Hydration-dehydration behavior and thermodynamics of chabazite

Geochimica et Cosmochimica Acta, Vol. 69, No. 9, pp. 2293–2308, 2005 Copyright © 2005 Elsevier Ltd Printed in the USA. All rights reserved 0016-7037/05 $30.00 ⫹ .00

doi:10.1016/j.gca.2004.11.007

Hydration-dehydration behavior and thermodynamics of chabazite CLAIRE I. FIALIPS,1,2,* J. WILLIAM CAREY,2 and DAVID L. BISH3 1

University of Newcastle upon Tyne, Drummond Bldg., School of Civil Engineering and Geosciences, Newcastle upon Tyne, NE1 7RU, UK 2 Los Alamos National Laboratory, EES-6, Hydrology, Geochemistry, and Geology, MS D469, Los Alamos, NM 87545, USA 3 Department of Geological Sciences, Indiana University, 1001 E 10th St., Bloomington, IN 47405, USA (Received May 4, 2004; accepted in revised form November 5, 2004)

Abstract—Equilibrium in the chabazite-H2O system was investigated by isothermal thermogravimetric analysis over a large range of temperatures (from 23 to 315°C) and H2O-vapor pressures (from 0.03 to 28 mbar). Thermodynamic analysis of the phase-equilibrium data revealed the existence of three energetically distinct types of H2O, referred to as S-1, S-2, and S-3. At 23°C and 26 mbar of H2O-vapor pressure, chabazite has maximum H2O occupancies of 8.2, 11.1, and 3.1 wt.% for S-1, S-2, and S-3, respectively. During dehydration, S-1 H2O is lost first, followed by S-2 H2O and then S-3 H2O, with significant overlap for S-1 and S-2 as well as S-2 and S-3. The thermodynamics of chabazite-H2O were modeled using three independent equilibrium formulations for S-1, S-2, and S-3. These formulations yielded standard-state molar Gibbs free energy of hydration of ⫺21.8 ⫾ 0.6, ⫺52.1 ⫾ 1.8, and ⫺111.7 ⫾ 6.7 kJ/mol for S-1, S-2, and S-3. Standard-state molar enthalpies of hydration for each type of H2O are ⫺65.6 ⫾ 0.5, ⫺100.1 ⫾ 1.6, and ⫺156.9 ⫾ 6.2 kJ/mol, respectively. Integral molar values for the Gibbs free energy of hydration for each type of H2O are ⫺19.0 ⫾ 0.7, ⫺40.1 ⫾ 2.1, and ⫺76.9 ⫾ 9.6 kJ/mol, respectively. Integral molar values for the enthalpy of hydration for each type of H2O are ⫺62.8 ⫾ 0.6, ⫺88.1 ⫾ 1.9, and ⫺122.2 ⫾ 9.3 kJ/mol, respectively. Integration of the predicted total partial molar enthalpy of hydration for all three types of H2O over the full H2O content of chabazite gave an integral molar enthalpy of ⫺39.65 ⫾ 9.3 kJ/mol relative to liquid water. The thermodynamic data obtained for the hydration of natural chabazite were used to predict the hydration state of chemically similar chabazites under various temperatures and PH2O, ranging from 25 to 400°C and from 10⫺5 to 104 bars. Copyright © 2005 Elsevier Ltd example, clinoptilolite has 5 distinct crystallographic H2O sites but phase equilibria reveals only one distinct energetic type of H2O (Armbruster and Gunter, 1991; Carey and Bish, 1996); while laumontite shows a direct correspondence between its 4 H2O crystallographic sites and its 4 energetic types of H2O (Fridriksson et al., 2003a, 2003b). In general one has to analyze the equilibrium data to determine the number of energetic sites for a given zeolite. Several authors have studied the hydration-dehydration thermodynamics of chabazite (Barrer and Cram, 1971; Valueva and Goryainov, 1992; Drebushchak, 1999, Shim et al., 1999; Ogorodova et al., 2002). In particular, Valueva and Goryainov (1992), using immersion calorimetry and Raman spectroscopy, observed stepwise behavior in the enthalpy of hydration of chabazite vs. H2O content and changes in Raman spectra that suggested the presence of three different types of H2O with different energetics. However, their analysis of energetics was based on experimental data showing considerable scatter. Later, Shim et al. (1999) assessed the enthalpy of formation and enthalpy of hydration of several synthetic cation-exchanged chabazites using high-temperature reaction calorimetry and showed that the experimental data presented by Valueva and Goryainov (1992), like their own calorimetric data, could be analyzed assuming a single energetic type of H2O. However, Shim et al. (1999) do not have available calorimetric data for n ⬍1 mol of H2O per formula unit (pfu) and Valueva and Goryainov (1992) have few data for n ⬍1. In the present study, isothermal thermogravimetric analyses (IsoT TGA) were conducted over a range of temperature and PH2O to better understand the hydration-dehydration behavior

1. INTRODUCTION .

Chabazite, ideally (Ca0.5,Na,K)4Al4Si8O24 12H2O, is one of the most widespread natural zeolites. It occurs in volcanic and metamorphic rocks and can be abundant in sedimentary beds (Bish and Ming, 2001). Zeolites are microporous minerals with large extra-framework sites that incorporate both exchangeable cations and water molecules. Exchangeable cations and water molecules in chabazite are spread over a large number of sites and the location and number of these sites depend on the nature of the exchangeable cations. The Alberti et al. (1982) crystal structure refinement of hydrated Na-, Ca-, Sr-, and K-exchanged chabazite identified up to four cation sites and seven water sites. Zeolite—H2O equilibria affects zeolite stability and efficiency as an adsorbing agent, cation-exchanger, catalyst, or molecular sieve. The response of zeolites to changes in temperature and/or water-vapor pressure (PH2O) is thus of great importance in a large variety of natural and artificial environments. Studies of the hydration-dehydration behavior of zeolites can be used to determine hydration energetics and to predict hydration state as a function of temperature and PH2O. In the present paper we make a distinction between energetic types of H2O identified calorimetrically or through phase equilibria and crystallographic sites for H2O as identified in diffraction studies (Bish and Carey, 2001): in general, the number of distinct types of H2O that can be identified in studies of hydration energetics is not the same as the number of distinct sites of H2O that can be identified crystallographically. For * Author to whom correspondence ([email protected]).

