European Polymer Journal 57 (2014) 47–57

Contents lists available at ScienceDirect

European Polymer Journal

journal homepage: www.elsevier .com/locate /europol j

The IR and Raman spectra of polyaniline adsorbed on the glasssurface; comparison of experimental, empirical force field,and quantum chemical results

http://dx.doi.org/10.1016/j.eurpolymj.2014.04.0230014-3057/� 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author. Tel.: +420 221 911 246; fax: +420 221 911 248.E-mail address: burda@karlov.mff.cuni.cz (J.V. Burda).

Jonáš Tokarsky a,b, Michal Maixner c, Pavlína Peikertová a, Lenka Kulhánková a,Jaroslav V. Burda c,⇑a Nanotechnology Centre, VŠB-Technical University of Ostrava, 17. listopadu 15/2172, 708 33 Ostrava-Poruba, Czech Republicb IT4Innovations Centre of Excellence, VŠB-Technical University of Ostrava, 17. listopadu 15, CZ-708 33 Ostrava-Poruba, Czech Republicc Department of Chemical Physics and Optics, Faculty of Mathematics and Physics, Charles University, Ke Karlovu 3, 121 16 Prague, Czech Republic

a r t i c l e i n f o

Article history:Received 20 October 2013Received in revised form 30 January 2014Accepted 29 April 2014Available online 9 May 2014

Keywords:PolyanilineIR and Raman spectroscopyQM calculationsMM simulations

a b s t r a c t

Vibrational spectra of the oligomeric models of polyaniline (PANI) were explored by quan-tum chemical and molecular mechanical tools. First, calibration calculations were per-formed on aniline sample where several computational models were compared withmeasured IR and Raman spectra. Based on this ‘calibration’, the xB97XD/6-31G(d,p) levelwas used for optimization and spectra determination of powder polymer and PANI depos-ited on the glass surface. From the individual models of PANI it was found that spectralshape converge relatively fast with the length of oligomeric chain and octamers can be con-sidered as the structures, which recover most of the spectral properties – both shape andintensities of individual peaks. As suggested by Stejskal, vibrational spectra provided bysaturated chains consisting purely from aniline building blocks do not correlate withexperimental values. However, the increasing amount of quinonic structures improvesthe agreement of computed spectra with experimental one substantially. The best corre-spondence occurs for the quinone:aniline ratio 1:3. This conclusion also follows from thefitted Raman spectrum calculated for the set of decameric structures – saturated andoxidized to three subsequent states (0, 2+, 4+, and 6+) where expansion coefficient ofthe 4+ state, which corresponds to two quinonic units clearly dominate.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction and optical properties, good redox reversibility and high

Recent development of conducting polymers (CP)represents one of the most vivid research disciplines andthese polymers are systematically studied for more than20 years [1–5]. They attract attention for many reasons.They can be relatively easily produced and can be usedas cheap microelectronics, sensors, markers or varioustools for analytical separations. PANI is nowadays the moststudied conducting polymer exhibiting unique electrical

environmental stability [6,7]. These features, together withthe low cost and wide availability of aniline and itsderivatives, make it an ideal candidate in many practicalapplications such as gas sensing [8–10], pH sensing [11],electrochromic devices [12,13], light-emitting diodes [14]or corrosion protection [15,16].

Nanocomposites of the CP/phyllosilicate type belong toone of the new fields in materials design, emerging in lastseveral years [17–20]. The phyllosilicates comprise veryuseful group of materials with a unique physico-chemicaland optical properties. Silicate layers are negativelycharged due to ionic substitutions at various active sites

48 J. Tokarsky et al. / European Polymer Journal 57 (2014) 47–57

in their structures. This layer charge is compensated bymetallic mono- or bivalent cations located on the surfaceor between individual layers. Also, the intercalation iscontrolled by the extent of total charge of the layer.

Recently published papers on CP/phyllosilicate nano-composites deal with various preparation techniquesusually in water solution [2–5] or by in situ deposition ofCP on phyllosilicate [21–24]. X-ray diffraction is commonlyused for the characterization of CP/phyllosilicate structuresor for verification of intercalation of the polymer chainsbetween the phyllosilicate layers. Electric conductivity istreated for measurement of the extent how much of thepolymeric material is intercalated into the phyllosilicate[2–5,17].

The most frequently studied form of PANI are thin lay-ers, which have been deposited on various supports: gold[25], quartz [26], zinc selenide [27], polyvinyl chloride[28], polyamide [29], and many others. For more informa-tions see, e.g. the review published by Stejskal et al. [30].

Besides the above-mentioned supports, many researchgroups used glass slides. The coating of glass surface witha thin PANI film was achieved by using various methods,e.g. Refs. [10,31]. The main reason for choosing glass as asubstrate (in addition to wide availability and chemicalstability) is its transparency, which allows researchers touse optical methods for the characterization and to changethe transmittance by changing the thickness of PANI film.

