Study of the Adsorption of Organic Molecules on Transition Metal

9630

J. Phys. Chem. B 2008, 112, 9630–9640

Study of the Adsorption of Organic Molecules on Transition Metal Exchanged Zeolites via Solid State NMR. Part 2: Adsorption of Organic Molecules on Zeolite NaX, CaX, and CaCoX Kristof Houthoofd,* Piet J. Grobet, and Pierre A. Jacobs Center for Surface Chemistry and Catalysis, Katholieke UniVersiteit LeuVen, Kasteelpark Arenberg 23, 3001 HeVerlee (LeuVen), Belgium ReceiVed: February 7, 2008; ReVised Manuscript ReceiVed: May 26, 2008

Literature [Denayer et al. Microporous Mesoporous Mater. 2007, 103, 1 and Denayer et al. Microporous Mesoporous Mater. 2007, 103, 11] shows that zeolite NaX exchanged with Ca2+ and Co2+ ions is able to remove cyclopentadiene (CPD) impurities from a 1-octene feed with high selectivity. In the present work, the adsorption of dicyclopentadiene (DCPD), CPD, 1-octene, and n-octane on zeolite X, exchanged with Ca2+ and/or Co2+ ions, has been investigated via 1H magic-angle spinning (MAS) NMR spectroscopy. The liquid adsorbate was dosed under inert atmosphere in an MAS rotor filled with dry adsorbent, at a pore filling degree of 70%. Next, the evolution in time was recorded of the 1H MAS NMR spectrum and the 1H spin-lattice and spin-spin relaxation times of the adsorbed components. For the various adsorbate-adsorbent systems, a plot is made of the signal intensity versus the square root of the contact time. It is found that, over the considered time interval, Fickian diffusion takes place. On the basis of the change in time of the spin-lattice relaxation time, a transport diffusion coefficient ranging between 1 and 2 × 10-15 m2 · s-1 is calculated. Moreover, there appear to be two sorption regimes, with different diffusivities. A comparison is made between the 1H spin-lattice relaxation behavior of DCPD, 1-octene, and n-octane, indicating that 1-octene and n-octane are located closer to the paramagnetic ions than DCPD. The average distance between the adsorbate molecules and the paramagnetic ions is derived from relaxometric data. By analyzing the chemical shifts of the resonance lines, it is found that the π-interaction of CPD and 1-octene is stronger than that of DCPD. 1. Introduction Modern polymerization processes require olefin feeds with high purity. Cyclic conjugated alkadienes may be present as impurities in commercial olefins and can poison the catalysts. Recently it has been reported that zeolite NaX successively exchanged with Ca2+ and Co2+ ions is a highly selective adsorbent for the removal of cyclopentadiene (CPD) impurities from 1-octene feed.1 For the described application as CPD sorbent, zeolite X has favorable geometrical properties. Whereas its minimum pore diameter is 740 pm, CPD has a kinetic diameter of 660 pm. CPD is thus able to interact with Co2+ ions positioned in the supercage. The structure of zeolite X, having the faujasite (FAU) framework topology, is represented in Figure 1.3–6 The framework is composed of sodalite cages, linked through double six-membered rings. The sodalite units encompass large cages or supercavities, with a diameter of 1270 pm, having 12, 6, and 4 faces. The 12-membered rings, with a free diameter of 740 pm, control the diffusion through the channels. The single 6-membered rings have a diameter of only 280 pm and are too small to admit guest molecules, with the exception of H2O, CO2, and NH3. As indicated in Figure 1, several cation sites are present in the zeolite.7–9 Only sites SII, SII*, and SIII are located inside the supercage, and are therefore available for interaction with guest molecules. The good performance of zeolite CaCoX for removal of traces of CPD from 1-octene can be ascribed to the accessibility of the transition metal ions and the stabilizing effect of the Ca2+ ions on the framework.1,10 By means of X-ray diffraction, it * Corresponding author. E-mail: [email protected].

Figure 1. Structure of zeolite X. The vertices represent the framework Si and Al atoms, and the lines represent the diameter of a lattice oxygen atom.3–9

has been demonstrated that, in fully dehydrated zeolite CaCoX, the Ca2+ ions are located at hidden sites, and the Co2+ ions occupy the SI′ and SII sites. Infrared spectroscopy also pointed to the existence of a high number of accessible Co2+ ions. The XRD spectrum of dry zeolite CaCoX revealed that the material remained crystalline. The SI Ca2+ ions proved to have a structure-stabilizing effect, since their coordination distance of 240 pm closely corresponds to that for the Ca2+ hexaquo complex. The Ca2+ ions thus neutralize the negative effect of Co2+ on the lattice stability. As Co2+ aquo complexes dissociate into H3O+ and CoOH+, the in-parallel-formed hydroxonium ions can give rise to dealumination and dissolution of the lattice.11 Moreover, cobalt ions are able to distort single 6-membered rings in zeolites, as demonstrated in a density functional theory (DFT) study.12

10.1021/jp801133x CCC: $40.75  2008 American Chemical Society Published on Web 07/17/2008