should

be

addressed 2293

2294

C. I. Fialips, J. W. Carey, and D. L. Bish

of natural chabazite, to determine the number of energetically distinct types of H2O, to determine the partial and integral molar enthalpies and Gibbs free energy of hydration, and to provide a basis for predicting the hydration state of chemically similar chabazites at any temperature and PH2O. 2. EXPERIMENTAL METHODS The chabazite sample used in this study is from Wasson’s Bluff, near Parrsboro, in Nova Scotia, Canada. The chemical composition of the chabazite was determined using a Cameca SX50 electron microprobe with accelerating voltage of 15 keV, a beam current of 15 nA, and a beam diameter of 10 ␮m. The chabazite unit-cell formula is Ca1.40Na0.29Sr0.03K0.22Al3.47Si8.55O24 · nH2O, based on an average of 27 microprobe analyses on 1 to 4-mm crystals. This composition differs significantly from the ideal unit-cell formula of fully hydrated chabazite ((Ca0.5,Na,K)4Al4Si8O24 · 12H2O). In particular, the Si/Al ratio of the present sample (2.46) is 23% higher than ideal. Such a high proportion of Si is rather unusual for hydrothermal chabazite but has been reported for other chabazite samples (e.g., Gottardi and Galli, 1985; Shim et al., 1999). Chabazite crystals were handpicked from the natural specimen, ground by hand under acetone in an agate mortar, and passed through a 400 mesh sieve to obtain a ⬍38 ␮m fraction. Powder X-ray diffraction (XRD) data for the chabazite sample were acquired on a Siemens D-500 diffractometer using Cu K␣ radiation (45 kV, 35 mA), incidentand diffracted-beam Soller slits, and a Kevex solid-state Si(Li) detector. The XRD measurement was performed on a front-packed mount of the ⬍38 ␮m fraction by step scanning from 2 to 70° 2⌰ using a step interval of 0.02° 2⌰ and a counting time of 10s per step. The relative humidity (RH) surrounding the sample was maintained at 30%RH for the duration of the XRD measurement. The XRD data showed only the presence of chabazite and no crystalline impurity was observed. Lattice parameters (space group R3៮ m) were determined using the Rietveld refinement program GSAS (Larson and Von Dreele, 1994) with the EXPGUI graphical interface (Toby, 2001). The structure determined by Alberti et al. (1982) was used as the starting model. A three-term Fourier series was used to model the background and the convolution function described by Howard (1982; Larson and Von Dreele, 1994) was used to model the experimental profile. The unit-cell data obtained for Wasson’s Bluff chabazite were a ⫽ 9.390 Å, ␣ ⫽ 94.356°, and V ⫽ 820.37 Å3. They are similar but slightly smaller than those obtained by Alberti et al. (1982) and Calligaris et al. (1982) for other natural chabazites (a ⫽ 9.42–9.43 Å, ␣ ⫽ 94.14 –94.20°, and V ⫽ 829 – 831 Å3). Chabazite-H2O equilibrium measurements were performed using a DuPont 951 thermogravimetric analyzer (TGA) equipped with automated RH control (e.g., Carey and Bish, 1996). The RH and corresponding partial pressure of H2O were controlled by mixing dry N2 gas with water-saturated N2 using two mass-flow controllers. A total mixed-gas flux of 100 cm3/min was maintained during all experiments. The actual RH of the gas mixture was measured before entering the TGA using a calibrated Vaisala 36B humidity probe. The RH and temperature data were converted into PH2O using a fit to the steam tables of Haar et al. (1984). For each experiment, ⬃10 –15 mg of sample was placed on the calibrated balance and first equilibrated at 23°C and 5%RH for 2 h. The isothermal TGA measurements involved holding the temperature constant while varying PH2O from ⬃1.3 mbar up to ⬃26 mbar (hydration) and from ⬃26 mbar down to ⬃0.05 mbar (dehydration), in a sequence of isobaric steps of 5.0%RH. For each isobaric stage, the sample was held at constant PH2O for 60 –120 min to achieve complete equilibrium. Isothermal TGA data were collected from 23 to 316°C. Conventional isobaric TGA measurements were also conducted to observe the dehydration process and to measure the total H2O content of the sample at 5.0%RH and 23°C (X5%RH, in grams of H2O per gram of dry chabazite, g-w/g-c) and the maximum water content of the sample at 100%RH and 23°C (Xmax, in g-w/g-c). The sample was first equilibrated at 23°C and the desired RH for 2 h. Weight losses were measured while heating to 900°C at 10°C/min, under constant PH2O. The dry mass was measured after complete dehydration of the sample, allowing calculation of X5%RH and Xmax. The value of Xmax was also

measured by isobaric TGA (100%RH and 23°C) for samples pretreated at various temperatures from 25 to 315°C to investigate the possibility of irreversible loss of sorption capacity following treatment at high temperatures. Errors associated with the isothermal and isobaric TGA experimental data (one standard deviation, 1␴) are in the order of ⬃0.001– 0.005 mg for the measured mass, 0.5–1.5% of the measured PH2O between 95 and 10%RH, 1.5%–20% of the measured PH2O under 5%RH, and ⬃1°C for the temperature. 3. THERMODYNAMIC METHODS The sorption of water-vapor by chabazite (Chab) and the corresponding equilibrium constant (K) are defined by the following relations: H2Ovap ⇔ H2Ochab

(1)

冉 冊

(2)

ln K ⫽ ln

aHChab 2O f Hvap2O

Chab Vap in which aH is the activity of H2O in chabazite and fH is the 2O 2O fugacity of H2O vapor. Assuming ideal Langmuir sorption, the activity of H2O can be expressed as

aHChab ⫽ 2O·Lang

冉 冊冉 ␪

⫽

(1 ⫺ ␪)

x (Xmax ⫺ x)

冊

,

(3)

in which ␪ is the fractional water content (␪ ⫽ x/Xmax), x is the actual water content of chabazite in g-w/g-c (x ⬍Xmax), and Xmax is the maximum water content of chabazite (0.2882 ⫾ 0.0009 g-w/g-c; see below). The relationship of lnK to H2O content is known to provide a sensitive method of examining experimental isothermal TGA data (Carey and Bish, 1996). For ideal Langmuir sorption, lnKLang is defined by the following relation: ln KLang ⫽ ln

冉

x (Xmax ⫺ x)P(H2O)

冊

(4)

