Colloids and Surfaces

A: Physicochemical and Engineering Aspects 187–188 (2001) 509–521

Relaxation of polymers in 2 nm slit-pores: confinementinduced segmental dynamics and suppression of the glass

transition

E. Manias a,*, V. Kuppa a, D.-K. Yang b, D.B. Zax b

a Materials Science and Engineering department, The Pennsyl�ania State Uni�ersity, 325-D Steidle Building, Uni�ersity Park,PA 16802, USA

b Department of Chemistry and Chemical Biology, Cornell Uni�ersity, Ithaca, NY, USA

Abstract

Molecular Dynamics (MD) simulations are used to explore the structure and dynamics of polystyrene confined in2 nm slit pores, between parallel, crystalline, mica-type surfaces. The systems simulated resemble experimentallystudied intercalated nanocomposites, where polystyrene is inserted between layered-silicate layers. The molecularmodeling perspective complements the experimental findings and provides insight into the nature of polymers innanoscopic confinements, especially into the molecular origins of their macroscopic behavior. Namely, a comparisonbetween simulation and NMR studies shows a coexistence of extremely faster and much slower segmental motionsthan the ones found in the corresponding bulk polymer at the same temperature. The origins of these dynamicalinhomogeneities are traced to the confinement induced density modulations inside the 2 nm slits. Fast relaxing phenyland backbone moieties are found in low density regions across the film, and preferentially in the center, whereas slowrelaxing moieties are concentrated in denser regions in the immediate vicinity of the confining surfaces. At the sametime, the temperature dependence of the segmental relaxations suggests that the glass transition is suppressed insidethe confined films, an observation confirmed by scanning calorimetry. © 2001 Elsevier Science B.V. All rightsreserved.

Keywords: Dynamics; Glass transition; Calorimetry

www.elsevier.nl/locate/colsurfa

1. Introduction

Polymers and simple fluids in nanoscopic confi-nements possess a variety of unusual properties,and in particular remarkable dynamical hetero-

geneities which vary on length scales as short as afew Angstroms [1–13]. While the Surface ForcesApparatus (SFA) provides an experimental probeof macroscopic properties of a single fluid filmbetween two atomically-smooth mica surfaces[14–16], few experimental probes are availablewhich probe the microscopic origins of these het-erogeneities in nanoscopically confined films.However, ultra-thin (�2–5 nm) polymer films

* Corresponding author. Tel.: +1-814-8632980; fax: +1-814-8652917.

E-mail address: manias@psu.edu (E. Manias).

0927-7757/01/$ - see front matter © 2001 Elsevier Science B.V. All rights reserved.

PII: S0927 -7757 (01 )00634 -3

E. Manias et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 187–188 (2001) 509–521510

confined between solid surfaces can be created byintercalating high molecular weight polymers inbetween allumino-phyllosilicates (mica-type lay-ered crystals). Especially where appropriatelymodified silicates are used to form intercalatednanocomposites very homogeneous ultra-thinconfined polymer structures self-assemblethroughout the composite [17,18]. In such cases,careful choice of the polymer/inorganic stoi-chiometry can provide macroscopic quantities ofself-assembled multilayer systems, which consistexclusively of identical nanoconfined films. Thisoffers great opportunity to study the properties ofpolymers and liquids in extreme confinements (1–2 nm) by employing conventional analytical tech-niques [17].

For this reason, intercalated polymer/mica-typesilicate nanocomposites have been recently inves-tigated with the emphasis on the fundamentals ofnanoscopically confined polymers [19,20]. It hasbeen found that the nanoscopic confinement be-tween solid surfaces affects the polymer behaviorin ways that are not intuitively expected. Addi-tionally, in many cases this confinement-modifiedpolymeric structure and dynamics have beenshown to dictate the materials response at themacroscopic level [17], causing remarkable en-hancements of the materials properties [18].

Here, we report a molecular modeling ap-proach, combined with NMR experimental find-ings, aiming in obtaining insight into the

confinement induced dynamical processes of ul-tra-thin polystyrene films in 2 nm slit pores.Studying the formation, structure and dynamicsof these intercalated structures can lead to a betterunderstanding of polymers in a confined environ-ment or at a solid interface, and at the same timeprovide the necessary fundamental level knowl-edge towards the molecular design of polymer/in-organic hybrid materials with desirable properties.

2. Simulation model and details

Inspired by striking new experimental findingsof polymer films in 2 nm slit pores between alkyl-ammonium modified mica-type surfaces [19,20],we performed extensive Molecular Dynamics(MD) computer simulations of polystyrene (PS) in2 nm slits. Following we will mention some of theessential details of the simulations, more in [21].

