Exposing a polymer to sequential organometallic vapor infiltration (SVI) under low pressure conditions can significantly modify the polymer's chemical, mechanical, and optical properties. We demonstrate that SVI of trimethylaluminum into polyethylene terephthalate (PET) can also proceed readily at atmospheric pressure, and at 60 °C the extent of reaction determined by mass uptake is independent of pressure between 2.5 Torr and 760 Torr. At 120 °C, however, the mass gain is 50% larger at 2.5 Torr relative to that at 760 Torr, indicating that the precursor diffusion in the chamber and fiber matrix decreases at higher source pressure. Mass gain decreases, in general, as the SVI process temperature increases both at 2.5 Torr and 760 Torr attributed to the faster reaction kinetics forming a barrier layer, which prevents further diffusion of the reactive species. The resulting PET/Al-Ox product shows high photoluminescence compared to untreated fibers. A physical mask on the polymer during infiltration at 760 Torr is replicated in the underlying polymer, producing an image in the polymer that is visible under UV illumination. Because of the reduced precursor diffusivity during exposure at 760 Torr, the image shows improved resolution compared to SVI performed under typical 2.5 Torr conditions.
Organic-inorganic hybrid materials have drawn attention for advanced electronics,1 biotechnology,2 optics,3,4 and catalysis5 applications. Sequential vapor infiltration (SVI) is one of the techniques, where hybrid materials are formed within organic materials by exposure to an organometallic precursor vapor. The technique is inspired by observations from atomic layer deposition (ALD) on polymers with moderate reactive functional groups (i.e., C=O) toward the ALD precursors.6–8 During ALD, an organic-inorganic hybrid material layer is observed at the interface of the inorganic film and polymer substrate.6–8 Motivated by these hybrid materials formation, techniques using the idea of extended ALD precursor exposures to form hybrid materials such as multiple pulsed vapor-phase infiltration (MPI),9–11 sequential infiltration synthesis (SIS),12,13 and SVI14,15 have been proposed. While the techniques vary by slight differences in processing, to all incorporate a common elongated exposure in order to enhance the infiltration. Biomolecules such as spider silk and collagens processed with MPI demonstrated improved strength by this hybridization.9–11,16 SIS has been studied for the formation of nano-patterning by selective infiltration into self-assembly block copolymers composed of one polymer that can react with the ALD precursor to form hybrid material and the other is inert towards the same precursor. The unreacted polymer is removed by plasma etching process, while hybrid material remains and hence nano-sized patterns are created.12,17–19
SVI utilizes a series of precursor exposures in a separated manner different from other infiltration methods. An exposure consists of an organometallic vapor dose into a reactor volume, whereafter the vapor is held in the reactor for a period of time and then purged from the reactor volume with inert gas. This dose-hold-purge can be repeated to increase the overall exposure of the organometallic vapor within the reactor. Afterwards, an oxidation agent (i.e., H2O) is also cycled in the same sequential order to oxidize the material.14 The SVI mechanism has been studied as a function of temperature, exposure conditions and surface area on C=O containing polymers such as polyamides and polyesters.14,20 SVI on polyethylene terephthalate (PET) and polybutylene terephthalate (PBT) fibers have been studied to create mesoporous hybrid materials for catalysis applications.15 Also, SVI on PET fibers showed the ability to tune optical properties of organic materials by formation of PET-alumina coordination complexes.20
Sol-gel is the common technique for organic-inorganic hybrid material formation, in which precursors react in an appropriate solvent.21 Solvent removal during the sol-gel process adds time and expense. One key advantage of SVI is that it utilizes precursors in the vapor phase thus allowing dry processing of the polymers and eliminates liquid solvents post processing. Furthermore, hybrid materials formation can be performed on a polymeric material that is already shaped as film or fiber. Therefore, the integration of the SVI process into polymer production lines, which is known for high throughput production, can produce hybrid materials or modify polymers with high throughput rates. Previous studies have shown that shape of the pristine polymeric material is mostly preserved even after high temperature