At chabazite-H2O equilibrium, the Gibbs free energy of the hydration reaction (⌬␮hydr; Reaction 1) is zero and the thermodynamics of chabazite hydration as a function of temperature and PH2O can be described by the following relation (Carey and Bish, 1996): ⌬␮hydr T

⫽

0 ⌬␮hydr

T0

0 ៮ hydr ⫹ ⌬H

冉 冊 1

T

⫺

1

T0

⫺

兰 T 兰 ⌬C៮ p . dT . dT

⫹

1

2

兰

V៮ H0,chab 2O T

dP ⫹ R ln K ⫽ 0

(5)

0 ៮ hydr in which ⌬H is the standard-state partial molar enthalpy of hydration, ⌬C៮ p is the standard-state heat capacity of the hydration reaction, 0,chab V៮ H is the standard-state partial molar volume of H2O in chabazite, 2O R is the gas constant, and T0 is the reference-state temperature (298K). To address the non-ideality of H2O sorption onto chabazite, we formulated the H2O activity using temperature-independent polynomial factors of expansion of the H2O content, W1, W2, etc. (Carey and Bish, 1996):

aHChab ⫽ 2O

冉

x (Xmax ⫺ x)

冊

exp共兺 Wi␪i兲

(6)

0,chab If we neglect the possible contribution of V៮ H upon dehydration or 2O hydration (the partial molar volume of H2O in zeolites is generally small and has a negligible energetic effect at shallow crustal pressures; e.g., Bish, 1984; Carey and Bish, 1996; Bish and Carey, 2001) and if we assume that ⌬C៮ p is a constant and equals 3R (e.g., Johnson et al., 1991; Carey, 1993; Carey and Bish, 1996), where R is the gas constant, Eqn. 5 becomes:

Hydration-dehydration behavior of chabazite ⌬␮hydr T

⫽

0 ⌬␮hydr

T0

0 ៮ hydr ⫹ ⌬H

冉 冊 冋 冉 冊 冉 冊册 冉 冊兺 1

T

⫺

1

T0

⫹ R ln

T

⫺ 3R ln

T0

x

(Xmax ⫺ x)P

⫹

⫹

T0 T

Wi T

2295

⫺1

␪i ⫽ 0

(7)

Eqn. 7 can be rearranged as

冋 冉 冊 冉 冊册

ln (KLang) ⫺ 3 ln

T

T0

⫹

T0 T

⫺1

⫽A⫹

C D ⫹ ␪ ⫹ ␪2 ⫹ . . . , T T T

B

(8) where A, B, C, D, etc. are temperature-independent and ␪-independent parameters. It should thus be possible to fit the experimental data of lnKLang vs. ␪ by multiple linear regression using Eqn. 8. For i ⫽ 2, the standard-state partial molar thermodynamic values for chabazite hydration (Eqn. 7) can be calculated from the fit parameters (A, B, C, and D, Eqn. 8) using the following relations (Carey and Bish, 1996): 0 ⌬␮hydr ⫽ ⫺R(T0A ⫹ B)

(9)

៮ 0 ⫽ ⫺RB ⌬H hydr

(10)

៮0 ⫽ RA ⌬S hydr

(11)

W1 ⫽ ⫺RC

(12)

W2 ⫽ ⫺RD.

(13)

For i ⫽ 2, the integral molar thermodynamic values as a function of water content are given by the following relations (Carey and Bish, 1996):

冉 冊 冉 冊 冉冊

˜ ⫽ ⌬H ៮ 0 ⫹ W1 ␪ ⫹ W2 ␪2 ⫹ 3R(T ⫺ T ) ⌬H hydr 0 hydr 2 3 ˜ ⫽ ⌬S៮ 0 ⫹ 3R ln ⌬S hydr hydr

T

T0

(14)

R ⫺ [␪ ln ␪ ⫹ (1 ⫺ ␪) ln (1 ⫺ ␪) ␪ ⫺ ␪ ln P]

˜ ⫽ ⌬H ˜ ⫺ T⌬S ˜ . ⌬G hydr hydr hydr

Fig. 1. Isobaric thermogravimetric curves for chabazite obtained at 5 and 100%RH (1.34 and 26.28 mbar PH2O respectively) in terms of x (gram of H2O per gram of dry chabazite, g-w/g-c) as a function of temperature and corresponding derivatives dx/dT (open symbols).

(15) (16)

4. RESULTS

The isobaric TGA data and their first derivative suggest three stages of dehydration in chabazite (Fig. 1). Both the 100%RH and the 5%RH curves show a distinct inflection at ⬃245°C that separates high and moderate temperature phases of dehydration. Both curves also show a less distinct inflection at lower temperature (80 to 100°C) that separates low and moderate phases of dehydration. A theoretical analysis of TGA dehydration curves made by Bish and Carey (2001) showed that inflections in TGA curves indicated the presence of distinct energetic types of H2O that could most easily be modeled by combining separate thermodynamic formulations for each energetic type of H2O. The TGA data for chabazite indicate the presence of at least two energetically different H2O types (one clearly resolved inflection at T ⬃245°C) and probably a third one (slight change in TGA curves inflexion at T ⬃90°C). The maximum H2O content of the sample at 23°C and 100%RH (Xmax) was 0.2882 ⫾ 0.0009 g-w/g-c, corresponding to 12.67 mol of H2O pfu. This high H2O content is consistent with the large proportion of calcium in the sample (Alberti et al., 1982). The H2O content of the sample at 23°C and 5%RH (X5%RH) was 0.2527 ⫾ 0.0010 g-w/g-c, corresponding to 11.11

