Silane Modification of Cellulose Acetate Dense ... - ACS Publications

Article pubs.acs.org/Macromolecules

Silane Modification of Cellulose Acetate Dense Films as Materials for Acid Gas Removal Carine S. K. Achoundong,* Nitesh Bhuwania, Steven K. Burgess, Oguz Karvan, Justin R. Johnson, and William J. Koros* School of Chemical and Biomolecular Engineering, Georgia Institute of Technology, 778 Atlantic Drive, Atlanta, Georgia 30332-0100, United States S Supporting Information *

ABSTRACT: The modification of cellulose acetate (CA) films via grafting of vinyltrimethoxysilane (VTMS) to −OH groups, with subsequent condensation of hydrolyzed methoxy groups on the silane to form a polymer network is presented. The technique is referred to as GCV-modif ication. The modified material maintains similar H2S/CH4 and CO2/CH4 selectivities compared to the unmodified material; however the pure CO2 and H2S permeabilities are 139 and 165 barrers, respectively, which are more than an order of magnitude higher than the neat polymer. The membranes were tested at feed pressures of up to 700 psia in a ternary 20 vol. %H2S/20 vol. % CO2/60 vol. % CH4 mixture. Even under aggressive feed conditions, GCV-modified CA showed comparable selectivities and significantly higher permeabilities. Furthermore, GCVmodified membrane had a lower Tg, lower crystallinity, and higher flexibility than neat CA. The higher flexibility is due to the vinyl substituent provided by VTMS, thereby reducing brittleness, which could be helpful in an asymmetric membrane structure.

■

INTRODUCTION Natural gas is one of the cleanest and most useful of all fossil energy sources, with consumption expected to grow by 39% from 2010 to 2035.1 Despite its advantages, natural gas can contain high amounts of acid gases (H2S and CO2) that require removal before distribution. Typical pipeline specifications in the US require that natural gas be purified below 4 ppm and below 2 mol % for H2S and CO2, respectively.2 In the US, over 30% of the gas produced and over 40% of proven raw natural gas reserves are subquality, due to the presence of these impurities in higher than desired concentrations.3 Recently, H2S removal has received more attention, because in certain areas in the US, Canada, and the Middle East, oil and gas reservoirs contain high levels of H2S as well as CO2 at varying levels. The removal of such acid gases is conventionally achieved by absorption in various amine-based solvents.4 While this technology is workable, it is energy intensive, especially for gas streams of high acid gas content.5 On the other hand, membrane processes have a small footprint, are energy efficient, are easy to scale-up due to modular design, require low maintenance, and offer operational flexibility, especially when dealing with feeds streams of varying compositions and/or flow rates. Moreover, membrane processes are environmentally safe and typically operate at ambient temperature, thereby minimizing extensive heat exchange.5 Despite these intrinsic advantages, membrane technology requires the development of higher efficiency membranes materials with both high selectivity and permeability. Most membranes lose selectivity © XXXX American Chemical Society

and sometimes even productivity in the presence of highly condensable, plasticizing gases such as H2S and CO2, so economical processable materials that can withstand such aggressive feed conditions are high priorities. Asymmetric cellulose acetate (CA) membranes are the current industrial standard for the removal of CO2 from natural gas; however, they suffer selectivity loss under aggressive feed conditions. Only a few studies have considered simultaneous removal of CO2 and H2S; moreover, due to the hazardous nature of H2S, prior studies have generally focused on low concentrations of H2S. Chatterjee et al.6 studied the permeation properties of membranes based on polyurethanes (PU) and polyurethaneurea (PUU). Although they obtained selectivities as high as 100 for H2S/CH4, total feed pressure did not exceed 200 psia and with maximum H2S molar concentrations of 12.5%. Since many gas wells pressures can reach pressures well above 1000 psia, more aggressive feeds need to be considered, and this is a focus of our work. Mohammadi et al7 studied the acid gas permeation behavior through poly(ester urethane urea) membrane. Although they achieved selectivities of 43 for H2S/CH4, their studies were performed only at 3% H2S concentration and for feed pressures up to 435 psia. Lokhandwala et al8 measured the performance of silicone rubber (PDMS) fibers at 95 psig in a ternary mixture of 650 ppm H2S/4%CO2/bal CH4. They Received: May 21, 2013 Revised: June 25, 2013

A

dx.doi.org/10.1021/ma4010583 | Macromolecules XXXX, XXX, XXX−XXX

Macromolecules

Article

obtained a CO2/CH4 selectivity of 3.3 and a H2S/CH4 selectivity of 6.9. In the same study, a CO2/CH4 selectivity of 10.6 and a H2S/CH4 selectivity of 9.5 in Matrimid at 390 psig were reported in 800 ppm H2S/4% CO2/ balance CH4. In the above studies using rubbery polymers, the permeability of H2S is attractive, but CO2 permeabilities are not impressive, and the CO2/CH4 selectivities are lower than those achieved by glassy polymers. In addition, the lack of tests at higher pressures limits the ability to truly assess the membrane performance under realistic use conditions. Common techniques used to improve polymer separation properties for a given starting material include grafting and cross-linking. The latter is mainly used to stabilize polymer membranes against plasticization. Among all the cross-linking techniques, photo-cross-linking and chemical cross-linking are the most common. Many 6FDA-based polymers are chemically cross-linked using 3, 5-diaminobenzoic acid or a diamine as the key cross-linking agent.9−12 In addition to cross-linking, grafting, a method by which components are attached onto polymer membranes, allows tuning membrane properties for specific applications using chemical, radiation, photochemical, and plasma-induced techniques.13,14 If the molecules of a reagent contain a functional group that can react with the polymer, grafting by chemical reaction can occur. In this paper, we seek to improve the performance of cellulose acetate membranes via silane grafting and crosslinking to enhance the polymer properties for membrane-based gas separations. Specifically, we test the performance of the modified CA membrane with aggressive high concentrations of CO2 and H2S feeds.

