Direct Measurement of Glass Transition Temperature in Exposed andBuried Adsorbed Polymer NanolayersMary J. Burroughs,† Simone Napolitano,§ Daniele Cangialosi,∥ and Rodney D. Priestley*,†,‡

†Department of Chemical and Biological Engineering and ‡Princeton Institute for the Science and Technology of Materials, PrincetonUniversity, Princeton, New Jersey 08544, United States§Laboratory of Polymer and Soft Matter Dynamics Faculte ́ des Sciences, Universite ́ Libre de Bruxelles (ULB), Boulevard duTriomphe, Bat̂iment NO, Bruxelles 1050, Belgium∥Centro de Física de Materiales (CSIC-UPV/EHU), Paseo Manuel de Lardizab́al 5, San Sebastiań 20018, Spain

*S Supporting Information

ABSTRACT: We employ a fluorescence bilayer method todirectly measure the glass transition temperature (Tg) of theirreversibly adsorbed layer of polystyrene (PS) buried in bulkfilms as a function of adsorption time, tads. This bilayergeometry allows for the examination of interfacial effects on Tgof the adsorbed nanolayer. In the presence of a free surface, weobserve a substantial reduction in Tg from bulk that lessenswith tads as a result of increased chain adsorption at thesubstrate. Submerging the adsorbed layer and effectivelyremoving the free surface results in a suppression of the Tgdeviation at early tads, suggesting chain adsorption dictates Tgat long tads. Annealing in the bilayer geometry promotes recovery of bulk Tg on a time scale reflecting the degree of adsorption.Our data are quantitatively rationalized via the free volume holes diffusion model, which explains adsorbed nanolayer Tg in termsof the diffusion of free volume pockets toward interfacial sinks.

■ INTRODUCTIONThin polymer films in contact with a substrate serve as theenabling material for a range of emerging technologies,including nanoimprint and block copolymer lithography formicroelectronics,1,2 membranes for efficient separations3 anddrug delivery,4 and semiconductors for organic solar cells5 andelectronics.6 In these applications, the polymer thin film is oftenprepared by solution casting directly atop the substrate.6 Afterinitial film formation, often a critical step in production isthermal annealing the polymer film in the melt state to removeexcess solvent,7 relax residual stresses and thermal historyinduced during the casting,8 andin the case of blockcopolymer filmsinduce self-assembly.9,10 During prolongedmelt-state annealing, monomer−substrate interactions on theorder of kBT can lead to the formation of an irreversiblyadsorbed (or physically bound) nanolayer.11−13 The long-chainnature of macromolecules, whereby multiple interaction sitesexist, stabilizes this adsorbed nanolayer against desorption.14,15

Though the nature of polymer adsorption onto surfaces fromsolutions has been well-established,16−20 adsorption from themelt state is still being studied in terms of its mechanism,21−23

structure,24 and influence on thin film properties.12,25,26

Recent work has focused on examining a connection betweenthe development of irreversibly adsorbed nanolayers duringmelt-state annealing and the well-documented significantdeviations from bulk physical properties in ultrathin polymerfilms.8,27−34 Several studies have demonstrated a correlation

between the growth of an irreversibly adsorbed nanolayer anddeviations in thin film properties from the bulk, including glasstransition temperature (Tg),

12 viscosity,26 dynamics,24 dewet-ting, and diffusion.25 Additionally, these adsorbed layers havebeen studied in terms of their stability.35 In particular, it wasshown that the Tg in polystyrene thin films could be controlledsolely by adjusting the annealing time and temperature, whichinfluenced the development of the adsorbed nanolayer.12 Thecurrent understanding of the properties of the irreversiblyadsorbed nanolayer (including Tg) has been obtained byperforming studies on the exposed nanolayer, i.e., after theremoval of unadsorbed polymer and in a capped geometry.12

Unfortunately, experimental challenges have prevented theanalysis of adsorbed nanolayer properties buried within the thinfilm, i.e., in situ analysis, and the examination of interfacialinfluences. As a result, understanding the critical role that theformation of adsorbed nanolayers plays in determining theproperties of ultrathin polymer films remains a fundamentalchallenge.To address this challenge, we employed a unique

fluorescence bilayer method to directly measure Tg of theirreversibly adsorbed polymer nanolayers. This measurementwas achieved by forming an adsorbed nanolayer covalently

Received: February 24, 2016Revised: April 26, 2016Published: June 7, 2016

Article

pubs.acs.org/Macromolecules

© 2016 American Chemical Society 4647 DOI: 10.1021/acs.macromol.6b00400Macromolecules 2016, 49, 4647−4655

pubs.acs.org/Macromoleculeshttp://dx.doi.org/10.1021/acs.macromol.6b00400

labeled with a fluorescence dye molecule capable of reportingpolymer Tg

36 and incorporating it into a bulk polymer film. Thestudy was undertaken using polystyrene (PS) as a modelpolymer, for which Tg in the thin film geometry has beenthoroughly investigated.8,27,36,37 By performing studies in boththe exposed and buried adsorbed nanolayer geometries, we arealso able to examine the influence of the free surface on thenanolayer Tg. Our results reveal a suppression of the freesurface influence and increased deviation of Tg in submergedadsorbed layers with longer annealing. The development ofthese trends is attributed to increased chain adsorption at thesubstrate interface. Furthermore, bilayer-annealing studies showa marked increase in the time scale of bulk Tg recovery withenhanced degree of adsorption. These findings stand to expandthe understanding of how irreversible adsorption contributes tothin film Tg through in situ experiments and the examination ofinterfacial effects. An appreciation for this relationship enablesprediction of the impacts of melt-state annealing ontechnologically relevant polymer thin film properties.