The simulation geometry resembles closely thesystems studied by NMR [20]. Our simulationboxes are 4.224×3.656×3.05 and 2.112×1.828×3.05 nm in dimensions, with periodicboundary conditions in all three directions (Fig.1). A solid surface with the crystal structure of asynthetic 2:1 mica-type silicate is placed parallelto the xy plane, thus confining the organicmolecules in the z direction. The confinement size(3.05 nm) was selected to correspond to the exper-imentally measured X-ray diffraction d-spacing

Fig. 1. The system under investigation. From top left clockwise: a bright-field TEM detail of the PS/octadecyl-ammonium silicateintercalate; a snapshot of the simulation box; a styrene 12 monomers (the smaller of the two PS molecules simulated) and theoctadecyl-ammonium surfactant molecule.

E. Manias et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 187–188 (2001) 509–521511

Table 1The force-field parameters used in the MD simulations; bonded and non-bonded interactions are given for the organic species in thesystems

Non-bonded interactions V(rij)=4� [(�/rij)12−(�/rij)

6]+(1/4��)qiqj/rij

� (kJ mol−1)� (nm) q (e)

0.35190.3207 0.000CCph 0.3550 0.2940 0.115

0.29400.3550 0.000Cpha

0.2318H 0.3182 0.0000.1260Hph −0.1150.24200.8767 0.0000.2976N

Bond-stretching potentials V(r)= (kb/2)(r−bo)2

kb (kJ mol−1)bo (nm)

C�C 0.153 334720418400Cph�C 0.1514184000.139Cph�Cph

0.133C�N 376560292880C�H 0.1092928800.108Cph�H

0.100N�H 374468

Bond-angle potentials V(�)= (k�/2)(�ijh−�o)2

k� (kJ per (mol rad2))�o (°)

H�C�H 306.40109.45366.90109.45C�C�H

109.45Cph�C�H 366.90482.30C�C�C 109.45482.30109.45C�C�Cph

120.00C�Cph�Cph 376.60376.60Cph�Cph�Cph 120.00418.80120.00Cph�Cph�H

109.45H�N�H 306.40366.90N�C�H 109.45482.30109.45C�C�N

Vijkl(�)=k� [1+cos (n�−�o)], cis at 0° and ��(ijk)�( jkl)Proper dihedrals

k� (kJ mol−1) N�o (°)

5.88C�C�C�C 30.05.88C�C�C�N 30.05.880.0 3C�C�C�Hb

C�C�N�Hb 0.0 5.88 3

Planar (improper) Vijkl(�)=k�(�−�o)2, i central Cph and ��(ijk)�( jkl)

� (°) k� (kJ mol−1)

167.36Cph�Cph�Cph�Cph 0.0Cph

i �Cph�Cph�Hph (on i ) 0.0 167.360.0Cph

i �Cph�Cph�C (on i ) 167.36

a The phenyl carbon connected to the backbone methine.b The hydrogens of the chain-end methyls and ammoniums.

E. Manias et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 187–188 (2001) 509–521 512

Fig. 2. The two simulation boxes used in this study: with dimensions 2.112×1.828×3.05 nm (a), and 4.224×3.656×3.05 nm (b).For all the results discussed herein (density profiles, segmental dynamics, and density/dynamics covariances) there were nodifferences between the two systems.

[17] and, after subtracting the wall thickness, itallows for a 2.05 nm thin film of organics betweenthe two solid surfaces. The MD runs were carriedout in an NVT ensemble [22], where the numberof particles (N) is chosen based on the experimen-tally measured material amounts. The graftingdensity of the surfactant chains (which are octade-cyl-ammonium molecules) is well-known andcharacteristic of the silicate surface [23]. Ourgrafting density (�graft=1/0.97 nm2) correspondsto fluorohectorite [23], which is the silicate used inthe NMR studies, and dictates 32 (eight for thesmall box) octadecyl-ammonium cations over thetwo solid surfaces. Furthermore, from ther-mogravimetric analysis (TGA), the polymer tosurfactant weight ratio can be measured [17], andnecessitates for 144 (36 for the smaller box) sty-rene monomers in the slit. Two systems wereextensively studied, the first with styrene chains of12 monomers each, and the second with chains of24 monomers (in this case, the small box size is2.112×3.656×3.05 nm). Although these chainsmay be too short to capture the genuine responseof long PS macromolecules, they are both longerthan the average physisorbed train size of PS inthese systems, and also allow for the existence ofbridges between the two solid surfaces. Conse-

quently, all the possible segmental dynamics thatcan develop inside these slit pores can be exploredby our MD. Moreover, comparing the two sys-tems with different PS chain lengths does notshow any differences in the segmental dynamicsdiscussed herein, albeit the substantially longerchain relaxation times of the confined 24monomers compared with the 12 monomers.