H2O pfu. As all the isothermal TGA experiments were started at 23°C and after equilibrium at 5%RH, the value for X5%RH was used to normalize all the isothermal TGA data. Isothermal TGA data were collected at 23.0, 36.2, 50.0, 69.2, 83.8, 98.4, 123.4, 148.3, 173.7, 198.7, 248.3, 272.8, 296.7, and 315.7°C, for PH2O between 0.03 and 28 mbar during both hydration and dehydration (Fig. 2; The raw experimental data are available in Appendix I as an electronic annex). No hysteresis was observed during the experiments, indicating that chabazite hydration was reversible and that equilibrium was reached for each isobaric stage during both hydration and dehydration. Unlike the results for clinoptilolite obtained by Carey and Bish (1996), no condensation effects were observed at 23°C as saturation water-vapor pressure was approached. By extrapolating the chabazite water content (x in g-w/g-c) vs. PH2O data at 23°C to the saturation water-vapor pressure (26.28 mbar), we obtained an Xmax value equal to 0.2878 g-w/g-c that is very close to and within errors for the Xmax value measured directly at 23°C and 100%RH. The isothermal TGA data were used to calculate lnKLang vs. ␪ isotherms (Fig. 3).The isothermal lnKLang values are not independent of ␪ , which indicates non-ideal sorption behavior (Eqn. 7). The shape of the isotherms clearly indicates that lnKLang is a function of at least ␪2. In contrast to clinoptilolite (Carey and Bish, 1996), the trend in the lnKLang isotherms for chabazite is not uniform over the entire range of temperature and ␪ values. However, we can distinguish three individual regions within which the data display similar and parallel trends as temperature increases and ␪ decreases. These regions are enclosed by dashed lines in Figure

2296

C. I. Fialips, J. W. Carey, and D. L. Bish

and thus calculated KLang values are very sensitive to the value of Xmax (cf. Eqn. 4). Choosing a slightly larger value for Xmax would remove this problem but we follow convention and assign Xmax to the water content at 23°C and 100%RH. The data obtained for ␪ ⬎0.96 and 23°C were not used for the analysis. The value of Xmax is also a sensitive measure of irreversible modifications to the chabazite structure. Values of Xmax were measured by isobaric TGA at 100%RH and 23°C following pretreatment of chabazite samples at temperatures from 25 to 315°C (Fig. 4). These results indicate that chabazite loses sorption capacity at temperatures ⬎175°C and that loss increases with the temperature of pretreatment to 4% at 315°C.

5. DATA ANALYSIS

Fig. 2. Isothermal thermogravimetric curves of hydration and dehydration in g-w/g-c (x) as a function of the water-vapor pressure (PH2O). Closed symbols represent data collected during hydration and open symbols represent data collected during dehydration.

3 and represent a low-temperature/high-␪ region (23 to 98°C, area I), an intermediate temperature region (69 to 273°C, area II), and a high-temperature/low-␪ region (199 to 316°C, area III). At high ␪ values (␪ ⬎0.96) and 23°C, lnKLang rises sharply (Fig. 3). This behavior occurs because (Xmax-x) is nearing 0,

For Na-, Ca-, and K-exchanged clinoptilolite, Carey and Bish (1996) were able to fit the experimental data over the entire measured range of lnKLang and ␪ values using Eqn. 7 and 8 with a second-order polynomial (i.e., i ⫽ 2), and they obtained r2 values greater than 0.91 and standard errors for lnKLang smaller than 0.45 (1␴). In contrast, regardless of the degree of the polynomial (i), any attempt to fit all chabazite experimental data gave systematic differences between the calculated and experimental data (e.g., Fig. 5; r2 ⫽ 0.92, 1␴ error for lnKLang ⫽ 0.61). This shows that chabazite hydration behavior is quite different than that of clinoptilolite. We hypothesize, based on the complex shape of the lnKLang vs. ␪ curves for chabazite (Fig. 3) and the apparent presence of three homogeneously behaving data regions (Fig. 3, dashed regions), that chabazite has three energetically distinct types of H2O, referred to as S-1, S-2, and S-3. During dehydration of chabazite, the three types of H2O would be progressively dehydrated

Fig. 3. Equilibrium relations for chabazite-H2O plotted in terms of the Langmuir equilibrium constant, lnKLang, vs. the fractional water content, ␪. Dashed lines delimit three data areas, I, II, and III, which are discussed in the text.

Hydration-dehydration behavior of chabazite

ln

冉

xj (X j max ⫺ x j)P

2297

冊 冋 冉 冊 冉 冊册 ⫺ 3 ln

⫽ AS⫺j ⫹

BS⫺j T

T

T0

⫹

T0

⫹

CS⫺j T

T

␪j ⫹

⫺1

DS⫺j T

␪2j ⫹ . . . .

where j is the H2O index number (1, 2, or 3), xj is the S-j water content in g-w/g-c, Xjmax is the maximum amount of S-j water in g-w/g-c, and ␪j is the fractional content of S-j water (␪j ⫽ xj/Xj max). The thermodynamic model in Eqn. 17 and 18 represents hydration/dehydration equilibria in chabazite as the sum of three independently acting types of H2O (S-1, S-2, and S-3). Eqn. 17 and 18 introduce six additional unknowns into the thermodynamic description of chabazite hydration (X1max, X2max, X3max, x1, x2, and x3), although only four are independent 3 3 because of the relations Xmax ⫽ j⫽1 X j max and x ⫽ j⫽1 x j. Unfortunately, direct observations of Xjmax and xj values are not possible, which significantly complicates the thermodynamic analysis. Thus we found it was not possible to analyze these data using the straightforward linear regression techniques of Carey and Bish (1996). One possible approach is to use implicit nonlinear regression analysis to simultaneously estimate both the unobservable independent variables Xjmax and xj and the fit parameters (AS-j, BS-j, CS-j. . . of Eqn. 18; e.g., Sachs, 1976). However, we found that a simpler approach could be used by constraining the possible values of Xjmax and xj (Table 1). Regions I, II, and III of Figure 3 reflect P-T conditions in which the hydration/ dehydration reactions are dominated by a single type of H2O. This suggests that it may be possible to extract reasonable estimates of thermodynamic parameters by sequentially analyzing the energetics of each type of H2O corresponding to regions I, II, and III.

兺

Fig. 4. Change in the chabazite H2O sorption capacity (Xmax in g-w/g-c) measured at 25°C and 100%RH as a function of pretreatment temperature.

following the sequence S-1 ¡ S-1 ⫹ S-2 ¡ S-2 ¡ S-2 ⫹ S-3 ¡ S-3 (Fig. 3). For a three-type model, Eqn. 7 and 8 become: 0 ⌬␮hydr,S⫺j

T0

冉 冊 冋 冉 冊 冉 冊册 冉 冊兺

0 ៮ hydr,S⫺j ⫹ ⌬H

⫹ R ln

1

T

⫺

1

T0

⫺ 3R ln

xj

(X j max ⫺ x j)P

⫹

T

T0

⫹

n

Wi,S⫺j

i⫽1

T

T0 T

⫺1

␪ij ⫽ 0

(17)

(18)

Fig. 5. Comparison of the experimental data for chabazite (symbols) with a calculated fit (solid lines) obtained by multiple linear regression of all the data and using a second-order polynomial to characterize the non-ideality of H2O in chabazite (Eqn. 8).