SA =

ideal αA/B =

SF =

SF # =

yA /yB xA /x B

(6)

yA /yB fA /fB

(7)

where fA and f B are the fugacity coefficients of components A and B on the feed side of the membrane.

■

EXPERIMENTAL SECTION

Materials. Cellulose acetate with 39.7 wt % acetyl content (degree of substitution (DS) of 2.45) and an average Mn of ∼50 000 was obtained from Sigma-Aldrich. All solvents, silanes, and catalyst used were also obtained from Sigma-Aldrich. Acetone and NMP solvents had a purity ≥99.5%. The silane agent, VTMS (vinyltrimethoxysilane), was a 98% pure liquid, Sure-Seal, with a boiling point of 123 °C and a density of 0.968 g/cm3. Trimethoxysilane (TMS) was 95% pure, and the catalyst dicumyl peroxide (DCP) was 98% pure. CO2 and CH4 gases with 99.999% purity were obtained from Airgas, whereas H2S gas was obtained from Praxair with 99.6% purity. The ternary gas mixture used was also obtained from Praxair with a composition of 20 vol. % H2S, 20 vol. % CO2, and 60 vol. % CH4. Dense Film Membrane Formation. Cellulose acetate powder was first dried under vacuum at 100 °C for 24 h to remove sorbed moisture. The polymer powder was then dissolved in acetone to make a 20 wt % solution. The formation of the CA dense film membrane was done using a knife casting method in a sealed glovebag. The bag was first swept with N2 gas to eliminate any impurities; it was then saturated with acetone for 4 h prior to casting to reduce the rate of solvent evaporation upon casting.16 This presaturation of the vapor space prevents irregularities or defects caused by excessively rapid evaporation. After this step, the polymer solution was cast onto a clean glass plate using a 500 μm gap knife, and the solvent was allowed to evaporate for 24 h before removing the film from the plate. The film was then dried under vacuum at 100 °C for 24 h to remove residual solvent. Film thicknesses were measured using a micrometer and were typically in the range of 77 μm.

(1)

cm 3(STP) cm cm 2 s cmHg

The permeability can also be written as the product of a kinetic (diffusion) and a thermodynamic factor (sorption), viz., PA = DA × SA

(5)

where yA and xA refer to the mole fraction of component A in the gas phase at the downstream and upstream faces of the membrane, respectively. In fact, this separation factor does not fully account for the deviation from ideal behavior that highly condensable gases such as CO2 and H2S might exhibit, and a better indicator of the intrinsic membrane performance can be achieved as done by Kosuri et al.15 to account for fugacity:

Permeability is commonly expressed in units of Barrer, where 1 Barrer = 1 × 10−10

⎡D ⎤ ⎡S ⎤ PA = ⎢ A ⎥ × ⎢ A ⎥ = αdiffusivity × αsolubility PB ⎣ DB ⎦ ⎣ SB ⎦

In eq 5, αdiffusivity is the diffusivity or mobility selectivity and αsolubility is the sorption selectivity. In the case of mixed gas feeds where plasticization, as well as competitive interactions between the permeating gases and the polymer and even when the downstream pressure are not negligible relative to the upstream, the separation factor (SF) is used to describe the separation, viz.

BACKGROUND Gas Permeation. Gas permeation in polymers involves a sorption−diffusion mechanism. The penetrant first sorbs at the high pressure (upstream) side of the membrane and then diffuses down a concentration gradient to the low pressure (downstream) side of the membrane, where it finally desorbs downstream. The rate of gas permeation through a membrane is characterized by the permeability coefficient, P. The permeability of a given gas A is defined as the observed steady-state flux NA divided by the driving partial pressure (ΔpA) across the membrane normalized by the membrane thickness l. NA ΔpA /l

(4)

where CA is the concentration of sorbed gas per unit volume and pA is the external penetrant pressure, or fugacity for cases of nonideal feeds. The separation performance of a membrane is characterized by its permselectivity, which can be written as a ratio of permeabilities, when the downstream pressure is negligible relative to the upstream pressure, as in this study.

■

PA =

CA pA

(3)

where DA is the average diffusivity or diffusion coefficient and SA is the average solubility coefficient of a given gas A. The diffusion coefficient characterizes the mobility of the penetrant molecule in the membrane while the sorption coefficient characterizes uptake concentration of the penetrant at a given fugacity or partial pressure, viz. B


Macromolecules

Article

Scheme 1. GCV-Modification Reaction Mechanism

GCV-Modification of Cellulose Acetate. In the first step, the dried CA dense film was immersed in neat VTMS in a 1:100 by weight CA: VTMS ratio and placed in a 40 mL Teflon bomb, which was placed in a sealed stainless-steel autoclave. The autoclave was put in the oven at 200 °C for 24 h. After the 24 h reaction, the vessel was cooled down to room temperature, and the modified film was removed and dried in a vacuum oven at 150 °C for 24 h, to remove excess VTMS. The grafted polymer was then exposed to moisture in ambient air with a relative humidity of ∼25% for 48 h. This likely caused the unreacted methoxy groups to hydrolyze and to form silanols, which subsequently condensed to form siloxane bonds, inducing crosslinking of the polymer chains as illustrated in Scheme 1. As noted above, the film thickness of neat CA prior to the reaction was 77 μm and was found to increase by 30% to 100 μm after this reaction. The film retained its transparency and increased in weight by 16.7%. The density of neat CA and GCV-modified CA films were measured using a density column gradient and were found to be 1.312 ± 0.0024 g/cm3 and 1.272 ± 0.0026 g/cm3, respectively. Since the modification is believed to include a combination of grafting and cross-linking, we use “GCV” to denote the modification of CA by grafting and cross-linking using VTMS. While we will focus mostly on the GCV-modified CA in this paper, many other conditions were explored as well before identifying the preferred protocol. The details of those are presented in Supporting Information as these initial studies will only be discussed briefly to provide a perspective and comparison to the preferred approach. These less desired approaches include the use of a DCP catalyst in the GCV-Modification procedure, and the use of a different grafting agent, trimethoxysilane (TMS). Reaction Conditions Selection. Initially, we envisioned that the vinyl bond on VTMS would open to induce grafting on that side of the VTMS (as is done in the case of PE cross-linking17,18). With this in mind, we decided to use a 150 °C reaction temperature for 24 h using the DCP catalyst in a 1:100:0.1 ratios. The 24 h reaction time was selected to allow adequate time for the reaction with minimal diffusion limitations. It appeared that the resulting film from this initial study opened the vinyl group and the resulting film was brittle, cloudy, and had undesirable properties. To consider the effect of eliminating possible cross-linking through the olefin functionality, we explored the use of trimethoxysilane (TMS) instead of VTMS. We found that the resulting film was even more brittle and not useful as a membrane material. These studies showed that the vinyl group added flexibility to the polymer. Therefore, we proceeded with the GCV-Modification protocol described above as it produced transparent, more flexible films with the best membrane performance.