In this study we concentrate on the comparison of purepolyaniline (PANI) vibration spectra (infrared and Raman)both in powder state and as a thin film deposited on glassslide. Since the vibrational spectra can be very specificallyinfluenced by various surrounding the main goal of thisstudy is to explore how accurately this environmentaleffect can be described by computational techniques andwhether these techniques can be possibly used for predic-tion of some characteristic (specific) shifts. This introduc-tory study will be later extended by the relationshipsbetween the structure and electrical and optical propertieson the basis of both experimental and computational toolscombining UV–VIS and Raman spectroscopy with resultsobtained by computational methods for deeper insight intostructural information and determinations of electronicproperties of the ground and excited states.

2. Experimental and computational details

2.1. Preparation of the samples

Aniline, sulfuric acid, and ammonium peroxydisulfatewere purchased from Lach-Ner, Czech Republic, and usedas received.

PANI powder was prepared using simple oxidativechemical polymerization of the solution of aniline in con-centrated sulfuric acid by strong oxidizing agent at roomtemperature. The method is based on the mixing of 0.2 Maniline solution in 0.5 M sulfuric acid (cationic source),and 0.2 M solution of ammonium peroxodisulfate in dis-tilled water (oxidation agent). After the mixing of bothsolutions the polymerization of the aniline, signalized bythe dark green color of the reaction mixture, proceeds.

The polymerization time was 40 min. The green solid wascollected on a filter by rinsing with distilled water, stabi-lized by HCl solution and dried at 40 �C in a kiln.

The glass slides (76 mm � 26 mm � 1 mm) werewashed in a soap solution, rinsed with distilled water, fur-ther with ethanol and finally dried. In order to obtain onlyone-sided coating of glass slides with PANI film, one side ofthe slides was covered by a scotch tape. PANI films on suchtreated glass slides were prepared by in situ chemicalpolymerization of aniline by mixing two solutions. The firstone (cationic source) contained 0.2 M anilinium sulfateprepared in 0.5 M sulfuric acid, and the second one (oxidiz-ing agent) contained 0.1 M ammonium peroxydisulfatedissolved in distilled water at room temperature. Thepolymerization time was 20 min. Then, the glass slide withPANI film was stabilized by HCl solution, washed bydistilled water and dried at room temperature.

2.2. Measurement tools

Fourier transform infrared (FTIR) spectra were mea-sured by Nicolet 6700 FT-IR (Thermo Scientific, USA) byATR technique with diamond crystal and 256 scans.

Raman spectra were recorded by Smart SystemXploRA™ (Horiba Jobin Yvon, France). The 532 nm lasersource and grating 1200 gr./mm were used. Acquisitiontime of signal was chosen 120 s and number of accumula-tions was 50, to reduce signal/noise ratio.

Thermogravimetry analysis (TGA) was used to deter-mine the water content in real PANI samples. Measure-ment was carried out with NETZSCH STA 409 EP(Netzsch-Gerätebau GmbH, Germany). The sample (ca10 mg) was heated up to 1000 �C (speed 10 �C min�1) ina-Al2O3 crucible in dynamic air atmosphere with a flowof 100 cm3 min�1.

2.3. Models and strategy of modeling

First, a structure of the isolated aniline and short oligo-mer sequences of PANI, containing 2, 3, 4, 5, 7, 8, and 10monomer units were optimized. For revealing the effectof quinonic units on spectra, calculations of fully saturatedoligomers were performed first. Then, the oxidized oligo-mers were constructed by lowering number of electronsby two for one quinonic monomer at the quantum chemi-cal (QM) level as suggested by Stejskal [30]. He also sug-gested that the quinone:aniline ratio should be 1:3 asdepicted in Scheme 1.

For the simplicity dimer, trimer and tetramer of PANIwere considered to contain just one quinonic ring, penta-mer to octamer one or two and decamer up to three rings(with the total charge 6+). For the molecular modeling(MM) simulations, the investigation of these models wasrepeated for both saturated and unsaturated oligomericmodels of PANI. More over the both sets of models havebeen extended in order to investigate the effect of longerchains that cannot be treated by QM approach due to thetoo high computational demands. PANI chains containing2–5, 7, 8, 10, 12, 16, 20 and 32 monomer units wereprepared and optimized. On the contrary to QM calcula-tions, different strategy for construction of unsaturated

Scheme 1. Tetrameric quinolone–aniline unit used in MM calculations.

J. Tokarsky et al. / European Polymer Journal 57 (2014) 47–57 49

structures had to be used since MM simulations requirethe quinonic units to be localized in exactly given positionsof the PANI chain due to different mapping of force-fieldparameters for regular aniline and quinonic units. Dimer,trimer and tetramer were analogously considered withone quinonic ring, pentamer, heptamer and octamer withone or two quinonic rings. In the case of decamer, one,two or three rings were modeled. Higher oligomers:dodecamer, hexadecamer, icosamer and dotriacontamercontained three, four, five, and eight quinonic rings,respectively. Finally, in order to simulate the real sampleof PANI powder as closely as possible, the single-moleculeoligomeric model was extended to model containing fivedotriacontamers compensated by eighty Cl� anions and160 water molecules. The amount of water was chosenbased on the results of TGA that proved presence ofapprox. 14 wt.% of water in real samples as can be seenin Fig. 1. This model has been optimized under the periodicboundary conditions.