Adsorption of Organic Molecules on Zeolites Part 2

J. Phys. Chem. B, Vol. 112, No. 32, 2008 9631

TABLE 1: Cation Distribution (Cations Per Unit Cell) of the Water-Free Samples (According to Refs 10 and 38) sample NaX CaX CaCoX(1) CaCoX(5)

hexagonal prism

sodalite cage +

32 Na 11 Ca2+ 11 Ca2+ 11 Ca2+

8 Ca2+ + 9 Na+ 8 Ca2+ + 9 Na+ 8 Ca2+ + 9 Na+

In the present work, we have investigated the adsorption of a number of hydrocarbons on zeolite X by means of 1H magicangle spinning (MAS) NMR spectroscopy. The following compounds were used as adsorbate: dicyclopentadiene (DCPD), cyclopentadiene (CPD), 1-octene and n-octane. Four adsorbents were used, namely, zeolite NaX, CaX, CaCoX(1), and CaCoX(5), the number in brackets representing the percentage of the cation exchange capacity occupied by the paramagnetic Co2+ ions. Zeolite NaX and CaX also contain paramagnetic species, namely, iron and manganese impurities. 2. Experimental Methodology Experimental Procedure. A time-resolved in situ MAS NMR method has been applied, according to the following

supercavity Na+

54 39 Na+ 38 Na+ + 0.5 Co2+ 34.6 Na+ + 2.2 Co2+

residual Fe (ppm) 350 75 80 60

procedure. The zeolite sample was dried under a nitrogen flow at 350 °C for 12.5 h. A 4 mm MAS rotor was filled under inert atmosphere with the dry zeolite, and the 1H MAS NMR spectrum of the dry zeolite was recorded. Subsequently, an amount of adsorbate was added under inert atmosphere to the dry zeolite, using a pore filling degree of 70% (70% of the theoretical pore volume of 0.357 mL per gram zeolite). The rotor was tightly closed in order to avoid the intrusion of moisture. The 1H MAS NMR spectrum of the sample was recorded as rapidly as possible (11 min) after the addition of the adsorbate, and the proton spin-lattice and spin-spin relaxation times were measured. Next, the evolution in time of the spectrum, the spin-lattice relaxation time (T1), and the spin-spin relaxation time (T2) was recorded. All NMR mea-

Figure 2. 1H MAS NMR spectra of 1-octene adsorbed on several zeolites. The initial spectrum is represented in blue, and the final spectrum is shown in green. t represents the time that has elapsed between the start of the addition of the adsorbate, and the start of the acquisition of the spectrum.

9632 J. Phys. Chem. B, Vol. 112, No. 32, 2008

Figure 3.

1

Houthoofd et al.

H MAS NMR spectra of DCPD adsorbed on several zeolites.

surements were performed at room temperature. DCPD, 1-octene, and n-octane were obtained from Acros Organics; the purities were 95, 97, and 99%, respectively. CPD was prepared by thermal cracking of DCPD at 170 °C under nitrogen atmosphere. The measurements were repeated two times with fresh samples. The NMR measurements were performed with a Bruker AMX300 spectrometer (7.0 T). For the acquisition of the 1H MAS NMR spectra, a single-pulse excitation method was used, accumulating 32 scans with a recycle delay of 10 s. Tetramethylsilane (Fluka) was used as a shift reference. The spin-lattice relaxation times were measured by means of a sequence of the form: [π/2-τ-π/2-aq].13 For the measurement of the spin-spin relaxation times, a [π/2-τ-π-τ-aq] sequence was used.14 Preparation of Adsorbents. Zeolite NaX, with unit cell composition Na86Al86Si106O384, was purchased from Ueticon. Zeolite CaX and CaCoX were prepared by ion exchange of NaX. Two different CaCoX adsorbents were prepared, namely, CaCoX(1) and CaCoX(5). The number between brackets is the percentage of the cation exchange capacity that is occupied by Co2+. The Co2+ loadings in CaCoX(1) and CaCoX(5) correspond to 0.5 and 2.2 Co2+ ions per unit cell, respectively.

ICP determination of the iron content of the samples, shows that zeolites NaX, CaX, CaCoX(1), and CaCoX(5) contain 350, 75, 80, and 60 ppm of iron, respectively. It seems that ion exchange with Ca2+ removes part of the residual iron in NaX zeolite. The following procedures were used for the preparation of the zeolite X adsorbents. For the synthesis of CaX, a solution of 1.731 g CaCl2 · 2H2O (Riedel-de-Hae¨n) and 3 L of distilled water was mixed for 24 h with 22 g of NaX at room temperature. The mixture was centrifugated at 3500 rpm for 20 min and washed with distilled water. The material was dried overnight at 60 °C, and subsequently calcined (100 °C, 1 °C/min for 1 h; 200 °C, 1 °C/min for 1 h; 500 °C, 1 °C/min for 20 h). For the synthesis of CaCoX(1), a solution of 50 mL of distilled water and 11.4 mg of CoCl2 · 6H2O (Alfa) was mixed at room temperature for 24 h with 3.0 g of calcined CaX. The mixture was filtered with a Bu¨chner and washed with distilled water. The material was dried overnight at 60 °C. For the synthesis of CaCoX(5), a solution of 50 mL of distilled water and 57.6 mg of CoCl2.6H2O (Alfa) was mixed at room temperature for 24 h with 3.0 g of calcined CaX. The



Figure 4. Comparison between the initial spin-lattice relaxation behavior of the adsorbates DCPD, 1-octene, and n-octane. t1 represents the initial time, i.e., 11 min after the addition of the adsorbate. The numbers above the bars refer to the distinct proton types in the adsorbate.

mixture was filtered with a Bu¨chner and washed with distilled water. The material was dried overnight at 60 °C. The dry Ca-exchanged sample, denoted further as CaX, contains 19 Ca2+ and 48 Na+ ions per unit cell. A Rietveld refined cation distribution in the dry sample showed the presence of 19 Ca2+ and 9 Na+ in the hidden cages and 39 Na+ ions in the accessible supercavities.10 All subsequently exchanged Co2+ ions replace residual Na+ cations, and are located in the supercavities. The cation distribution among the different sites for the samples used is given in Table 1. All Ca2+ ions remain in sites hidden for direct interaction with the sorbate molecules, while all exchanged Co2+ ions remain accessible for direct interaction with adsorbate. The population of supercavity sites with Na+ decreases in the following sequence: NaX > CaX ≈ CaCoX(1) > CaCoX(5). The available space for sorbate diffusivity should increase in this sequence.15 3. Results and Discussion. The 1H MAS NMR spectra of adsorbed DCPD, CPD, 1-octene, and n-octane after initial contact (11 min) and equilibration (431 min) on the four zeolite samples used were scanned. As an example, the data for