Characterization Methods. The gel content of the GCVmodified CA was determined by first drying and weighing the modified film, then immersing it in the NMP solvent and heating it to 100 °C for 24 h. The solution was filtered and the insoluble gel was dried at 150 °C overnight to remove excess solvent, and the film residue was weighed. The gel content was found to be 96.6 ± 0.86% for the GCV-modified CA. Infrared spectra of the GCV-modified CA and neat CA were measured using a Bruker Tensor 27 FT-IR spectrometer. Each sample was analyzed using 128 scans with a scanner velocity of 10 kHz with a resolution of 4 cm−1 and 6 mm aperture setting. Each sample was scanned from 370 to 4000 cm−1. The NMR spectra were recorded using a high resolution Bruker DSX300 solid-state spectrometer. 13C solid state NMR spectra were recorded at a 13C frequency of 75 MHz, 1H 90 degree pulse length of 5 μs, MAS at 10 kHz, VACP (variable amplitude CP) with a ramp on 1H ranging from 85 to 100%, 1 ms contact time, and repetition delay of 4 s. The 29Si studies were carried out with a 29Si frequency of 59.6 MHz, 1 H 90 degree pulse length 5 μs, MAS at 10 kHz with regular CP (no ramp in contact pulse), a contact time of 3000 μs and repetition delay of 5s. TG curves were obtained from STA 409PC Luxx from NETZSCH. The samples were heated at a rate of 10 °C/min from 25 to 800 °C in argon at a flow rate of 30 mL/min. DSC curves were obtained from a DSC Q200 from TA Instruments. Aluminum pans were used for the samples and reference, under a nitrogen atmosphere with a flow rate of 50 mL/min. The samples were first heated from 40 to 250 °C at a rate of 10 °C/min, then were kept at 250 °C for 2 min, and finally cooled to 40 °C at 10 °C/min. Dynamic mechanical analysis was done using a DMA Q800 from TA Instruments. Operations conditions were: heating rate of 3 °C/min from 100 °C250 °C, a fixed frequency of 3 Hz in air using the tensile geometry clamp. X-rays diffraction patterns were obtained using a PANalytical X’Pert Pro with a Ni filter and Cu Kα radiation from 5° to 60°. Elemental analysis of the GCV-modified and the neat CA samples were performed by Columbia Analytical Services, Inc., located in Tucson, AZ. The samples were ground prior to analysis. The Si wt % was obtained using the total dissolution method. The C and H wt % were obtained by combustion, and the O wt % was obtained by pyrolysis. XPS results presented here were measured using a Thermo K-alpha XPS using aluminum (Al) Kα radiation, a 100 μm spot size, and a 50 eV pass energy. C


Macromolecules

Article

Figure 1. ATR-IR spectra of CA (top), VTMS liquid (middle), and GCV-modified CA (bottom).

■

this case, we used the bands at 1191 and 2842 cm−1 since the band at 1080 cm−1 appears to have merged with the one at 1039 cm−1 in neat CA and was therefore not useful. These results indicate that approximately 85% of the methoxy groups in VTMS have reacted.

RESULTS AND DISCUSSION The IR spectrum of the GCV-modified CA film is shown in Figure 1. Some important features of the GCV-modified CA film in Figure 1 include: (1) a disappearance of the −OH band around 3469 cm−1, (2) the appearance of bands at 2842 and 1191 cm−1, characteristic of the unreacted −OCH3 groups, (3) the appearance of the doublet at 811 and 769 cm−1, characteristic of the vinyl group. All spectra were normalized with respect to one another for comparison, and the same peak sensitivity setting was used for each spectrum for peak detection. Therefore, it is reasonable to determine that some peaks are not being detected or have decreased in intensity relative to the others; however, to better quantify those observations, we used a band ratio analysis approach. The neat VTMS liquid has three bands that correspond to the Si− OCHH3 group at 2842, 1191, and 1080 cm−1. Therefore, we could take the ratio of any two of those peaks and compare it to the ratio of the same peaks in the GCV-modified CA film. In

⎡ Si−OCH3| ⎤ at 1191 cm−1 ⎢ ⎥ = 1.25 ⎢⎣ Si−OCH3|at 2842 cm−1 ⎥⎦ in neat VTMS ⎡ Si−OCH3| ⎤ at 1191 cm−1 ⎢ ⎥ ⎢⎣ Si−OCH3|at 2842 cm−1 ⎥⎦

and

= 0.19 in GCV‐modified CA

We used the same approach for the vinyl bands. The neat VTMS liquid has bands that correspond to the Si−CHCH2 group at 1599, 1410, 1011, and 968 cm−1. In addition, CH and CH2 bending bands occur at 811 and 769 cm−1. In this case, we took the ratio of the bands at 811 and 769 cm−1 to the band at 1410 cm−1 and compared with similar bands in the D


Macromolecules

Article

Figure 2. 13C solid state NMR of neat CA dense film membrane.

