Controlling surface morphology of spin coated epoxy composites using the Marangoni instability

2.1 Materials

EPON® Resin 828, is a diglycidyl ether of bisphenol-A (DGEBA) with an epoxy equivalent weight (EEW) of approximately 190 g/eq. The curing agent used was Epikure® 3140, a polyamide which contains less than 10 % of primary amine triethylenetetramine (TETA) and has an amine value of approximately 370 mg KOH/g with amine hydrogen equivalent weight (AHEW) of approximately 95. The stoichiometric equivalent parts hundred epoxy resin (PHR) is 50 g. The epoxy and curing agent chemicals and their associated data were obtained from Miller-Stephen Chemical Co. ³⁵ Bubble formation was reduced though not eliminated by processing under vacuum after shearing and mixing by degassing at a vacuum of greater than – 85,000 Pa for approximately five min.

Nanoclay Cloisite Clay 20A and 93A, with platelets of about 1 nm, were used in this study. Data sample specifications were provided by Southern Clays. All the clays were heated in a vacuum oven at temperatures over 110 °C for 48 h before being immediately stored in a desiccator until use. ^{36 , 37 , 38}

Substrates were copier transparency film IVR-65210 from Innovera_TM Technology Essentials. It is a poly(ethylene terephthalate) (PET) base approximate 100 μm in thickness with a proprietary acrylic co-polymer toner receptive coating. ³⁹

2.2 Equipment and material characterization

An IKA T25 digital Ultra-Turrax® High Shear (HS) was used for the intercalating and exfoliation of the clays in the epoxy resin. Samples were high sheared for 2 h unless specified differently below. A Laurell WS-400B Spin Coater with a four-inch chuck was used for spin coating thin films. Hand mixed epoxies were deposited manually after degassing onto the chuck. Dynamic mechanical analysis (DMA) of the epoxy composites were performed using a Q800 DMA (TA Instruments, Newcastle, DE) with a multi-strain recipe using a temperature ramp of 3 °C/min, a frequency of 1 Hz and a thin film/fiber clamp. Humidity control was controlled using a Suncore Humidifier in a 3.5 ft × 3.5 ft × 4 ft glove box, and the relative humidity (RH) was measured using an Omega CP.

2.3 Polymer and composite processing

3.1 Spin coater parameters impacting Marangoni cell formation

In order to test the impact of spin coating on the epoxy systems, 90 PHR samples were spin coated onto substrates at room temperature and then cured, as described in the methods section. Surface Figure 1(a) provides an image of a neat epoxy thin film. Due to its translucent nature, actual patterns visible to the eye are difficult to capture through a photograph. Concentration gradients developed through centripetal and elongation forces and the development of Marangoni cells can result in lower density cross-linking along Marangoni edges during spin coating. ^{42 , 43 , 44} So a sample was ‘rolled' in order to induce cracks to make them more visible under a microscope as can be seen in Figure 1(b). It should be noted that even though hexagonal formation in Marangoni-Benard convection is the likely outcome in uniform systems, due to other forces at play such as elongation, striation, and centripetal forces along with localized concentration gradient variation, surface deformation results in multiple polygons and shapes. ^{9 , 45 , 46}

Figure 1:

Marangoni surface morphology images of neat epoxy. In (a), photo is of thin film after removed from substrate. Due to translucent nature, the hexagonal pattern is very difficult to see. In (b), sample was ‘rolled’ to enhance cracking along edge of hexagons to improve microscope image of Marangoni effect.

The thickness of samples was measured after three days of cure. The Marangoni effect was visible to the naked eye at these conditions. The final thickness and percent error with respect to RPM is tabulated in Tables 1 and 2 below.

Emslie’s equation does not approximate the final film thickness due to Marangoni cell formation on the surface. To see if Washo et al. ²⁴ conditions were pertinent, these epoxy-amine systems samples were processed under conditions where one system would generate Marangoni cell formation and the other would inhibit formation for comparison. Samples prepared at 60 % ± 2 RH and spun for 40 s did not produce visible Marangoni cells while those prepared at 33 % ± 5 % RH and spun for 40 s did generate cells. Thicknesses were measured using a standard micrometer. For samples at 60 % RH, the difference between Emslie’s Equation prediction and measured values were 2.13 %, 7.9 %, and 20.7 % for RPM of 2,000, 4,000, and 6,000 respectively. Obviously, Emslie’s Equation did not approximate samples with Marangoni cells (33 % RH). Out of curiosity, if one measured the center thickness (taken as the difference of film thickness first measured with a standard micrometer and then substracting the thickness measured at the center of the Marangoni cell using a point micrometer) Emslie’s equation predictions were significantly improved to −8.2 %, +16.6 %, and 26.1 % error for RPM’s of 2,000, 4,000, and 6,000 respectively, which is on a similar order of magnitude error to samples without Marangoni cell formation. The comparison of the film thickness for samples is provided in Figure 2. From Emslie’s equation calculations, it would be expected that shear thinning polymers would overestimate the thickness, which is not observed. However, this is most likely due to the fact that Emslie’s equation assumes an initial constant film thickness which was approximated by the volume of sample deposited, divided by substrate surface area. However, system limitations required the deposit of sample onto the center of the spin coater, which then was initiated. The time required for the polymer to flow across the substrate before it is flung off could account for the underpredicted thickness instead of overprediction. For the Marangoni cell depth, final thickness measurements were fitted to Equation (4), for which n = −0.27 ± 0.01 and A = 14.03 ± 1.04. Both A and n are in excellent agreement with other similar published works. ^{31 , 47}