1-octene and DCPD adsorption are shown in Figures 2 and 3, respectively. The assignment of the different lines based on the spectra of the liquid sorbates is indicated as well. The same resonance lines are present for the liquid and zeolite occluded adsorbates, although, with respect to the bulk liquid phase, they are significantly broadened for the adsorbate in the zeolite cages. Only for zeolite adsorbed 1-octene, new resonance lines (d′ and e′) occur compared to the liquid phase. Both lines correspond to the resonance of the double bond H atoms, involved in an increased deshielding. Possibly, they correspond to strongly adsorbed alkenes on supercage Co2+ ions and/or paramagnetic impurities. They will not be considered further on. For n-octane adsorption, the spectral changes after equilibration are minor compared to the initial states. On CaX, the initial spectrum shows a resolution similar to that of the liquid phase. On the other zeolite adsorbents, significant line broadening has already occurred in the initial state. Clearly, the sorbent that causes the lowest degree of broadening (CaX) also has the lowest number of paramagnetic centers (Fe or Co) (Table 1). Upon 1-octene adsorption, the line extinction and broadening


Houthoofd et al.

TABLE 2: Values for δ′, the Difference in Line Position of the Adsorbates in the Adsorbed and Liquid Phase δ′ ) δads - δliq (ppm) adsorbate

adsorbent

line g

line f

line e

line d

line c

line b

line a

DCPD

NaX CaX CaCoX(1) CaCoX(5)

-0.01 -0.02 -0.01 -0.02

-0.02 -0.03 -0.02 -0.03

0.00 0.00 0.00 -0.01

0.01 0.00 0.00 -0.02

0.02 0.02 0.02 0.02

0.00 0.00 -0.01 0.00

0.00 -0.01 -0.01 0.00

adsorbate

adsorbent

line c

line b

line a

CPD


0.49 0.44 1.43

0.49 0.50 1.43

0.47 0.49 1.59

adsorbate

adsorbent

line e′

line e

line d′

line d

line c

line b

line a

l-octene


0.80 0.75 0.80

0.09 0.11 0.09 0.07

0.52 0.42 0.52

0.05 0.13 0.05 -0.02

0.05 0.17 0.05 0.05

0.05 0.18 0.05 0.05

0.05 0.20 0.05 0.01

adsorbate

adsorbent

line b

line a

n-octane


0.09 0.11 0.27 0.69

0.01 0.11 0.28 0.36

is most pronounced for the CaCoX samples, compared to that of CaX (Figure 2). Adsorption of DCPD shows the most pronounced line broadening and line intensity decrease on CaCoX(5) (Figure 3), the sample with the highest concentration of paramagnetic Co centers. These results on a semiquantitative basis indicate that the presence of paramagnetic sites is at the basis of reduction of intensity and broadening of the NMR lines of an adsorbate. The initially measured 1H spin-lattice relaxation times (T1) of the adsorbed components are represented in Figure 4. The whole numerical data set is added as Supporting Information. Analysis of the Chemical Shifts of Adsorbate in the Initial Spectra. The chemical shifts of the resonance lines in the initial spectrum of the adsorbed components can be compared with those for the pure liquid components. For each line, the difference δ′ ) δads - δliq has been calculated, with δads representing the position of a line in the initial spectrum of the adsorbed component, and δliq being the position in the spectrum of the liquid component. The obtained values are shown in Table 2. A downfield shift of the olefinic signals is observed with CPD and 1-octene, but not with DCPD. In case of CPD, the shift increases with increasing cobalt loading. These shifts are most

probably due to the formation of a strong complex between the paramagnetic ions and the surrounding π-electron system. It can thus be concluded that both CPD and 1-octene have a relatively strong π-interaction with the paramagnetic ions, while DCPD shows a relatively weak π-interaction. The relaxometric data also indicate that 1-octene interacts more strongly with the paramagnetic ions (Co2+ and/or paramagnetic iron impurities) than DCPD. Indeed, the spin-lattice relaxation times of 1-octene (Figure 4) are considerably shorter than those of DCPD. Diffusivity Aspects of Adsorbate Molecules in Zeolites Derived from Line Intensities. On the basis of the NMR data, it will now be attempted to derive information on the diffusion behavior of the adsorbates. In general, two types of zeolitic diffusion are described in the literature, namely, transport and self-diffusion. The former process refers to the transport of adsorbate molecules under the influence of a concentration gradient, and can be described by Fick’s laws. The latter refers to the mutual exchange of position of the molecules at equilibrium, i.e., in the absence of a concentration gradient.16,17 Conventional methods for measuring the transport diffusivity include uptake rate measurements and frequency response analyses.18–20 In the first method, the sample is subjected to a step change in the external gas phase pressure, while the change

Figure 5. Plot of the slope of I′ ) f(t1/2) (eq 4a) for the various adsorbate-adsorbent combinations.



Figure 6. Plot of the difference y ) (1/T1)∞ - (1/T1)t as a function of the square root of time, for the system CaCoX(5) + DCPD (according to eq 7).

