The effect of chemical structure on the stability of physical vapor deposited glasses of 1,3,5-triarylbenzene

Amorphous organic materials have important applications in pharmaceutical and organic electronics industries. However, their widespread use has been limited by the changes in properties due to aging,¹ crystallization,² or dewetting.³ Unlike their crystalline counterparts, organic glasses lack thermal and kinetic stability.^4–6 Over the past few years, physical vapor deposition (PVD) has been demonstrated as an effective method to prepare organic glasses with exceptional properties including high density,^7–10 high thermal stability,^11,12 low water vapor uptake,¹³ and high kinetic stability.^14,15 Glasses that manifest the above properties are denoted as stable glasses (SG).¹⁶ Many remarkable characteristics of SGs, such as low enthalpy,^11,14 low heat capacity,^12,15 and high density imply that SGs are in the near-equilibrium states at temperatures well below their glass transition temperatures (T_g). Other properties, including transformation growth fronts^17,18 and the suppression of tunneling two-level systems,^19,20 even resemble properties of crystals, suggesting that SGs are well-packed structures. These exceptional properties make SGs excellent candidates for applications in pharmaceuticals,²¹ organic electronics,²² nanoimprint lithography, and tip-based nanomanucturing.³ 

Physical aging can also lower the position of a glass on the energy landscape and lead to improved properties. However, aging is slow and the degree of density change and enhancement of stability is not comparable to those prepared by PVD within accessible laboratory time scales.^7,11,16,23 The largest reported density increase due to aging is around 2% in a million-year-old amber,^24,25 while PVD organic glasses prepared within an hour can yield density changes as large as 1.5%.⁸ Other glass preparation methods, such as cooling or quenching a liquid, cannot produce glasses with nearly the same degree of stability as PVD glasses.¹⁴ It has been hypothesized, but not systematically studied, that the accessibility to such low energy states is due to enhanced surface mobility during PVD.^26–28 During vapor deposition, molecules at the free surface may be able to adopt more stable conformations before being buried by the incoming flux of molecules. Deposition rate and the substrate temperature (T_dep) then become the two most important parameters to control properties of PVD glasses.^10,11,14,29 The deposition rate dictates the time spent at the free surface during which each layer of molecules can rearrange.²⁷ The deposition temperature (T_dep) affects the relaxation time at the free surface and the energy difference between glass and equilibrium super-cooled liquid (SCL) states. Glasses with various degrees of stability can be prepared by modifying these deposition parameters.^10,11,29,30

To date, many compounds including metallic,^31,32 polymeric,³³ and organic^12,30,34 materials have been shown to form stable glasses upon PVD. Isomers³⁵ and mixtures³⁶ of organic SG formers also shown to form SGs. Simulations on both metallic and organic SGs have been reported to access SG states via enhanced surface mobility.^37,38 These studies indicate that SG formation may be universal to all glass-formers, but the criteria are not yet well defined. In order to elucidate the origins of SG formation, it is important to study detailed effects of various parameters, such as the fragility of SCL,³⁹ the bulk relaxation times, and the fragility of the mobile surface layer on the stability of PVD systems.

Among many choices of materials, small organic molecules are appealing because their chemical and physical properties can be easily tailored by modifying the substituents. α,α,β-Trisnaphthylbenzene (α,α,β-TNB) is a well studied molecule that provides a reliable frame of reference to systematically study the effect of chemical structure on SG formation.^8,16,34,40 We recently developed a versatile method to synthesize various structural analogues of TNB in order to probe the relationship between structure and ability to form SG.⁴¹ We also demonstrated that these TNB analogues are amenable to vapor deposition and can form amorphous layers at the deposition temperatures of interest. Herein, we explore the influence of chemical structure on the SG formation of these molecules. Our series includes α,α,β-TNB and four new compounds: 3,5-di(naphthalen-1-yl)-1-phenylbenzene (α,α-P), 9-(3,5-di(naphthalen-1-yl)phenyl)anthracene (α,α-A), 9,9′-(5-(naphthalen-2-yl)-1,3-phenylene)dianthracene (β-AA), and 3,3′,5,5′-tetra(naphthalen-1-yl)-1,1′-biphenyl (α,α,α,α -TNBP) as shown in Figure 1.