■ EXPERIMENTAL SECTIONPolymer Synthesis and Characterization. The fluorescent

probe incorporated into labeled polystyrene, pyrenebutyl methacrylate,was synthesized via an esterification reaction of methacryloyl chloride(purchased from Sigma-Aldrich) with 1-pyrenebutanol (Aldrich) intetrahydrofuran (THF) (Sigma-Aldrich) and purified by recrystalliza-tion in ethanol (Fisher Scientific). It was characterized by 1H NMR ona Bruker Avance III 500 MHz spectrometer. Pyrene-labeled PS (lPS)(Mw = 155 × 10

3 g/mol; Mw/Mn = 1.9) was synthesized by bulk freeradical polymerization of styrene (Sigma-Aldrich) with 0.16 mol %pyrenebutyl methacrylate at 75 °C under nitrogen with benzoylperoxide as the initiator. Afterward, the polymer was dissolved intoluene (Fisher Chemical) and precipitated in methanol (FisherChemical) three times to remove any excess monomer. Molecularweight and dispersity were measured by size exclusion chromatog-raphy (SEC) in THF relative to PS standards. The labeling amountwas determined to be 0.2 mol % by absorbance measurements on anAgilent Technologies Cary 5000 UV−vis−NIR spectrophotometer.Unlabeled PS (116 × 103 g/mol; Mw/Mn = 1.9) was synthesized andcharacterized similarly without the addition of pyrenebutyl meth-acrylate. Similar synthesis routes were followed to achieve identicalend functionality and comparable dispersity for both polymers. BulkTg for lPS and neat PS were measured to be 106 and 104 °C,respectively, via differential scanning calorimetry (DSC) on a TAInstruments Q2000 DSC on second heating at 1 °C/min. Thesemeasurements were in agreement with the bulk Tg of 106 °Cmeasured by fluorescence.Irreversibly Adsorbed Nanolayer Preparation. To prepare the

irreversibly adsorbed layer at different stages of development, 200 nmthick films of lPS and neat PS were spin-coated from a polymer−toluene solution onto piranha-treated (soaked in a 70:30 solution ofH2SO4 to H2O2, at 90 °C for 1 h) substrates. Films for fluorescencemeasurements were spin-coated on silica and films for ellipsometrymeasurements were spin-coated on silicon substrates. These films werethen annealed at 150 °C under vacuum for various times tads (0.5−20h) to promote chain adsorption and the resulting growth of adsorbedlayers. After the annealed films were rapidly cooled to roomtemperature, they were rinsed with toluene and soaked in freshtoluene for 10 min. This rinsing and soaking procedure was repeatedtwice more to thoroughly remove any unadsorbed chains, for a total ofthree wash cycles. Figure 1 shows that the fluorescence intensity of thetoluene soaking solutions decreases below that of pure toluene aftertwo soaks, indicating three soaking cycles are sufficient to removeunadsorbed lPS. In the remainder of the text, adsorbed layers will bereferred to by the initial annealing time required to make them,adsorption time tads. For example, an adsorbed layer produced afterannealing a 200 nm spin-cast film for 8 h at 150 °C and then washing

it in toluene is an 8 h adsorbed layer. This classification will persistregardless of further treatment (see bilayer film preparation). Theseexposed nanolayers were then dried under vacuum at roomtemperature overnight to remove any excess solvent beforefluorescence, thickness, or surface topography measurements.

Bilayer Film Preparation. In order to prepare bilayer films, 400nm neat PS overlayers were spin-coated from a toluene solution ontomica and annealed above Tg to remove residual solvent and stresses.These unlabeled bulk films were then floated onto previously preparedlPS adsorbed layers in a bath of deionized water. After dryingthoroughly, the stacked films were annealed at 120 °C for 20 min inorder to remove wrinkles induced during floating and create aconsolidated bilayer film. We note that both this consolidationtemperature and annealing time are considerably less than those usedto make the adsorbed nanolayer, thus ensuring minimal modificationof the adsorbed amount during consolidation. We therefore do notcount consolidation steps toward bilayer annealing time. The bilayerfilms at this stage are referred to as containing buried, or submerged,adsorbed layers of tads (shown in Figure 6).

Several of these bilayer films were then later annealed further at 150°C under vacuum for bilayer annealing time tBann (shown in Figure 7),where tBann = 0 is the state at which bilayer films are consolidated butnot further annealed. These further annealed bilayer systems arereferred to by both tads and tBann. Hence, a 6 h irreversibly adsorbed

Figure 1. (A) Fluorescence spectra of subsequent soaking solutions foran 8 h irreversibly adsorbed layer compared to that of toluene. (B)Integrated fluorescence intensity of soaking solutions for 8 hirreversibly adsorbed layers. Error bars (smaller than data points)represent ±1 standard deviation of the integrated fluorescenceintensity of soaking solutions from identically prepared films.

Macromolecules Article

DOI: 10.1021/acs.macromol.6b00400Macromolecules 2016, 49, 4647−4655

4648

http://dx.doi.org/10.1021/acs.macromol.6b00400

layer (tads = 6) that has been capped with neat PS, consolidated, andfurther annealed for 2 h is a 6 h bilayer system with tBann = 2.Thickness and Morphology Measurements. Thicknesses of

lPS and neat PS adsorbed layers were determined using a Woollam M-2000 spectroscopic ellipsometer and fitting with a two-layer model(native oxide and PS layers). The surface topography of lPS and neatPS irreversibly adsorbed layers was studied using an Asylum ResearchMFP-3D-SA atomic force microscope (AFM) in tapping mode. TheAFM probes were silicon with force constants = 11−54 N/m andresonance frequencies = 200−400 kHz. All images were taken at roomtemperature with 512 × 512 pixel resolution and analyzed usingGwyddion software.Glass Transition Temperature Measurement. A Horiba