TABLE 3: Average Diffusion Coefficient over All Individual Resonance Lines of the Adsorbate Molecules for the First Regime Dc (10-15 m2 · s-1) adsorbate DCPD 1-octene n-octane

NaX

CaX

CaCoX(1)

CaCoX(5)

1.745 ( 0.035 1.500 ( 0.030 1.558 ( 0.031 1.424 ( 0.028 1.661 ( 0.033 1.964 ( 0.039 1.849 ( 0.037 1.754 ( 0.035 1.363 ( 0.027 1.343 ( 0.027 1.070 ( 0.021 1.908 ( 0.038

of the sample weight is monitored in time. The second method consists of applying a sinusoidal modulation to the gas phase volume, and recording of the induced pressure response. In order to measure the self-diffusivity, the so-called pulsed field gradient (PFG) NMR method can be used.24–28 Unfortunately, NMR methods do not allow to determine diffusional characteristics in zeolite samples containing paramagnetic ions or occlusions. For 1H MAS NMR spectra of the various adsorbed components on the zeolite samples mentioned earlier, it is observed that the resonance lines become significantly broader and less intense as a function of the contact time with the adsorbate. These effects can be ascribed to the diffusive transport of the

adsorbate. Indeed, when this type of diffusion takes place, the adsorbate proton ensembles will approach the paramagnetic ions more closely as a function of time. As can be shown theoretically, this gives rise to a broadening and a reduction of the intensity of the resonance line of the composing ensembles.21 It is reasonable to assume that the linear structure of 1-octene and n-octane, as well as the cyclic planar structure of CPD, allows faster diffusion, while the shape of DCPD should cause reduced diffusivity. In the case of n-octane, the resonance lines do not change considerably in time. This may indicate that the diffusion of n-octane has already taken place to a large extent during the initial time interval (i.e., 11 min after the addition of the liquid to the rotor). The occurrence of an extinction indicates that the fraction of the adsorbate protons located inside a critical volume gradually increases in time.21 This implies that protons are located at a distance less than about 1 nm from the paramagnetic ions. This distance is close to the supercavity diameter of zeolite X (1.27 nm). It should be realized that the maximum number of paramagnetic sites (2.2 Co2+ ions per unit cell) in the CaCoX(5) sample corresponds to an occupation of about one Co2+ ion for


Houthoofd et al.

TABLE 4: Values for the Average of the Parameter 〈R〉 over All Individual Resonance Lines of a Given Adsorbate 〈R〉avg (nm) sorbent NaX

CaX

CaCoX(1)

CaCoX(5)

t (min.)

DCPD

1-octene

n-octane

11 71 131 191 251 311 371 431 11 71 131 191 251 311 371 431 11 71 131 191 251 311 371 431 11 71 131 191 251 311 371 431

0.838 ( 0.003 0.802 ( 0.005 0.797 ( 0.006 0.796 ( 0.007 0.794 ( 0.003 0.792 ( 0.003 0.786 ( 0.009 0.790 ( 0.003 0.877 ( 0.011 0.842 ( 0.002 0.832 ( 0.004 0.830 ( 0.005 0.827 ( 0.005 0.825 ( 0.004 0.824 ( 0.003 0.821 ( 0.003 0.846 ( 0.006 0.808 ( 0.005 0.798 ( 0.007 0.796 ( 0.007 0.793 ( 0.007 0.792 ( 0.008 0.790 ( 0.007 0.787 ( 0.006 0.838 ( 0.015 0.766 ( 0.006 0.744 ( 0.002 0.736 ( 0.005 0.732 ( 0.003 0.728 ( 0.003 0.726 ( 0.005 0.721 ( 0.003

0.650 ( 0.006 0.616 ( 0.004 0.606 ( 0.006 0.601 ( 0.006 0.599 ( 0.005 0.598 ( 0.005 0.597 ( 0.005 0.597 ( 0.005 0.702 ( 0.002 0.683 ( 0.001 0.675 ( 0.002 0.673 ( 0.003 0.673 ( 0.002 0.673 ( 0.002 0.672 ( 0.004 0.672 ( 0.003 0.689 ( 0.002 0.658 ( 0.005 0.649 ( 0.006 0.643 ( 0.007 0.643 ( 0.008 0.642 ( 0.006 0.643 ( 0.004 0.642 ( 0.005 0.621 ( 0.007 0.583 ( 0.010 0.571 ( 0.008 0.566 ( 0.007 0.565 ( 0.006 0.564 ( 0.006 0.563 ( 0.006 0.562 ( 0.006

0.601 ( 0.002 0.586 ( 0.002 0.581 ( 0.002 0.579 ( 0.003 0.577 ( 0.002 0.576 ( 0.002 0.574 ( 0.001 0.573 ( 0.001 0.679 ( 0.006 0.674 ( 0.006 0.667 ( 0.006 0.665 ( 0.006 0.663 ( 0.006 0.661 ( 0.006 0.661 ( 0.007 0.662 ( 0.008 0.654 ( 0.001 0.646 ( 0.002 0.643 ( 0.002 0.639 ( 0.002 0.636 ( 0.003 0.637 ( 0.003 0.635 ( 0.003 0.635 ( 0.003 0.610 ( 0.013 0.581 ( 0.003 0.573 ( 0.003 0.570 ( 0.004 0.567 ( 0.002 0.565 ( 0.001 0.573 ( 0.005 0.568 ( 0.002

every four supercages. Thus the adsorbate on the average has to cross at least one empty supercage before it enters a “critical” supercavity occupied by a paramagnetic center. For all the systems studied, the spin-lattice relaxation time, T1, of the resonance lines decreases in time (Supporting Information). The observed changes can be ascribed to the diffusion of the adsorbate toward the paramagnetic centers. An exception is the behavior of the shifted double bond lines (lines d′ and e′) of 1-octene, which show a constant T1 value. These lines probably originate from molecules that form a strong complex with paramagnetic ions. The invariance of the T1 values possibly indicates that the diffusion of these molecules is fast, and has already taken place before the first measurement is done. By plotting the intensity of the resonance lines as a function of the square root of time, it can be appreciated whether the diffusion is Fickian in nature. This can be rationalized in the following way: The global proton spin system consists of strongly and weakly adsorbed protons, denoted further as “bound” and “free” protons, respectively. The intensity of a resonance line decreases with the number of bound ensembles, since there exists a paramagnetic extinction effect. When it is assumed that this decrease is strictly linear, it follows that