兺

2298

C. I. Fialips, J. W. Carey, and D. L. Bish

6a). The fitted parameters (AS-1, BS-1, CS-1, DS-1; Table 3) were used to determine the corresponding standard-state partial molar 0 0 ៮ hydr,S⫺1 thermodynamic values (⌬␮hydr,S⫺1 , ⌬H , W1,S-1, and W2,S-1; Table 4) using Eqn. 9 to 13. The calculated thermodynamic properties for S-1 provide a complete description of the hydration/ dehydration behavior of type S-1 H2O, allow prediction of the water occupancy of S-1 (x1) at any T and PH2O, and are independent of the behavior of S-2 and S-3. We continued the second stage of analysis by fitting the experimental data within area II (S-2) plus experimental data between area I and II assuming 100% occupancy for S-3, but instead of assuming 0% occupancy for S-1 (as in the first stage analysis), values of x1 were predicted using the previously calculated thermodynamic properties of S-1. The nonlinear regression resulted in an r2 ⫽ 0.99 and closely reproduced the data within region II (Fig. 6b). The standard-state partial molar thermodynamic values for S-2 and the predicted water occupancy of S-2 (x2) were determined using the same method as for S-1 (Table 4). Finally, we fit the experimental data within area III (hydration site S-3) plus the experimental data between area II and III using predicted x1 and x2 values instead of assuming 0% occupancy for S-1 and S-2. The nonlinear regression yielded an r2 ⫽ 0.91 and provided an adequate fit of most of the data in region III (Fig. 6c) but does not reproduce the lowest temperature isotherm (199°C). This is probably due to the difficulty in resolving the behavior of S-2 and S-3 in this temperature range (see below). The standard-state partial molar thermodynamic values and the predicted water occupancy of S-3 (x3) were determined using the same methods as for S-1 and S-2 (Table 4). Integral molar thermodynamic values for each type of H2O of chabazite were obtained using Eqn. 14, 15, and 16 with ␪ ⫽ 1 and are summarized in Table 5. The predicted amounts of the three types of H2O for all of the experiments as a function of the observed total water content shows that as chabazite dehydrates H2O is lost progressively from S-1 to S-2 to S-3 (Fig. 7a). For x values greater than ⬃0.23 g-w/g-c, S-2 and S-3 have a 100% occupancy and changes in the H2O content of chabazite are solely due to S-1. For x values between ⬃0.1 and ⬃0.24 g-w-g-c, S-3 has a 100% occupancy and changes in the H2O content of chabazite are dominated by S-2 but also include significant contributions from S-1. For x values ⬍⬃0.07 g-c/ g-w, S-1 has a 0% occupancy and changes in the H2O content

Table 1. Predicted amount of H2O for S-1 (x1), S-2 (x2), and S-3 (x3) in chabazite for the areas (I, II, and III) shown in Figure 3, where x is the total observed H2O content and Xjmax is the maximum amount of H2O associated with each type of H2O.

x1 x2 x3

Area I

Area II

Area III

x-X2max-X3max X2max X3max

0 x-X3max X3max

0 0 x

A two-stage procedure was used to calculate the variables and parameters of Eqn. 18. In the first stage, nonlinear regression (using the SOLVER function of Microsoft EXCEL) was used to estimate values of Xjmax. Data in each of the regions, I, II, and III, was sequentially regressed using Eqn. 17 with the assumption that the variations in H2O content in the region were solely due to one type of H2O (S-1, S-2, or S-3) and that in each region the non-varying H2O-types were present in the structure at either 0 or 100% (Table 1). This simplification 0 allowed us to obtain preliminary estimates of ⌬␮hydr,S⫺j , 0 ៮ ⌬Hhydr,S⫺j, Wi,S⫺j, Wi⫺l,S⫺j and final values of Xjmax (Table 2). In this stage, we also investigated the required degree of polynomial (Eqn. 6) and found that i ⫽ 2 gave reasonable estimates with r2 ⬎0.94 and that increasing i did not improve r2 values and gave poor F statistics (Table 2; inclusion of an additional term is statistically nonsignificant at ⬎45% level). The first stage of nonlinear regression yielded values of X1max, X2max, and X3max equal to 0.1050, 0.1429, and 0.0403 g-w/g-c, respectively. Values of Xjmax obtained in the first stage were used in the second stage of nonlinear regression in which we relaxed the assumptions that the variations in H2O content in each region were due solely to one type of H2O. In outline, the second stage of regression was done by progressively applying Eqn. 18 to regions I, II, and III and applying the results from each region to the analysis of the subsequent region. All analyses were done with polynomials of degree 2 (i ⫽ 2) as determined in the first stage analysis. The second stage analysis began with fitting the experimental data within area I (S-1) using the Xjmax previously determined and assuming that S-2 and S-3 were 100% occupied and refining only the parameters relevant to S-1. The nonlinear regression resulted in r2 ⬎0.99 and closely reproduced the data within region I (Fig.

Table 2. Nonlinear regression estimates of the unknown parameters of Equation 17 (using Microsoft EXCEL Solver tool) for the hydration sites S-3, S-2, and S-1 of chabazite and for different degrees (i) of the non-ideality polynomial (Equation 6). The F statistic represents the calculated probability that the i ⫽ 3 model is statistically significant compared to the i ⫽ 2 model.

i Xjmax ⌬␮hydr0 (J/mol) ៮ hydr0 (J/mol) ⌬H W1 W2 W3 r2 NEDP F

S-3

S-2

S-1

S-3

S-2

S-1

2 0.0403 ⫺115748.49 ⫺169088.85 106905.89 ⫺70965.24

2 0.1429 ⫺50906.47 ⫺99115.41 34156.51 ⫺19584.21

2 0.1050 ⫺24972.33 ⫺71137.74 13143.05 ⫺4288.60

0.9439 35

0.9864 118

0.9716 118

3 0.0403 ⫺115868.06 ⫺169215.59 109051.90 ⫺76900.80 4218.99 0.9364 35 29.1%

3 0.1430 ⫺50357.61 ⫺98390.27 30060.52 ⫺9556.24 ⫺7067.77 0.9848 118 53.0%

3 0.1050 ⫺21624.38 ⫺71200.95 ⫺11331.10 42214.58 ⫺83.38 0.8995 118 27.9%

Note: NEDP ⫽ number of experimental data point used for the estimation.