Scientific Fluorolog-3 spectrofluorometer was employed to measureTg of the nanolayers. The lPS adsorbed layers were excited at λ = 340nm (with 3.8 nm band-pass), and emission spectra were collected fromλ = 350 to 500 nm. The fluorescence emission spectra of lPS adsorbedlayers were obtained upon cooling from 140 to 30 °C in 5 °Cincrements after initially holding at 140 °C for 20 min to removethermal history. To ensure equilibration and isothermal measure-ments, films were stabilized for 5 min at each temperature prior tocollecting spectra; this resulted in an effective average cooling rate of 1°C/min. The integrated spectra were plotted as a function oftemperature, and Tg was identified as the intersection of linear fits tothe rubbery and glassy regimes (R2 > 0.99), i.e., a pseudo-thermodynamic approach.38−40

■ RESULTS AND ANALYSISGrowth of Adsorbed Nanolayers. In recognition of

structure−property relationships inherent to adsorbedlayers,12,41 an understanding of their surface topography andthickness at different stages of development is essential torelating trends in Tg and interfacial influences to their growth.As such, we characterized the growth of our adsorbed layers incontext of the literature. Figure 2 shows AFM topologicalimages of lPS after annealing for different times, tads. Inagreement with prior literature studies, a smooth nanolayer wasobserved after a critical adsorption time.11 In line with the workof Bal et al. on the stability of PS adsorbed layers,35 we observethat immediately after rinsing, adsorbed layers with tads less than2 h (corresponding to a thickness of 3.1 ± 0.2 nm) wereunstable and spinodally dewetted, as recognizable by thecharacteristic pattern shown in Figure 2. Because of thisdewetting, adsorbed layers formed after less than 2 h ofannealing have an enhanced surface area due to this instability;the implications of this on Tg will be discussed later.Adsorbed layer growth has been shown to be dependent on

both polymer molecular weight and annealing temper-ature.11,12,42 The thickness of the irreversibly adsorbednanolayer, hads, for both lPS and neat PS as a function ofadsorption time is shown in Figure 3. Films were annealed onsilicon and thickness measurements were obtained byellipsometry. For films annealed less than 2 h, in whichincomplete nanolayers were formed, the average film thicknesswas determined by averaging the thickness across the film, tomatch average thicknesses reported by ellipsometry. Withinexperimental error, the growth rates of the adsorbed nanolayersfor both the labeled and neat PS are identical and follow a trendconsistent with the literaturesmall differences due to theirrespective molecular weights are expected.11 The dotted lines inFigure 3 indicate plateau thicknesses, h∞, for each polymer,representing average thicknesses of adsorbed layers afterleveling out at tads = 6 h. The Rg of labeled and neat PS usedin this investigation were calculated according to the correlation

reported by Fetters et al. for PS in cyclohexane at 34.5 °C(theta solvent and temperature):43

Figure 2. 5 × 5 μm AFM surface topography images of lPS irreversiblyadsorbed layers annealed for different times: (A) 0.5 h (5 nm heightscale), (B) 1 h (5 nm scale), (C) 2 h (2 nm height scale), and (D) 4 h(2 nm height scale). White lines indicate the section of the imagerepresented by the inset profiles.

Figure 3. Adsorbed layer thickness measured via ellipsometry as afunction of tads for lPS and neat PS on silicon. Thicknesses of lPSadsorbed layers on silica substrates (green circles) were measured viaAFM. Error bars represent ±1 standard deviation of hads for multiplesamples. Lines show h∞ for each polymer, obtained by averagingadsorbed layer thicknesses after 6 h of annealing at 150 °C.

Macromolecules Article

DOI: 10.1021/acs.macromol.6b00400Macromolecules 2016, 49, 4647−4655

4649

http://dx.doi.org/10.1021/acs.macromol.6b00400

= × −R M2.42 10g2 0.512

(1)

Using the weight-average molecular weights measured by SEC,Rg is 11.0 and 9.5 nm for lPS and neat PS, respectively. Dividingh∞ by Rg for each polymer gives scaling factors of 0.49 ± 0.05and 0.55 ± 0.03 for lPS and PS, respectively. These scalingfactors are within error of each other and are in agreement withthe scaling relationship h∞ = 0.47Rg reported by Fujii et al. forPS adsorbed layers on SiOx.

42 These growth profiles arepresented in Figure 3 and confirm that the pyrene moleculeslabeled to PS (at less than 0.2 mol %) do not alter its adsorbedlayer growth kinetics or h∞.We note that also plotted in Figure 3 are hads for select

nanolayers formed on silica in which film thickness wasmeasured via AFM. The thicknesses of select adsorbed layersformed on silica were within error of corresponding thicknesseson silicon, indicating no discernible difference between theinfluences of the two substrates on irreversibly adsorbednanolayer formation.Tg of Exposed Adsorbed Layers. Once the growth and

development of these adsorbed layers were characterized, weexploited the fluorescence of lPS to measure their Tg viafluorescence spectroscopy.36,44,45 Figure 4A plots the normal-ized integrated fluorescence intensity as a function oftemperature for an irreversibly adsorbed nanolayer obtainedby washing a 200 nm thick film of lPS previously annealed for10 h at 150 °C. Linear fits of the fluorescence data at high andlow temperatures (corresponding to rubbery and glassyregimes, respectively) intersect at 90 °C. This intersectionhas been previously shown to be an accurate measurement ofTg in bulk and confined polymers.