I(t) ) A - B

n(t) N

I(t) ) I(0) - B

n(t) N

(2)

When the diffusion is Fickian, the number of bound protons is given by eq 3.

n(t) ) n(∞)

()

6 Dc √π rc2

1⁄2

√t

(3)

where n(∞) represents the number of bound protons at equilibrium, Dc is the intracrystalline transport diffusion coefficient, and rc is the crystal radius.22,23 By substitution into eq 2, eqs 4 are obtained:

I(t) ) I(0) - B k)

n(∞) k√t N

()

6 Dc √π rc2

(4a)

1⁄2

(4b)

Equation 4a suggests that the slope of the function I ′ ) f(t1/2) is a measure of the diffusivity. The values for the slope of the function I ′ ) f(t1/2) are represented in Figure 5. It should be noted that the obtained values for the slope are related to the diffusivity during the time interval over which the spectra were recorded, i.e., the interval between 11 and 431 min after addition of the adsorbate. The data indicate that CPD has a significantly reduced diffusivity in all zeolite samples, while for DCPD and n-octane comparable values are found, except perhaps on CaCoX(5), for which n-octane diffusion seems to be faster than for DCPD. 1-Octene on all samples shows intermediate values between those of CPD and DCPD/n-octane. The influence of the type of adsorbent on the diffusivity of a given adsorbate can also be analyzed. The cation distribution data for the four zeolite samples have been shown in Table 1. The diffusivity in the supercavity probably will be determined by the number of the available cations. Therefore, it seems that diffusion is slower in NaX, compared to all other samples. The correlation with the high number of supercavity Na+ ions is obvious. For n-octane and DCPD, there is the same correlation between diffusivity and number of supercavity Na+ ions; the CaCoX(5) sample with the lowest number of such ions shows the fastest diffusion. This hypothesis rationalizes why comparable values are obtained for CaX and CaCoX(1). For 1-octene a similar trend is observed: NaX < CaX ≈ CaCoX(1), the reduced value on CaCoX(5) pointing to the existence of a strong interaction with supercavity Co2+ ions. For CPD, the available data are in line with this hypothesis, at least when the reduced diffusivity values for CaCoX(5) are considered. The lower value on CaCoX(5) compared to CaCoX(1) is not unexpected. The very low value on CaX is difficult to rationalize. Calculation of the Diffusional Time Constants and Diffusion Coefficients from T1 Values. A method will now be described that allows one to determine the diffusional time constant, based on the time evolution of the spin-lattice relaxation time of an adsorbate. The spin-lattice relaxation rate constant (1/T1) increases with the number of bound ensembles. By assuming a linear increase, it follows that

(1)

where n(t) represents the number of bound protons at time t, and N is the number of protons in an ensemble. At time t ) 0, no protons are bound, so that A ) I(0). Consequently, eq 2 is obtained.

1 n(t) )A+B T1 N

(5)

In the case of Fickian diffusion, the number of bound protons obeys eq 6,


n(t) ) n(∞)

()

6 Dc √π rc2

1⁄2

√t


(6)

ensemble, due to the interaction with the paramagnetic ions, is given by eq 8:21

() 1 T1

where Dc represents the intracrystalline transport diffusion coefficient. Combining these equations yields

() () 1 T1

-

∞

()

1 1 6 Dc ) -B n(∞) T1 t N √π rc2

1⁄2

1 √t + B n(∞) (7) N

The diffusional time constant can thus be determined by plotting the difference y ) (1/T1)∞ - (1/T1)t as a function of t1/2, and calculating the line of least-squares fitting the data points. The numerical value of B/N n(∞) is obtained from the section of the straight line on the ordinate. It allows to calculate values of (Dc/rc2) from the slope. Graphs of y ) f(t1/2) have been made for all measured adsorbate-adsorbent systems; a typical example is shown in Figure 6. During the first three hours of the experiments, i.e., up to t ) 191 min, a linear dependence was observed, indicating that, in this time interval, the diffusion is Fickian in nature. Data points taken at later adsorption times with low y values can also be fitted to a straight line with significantly reduced slope (see later). The correlation coefficients for the calculated regression lines are close to 1 (>0.931). In all cases, it is immaterial which line of the spectrum is taken to determine diffusional time constants from the T1 values. For each adsorbate, the average of the diffusional time constant over all individual resonance lines has been calculated. The diffusion coefficient has been calculated from the average diffusional time constant, using a crystal radius of 2.0 µm. The initial diffusion coefficients, Dc, which can be extracted from the initial slope, are shown in Table 3. It is clear that two different regimes can be considered. From the data obtained initially, i.e., in the first regime, the following information can be obtained: 1. Average values of the diffusion coefficients using the different resonance lines of an adsorbate show acceptable accuracies. 2. The order of magnitude of diffusion coefficients (1 to 2 × 10-15 m2/s) is in good agreement with values determined by conventional methods for probing diffusive transport on the same zeolites free of paramagnetic ions.29,30 3. In all cases, combining different sorbents and sorbates, the values of the initial Dc are significantly higher than those derived from the adsorption data at longer contact times. In other words, the values for Dc are coverage dependent. It has indeed been reported that the Fickian diffusivity for diffusion in zeolites is generally a function of the adsorbate concentration.31 4. The Dc value obtained during the initial sorption phase of n-octane is higher on the zeolite sample with the lowest number of ions in the supercavities (CaCoX(5)) (Table 1). 5. The Dc values for DCPD on CaX and CaCoX(1) are comparable. This is consistent with the arguments of space availability determined by the number of Na+ ions in the supercavities, being approximately equal for CaX and CaCoX(1). 6. For 1-octene adsorption, diffusion in zeolite NaX in the initial regime is still slowest. The high diffusivity in CaX is in line with the absence of Co2+ compared to CaCoX and the lower number of supercavity Na+ compared to NaX. Average Distance between Adsorbate Molecules and Paramagnetic Ions from T1 Values. The relaxometric data can be used for estimating the average distance between an adsorbate molecule and the paramagnetic ions of the sorbent. The spin-lattice relaxation rate constant of a uniform proton