One aspect that was scrutinized was the size ofthe box, especially in connection with the segmen-tal dynamics correlation lengths in these systems.Comparison between two systems (a smaller ver-sus a larger, Fig. 2) did not reveal any differencesin the segmental dynamics, and in the other re-sults discussed herein (density profiles, correla-tions of segmental dynamics, and covariancebetween local density and segmental dynamics);this is why we decided to use the (computationallyless intensive) smaller box for the longer timessimulated, and compare against shorter runs per-formed for the larger box. Another more subtleeffect of the finite size is the pbc artificial correla-tions between fast-relaxing moieties that are inclose spatial proximity (�1/2 box size) acrossperiodic boundary conditions. In our systems, wehave checked extensively for such correlations andwe found none. The reason is that such fast-relax-

E. Manias et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 187–188 (2001) 509–521513

ing elements (e.g. fast-phenyls) are separated byslow-relaxing moieties (e.g. kinetically hinderedPS segments), which are effectively immobilizedfor the duration of each productive MD run.These ‘immobile’ molecule segments shield againstany density or dynamical correlations betweenspecies that are on the opposite sides of them.

The force-fields for polystyrene, and its interac-tions with the alkyl-ammoniums (tabulated inTable 1), are the same as recent PS simulations byMuller-Plathe [24]. The interactions between thesilicate and the organic molecules (styrene andaliphatic chains) is based on our earlier simulationwork investigating the structure of alkyl-ammo-nium modified silicates [23]. The silicate partialcharges, crystal structure and silicate-cation inter-actions were developed by Chan and Panagioto-poulos [25], and were proven to reproducefaithfully the hydration behavior of these layeredinorganic crystals.

The solid surface atoms were constrained totheir equilibrium positions via harmonic springs,which allowed for thermal vibration. The temper-ature of the surfactant and the polymer was stabi-lized via a weak coupling to a heat bath by theBerendsen thermostat [26]. Neighbor lists wereused for the LJ interactions, and a particle–parti-cle/particle–mesh scheme with Ewald sums forthe coulombic interactions. Since the silicate sur-face is negatively charged (for fluorohectorite −1qe/0.97 nm2), the positive ammoniums have astrong Coulombic attraction to the surfaces; onlyonce, and for the highest temperature studied (470K), did we observe one of the ammoniums tomomentarily desorb from the surface and re-phy-sisorb on a different position on the surface.Thus, for all practical purposes, one can think ofthe octadecyl-ammoniums as end-tethered on thesurfaces.

Probably the most crucial and important aspectof these simulations is the way we selected tosample the configuration phase-space. A simpleConfigurational Biased Monte Carlo (CBMC)scheme was developed to produce intercalatedalkane and alkane surfactant configurations [27].For the polystyrene systems, however, — wherethe phenyl sidechains prohibit the use of a simpleCBMC approach — a more elaborate route was

necessary. The rotational-isomeric-state (RIS)model was used to create initial polymer confor-mations of PS [28]. Conformations that fit in theinterlayer gallery are chosen, and equilibrated byan off-lattice Monte Carlo scheme that employsrandom displacements of the backbone atoms andOrientational Biased Monte Carlo rotations of thephenyl rings; the surfactants were equilibrated byCBMC attempts in coexistence with the polymerchains. It is critical to have many independent,well-equilibrated initial configurations of the sys-tems to be simulated. As the polymer is heldtogether by strong covalent bonds, the relaxationof the chains is determined by the slowest movingsegments along the polymer. Even though mobilesegments are expected to exist in large numbers[20], they will be bonded to portions of the poly-mer that remain ‘frozen’ for timescales vastlylonger than what is accessible through MD.Therefore, the only realistic way to explore a largeportion of the configurational space is to startwith many independent initial system conforma-tions and perform productive MD simulations ofa few (10–100) ns. Typically, we used ten inde-pendent initial configurations for each tempera-ture and we followed each system for 100 ns atT=350 K, 50 ns at T=370 and 390 K, and 10 nsat higher temperatures.

We feel that this computer simulation approachimproves our insight in the molecular processesthat underlie the striking behavior of the polymer/silicate nanocomposites, and thus can further ourmolecular level understanding of the nature ofpolymers in nanometer confinements in general[17]. From this standpoint, we will proceed todescribe the behavior of the confined PS systemsin a comparative way between simulation andexperimental findings.

3. Polystyrene in 2 nm thin slit pores

3.1. Structure of the confined film

In Fig. 3, the density profiles across the slit poreare shown for different components of theconfined organic film. These density profiles cor-respond to the probability of finding the corre-

E. Manias et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 187–188 (2001) 509–521 514

Fig. 3. Density profiles versus distance normal to the confining surfaces. Partial densities across the slit pore are shown,corresponding to the surfactant carbons and nitrogens (dashed line), the styrene phenyl carbons (light continuous line), styrenemethyl, methylene and methine carbons (heavy continuous line), and the total organic density (dotted line).

sponding moieties at a certain distance away fromthe confining solid surfaces, and are calculatedthrough NVT ensemble averages of productive MDruns starting from distinctly different initial systemconfigurations. The usage of several — typicallyten — different initial configurations for eachsimulated temperature is essential, because therelaxation times of both the surfactant moleculesand styrene chains reach far beyond the 10–100 nstime periods simulated in each MD run. The samedensity profiles can also be calculated based solelyon the Monte Carlo scheme used to create theinitial systems for the MD, and are the same withthe MD profiles within the data accuracy. Thisagreement between the MC and the MD ensemblesis an indication that our MD’s actually sample arepresentative portion of phase-space.