SGs have been characterized using various methods such as differential scanning calorimetry (DSC),^11,14 nanocalorimetry,^15,30,42 Brillouin light scattering,²⁹ dielectric relaxation,^39,43 and dilatometry using ellipsometry.^8–10 Ellipsometry is a precise method to study properties of thin films of glasses.^8,10,44,45 This method has been extensively employed to measure thickness,^46–48 as well as the index of refraction and birefringence of thin polymer and organic films.^45,49–51 The advantage of using ellipsometry is its ability to probe a wide array of structural and kinetic properties in a high-throughput manner,^10,47 which makes it possible to characterize a large number of molecules with small amounts of material. Here, ellipsometry is used to measure the thermal expansion coefficient and the transformation kinetics of as-deposited glasses. The stability of glasses were interpreted from the fictive temperature (T_f)⁵² and changes in the density as a function of T_dep with respect to T_g. T_dep-dependent stability was examined in-depth for three molecules with well-defined bulk T_gs; α,α-P, α,α,β-TNB, and α,α-A. At the deposition rate chosen for this study (0.2 nm/s), T_dep around 0.8-0.85 T_g resulted in the most stable structures. Under similar deposition conditions relative to T_g, α,α-P, and α,α,β-TNB, SGs show similar degrees of stability, while α,α-A SGs display a clear reduction in stability compared to the other compounds.

Silicon (100) wafers (Virginia Semiconductor, Inc.) with ∼1.4 nm native oxide layer were cut into 1 × 1 cm squares, cleaned with dry nitrogen gas on the polished side and were adhered to the copper sample holder using vacuum grease (Apiezon H). Physical vapor deposition was performed in a custom ultra-high vacuum chamber with a base pressure of less than 2.0 × 10⁻⁷ Torr. The temperature of the sample holder was controlled using an external thermoelectric temperature control setup mounted onto an aluminum flange, which was in thermal contact with the copper sample holder stage (Figure S1 of the supplementary material).⁵³ The compounds were loaded into an alumina crucible and were thermally evaporated. The input power for thermal evaporation was controlled using a DC power supply (TDK-Lambda Gen8-90-U). The deposition rate was manually controlled by the input power and monitored using a quartz crystal microbalance (Sycon Instrument STM-1). All PVD depositions reported here were carried out at a deposition rate of 0.20 ± 0.03 nm/s. Most of the films used in this study were prepared in a thickness range of 180-210 nm. Details of the vacuum chamber setup and additional deposition procedures are presented in the supplementary material.⁵³ Table SII lists all deposition temperatures and film thicknesses for various compounds. Due to the limited range of deposition temperature available in the custom chamber, two depositions of the α,α-P compound at low temperatures were carried out at the Center for Functional Nanomaterials, Brookhaven National Laboratory (BNL). These measurements are highlighted in green color in Table SII.⁵³ 

Dilatometry measurements on all vapor-deposited samples were performed on a spectroscopic ellipsometer (J.A. Woolam spectroscopic ellipsometer M-2000V) with 70° acquisition angle and 500–1600 nm wavelength range. The acquisition time was chosen to be 3 seconds with zone averaging enabled to prevent potential systematic errors due to polarizer misalignment. After an initial alignment, focusing optics were installed to reduce the size of beam spot to 30 μm. Calibrations with focusing optics were performed on silicon with thermally grown oxide films of known thicknesses. Each sample was adhered onto a temperature-controlled stage (Linkam THMSEL350V, 77 K to 623 K) using thermal paste (Arctic Silver Ceramic polysynthetic thermal compound), and was secured by two Teflon screws. During measurements, nitrogen gas was used to purge the humidity from the sample stage. The temperature of the sample was controlled using a Linkam temperature stage and controller (Linkam PE5/T95). Temperature profiles used for dilatometry measurement were written using temperature control and video capture software (Linkam Linksys32). Melting point of indium (Sigma Aldrich) was tested on the same stage to calibrate temperature, and the error was determined to be ±0.6K.