) (C1)IS IS

with

(C1)IS )

2τz 1 + ΩI2τz2

∑ sin2 θis cos2 θisris-6

(8)

s

( )

3 µ0 2 2 2 2 2 γ g γ p S(S + 1) 2 4π I e e

(9)

This expression follows directly from the quantum mechanical formula for the transition probability in a nuclear spin system.32 When it is assumed that each proton is influenced by only one paramagnetic ion, and the angular-dependent factor is replaced by its average value of 2/15,33–36 this expression reduces to

1 ) (C1 ′)ISris-6 T1 with

(C1 ′)IS )

( )

(10)

2τz 1 µ0 2 2 2 2 2 γI ge γe p S(S + 1) ) 5 4π 1 + Ω 2τ 2 I

z

2τz

2 (C ) (11) 15 1 + Ω 2τ 2 1 IS I

z

Therefore, for a uniform proton ensemble, the following relation exists between the I-S distance and the spin-lattice relaxation time:

ris ) [(C1 ′)IST1]1⁄6

(12)

When a value of 1.0 × 10-12 s is used for τz,37 and a value of 3/2 is used for S, the factor C1′ is equal to 3.70 × 10-55 m6 · s-1. The estimation of τz is based on the literature, where the spin-lattice relaxation of cobalt ions in an aluminophosphate molecular sieve is discussed.37 As previously explained, the spin-lattice relaxation time of an overall proton system can be approached by averaging over the composing ensembles. Therefore, an estimate of the average distance between an overall proton system and the paramagnetic ions is given by

〈R〉 ) [(C1 ′)IST1]1⁄6

(13)

where T1 is the measured spin-lattice relaxation time. The calculated distances (Table 4) are in a range between 0.5 and 1.0 nm. These values have the same order of magnitude as the supercavity diameter (1.27 nm), indicating that the adsorbate molecules do not remain at the external surface of the zeolite grains, but effectively enter the micropores. When a value of 1/2 or 5/2 is used for S, the values obtained for 〈R〉 are again of the order of the supercage diameter. The values for 1-octene (Table 4) are of the same order of magnitude as the values obtained from a cation position localization by XRD on CoCaX zeolites.10 The applied method for calculating the average distance is only valid when the adsorbate protons are influenced by only one paramagnetic ion. It may indeed be assumed that this is the case, since the used adsorbents have a low paramagnetic ion content. As an example, for zeolite CaCoX(1) and CaCoX(5), the separation between two neighboring paramagnetic centers is equal to 2.00 and 1.17 nm, respectively. The distance between two neighboring paramagnetic ions (rss) can be calculated in the following way. With N, the number of paramagnetic ions per unit volume, there will be one


Houthoofd et al.

Figure 7. Comparison between the initial values of 〈R〉 for the adsorbates DCPD, 1-octene, and n-octane. t1 represents the initial time.

paramagnetic ion in 1/N volume units. This volume can be considered as a volume of a sphere with radius equal to rss:

4 1 πr 3 ) 3 ss N

(14)

From this equation, it follows that

rss )

( N1 4π3 )

1⁄3

(15)

For samples CaCoX(1) and CaCoX(5), the value of N is equal to 0.030 and 0.149 Co2+ ions/nm3, respectively, since the unit cell has a volume of 14.43 nm3. Inserting these values in eq 9 allows one to calculate rss. The consistency of the data set obtained is obvious when the values of R〈R〉 are considered, i.e., the ratio of the steady and initial values for 〈R〉. Indeed, for a given adsorbent and adsorbate, every value for this ratio, within experimental error, is identical. The average distance, 〈R〉, between a proton system (the adsorbate) and a paramagnetic center (in the adsorbent), in

all cases does not exceed the free diameter of a supercavity (around 1.2 nm). For a given adsorbate, the 〈R〉 value of a given resonance line is also hardly dependent on the nature of the adsorbent. The 〈R〉 value for a given adsorbate depends on its nature. For n-octane, this distance is shorter than in case of DCPD (Figure 7). On the other hand, comparing n-octane with 1-octene, this difference in distance is much reduced. Thus it seems that the double bonds of the bulky DCPD molecule have difficulties interacting with residual Na ions in the large cages, in contrast to what happens with 1-octene. It seems that, for the 〈R〉 values, no significant changes are obtained depending on the nature of the resonance line of the adsorbate, viz., no difference is found for methyl- or methylene H-atoms. Relaxation Behavior of Adsorbates from T1 Values. The data allow one to make a comparison between the spin-lattice relaxation behavior of the different adsorbates (Figure 4). It is observed that the T1 values of 1-octene and n-octane are considerably lower than those of DCPD. The T1 values of



Figure 8. Comparison between the initial spin-spin relaxation behavior of the adsorbates DCPD, 1-octene, and n-octane. t1 represents the initial time.