From Fig. 3, it is obvious that the mass isdistributed across the slit pore in a highly inhomo-geneous layered fashion. This kind of layering isseen in all simulations of confined systems, rangingfrom small molecule simple liquids, to alkaneoligomers, to grafted alkanes, to generic LennardJones chains and realistic polymers. This layeringnormal to confining surfaces is attributed to the

steric interactions between the walls and theconfined fluids, and is macroscopically manifestedin SFA experiments by wildly oscillating solvationforces when confined fluids are squeezed betweenmica surfaces [1–9]. Finally, they are so robust thatthey remain unaffected even under severe shearflows [11].

What is new about our systems, is that there existseveral species across the pore (such as surfactantammonium and aliphatic groups, as well asaliphatic and aromatic styrene groups) and a pref-erential layering with respect to the walls is devel-oped. The positively charged ammoniums arefound in direct contact with the silicate walls —even at the highest temperatures studied —whereas most of the aliphatic part of the surfactantis preferentially located at the center of the slit (Fig.3). At the same time, we find that the polymerbackbone is predominately located near the confi-ning walls, and at distances of about 0.3–0.7 nmfrom the solid surface, but more removed, onaverage, than most of the phenyl sidechains. Often,a short strand of polymer adopts a bridgingconfiguration perpendicular to, and connectingbetween, both walls. The phenyl rings, which attach

E. Manias et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 187–188 (2001) 509–521 515

to the backbone at each methine, are similarlyfound on either side of the backbone, and dominatethe organic matter nearest the silicate surface.Finally, while surfactant methylenes are founddistributed all across the slit pore, they bunchpreferentially in the middle layer. These layereddensity profiles provide some clues as to the originof the confined film structure. The polar phenylgroups that interact strongly with the silicate sur-face are preferentially located closest to the surface,whereas the aliphatic groups are displaced towardsthe center of the pore. Obviously, due to thecovalent bonds between the styrene phenyls andmethines, the backbone cannot be displaced com-pletely in the center of the pore. Similarly, as thehead group of the surfactant is tethered via an ionicinteraction to the negatively charged silicate, thesurfactant cannot be completely displaced from thesurface either. As a result, only its aliphatic tails aredisplaced by the PS in the center of the slit(Fig. 3).

At first glance, all this detail may seem tootheoretical. However, the surface sensitive 1H-29Sicross-polarization NMR does provide a directcheck of the validity of this layered structure (Fig.4 in [20]). This NMR approach does support theabove structural description, and finds the styrenesegments — phenyls and backbones — close to thesurfaces and the surfactant tails removed towardsthe center of the slit [20]. Moreover, NMR spec-troscopy provides insight on the segmental dynam-ics of the various moieties in the slit pore, asdiscussed in the next paragraph.

3.2. Segmental dynamics of PS in 2 nm slit pores

A combination of surface sensitive 1H-29Si cross-polarization NMR and quadrupolar echo 2H NMRwas used to study the segmental dynamics ofselectively deuterated PS in 2 nm slit pores [20].From these recent NMR studies a new and quiteunexpected picture is emerging for the dynamics ofnanoscopically confined polymers, which was fur-ther confirmed by dielectric spectroscopy investiga-tions on a polysiloxane system [19]. Namely, acoexistence of fast and slow segmental dynamics isobserved over a wide range of temperatures, belowand above the bulk Tg [20]. This very wide distri-

bution of segmental dynamics spanned relaxationtimes that ranged from much faster to much slowerthan the ones observed in the bulk, and thisbehavior was recorded for all the temperaturesstudied. In other words, very fast motions wereobserved in the confined polymer for temperaturesdeep below the bulk Tg, where the correspondingbulk polymer was essentially in a solid-like, immo-bilized, glassy state (Fig. 4); while for temperaturesabove the bulk Tg, there still exist many segmentsthat remain solid-like and immobilized in theconfined systems [20], when at the same tempera-tures the bulk PS is in a molten, liquid-like state.

Our molecular modeling approach is a conscien-tious effort to unveil the atomic scale origins ofthese dynamical inhomogeneities. To this end, wesimulated the system depicted in Fig. 1 for severaltemperatures in the same T-range of the 2H NMRexperiments. We monitor segmental mobility

Fig. 4. Quadrupolar-echo 2H NMR intensities multiplied by T,which corrects for the Curie temperature decrease of observ-able intensity with increased T. Data are shown for selectivelydeuterated PS (at the backbone and at the phenyl rings, d-3and d-5 PS, respectively), in bulk and intercalated in organi-cally-modified FH. Intensities less than one correspond to 2Hsites mobile on the time scale of the spin echo experiment, asis indicated in the right-hand axis. In bulk, ring modes areenabled somewhat below Tg, while backbone modes becomesignificant only near Tg. In intercalated samples, the twomodes are coupled and grow in over a broad temperaturerange from well below Tg, suggesting the existence of fastermodes in confined PS than are found in bulk PS (from [20]).