The film thickness and index of refraction were determined by fitting the measured ellipsometric angles Ψ(λ) (amplitude) and Δ(λ) (phase) at each temperature to a three-layer model; layer 1, a temperature-dependent model for optical properties of silicon, layer 2, a 1.4 nm native oxide layer, and layer 3, a Cauchy uniaxial anisotropic model for the PVD film (an example is shown in Figure S2).⁵³ In the spectroscopic range of the experiment (λ = 500–1600 nm), all compounds used in this study were transparent and therefore this model was suitable for determining the optical properties and film thickness. The anisotropic model fits the optical properties of the stable glass better than an isotropic model (Figure S2).⁵³ This is consistent with previous studies that indicate that PVD glasses are birefringent.^8,49,51,54 The relationship between index of refraction and wavelength in Cauchy model is described by Equation (1),

where $n x y λ$ is the index of refraction in the xy-plane and A and B are fit parameters. Since the anisotropy of the glass is small, a wavelength-independent value of anisotropy was used as a third fit parameter,

where n_z(λ) is the index of refraction normal to the plane of the sample. Overall, 4 fit parameters A, B, Δn_z, and film thickness, h, were used to fit 486 data points for Ψ(λ) and Δ(λ) in this spectral range. Figure S2(a)⁵³ shows the quality of the fit for an as-deposited 200 nm film of α,α-A deposited at T_dep = 277 K = 0.77 T_g. Figure S2(b)⁵³ shows the residual error between the fit and the data for both an isotropic model (⁠$Δ n z λ =0$⁠) and anisotropic Cauchy models for comparison.

After determining the properties of the as-deposited glass, dilatometry measurements were carried out by heating the sample at 1 K/min to a set temperature of 15–30 K above the ordinary glass (OG) T_g (shaded as orange in Figure 2) holding isothermally until the transformation into OG was complete (shaded as pink in Figure 2), and then cooling back to the initial temperature at 1 K/min to determine the ellipsometric T_g of the OG (purple and blue regions in Figure 2). It is worth noting that not all of the films were heated to high enough temperatures to observe the onset of transformation as performed in some other studies.^8,11 Instead, a set temperature above T_g was chosen to isothermally transform the glasses before cooling. This was done primarily to avoid dewetting and to minimize the number of experimental trials needed with limited amounts of synthesized compounds. Figure 2 shows the thickness change in a thermal cycle of a 204 nm TNB film described above. The solid black lines are the calculated thickness (Figure 2(a)) and the recorded temperatures (Figure 2(b)). During isothermal transformation above T_g, the film continues to expand and eventually reaches a constant thickness when transformation is complete. The blue shaded area highlights the OG formed when the SCL falls out of equilibrium at T_g upon cooling. The slopes of stable glass, SCL, and OG lines are used to determine the thermal expansion coefficient of these states.

Figure 3(a) presents the data in Figure 2 as film thickness as a function of temperature and highlights different stages of the transformation. By extrapolating the SCL line to intersect with the SG line, the fictive temperature, T_f, can be defined for each film. The fictive temperature is a measure of the degree of equilibrium reached during vapor-deposition.^9,10 For samples that had T_f lower than 303 K, the lowest reliable temperature accessible with this setup, the stable glass line was extrapolated to lower temperatures assuming a constant expansion coefficient. Density changes between the PVD glass and the ordinary glass were calculated according to Equation (3),

where h_stable(303 K) is the thickness of the as-deposited SG film at 303 K before transformation, and h_ordinary(303 K) is the thickness of the transformed (OG) film at 303 K. This equation is valid under the assumption that expansion in the xy plane is negligible due to a small expansion coefficient of silicon substrate. Since both OG and SG are out-of-equilibrium at the experimental temperatures and their expansion coefficients are much larger than that of the substrate, these glasses may be under compressive or tensile stress depending on T_dep. However, no experimental data exist for the Poisson’s ratio of either SG or OG of these molecules, and the values are not expected to be significantly different in the two states.⁵⁵ As such, Equation (3) should present a reasonable measure of the relative density difference.

Figure 3(b) shows the calculated in-plane (n_xy) and out-of-plane (n_z) indices of refractions as a function of temperature. Under the deposition conditions described above, the as-deposited PVD film was anisotropic. The glass becomes isotropic upon transformation into OG. The slight residual anisotropy can be attributed to stress due to the mismatch between thermal expansion coefficients of silicon and glass. It is important to note that these stresses may also affect the expansion coefficient of the glassy films, but more reliable estimates are not permitted without a knowledge of the Poisson’s ratio^55,56 or the ability to vapor-deposit much thicker films. More details on the anisotropy of vapor-deposited films will be discussed in future publications.