In the last step the influence of glass surface was con-sidered in calculations. An amorphous structure of theglass surface, which is available in many MM databases,is not suitable for small quantum chemical models due totoo large inhomogeneity. Therefore, a surface based onSiO2 twelve-membered rings with six Si and six O in theupper layer was constructed. Its size is comparable withthe size of aniline unit (cf. Fig. 2a). The oxygen freevalences were anchored by hydrogen atoms so that thewhole glass-pad was electroneutral. Since such a surfaceis very smooth it can be considered that the adhesion (ornegatively taken interaction) energy between PANI andglass is relatively low (underestimated) in this model since

Fig. 1. Thermal analysis of pur

there are not any active sites or ‘dislocations’ where theuncompensated valence of such active sites could formstronger interactions by chemical bonding. In this waythe influence on vibrational spectra can be reduced incomparison with real samples. The whole system was opti-mized for several different arrangements where the oligo-mer and glass were mutually shifted to various extents andpossibly slightly rotated. The energetically lowest struc-tural conformations were used for determination of theRaman and IR vibrational spectra.

Evaluation of this QM model was repeated at MM levelfor the same set of models and subsequently extendedwith longer chains (to dotriacontamer) lying on the ‘QM’regular and amorphous glass substrates. The lattersubstrate was of the same size for all models, i.e. approxi-mately five time wider than diameter of aniline unit(�3 nm) and slightly longer than dotriacontamer PANIchain (�23 nm). Finally, two larger models containingeither six icosamers or three dotriacontamers as a mono-layer has been studied.

The quantum chemical calculations were performed byGaussian 09 program package. Individual structures wereoptimized using the xB97XD/6-31G(d,p) model with sub-sequent diagonalization of Hessian matrix at the samelevel. In calculations of the vibrational modes for PANIdeposited on glass surface, atoms of the glass were keptfrozen in order not to mix with frequencies of the oligomermodels. For the visualization of IR and Raman spectra theMolden v. 5.0 program [32] was used.

In MM calculations, Materials Studio (MS) modelingenvironment/Forcite module was used for the geometryoptimization procedure with Universal force field [33].

e PANI powder sample.

Fig. 2. (a) Single twelve-member ring of regular glass; (b) regular glass surface for PANI trimer. Oxygen atoms are red, silicon atoms are beige, hydrogenatoms are white. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

50 J. Tokarsky et al. / European Polymer Journal 57 (2014) 47–57

Charges of PANI chains and SiO2 matrices are assigned byGasteiger and Marsili [34] and QEq method [35], respec-tively. Gast_polygraph 1.0 parameter set containingenhancements for the treatment of four-valent nitrogenand QEq_charged 1.1 parameter set suitable for calculationof silicate charges were used. In the case of molecular sim-ulations, vibrational frequencies were transformed to spec-tral lines with Lorentzian distribution functions with thesame broadening as used in Molden program (the half-width at half-maximum: c = 20 cm�1).

3. Results and discussion

3.1. Vibrational spectra of pure aniline

Before exploring such a complex system like oligomericchain deposited on a surface we need to know how accu-rate the computational tools are for related well-defined(small) molecules. Therefore, the monomer building blockof PANI was used for this calibration purpose.

As to quantum chemical approach, aniline was opti-mized with consequent spectra determination using sev-eral levels of calculations: MP2/6-31G(d,p) (hereafterlabeled as BS_I), MP2/6-311++G(2df,2pd) signed as BS_II,B3LYP/BS_I, B3LYP/BS_II, xB97XD/BS_I and xB97XD/BS_II.For comparison also semiempirical PM3 calculations andMM with Universal force field in MS/Discover module wereperformed. In the MM case only determination of IR spec-tra is available. Experimental spectra were recorded asmentioned above.

All the obtained spectra are collected graphically inFig. 3a and b for both IR and Raman methods, respectively.At a first glance it can be seen that semiempirical approachfails to predict any spectra even qualitatively, not only asto shifting individual peaks but also the shape of wholespectra is completely different. On the other hand theMM spectra are systematically shifted in comparison withthe both experimental and MP2 or DFT results.

From the closer inspection of IR spectra it can benoticed that DFT and MP2 results fairly agree with experi-mental spectra. However, in the area of hydrogen stretch-ing modes the calculated intensities deviates. Also thesplitting of individual hydrogen stretching modes of ben-zene ring is not so pronounced. Only two peaks can be dis-tinguished when half-width of Lorentzian distribution

used for peak broadening is lower than 10 (half-width = 20was used for broadening of spectral lines in Fig. 3). Never-theless, the ‘average’ position of this band correspondswell with measurements. Also, estimations of next threelower peaks, which belong to CAN and CAC stretching(or valence), and HANAH bending vibrations is fairly good(cf. Fig. 3a). In the range between 1100 and 800 cm�1 thecalculated intensities are again quite low in comparisonwith experimental data. From ca 750 cm�1 lower in fre-quencies, the out-of-plane hydrogen vibrations occur.Their intensities match well. While the experimentallyobserved peak at 747 cm�1 fits with calculated values nice(within 20 cm�1) the next lower peak at 689 cm�1 deviatesquite substantially (ca by 80 cm�1).