Hydration-dehydration behavior of chabazite

2299

Fig. 6. Comparison of the experimental data for chabazite (symbols) with a calculated fit (solid lines) for each of three data areas using a second-order polynomial to characterize the non-ideality of H2O in chabazite. lnKj ⫽ ln[xj/((Xjmax-xj)P)], ␪j ⫽ xj/Xjmax, with j ⫽ index number of the H2O type ⫽ S-1, S-2, or S-3.

of chabazite are controlled by both S-3 and S-2. The result shows that there are ranges of x in which multiple types of H2O contribute to sorption behavior and illustrates why we had to use a two-stage procedure to obtain regression coefficients. One measure of the adequacy of the thermodynamic model is the difference between the sum of the amount of H2O for S-1, S-2, and S-3 and the observed total H2O content (x). The differences are small with an average absolute deviation of 0.0034 g-w/g-c and a range of ⌬x from ⫺0.0174 to ⫹ 0.0135

g-w/g-c (Fig. 7a) and show that the model provides good predictions of the total H2O content of chabazite. The predicted H2O distributions show deviations from homogeneous curves (Fig. 7a). This is expected because the calculated distributions are a function of temperature (and water-vapor pressure) and the experimental conditions used to generate Figure 7a range from 23 to 315°C. Predicted H2O distributions under isothermal conditions at 25, 150, and 315°C were obtained by systematically varying PH2O

2300

C. I. Fialips, J. W. Carey, and D. L. Bish

Table 3. Summary of linear regression results for each type of H2O in chabazite (S-1, S-2, and S-3) using the model given by Equation 18 and the Xjmax values from Table 2

A (no unit) B (K) C (K) D (K) r2 lnKj Std. error (1␴) NEDP

S-1

S-2

S-3

⫺17.7 (0.1) 7894 (55) ⫺219 (65) ⫺707 (60) 0.9971 0.14 105

⫺19.4 (0.3) 12039 (193) ⫺4994 (204) 3179 (163) 0.9855 0.20 135

⫺18.3 (1.0) 18875 (744) ⫺14597 (1468) 9350 (1152) 0.9127 0.35 71

Note: Numbers in parentheses represent standard errors (1␴) derived from the multiple linear regressions analyses. NEDP ⫽ number of experimental data points used for the regression.

until the entire range of H2O content was obtained (Fig. 7b). The significant differences in distribution as a function of temperature reflect the different responses of S-1, S-2, and S-3 to temperature as embodied in their calculated thermodynamic properties. On the other hand, the predicted H2O distributions are not a particularly strong function of watervapor pressure as shown in Figure 7c, which was calculated at 27 and 0.027 mbar H2O and by varying temperature until the entire range of H2O content was obtained. The isobar calculation at 27 mbar (Fig. 7c) closely reproduces the experimental data (Fig. 7a) because the experimental data were collected over a relatively wide temperature and narrow water-vapor pressure range. The fidelity of the thermodynamic model (Eqn. 17; Table 4) was also investigated by examination of fit to the lnKLang vs. ␪ data (Fig. 8). This comparison was made by summing the results for S-1, S-2, and S-3 at each experimental T and PH2O to calculate a value of x in Eqn. 4. One quantitative measure of the quality of the fit is the average absolute deviation of lnKLang and ␪ (␦lnKLang and ␦␪) for the entire PH2O-T data range. The ␦lnKLang and ␦␪ values are small (0.0615 and 0.0068, respectively), indicating a reasonable fit. The model provides accurate predictions except at low PH2O (less than 0.07 mbar; cf. symbols at the highest lnKLang along an isotherm in Fig. 8) and at 84, 98, and 199°C for PH2O ⬍3 mbar. The deviations at low PH2O could be explained in part by the relatively large errors associated with measurements of the low PH2O of the applied atmosphere. However, the deviations may also reflect interchange of H2O among S-1, S-2, and S-3 that cannot be represented by our thermodynamic model. A more direct method of evaluating the model is to

compare measured and predicted H2O contents as a function of PH2O and T (Fig. 9). These results demonstrate that the thermodynamic model is capable of reproducing observed H2O contents quite accurately over the entire range of experimental conditions (r2 ⫽ 0.9942). There are small, systematic deviations evident in the 98 and 123°C isotherms that are likely to represent the previously discussed interchange of H2O among S-1 and S-2. 6. DISCUSSION

The experimental data obtained for chabazite cannot be fit without assuming the presence of multiple, energetically distinct types of H2O. The hydration behavior and thermodynamics of chabazite are thus very different and far more complex than those of other zeolites, such as clinoptilolite, whose hydration behavior was adequately modeled assuming one energetic type of H2O (Carey and Bish, 1996). Analysis of chabazite hydration/dehydration is also more complex than the case of laumontite (Fridriksson et al., 2003a, 2003b). Laumontite has four energetically distinct types of H2O, but these are readily assigned to four H2O sites in the crystal structure. Moreover, each of the sites is clearly resolved in isobaric TGA analyses facilitating independent thermodynamic analysis (Fridriksson et al., 2003b). The thermodynamic model for chabazite (Eqn. 17; Table 5) allows calculation of the partial molar enthalpy of hydration for each type of H2O at 25°C using the following relation (modified from Carey and Bish, 1996): 2 ៮ ៮0 ⌬H hydr,S⫺j ⫽ ⌬H hydr,S⫺j ⫹ W 1,S⫺j␪ j ⫹ W 2,S⫺j␪ j .

The total partial molar enthalpy of hydration of chabazite can be obtained by summing each of the contributions in Eqn. 19 as follows: ៮ ⫽ ⌬H hydr

⭸x j

3

兺 ⌬H៮

hydr,S⫺j

j⫽1

⭸x

.