It is important to note that during transformation, the three-layer model did not properly fit some data points. These data points are illustrated in Figure S3⁵³ as increased mean square error (MSE) during the transformation. We attribute this observation to the transformation growth fronts.^17,57,58 The layers behind the growth front may have different optical properties to an extent that a single-layer model is no longer suitable. Since the transformation fronts are not the focus of this study, we leave details of the fits during these stages to future studies. These data points are marked with open symbols throughout this manuscript.

Figure 4 shows the results of ellipsometry based dilatometry measurements on PVD films of α,α-P (a), α,α,β-TNB (b), and α,α-A (c) deposited at various deposition temperatures. Most film thicknesses are around 200 nm with the exceptions described in Table SII.⁵³ Based on Figure 4, T_g(α, α-P) = 328 ± 2 K, T_g(α, α, β-TNB) = 338 ± 2 K, and T_g (α, α-A) = 359 ± 1 K. These values are in good agreement with DSC measurements reported in another study from our group⁴¹ as 330 K, 343 K, and 360 K, respectively. The differences in T_gs between these two methods are primarily due to the different cooling rates used (10 K/min for DSC vs. 1 K/min for ellipsometry measurements). We have shown in another manuscript⁴¹ that ellipsometry measurements at 10 K/min yield values closer to those measured by DSC.

The data in Figure 4 show that all transformed OGs of each compound (with the exception of α,α-P films deposited at T_dep = 0.73 T_g) have similar apparent expansion coefficients and T_gs across different films. This is a strong indication that the transformed OGs are the same within experimental error. Therefore, the properties of SGs deposited at various temperatures can be directly compared with the common OG state. It is not clear why the transformed glass of α,α-P film deposited at T_dep = 0.73 T_g showed a higher apparent expansion coefficient. Unfortunately, these measurements could not be repeated due to the limitations in the amount of synthesized material and access to equipment.

All as-deposited glasses in the experimental range had higher density, as measured by lower initial thickness, compared to the transformed glass at the same temperature. Furthermore, most glasses showed increased kinetic stability and did not transform into ordinary glass until the temperature was raised well above T_g. As shown in Figure 4(a), the onset temperatures of the α,α-P glasses deposited at T_dep = 0.73 and 0.78 T_g were below the isothermal holding temperature, as observed by a rapid change in thickness at 340 K and 337 K, respectively. Similarly, for α,α,β-TNB (Figure 4(b)), the onset temperature of the film deposited at T_dep = 0.96 T_g was observed and the isothermal transformation temperature was lowered by 5 K to prevent dewetting. In Figure 4(c), isothermal holding temperature of α,α-A deposited at T_dep = 0.77 and 0.95 T_g glasses was also lowered by 5 K due to the appearance of the onset of transformation. In all the other films, the onset temperature was either above or equal to the isothermal holding temperature. These observations are in good agreement with previous experiments on α,α,β-TNB.⁸ Ellipsometry measurements were also performed on ∼200 nm films of β-AA (T_g = 391 K and T_dep = 0.85 T_g) and α,α,α,α-TNBP (T_g = 383 K and T_dep = 0.90 T_g) as shown in Figures S4 and S5 in the supplementary material,⁵³ respectively. Due to the limited amount of synthesized material, only one film was deposited for each molecule. Nevertheless, both molecules showed the ability to form SGs upon PVD, as shown by increased density, 1.65% for β-AA and 1.20% for α,α,α,α-TNBP, and lowered fictive temperature compared to T_g, T_f = 325 K = 0.83 T_g for β-AA and T_f = 354 K = 0.92 T_g for α,α,α,α-TNBP.

