13488
J. Phys. Chem. C 2009, 113, 13488–13497
Molecular Length Dependence of the Melting Temperature Elevation for the Encapsulated Linear Monocarboxylic Acids inside Titania Nanotubes Xiao-Ping Tang,* Eddie Z. Y. Ding, Nitzia C. Ng, and Benjamin K. Mezick Department of Physics and Astronomy, UniVersity of LouisVille, LouisVille, Kentucky 40292 ReceiVed: April 17, 2009; ReVised Manuscript ReceiVed: June 12, 2009
A series of linear monocarboxylic acids (specifically, decanoic, dodecanoic, tetradecanoic, hexadecanoic, and octadecanoic acids) encapsulated in titania nanotubes was studied by 13C nuclear magnetic resonance. The 13C CP-MAS spectra for the encapsulated acids display a broad peak representing the molecules chemisorbed on the nanotube surface and two new, sharp peaks representing the molecules trapped in the pore shielded by the chemisorbed molecules. The two new, sharp CP peaks for the trapped acids are readily differentiable from the sole sharp CP peak for the bulk neat acids. They were assigned to the trapped dimers and monomers, which are alternately distributed in the pore. The trapped acids thus possess a novel molecular arrangement, in contrast to the bulk neat acids’ containing solely dimers. The melting temperatures of all the trapped acids were measured higher than the corresponding bulk melting temperatures. The degree of the melting temperature elevation was observed inversely proportional to the effective pore diameter for the trapped acids, bearing a resemblance to the Gibbs-Thomson equation. With respect to using encapsulated fatty acids for thermal energy storage, the present work gives insight to lessening the latent heat loss induced by encapsulation. Introduction Confined materials in narrow pores exhibit1-12 interesting physicochemical phenomena and are of potential applications. Of the encapsulated molecules in the pore, these adsorbed on the pore surface are subject to direct interaction with the pore wall and may be used to passivate13 the pore surface. If the adsorbed molecules are big enough, the other encapsulated molecules in the pore are effectively far away from the native pore wall and, thus, experience negligible interaction with the native pore wall. These encapsulated molecules thus can be viewed as merely trapped in the narrowed pore with the adsorbed molecules acting as an “effective pore wall”. When the adsorbed molecules are of the same species as the trapped molecules, the “effective pore wall” for the latter is built with the same kind of molecules. The trapped molecules then provide a unique system to study the confinement effect on chemical bonding, molecular arrangement, and phase behavior, if they are experimentally differentiable from the adsorbed molecules. Previously, hexadecanoic and octadecanoic acids encapsulated in titanate nanotubes (TiNTs) were revealed as such examples.14,15 Hexadecanoic and octadecanoic acids are linear monocarboxylic acids that can passivate the pore surface by forming hydrogen bonds between their carboxyls and the hydroxyls on the pore wall. For the medium- to long-molecular-chain acids, the molecular length is >1 nm, and the chemisorbed molecules provide good isolation for the trapped molecules. For these acids, solid, bulk, neat acids contain solely dimers,16,17 each with two hydrogen bonds of CdO · · · H-O-C connecting two carboxyls. However, the solid octadecanoic acid trapped in TiNTs was observed15 to consist of not only dimers but also monomers, signifying a novel molecular arrangement under confinement. For the trapped hexadecanoic acid in TiNTs, the melting * Corresponding author. Phone: 502-8520917. Fax: 502-8520742. E-mail:
[email protected].
temperature was observed14 significantly higher than the bulk melting temperature of hexadecanoic acid. In this work, we investigated decanoic acid (CH3(CH2)8COOH), dodecanoic acid (CH3(CH2)10COOH), tetradecanoic acid (CH3(CH2)12COOH), hexadecanoic acid (CH3(CH2)14COOH), and octadecanoic acid (CH3(CH2)16COOH) trapped in TiNTs. Using the series of acids of similar molecular structure but different molecular lengths, we try to determine whether the observed novel molecular arrangement15 is generic for the trapped linear monocarboxylic acids and measure how the melting temperature elevation14 depends on the molecular length. Fatty acids such as dodecanoic acid are under active research18,19 as noncorrosive phase change materials for high-efficiency thermal energy storage. For such applications, fatty acids are often encapsulated18,19 in porous materials to form thermal composite structures, but the encapsulated fatty acids have been shown18,19 with reduction in the latent heat and, thus, loss of efficiency as thermal energy storage. The present work seeks to understand how encapsulation changes chemical bonding and molecular arrangement, which govern the latent heat, and thus may gain molecular-level information on how to lessen the latent heat loss for the encapsulated fatty acids. Materials and Methods NMR measurements were conducted using a 300-MHz Bruker Avance console and a Bruker CP-MAS20 (crosspolarization/magic angle spinning) probe. Samples were sealed in Pyrex inserts which fit in 4-mm zirconium rotors. Because the 13C peak of the carboxylic carbon is sensitive to the chemical bonding near the carboxyl and is well-separated from the 13C peaks of the methylene and methyl carbons,21 it was measured in this work by a variety of pulse sequences to characterize the chemical bonds and to measure the melting transition for the investigated acids. Figure 1a displays the standard one-pulse sequence22 (OPS) to record 13C spectra for liquid acids. The
10.1021/jp903553d CCC: $40.75 2009 American Chemical Society Published on Web 07/01/2009
Monocarboxylic Acids inside Titania Nanotubes
J. Phys. Chem. C, Vol. 113, No. 31, 2009 13489
Figure 1. (a) The standard one-pulse sequence to measure 13C spectra for liquid acids. (b) The CP pulse sequence to measure 13C spectra for solid acids. During CP, the RF power applied to the 1H spins is ramped up, and the spectrum is recorded under high-power 1H-decoupling using the TPPM25 scheme. (c) The Torchia pulse sequence26 to measure 13C T1 for solid acids. (d) Pulse sequence to measure the 1H T1 corresponding to the 13 C CP peak. (e) Pulse sequence27 to measure 13C homonuclear double quantum. High-power 1H-decoupling is applied during the excitation and reconversion periods and during t1 using the CW scheme, and during t2 using the TPPM scheme. (f) The SUPER28 sequence to measure 13C chemical-shift anisotropy. Tr is the rotor period.