36,44,45 Hence, we define Tgfor the 10 h adsorbed nanolayer to be 90 °C. Insets in Figure4A show the temperature-dependent fluorescence spectra forthis 5.5 nm thick adsorbed nanolayer and structure of thepyrenebutyl methacrylate label incorporated into lPS.Using this fluorescence technique, we determined the Tg of

the exposed adsorbed nanolayers as a function of tads. Theresults are presented in Figure 4B and are represented by theblue diamonds. At tads = 1 h, Tg is reduced by 30 ± 4 °C

compared to the bulk. Upon continued annealing up to tads = 6h, Tg of the adsorbed nanolayer gradually increases to 89 ± 3°C, a 17 ± 3 °C reduction from bulk Tg. Interestingly, themagnitude of the Tg deviations reported for the adsorbednanolayers are quite different from those of spin-cast films. Forinstance, Tg − Tg(bulk) ∼ −30 °C for adsorbed nanolayerswith hads ∼ 1.5 nm and for spin-cast films when film thickness h∼ 15 nm.36 This difference is even more noteworthy when oneconsiders that the 1.5 nm thick film is not homogeneous buttextured and, hence, exposed to additional free surfaces whichare considered to be a key requirement to observe a Tgreduction in confined polymer systems. Hence, the adsorbednanolayers have a greater Tg in comparison to spin-cast films ofcomparable thickness. We believe this difference is due to theimmobilization of the chains at the substrate from irreversibleadsorption. We note the elevation of Tg for the adsorbednanolayers is reminiscent of the rigid amorphous fraction insemicrystalline polymers that possesses a higher Tg due to chainimmobilization.46,47

Further annealing beyond 6 h results in no change in Tg,within error. The Tg and thickness of the irreversibly adsorbedlayers show strikingly similar development with tads = 6 h alsobeing the approximate time at which hads ceased to grow. Todemonstrate the similarity, in Figure 5 we have plotted Tgagainst hads. Here, a positive correlation between the twoparameters is observed.Also plotted in Figure 4B (squares) are the only previous Tg

measurements of irreversibly adsorbed PS nanolayers, reportedby Napolitano and Wübbenhorst.12 Here, PS (M = 97 kg/mol)adsorbed layers were sandwiched between two aluminumsubstrates with no free surface present. The initial reduction inadsorbed layer Tg was attributed to additional free volumeresulting from packing frustration during low-density chainadsorption and the increase in Tg of the adsorbed layers uponannealing to the gradual densification of chains at thesubstrate.21 Remarkably, Tg of these capped irreversiblyadsorbed nanolayers completely recovered (and exceeded)bulk Tg.

12 In our system, Tg(bulk) is never recovereda

Figure 4. (A) Temperature dependence of fluorescence intensity for a 10 h irreversibly adsorbed layer, with solid lines being linear fits to the highand low temperature data. Lower inset shows temperature dependence of its spectra. Upper inset shows molecular structure of pyrenebutylmethacrylate monomer copolymerized with styrene at low concentrations in lPS. (B) Exposed adsorbed layer Tg at different tads as measured byfluorescence (diamonds) and as reported for capped adsorbed layers by Napolitano and Wübbenhorst12 (squares), shifted to match respectiveTg(bulk) values. Error bars on fluorescence data correspond to ±1 standard deviation of Tg for multiple samples. Dashed line illustrates Tg predictedby FVHD with fitting parameters (Afree/Atotal)max = 0.55 and a = 0.0083 h

−1.

Macromolecules Article

DOI: 10.1021/acs.macromol.6b00400Macromolecules 2016, 49, 4647−4655

4650

http://dx.doi.org/10.1021/acs.macromol.6b00400

striking difference in behavior that highlights the key role of thefree surface in modifying adsorbed nanolayer Tg. This largeinfluence of the free surface on confined Tg is consistent with amultitude of studies reporting negative Tg deviations in thepresence of a free surface.8,27,28 However, similarities in growthkinetics suggest that free volume also contributes to adsorbedlayer Tg in the presence of a free surface. These comparisonswith supported films and capped adsorbed layers suggest thatthe Tg behavior we observe is due to competition between theinfluences of chain adsorption and free volume at the substrateand enhanced mobility at the free interface.Tg of Buried Adsorbed Nanolayers. To better under-

stand the origins of Tg depression in the exposed adsorbed layerand remove any competing free surface effects, we focused onstudying buried layer dynamics, i.e., in situ analysis of theadsorbed layer, by fabricating bilayer films in which only theadsorbed nanolayer is labeled with the fluorescent dye moleculecapable of reporting Tg; see Experimental Section for theprocedure to make bilayer films. Figure 6 compares the Tg ofthe adsorbed nanolayer in the exposed and buried-layergeometries as a function of tads, where again tads representsthe annealing time directly after spin-casting used to grow theadsorbed nanolayer. The insets in Figure 6 illustrate the twodifferent geometries investigated: one of the exposed nanolayerand the other of the bilayer film consisting of the adsorbednanolayer and neat PS overlayer. At shorter tads, there is adramatic difference in Tg of the exposed and buried nanolayers.Placement of the neat PS overlayer atop the adsorbednanolayer (to create the bilayer film) resulted in the effectiveremoval of the free surface and a near complete recovery ofbulk Tg, within experimental error. Therefore, the bilayergeometry can be considered as effective model for in situanalysis of the adsorbed nanolayer Tg. In this case, the resultsindicate that removal of the free surface leads to recovery ofbulk Tg. We argue that this is due to relatively loose binding ofthe adsorbed nanolayer formed at short tads that can beinfiltrated by the overlayer film.In contrast, placement of the neat PS overlayer atop the

adsorbed nanolayer formed with tads greater than 6 h had aminor effect on the adsorbed nanolayer Tg. While placement of

the overlayer atop the adsorbed nanolayers formed at larger tadsresulted in the removal of the free surface, we argue thatoverlayer could not effectively infiltrate the nanolayer. This isbecause nanolayers formed at long annealing times are tightlybound to the substrate, thus preventing interdiffusion of chainsand interchange of dynamics between the two layers. Thisfinding also suggests the gradual decrease in the influence of thefree surface on the buried nanolayer Tg with increasing degreeof adsorption. That is, Tg of the adsorbed nanolayer isunchanged with and without the overlayer film. To the best ofour knowledge, this is the first study to demonstrate thisphenomenon. Since prior work suggested an increase inadsorption density and decrease in free volume within adsorbedlayers at longer adsorption times,22,41,48 it is reasonable todeduce that strong chain adsorption likely reduces the influenceof the free surface on nanolayer Tg. Hence, excess free volumeresulting from chain packing frustration within the adsorbedlayer exerts a stronger influence on nanolayer Tg than the freesurface in the long-anneal regime.