Figure 5 shows the relative fictive temperature, T_f/T_g, (Figure 5(a)) and changes in density, Δρ, (Figure 5(b)) obtained from data presented in Figure 2 for α,α-P, α,α,β-TNB, and α,α-A SGs. The horizontal error bars are estimated based on uncertainties in determining T_g and T_dep. The error in determining T_g is evaluated using the standard deviations of the fits to OG and SCL lines. The lower limit of this error is ±0.6 K, which is the uncertainty of the absolute value of temperature from the temperature stage. Uncertainties in T_deps were determined according to Tables SI and SII in the supplementary material.⁵³ The vertical error bars in Figure 5(a) are determined based on the uncertainty in T_g (as described above) and the uncertainty in T_f (standard deviations of fitted SG and SCL lines). The dashed line in Figure 5(a) has a slope equal to one, where T_dep = T_f. This line represents the lowest attainable fictive temperature at any deposition temperature. A glassy state with lower T_f corresponds to a lower energy state on the energy landscape, indicating higher stability. Figure 3(a) shows similar trends in stabilities of all three molecules. Glasses deposited near 0.95 T_g yield near equilibrium structures. The maximum stability for all three compounds is achieved around T_dep = 0.8 T_g–0.85 T_g. As the deposition temperature is reduced below T_dep = 0.8 T_g, less stable glasses are produced. These results are consistent with previous measurements on α,α,β-TNB and other organic compounds.^9,11,29 Interestingly, although the trends are similar among all three compounds, α,α-A produces less stable glasses at the same relative deposition temperature compared to the other two compounds beyond experimental error.

Figure 5(b) shows the percent change in densities upon transformation described by Equation (3) as a function of relative deposition temperature, T_dep/T_g. Vertical error bars are calculated based on the standard deviations in fits to the OG and SG lines. This is used as a measure of errors in the ellipsometry measurements instead of the reproducibility because it was not always possible to repeat the experiment at a given deposition temperature due to the limited amount of synthesized material. The dashed lines in this graph represent the extrapolated equilibrium line of each compound, where the density change is the maximum possible density increase to reach equilibrium at the deposition temperature. This line is calculated by connecting two points, one obtained by calculating the density change between the extrapolated SCL and OG at 303 K and the other by assuming that the density change is zero at T_g.¹⁰ The non-monotonic trend observed in fictive temperature measurements (Figure 5(a)) is also seen in the density increase as a function of relative T_dep/T_g. The densities are close to the equilibrium densities at deposition temperatures near T_dep = 0.95 T_g. The most stable glass, as measured by the highest initial density, is obtained around T_dep = 0.8 T_g–0.85 T_g. These observations are consistent with those previously reported for α, α, β-TNB.⁸ Again, we note that α,α-A shows a smaller degree of density change compared to the other two compounds at the same relative deposition temperature. Furthermore, the equilibrium density of α,α-A (blue dashed line) is higher at all relative T_deps, and therefore a larger driving force towards the equilibrium would be expected. This also indicates that α,α-A is farther away from equilibrium than the other two compounds at the same relative T_deps.

Figure 6 shows the apparent thermal expansion coefficients (α_SG) of the as-deposited SGs at corresponding substrate temperatures. The dashed lines represent the apparent expansion coefficient of the ordinary glass, α_OG, for each molecule.⁴¹ As discussed in another manuscript,⁴¹ the π-conjugated interactions are enhanced as the number of fused aromatic rings increases in the substituent, resulting in a more harmonic potential. As the potential becomes more harmonic, the apparent expansion coefficient of OGs is further decreased.⁵⁹ Each compound’s apparent α_SG decreases with increased stability of the glass. The lowest α_SGs are measured around T_dep = 0.8 T_g–0.85 T_g for each compound, indicating more harmonic potentials as the packing density increases. As such, the stability of SGs can also be evaluated from the changes in the apparent expansion coefficient. The percentage change in expansion coefficient of the most stable SGs of each compound compared to its OG is $Δα 0.83 T g α , α - P =28±3%$⁠, $Δα 0.86 T g α , α , β - TNB =21±1%$⁠, and $Δα 0.82 T g α , α - A =19±1%$⁠, respectively. Thermal expansion coefficients of α,α,β-TNB glasses deposited between 0.8 T_g and 0.9 T_g are close to that of the crystal value, 1.25 × 10⁻⁴/K.^60–62 

