Langmuir 2001, 17, 5865-5871
5865
Surface Transitions by Shear Modulation Force Microscopy Y. Pu, Shouren Ge, M. Rafailovich,* and J. Sokolov Department of Materials Science and Engineering, SUNY at Stony Brook, Stony Brook, New York 11794-2275
Y. Duan and E. Pearce Department of Chemical Engineering & Chemistry, Polytechnic University, Brooklyn, New York 11201
V. Zaitsev and S. Schwarz Department of Physics, Queens College of CUNY, Flushing, New York 11367 Received November 21, 2000. In Final Form: June 20, 2001 With the increasing importance of thin film in various applications, there is a need for new techniques with high surface sensitivity to measure physical properties. In this paper, we report results using a recently developed technique based on atomic force microscopy, temperature-dependent shear modulation force microscopy (SMFM), to investigate the surface glass transition. We test the effects of pressure under the tip, modulation frequency, and driving amplitude, which have been the subject of some controversy. The glass transition measurements on polystyrene and poly(methyl methacrylate) with different sample geometries demonstrate that the active volume probed by this technique has lateral dimensions on the order of the tip-sample contact radius. Applications to thin film glass transition measurements and surface segregation in long-chain/short-chain blends demonstrate the general utility of this technique.
1. Introduction Modern technology has brought thin film applications into everyday life. When the surface-to-volume ratio become very large, the surface properties can dominate film behavior. Consequently, there is a great demand for novel techniques to characterize surface properties such as the glass transition, molecular mobility, spreading and flow dynamics, and mechanical response. Among the existing surface characterization instruments, scanning force microscopy (SFM) is one of the most commonly used techniques. In this paper, we investigate the use of SFM to determine surface transitions. There are a variety of approaches available to probe the glass transition temperature (Tg) of thin films by SFM. Most of them rely on changes in viscoelastic properties as a function of temperature and/or frequency. Four different types of SFM models have already been reported as the following. (1) Scanning viscoelasticity microscopy (SVM);1-4 The phase lag between a small sinusoidal perturbation (applied normally to the sample surface) and the cantilever response signal is recorded. Tg is determined as the onset of change in the phase as a function of temperature. (2) Lateral force microscopy (LFM):5-9 The sample is scanned at low load and the lateral force is measured as * To whom correspondence should be addressed. (1) Kajiyama, T.; Tanaka, K.; Ohki, I.; Ge, S. R.; Yoon, J. S.; Takahara, A. Macromolecules 1994, 27, 7932-7934. (2) Tanaka, K.; Taura, A.; Ge, S. R.; Takahara, A.; Kajiyama, T. Macromolecules 1996, 29, 3040-3042. (3) Kajiyama, T.; Tanaka, K.; Takahara, A. Macromolecules 1997, 30, 280-285. (4) Satomi, N.; Takahara, A.; Kajiyama, T. Macromolecules 1999, 32, 4474-4476. (5) Tanaka, K.; Takahara, A.; Kajiyama, T. Macromolecules 1997, 30, 6626-6632.
a function of the scan rate or the temperature. The scan rate and temperature dependence of the mean value of frictional distribution is correlated to the known glass transition and/or secondary relaxation mechanisms. (3) Stiffness10 and adhesion force11 method: The stiffness and adhesion of the polymer is determined by the slope and pull-off force from the force vs displacement curves. Stiff cantilevers with a deflection sensitive interferometer are used. Tg is defined as the temperature at which the slope and adhesion force changes. (4) Shear modulation force microscopy (SMFM):12 This is described in this paper. A sinusoidal drive signal is applied to the x-piezo to induce a small oscillatory motion of the tip parallel to the sample surface. Below Tg, the response signal does not change significantly as the temperature increases. Above Tg, the response signal increases dramatically. Previous measurements of Tg by the above techniques have produced a variety of contradictory results. A large depression of Tg on polystyrene (PS) free surfaces with molecular weight lower than 140K was observed using (6) Tanaka, K.; Jiang, X. Q.; Nakamura, K.; Takahara, A.; Kajiyama, T.; Ishizone, T.; Hirao, A.; Nakahama, S. Macromolecules 1998, 31, 5148-5149. (7) Kajiyama, T.; Tanaka, K.; Satomi, N.; Takahara, A. Macromolecules 1998, 31, 5150-5151. (8) Kajiyama, T.; Tanaka, K.; Takahara, A. Polymer 1998, 39, 46654673. (9) Hammerschmidt, J. A.; Gladfelter, W. L.; Haugstad, G. Macromolecules 1999, 32, 3360-3367. (10) Gracias, D. H.; Zhang, D.; Lianos, L.; Ibach, W.; Shen, Y. R.; Somorjai, G. A. Chem. Phys. 1999, 245, 277-284. (11) Tsui, O. K. C.; Wang, X. P.; Ho, J. Y. L.; Ng, T. K.; Xiao, X. D. Macromolecules 2000, 33, 4198-4204. (12) Ge, S. R.; Pu, Y.; Zhang, W.; Rafailovich, M.; Sokolov, J.; Buenviaje, C.; Buckmaster, R.; Overney, R. M. Phys. Rev. Lett. 2000, 85, 2340-2343.