Tg as a Function of Annealing Time of Bilayer Films.We aimed to provide further insight into the Tg of buriedadsorbed nanolayers by systematically annealing the previouslyconsolidated bilayer films at 150 °C under 10−2 Torr vacuum inorder to promote chain interpenetration and the recovery ofbulk dynamics. Hence, the bilayer films were annealedisothermally at 150 °C and Tg of the buried adsorbed layerwas measured as a function of bilayer annealing time, tBann.Figure 7 plots the Tg of the submerged adsorbed layer as afunction of tBann. The adsorbed nanolayer incorporated intoeach bilayer film in panels A, B, and C was originally preparedby 4, 6, and 10 h of respective annealing. Annealing each bilayersystem at 150 °C resulted in an increase in Tg as a function oftBann. In fact, for each bilayer film bulk Tg of the adsorbednanolayer was eventually fully recovered. Strikingly, the criticalbilayer annealing time required for the attainment of bulk Tg,tBann*, in adsorbed nanolayers was dependent on adsorption

Figure 5. Exposed adsorbed layer Tg (as measured by fluorescence)plotted against adsorbed layer thickness (as measured by ellipsom-etry). The dotted line serves as a guide to the eye. Error bars represent±1 standard deviation in thickness and Tg measurements for multiplefilms.

Figure 6. Irreversibly adsorbed layer Tg with and without a free surfaceas a function of tads. Error bars represent ±1 standard deviation of Tgfor multiple samples. Insets illustrate exposed and buried adsorbedlayer geometries. Lines show Tg predicted by FVHD for exposed andsubmerged adsorbed layers. Predictions for submerged adsorbed layersshown are with fitting parameter (Afree/Atotal)min = 0.05 and hads asmeasured (dash-dotted) and set at 5.4 nm (solid) to account forswelling.

Macromolecules Article

DOI: 10.1021/acs.macromol.6b00400Macromolecules 2016, 49, 4647−4655

4651

http://dx.doi.org/10.1021/acs.macromol.6b00400

time tads (4, 6, and 10 h) in which the adsorbed nanolayer wasformed. To obtain tBann*, we fit each curve in Figure 7A−C toan exponential function (see Supporting Information). InFigure 7D, we plot tBann* as a function tads and confirm thattBann* is strongly influenced by tads, with the 4 h adsorbed layerrequiring the least amount of bilayer annealing to recover bulkTg. However, we note that tBann* should not be interpreted asthe time for complete interpenetration of the upper layer withinthe adsorbed nanolayer. This is because the apparent diffusioncoefficients, based on the use of tBann*, are of order 10

−10 cm2/s,as predicted by the scaling argument:

∼*

Dh

tads

2

Bann (2)

These estimated apparent diffusion coefficients are much highervalues than to be expected for the bilayer films.49 Therefore, wesuggest that while chain interdiffusion likely contributes torecovery of bulk Tg within the adsorbed nanolayers duringbilayer annealing, it cannot be entirely responsible for theobservations presented in Figure 5. Previous studies haveshown that interfacial influences on Tg penetrate length scalesgreater than chain interpenetration,50,51 but the apparent

dependence of these interfacial influences on degree of chainadsorption warrants further explanation.We performed two sets of experiments in order to verify that

these bilayer annealing experiments did not result in anyadsorption of the bulk overlayer, desorption of the irreversiblyadsorbed underlayer, or chain exchange between the two layers.To confirm that additional adsorption of the lPS adsorbed layeror neat PS overlayer was not responsible for observed bulk Tgrecovery, bilayer films were soaked in toluene for 8 h tothoroughly remove unadsorbed chains and then Tg wasremeasured. As seen in Figure 8, no significant change in Tgof the adsorbed layers, within error, took place prior to bulk Tgrecovery for the 4 and 6 h bilayer systems. In films annealedlonger than tBann*, increased scatter led to uncertainty in theaccuracy of Tg measurements and prevented confident reportsof Tg. The fact that the increase in Tg toward Tg(bulk)commences much sooner than tBann* supports the claim thatadditional adsorption is not responsible for the gradual recoveryof bulk Tg in these systems. Additionally, the relative availabilityof interface at lower adsorption times implies that additionaladsorption or desorption would be more likely at lower tads, inthe 4 and 6 h systems.

Figure 7. Recovery of bulk Tg in (A) 4 h, (B) 6 h, and (C) 10 h irreversibly adsorbed layers when annealed in the bilayer geometry, as measured byfluorescence. Error bars represent ±1 standard deviation of Tg for multiple samples. (D) Bulk Tg recovery time as a function of tads. Errors arepropagated from exponential fit parameters. Solid lines in panels A−C show Tg predicted by FVHD.