E. Manias et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 187–188 (2001) 509–521516

Fig. 5. Density profiles versus distance for all the phenylcarbons (dashed line) and the carbons belonging to mobilephenyls (at 370 K light line, and 420 K heavy line). Withascending temperature the number of mobile phenyls in-creases, as can be enumerated by the area under the corre-sponding profiles, and at the same time, the mobile phenyls arepreferentially located in the center of the slit. The total organicdensity is also shown (dotted line) scaled by 0.75.

quantify the percentage of mobile phenyls as afunction of temperature. From these two figures,it becomes obvious that with increasing tempera-ture the number of mobile phenyls increases, andalso these mobile rings are located increasingly inthe center of the pore. This is not unexpected as itwas observed in the NMR experiments. Since themobile phenyls are located both in the center ofthe slit pore, as well as near the surfaces — albeitin smaller numbers — the fast segmental dynam-ics do not seem to originate from the location ofthe segments across the pore. Motivated fromMD simulations in glass forming simple liquids[29], we can check the correlation between thesegmental dynamics and the local structure inside

Fig. 6. Some aspects of the mobile phenyl rings as a functionof temperature: (a) the mobile fraction of phenyls, character-ized by fast orientational relaxations through ring ‘flips’ and/or axis ‘rocking’. (b) The covariance of the relaxation time andthe local density for the mobile phenyls, the positive covari-ance denotes that fast relaxing phenyls are located in lower-than-average local density regions (� flipping, � rocking). (c)The average distance between mobile phenyls in the confinedsystem, the decrease with T is due to the increased number ofmobile phenyls, and also their tendency to preferential concen-trate in the middle of the slit pore bunched in groups at highertemperatures.

much like the NMR is doing, by following therelaxation of the C�H bond reorientation. Beyondtranslation, one can define two types of phenylreorientations: the ‘flipping’ around the phenylaxis, and the ‘rocking’ of this axis. For the back-bone methylenes and methines, C�H reorientationtakes place by trans-to-gauche isomerization. ForNMR, ‘mobile’ are those C�H bonds that arecharacterized by relaxation times smaller than theNMR echo spacing (�=20 �s). In our simula-tions, we define as mobile those phenyls that arecharacterized by relaxation times shorter than �,as measured from the time autocorrelation func-tion of the corresponding C�H bond vectors [21].For both segmental moieties, we found an in-creasing number of mobile entities across the slitwith ascending temperature.

Focusing on the phenyl ring dynamics, we showin Fig. 5 the partial densities of the carbonsbelonging to ‘mobile’ phenyl rings, for two differ-ent temperatures (370 and 420 K). In Fig. 6(a), we

E. Manias et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 187–188 (2001) 509–521 517

the confined film. This can be realized through theestimation of the covariance of the segmentalrelaxation and the local density, defined as:

cov(�,�)��(�−���) · (�−���)�

��2(�)�2(�)(1)

where, � is the segmental relaxation time of aparticular phenyl, ��� is the ensemble averageover all phenyls, ��� is the average density acrossthe slit pore, and � is the local density around thephenyl in question. �2(�) and �2(�) are theensemble variances of the relaxation times anddensity, respectively. Following this definition,maximum correlation between fast segmental re-laxations and low local densities (or equivalentlyslow segmental relaxations and high local densi-ties) yields a value of cov(�,�)= +1. Or, for themobile phenyls of Fig. 6(a):

The relaxation time/local structure covariance forthe mobile phenyls is shown in Fig. 6(b). For thelower temperatures simulated there exists a strongcorrelation between the local structure and thesegmental relaxation. Namely, the fast relaxingphenyls are located in regions where the localdensity is low (independent of whether they areclose to the walls, or in the center of the slit; andindependent of whether the C�H relaxation takesplace through ‘flipping’ or ‘rocking’ motions). Asthe temperature increases this correlation seems todecay in magnitude, although it still remainspositive.

In fact, as the temperature increases, the num-ber of mobile phenyls also increases in such adegree that large volumes in the middle part ofthe slit consist solely of phenyls with fast segmen-tal dynamics. This reflects in a decreasing averagedistance between fast relaxing phenyls (as seen inFig. 6(c)). Furthermore, careful analysis of theMD trajectories reveals that at the lower tempera-tures there exist only a few segments with fastdynamics, which are isolated from each-other in-side the confined film. As the temperature in-

creases, ‘groups’ of mobile phenyls are formed,especially in the center of the slit pore (Fig. 7);however, despite their proximity, the motions ofthe phenyls within the same ‘group’ are not corre-lated nor cooperative. For all the temperaturessimulated (even above 420 K) there exists a largenumber of immobilized phenyl rings in the imme-diate vicinity of the confining surfaces, where theorganic film is locally densified by the silicatesurfaces. The relaxation of these slow phenyls ismarkedly temperature independent, at least forthe time scales that our MD simulations probe.This is in general agreement with the cross-polar-ization NMR studies, which suggest that styrenemoieties in the center of the slits are more mobile,while chain elements (phenyls, methylenes, andmethines) interacting with the surface are dynami-cally inhibited.