10.1021/la001619h CCC: $20.00 © 2001 American Chemical Society Published on Web 08/24/2001
5866
Langmuir, Vol. 17, No. 19, 2001
SVM and LFM by Kajiyama and co-workers.1-8 The shifts in Tg were attributed to chain end segregation, reduced entanglement, and decreased segment size of molecular motion due to the existence of the free surface. Similar shifts to lower Tg values were also found by Hammerschmidt et al. with temperature-controlled FFM9 on thinner PS films of lower molecular weight than in the above-mentioned studies by Kajiyama et al. Schmidt et al.13 studied the wear patterns induced by scanning under a large load and temperatures above Tg, where the tip penetrates to the silicon substrate through a 20 nm thick PS film of Mw ) 110K. By analyzing the temperature dependence (in the range of 415-433 K) and frequency dependence of the shapes of the wear patterns, Schmidt et al.13 deduced WLF parameters that differed from the bulk values. They ascribed the difference to a shift in Tg, which was estimated to be 386K. In comparing this to an earlier thin film measurement of Tg of 360 K by Keddie et al.,14,15 they attributed the difference of 26 K to the high pressure (84 MPa) under the tip, the induced change being consistent with bulk Tg measurements on samples subject to hydrostatic pressure. Gracias et al.10 reported measuring Tg of polypropylene surfaces via the temperature dependence of the slope of force-vs-distance curves. For tip radii of 1000 nm, they reported bulklike values of Tg, while, for radii of ∼50 nm, 20 K higher values of Tg were found. As in the experiment of Schmidt et al., they attributed the difference to increased pressure under the sharper tips. In the LFM mode, a very low scan velocity (about 100 nm/s) was used. In adhesive wear experiments with SFM it was found that ultralow velocity scanning can induce plastic deformation of the sample surface.13,16,17 Amorphous polymer systems are very likely to be plastically deformed by the sharp SFM tip. The degree of deformation depends on the materials properties (Mw, etc.), the scanning conditions (load and scan velocity), and external conditions (temperature and pressure). In LFM measurements, it is likely that, at a critical elevated temperature and low scan velocity, adhesion occurs and friction increases. Any decrease in scan velocity will cause a stepwise increase in the frictional value.17 It can be expected that adhesive wear occur tens of degrees below the glass transition temperature. That would concur with the onset of frictional change observed for high Mw and 200 nm thick PS film5-8,13 and could resolve the controversy of a shifted Tg to lower temperature. The optimal method would be one that directly measures a relative change in polymer properties associated with Tg but is independent of all of other experimental factors. In this paper, we describe the technique of shear modulation force microscopy (SMFM) and demonstrate that it is accurate and sensitive to surface changes in Tg and independent of variables such as load, modulation frequency, and scan amplitude over a substantial range. We describe in detail the principle of the operation and the calibration method and try to determine the active volume probed. Finally, we conclude by illustrating some general applications. 2. Materials and Instrumentation 2.1. Substrate and Sample Preparation. The polished silicon wafers used in this study were first cleaned in chromic(13) Schmidt, R. H.; Haugstad, G.; Gladfelter, W. L. Langmuir 1999, 15, 317-321. (14) Keddie, J. L.; Jones, R. A. L.; Cory, R. A. Europhys. Lett. 1994, 27, 59-64. (15) Keddie, J. L.; Jones, R. A. L.; Cory, R. A. Faraday Discuss. 1994, 219-230. (16) Overney, R.; Meyer, E. MRS Bull. 1993, 18, 26-34. (17) Overney, R. M.; Takano, H.; Fujihira, M.; Meyer, E.; Guntherodt, H. J. Thin Solid Films 1994, 240, 105-109.