Macromolecules Article

DOI: 10.1021/acs.macromol.6b00400Macromolecules 2016, 49, 4647−4655

4652

http://pubs.acs.org/doi/suppl/10.1021/acs.macromol.6b00400/suppl_file/ma6b00400_si_001.pdfhttp://dx.doi.org/10.1021/acs.macromol.6b00400

In order to investigate if chain exchange (desorption of alabeled chain paired with adsorption of a neat chain) wasoccurring during bilayer annealing, we analyzed the fluores-cence content of the rinsing and soaking solutions to detectdesorption of lPS. Desorption would be indicated by afluorescence response of the toluene soaking solutions resultingfrom desorbed lPS. Again, more desorption would be expectedin loosely bound adsorbed layers characteristic of low tads.Therefore, we examined a 2 h bilayer system. Figure 9 showsoverlays of the spectra of toluene soaking solutions aftersoaking the bilayer films for 8 h to remove the bulk overlayer.Minimal change in the fluorescence response is detected, withno trends present. One would expect a higher fluorescenceresponse to correlate with longer tBann in the presence of chainexchange. Minimal fluorescence response and lack of trendscombine to indicate a lack of chain exchange or lPS desorption.Interpretation by Free Volume Holes Diffusion Model.

Deviations in Tg of irreversibly adsorbed layers from bulk havebeen attributed to the presence of excess free volume and itsgradual displacement at longer tads due to an increased degreeof chain adsorption at the substrate. The amount of free volume

at the substrate interface has been quantified in cappedadsorbed layers using the free volume holes diffusion (FVHD)model, which defines a relationship between the diffusion offree volume out of a film and the Tg:

48,52

= −⎛⎝⎜

⎞⎠⎟

hD T q

22 ( )eff

2

g1

(3)

where, for our system, heff is an effective length scale of theadsorbed layers, D(Tg) is the diffusion coefficient of freevolume holes at adsorbed layer Tg, and q is the experimentalcooling rate. Comparing our Tg data at long adsorption times(tads > 6 h) to the diffusion coefficient of free volume holes in abulk PS matrix48,52 reveals D(Tg)ads ≈ D(Tg)bulk/350: a 2 orderof magnitude difference, consistent with the literature.53 Detailson FVHD calculations are outlined in the SupportingInformation. The model allows us to predict how Tg changeswith tads, employing the relationship between the diffusion offree volume holes within adsorbed layers and the amount of“free interface” (free volume present at the free surface andsubstrate interface), Afree, available as a sink for said holes(which is related to heff). Incorporating a linear decrease in thefractional free interface with adsorption time, as consistent withprevious studies,41 we are able to describe its progression usingthe equation

= −⎛⎝⎜

⎞⎠⎟

AA

AA

atfreetotal

free

total maxann

(4)

where Afree/Atotal is the fractional free interface, a (≈0.01 h−1) isthe rate of free interface filling, and (Afree/Atotal)max is themaximum amount of free interfaceachieved at tads = 0. Wethen evaluated eqs 3 and 4 using measured hads and optimizingfitting parameters to obtain Tg values. Figure 4 shows Tgpredicted by FVHD overlaying experimental Tg. Closeagreement of the data suggests that the reduced ability offree volume to diffuse to sinks at the substrate and free surfacewith tads is a plausible explanation of our observations. It isworth remarking that applying the FVHD model requiresadjusting the parameter (Afree/Atotal)max, which gives (Afree/

Figure 8. Change in irreversibly adsorbed layer Tg after conductingbilayer annealing experiments and rewashing to remove the bulkoverlayer (postwash Tg) from initial exposed irreversibly adsorbedlayer Tg (original Tg) for (A) 4 h and (B) 6 h irreversibly adsorbedlayers as a function of bilayer annealing time (tBann). Shaded areasindicate when Tg(bulk) is recovered for each system (tBann*), reportedas the lower error limit in Figure 7. Error bars represent ±1 standarddeviation of postwash Tg minus original Tg from multiple samples withthe same annealing conditions.

Figure 9. Fluorescence spectra for the toluene soaking solutions usedto remove neat PS overlayers as a function of bilayer annealing timefor a 2 h irreversibly adsorbed bilayer system. The spectrum for atoluene blank is shown in red.

Macromolecules Article

DOI: 10.1021/acs.macromol.6b00400Macromolecules 2016, 49, 4647−4655

4653

http://pubs.acs.org/doi/suppl/10.1021/acs.macromol.6b00400/suppl_file/ma6b00400_si_001.pdfhttp://pubs.acs.org/doi/suppl/10.1021/acs.macromol.6b00400/suppl_file/ma6b00400_si_001.pdfhttp://dx.doi.org/10.1021/acs.macromol.6b00400

Atotal)max = 0.55 for exposed adsorbed layers. Given theboundary condition (Afree/Atotal) = 0.5 at long adsorptiontime, it is therefore worth pointing out that at tads = 0 thesubstrate interface provides a source of free volume that isabout one-tenth as strong as that being provided by the freesurface. This result is in the same order as that found in ref 48,where the FVHD was applied to Tg depression data of Al-capped PS films.Following similar logic, the FVHD model was used to

describe both removing the free surface (Figure 6) and bilayerannealing experiments (Figure 7). Figure 6 shows that theFVHD model is not consistent with measured Tg of submergedadsorbed layers when using heff derived from measured valuesof hads, likely due to swelling of the submerged adsorbed layersresulting from interpenetration of the two layers. In order toaccount for this swelling, hads was fixed at the average plateauthickness (5.4 nm), thereby representing a decrease in theamount of swelling to zero at tads = 6 h, when nointerpenetration is expected. This modification, also picturedin Figure 6, results in much better agreement between themodel and experimental Tg. The FVHD fits to Figure 7 reflectthe recovery of bulk Tg with bilayer annealing. Curves in the fitsapproaching tBann* result from the combination of VFT andArrhenius components to D(T) in the model.43,47 At tBann* thefit is truncated due to diverging heff when (Afree/Atotal)approaches zero (see Supporting Information). Despitequalitative agreement between FVHD predictions and exper-imental data, differences in the approach to Tg(bulk) indicateadditional levels of complexity in the system. For one, theFVHD model assumes that Dads(Tg) is consistent for all degreesof chain adsorption and is independent of the molecular weightof the polymer matrix. Additionally, in the Tg(bulk) recoveryexperiments it is likely that some degree of chain interdiffusion,suggested by the swelling of the submerged adsorbed layerpredicted by FVHD (Figure 6), also contributes to the rate ofrecovery.