When the average match of all peaks in individualmethods (minimizing RMSD) is determined the bestperformance occurs for the B3LYP/BS_I level, with the leastsquare optimized scaling factor a = 0.962 and RMSD =328 cm�1, followed by B3LYP/BS_II (a = 0.965) andxB97XD/BS_I (a = 0.950). The alpha parameter for IR spec-tra with MM method is 1.04. All these values are collectedin Table 1.

The difference between single-molecule IR spectrumand weakly (LJ) interacting 8 molecules randomly opti-mized in one box is depicted in Fig. 3c. It can be seen thatall spectral lines are split due to van der Waals interactionsand for instance the shape of stretching modes of benzeneCAH vibrations is closer to experimental spectra.

Similar comparison was also conducted for Raman spec-tra. From Fig. 3b it follows that all the discussion for IR spec-tra can be repeated here including the order of the bestperformed methods and their optimal a scale parameter.The relative error of the predicted spectral line can be deter-mined from the RMSD value. In this way the accuracy is±20 cm�1 for both IR and Raman spectra while visibly worseaccuracy was obtained for MM simulations (ca ±30 cm�1).For further evaluation the xB97XD/BS_I level was usedsince oligomeric models of PANI are ‘attached’ to thesurface mainly due to the dispersion interaction, whichrequires using corresponding term at least qualitatively.

3.2. IR spectra of PANI powder

As specified in computational details two kinds of PANImodels were considered. Saturated oligomeric chains and

Fig. 3. (a) IR spectra of aniline at various computational levels (from bottom MP2/BS_I, MP2/BS_II, B3LYP/BS_I, B3LYP/BS_II, xB97XD/BS_I, and xB97XD/BS_II, PM3, MM); (b) Raman spectrum of aniline; (MM results are not available) (c) IR spectrum of one and eight molecules of aniline under Lennard–Jonesinteraction.

J. Tokarsky et al. / European Polymer Journal 57 (2014) 47–57 51

unsaturated (oxidized) chains, which were due to the prep-aration conditions [30] regarded in cationic form. Accord-ing to the Stejskal study [30], such a structural patterncorresponds to real sample when approximately everyfourth aniline building unit is replaced by oxidized quinon-ic structure. In the case of oligomeric cationic structures ofpartially oxidized linear chains, pronounced quinonic

character was really obtained after optimization at theDFT level even if the optimization started from regular(aniline) structure.

First, we concentrate on saturated models of polyaniline– dimer to decamer. Since experimental IR spectra wererecorded only in the range of 500–1800 cm�1, also thecomputational data will be considered in the same extent.

Table 1RMSD, and accuracy of the spectra lines (r), and optimized scalingparameter a for IR and Raman spectra.

IR Raman

RMSD r a RMSD r a

B3LYP/Opt 328 19.3 0.962 327 18.2 0.964B3LYP/SP 334 19.7 0.965 342 19.0 0.967MP2/Opt 412 24.3 0.946 351 19.5 0.947wB97/Opt 336 19.8 0.950 347 19.3 0.952wB97/SP 339 19.9 0.998 356 19.8 0.957MP2/SP 362 21.3 0.998 353 19.6 0.957MM 538 31.7 1.048

52 J. Tokarsky et al. / European Polymer Journal 57 (2014) 47–57

In Fig. 4a and b the effect of the chain length is presentedfor DFT and MM method, respectively. It is known alsofrom literature [36–39] that all the characteristic features

Fig. 4. (a) Calculated IR spectra of saturated PANI chains from dimer to decamerby MM tools; (c) influence of various oxidation states on IR spectra of PANI octammodels determined by DFT and IR spectrum obtained from MM model containingIR spectrum of PANI powder; (f) IR spectra obtained from MM calculations of stogether with two models of five octamers without (MM1) and with (MM2) waPANI powder.

from long oligomeric chains are already fully presentedin shorter oligomeric (ca. heptameric) structures especiallyfor chains containing ring building blocks. Only the inten-sity ratio of some peaks mildly changes. Spectra of longermodels (with 7, 8, and 10 units) have practically the sameshape. Comparing the DFT and MM approaches, some seri-ous differences basically in the whole spectral extent canbe noticed. For instance, two main peaks occur in both ser-ies at ca 1500 and 1250 cm�1 (CAC and CAN stretchingmodes). Nevertheless, their shape is different – intensityratio of twins at 1500 cm�1 is opposite and also the lowerpeak at 1250 cm�1 is much less pronounced using the MMmethod than according to DFT results. The most seriousdifferences concern the area between 500 and 1000 cm�1

where the most intensive peak in MM spectra occurs (at750 cm�1). The group of these peaks remains intensive also

using DFT methods; (b) spectra from dimer to dotriacontamer determineder determined by DFT and (d) MM calculations; (e) IR spectra of decameralso residual water and Cl� counterions in comparison with experimental

aturated and unsaturated octamers, hexadecamers, and dotriacontamerster and Cl� counterions in comparison with experimental IR spectrum of

J. Tokarsky et al. / European Polymer Journal 57 (2014) 47–57 53

in MM spectra of oxidized chains and since this frequencyrange in experimental spectra does not contain so inten-sive bands we can conclude that the vibrational spectraobtained by MM approach has to be taken with care.