Figure 3. 13C solid state NMR of GCV-modified CA dense film membrane.

Figure 4. 29Si solid state NMR of GCV-modified CA Film.

assess whether the band in GCV-modified at 1044 cm−1 confirms the presence of the Si−O−Si bond. Therefore, the gel content is the only definitive evidence that cross-linking has occurred in the GCV-modified CA; and this value (96.6 ± 0.86%) clearly indicates that considerable cross-linking has occurred. The 13C NMR spectrum of neat CA is shown in Figure 2, while that of the modified film is shown in Figure 3. The 13C NMR spectrum of the GCV-modified CA generated two new signals around 130−140 ppm, corresponding to the carbons at either end of the carbon−carbon double bond. The signal around 50 ppm corresponds to the C−O bond, which belongs to the unreacted −OCH3 groups. This result further supports our proposed reaction pathway of Scheme 1. Since VTMS has a tendency to react with moisture and cross-link with itself to form a gel, 29Si NMR measurements were used to determine whether this was the case in the GCVmodified film. This technique has been used to generate reaction profiles of hydrolysis and condensation of various alkoxysilanes that are used in sol−gel processes19 as it provides insight into the chemistry of sol−gel processes. 29Si has a

GCV-modified CA film. These two ratios are similar, indicating that the vinyl groups did not react, which was desirable in this case since it adds flexibility in the polymer and has a positive effect on the permeation rate as we will discuss later. A follow up study with ethyl substituted analogue of VTMS could further probe this issue; however, it is beyond the scope of this study. ⎡ −CH and CH 2| ⎤ at 769 cm‐1 and 811 cm−1 ⎢ ⎥ = 12.95 ⎢⎣ ⎥⎦ Si−CHCH 2|at 1410 cm‐1 in neat VTMS and

⎡ −CH and CH ⎤ 2|at 7 cm‐1 and 811 cm‐1 ⎢ ⎥ ⎢⎣ ⎥⎦ Si−CHCH 2|at 1409 cm‐1 in GCV‐modified CA

= 12.44

The gel content test done on this GCV-modified CA film as described earlier suggests that cross-linking occurred. However, since the siloxane (Si−O−Si) band usually occurs between 1000 and 1130 cm−1 and both CA (1039 cm−1) and VTMS (1080 cm−1) have bands around that region, it is difficult to E


Macromolecules

Article

natural isotropic abundance of 4.7%, a nuclear spin of 1/2, and a magnetic moment that is slightly lower than that of 13C, leading to a lower resonance frequency. The referencing in 29Si NMR is mostly done relative to tetramethylsilane (TMS) due to their low boiling point, inert nature, and short relaxation time.20 Negative values of the 29Si shift are due to low frequency and high field compared to TMS. The literature data used to compare our data with were in solution rather than in solid state as it is the case here, so there might be a slight difference in the chemical shift. Figure 4 shows the 29Si NMR spectrum of the GCV-modified CA. As Douskey et al21 showed, neat VTMS has a signal around −56 ppm, whereas the ethyltrimethoxysilane (ETMO) has a signal around −41 ppm. In Figure 4, there is a strong signal around −58 ppm and a low signal around −43 ppm, which means that characteristic VTMS peak is still present in the GCV-modified structure as expected from the IR results. Furthermore, these results suggest that VTMS does not just swell the polymer and cross-link with itself to form a gel, since siloxanes bridges are not a significant part of the structure; at least the intensities may not be high enough to be detected relative to the other signals. For example, if a silsesquioxane with a vinyl as the end group had formed, there would have been a signal −80 ppm.22 We should note that this does not mean that the grafted VTMS molecules have not grafted and cross-linked as shown in Scheme 1, it simply means that a silica gel has not formed in this case, as it is generally the case with alkoxysilanes. A comparison of the TG curves of the neat CA and the GCV-modified CA film are shown in Figure 5. The initial

Unlike all other measurements, the DSC thermograms were done on powder as well as on the film. The GCV-Modification procedure for the powder was the same as used in the film as described above. The DSC scan of the neat CA and GCVmodified CA films is shown in Figure 6, while the powder scan

Figure 6. DSC curve of the Neat CA and GCV-modified CA films.

can be found in Supporting Information. The average values of all runs are shown in Table 1. Multiple effects lead to the Table 1. Average Tg (°C) Values of Neat CA and GCVModified CA using DSC and DMA neat CA Tg (°C) DSC DMA (E′ onset) DMA (E″ peak) DMA (tan δ peak)

202.1 196.5 205.9 220.1

± ± ± ±

0.11 0.15 0.00 0.00

GCV-modified CA Tg (°C) 162.0 153.9 157.5 170.0

± ± ± ±

0.16 0.08 0.93 0.64

observed properties seen in this study. There is a roughly 40 °C decrease in the Tg from the neat CA to the GCV-modified CA. While Maeda and Paul23 showed that plasticizers tend to lower the Tg of the polymer, this decrease does not suggest that VTMS acts as an internal plasticizer. Rather, the main effect of grafting VTMS is to substitute the hydrogen-bonding hydroxyl groups in the neat CA which, as Kamide and Saito24 showed, are mainly responsible for the stiff nature of CA. Furthermore, bulky side groups tend to lower the Tg by inhibiting segmental packing. While polymer cross-linking can lead to increased Tg’s, as the reduced Tg observed here suggests that factors tending to lower the Tg override those promoting chain rigidity. The large endothermal peak between 50 and 150 °C in neat CA presumably represents evaporation of sorbed water,25 and in GCV-modified CA represents methanol and water evolving during hydrolysis and condensation. By dividing the heat of vaporization of water (540 cal/g) by the area under the sorbed water endothermal peak in neat CA, the water content is estimated at about 3.3% prior to heating. This water loss is supported by TG curve in Figure 5. Using the heat of fusion for a perfect cellulose triacetate crystallite (8.2 cal/g),26 the crystallinity of neat CA was estimated to be about 33%. This melting transition observed in neat CA around 231 °C is not evident in the GCV-modified CA, and since melting characterizes crystalline polymers, this reveals the more amorphous nature of the GCV-modified CA as we show later using XRD (Figure 8). This may lead to permeability increases; however,