Similar analysis for the backbone segmentaldynamics reveals the same trends as discussedabove for the phenyls [21]. That is, fast backbonesegmental dynamics are predominately found inregions where the local density is low, and withincreasing temperature mostly in the center of theslit pore. Furthermore, an interesting observationis that the fastest relaxing phenyls are covalentlybonded to the fastest backbone segments, and thesame for the slowest moving backbones andphenyls. This was not obvious, or intuitively ex-pected, as the phenyl segmental relaxation (e.g.through flipping) does not require backbone reori-entation. For example, in bulk — unconfined —PS chains phenyl modes are enabled at tempera-tures below Tg where the backbone remains rigid;backbone modes start only near Tg. Although amore detailed analysis is necessary, this connectiv-ity between moieties with fast segmental relax-ations could be the reason that the phenyl and thebackbone dynamics are strongly coupled in theconfined films (Fig. 4).

Due to the existence of a large number ofalkyl-ammonium molecules in our confined sys-

cov=

�����

+1 all fast segmental dynamics at local density minima0 nocorrelation between segmental dynamics and local density

−1 all fast segmental dynamics at local density maxima(2)

E. Manias et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 187–188 (2001) 509–521518

Fig. 7. All the PS phenyls are shown and the ones with fast segmental dynamics are highlighted. The hydrogen atoms, the polymerbackbones, and the surfactant molecules are omitted for clarity. To promote comparison, the same system configuration at twodifferent temperatures is shown (left: 350 K, right: 420 K). For the lower temperature the mobile phenyls are isolated, and arelocated in parts of the system where the local density is low. At 420 K, regions of closely bunched phenyls with fast segmentaldynamics develop, especially in the center of the slit, but there also still exist isolated mobile phenyls especially in the low localdensity regions near the confining surfaces.

tems, that can ‘plasticize’ the polymer, it was notinitially clear [20] whether these dynamical inhomo-geneities arise from plasticization effects, orwhether are confinement induced. On this point,our simulations show that the fast segmental relax-ations are not correlated to the surfactant proxim-ity and/or dynamics [21], but originate from theconfinement-induced density inhomogeneities. Thisis further supported by other experimental studiesof different systems, which also provide strongindications that these segmental dynamical areassociated with the polymer confinement, (i ) 2HNMR lineshape studies by Zax and coworkers, alsoobserved coexisting fast and slow modes inpoly(ethylene oxide), which was intercalated in thesame silicate as the PS herein, but in absence of anysurfactant whatsoever [30,31]. In this case, despitethe much narrower confinements (0.8 nm), theintercalated PEO chains exhibit substantial seg-mental motion [30,31]. (ii ) Dielectric spectroscopy

of poly(methyl phenyl siloxanes) (with a Tg morethan 150 K below the PS Tg) report a similarcoexistence of ultra-fast and solid-like (slow) relax-ations over a wide range of temperatures [19], andmeasure an extraordinarily fast relaxation, which iscompletely absent in the bulk unconfined polymer.

Finally, we would like to commend on thequantitative comparison between the molecularsimulations and the NMR experiments. The assid-uous reader must have noticed that the simulationenumeration of the mobile fraction (Fig. 6(a))underestimates the NMR measured one (Fig. 4),albeit capturing the correct temperature depen-dence. This discrepancy can result from severalsources, such as the uncertainties in the relaxationtime calculation, and phenomena that becomeimportant after the 100 ns probed by our MD runs(such as polymer diffusion or desorption). How-ever, none of these two reasons can adequatelyaccount for the deviation between the experimental

E. Manias et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 187–188 (2001) 509–521 519

and simulation values. We believe that the causeof this difference lies in the styrene and alkyl-am-monium densities used in our simulations, whichmust be higher than the ones in the intercalatedexperimental systems. Even though we based ourdensities on the TGA estimated quantities, thegravimetric method may also measure polymersphysisorbed on the outside of the silicate stacks,i.e. not in the confined film. Since there exists avery strong dependence between the density andthe fast segmental dynamics, minute deviations indensity can result in large variations in the systemdynamics. Nevertheless, the aim of the simulationsis not to quantitatively reproduce the experimentalobservations, but to reveal the origins of theexperimentally observed behavior, and provideinsight to the relevant molecular processes.