Figure 2:

Comparison of spin coated thickness at 40 s at different relative humidities and the Marangoni cell depth measurement (33RH_C) compared to Emslie’s (Equation (3)) approximation.

To better understand how RPM and spin time impact Marangoni cell formation of epoxy spin coated nanocomposites, neat and 2 wt% MMT samples were spun at different RPM. Figure 3 displays the impact on cell diameter, measured manually with a micrometer, as RPM was varied for both neat (Nxx) and 2 wt% 20A MMT composite (Cxx) epoxies spun for either 45 or 75 s. Lines are provided to guide the eyes. Further data provided in the Supplementary Material.

Figure 3:

Marangoni cell width dependence on RPM and spin time for 90 PHR cured neat (N) or 2 wt% MMT composite (C) epoxies spun for 45 or 75 s.

Except for the N75, the difference between 1,000 and 2,000 is within one standard deviation, which all have large ranges (a quarter of the average size in some instances). With 45 s spin time, there is less than 2.5 % change in cell dimensions between 1,000 and 2,000 RPM. As RPM or spin time increases, Marangoni cell width, in general, decreases to a limit of approximately one mm with one σ of approximately 0.3 mm. Additionally, the composite samples appear to result in larger cell formation when compared with the neat epoxies for a given spin time. The exact cause of the larger cell formation is uncertain. It might be due to the nanoparticles’ impact on flows due to differences in viscosities or the change induced by a concentration profile and surface tension.

When considering the temperature effect on the Marangoni cell formation, even if h_f was assumed to be 0.02 cm (thickness at initial substate coverage before draw down), and ∇T was assumed to be 2 K (high for this exothermic reaction), the temperature dependent Ma number <1. This is well below 80, indicating that temperature is not a significant driving force in the formation of Marangoni cells. The relatively thin sample thickness results in the temperature gradient having minimal effect. ^{5 , 15} As such, the concentration gradient is the primary driving force for the development of the Marangoni cells. ^{18 , 19}

To further explore the impact of changing composition on the Marangoni cell diameter, experiments investigating the role of extent of reaction and RPM and their impact on cell formation were conducted. The RPM was varied from 2,000 to 6,000, and for PHR 90 neat and composite samples. Samples were spun and placed in the vacuum oven and cured at 40 °C. Results are presented in Figure 4. As the extent of reaction increases, so does the Marangoni cell width. In prior work, after an initial decrease, viscosity increases as the extent of reaction progresses. ⁴³ This suggests Marangoni cell width has a strong dependence on the viscosity of the reacting solution. The lower RPM (2,000–4,000) had cell widths increase more than 200 %, while at 6,000 RPM, the neat samples increased only 60 % and composite samples only 23 % respectively. The greater the RPM, the smaller the change in cell width. Centripetal forces may play a role in the alignment of molecules and the movement of concentration gradients, impacting cell width. Shenk et al. ⁴³ demonstrated that cell edges have lower cross-linking bonding and could be etched away, leading to new composite structures and possibilities. Additionally, they demonstrated that post curing can effectively increase composite strength. It was hypothesized that this was due to the reverse reaction of carbamate back into its amine constituents, leading to additional cross-linking occurring. ^{43 , 48} Using their hypothesis, it could be suggested that future work might focus on controlling geometries and shapes to target multilayer composite systems. If etching is utilized to remove cell edges, then one could choose different filler materials with targeted properties such as strength or conductivity to produce composites to meet desired properties.

Figure 4:

Marangoni cell width dependence on reaction time for 2 wt% MMT composites (C) and neat (N) epoxies spun at 2,000 (2 k) 4,000 (4 k) and 6,000 (6 k) RPM.