Figure 5. TG curve of neat CA and GCV-modified CA dense film membranes.

weight loss in CA that occurs below 100 °C is mainly due to moisture, and that occurring in GCV-modified CA around180 °C may be due to the methanol evolving from unreacted methoxy groups. Therefore, the onset of weight loss of CA and GCV-modified CA occurs around 340 and 300 °C, respectively for both materials. The TG curve also shows that CA loses up to ∼90% of its weight when heated to 800 °C compared to only ∼73% for the GCV-modified CA, reflecting additional residual mass introduced by the GCV technique. However, both materials show similar degradation temperatures, slightly above 360 °C, as reflected by the first derivative of the weight loss curve. The peak of the first derivative indicates the point of greatest rate of change on the weight loss curve. F


Macromolecules

Article

the cross-linking prevents excessive swelling that might occur otherwise. In addition, the GCV-modified CA film does not melt and is not etched upon reacting. The Tg of the polymer was measured as the tan δ peak as it is the one mostly used in literature; it occurs at the highest temperature. However, the loss modulus (E”) peak and the onset of the storage modulus are also reported. The storage modulus (E’) onset occurs at the lowest temperature and is related to the mechanical rigidity of the material, while the loss modulus peak is related to physical property changes (onset of segmental motion) in the material. Figure 7 show the DMA

Figure 8. X-ray diffraction pattern of neat CA and GCV-modified CA dense film membrane.

Table 2 shows that the silicon loading increases with increasing temperature and time. Samples D and E show an Table 2. Elemental Composition of GCV-modified CA Films by Columbia Analytical sample

Figure 7. Dynamic Mechanical spectrum of neat CA and GCVmodified CA dense films membrane at 3 Hz.

neat CA sample A

scan of the neat CA and GCV-modified CA films, respectively. Duplicate DMA scans can be found in the Supporting Information. As expected, the Tg values measured by the DMA were higher than those measured using DSC. In addition, the higher frequency of 3 Hz at which the measurements were made in this case also yielded slightly higher values than the values reported at 1 Hz in literature, as done by Puleo et al25 for example. The average Tg value as measured by the tan δ peak was found to be 220.1 °C for neat CA and 170 °C for the GCV-modified CA film. This 50 °C reduction in the Tg is in good agreement with the DSC measurements reported above. The magnitude of the tan delta peak at the Tg is higher for the GCV-modified CA compared to neat CA, indicating a higher fraction of the amorphous phase in the modified material. The higher frequency was used to improve the signal-to-noise ratio. The XRD patterns of the neat CA and GCV-modified CA are shown in Figure 8 and it also shows that there is a large difference in chain packing between the two materials. Neat CA has 2 diffraction peaks at 2θ angles of 9.4 and 17.3°, which correspond to characteristic d-spacings of 9.44 and 5.13 Å, respectively. The peak at 9.4° is quite sharp and more intense, revealing higher levels of crystallinity and possibly more perfect crystallites. However, the GCV-modified pattern has only one peak at 23.4°, which corresponds to a lower characteristic dspacing of 3.80 Å. This peak presumably represents a shift from the peak at 17.3° in neat CA and it is also broader, indicating that the modified material is more amorphous than the neat CA consistent with higher permeation rate discussed later, which is explained by the addition of the VTMS containing the Si−O bulky groups that may make close chain packing more difficult. Of course, it is well-known that crystallinity reduces the overall rate of permeation of semicrystalline polymers,27 due to suppression of both DA and SA factors in eq 3.

sample B sample C sample D sample E

reaction conditions − 1 day at 150 °C 3 days at 150 °C 4 days at 150 °C 1 day at 200 °C 2 days at 200 °C

carbon wt %

hydrogen wt %

oxygen wt %

silicon wt %

48.65 48.75

5.31 5.55

46.04 45.12

− 0.58

48.61

5.41

45.36

0.62

48.12

5.53

45.35

1.00

48.84

5.68

43.74

1.74

45.43

5.96

42.30

6.31

increase in Si wt % of 72% when the reaction time increases by 1 day; however, sample E was found to be unstable under those conditions. Theoretically, if 1 VTMS molecule grafts to every −OH group available per CA monomer, the Si wt % will be 2.04%. This value is the optimal loading if the VTMS molecule only graft at the hydroxyl end. Sample D, which is the case mostly discussed in this paper, has a Si wt % of 1.74%, which is close to the maximum loading of 2.04%, and the material maintained a desirable appearance and flexibility. By taking the ratio of these wt %, we estimated that roughly 85% of the −OH groups were grafted with VTMS, which matches the value obtained with IR ratio analysis of the methoxy groups that have reacted. To summarize, there is an optimum loading at which the film retains its appearance and stability for practical membrane purposes. The conditions that we’ve studied lead to the conclusion that each reaction temperature will have an optimum reaction time to reach the desired Si loading. For example as we go from 1 to 4 days of reaction at 150 °C, the Si wt % increases by 72%, indicating that by tuning reaction time, an optimum loading can be achieved. It also shows that the modification could be done at much lower temperatures. While long reaction times are shown in this study, it should be noted that these studies were performed on dense films, which have diffusion limitations. It is anticipated that this modification could be done at lower temperatures and shorter reaction times G