calcination process or plasma etching.15,17 This offers SVI additional advantages from a materials processing perspective since polymers may be molded into more complex shapes and structures that can subsequently be modified with the hybrid materials afforded through SVI. To this point, one major limitation of the SVI process is that it has been only conducted under low to medium vacuum conditions with batch processing, which reduces the materials production speed. It is highly desired to have atmospheric pressure conditions, to apply the process for roll-to-roll manufacturing of hybrid materials. Operating at atmospheric pressure provides the opportunity to eliminate the time needed for reactor evacuation and pressure control, and facilitates integration for in-line manufacturing. This work provides the first investigation of the SVI mechanism as a function of pressure and aims to investigate feasibility of the roll-to-roll SVI under atmospheric pressure conditions. PET is one of the most common polymers used in textiles, packaging, and electrical insulation applications. Due to its high chemical and thermal resistance and low moisture permeability, PET has shown application as a substrate for flexible electronics.22–24 With respect to optical applications, PET shows a weak photoluminescence by absorption of UV light both in solid state and in solution due to the presence of pi electrons on the polymer backbone.25–27 Prior work by our team has shown SVI processing of PET fibers to increase the photoluminescence intensity by an order of magnitude.20 Here along with the mechanistic analysis of SVI process at various pressures, we also demonstrate advantages for atmospheric pressure SVI to selectively modify optical properties of the substrate with improved resolution.
PET fabrics were obtained with 2/1 twill woven structure, made of 100% PET round fibers in a 330/70 denier multifilament yarn and were used as received. Fabrics are cut into 2 × 1 in. pieces and the 2–3 yarns from each side are pulled out to make sure they are not separated during later handling of the samples, hence the initial mass for the samples remains the same. Fabric samples are weighed prior to SVI processing and they were in the range of 0.25–0.3 g. SVI treatments were conducted in a custom made viscous flow type ALD reactor described previously that has special system design, which allows variable pressure operation up to 760 Torr.28–30 The ability of precursor delivery at pressures higher than the vapor pressure of the precursor is enabled by pressurization of a hold cell on the gas delivery line. The evacuated hold cell is filled with the precursor or oxygen source (charge time) and then pressurized with N2 (pressurizing time). Afterwards, the hold cell is opened to the reactor through the main N2 line (dose time). In order to conduct the SVI hold steps (hold time), a pneumatic gate valve is added to the end of the reactor, either on vacuum side or atmospheric pressure exhaust side depending on the SVI operation pressure. The gate valve is opened during the reactor purging after each hold step (purge time).
Following the sample loading, the reactor chamber was evacuated for 5 min by opening the pump side of the exhaust before every run. The chamber was then purged at the operation pressure for 10 min. To achieve 2.5 Torr operation pressure, the N2 flow rate was 0.5 slm and for 760 Torr the flow rate was 5 slm. SVI process is conducted by first cycling trimethylaluminum (min. 98%, Strem Chemicals) in a dose/hold/purge sequence, followed by H2O (deionized) cycles in the same sequence. An inert purge step for 5 min is conducted in between trimethylaluminum (TMA) and H2O cycles. After the process is completed, 2 min of purging is conducted and the samples are removed to ambient conditions. For the mass gain calculations, the PET samples were weighed before loading to the reactor and after sitting at ambient conditions for 30 min after infiltration. For masking during the SVI processing, premium vinyl adhesive films (from Silhouette) were cut to a desired pattern using a Silhouette Cameo die cutting tool. The adhesive-backed films were stuck on the fabric surface and pressed manually.
In order to analyze the changes in chemical structure of the fibers, Fourier transform infrared (FTIR) spectra of the untreated and treated fabrics are obtained using a Nicolet Nexus 470 FTIR spectrometer with a germanium crystal attenuated total reflectance (ATR) attachment. Samples were placed on a germanium crystal and a force normal to the crystal surface is applied using a 3 mm diameter metal tip. Each FTIR-ATR measurement consisted of 64 scans from 700 to 4000 cm−1.