Pu et al. Table 1. Glass Transition Temperature Measurement Results (K) for PHEPS and PVPy by DSC and SMFM PEPS PVPy bulk (DSC) surf (SMFM) a
313 315
PVPy (50%) + PEPS (50%) blend calcd result
423 426
368 353a
360 345
Annealed at 383 K for 3 days.
sulfuric acid solution to remove organic contaminants followed by a HF treatment. Monodisperse polystyrenes (PS) (Mw/Mn < 1.1) and poly(methyl methacrylate) (PMMA) (Mw/Mn < 1.05) of various molecular weights were dissolved in toluene and spuncast onto these wafers. The film thickness which varied from 8 to 500 nm, as measured by ellipsometry, was controlled by the concentration of the solution. Bilayer samples were made by spinning the bottom layer directly on the silicon wafer. The second layer was spun cast on a glass slide and floated onto the first layer from a deionized water bath. Surface segregation samples were made by blending the mixtures (low-Mw PS and high-Mw PS/poly(4-ethenylphenolmethylsiloxane) (PEPS) and poly(vinylpyridine) (PVPy)) as shown in Table 1 in toluene solutions and spinning on a clean silicon wafer. These films were then annealed at 413 and 383 K, respectively, until equilibrium was reached. 2.2. Shear Modulation Force Microscopy. Shear modulation force microscopy (SMFM) experiments were done using a Dimension 3000 (Digital Instruments, Santa Barbara, CA) equipped with a standard 100 µm XY and 10 µm Z piezo scanner. A schematic diagram of the electrical connection is given in Figure 1. To quantify the results, a commercially available etched silicon contact probe (Digital Instruments) consisting of an integrated single-crystal silicon cantilever and tip was used. The cantilever has a spring constant of 0.02-0.1 N/m. The vertical deflection of the cantilever is measured by a position-sensitive four-segment photodiode (PSD) and fed into the feedback loop that controls the contact force. The x-piezo was modulated by applying a sinusoidal voltage from a function generator. The modulated lateral force was detected by the lateral deflection signal of the cantilever generated by the PSD and fed into a dual-phase lockin amplifier. The reference signal used was the sinusoidal signal from the function generator. As the sample temperature was increased, the output signal of amplitude and phase were fed into the computer and recorded. The sample was mounted on a heating stage (MMR Technologies model R2700-2) where the temperature could be varied from 250 to 430 K with stability (0.05 K. The SMFM head and the heating stage were located in a sealed glovebox, which was purged with dry nitrogen to reduce capillary condensation ( Tg soft surface).
region of microslip can extend toward the center of the contact until the force of limiting friction is exceeded and the lateral deflection will decrease as the tip begins to slide. Figure 5 shows the relation between the drive signal amplitude and the response signal amplitude ∆X for a typical load applied on the tip. A linear relation can be seen clearly from both T < Tg and T > Tg. This means that no sliding of the tip occurs over the entire range of drive signals.24,25 For a given measurement at a fixed driving amplitude, the lateral deflection of the tip, ∆X, is the largest when the slip is a minimum. When the tip is in contact with the surface, the degree of slip decreases as the adhesion force or contact area between the sample and the tip increase. Since polymers are viscoelastic, we expect a time-dependent creep component to the tip response. The effect of creep can be measured by observing ∆X as a function of time at constant temperature. In Figure 6, we plot ∆X vs time at temperatures T ) 310, 385, and 420 K for a 200 nm thick film of PS (Mw ) 1230K). From the figure we can see that for a given molecular weight there is almost no creep for T < Tg where the internal relaxation times are very long, whereas for T ∼ Tg we see very slow creep such that it takes nearly 20 min to reach (23) Lantz, M. A.; Oshea, S. J.; Welland, M. E.; Johnson, K. L. Phys. Rev. B: Condens. Matter 1997, 55, 10776-10785. (24) Carpick, R. W.; Ogletree, D. F.; Salmeron, M. Appl. Phys. Lett. 1997, 70, 1548-1550. (25) Overney, R. M.; Buenviaje, C.; Luginbuhl, R.; Dinelli, F. J. Therm. Anal. 2000, 59, 205-225.