Fig. 4c and d shows the changes of the spectra shapewhen the model of octameric chains is subsequently oxi-dized (containing 0, 1, and 2 quinonic units) determinedwith the DFT and MM method, respectively. Broadeningof stretching modes of CAC and CAN bonds due to differ-entiation of their bond order is clearly described by bothapproaches. Similar effect occurs also at lower frequenciesbut these peaks are much less pronounced at the DFT level.Fig. 4e displays similar behavior of increasing quinoniccharacter for the decameric chain. While the spectrum ofsaturated model contains four very intensive stretchingCAC and CAN vibration modes between 1295 and1515 cm�1 broadening to the range of 1185–1640 cm�1

occurs already for 2+ charged (partially oxidized) model.In the case of further oxidation of the chain, the more com-plex structure of peaks occurs especially in the area of1500–1650 cm�1 where quite broad band with many visi-ble peaks is remarkable (blue line) as a clear consequenceof quinonic structure present in the chain. This feature is inaccord with experimental data despite the fact that inmeasured samples residual water is present and intensivebending vibration modes partially mask the vibrationalmodes of PANI chains, cf. below. In the same plot alsoresults of molecular simulations are drawn (pink line).Despite of the visibly broader range of the stretching vibra-tions (1190–1725 cm�1) the qualitative description of themain peaks is similar. Worse correspondence of calculatedand experimental spectra can be seen below 1000 cm�1

where much lower intensities of isolated linear chainswere obtained.

Deficiencies like missing mutual interchain interac-tions, bending or turnings on the chains, counterion effects,trapped solvent molecules, etc. affect vibrational spectraand especially the weaker vibrational modes (under1000 cm�1). We tried to address such effects in the morecomplex model system using MM simulations in periodicboundary conditions with five octameric fully oxidized

Fig. 5. Comparison of Raman spectra determined for decamers (with various spowder sample.

chains (4+ charged structure with two quinonic units) neu-tralized by Cl� anions and water molecules as described incomputational details. In this box the originally linearchains bended and final density of the bulk was 1.18 g/cm3 in reasonable agreement with the density of real sam-ple. Comparison of various MM models is summarized inFig. 4f. The displayed MM spectra match relatively wellat higher frequencies (above 1250 cm�1). Neverthelesssimilarly to the DFT results, MM spectra do not exhibitany clear correspondence with the experimental peak at1033 cm�1. Interesting part in Fig. 4f represents the uppergreen spectrum, which correspond to five octamers in peri-odic box without (MM1) and with (MM2) water and coun-terions. Comparing both curves, the role of water is clearlyvisible – very intense peak at 1660 cm�1 mainly belongs towater bending mode. New very broad band is visiblebetween ca 500–700 cm�1. It is present also in the modelwithout solvent, which means that mutual interchaininteractions play important role in this part of spectrum.

3.3. Raman spectra of powder PANI

In the case of Raman spectra there is no possibility toemploy MM models with longer chains or systems withmore molecules, residual water and counterions. There-fore, only comparison of DFT calculated vibration modesof shorter chains with experimental spectra is possible.Similarly to IR spectrum the effect of elongated chainscan be observed to heptameric structures. For longerchains only some marginal increase of intensities occursbut the frequencies remain practically at the same value.The influence of the oxidation stage is demonstrated ondecameric model of PANI in Fig. 5. Obviously, the spectrumof saturated system fails in comparison with measuredones. Nevertheless, except of peak at 1600 cm�1 all theother have some corresponding intensities also withinthe spectra of oxidized models. More over the peak at1600 cm�1 seems to be clearly visible also in measuredspectrum as the most right peak. As to oxidized chainswith two or three quinonic units it can be stated that theymatch nearly perfectly with experimental values above

tage of chain oxidation) at DFT level and experimental data of the PANI

54 J. Tokarsky et al. / European Polymer Journal 57 (2014) 47–57

1000 cm�1 (blue and pink lines). On the other hand the cal-culated vibration modes between 600 and 660 cm�1 do notfit with experimental spectrum so nicely. The explanationis similar as already mentioned in discussion of IR spectra:lack of intermolecular interactions and single linear struc-ture. These vibrations are more sensitive to weak interac-tion. The valence CAC and CAN stretching modes arerelatively much more pronounced in Raman spectrum thanin IR spectrum where out-of-plane wagging vibrations aremore intensive than in Raman spectrum.

3.4. Prediction of IR spectra on the glass slide

Experimental IR spectra are available only for PANI inpowder form since the reflection of the glass pad doesnot allow spectra recording when polymer is depositedon the glass surface. Despite the experimental results forIR spectra are not available for PANI samples on the glasssurface we performed also this kind of calculations notonly for predicative power of the theoretical tool but alsofor the confirmation that our model of glassy surface isplausible. Due to high demands of DFT or electronic calcu-lations generally we could not use large model of amor-phous glass. Therefore we used much smaller regularsilicate surface layer. In this section we try to justify sucha model of glass surface against amorphous glass sincethe IR spectra of PANI on the both surfaces can be com-pared at the MM level. In this way we can estimate theinfluence of the type of glass surface on the frequenciesand intensities of the calculated spectral lines. Similarlyto the discussion in the section on the powder PANI, wecompared the IR spectra obtained for shorter PANI chainson the regular silicate surface at the DFT and MM levels.From Fig. 6 it follows that basically the same spectral lineswere obtained for both models of glass. The only visibledifference occurs at ca 500–700 cm�1 area. While rela-tively broad range of peaks (about three peaks with similarintensity) can be observed for the regular surface in amor-phous surface very intense spectra line at 710 cm�1 domi-nate this area. Nevertheless, remaining two peaks can bealso traced in this spectrum. In comparison of MM and