III. RESULTS AND DISCUSSION
In order to be consistent between low pressure and high pressure conditions, we first investigated the optimum dosing conditions that allowed for consistency in the amount of TMA delivered to the reactor. Since the hold cell is always charged with TMA up to 18 Torr, the pressurizing and dose times need to be adjusted to ensure that the maximum amount of TMA molecules are delivered to the reactor under both pressure conditions to achieve a saturated reaction between the sample and TMA (as observed by mass gain). For an ideal gas, the known hold cell temperature, volume, and pressure allows for calculating the number of TMA moles in the hold cell, n ∼ 2.8 × 10−5 mol. At 2.5 Torr, 0.5 s N2 charge and 0.5 s dose conditions are sufficient for saturation for SVI at 120 °C. At 760 Torr and 120 °C, a 10 s N2 charge and a 6 s dose is required for saturation. Furthermore, we investigated effect of purge time at different temperatures, with 6 s and 35 s purging at 2.5 Torr and with 15 s and 45 s purging at 760 Torr. At both pressures, longer purge times resulted in slightly lower mass gain values. In order to minimize the effect of left over TMA in the reactor, we fixed the purging time to >35 s. We therefore fixed these exposure conditions at the temperature and pressure values studied.
Figures 1(a) and 1(b) present the mass gain observed at 2.5 Torr and 760 Torr, respectively, as a function of number of TMA cycles in a temperature range of 60–150 °C. At 2.5 Torr, highest mass gain values are observed at 90 °C with a linear relationship to the number of TMA cycles. At 60 °C, a slightly lower mass gain is observed, also with a linear trend. However as the temperature increases, the mass gain deviates from linearity and higher mass gain is observed at 120 °C than that at 150 °C. In Fig. 1(b), 760 Torr samples show a pronounced mass gain decrease as the process temperatures increases. At 60 and 90 °C, a linear mass gain increase is observed as a function of the number of TMA cycles. As temperature is increased, a saturation behavior begins to be observed. It is interesting to note that the mass gain trend at 60 °C is independent of pressure. As the temperature increases, the differences in mass gain due to pressure arise and become more pronounced. This result is important for determining the feasibility of the process for atmospheric pressure as temperature is varied. The scale of the mass gain is very similar for both pressures (between 2 and 10 wt. %), which is promising for applying the SVI process at high pressures and low temperatures.
In Figure 2, the mass gain of the PET fabrics is compared as a function of temperature after 60 TMA SVI cycles at 2.5 and 760 Torr. The first observation is the marginal difference in mass gain of the samples processed at 2.5 and 760 Torr at 60 and 150 °C. Second, mass gain is highest at 90 °C at 2.5 Torr, whereas the maximum mass gain occurred at 60 °C at 760 Torr. Moreover, the mass gain of the samples at 2.5 Torr and 760 Torr shows a ∼3 wt. % difference between the 90 °C and 120 °C samples. This behavior is a consequence of the process being a combination of precursor diffusion along the reactor, through the fiber matrix (between fibers), and through the polymer bulk as well as chemical reactions taking place between precursor and polymer functional groups at the surface and in the bulk of the polymer. As for reaction rate, it is only affected by temperature, for constant TMA partial pressure. However, diffusion of the precursor in the gas phase is affected by both temperature and pressure, which will affect the speed of precursor delivery to the surface of the fibers. As discussed in previous studies,8,14,20 for constant pressure, the reaction rate between precursor and the polymer is reduced at low temperatures, which permits more diffusion, and is observed as higher mass gain in the sample. This statement remains true for this study as at 60 °C for both pressures results in high mass gain. Similarity of the mass gains at both pressures at 60 °C suggests that, despite the diffusivity of the TMA in N2 is changing with the pressure; the diffusivity in the bulk polymer is for the most part dependent on temperature. As the temperature increases, it is expected that the reaction rate between the precursor and C=O groups