Surface Transitions by SMFM
Figure 6. ∆X vs time curves for PS of Mw ) 1230K thick film at different temperatures T ) 310, 385, and 420 K.
the saturation amplitude. For T . Tg the relaxation times are very fast and the amplitude saturates in less than 1 min. These data are consistent with the findings in ref 12, where we showed that the relaxation times or creep rate scaled with the molecular weight of the polymer or the viscosity. Hence the precise nature of the slope of the ∆X vs time curves depend on the details of the creep function at a given temperature since for polymers of high viscosity the ramping interval of 1 min is clearly not sufficient to reach the saturation amplitude for that temperature. Hence, in principle, one can extract information on the temperature dependence of the viscosity as a function of film thickness, etc., from the slopes of the data. However, since the contact area and adhesion may also be changing in an undetermined manner, this type of analysis is very difficult and will not be discussed any further here. This is a rough analysis showing the qualitative relationship between the input parameters, material properties, and observables. A more rigorous analysis would also consider time dependence by determining the full viscoelastic response of the material on the penetration of the tip, including the complicated interfacial behavior of the tip-surface contact. However, for these experiments, even these crude estimates are sufficient since we are only concerned with the change in ∆X rather than its magnitude or precise scaling relation. Hence, this method allows for a clear measure of Tg despite a lack of precise knowledge of tip-surface interaction and sample response. The contact area, tip shape, and tip surface will all vary from run to run and thus influence the absolute values of ∆X but not the position of the kink. (b) Active Volume Probed by SMFM. As mentioned in the previous section, we can observe creep, but we do not observe the deformation caused by the tip directly. Hence, we must detect indirectly the volume probed by this method. This was accomplished by making bilayer films where the Tg of the outer layer is higher than the sublayer, as shown in the inset of Figure 7. In our case, the outer layer, PMMA (Tg ) 400 K), which is in contact with the tip, has a higher Tg than the sublayer, PS (Tg ) 375 K). In Figure 7, we plot ∆X vs temperature for a series of bilayer films where the thickness of the PMMA layer is varied from 40 to 100 nm while the PS is fixed at 100 nm. From the figure, we can see that both Tgs are sensed until the top layer thickness is greater than about 65 nm. The penetration depth of the deformation zone under the tip can be estimated from the Hertzian model19 for a rigid sphere in contact with an elastic plane to be approximately equal to the diameter of the contact region. Below the Tg of both films, the contact radius is very small, ∼1 nm, and the ∆X response is small. Above the lower Tg, if the two
Langmuir, Vol. 17, No. 19, 2001 5869
Figure 7. Active volume probed by SMFM of double layer (PMMA with Mw ) 247K on PS with Mw ) 697K) sample with various PMMA thicknesses.
Figure 8. Amplitude vs temperature for PS (Mw ) 697K) on PMMA (Mw ) 247K) with different PS thicknesses.
layers are treated as springs in series with 1/Etotal ) 1/E1 + 1/E2, the Etotal will be determined mostly by the lower modulus molten layer (here PS) and the deformation zone will be large, ∼rtip. If rtip g thickness of the outer layer, touter layer, then the second Tg will be sensed as the temperature is increased, as observed. For top layers much thicker than rtip, equal to about 40 nm for our experiments, we expect not to be sensitive to the lower layer Tg, consistent with the results of Figure 7. When the temperature reaches the Tg of the sublayer, the indentation of the tip will increase due to the decrease of the total elastic modulus. ∆X will then increase as a result of larger contact area. As the temperature is further increased, the outer layer (PMMA) goes through its Tg and ∆X will experience a second kink as a result of the sudden decrease of E2 and Etotal. One can also probe the active volume when the Tg of the outer layer is lower than that of the sublayer. The minimum obtainable outer layer thickness of 17 nm was determined by the need to float this layer onto the first layer. Another method using the spin casting technique will be discussed later. Figure 8 plots ∆X vs temperature for a bilayer sample where the PS thickness is varied and the PMMA thickness is fixed at 100 nm. From the figure, we can see only one Tg ) 377 K corresponding to that of bulk PS is detected in the two upper traces. The lower trace with Tg ) 400 K corresponds to placing the tip on the uncovered PMMA substrate. In the bilayers, when T > Tg of PS, the ∆X response saturates, corresponding to the tip being nearly pinned at the surface. Since this occurs before Tg of PMMA is reached, the higher Tg is not sensed.