Fig. 6. Comparison of IR spectra for two types of glass surface models interactingmodel both DFT and MM methods were used. For amorphous glass (ag) surface

DFT spectra, analogous conclusions like in previous partdevoted to the IR spectra on powder PANI can be drawn.The MM spectra are red-shifted in the 1500–1750 cm�1

range in comparison to the DFT and experimental spectraand they are also less intensive. On the contrary in500–1000 cm�1, the intensities of the MM peaks are visiblyoverestimated. Nevertheless, the agreement between bothcomputational tools (DFT and MM) is at least partiallyestablished.

On the contrary to previous spectral simulations inpowder PANI, a comparison of octamer and decamerunsaturated models do not agree so well for the oxidationto higher degree (4+, 6+) between 1500 and 1600 cm�1

where intensities of the 1520 cm�1 and 1560 cm�1 peaksobtained for octameric spectra have inverted intensitiesthan in the case of decameric models. They are also simul-taneously shifted by ca. 15 cm�1. On the other side satu-rated and partially oxidized structures with one quinonicunit (i.e. octamer 2+ and decamer 2+) can be consideredas already converged (analogous shape of dark blue andviolet curves) as can be seen from Fig. 7. These curves ofpartially oxidized oligomeric chains exhibit more or lessanalogous spectral characteristics like fully oxidized mod-els in the range 1000–1500 cm�1. Nevertheless at higherfrequencies both curves deviate from spectrum of fully oxi-dized chain. This fact can be easily understood from thecharacter of the peaks in this area due to the increasingquinonic character of the oxidized oligomeric chains.

3.5. Raman spectra of PANI on the glass slide

In the final part we concentrate on the comparison ofmeasured and calculated Raman spectra on the surface ofthe glass slide. As it was already mentioned above onlyQM calculations are available for this kind of spectra. InFig. 8 results for unsaturated models are drawn. At the firstglance it can be noticed that most of the predicted spectrallines very closely correspond with experiment data whenoxidized forms of oligomers: decamer 4+ (upper blue line)and decamer 6+ (green line), are considered. In these casespractically all the measured peaks have their analogues

with octameric (fully oxidized) PANI chain. In the case of regular glass (rg)only MM method was applied.

Fig. 7. The IR spectra obtained from DFT models of oxidized PANI chains (from dimer to decamer) on regular glass (rg) surface.

Fig. 8. Comparison of experimental Raman spectrum measured after 20 min of polymerization and Raman spectra determined by DFT calculations foroxidized PANI chains (dimer to decamer) on the regular glass surface.

J. Tokarsky et al. / European Polymer Journal 57 (2014) 47–57 55

with similar intensity also in predicted spectra (after scal-ing by a=0.94). On the contrary to results obtained forPANI powder samples, in the case of glass slides signifi-cantly better correspondence with measurement existsfor decamer 4+ model than decamer 6+ model. Also modelof partially oxidized decamer 2+ match reasonably, at leastat the range of valence CAC and CAN stretching. Similarlywith calculations for PANI powder models the resultsobtained also for octameric chains (lower green/yellowline) are in acceptable agreement with experimentallyobtained Raman spectra. Similarly to decameric models,partially oxidized structure (with one quinonic unit)reproduce experimental peaks better both for higherfrequencies – where in octamer 4+ structure (green line)the peak at ca 1600 cm�1 is missing as well as for lowerfrequencies (the out-of-plane wagging modes). In analogyto IR spectra in Fig. 7, for the higher oxidations of the PANIchains (4+ and 6+) the chain length does not seem to befully converged yet.

Since spectra of several oxidized PANI structures weredetermined we can ask, which one is the closest and

eventually which combination (or concentration ration)of the individual PANI chains gives the best agreementwith measured spectrum. The fitted spectrum resultedfrom saturated and all three oxidized decameric structuresis compared with experimental spectrum in Fig. 9 and theoptimal coefficients are c(0) = 0.12, c(2) = 0.02, c(4) = 0.54,and (6) = 0.31. More over frequency scaling was left as fifthvariational parameter. Its optimal value is 0.940, which isin nice match with results analyzed for pure aniline sam-ples in previous part of the study. From the expansion coef-ficients it clearly follows the closest resemblance of thespectrum obtained for decamer 4+ oxidized structure tothe real PANI sample. Nevertheless, some portion ofdecamer 6+ ‘admixture’ improves the spectral shape visi-bly. This also confirms Stejskal’s conclusion on structuralfeatures of the PANI samples that every fourth unit isoxidized.

Finally the difference between spectra of the powderPANI and PANI deposited on the glass surface should bementioned. This comparison is drawn in Fig. 10 where bothexperimental and predicted Raman spectra are displayed.

Fig. 9. Comparison of experimental Raman spectra with scaled (0.943)fitted combination of all four decameric structures (saturated and threeoxidized chains whose spectra are placed in lower part).