■ CONCLUSIONSWe have examined the glassy dynamics of the buriedirreversibly adsorbed nanolayer within PS thin films byselectively labeling adsorbed layers at different stages ofdevelopment within bilayer films and employing fluorescenceto measure their Tg. Our approach allowed us to report the firstinvestigation of the influence of a free surface on adsorbed layerTg. Through this study, we were able to observe thecompetition of interfaces and how the free surface largelyinfluences adsorbed layer Tg at short tads, but its influence isdiminished with increased degree of adsorption at longer tads.Further examination of bulk Tg recovery through bilayerannealing experiments indicates that an increased degree ofchain adsorption extends the time for bulk Tg recovery. Ourinvestigation of the influence of interfaces and free volumediffusion on irreversibly adsorbed layer Tg serves as afoundation to further the understanding of the interplaybetween processing, structure, and dynamics in polymersconfined to the nanoscale.

■ ASSOCIATED CONTENT*S Supporting InformationThe Supporting Information is available free of charge on theACS Publications website at DOI: 10.1021/acs.macro-mol.6b00400.

Exponential fits to Tg(bulk) recovery and FVHDderivations (PDF)

■ AUTHOR INFORMATIONCorresponding Author*E-mail rpriestl@princeton.edu (R.D.P.).NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSR.D.P. acknowledges support of the NSF Materials ResearchScience and Engineering Center program through the Prince-ton Center for Complex Materials (DMR-1420541) and the AirForce Office of Scientific Research YIP Award (FA9550-12-1-0223). D.C. acknowledges the Spanish Ministry of Education(Project MAT2015-63704-P). S.N. acknowledges the Fonds dela Recherche ScientifiqueFNRS under Grant no. T.0037.15“INCODYNCO”.

■ REFERENCES(1) Bates, C. M.; Maher, M. J.; Janes, D. W.; Ellison, C. J.; Willson, C.G. Block Copolymer Lithography. Macromolecules 2014, 47, 2−12.(2) Hardy, C. G.; Tang, C. Advances in square arrays through self-assembly and directed self-assembly of block copolymers. J. Polym. Sci.,Part B: Polym. Phys. 2013, 51, 2−15.(3) Jackson, E. A.; Hillmyer, M. A. Nanoporous Membranes Derivedfrom Block Copolymers: From Drug Delivery to Water Filtration. ACSNano 2010, 4, 3548−3553.(4) Yang, S. Y.; Yang, J.-A.; Kim, E.-S.; Jeon, G.; Oh, E. J.; Choi, K.Y.; Hahn, S. K.; Kim, J. K. Single-File Diffusion of Protein Drugsthrough Cylindrical Nanochannels. ACS Nano 2010, 4 (7), 3817−3822.(5) Li, G.; Zhu, R.; Yang, Y. Polymer solar cells. Nat. Photonics 2012,6 (3), 153−161.(6) Forrest, S. R. The path to ubiquitous and low-cost organicelectronic appliances on plastic. Nature 2004, 428, 911−918.(7) Forrest, J. A.; Dalnoki-Veress, K.; Dutcher, J. R. Interface andchain confinement effects on the glass transition temperature of thinpolymer films. Phys. Rev. E: Stat. Phys., Plasmas, Fluids, Relat. Interdiscip.Top. 1997, 56, 5705−5716.(8) Keddie, J. L.; Jones, R. a. L.; Cory, R. A. Size-DependentDepression of the Glass Transition Temperature in Polymer Films.EPL (Europhysics Letters) 1994, 27, 59.(9) Davis, R. L.; Chaikin, P. M.; Register, R. A. Cylinder Orientationand Shear Alignment in Thin Films of Polystyrene−Poly(n-hexylmethacrylate) Diblock Copolymers. Macromolecules 2014, 47 (15),5277−5285.(10) Davis, R. L.; Michal, B. T.; Chaikin, P. M.; Register, R. A.Progression of Alignment in Thin Films of Cylinder-Forming BlockCopolymers upon Shearing. Macromolecules 2015, 48 (15), 5339−5347.(11) Housmans, C.; Sferrazza, M.; Napolitano, S. Kinetics ofIrreversible Chain Adsorption. Macromolecules 2014, 47, 3390−3393.(12) Napolitano, S.; Wübbenhorst, M. The lifetime of the deviationsfrom bulk behaviour in polymers confined at the nanoscale. Nat.Commun. 2011, 2, 260.(13) Napolitano, S.; Capponi, S.; Vanroy, B. Glassy dynamics of softmatter under 1D confinement: How irreversible adsorption affectsmolecular packing, mobility gradients and orientational polarization inthin films. Eur. Phys. J. E: Soft Matter Biol. Phys. 2013, 36, 1−37.(14) Rubinstein, M.; Colby, R. Polymer Physics; Oxford UniversityPress Inc.: New York, 2003; p 440.(15) O’Shaughnessy, B.; Vavylonis, D. Irreversibility and PolymerAdsorption. Phys. Rev. Lett. 2003, 90, 056103.(16) Schneider, H. M.; Frantz, P.; Granick, S. The Bimodal EnergyLandscape When Polymers Adsorb. Langmuir 1996, 12, 994−996.