increases, as well as the diffusion rate of the TMA in N2 and polymer. While an increase in mass gain is observed at 2.5 Torr at 90 °C as compared to that at 60 °C, the mass gain slightly decreases at 760 Torr. Considering that the diffusion of the precursor in polymer matrix is primarily dependent on the temperature, we can attribute this difference to the pressure dependency of the precursor diffusion in N2. Faster diffusion will keep the surface concentration of TMA higher at 2.5 Torr resulting in higher mass gain. At 760 Torr, the diffusion rate does not increase as much as, resulting in lower mass gain. Furthermore, faster reaction kinetics at 90 °C can lead to a barrier layer close to the surface and result in less mass gain as noted at the 760 Torr sample. As the temperature increases to 120 °C, mass gain decreases at both pressures, which indicates that the reaction rate starts to dominate over the diffusion rate independent of pressure. Finally, as the process temperature reaches 150 °C, the reaction rate dominates the diffusion rate effect. The formed barrier layer prevents further diffusion resulting in similar mass gain for both pressures. However, the evolution of the mass gain at 150 °C by increasing number of the TMA cycles is very different for different pressures, as can be seen in Fig. 1. At low pressure, this barrier layer starts to form at initial cycles whereas at high pressure, it forms gradually. This can be attributed to slower kinetics at the surface at high pressures due to the lower partial pressure of the precursor as a result of slower diffusion.
To examine the reactivity of the TMA and PET as a function of pressure and temperature, FTIR analysis was conducted on samples using an ATR set up, the spectra of which are provided in Figure 3. All spectra show two common features, independent of the SVI temperature and pressure that are related to the TMA-PET reaction, and have been previously determined in literature.14,15,20,31,32 The first feature is the peak at 1716 cm−1 that upon SVI processing is decreased due to the ligand exchange reaction taking place between TMA and C=O groups. The second feature is a broad peak from 3000 to 3600 cm−1 wavenumber, which is attributed to the formation of OH groups by oxidation of Al-CH3 groups during water exposures.
As an indication of the reaction extent in the probing depth of the ATR setup, we calculated the consumption of the C=O peaks in Figure 3 following: Carbonyl Consumption (%) = 100 × (Peak Intensity(control) – Peak Intensity(sample))/Peak Intensity(control). In Figures 4(a) and 4(b), the C=O consumption of the SVI treated samples are provided as a function of temperature at 2.5 Torr and 760 Torr, respectively. After 30 cycles at 2.5 Torr (Fig. 4(a)), a maximum consumption of C=O peak is observed for the 120 °C sample followed by 90, 150, and 60 °C, respectively. In contrast, the mass gain analysis in Fig. 1(a) is quite different than the ATR analysis. Recall that in the mass gain analysis after 30 TMA cycles, the samples processed at 60, 120, and 150 °C samples show similar mass gain and the 90 °C sample shows higher mass gain. The difference in both trends can be attributed to the surface sensitivity in the ATR method and the limitation of its ultimate probe depth. This surface sensitivity provides important inferences as to the characteristics of the SVI materials modification to the PET at different pressures. First, the discrepancy in the relative magnitude of the mass gain and C=O consumption at 120 °C can be attributed to the reaction occurring close to the surface of the PET fiber. At low temperatures, the higher mass gain means that precursor diffuses more into the material and the consumption of the C=O groups close to the surface is lower. After 60 cycles at 2.5 Torr, the trends for mass gain and ATR analysis of the C=O consumption changes. As shown in Fig. 4(a), a marked decrease in the C=O concentration is observed at all temperatures except 60 °C. Even though after 60 cycles of SVI, the mass gain increased significantly at 60 °C, the C=O concentration (at the surface) is not considerably higher than 30 cycle SVI. In comparison, at all other temperatures, the C=O peak is nearly all consumed. This can be interpreted that the surface is nearly saturated with AlOx at higher temperatures.