The Marangoni cell can also be quantified by overall sample thickness. In Figure 5, the differences in sample thicknesses with respect to the extent of reaction for both neat and composite samples at 4,000 and 6,000 RPM are presented. The extent of reaction is measured as the time from the end of stirring until the epoxy is deposited on the substrate and spun. Samples deposited within the first 20 min of the reaction show a decrease in thickness up to 41 %. This might be explained in the nearly 10 % decrease in viscosity occurring during initial 20 min of a reaction before viscosity begins to increase as reported elsewhere. ^{43 , 49} After 40 min, the thickness of neat and composite samples spun at 4,000 RPM increases drastically. This was due to bubble entrapment visible to the naked eye at the edges of the Marangoni cells. The bubbles were induced during the stirring of the epoxy and curing agent and were not completely removed during degassing. After spinning, they are visible to the naked eye. Upon curing, most bubbles are removed from the surface of the epoxy at the macro level. However, the samples at 4,000 RPM with extent of reaction past 40 min, bubbles are not removed. At 6,000 RPM, there is sufficient centripetal forces to minimize bubble formation, which disappear during the curing in the vacuum oven.

Figure 5:

Marangoni cell formation for neat (N) and 2 wt% MMT composite (C) epoxy with respect to extent of reaction.

The impact of bubble formation and thickness variation with changes in RPM as a function of time are also shown in Figure 6. The data represented is for neat (N) and 2 wt% MMT 20A composites (C) at RPM settings of 2,000 (2 k) and 6,000 (6 k). The setting of the ramp rate, r1, resulted in the RPM set point being reached at the end of the spin time. The max ramp rate, r20, resulted in the RPM set point being reached nearly instantaneously. Bubble formation occurred for composites processed at r1, even at 6,000 RPM, because the samples at lower ramp rates did not spin for significant amounts of time at the higher RPM. Samples at 2,000 RPM all had deeper well depths and bubble entrapment for reasons explained above. Samples spun at 6,000 RPM but at r1 were thicker overall. This is due to an “effective RPM” significantly below 6,000, and thus well depths of the same magnitude as those spun at 2,000 to 4,000 RPM. ^{22 , 50}

Figure 6:

Film thickness dependence for neat (N) and 2 wt% MMT composites (C) at different ramp rates r1 (slow) and r20 (instantaneous) for 2,000 (2 k) and 6,000 (6 k) RPM.

The objective of the characterization of the spin coater was to develop a more fundamental understanding of how the operation of this instrument might impact material morphology and Marangoni morphology development. Marangoni cell width decreases as RPM is increased and increases as the extent of reaction increases. The Marangoni cell depth decreases with decreasing viscosity during the initial stages of curing. However, as the extent of reaction and viscosity of the system increases, well depth increases due to bubble formation for the lower spin rates developed around the cell edges.

3.2 Relative humidity, modulus and the Marangoni cell

The Marangoni instability not only influences the surface roughness, but it also can impact material properties. Shenk et al. ⁴⁵ demonstrated that Marangoni cell formation occurs at the two lower relative humidity conditions for this epoxy-amine system due to the reaction of the amine with CO₂ in the presence of water vapor resulting in carbamate. Carbamate formation lowers the surface tension, creating localized concentration gradients. As the RH increases, however, a saturation of carbamate formation across the surface results in a more uniform concentration resulting in a reduction of the surface tension gradient. Thus, the mechanism for the instability is removed, and uniform surfaces result. Neat, 1.0 wt% and 2.5 wt% 93A MMT epoxy samples were spin coated at low (22 % RH), mid (40 % RH) and high (67 % RH) humidity to analyze material properties under conditions when Marangoni cells are generated (22 and 40 %) compared to those which are not (67 %). Measurements of the materials’s storage modulus (G′) and glass transition temperature (T _g ) were recorded. Results are provided in Figure 7.

Figure 7:

Modulus of neat, 1 wt% and 2.5 wt% MMT composites with respect to humidity for a 90 PHR epoxy spin coated film.

The 2.5 wt% MMT composites’ storage modulus outperformed the neat and 1 wt% epoxy composite. This was anticipated, and is due to better partial exfoliation obtained from an increase in solution viscosity during high shear, enabling the shearing of the platelets and the larger numbers of clay platelets interacting with the epoxy and enhancement of composite modulus. ^{40 , 51 , 52 , 53} Figure is provided in the Supplementary Material as an example of partial exfoliation of the clay using our high shearing method. An in-depth analysis of the impact of high shear pre-mixing treatments on enhanced exfoliation and the improvement of mechanical properties for this epoxy-amine system is beyond the scope of this analysis and is given in Shenk et al. ^{45 , 48}