Macromolecules

Article

in hollow fibers since they do not have those diffusion limitations as the actual separation skin is much smaller. Table 3 shows the elemental composition of the GCVmodified CA using XPS. Spot sizes of 100 μm were measured at Table 3. Elemental Composition of GCV-Modified CA Dense Film by XPS (Atomic %) O 1s C 1s Si 2p

area 1 (top)

area 2 (top)

area 3 (bottom)

area 4 (bottom)

29.11 54.70 16.19

28.15 56.66 15.19

27.40 58.59 14.01

29.75 55.49 14.76

two different areas of the sample, both on the top surface and at the bottom surface to look at the distribution of the silicon atoms across the film surface. As shown in Table 3, there is an incorporation of 15.0 ± 1.1% Si in the GCV-modified CA material. The GCV-Modification was also done on powder as mentioned previously and the elemental analysis showed a similar Si content (2.14 wt %) as the film. The reaction was done on powder to ensure that the modification was done throughout the samples for comparison to the film composition, since the powder has less diffusion limitations than the dense film. Furthermore, a gel content of 96.6% as shown previously also supports the fact that the modification occurred throughout the film. Pure Gas Permeation. Pure gas permeation measurements were made using both neat CA and GCV-modified CA at 65 psia and 35 °C with the constant-volume apparatus, with some deviations from the previously reported systems28−30 as described in the Supporting Information in order to ensure high safety when handling H2S gas mixtures. The pure gas results of the GCV-modified CA films and other polymer materials were plotted on a permeability-selectivity trade-off curves31,32 for CO2/CH4 in Figure 9 and H2S/CH4 in Figure 10. Even though the H2S/CH4 upper-bound line does not exist yet due to the limited amount of data available in literature for this gas pair, we nevertheless show our current polymer performance relative to some others. The GCV-modified CA film shows an overall increase in membrane productivity. The CO2 permeability of the GCV-modified CA sample is an

Figure 10. H2S/CH4 Permeability-selectivity trade-off of GCVmodified CA in pure gas feeds at 65 psia and 35 °C.

impressive ∼28× higher than pure CA. Similarly, the H2S permeability is ∼37× higher than pure CA. The ideal CO2/ CH4 selectivity of the GCV-CA is similar to that of pure CA and comparable to high performance polyimides such as PDMC,33 6FDA-DAM: DABA (3:2)34 and 6F-PAI,35 6FDADAM: DABA (2:1).36 The pure H2S/CH4 selectivity of the GCV-modified sample is also 34% higher than that of pure CA. Morita et al37 have synthesized and characterized silyl derivates of CA. While they obtained CO2 permeabilities of 110 and 160 barrers, the corresponding CO2/CH4 selectivities were 5.5 and 6.7, respectively. However, in our case, the H2S/CH4 and CO2/ CH4 selectivities are much higher. The pure gas performance of this GCV-modified CA material is very promising as it competes well with more expensive high performance polymers. While the final outcome of the VTMS treatment is very attractive, explaining its detailed impact requires addressing many effects. Kamide et al24 showed that strong intermolecular interactions may exist between hydroxyl and acetyl groups, between two acetyl or two hydroxyl groups, and Puleo et al25 showed that for CA with high DS, the acetyl− acetyl interactions are the most prevalent. Those interactions contribute to the motions and packing of CA chains, thereby affecting transport properties. The acetyl groups reduce the amount of hydrogen bonding, which leads to an increase in chain flexibility and mobility, and therefore an increase in permeability. Using this analogy, since the presence of hydroxyls groups in the virgin CA provides hydrogen bonding sites that stabilize and rigidify the matrix, the replacement of the simple hydrogen bonding with 3-D covalent cross-links through the silanols sites tends to greatly rigidify the matrix but also to embrittle it. In addition, the presence of the vinyl substituent on the VTMS has a moderating effect on the brittleness, and their absence makes the silanols cross-linking unworkable for practical membranes. Both CO2 and H2S have the ability to disrupt hydrogen bonds at high sorption levels. Finally, the combination of crystallinity disruption and cross-linking stabilization creates an excellent synergistic effect. All these effects make the overall VTMS treatment challenging to explain, but highly attractive in terms of ultimate outcome as a novel membrane modification technique. The presence of the Si−O−Si bridge may also contribute to the high selectivity observed. Mixed Gas Permeation. Even though the GCV-modified CA showed promising results in pure gas feeds, multicomponent mixture tests were clearly needed to assess its

Figure 9. CO2/CH4 permeability−selectivity trade-off curve31,32 comparison of GCV-modified CA to other polymers materials33−36 in pure gas feeds at 65 psia and 35 °C. H


Macromolecules

Article

Figure 11. CO2 permeability (a), H2S permeability (b), H2S/CH4 selectivity (c), and CO2/CH4 selectivity (d) of neat CA and GCV-modified CA in a 20/20/60 vol. % H2S/CO2/CH4 feed mixture at 35 °C.

Table 4. Comparison of GCV-Modified CA with Other Polymers in Mixed Gas Feeds at 35°C polymeric membrane

pressure (psi)

pure CA pure CA GCV-modified CA GCV-modified CA pure CA 6FDA-DAM:DABA (3:2) annealed 180 °C 6FDA-DAM:DABA (3:2) Annealed 230 °C Pebax 1074 Pebax 1657 PU1 PU2 PU3 PU4

500 700 500 700 145 700 700 145 145 145 145 145 145

PH2S

PCO2

αCO2/CH4

αH2S/CH4

source

20/20/60 20/20/60 20/20/60 20/20/60 6/29/65

8.71 39.7 204 190 2.1

8.66 27.5 129.4 136 2.4

29.5 19.1 21.8 20 22.1

29.7 27.4 34.3 27.5 19.4

this study this study this study this study Chatterjee6

10/20/70 10/20/70 12.5/18.1/69.4 1.30/27.9/70.8 12.5/18.1/69.4 12.5/18.1/69.4 12.5/18.1/69.4 12.5/18.1/69.4