90° pulse applied to 13C spins directly generates 13C transverse magnetization,22 which is recorded without 1H-decoupling because fast molecular motion averages out the direct 1H-13C nuclear spin coupling.20,22 For solid acids, because they, as shown later, possess extremely long 13C nuclear spin-lattice relaxation time (T1), CP is the only viable method. Figure 1b displays the employed CP pulse sequence.20,23 The 1H 90° pulse generates 1H transverse magnetization, which is transferred to 13 C transverse magnetization during CP. For this work, the RF power applied to 1H spins is ramped up24 during CP. To resolve the 13C peaks for different solid acid components, it is also necessary to apply high-power 1H-decoupling during spectral acquisition to remove the dipolar coupling between the 13C spins and the abundant 1H spins. The two-pulse-phase-modulation (TPPM) scheme25 was used to optimize 1H decoupling. CP is induced20,23 by the through-space dipolar coupling between 1H and 13C spins by transferring spin magnetization from the former to the latter. For liquid acids, because the fast molecular motion reduces such dipolar coupling, CP becomes ineffective. On the other hand, the OPS pulse sequence is not efficient for the measuremetn of solid acids because the extremely long 13C T1 requires unaffordably long experimental time, and the strong 1 H-13C dipolar coupling also requires high-power 1H decoupling. Thus, in this work, the OPS and CP pulse sequences were used to measure liquid and solid acids, respectively. Different solid acid components, apart from displaying different 13C CP peaks, can also present different 1H and 13C T1 values. The 13C T1 for each 13C CP peak was determined by the Torchia pulse sequence26 shown in Figure 1c. The CP pulse
segment creates 13C transverse magnetization, which is then transferred to 13C longitudinal magnetization by the immediately applied 13C 90° pulse. The 13C longitudinal magnetization decays during τ, and the second 13C 90° pulse transfers the remaining component to 13C transverse magnetization for measurement. The 13C T1 is determined26 by analyzing how the generated 13C longitudinal magnetization decays with respect to τ. The 1H T1 corresponding to each 13C CP peak was determined by the pulse sequence shown in Figure 1d. The comb sequence applied to 1 H spins reduces 1H longitudinal magnetization to zero. During τ′, the 1H longitudinal magnetization recovers and is then transferred to 1H transverse magnetization by the 1H 90° pulse. CP then transfers the generated 1H transverse magnetization to 13 C transverse magnetization for measurement. The 1H T1 is determined by analyzing how the 1H longitudinal magnetization recovers during τ′. Figure 1e displays the pulse sequence27 to measure 13C homonuclear double quantum20,22 (DQ), which is generated by the through-space 13C homonuclear dipolar coupling,20,22 ∝ (I1 · I2)/r3 - ((I1 · r)(I2 · r))/r5, where I1 and I2 are two 13C spins and r is the distance between them. Because of the 1/r3 dependence, when the carboxylic carbons are enriched with the 13 C isotope, the 13C homonuclear DQ reveals the spatial connectivity between them. The back-to-back (BABA) pulse segment27 reintroduces the through-space homonuclear dipolar coupling under MAS. The excitation and reconversion periods each consist of four supercycles. Each supercycle covers four rotor periods and consists of two differently and appropriately phased BABA pulse segments.27 DQ is generated during the
13490
J. Phys. Chem. C, Vol. 113, No. 31, 2009
excitation period, and the generated DQ is encoded during t1. During the reconversion period, the generated DQ is transferred to 13C transverse magnetization to be recorded during t2. The DQ spectrum is obtained by Fourier transform with respect to t1, and the 13C single-quantum peaks are obtained by Fourier transform with respect to t2 which appear at the corresponding isotropic chemical shifts (σiso). Figure 1f displays the SUPER sequence28 to measure the powder pattern corresponding to each 13 C peak. The chemical-shift anisotropy (CSA) is obtained by fitting the powder pattern. When the carboxyl forms different bonds producing different electron-density distributions, the carboxylic carbon presents21,28-30 characteristically different powder patterns. The powder pattern measurement was thus conducted to investigate the chemical bonds for solid acids. Decanoic, dodecanoic, tetradecanoic, hexadecanoic, and octadecanoic acids (Cambridge Isotope Laboratory) were enriched with the 13C isotope for the carboxylic carbons with enrichment >99%. The TiNTs for the present work were produced by alkaline hydrothermal synthesis as described in refs 31 and 32. Specifically, 4 g of 