A similar analysis is performed for the samples processed at 760 Torr, as provided in Figure 4(b) for 30 and 60 SVI cycles. After 30 cycles, the highest C=O consumption occurs at 150 °C then 90, 120, and 60 °C, respectively. In comparison, mass gain of 150 °C sample after 30 cycles is slightly lower whereas other temperatures are very similar. As compared to the 2.5 Torr, the C=O consumption values after 30 cycles in Fig. 4(a) show a wider dispersion then 760 Torr samples. This suggests that the change of diffusion rate of the precursor shows little dependence on temperature at high pressure and much higher dependence at low pressure. After 60 cycles, the ATR spectra in Fig. 3(d) suggest that all the samples have similar C=O peak intensities. However, mass gain of the samples increase as the processing temperature decreases. By comparing the trend of C=O consumption to the mass gain trend, it can be observed that the diffusion rate of the TMA in the samples is less temperature dependent at high pressure.
In order to support the above discussion of the operating pressure effect on the TMA diffusion rate along the reactor and, consequently, its concentration at the polymer surface, diffusion coefficients of TMA in N2 are calculated using the equation suggested by Fuller et al.33 as derived from Chapman–Enskog
where T is the gas mixture temperature, p is the total pressure, M are the molecular weights of the gas components, and vi are diffusion parameters considering the molecule structure of the diffusing species.
Calculated diffusion coefficients of TMA in N2 as a function of temperature at the low and high pressures during the hold step, 14 and 760 Torr, respectively, are given in Figure 5. At high pressure, the diffusion coefficients show significant lower values as compared to those at low pressure. Therefore, the overall mass gain at high pressure is expected to be lower than that observed at low pressure. At low pressure, TMA molecules have higher diffusion rates in N2 along the reactor and between fibers than at atmospheric pressure. This is especially true at lower temperatures, where at 60 °C the time required to diffuse 1 cm is ∼0.04 s at low pressure and ∼2.04 s at 760 Torr. The diffusion times were calculated based on the equation derived from Fick's first law as calculated by Mousa et al.29 Due to the faster diffusion along the reactor and between fibers at low pressure, a constant TMA partial pressure is maintained at the polymer surface, especially at high temperature, while a much lower TMA partial pressure is expected at polymer surface at atmospheric pressure. The TMA molecules that will react with the polymer surface will not be quickly substituted by new TMA molecules at atmospheric pressure due to the very slow diffusion rate along the reactor and between fibers. Such low TMA partial pressure at polymer surface will consequently lead to low TMA diffusion through the polymer bulk due to the low TMA concentration gradient between polymer surface and bulk.
It can also be noted from Fig. 5 that the change in diffusion coefficients with temperature at 760 Torr (0.74 cm2/s at 150 °C to 0.49 cm2/s at 60 °C) is much lower than the change at low pressure (40.4 cm2/s to 26.6 cm2/s). The difference in TMA diffusion coefficients can be used as a good estimate for the difference in the precursor diffusion through the polymer as discussed earlier. Accordingly, the difference in diffusion coefficients shown in Fig. 5 can explain the temperature dependence of mass gain observed in Figs. 1 and 2. An increase in temperature from 60 °C to 90 °C at 2.5 Torr shows an increase in mass gain as a result of the higher TMA diffusion in the polymer. However for the same temperature values at 760 Torr, the mass gain is low as a result of a small increase in the diffusion coefficient. As the temperature increased to 120 °C and 150 °C, the decrease in mass gains at both high and low pressures is a result of change in reaction rate between the polymer and the precursor rather than change in diffusion rates. Higher reaction rates at these temperatures lead to the formation of a barrier layer close to the fiber surface, which eventually reduces the diffusion of the precursor. Samples treated at 150 °C show similar mass gain and C=O consumption in the FTIR spectra at both pressures, which is attributed to the reaction rate being the primary factor that determines the mass gain values at this temperature.