S-1 ⌬␮ (J/mol) ៮ ⌬H (J/mol) ⌬S៮ (J/mol.K) W1 (J/mol) W2 (J/mol)

⫺21835 ⫺65629 ⫺146.96 1819 5876

S-2 (580) (457) (1.20) (540) (495)

(20)

៮ hydr,S-j, is readily calcuAlthough the first term in Eqn. 20, ⌬H lated from Eqn. 19, there is no simple method of calculating ⭸xj/⭸x because the xj are only defined implicitly by Eqn. 17. However, there is a clear relation between xj and x as illustrated in Figure 7. Since ⭸xj/⭸x depends on the reaction path (Fig. 7), calculation of the total partial molar enthalpy, requires choosing a specific reaction path. Because our experimental conditions are most closely reproduced by the isobaric calculation at 27 mbar of Figure 7c, we have used this isobaric dehydration

Table 4. Standard-state partial molar values obtained for the thermodynamics of each type of H2O in chabazite based on the results in Table 3

0 hydr 0 hydr 0 hydr

(19)

⫺52070 ⫺100095 ⫺161.16 41523 ⫺26430

(1824) (1608) (2.89) (1699) (1355)

S-3 ⫺111688 ⫺156930 ⫺151.82 121361 ⫺77732

(6657) (6186) (8.26) (12208) (9577)

Note: numbers in parentheses represent uncertainties propagated from the standard errors (1␴) of the regression coefficients.

Hydration-dehydration behavior of chabazite

2301

Table 5. Integral molar values obtained for the thermodynamics of each type of H2O in chabazite based on Table 4 values.

˜ hydr (J/mol) ⌬G ˜ hydr (J/mol) ⌬H ⌬S˜hydr (J/mol.K)

S-1

S-2

S-3

⫺18967 (661) ⫺62761 (555) ⫺146.96 (1.20)

⫺40119 (2063) ⫺88144 (1874) ⫺161.16 (2.89)

⫺76918 (9580) ⫺122160 (9258) ⫺151.82 (8.26)

Note: numbers in parentheses represent the uncertainties propagated from the standard errors (1␴) of the regression coefficients.

path to calculate the total partial molar enthalpy using Eqn. 20. Our method requires the numerical calculation of ⭸xj/⭸x by determining xj along our chosen reaction path at sufficiently small intervals that the differential relation can be adequately ៮ hydr show a decrease approximated. Our calculated values of ⌬H in partial molar enthalpy as chabazite dehydrates and clearly shows the energetic consequences of three types of H2O (Fig. 10). The nearly step-like features in the calculated partial molar enthalpy reflect transitions from regions dominated by one type ៮ hydr calculated along other of H2O to the next. The values of ⌬H P-T paths (Fig. 7) are similar to Figure 10. ˜ hydr, can be obtained by The integral molar enthalpy, ⌬H ៮ integrating ⌬Hhydr from 0 to 12.67 mol of H2O pfu. Again, because Eqn. 20 cannot be obtained in an analytical form the integration must be performed numerically. We used Mathematica to integrate a splined representation of the partial molar enthalpy data of Figure 10 to obtain an integral value of ⫺39.65 ⫾ 2.41 kJ/mol-H2O. Our results for enthalpy may be compared with several other studies of Ca-chabazite hydration energetics. Valueva and Goryainov (1992) studied the hydration/dehydration behavior of a natural Ca-rich chabazite by immersion calorimetry and Raman spectroscopy. They reported an enthalpy of immersion, Himm, vs. H2O content, n, which, similar to our results, indicates a nonlinear dependence of Himm on n and suggests the presence of three distinct types of H2O with different energetics. However, the data presented by Valueva and Goryainov (1992) are integral enthalpies of hydration and show considerable scatter, hindering a conclusive analysis (also see Shim et al., 1999). Shim et al. (1999) fit the Himm vs. n data of Valueva and Goryainov (1992) to a single quadratic polynomial from n ⫽ 0 to 7 and reported that the ៮ hydr (in corresponding partial molar enthalpy of hydration, ⌬H kJ/mol relative to liquid water) is: ៮ ⫽ ⫺40.7 ⫹ 1.408n ⌬H hydr

(21)

˜ hydr, of which gives an average enthalpy of hydration, ⌬H ⫺31.7 kJ/mol-H2O when integrated from n ⫽ 0 to 12.8. This value is ⬃8 kJ/mol less energetic than our model predicts. Shim et al. (1999) studied the hydration energetics of several synthetic cation-exchanged chabazites by transposed-temperature drop calorimetry (TTD) to 700°C. They did not resolve distinct energetic types of H2O. Their analysis of the TTD data obtained for Ca-chabazite using a quadratic fit gives: ៮ ⫽ ⫺52.97 ⫹ 2.94n , ⌬H hydr

(22)

˜ hydr ⫽ ⫺34.6 ⫾ 1.2 kJ/mol-H2O, which is ⬃5 yielding ⌬H ˜ hydr kJ/mol less energetic than our results. Our value of ⌬H (⫺39.65 kJ/mol-H2O) is also more energetic than those ob-

tained by Ogorodova et al. (2002; ⫺31 ⫾ 4 kJ/mol-H2O) using transposed temperature drop calorimetry and Barrer and Cram (1971; ⫺27.3 kJ/mol-H2O) using immersion calorimetry. Some insight into possible reasons for the difference in energetics we have obtained by phase equilibria and those obtained by calorimetric techniques may be found by considering results on ៮ hydr calculations partial molar enthalpies of hydration. Our ⌬H agree well with those of Shim et al. (1999) for n ⱖ2 mol of H2O (Fig. 10). However, they do not agree for n ⬍2 where we have ៮ hydr from n obtained much more energetic H2O. Integration of ⌬H ⫽ 0 to 2 shows that ⬃4 kJ/mol-H2O of the difference between our results can be attributed to this low H2O content region. Our partial molar enthalpy results can also be integrated and compared with the enthalpy of immersion data of Barrer and Cram (1971) and Valueva and Goryainov (1992; Fig. 11). For n ⬎4 mol of H2O pfu our results are in good agreement with the immersion calorimetry but again our results are significantly more negative at low H2O contents. These observations show that the differences between our phase-equilibria measurements and the calorimetric data are dominated by differing estimates of the energetics at low H2O contents. All of our hydration-dehydration data are reversible (Fig. 2), indicating equilibrium that should yield accurate energetics. However, our fit of the S-3 type H2O (lnK3 vs. ␪3; Fig. 6) is not particularly good (r2 ⫽ 0.91, Table 3) and this type of H2O dominates the calculations of energetics at low values of n. This is reflected in the larger errors at n ⬍2, which correspond to the hydration of S-3 (Fig. 10). Thus, our calculated energetics at low n have considerable uncertainty. It is also important to note that our results are modeldependent. The assumption that the heat capacity of reaction, ⌬C៮ p, is a constant and equals 3R (Eqn. 5 and 7) may not be negligible. To test this effect, calculations were also done assuming that ⌬C៮ p equals 0 instead of 3R (Table 6, Figs. 10 and 11). The results indicate that the estimated maximum amounts of water in each site, Xjmax, are not significantly affected by the chosen ⌬C៮ p value (Table 2 and Table 6; less than 0.2% difference in Xjmax values). The quality of the fits of the experimental data are not improved by assuming ⌬C៮ p ⫽ 0 instead of ⌬C៮ p ⫽ 3R (r2 and standard errors of lnKj in Table 6 0 did not improve compared to results in Table 3) and the ⌬␮hydr ៮ values are similar within errors to those obtained for ⌬Cp ⫽ 3R 0 0 ៮ hydr (Table 4). However, the standard state ⌬H and ⌬S៮ hydr values are significantly less negative for ⌬C៮ p ⫽ 0 than for ⌬C៮ p ⫽ 3R. Though the predicted partial molar enthalpy of hydra៮ hydr (Fig. 10) and enthalpy of immersion, Himm (Fig. tion, ⌬H 11) are less negative for ⌬C៮ p ⫽ 0 than for ⌬C៮ p ⫽ 3R, they remain significantly more negative than the results of Shim et al. (1999), Valueva and Goryainov (1992), and Barrer and

