narrow band profile from a broader one leadsto maxima, minima, and points of inflection that do notcorrespond to normal modes of vibration, as we showed


FIG. 12. (A) Synchronous and (B) asynchronous contour plots in theregion 3600 to 3100 cm21 of the data obtained from four blends ofcomposition (by weight): 40/60, 30/70, 20/80, and 10/90, PVPh/PMMA(no mean normalization).

FIG. 13. Difference spectra of PVPh/PMMA blends with low PVPhconcentrations.

previously.1 Similarly, subtracting a band that is frequen-cy shifted from its counterpart in another spectrum leadsto a bipolar difference band, one with a negative andpositive component. These two difference peaks, whichagain are not the actual normal modes of vibration, giverise to asynchronous peaks.2 All these effects appear toinfluence the 2D correlation plots and it is these featuresthat give rise to the ‘‘new’’ bands in the asynchronousplots.

These difference spectra change with the data set usedfor the calculation of the 2D plots, because the relativeintensities of the bands are different. Figure 12 shows the

OH stretching region of the synchronous and asynchro-nous plots obtained from PVPh/PMMA blends withPVPh content between 40 and 10%. Now, only one bandat 3450 cm21 is found in the synchronous plot, and threebands are revealed in the asynchronous plot. These threebands are located around 3540, 3475, and 3375 cm21,respectively. In other words, four bands are apparentlyrevealed in PVPh/PMMA blends with a PVPh contentbetween 40 and 10%. Again, it is the difference spectrathat give rise to some of these features, as shown in Fig.13. However, compared to PVPh/PMMA blends withPVPh content between 90 and 60%, the number of bandsrevealed and the band positions are obviously different.One might expect that certain bands would be more eas-ily resolved in, say, high PVPh content blends as opposedto low PVPh content mixtures, but the fact that some ofthe new features are at very different frequencies suggeststhat they are artifacts. In fact, it is easy to show that themaxima, minima, and points of inflection generated inthe difference spectra shift with relative band intensitiesand it is these features that generate bands in the asyn-chronous spectra.

SUMMARY

Generalized 2D correlation infrared spectroscopy hasbeen applied to a study of miscible blends with strongintermolecular interactions, hydrogen-bonded PVPh/PMMA blends. As in the preceding studies of immisciblesystems and miscible systems with weak molecular in-teractions, new features have been found in various re-gions of the spectra. But these features depend on thedata sets used and correspond to features in the differencespectra used to generate the 2D plots. This demonstratesthat these new features are not the normal modes of vi-bration of specific functional groups, nor previously hid-den bands. Instead, they are characteristic of bandwidthchanges and/or frequency shifts, both important in thesemixtures. One must take great care in interpreting fea-tures, particularly in asynchronous spectra, in terms ofthe presence of new bands. Often these are artifacts.

ACKNOWLEDGMENTS

This material is based upon work supported by the National ScienceFoundation under Grant No. 0100818 and the Office of Chemical Sci-


ences, U.S. Department of Energy, under grant No. DE-FG02-86ER13537.

1. H. Huang, S. Malkov, M. M. Coleman, and P. C. Painter, Macro-molecules 36, 8148 (2003).

2. H. Huang, S. Malkov, M. M. Coleman, and P. C. Painter, Macro-molecules 36, 8156 (2003).

3. H. Huang, S. Malkov, M. M. Coleman, and P. C. Painter, J. Phys.Chem. A 107, 7697 (2003).

4. Y. Ren, A. Matsushita, K. Matsukawa, H. Inoue, Y. Minami, I.Noda, and Y. Ozaki, Vib. Spectrosc. 23, 207 (2000).

5. H. Huang and P. C. Painter, unpublished results, Penn State Uni-versity (2000).

6. S. R. Fahrenholtz and T. K. Kwei, Macromolecules 14, 1076(1981).

7. S. H. Goh and K. S. Siow, Polym. Bull. 17, 453 (1987).8. C. J. Serman, Y. Xu, P. C. Painter, and M. M. Coleman, Macro-

molecules 22, 2015 (1989).9. C. J. Serman, Ph.D. Thesis, The Pennsylvania State University,

Pennsylvania (1989).

10. Y. Xu, Ph.D. Thesis, The Pennsylvania State University, Pennsyl-vania (1989).

11. J. A. Pomposo, I. Eguiazabal, E. Calahorra, and M. Cortazar, Poly-mer 34, 95 (1993).

12. C. J. Serman, P. C. Painter, and M. M. Coleman, Polymer 32, 1049(1991).

13. X. Zhang, K. Takegoshi, and K. Hikichi, Macromolecules 24, 5756(1991).

14. C. Landry and D. Teegarden, Macromolecules 24, 4310 (1991).15. J. A. Pomposo, I. Eguiazabal, E. Calahorra, and M. Cortazar, Mac-

romolecules 26, 2104 (1993).16. D. Li and J. Brisson, Macromolecules 29, 868 (1996).17. D. Li and J. Brisson, Macromolecules 30, 8425 (1997).18. J. Dong and Y. Ozaki, Macromolecules 30, 286 (1997).19. D. Li and J. Brisson, Polymer 39, 793 (1998).20. D. Li and J. Brisson, Polymer 39, 801 (1998).21. S. Sasic, A. Muszynski, and Y. Ozaki, Appl. Spectrosc. 55, 343

(2001).22. M. M. Coleman, J. Graf, and P. C. Painter, Specific Interactions and

the Miscibility of Polymer Blends (Technomic, Lancaster, PA,1991).

23. J. Moskala, S. E. Howe, P. C. Painter, and M. M. Coleman, Mac-romolecules 17, 1671 (1984).

Documents

Application of Generalized Two-Dimensional Infrared Correlation Spectroscopy to the Study of a Hydrogen-Bonded Blend