While it has been shown that the lower diffusion rate of TMA through the polymer matrix at high pressure affects the mass uptake, the diffusion in the fiber matrix (between fibers) will also be altered. We investigated the effect of TMA diffusion through the fiber matrix by noting the percentage decrease in mass uptake after masking ∼82% of the sample upper surface. The change in mass uptake due to masking was noted by comparing the values for masked and bare samples
The samples were masked using a film with a vinyl adhesive and subsequently treated with SVI at different pressures. The sample processed at low pressure at 60 °C with 30 TMA cycles showed ∼20% lower mass uptake compared to the unmasked sample. On the other side, the masked sample processed under similar conditions but at high pressure (low TMA diffusion) showed ∼80% lower mass uptake as compared to the unmasked sample. The lower TMA diffusion through the fiber matrix at higher pressure may have strategic benefits in patterning. To demonstrate the diffusion differences at low and high pressures, the pattern resolution of masked fabric was investigated. As previously demonstrated, SVI of TMA can induce an increase in the photoluminescence of PET.20 Optical images of the treated fabrics with and without UV illumination are provided in Figure 6. It is observed that all of the treated samples do not show any indication of the patterned SVI process under normal light conditions. When the samples are exposed to UV light in dark room, a photoluminescent pattern appears, as observed in Figs. 6(a)–6(d). The resolution of the luminescent pattern is a direct indicator of the diffusion behavior of the TMA into the fiber matrix as pressure is altered. The sample treated at 2.5 Torr and 60 °C is not readable because precursor reached the masked sites due to the high diffusion rate at low pressure. By increasing the pressure to 760 Torr, the resolution of the pattern is greatly improved due to the lower diffusion into the masked regions because of the low precursor diffusion rate.
Assuming an infinitely large mask on top of a fabric surface on which there is a finite opening for precursor diffusion into the substrate; the cases of ideal, low, and good resolution are given schematically in Figure 7. For best pattern resolution, precursor should diffuse into the fiber matrix through only the exposed part of the mask. Penetration depth through the fiber matrix should also be very limited and hybrid modification happens only in the fibers close to the surface (Fig. 7(a)). For the demonstrated masked sample at 2.5 Torr, masking is limited to small areas under the label and precursor diffuses beyond the pattern definition and the resolution is very low (Fig. 7(b)). At 760 Torr, diffusion coefficients are much lower and the masking is significantly more effective (Fig. 7(c)). Limited and very low diffusion through the pattern definition results in better resolution. The ability to have fine photoluminescent patterning using the SVI process at atmospheric pressure has the potential for unique applications in flexible electronics and protective identification applications.
This research demonstrates the mechanistic differences for the synthesis of hybrid materials formed via sequential vapor infiltration of TMA at vacuum and atmospheric pressure. Results showed that SVI can be conducted at atmospheric pressure, which is very important for utilization of the process for roll-to-roll polymer material (fiber and film) modification applications. Differences in the extent of infiltration into the polymer matrix, as determined through mass gain and FT-IR analysis, are related to the precursor diffusion in N2 along the reactor and between fibers as well as its diffusion through the polymer and the reaction rate with the functional groups on polymer surface and bulk. The much slower diffusion of TMA in N2 at atmospheric pressure lead to lower TMA partial pressure at the polymer surface and, thus, lower diffusion rate through the polymer bulk as compared to the processing under lower pressure. Also, at higher temperatures, faster reaction kinetics leads to the formation of a barrier hybrid layer that prevents the reactants diffusion into and through the samples resulting in a lower mass gain values in general. Finally, the effect of diffusion through the fiber matrix (between fibers) on the ability to pattern the photoluminescent behavior of the hybrid materials is shown to improve with increasing pressure. Patterns created at higher pressure showed more defined edges, which shows great advantage for the feasibility of the pattern-based roll-to-roll processing of SVI.
Authors would like to thank Professor Gregory Parsons at Chemical and Biomolecular Engineering department at North Carolina State University for his help in this work. Authors also acknowledge partial funding from National Science Foundation Industry & University Cooperative Research Program: Center for Dielectrics and Piezoelectrics (CDP) (Grant No. 1361503), National Science Foundation CMMI Project No. 1000382 and Republic of Turkey Ministry of National Education.