32-nm anatase nanocrystals (Alfa Aesar) were mixed with 160 g of NaOH and 400 mL of distilled water. The mixture was placed in a Teflon container and inserted into a steel autoclave. The autoclave was placed in an oven at 140 °C for 72 h. The produced solid precipitate was collected, ground, and then placed in a centrifuge cylinder and topped off with fresh distilled water to produce water-washed TiNTs. The centrifuge cylinder was stirred and centrifuged, and the top solution was then sifted off. This water-wash process was repeated for 20-25 times until the pH value of the top water solution reached 5.5-6. The precipitated white powder was then dried in oven at 50 °C and was then ground using a mortar and pestle. The signature X-ray and Raman spectra on the resultant material confirmed that it contains exclusively TiNTs. As characterized by transmission electron microscopy, the waterwashed TiNTs are open-ended with hollow cylindrical pores. The inner and outer diameters are about 6.0 and 12.0 nm, respectively. To induce encapsulation of carboxylic acids, the TiNTs were mixed at room temperature with one of the five acids at a specific mass ratio. The mixture was then sealed in a Pyrex insert and thermally annealed near the bulk melting temperature of the acid (Tm,bulk). During annealing, the acid sublimed and was encapsulated in the nanotubes. 13C spectra were measured during annealing to monitor spectral change when encapsulation took place. When the spectra stopped changing upon further annealing, the mixture was considered to have attained the maximum encapsulation for the mass ratio. The fully annealed mixture, sealed in Pyrex inserts, was kept at room temperature for over 30 days before the measurement of melting behavior. Melting behavior was measured during warm-up for the encapsulated acids and, for comparison, for the bulk neat acids. Results and Discussions Figure 2 displays the observed changes of the carboxylic 13C CP spectrum versus the thermal annealing time at 315.5 K for a dodecanoic acid/TiNT mixture of 0.55:1 mass ratio. The bottom spectrum was recorded at 300.0 K before annealing. The large peak at 181.9 ppm, marked as “bulk peak” in Figure 2, shows the same σiso and the same 13C T1 ≈ 240 s as measured for the carboxylic 13C peak of the bulk neat dodecanoic acid. Thus, before thermal annealing, the mixture contained predominantly bulk dodecanoic acid. In addition to this bulk peak, the bottom spectrum also displays two small, new peaks with one peak shifting upfield and the other shifting slightly downfield.
Tang et al.
Figure 2. 13C CP spectra recorded for the dodecanoic acid/TiNT mixture of 0.55:1 mass ratio. The bottom spectrum was recorded at 300.0 K before annealing. The other spectra were recorded after 10, 35, 72, 108 minutes of annealing the mixture at 315.5 K. The bulk peak and the two new, sharp peaks are marked by arrows. The spectra are referenced externally to tetramethylsilane.
This suggests that a small fraction of dodecanoic acid was already sublimed and encapsulated during mixing, which is not surprising, given the low Tm,bulk of dodecanoic acid. The other spectra in Figure 2 were recorded after the specified durations of thermal annealing at 315.5 K. During annealing, the bulk peak slowly decreased while the two new peaks at 182.7 and 179.6 ppm, marked as “two new, sharp peaks” in Figure 2, slowly grew at the same rate. After annealing for over 100 min, as shown in the top spectrum, the bulk peak became invisible. The spectrum stopped changing. The pulse sequence shown in Figure 1d was used to measure the fully annealed mixture with the recovery time τ′ set to a series of different values from 0.01 to 90.0 s. Figure 3 displays the 13C spectra recorded with three representative τ′ values. As observed, the two new, sharp peaks at 182.7 and 179.6 ppm attain full intensity only when τ′ is set to tens of seconds. By contrast, a broad peak attains full intensity even for short τ′ and is thus relatively pronounced at short τ′. As determined by such measurement, the 1H T1 corresponding to the broad peak is ∼0.06 s. The 1H T1 values corresponding to the two new, sharp peaks within the experimental uncertainty are identical as 12.0 ( 2.0 s, which is similar to the determined 1H T1 corresponding to the bulk peak. Figure 4a and 4b displays the 13C spectra, recorded by the Torchia pulse sequence (Figure 1c) using the specified τ for the fully annealed mixture and for the bulk neat dodecanoic acid. As determined for the latter, the 13C T1 for the bulk peak is 240 ( 10 s. For the fully annealed mixture, the 13C T1 values for the two new, sharp peaks are approximately identical at 1025 ( 40 s, whereas the 13C T1 for the broad peak is