3.3. The glass transition in 2 nm confinements

The glass transition in thin (10–150 nm) sup-ported polystyrene films has been extensively stud-ied by several methods, including ellipsometry,X-ray reflectivity, positron annihilation lifetimespectroscopy (PALS), and Brillouin light scatter-ing [32–37]. Although many of the findings stillremain controversial, some consensus has beenreached [38]. Specifically, it is believed that wherethere exist strong specific interactions between thepolymer and the surface, the glass transition tem-perature (Tg) increases, whereas near weak inter-acting SiOx surfaces Tg decreases. In efforts toexplain these results some studies [37] propose theexistence of a 50 nm immobilized, rigid, ‘dead’layer of polymer adsorbed on the solid surfaces, inwhere no Tg is observed and the polymer andsegmental dynamics are believed to be inhibited.

Our intercalated polymers offer a great modelsystem to study the glass transition in a muchmore severe confinement of 2 nm, where all thepolymer segments are adsorbed on the confiningsurfaces. These systems provide the means todirectly explore the behavior of the adsorbed poly-mers, without the existence of any non-adsorbed— bulk — material. In Fig. 4 any evidence of aglass transition is completely absent in theconfined polymer systems, whereas is clearly ob-served in the corresponding bulk. In order to

further explore this response, we used DSC tostudy the same intercalated material used in theNMR, and no glass transition was detected in theconfined systems (Fig. 8). Considering both theNMR and the DSC data, it becomes obvious thatthere seems to be no clear glass transition in theseultra-confined polymers inside the 2 nm slit pores.This is in concert with extensive studies investigat-ing the glass transition of poly(ethylene oxide)intercalated in similar slit pores, employing DSCand thermally stimulated dielectric depolarization[39].

If we briefly consider the theoretical descrip-tions of the glass transition, we find that manymodern approaches emphasize the existence of a‘cooperativity volume’ that increases in size as Tapproaches Tg. This volume consists of moietieswith cooperatively rearranging segmental dynam-ics, and characteristic space dimensions of about 4nm or larger. Apparently, both simulations andexperiments suggest that in our 2 nm slit pores thecreation of such cooperativity regions is strongly

Fig. 8. Differential scanning calorimetry study of bulkpolystyrene, PS/octadecyl-ammonium FH (F18) physicallymixed at room temperature (RT), and PS intercalated in F18(the same system used in the NMR studies). Although in theneat polymer and the polymer/silicate mixture the glass transi-tion is evident, it seems to be suppressed when the samepolymer is confined in 2 nm slit pores (figure adopted from[17]).

E. Manias et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 187–188 (2001) 509–521520

perturbed by the severe confinement size. Thisraises the issue whether in these slit pore ge-ometries bulk-scale cooperative relaxations —traditionally associated with the glass transition— are even possible. However, the assumptionthat the adsorbed polymer segments create a rigid,‘dead’ layer seems to be at odds with our simula-tion findings discussed in the earlier paragraph.On the contrary, we find the adsorbed polymersegments to have a very rich dynamical behaviorthat simultaneously includes ultra-fast and solid-like slow relaxations.

4. Conclusions

Complementary NMR experiments and MDmolecular modeling approaches are employed toexplore polystyrene molecules confined in 2 nmslit pores, defined by alkyl-ammonium modifiedlayered-silicate surfaces.

The simulations suggest that the confined filmadopts a layered structure normal to the solidsurfaces, with the polar phenyls dominating theorganic material adsorbed on the walls, and thealiphatic groups predominately in the center ofthe pore. This layered structure, with the polarstyrene groups preferentially on the surfaces,reflects on 1H-29Si cross-polarization NMRmeasurements.

NMR reveals a coexistence of ultra-fast andsolid-like slow segmental dynamics throughout awide temperature range, below and above Tg, forboth the styrene phenyl and the backbone groups.The mobile moieties concentrate at the center ofthe slit pore, especially for the higher tempera-tures. The molecular simulations attribute thesedynamical inhomogeneities to confinement-in-duced density inhomogeneities. Namely, the moi-eties with fast segmental dynamics are located inlocal density minima. Moreover, they show thatat low temperatures the fast moving species areisolated throughout the confined film, whereaswith increasing temperature, the mobile moietiesincrease in number and bunch up in groups espe-cially in the middle of the pore.

Finally, NMR suggests, and DSC confirms,that there is no detectable glass transition inside

the 2 nm confined polymer. This leads us to theconclusion that such severe confinements hinderthe creation of extended regions with cooperativesegmental dynamics, which underlie the glasstransition.

Acknowledgements

This work was supported by a Wilson ResearchInitiation grant from the EMS College at PennState. We would like to thank Dr R. Vaia forinsightful discussions and for providing the DSCdata in Fig. 8.