25.4 23.6 695 248 183 618 280 223

55.6 50.8 155 69.1 55.8 195 62.2 50.8

32.1 31.1 11.2 14.1 6.9 5.6 12.2 14.9

14.7 14.4 50.4 50.6 22.6 17.8 54.9 65.6

Kraftschik34 Kraftschik34 Chatterjee6 Chatterjee6 Chatterjee6 Chatterjee6 Chatterjee6 Chatterjee6

feed composition (H2S/CO2/CH4)

increases. This somewhat surprising opposite trend presumably occurs because in a ternary system, the highly condensable gases, CO2 and H2S, now have to compete for the Langmuir sorption sites. Since H2S has a better affinity for the sites, the sorption of CO2 is presumably greatly reduced, leading to lower CO2 permeability increase than H2S. Methane is not greatly affected in this case because it already has lower affinity for the Langmuir sorption sites than both CO2 and H2S. These results are reflected in the productivity of the membrane. One important point to note is that at high feed pressures of 700 psia, the H2S/CH4 selectivity is still 27 in the GCV-modified CA, which is much higher compared to other glassy polymers and the CO2 /CH4 selectivity (∼20) did not suffer significant loss. This stability at high pressures is impressive, since most literature reports focus only on low H2S concentrations and low pressures. Even though we cannot make a direct comparison

true performance. Ternary gas measurements were made with a 20/20/60 vol. % H2S/CO2/CH4 mixture at upstream pressures ranging from 100 to 700 psia and with vacuum downstream. Parts a and b of Figure 11 show the H 2 S and CO 2 permeabilities of CA and GCV-modified CA in the ternary mixture, while parts c and d of Figure 11 show the H2S/CH4 and the CO2/CH4 selectivities, respectively. Parts a and b of Figure 11 show that there is a very large increase in productivity in the GCV-modified CA compared to pure CA, which is expected given the results observed in the pure gas feed. We also note that the rapid upswing in permeability is suppressed for both H2S and CO2 isotherms of the GCV-modified CA compared to pure CA, demonstrating that the GCV-modified CA is much more plasticization resistant. Parts c and d of Figure 11 show that when the CO2/CH4 selectivity decreases, the H2S/CH4 selectivity I


Macromolecules

Article

The actual ternary mixtures plotted in Figures 12 and 13 are shown in Table 4, only the H2S content of the mixture is shown on the plot along with the upstream pressures. The GCVmodified CA result shown is at 700 psia and the other polymers shown are at pressures much below that except for 6FDADAM: DABA (3:2) that is compared at the same pressure. All other pressures are much lower than 700 psia and the performance of the GCV-modified CA is still better than most polymers in this ternary mixture.

due to the difference in testing conditions, we can nevertheless note that our measurements were done in much more aggressive (higher H2S concentration, higher pressures) environments than those previously reported. Table 4 below summarizes most of the mixed gas studies that have been conducted on polymeric dense films for natural gas sweetening. We can see from this table that the performance of the GCV-modified CA in presence of H2S is very impressive compared to the other polymers shown in the table. The H2S/ CH4 selectivity is 34 at 500 psia, which is much higher compared to other glassy polymers. We should note that the result we present contains 20% H2S, whereas the other measurements report lower H2S feed content, which indicates the robustness of this GCV-modified CA membrane. Moreover, our GCV-modified CA has a higher CO2/CH4 and H2S/CH4 selectivities overall in the aggressive ternary feed compared to PU1 and PU2. To put the results of Table 4 in perspective, the productivityefficiency trade-off curve for CO2/CH4 of selected polymers is shown in Figure 12 and that of H2S/CH4 is shown in Figure 13.

■

CONCLUSION This study presents the development of a novel modification method for the low cost, industrial standard membrane material (cellulose acetate) to create a material capable of separating aggressive acid gas feeds. In this technique, cellulose acetate is modified via grafting of vinyltrimethoxysilane (VTMS) to hydroxyl groups, followed by hydrolysis of the methoxy groups, with subsequent condensation of silanols to create a polymer network. The GCV-modified CA membrane had CO2 and H2S productivities more than 1 order of magnitude higher than neat CA, with a much higher H2S/CH4 selectivity compared to other glassy polymers and some rubbery polymers, and is quite stable under high aggressive feed conditions. Previous studies who used a similar approach obtained low H2S/CH4 and/or CO2/CH4 selectivities. This non expensive polymer is the benchmark for membrane gas separations, and this approach may have an immediate application in the field. These findings could open new avenues for CA as a lower cost, fairly high performance alternative even in non membrane applications. This GCV-modified CA is capable of separating feed streams with high acid gas content. It was found that the GCV-modified film was more amorphous and had a lower glass transition temperature (Tg) than the neat polymer. The presence of the vinyl substituent on the VTMS helped the membrane retain its transparency and increased its flexibility, and the substitution of the hydroxyl groups, which were mainly responsible for the stiffness of CA, both contributed to the overall permeability increase of the modified polymer.

■

Figure 12. CO2/CH4 permeability−selectivity trade-off31,32 comparison of GCV-modified CA and other polymers materials6,34 in ternary gas feeds at 65 psia and 35 °C.

ASSOCIATED CONTENT

S Supporting Information *

The permeation apparatus, FTIR spectra of neat CA and VTMS (with band assignments), GCV-modified CA using DCP (GCV-DCP), modified CA using TMS (GCT-modified), the DSC scan of CA powder, the duplicate DMA scans of CA and GCV-modified CA, 13C and 29Si NMR spectra of GCVDCP-modified CA, and GC-MS analysis of GCV-modified CA liquid residue. This material is available free of charge via the Internet at http://pubs.acs.org/.