1-octene and n-octane have approximately the same order of magnitude. This is found to be the case for each adsorbent, and at both the initial and final time. These observations indicate that 1-octene and n-octane are located closer to the paramagnetic ions than DCPD. Indeed, it can be shown that the spin-lattice relaxation time of a proton system increases with the average distance from the paramagnetic ions.21 Relaxation Behavior of Adsorbates from T2 Values. Physicochemically, spin-lattice relaxation represents the interaction of the adsorbate molecules with the paramagnetic centers. On the other hand, spin-spin relaxation represents both the interaction of the adsorbate molecules with the paramagnetic centers, and their exchange. The calculated values for T2, after initial contact of sorbates with the zeolites, are shown in Figure 8. It is observed on all adsorbents that the T2 values of 1-octene are lower than those of DCPD and n-octane. The T2 values of n-octane are slightly higher than or approximately equal to those of DCPD. This indicates that, in the case of n-octane, molecular exchange takes place to a larger extent than with 1-octene and DCPD. In order to clarify this, we recall that the spin-lattice and spin-spin relaxation time increase with the average distance

between the adsorbate and the paramagnetic ions.21 However, the T2 value also increases with the degree of molecular exchange.21 This implies that if a given adsorbate is located closer to the paramagnetic ions, but has a higher degree of exchange, it may obtain a longer T2 value. Conclusions In the present work, a MAS NMR study has been performed of the adsorption of DCPD, CPD, 1-octene, and n-octane on zeolite NaX exchanged with Ca2+ and/or Co2+ ions. The 1H MAS NMR spectrum and the 1H spin-lattice and spin-spin relaxation times of the adsorbed components have been measured. The change in time of the NMR signal and the relaxation times has been analyzed in detail. For a number of systems, the resonance lines become significantly broader and less intense in time. The following systems exhibit this behavior: DCPD adsorbed on NaX, CaX, and CaCoX; CPD adsorbed on CaCoX; 1-octene adsorbed on NaX and CaCoX. These effects can be ascribed to diffusive transport. Upon adsorption of n-octane, only minor spectral changes are observed. This may indicate

9640 J. Phys. Chem. B, Vol. 112, No. 32, 2008 that the diffusion has already taken place before the first spectrum was recorded. Indeed, it can be expected that the diffusion of n-octane is fast because of its structure. From the variation of the spectral intensity as a function of the square root of time, it is found that for the studied systems, the diffusion is Fickian. For the four zeolite samples, the diffusivity decreases in the following sequence: n-octane ≈ DCPD > 1-octene > CPD. Further, the data indicate that, for a given adsorbate, the transport diffusivity is determined by the type of adsorbent. This can be interpreted in terms of the population of the cation sites in the supercavities with Na+ ions. Based on the evolution in time of the 1H spin-lattice relaxation time, the transport diffusion coefficient (Dc) has also been calculated. The method consists of plotting the difference between 1/T1 at equilibrium and 1/T1 at a given contact time, versus the square root of the contact time. The diffusional time constant can then be determined from the slope and intercept of the straight line. In the experimental graphs, two different sorption regimes could be discerned. All Dc values obtained from the line slopes in the second regime are smaller than the corresponding values in the initial regime, due to a higher degree of pore filling. The order of magnitude of the calculated values (1 to 2 × 10-15 m2/s) corresponds well with values reported in the literature, determined via conventional methods on similar samples devoid of paramagnetic species. By comparing the initial 1H spin-lattice relaxation times of the different adsorbates, it is found that 1-octene and n-octane are situated closer to the paramagnetic ions than DCPD. This is most probably due to sterical factors. From the magnitude of the 1H spin-lattice relaxation time, an estimate has been made of the average distance between an adsorbate molecule and the paramagnetic ions. The obtained values are in a range between 0.5 and 1.0 nm, and thus have the same order of magnitude as the supercage diameter. As an example, on the adsorbent CaCoX(5), the initial distance is 0.84 nm for DCPD, 0.62 nm for 1-octene, and 0.61 nm for n-octane. It has been examined whether in the adsorbed state, the resonance lines are shifted with respect to the position in the liquid state. Significant shifts were observed for the olefinic protons of CPD and 1-octene, but not for those of DCPD. This indicates that CPD and 1-octene should form stronger complexes with the transition metal ions. Acknowledgment. The work is sponsored in the frame of the IAP network (IAP-PAI G/26) and the GOA action. K.H. acknowledges a grant from the same program. Supporting Information Available: T1 values of the protons in adsorbed DCPD, 1-octene, and n-octane. This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes (1) Denayer, J. F. M.; Huybrechts, B.; Depla, A.; Hermans, Y.; Gemoets, F.; van Buren, F. R.; Kirschhock, C.; Baron, G. V.; Jacobs, P. Microporous Mesoporous Mater. 2007, 103, 1. (2) Denayer, J. F. M.; Depla, A.; Vermandel, W.; Gemoets, F.; van Buren, F. R.; Martens, J.; Kirschhock, C.; Baron, G. V.; Jacobs, P. A. Microporous Mesoporous Mater. 2007, 103, 11. (3) http://www.iza-structure.org, Sept. 2007. (4) Klinowski, J. Prog. NMR Spectrosc. 1984, 16, 237.