5870
Langmuir, Vol. 17, No. 19, 2001
Pu et al.
Figure 9. Amplitude vs temperature for PMMA (Mw ) 247K) on PS (Mw ) 90K) with different PS thicknesses.
3.2. Applications. (a) Tg Measurements of Thin Films. The Tg of PS films as a function of film thickness has been studied using numerous techniques including SMFM.5 The value of Tg measured on a substrate was postulated to be a combination of effects from a reduced Tg layer with excess free volume at the free surface and a pinned layer of higher Tg due to the interaction between polymer and substrate by annealing. The exact combination of these effects was unknown. To eliminate the “free surface” effect and taking advantage of the ability to detect the Tg of a buried layer using this technique, we detect the Tg of PS as a function of film thickness using the PMMA/PS/silicon bilayer sample geometry. The advantage of this method is that the PS is spun cast directly on the silicon and no limit on film thickness is imposed due to floating. Figure 9 plots ∆X vs temperature for the PMMA/PS/silicon samples with PS thickness ranging from 0 to 24 nm while the PMMA thickness is fixed at 45 nm. We can see that Tg ) 377 K, the bulk value of PS, and it is independent of PS film thickness. Hence, this supports the conclusion of a previous paper5 that the surface Tg is not different from its bulk value under the condition investigated. (b) Detection of Surface Segregation. (1) Surface Segregation of Low-Mw PS. When short chains are blended with chemically identical long chains, it is well established that the shorter chains prefer to segregate to the free surface.26 This is consistent with numerous theoretical models that have shown that the conformational entropic penalty at an interface is less for the shorter chains. It is also well-known that as the polymer chain length decreases, the concentration of chain ends increases and the excess of free volume results in a decrease of Tg.27 Therefore, by detecting the Tg at the surface of a film, we can detect if segregation has occurred. There are several empirical relationships proposed to relate the glass transition temperature of compatible blends to their composition, e.g., Fox, Gordon-Taylor, Couchman, and Lu-Weiss equations.28 Here the Tg of the blend is calculated from the Fox relation
1/Tg ) WA/TgA + WB/TgB
(4)
(26) Hariharan, A.; Kumar, S. K.; Russell, T. P. Macromolecules 1990, 23, 3584-3592. (27) Sperling, L. H. Introduction to physical polymer science, 2nd ed.; Wiley: New York, 1992. (28) Alfrey, T.; Miller, R. L.; Boyer, R. F.; Rieke, J. K. Turner Alfrey Symposium: milestones and trends in polymer science and technology: a tribute to Turner Alfrey; Dow Chemical Co., Michigan Molecular Institute; Wiley: New York, 1985.
Figure 10. ∆X vs temperature for the blend of short-chain (Mw ) 2.9K) and long-chain polymers (Mw ) 900K) before and after annealing.