Within experimental error, the values reported here for 200 nm α,α,β-TNB films agree well with those reported by Dalal et al. for 500 nm films of the same compound.⁸ We note that both reported α_SG and α_OG are thermal expansion coefficients calculated in the direction normal to the film, under the condition where the expansion of the film in the xy direction is constrained by the silicon substrate, with a much smaller expansion coefficient. As such, the measured expansion coefficients are expected to be closer to the values of the bulk expansion coefficient rather than the linear expansion coefficients. The actual values of the expansion coefficient can be calculated from these measured apparent values of expansion coefficient with a knowledge of the materials’ Poisson’s ratio.⁵⁶ Since the Poisson’s ratios of neither the OGs nor the SGs of these compounds are known, the correction cannot be applied to the current measurements. However, we expect that for a given compound, the relative difference between α_SG and α_OG upon transformation be minimally affected by the corrections that would be in the same direction for both phases. We also note that although the fictive temperature and density measurements indicate that all vapor-deposited glasses are more stable than the OGs of the corresponding compound, the apparent thermal expansion coefficients of some SGs are higher than α_OG. As explained earlier, this may be due to differences in the compressive vs. tensile stresses on the α,α-P glasses deposited at temperatures well below room temperature. Careful studies on the mechanical properties of these glasses are needed to further verify this point.

The results of this study overall agree well with previous studies on stable glasses of various organic compounds.^8,11,40,42 All five structural analogues can form stable glasses at the deposition temperature around 0.8 T_g–0.85 T_g. When deposition rate is kept constant, the relative deposition temperature, T_dep/T_g, is a strong predictor of the PVD glass properties. Similar to the results in other studies, the properties of glasses deposited at temperatures close to T_g are limited by the equilibrium line.¹⁰ In this regime, dynamics of the surface is fast enough so that properties of the glass are defined by the thermodynamic properties of the super-cooled liquid (equilibrium).⁸ When the temperature is decreased, the glass stability is increased as lower energy states become available. Around T_dep = 0.8 T_g–0.85 T_g, the most stable structures are obtained. At deposition rates below 0.8 T_g, surface mobility slows down and the glass structure is kinetically trapped in an intermediate state. As the temperature is further decreased, the mobility at the free surface decreases even more and less stable glasses are obtained. The interplay between the equilibrium structure, which is defined by the bulk properties of a glass, and the dynamics at the free surface are two important variables in defining the most stable structure at a constant deposition rate. This non-monotonic trend is observed in both measurements of the fictive temperature and change in the density of the three compounds α,α-P, α,α,β-TNB, and α,α-A. We also note that for glasses prepared at these deposition temperatures, their structures are consistent with an amorphous structure aging towards the equilibrium line.

Although all three molecules are able to form stable glasses, α,α-A glasses have lower density (Figure 5(b)) and higher fictive temperature (Figure 5(a)) at the same T_dep/T_g compared to α,α-P and α,α,β-TNB glasses at T_deps below 0.9 T_g. In the thermodynamic driven regime, the equilibrium density of α,α-A is higher than both α,α-P and α,α,β-TNB, suggesting that the driving force to reach equilibrium should be larger. To rationalize the reduction in stability, we compared apparent fragility and estimated aging times of three glasses. Studies in the past have suggested that fragility may be a dominant factor that leads to stable glass formation.^63,64 However, a few recent works on strong liquids have proposed that the fragility of a liquid does not directly correspond to the stable glass forming ability.^27,39 In a recent study, we used cooling rate–dependent T_g measurements to obtain the fragility of these compounds.⁴¹ The values of fragility are reported in Table I. As shown in Table I, bulk fragility of α,α-A is similar to that of the other two compounds,⁴¹ which indicates that the bulk relaxation times around T_g for these three compounds are similar. Therefore, bulk fragility or relaxation times cannot explain the differences observed in the stability of the PVD films between α,α-A and the other two compounds.

To demonstrate the large difference among the degrees of stability of these three glasses, it is useful to estimate the required aging time for an ordinary glass to achieve the same density as the most stable glass. Using the reported values of fragility, along with an Arrhenius extrapolation to the deposition temperatures of interest, we estimate the aging time required to age the system to the same state. Table I shows the fragilities and the estimated age of the most stable glasses of three molecules obtained in this study.