■

AUTHOR INFORMATION

Corresponding Author

*Telephone: (404) 385-2845. Fax: (404) 385-2683. E-mail: [email protected] (C.S.K.A); bill.koros@chbe. gatech.edu (W.J.K). Notes

The authors declare no competing financial interest.

■

Figure 13. H2S/CH4 permeability−selectivity trade-off31,32 comparison of GCV-modified CA and other polymers materials6,34 in ternary H2S/CO2/CH4 feed mixtures at 65 psia and 35 °C.

ACKNOWLEDGMENTS The authors would like to thank KAUST for funding this project. The authors would also like to thank Johannes E. Leisen at the Georgia Tech NMR Center for his help on solidJ


Macromolecules

Article

state NMR measurements and Jose Baltazar and Andac Armutlulu for their help with XPS measurements.

■

REFERENCES

(1) Annual Energy Outlook 2012: With Projections to 2035; U.S. Department of Energy: Washington, DC, 2012. (2) Baker, R. W.; Lokhandwala, K. Ind. Eng. Chem. Res. 2008, 47, 2109. (3) Meyer, H. S. GasTIPS 2000, 6, 10. (4) Kesting, R. E.; Fritzsche, A. K. Polymeric gas separation membranes; Wiley: New York, 1993. (5) Bhide, B. D.; Voskericyan, A.; Stern, S. A. J. Membr. Sci. 1998, 140, 27. (6) Chatterjee, G.; Houde, A. A.; Stern, S. A. J. Membr. Sci. 1997, 135, 99. (7) Mohammadi, T.; Moghadam, M. T.; Saeidi, M.; Mahdyarfar, M. Ind. Eng. Chem. Res. 2008, 47, 7361. (8) Lokhandwala, K. A.; Baker, R. W. Sour Gas Treatment Process Including Membrane and Mon-membrane Treatment Steps. U.S. Patent 5,407,466, April 18, 1995. (9) Koros, W. J.; Wind, J. D.; Staudt-Bickel, C.; Paul, D. R. Macromolecules 2003, 36, 1882. (10) Cao, C.; Chung, T. S.; Liu, Y.; Wang, R.; Pramoda, K. P. J. Membr. Sci. 2003, 216, 257. (11) Bos, A.; Punt, I. G. M.; Wessling, M.; Strathmann, H. Sep. Purif. Technol. 1998, 14, 27. (12) Liu, Y.; Chung, T. S.; Wang, R.; Li, D. F.; Chng, M. L. Ind. Eng. Chem. Res. 2003, 42, 1190. (13) Ulbricht, M. Polymer 2006, 47, 2217. (14) Bhattacharya, A.; Misra, B. N. Prog. Polym. Sci. 2004, 29, 767. (15) Kosuri, M. R.; Koros, W. J. J. Membr. Sci. 2008, 320, 65. (16) Baker, R. W. Membrane technology and applications; 3rd ed.; John Wiley & Sons: Chichester, U.K., and Hoboken, NJ, 2012. (17) Barzin, J.; Azizi, H.; Morshedian, J. Polym. Plast. Technol. Eng. 2006, 45, 979. (18) Barzin, J.; Azizi, H.; Morshedian, J. Polym. Plast. Technol. Eng. 2007, 46, 305. (19) Hook, R. J. J. Non-Cryst. Solids 1996, 195, 1. (20) Uhlig, F.; Marsmann, H. C. In Gelest Catalog; Gelest: Morrisville, PA, 2009. (21) Douskey, M. C.; Gebhard, M. S.; McCormick, A. V.; Lange, B. C.; Whitman, D. W.; Schure, M. R.; Beshah, K. Prog. Org. Coat. 2002, 45, 145. (22) Devreux, F.; Boilot, J. P.; Chaput, F.; Lecomte, A. Phys. Rev. A: At. Mol. Opt. Phys. 1990, 41, 6901. (23) Maeda, Y.; Paul, D. R. J. Membr. Sci. 1987, 30, 1. (24) Kamide, K.; Okajima, K.; Saito, M. Polym. J. 1981, 13, 115. (25) Puleo, A. C.; Paul, D. R.; Kelley, S. S. J. Membr. Sci. 1989, 47, 301. (26) Mandelkern, L.; Flory, P. J. J. Am. Chem. Soc. 1951, 73, 3206. (27) Michaels, A. S.; Parker, R. B. J. Polym. Sci. 1959, 41, 53. (28) Moore, T. T.; Damle, S.; Williams, P. J.; Koros, W. J. J. Membr. Sci. 2004, 245, 227. (29) Damle, S.; Koros, W. J. Ind. Eng. Chem. Res. 2003, 42, 6389. (30) O’Brien, K. C.; Koros, W. J.; Barbari, T. A.; Sanders, E. S. J. Membr. Sci. 1986, 29, 229. (31) Robeson, L. M. J. Membr. Sci. 1991, 62, 165. (32) Robeson, L. M. J. Membr. Sci. 2008, 320, 390. (33) Hillock, A. M. W.; Miller, S. J.; Koros, W. J. J. Membr. Sci. 2008, 314, 193. (34) Kraftschik, B.; Koros, W. J.; Johnson, J. R.; Karvan, O. J. Membr. Sci. 2013, 428, 608. (35) Vaughn, J. T.; Koros, W. J.; Johnson, J. R.; Karvan, O. J. Membr. Sci. 2012, 401−402, 163. (36) Wind, J. D.; Paul, D. R.; Koros, W. J. J. Membr. Sci. 2004, 228, 227. (37) Morita, R.; Khan, F. Z.; Sakaguchi, S.; Shiotsuki, M.; Nishio, Y.; Masuda, T. J. Membr. Sci. 2007, 305, 136. K


Silane Modification of Cellulose Acetate Dense ... - ACS Publications

Recommend Documents