where WA and WB refer to the weight fraction of components A and B, respectively. TgA, TgB, and Tg are the glass transition temperatures of components A and B and the blend. In the absence of segregation, addition of a 10% low-Mw PS (Mw ) 2.9K, Tg ) 342 K) into high-Mw PS (Mw ) 900K, Tg ) 376 K) yields Tg ) 372 K which is close to the Tg of the high-Mw component. The lower curve in Figure 10 is a plot of ∆X vs temperature for an unannealed sample. The measured value, Tg ) 370 K, is slightly below the calculated value and indicates that some segregation may have occurred in solution. The sample was then annealed at 413 K for 4 h in a vacuum of 10-3 Torr so that the chains reach equilibrium. The upper curve in Figure 10 shows the SMFM result for the annealed sample. Two transitions can be seen in the figure. The first corresponds to Tg ) 349 K, which is close to the glass transition temperature of the shorter chain polymer. The second transition corresponds to Tg ) 370 K, which is the same as the bulk value for the blend without segregation. These results show that a low-Tg layer has formed on the surface due to the segregation of the shorter chains to the film surface. A significant advantage of the SMFM measurement of the short chain segregation is that no potentially disturbing chain labeling is necessary, as in the deuterium labeling used for neutron scattering. Previous experiments and theoretical results29 have shown that for T . Tc, where Tc is the critical temperature, the thickness of the segregated layer is no larger than approximately 1-2 Rg of the segregated species. Hence we can estimate the thickness of the low-Tg layer on the free surface is about 15-30 Å, which gives a limit on the detection sensitivity of the technique when the Tg of the surface layer is less than the Tg of the lower layer. (2) Surface Segregation of PEPS. To determine whether we can quantify the degree of surface segregation from the measured value of Tg, we compared the results of SMFM to that of SIMS and XPS obtained from a miscible polymer blend. The polymers chosen are poly(4-ethenylphenol-methylsiloxane) (PEPS) and poly(vinylpyridine) (PVPy), which were synthesized at Polytechnic University.30 These polymers interact with each other via hydrogen bonding and are completely miscible for all compositions. This is confirmed by differential scanning (29) Zhao, X.; Zhao, W.; Sokolov, J.; Rafailovich, M. H.; Schwarz, S. A.; Wilkens, B. J.; Jones, R. A. L.; Kramer, E. J. Macromolecules 1991, 24, 5991-5996. (30) Duan, Y.; Pearce, E.; Kwei, T.; Hu, X.; Pu, Y.; Rafailovich, M.; Sokolov, J. Submitted for publication in Macromolecules.
Surface Transitions by SMFM
calorimetry (DSC) results for a symmetric mixture of PEPS and PVPy, which shows only one Tg. Since the surface energy of PEPS is smaller than that of PVPy, we expect enrichment of PEPS at the vacuum interface. Previous Tg measurements of PEPS, PVPy, and a 50/50 blend by DSC are shown in Table 1. Inserting the values of Tg for PEPS and PVPy obtained from DSC and the volume fraction, φ ) 0.5, into eq 3, we calculate Tg ) 360 K for the blend which is close to the measured value, Tg ) 368 K. The SMFM results are tabulated in Table 1 for the PEPS and PVPy homopolymers and the symmetric blend of the two components. From the table, we can see that Tg of each component is nearly identical to the value obtained from DSC. After annealing the blend at 383 K for 3 days in a vacuum of 10-3 Torr, we find that Tg of the blend decreases by 15 K as compared with the DSC result, indicating that preferential surface segregation of PEPS has occurred. The surface composition after annealing was measured by X-ray photoelectron spectrometers (XPS) and dynamic secondary ion mass spectrometry (DSIMS). The results show a surface enrichment layer with φPEPS ) 0.67 and φPVPy ) 0.33. Substituting the value into eq 3, we obtain Tg ) 345 K. This value is in reasonable agreement with the value of 353 K measured by SMFM. 4. Conclusion The principle of operation of the SMFM technique is described. We show that this method senses Tg from
Langmuir, Vol. 17, No. 19, 2001 5871
changes in the elastic modulus near the surface and the results are independent of modulation frequency, normal force, and driving amplitude over a substantial range. The SMFM technique was sensitive enough to detect a distinct Tg in segregation layers as thin as 2 nm. When Tg of the outer layer in a bilayer sample is higher than that of the sublayer, an effective zone of deformation on the order of the tip contact radius was observed and Tgs of both layers were measured if this zone includes the sublayer. Several applications of this technique are illustrated: (1) Tg was measured and found to be bulklike for a PS layer (>8 nm) sandwiched between two non attractive interfaces. (2) Surface segregation of short chains in a matrix of long chain was detected. (3) The Fox equation was used to calculate the amount of surface segregation in a compatible blend, and the volume was found to be in good agreement with concentration profiles detected by XPS and DSIMS. Acknowledgment. Support from the NSF MRSEC program (Grant No. DMR0080604) is gratefully acknowledged. The authors thank R. M. Overney from U. Washington for some useful discussion. LA001619H