Fig. 10. Raman spectra of +4 oxidized models scaled by 0.96 factor (solidlines) and experimental values of PANI (dashed lines) in powder form(lower pair – red) and deposited on the glass surface (upper pair – blue).(For interpretation of the references to color in this figure legend, thereader is referred to the web version of this article.)

56 J. Tokarsky et al. / European Polymer Journal 57 (2014) 47–57

From the solid lines (computed spectra) it can be noticedtwo visibly different ranges. The first and probably moreimportant is difference around 1400 cm�1. Here, invertedintensities of valence vibrations at 1390 and 1440 areapparent. The other range concerns to out-of-plane vibra-tions around 500 cm�1. Analogous changes can be noticedalso in experimental spectra. However, the differences invalence area are visibly more pronounced and the influ-ence of surface play much more important role than in cal-culated spectra. We can guess that this is probably theconsequence of the glass model used in the DFT approach.Similar differences can follow from the comparison of MMmethod with experimental IR spectra in Fig 6. It should bestressed that predicted spectra contain mainly features of4+ model and as follow from Fig. 9 the admixture of theother (lower and higher oxidized) chains visibly improvesthe similarity (shape) of experimental and calculatedspectra.

4. Conclusions

In the current contribution we studied the vibrationalspectra of various oligomeric models of PANI in powderform and deposited on glass surface. For these systemsexperimental spectra were also measured and used forcomparison.

At first the accuracy of spectra determination wastested for several quantum mechanical and molecular

mechanical methods on the ‘small’ and well defined mod-eling system – aniline, a basic building unit of longer olig-omeric PANI models. In this part it was found that bothDFT functional and MP2 method give reasonable agree-ment with experimental IR and Raman spectra of aniline.RMSD of spectral lines was about ±20 cm�1 after applyingscaling factor (usually 0.95–0.97). Based on these calcula-tions the xB97XD/6-31G(d,p) level of theory was chosenas a compromise between accuracy and computationaldemands. The choice of this functional was also affectedby the fact that dispersion interaction between PANI andglass surface has to be taken into consideration. From thespectral shape it can be clearly seen that semiempiricalmethods do not give sufficient accuracy of the predictedvibrational transitions.

In the second part of the study vibrational spectra ofPANI powder were investigated. Calculations were dividedinto several parts dealing with saturated and oxidizedoligomeric models in order to distinguish various featuresconnected with effect of the chain length. In the case of sat-urated models, the shape of spectral lines converges rela-tively very fast with the length of oligomeric chain. As tooxidized models it was found that partially unsaturatedstructures (with less than every fourth quinonic units inchain) do not provide an appropriate vibrational spectra.MM extension to longer PANI chains (up to dotriacont-amer) did not basically change the spectral shape, inaccord with previous QM calculations. Also, the accuracyof predicted transitions is similar to previously obtainedaniline spectra. However, considering the model withseveral PANI chains (with counteranions and watermolecules) using the MM method led to deterioration ofthe calculated spectra, especially in the range of 500–750 cm�1, where the role of mutual interaction is clearlyoverestimated. On the other hand the effect of water bend-ing mode dominates over PANI C@C vibrational modes ofbenzene/quinone rings in agreement with experimentalresults.

In the final part, the spectra of PANI deposited on glassslide were examined. In QM calculations regular type ofglass surface was employed due to the size of the exploredsystems. In MM calculations both regular and amorphousglass models were studied. It was found that the type ofglass surface affects the obtained shape of IR spectra onlymarginally (basically only at 750 cm�1 in non-scaledspectrum). From comparison of QM calculations ondecameric chains with experimental Raman spectrum itcan be seen fair agreement especially for decameric 4+oxidized structure. When spectra fitting of all the formsof decamer to the experimental shape was performed thedominating character of 4+ oxidized structure wasconfirmed. This can be also considered as a confirmationof Stejskal’s conclusion on structural pattern of PANIchains.

Comparing the frequencies and intensities of individualvibrational modes of PANI in powder form and as a thinlayer deposited on glass surface it can be concluded thatdespite some minor changes of peaks in lower frequenciesthe main difference can be assumed around 1400 cm�1.This is due to influence of valence CAC and CAN vibrationmodes of PANI chains by the glass surface.

J. Tokarsky et al. / European Polymer Journal 57 (2014) 47–57 57

Acknowledgements

Authors are grateful to GACR Project No. P108/11/1057for supporting this study. Authors wish to thank toDr. Vallová for TGA measurement. Also, the access to theMetaCentrum computing and storage facilities (GrantLM2010005) is highly appreciated.

References

[1] Gawryla MD, Liu L, Grunlan JC, Schiraldi DA. Macromol RapidCommun 2009;30:1669–75.

[2] Irfan A, Manwar H, Keum-Suk S, Young-Ho C. J Appl Polym Sci2010;116:314.