Macromolecules Article

DOI: 10.1021/acs.macromol.6b00400Macromolecules 2016, 49, 4647−4655

4654

http://pubs.acs.org/doi/suppl/10.1021/acs.macromol.6b00400/suppl_file/ma6b00400_si_001.pdfhttp://pubs.acs.orghttp://pubs.acs.org/doi/abs/10.1021/acs.macromol.6b00400http://pubs.acs.org/doi/abs/10.1021/acs.macromol.6b00400http://pubs.acs.org/doi/suppl/10.1021/acs.macromol.6b00400/suppl_file/ma6b00400_si_001.pdfmailto:rpriestl@princeton.eduhttp://dx.doi.org/10.1021/acs.macromol.6b00400

(17) Johnson, H. E.; Granick, S. New Mechanism of NonequilibriumPolymer Adsorption. Science 1992, 255, 966−968.(18) Fu, Z.; Santore, M. M. Kinetics of Competitive Adsorption ofPEO Chains with Different Molecular Weights. Macromolecules 1998,31, 7014−7022.(19) Mubarekyan, E.; Santore, M. M. Influence of Molecular Weightand Layer Age on Self-Exchange Kinetics for Saturated Layers of PEOin a Good Solvent. Macromolecules 2001, 34, 4978−4986.(20) Besio, G. J.; Prud’homme, R. K.; Benziger, J. B. Effect ofelongational flow on polymer adsorption. Macromolecules 1988, 21,1070−1074.(21) Jiang, N.; Shang, J.; Di, X.; Endoh, M. K.; Koga, T. FormationMechanism of High-Density, Flattened Polymer Nanolayers Adsorbedon Planar Solids. Macromolecules 2014, 47, 2682−2689.(22) Rotella, C.; Napolitano, S.; Vandendriessche, S.; Valev, V. K.;Verbiest, T.; Larkowska, M.; Kucharski, S.; Wübbenhorst, M.Adsorption Kinetics of Ultrathin Polymer Films in the Melt Probedby Dielectric Spectroscopy and Second-Harmonic Generation.Langmuir 2011, 27, 13533−13538.(23) Gin, P.; Jiang, N.; Liang, C.; Taniguchi, T.; Akgun, B.; Satija, S.K.; Endoh, M. K.; Koga, T. Revealed Architectures of AdsorbedPolymer Chains at Solid-Polymer Melt Interfaces. Phys. Rev. Lett.2012, 109, 265501.(24) Carrillo, J.-M. Y.; Cheng, S.; Kumar, R.; Goswami, M.; Sokolov,A. P.; Sumpter, B. G. Untangling the Effects of Chain Rigidity on theStructure and Dynamics of Strongly Adsorbed Polymer Melts.Macromolecules 2015, 48, 4207−4219.(25) Jiang, N.; Sendogdular, L.; Di, X.; Sen, M.; Gin, P.; Endoh, M.K.; Koga, T.; Akgun, B.; Dimitriou, M.; Satija, S. Effect of CO2 on aMobility Gradient of Polymer Chains near an Impenetrable Solid.Macromolecules 2015, 48, 1795−1803.(26) Koga, T.; Jiang, N.; Gin, P.; Endoh, M. K.; Narayanan, S.; Lurio,L. B.; Sinha, S. K. Impact of an Irreversibly Adsorbed Layer on LocalViscosity of Nanoconfined Polymer Melts. Phys. Rev. Lett. 2011, 107,225901.(27) Forrest, J. A.; Dalnoki-Veress, K.; Stevens, J. R.; Dutcher, J. R.Effect of Free Surfaces on the Glass Transition Temperature of ThinPolymer Films. Phys. Rev. Lett. 1996, 77, 2002−2005.(28) Keddie, J. L.; Jones, R. A. L.; Cory, R. A. Interface and surfaceeffects on the glass-transition temperature in thin polymer films.Faraday Discuss. 1994, 98, 219−230.(29) Kim, S.; Hewlett, S. A.; Roth, C. B.; Torkelson, J. M.Confinement effects on glass transition temperature, transitionbreadth, and expansivity: Comparison of ellipsometry and fluorescencemeasurements on polystyrene films. Eur. Phys. J. E: Soft Matter Biol.Phys. 2009, 30, 83−92.(30) Kim, S.; Roth, C. B.; Torkelson, J. M. Effect of nanoscaleconfinement on the glass transition temperature of free-standingpolymer films: Novel, self-referencing fluorescence method. J. Polym.Sci., Part B: Polym. Phys. 2008, 46, 2754−2764.(31) Roth, C. B.; Dutcher, J. R. Glass transition temperature of freely-standing films of atactic poly(methyl methacrylate). Eur. Phys. J. E: SoftMatter Biol. Phys. 2003, 12, 103−107.(32) Paeng, K.; Richert, R.; Ediger, M. D. Molecular mobility insupported thin films of polystyrene, poly(methyl methacrylate), andpoly(2-vinyl pyridine) probed by dye reorientation. Soft Matter 2011,8, 819−826.(33) Ediger, M. D.; Forrest, J. A. Dynamics near Free Surfaces andthe Glass Transition in Thin Polymer Films: A View to the Future.Macromolecules 2013, 47, 471−478.(34) Priestley, R. D.; Ellison, C. J.; Broadbelt, L. J.; Torkelson, J. M.Structural Relaxation of Polymer Glasses at Surfaces, Interfaces, and InBetween. Science 2005, 309, 456−459.(35) Bal, J. K.; Beuvier, T.; Unni, A. B.; Chavez Panduro, E. A.;Vignaud, G.; Delorme, N.; Chebil, M. S.; Grohens, Y.; Gibaud, A.Stability of Polymer Ultrathin Films (