References

[1] H.K. Christenson, J. Chem. Phys. 78 (1983) 6906.[2] D.Y.C. Chan, R.G. Horn, J. Chem. Phys. 83 (1985) 5311.[3] J.P. Montfort, G. Hadziioannou, J. Chem. Phys. 88

(1988) 7187.[4] M.L. Gee, et al., J. Chem. Phys. 93 (1990) 1895.[5] J. van Alsten, S. Granick, Macromolecules 23 (1990)

4856.[6] S. Granick, Science 253 (1991) 1374.[7] H. Hu, G.A. Carson, S. Granick, Phys. Rev. Lett. 66

(1991) 2758.[8] J. Klein, E. Kumacheva, Science 269 (1995) 816.[9] J. Israelachvili, Intermolecular and Surface Forces, Aca-

demic Press, 1991.[10] I. Bitsanis, C. Pan, J. Chem. Phys. 99 (1993) 5520.[11] E. Manias, G. Hadziioannou, G. ten Brinke, Langmuir 12

(1996) 4587.[12] E. Manias, G. Hadziioannou, I. Bitsanis, G. ten Brinke,

Europhys. Lett. 24 (1993) 99.[13] E. Manias, I. Bitsanis, G. Hadziioannou, G. ten Brinke,

Europhys. Lett. 33 (1996) 371.[14] D. Tabor, R.H. Winterton, Proc. R. Soc. London A312

(1969) 543.[15] D. Tabor, J.N. Israelachvilli, Proc. R. Soc. London A331

(1972) 19.[16] J.N. Israelachvilli, G.E. Adams, J. Chem. Soc. Faraday

Trans. I 74 (1978) 975.[17] E. Giannelis, R. Krishnamoorti, E. Manias, Adv. Poly-

mer Sci. 138 (1998) 107.[18] M. Alexandre, P. Dubois, Mat. Sci. Eng. R: Reports 28

(2000) 1.[19] S. Anastasiadis, K. Karatasos, G. Vlachos, E. Giannelis,

E. Manias, Phys. Rev. Lett. 84 (2000) 915.[20] D. Zax, D.K. Yang, R. Santos, H. Hegemann, E. Gian-

nelis, E. Manias, J. Chem. Phys. 112 (2000) 2945.[21] V. Kuppa, E. Manias, ACS Symp. Ser., in press (2001).

E. Manias et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 187–188 (2001) 509–521 521

[22] During the NVT ensemble simulations we have alsofollowed the pressure of the systems (through the stresstensor at the walls). The pressure is maintained constantfor each temperature, and there exists a slight only change(less than a few percent) across different temperatures.Thus, this NVT ensemble could also be considered as acollection of microstates from an isobaric (NPT) ensem-ble with a weak coupling to a pressure [26].

[23] E. Hackett, E. Manias, E.P. Giannelis, J. Chem. Phys.108 (1998) 7410.

[24] F. Muller-Plathe, Macromolecules 29 (1996) 4782 com-plete information on the PS force-field can be obtainedfrom Muller-Plathe’s web site at the Max-Planck Inst.

[25] Y. Chan, Computer Simulations of 2:1 Silicates Hydra-tion, Master Thesis (under A.Z. Panagiotopoulos), Cor-nell University, 1997.

[26] H.J.C. Berendsen, J.P.M. Postma, W.F. van Gunsteren,A.D. Nola, J.R. Haak, J. Chem. Phys. 81 (1984) 3684.

[27] E. Hackett, E.P. Giannelis, E. Manias, J. Chem. Phys., inpress (2001).

[28] W.L. Mattice, U.W. Suter, Conformational Theory ofLarge Molecules: The Rotational Isomeric State Model inMacromolecular Systems, Wiley, New York, 1994.

[29] J. Qian, R. Hentschke, A. Heuer, J. Chem. Phys. 111(1999) 10177.

[30] S. Wong, S. Vasudevan, R. Vaia, E. Giannelis, D. Zax, J.Am. Chem. Soc. 117 (1995) 7568.

[31] S. Wong, R. Vaia, E. Giannelis, D. Zax, Solid StateIonics 86 (1996) 547.

[32] J. Keddie, R. Jones, R. Cory, Europhys. Lett. 27 (1994)59.

[33] J. Keddie, R. Jones, J. Isr. Chem. Soc. 35 (1995) 21.[34] J. Forrest, K. Danloki-Veress, J. Dutcher, Phys. Rev. E

56 (1997) 5705.[35] W. Wallace, J. van Zanten, W. Wu, Phys. Rev. E 52

(1994) R3329.[36] W. Wallace, J. van Zanten, W. Wu, Phys. Rev. E 53

(1996) R2053.[37] G.B. DeMaggio, W.E. Frieze, D.W. Gidley, M. Zhu,

H.A. Hristov, A.F. Yee, Phys. Rev. Lett. 78 (1997) 1524.[38] J. Forrest, R. Jones, in: A. Karim, S. Kumar (Eds.),

Polymer Surfaces, Interfaces and Thin Films, World Sci-entific, 2000, p. 251.

[39] R. Vaia, B. Sauer, O. Tse, E. Giannelis, J. Polymer Sci.Part B: Polymer Phys. 35 (1997) 59.

.