Houthoofd et al. (5) Engelhardt, G.; Michel, D. High-Resolution Solid-State NMR of Silicates and Zeolites; John Wiley and Sons: New York,1987; pp 210211. (6) Dwyer, J.; Dyer, A. Chem. Ind. (London) 1984, 237. (7) Mortier, W. J. Compilation of Extra Framework Sites in Zeolites; Butterworth: London, 1982; p 19. (8) Yang, R. T. Adsorbents: Fundamentals and Applications; John Wiley and Sons: New York, 2003; p 177. (9) Kowenje, C.; Jones, B. R.; Doetschman, D. C.; Yang, S.-W.; Kanyi, C. W. Chem. Phys. 2006, 330, 401. (10) Denayer, J. F. M.; Depla, A.; Vermandel, W.; Gemoets, F.; van Buren, F. R.; Martens, J.; Kirschhock, C.; Baron, G. V.; Jacobs, P. A. Microporous Mesoporous Mater. 2007, 103, 11. (11) Bae, D.; Seff, K. Microporous Mesoporous Mater. 1999, 33, 265. (12) Pierloot, K.; Delabie, A.; Ribbing, C.; Verberckmoes, A. A.; Schoonheydt, R. A. J. Phys. Chem. B 1998, 102, 10789. (13) The sequence consists of two π/2 pulses, separated by a variable delay τ; acquisition of the free induction signal was started after the second pulse. A series of two-pulse spectra with variable delay τ was recorded, and a plot was made of the peak amplitude versus τ. Subsequently, the spin-lattice relaxation time was determined by means of the least-squares method. (14) The sequence consists of a π/2 and π pulse, separated by a variable delay τ. At a time t ) τ after the π pulse, the free induction signal was acquired. A series of two-pulse spectra with variable delay τ was recorded, and the peak amplitude was plotted versus τ. The spin-spin relaxation time was calculated using the least-squares method. (15) Chen, N. Y.; Degnan, T. F.; Smith, C. M. Molecular Transport and Reaction in Zeolites-Design and Application of Shape SelectiVe Catalysts; VCH Publishers, New York, 1994; pp 36-38. (16) Chen, N. Y.; Degnan, T. F.; Smith, C. M. Molecular Transport and Reaction in Zeolites-Design and Application of Shape SelectiVe Catalysts; VCH Publishers, New York, 1994; pp 76-77. (17) Crank, J. The Mathematics of Diffusion, 2nd ed.; Oxford University Press: Oxford, 1986; p 212. (18) Chen, N. Y.; Degnan, T. F.; Smith, C. M. Molecular Transport and Reaction in Zeolites-Design and Application of Shape SelectiVe Catalysts; VCH Publishers, New York, 1994; pp 139-152. (19) Ka¨rger, J.; Vasenkov, S. Microporous Mesoporous Mater. 2005, 85, 195. (20) Song, L.; Sun, Z.; Duan, L.; Gui, J.; McDougall, G. S. Microporous Mesoporous Mater. 2007, 104, 115. (21) Houthoofd, K.; Grobet, P. J.; Jacobs, P. A. Study of the Adsorption of Organic Molecules on Transition Metal Exchanged Zeolites via Solid State NMR. Part 1: Theoretical Aspects. J. Phys. Chem. B 2008, 112, 9625. (22) Chen, N. Y.; Degnan, T. F.; Smith, C. M. Molecular Transport and Reaction in Zeolites-Design and Application of Shape SelectiVe Catalysts; VCH Publishers, New York, 1994; pp 139-140. (23) Rodrı´guez, O.; Fornasiero, F.; Arce, A.; Radke, C. J.; Prausnitz, J. M. Polymer 2003, 44, 6323. (24) Stejskal, E. O.; Tanner, J. E. J. Chem. Phys. 1965, 42, 288. (25) Stejskal, E. O. J. Chem. Phys. 1965, 43, 3597. (26) Ka¨rger, J. ; Pfeifer, H. NMR Studies of Molecular Diffusion. In NMR Techniques in Catalysis; Bell, A. T., Pines, A. , Eds.; Marcel Dekker: New York, 1994; pp 84-94. (27) Chen, N. Y.; Degnan, T. F.; Smith, C. M. Molecular Transport and Reaction in Zeolites-Design and Application of Shape SelectiVe Catalysts; VCH Publishers: New York, 1994; pp 152-153. (28) Ka¨rger, J.; Pfeifer, H. Zeolites 1987, 7, 90. (29) Vavlitis, A. P.; Ruthven, D. M.; Loughlin, K. F. J. Colloid Interface Sci. 1981, 84, 526. (30) Schemmert, U.; Ka¨rger, J.; Weitkamp, J. Microporous Mesoporous Mater. 1999, 32, 101. (31) Yang, R. T. Adsorbents: Fundamentals and Applications; John Wiley and Sons: New York, 2003; p 24. (32) Abragam, A. The Principles of Nuclear Magnetism; Oxford University Press: Oxford, 1961; pp 272-274. (33) Bloembergen, N. Physica 1949, 15 (3-4), 386. (34) Rorschach, H. E. Physica 1964, 30, 38. (35) Blumberg, W. E. Phys. ReV. 1960, 119 (1), 79. (36) Abragam, A. The Principles of Nuclear Magnetism; Oxford University Press: Oxford, 1961; p 380. (37) Mali, G.; Kaucic, V. Solid State Nucl. Magn. Reson. 1998, 12, 243. (38) Mellot-Draznieks, C.; Buttefey, S.; Boutin, A.; Fuchs, A. H. Chem. Commun. 2001, 2200.

JP801133X

Study of the Adsorption of Organic Molecules on Transition Metal

Recommend Documents