Both 1 wt% MMT and neat epoxy thin films were hard to remove from the substrate without tearing while peeling. This is indicative of lower storage modulus and curing conditions not optimal for obtaining the highest storage modulus possible, resulting in a greater likelihood of more compliant (lower crosslinking) material. Though higher crosslinking would occur from further curing at higher temperatures, this study was interested in exploring the control of the Marangoni cell, and the impact on material properties. All epoxy thin films spun at 22 % RH had significant surface roughness and exhibited Marangoni cells. Samples spun at 40 % RH exhibited larger Marangoni hexagons than those spun at 22 % RH. Thin films produced in 67 % RH did not exhibit Marangoni patterns. It appears that as the RH increases, so does the carbamate/carbonate formation on the surface, and thus a reduction in the driving force (concentration gradient) for the formation of the Marangoni cells as reported by Shenk . ^{43 , 45} Shenk et al. also demonstrated more compliant material in the higher RH samples due to the formation of carbamates during the curing process. Additionally, at the curing temperature, the cure of the epoxy becomes diffusion limited, and thus conversion of the oxirane (epoxy functional group) is limited, resulting in a lower cross linking density and much more compliant material. ^{54 , 55} A combination of these effects probably explains the decrease in storage module between 67 % and 40 % for the neat and composites. The 1 wt% samples were more greatly impacted than the 2 wt% due to insufficient nanoparticle/polymer interactions to overcome the carbamate formation reducing composite crosslinking and chain interactions. It is not known why modulus increased from 22 % RH to 40 % RH, though the lower RH surfaces have a larger number of Marangoni cell formations. This will require future research.

The material’s T _g was also measured. Results are provided in Figure 8. For a given RH, 1 wt% composites demonstrated a maximum in T _g . The 2.5 wt% composites resulted in the lowest T _g .

Figure 8:

Glass transition temperatures of different compositions of 90 PHR amine cured epoxy.

The addition of MMT to nanocomposites can result in an increase in T _g  > 10 °C, and can also result in a decrease of T _g  > 10 °C. ^{53 , 56} The increase in T _g is normally attributed to an increase in exfoliation, which lowers the polymer chains segmental motion due to it being anchored to the more abundant clay surface. On the other hand, some research has demonstrated a decrease in T _g . The organoclay may homopolymerize with the epoxy resin during the mixing if at elevated temperatures, which results in the decrease of cross-linking density when the polymer is cured with the amine. This decrease in cross-linking density leads to a subsequent decrease in T _g . It may also be due to high temperatures degrading the organic modifier in the clay such that it separates and acts as a small chain lubricant, decreasing T _g . ^{57 , 58} In either case, each acts as a plasticizer. It would be understandable if one were only comparing the neat epoxy to 1 wt% composite and attribute the increase in T _g to intercalation and exfoliation, in agreement with Lu and Nutt, ⁵⁹ Kaya and Tanoglu ⁶⁰ and Miyagawa et al. ⁶¹ If one were comparing the loading between 1 wt% and 2.5 wt%, one could conclude plasticization is occurring due to the temperatures reached during high shear whether via homopolymerization or lubrication.

The increase in T _g from neat to 1 wt% and the observed decrease in T _g as the MMT concentration is further increased to 2.5 wt% suggests competing effects. For low nanoparticle loading, the increase in intercalation/exfoliation impact is greater than the homopolymerization or lubrication effects, such that 1 wt% shows a higher T _g to that of the neat polymer. This is believed to be the effect of polymer chains entering the clay gallery resulting in the restriction of mobility of the polymer chains. However, as the clay content increases, the impact of additional small molecules being added to the system by the temperature and high shear degradation function as lubricants, decreasing T _g . Additionally, as the loading is increased, there could be an increase in homopolymerization such that it dominates the effects that intercalation/exfoliation might have on increasing T _g , resulting in the overall decrease. Both lubrication and homopolymerization could be effects resulting in the overall decrease of T _g as the organo-modified nanoparticle loading is increased.

For this system, as the RH increased, T _g increased. Generally, for polymeric systems, as RH increases T _g decreases due to plasticization and lower crosslinking. At sufficiently high RH, a surface phenomenon called blushing occurs. Amine blushing is a side reaction occurring in epoxy-amine cured films when they are exposed to CO₂ in the presence of sufficient water vapor at the dew point. ^{62 , 63} This reaction results in carbamate/carbonate formation, decreasing the cross-linking of the network as amine reacts with the CO₂ in the environment, reducing the composite’s modulus and T _g . ^{64 , 65} Since the temperature during spin coating at the three different RHs were below 70 °C, blushing is not expected (70 °C is below the dew point for 67 % RH). Though the role of blushing is minimized, ⁶⁶ Shenk et al. ⁴³ demonstrated there is carbamate formation, which may still result in plasticization. The exact reason for the greater T _g at 67 % RH versus the lower RH is not known. Shenk et al. ⁴³ demonstrated after sufficient curing, cracking initiates at the edge of the cells when stress is applied, indicating lower cross-linking at the edges. ¹ Since Marangoni cell formation is inhibited when the spin coating environment is above 60 % RH, it is possible a lower cross-linking density along the Marangoni cell boundary would result in a lower T _g .