Clearly, α,α-A stable glasses are not as “aged” as the other two glasses. Since the apparent bulk relaxation times at T_g for these glasses are similar,⁴¹ and the driving force towards equilibrium is stronger for α,α-A, we hypothesize that the surface mobility of α,α-A glasses must be slower than that of α,α-P and α,α,β-TNB at the same temperature relative to T_g. Stable glasses of a wide range of molecules have been formed in the past, but a clear relationship between stability and the molecular structure and size has been lacking.^8,29,30,33 Few studies have directly correlated molecular structure to surface mobility. Brian and Yu have demonstrated an example of molecular weight or structure effect on the enhanced surface mobility.⁶⁵ In their study, nifedipine has a much faster surface diffusivity than indomethacin despite sharing similar bulk relaxation times and dynamics. They hypothesized that under similar deposition conditions, nifedipine would make a more stable glass than indomethacin due to the fast surface relaxation. Both hypotheses need to be verified to elucidate the relationships between molecular weight/structure, surface diffusivity, and the stability of PVD glasses with direct studies of the diffusion coefficients at free surfaces. The tailorable molecules presented in this study provide a great framework for such studies. This will be a topic of future investigations.

The other notable differences between α,α-A and the other two molecules studied here are the larger substituent size and stronger π-conjugation in α,α-A. The strong interactions result in more harmonic potentials as shown in Figure 6. Extended π-systems and the larger size of the substituent could increase the barriers to rearrangement and rotation, resulting in enhanced crystallization rate, more harmonic potentials, and reduced mobility at the free surface. Examples of enhanced intermolecular interaction affecting surface mobility include indomethacin, which has a slower surface diffusion rate than nifedipine, possibly due to hydrogen bonding.⁶⁵ Some simulations have shown that polymers containing more rigid backbones experience a reduction in surface mobility, which would also be consistent with our observations.⁶⁶ Few studies systematically probe the relationship between the molecular level interactions and the enhanced surface mobility. It would be interesting to directly study such correlations in the future to understand the origins of enhanced mobility and the variables influencing it. Extended π-systems are also of great interest in organic light emitting systems. The “face-to-face” π-stacking has been suggested to be crucial in promoting charge mobility and improving device performance.^67–70 Understanding the correlation between the enhanced surface mobility and π-conjugation can help predict material properties at nanoscale, which would be relevant to these applications.

We have synthesized a series of organic molecular analogues of α,α,β-TNB using a simple method to selectively replace naphthyl substituents. We have shown that all analogues form stable glasses upon physical vapor deposition at temperatures around T_dep = 0.8-0.85 T_g. In-depth studies on three of these compounds, α,α-P, α,α,β-TNB, and α,α-A, deposited at T_dep between 0.73 T_g and 0.96 T_g show that compared to the ordinary glasses of the same molecules, all vapor deposited glasses have higher density, enhanced stability and lower fictive temperature. Most PVD glasses also have lower apparent thermal expansion coefficients.

Relative to T_g, general trends of stability among α,α-P, α,α,β-TNB, and α,α-A were similar. At T_deps close to T_g, the corresponding equilibrium super-cooled liquid limits the properties of these glasses. At T_deps well below T_g, the glassy structures are kinetically trapped in an out-of-equilibrium state. The most stable glasses of all of these compounds were produced around T_dep = 0.8-0.85 T_g.

Despite these similarities, the anthracyl containing compound, α,α-A, formed a less stable glass at the same relative temperature compared to α,α-P and α,α,β-TNB. We hypothesize that this is due to reduced mobility at the free surface of α,α-A. Direct experiments on free surfaces are needed to verify these observations. We hypothesize that the introduction of larger substituents with extended π-conjugation affects the mobility at free surface, which will be studied in the near future.

Z.F. acknowledges funding from the University of Pennsylvania and seed funding by MRSEC program of the National Science Foundation under Award No. DMR-11-20901 at the University of Pennsylvania. P.J.W. acknowledges funding from NSF (No. CHE-1152488). F.G. thanks the Chinese Scholarship Council for financial support. Physical vapor depositions were carried out in part at the Center for Functional Nanomaterials, Brookhaven National Laboratory, which is supported by the U.S. Department of Energy, Office of Basic Energy Sciences, under Contract No. DE-AC02-98CH10886.