[3] Lee D, Char K, Lee SW, Park YW. J Mater Chem 2003;13:2942.[4] Yoshimoto S, Ohashi F, Ohnishi Y, Nonami T. Synth Met

2004;145:265.[5] Vilela SO, Soto-Oviedo M, Fonseca Albers AP, Faez R. Mater Res

2007;10:297–300.[6] MacDiarmid AD. Synth Met 2002;125:11–22.[7] Sapurina I, Stejskal J. Polym Int 2008;57:1295–325.[8] Agbor NE, Cresswell JP, Petty MC, Monkman AP. Sens Actuators

1997;B 41:137–41.[9] Janata J, Josowicz M. J Solid State Electrochem 2009;13:41–9.

[10] Menegazzo N, Herbert B, Banerji S, Booksh KS. Discourse on theutilization of polyaniline coatings for surface plasmon resonancesensing of ammonia vapor. Talanta 2011;85:1369–75.

[11] Ayad MM, Salahuddin NA, Alghaysh MO, Issa RM. Curr Appl Phys2010;10:235–40.

[12] Kobayashi T, Yonevama N, Tamura H. J Electroanal Chem1984;177:281–91.

[13] Batich CD, Laitinen HA, Zhou HC. J Electrochem Soc 1990;137:883–5.[14] Karg S, Scott JC, Salem JR, Angelopoulos M. Synth Met

1996;80:111–7.[15] Gonçalves GS, Baldissera AF, Rodrigues Jr LF, Martini EMA, Ferreira

CA. Synth Met 2011;161:313–23.[16] Lu WK, Elsenbaumer RL, Wessling B. Synth Met 1995;71:2163–6.[17] Thakur AK, Pradhan DK, Samantaray BK, Choudhary RNP. J Power

Sources 2006;159:272.[18] Mohapatra SR, Thakur AK, Choudbary RNP. Indian J Eng Mater Sci

2008;15:347.[19] Kurian M, Galvin ME, Trapa PE, Sadoway DR, Mayes AM. Electrochim

Acta 2006;50:2125.[20] Yeh JM, Liou SJ, Lai CY, Wu C. Chem Mater 2001;13:1131.[21] Bae WJ, Kim KH, Jo WH, Park YH. Macromolecules 2004;37:9850.[22] Nascimento GM, Constantino VRL, Temperini MLA. Macromolecules

2002;35:7535.

[23] Soundararajah QY, Karunaratne BSB, Rajapakse RMG.MaterChemPhys 2009;113:850.

[24] Yoshimoto S, Ohashi F, Kameyama T. J Polym Sci B: Polym Phys2005;43:2705.

[25] Liu CJ, Hayashi K, Toko K. A novel formation process of polyanilinemicro-/nanofiber network on solid substrates. Synth Met2009;159:1077–81.

[26] Ayad MM, Torad NL. Alcohol vapours sensor based on thinpolyaniline salt film and quartz crystal microbalance. Talanta2009;78:1280–5.

[27] Sedenková I, Trchová M, Blinova NV, Stejskal J. In situ polymerizedpolyaniline films. Preparation in solutions of hydrochloric, sulfuric,or phosphoric acid. Thin Solid Films 2006;515:1640–6.

[28] Shishkanova TV, Matejka P, Král P, Šedenková I, Trchová M, Stejskal J.Optimization of the thickness of a conducting polymer, polyaniline,deposited on the surface of poly(vinyl chloride) membranes: a newway to improve potentiometric response. Anal Chim Acta2008;624:238–46.

[29] Nasybulin E, Menshikova I, Levon K. Preparation of conductivepolyaniline/Nylon 6 composite films by polymerization of aniline inNylon-6 matrix. J Appl Polym Sci 2009;114:1643–7.

[30] Stejskal J, Sapurina I, Trchova M. Polyaniline nanostructures and therole of aniline oligomers in their formation. Prog Polymer Sci2010;25:1420–81.

[31] Li ZF, Ruckenstein E. Conductive surface via graft polymerization ofaniline on a modified glass surface. Synth Met 2002;129:73–83.

[32] Schaftenaar G. Molden. <http://wwwcmbikunnl/~schaft/molden/moldenhtml. 3.9 ed>.

[33] Rappé AK, Casewit CJ, Colwell KS, Goddard III WA, Skiff WM. UFF, afull periodic table force field for molecular mechanics and moleculardynamics simulation. J Am Chem Soc 1992;114:10024–35.

[34] Gasteiger J, Marsili M. Iterative partial equalization of orbitalelectronegativity: a rapid access to atomic charges. Tetrahedron1980;36:3219.

[35] Rappé AK, Goddard III WA. Charge equilibration for moleculardynamics simulations. J Phys Chem 1991;95:3358.

[36] Huang D, Simon SL, McKenna GB. Chain length dependence of thethermodynamic properties of linear and cyclic alkanes andpolymers. J Chem Phys 2005;122:084907.

[37] Szleifer I, Panagiotopoulos AZ. Chain length and density dependenceof the chemical potential of lattice polymers. J Chem Phys1992;97:6666–73.

[38] Weniger EJ, Kirtman B. Extrapolation methods for improving theconvergence of oligomer calculations to the infinite chain limit ofquasi-one-imensional stereoregular polymers. Comput Math Appl2003;45:189–215.

[39] Cornil J, Beljonne D, Calbert J-P, Brédas J-L. Interchain interactions inorganic Pi-conjugated materials:impact on electronic structure,optical response, and charge transport. Synth Met 2002;125:107–16.