2302

C. I. Fialips, J. W. Carey, and D. L. Bish

Fig. 7. The symbols represent the amount of S-1, S-2, and S-3 H2O as a function of the total H2O content in chabazite either measured (a) or calculated (b and c). The amount of H2O was calculated based on Eqn. 17 and Table 4 at each experimental T and PH2O (a), calculated along isotherms at 25, 150, and 315°C with PH2O varying from 10⫺21 to 10⫺1, 10⫺13 to 103, and 10⫺9 to 105 bar, respectively (b), and calculated along isobars at 27 and 0.027 mbar with T varying from 25 to 500 and ⫺120 to 350°C, respectively. The upper figure (a) also shows the difference between the observed total water content and the sum of the predicted S-1, S-2, and S-3 H2O content at each experimental T and PH2O.

Hydration-dehydration behavior of chabazite

Fig. 8. Comparison of the experimental data for chabazite (symbols) with the total calculated fit for lnKLang vs. ␪ based on the summation of S-1, S-2, and S-3. lnKLang ⫽ ln[x/((Xmax-x)P)] and ␪ ⫽ x/Xmax.

Fig. 9. Comparison of the experimental data for chabazite (symbols) with the calculated fit for total H2O content, ␪, as a function of temperature.

2303

2304

C. I. Fialips, J. W. Carey, and D. L. Bish

Fig. 10. Partial molar enthalpy of hydration of Ca-chabazite relative to liquid water as a function of H2O content as calculated using Eqn. 20 (see text); error bars are propagated errors using one standard deviations (1␴). The data of Shim et al. (1999) and Shim et al.’s quadratic fit to the data of Valueva and Goryainov (1992) are also shown. Data obtained assuming that ⌬C៮ p equals 0 instead of 3R are also presented for comparison.

Cram (1971) for n ⬍2, indicating that the difference between our data and the literature does not result from a ⌬C៮ p assumption. Another potential complication in our study (as well as the calorimetric studies) is the irreversible loss of H2O sorption capacity at T ⬎175°C (Fig. 4). Our data show that chabazite loses up to 4% of its H2O sorption capacity at 315°C. The loss of sorption capacity is consistent with the crystal structural studies of Woodcock and Lightfoot (1999) and Bish and Carey (2001) who observed significant contraction of several chabazite samples at temperatures from 25 to 700°C and 25 to 300°C, respectively. Woodcock and Lightfoot (1999) did not investigate the irreversibility of this behavior but Bish and Carey (2001) found that while one sample showed reversible structural behavior the other actually returned to room temperature with a 0.5% increase in unit-cell volume. The progressive loss of H2O sorption capacity of our chabazite upon heating above 175°C (Fig. 4) could reflect irreversible structural changes. These irreversible changes affect the calculation of the fractional amount of H2O (␪j), because it changes the effective Xmax, although we still observed reversible sorption. Though the resulting affect on our lnKLang data above 175°C cannot be quantified, we suspect that it is small and within experimental errors. All of the calorimetric studies require heat-treatment of the chabazite samples to generate partial or complete dehydration. Shim et al. (1999) heat-treated their hydrated samples from 300 to 700°C for 1 h to obtain various H2O contents (n ⫽ 1.04 to 11.68) and indicated that the most dehydrated samples (5–7

wt.% H2O) suffered a 33%–24% loss of H2O sorption capacity. Valueva and Goryainov (1992) prepared their dehydrated chabazite samples by heat treatment up to 400°C. In particular,

Fig. 11. Calculated enthalpy of immersion of chabazite as a function of H2O content; error bars represent propagated errors using one standard deviations (1␴). The experimental immersion data are those of Barrer and Cram (1971) and Valueva and Goryainov (1992). Data obtained assuming that ⌬C៮ p equals 0 instead of 3R are also presented for comparison.

Hydration-dehydration behavior of chabazite

2305

Table 6. Xjmax values estimated by nonlinear regression of Equation 17 excluding the heat capacity contribution (⌬C៮ p ⫽ 0; using Microsoft EXCEL Solver tool), linear regression results for each type of H2O in chabazite (S-1, S-2, and S-3) using the new Xjmax values and the model given by Equation 18 excluding the heat capacity contribution (⌬៮ Cp ⫽ 0), and corresponding standard-state partial molar values.

Xjmax A (no unit) B (K) C (K) D (K) r2 lnKj Std. error (1␴) 0 ⌬␮hydr (J/mol) 0 ៮ hydr ⌬H (J/mol) 0 ⌬S៮ hydr (J/mol.K) W1 (J/mol) W2 (J/mol)

S-1

S-2

S-3

0.1048 ⫺17.3 (0.1) 7774 (56) ⫺232 (65) ⫺690 (60) 0.9969 0.15 ⫺21786 (587) ⫺64636 (462) ⫺143.79 (1.22) 1928 (544) 5739 (499)

0.1431 ⫺18.2 (0.3) 11623 (192) ⫺5071 (203) 3262 (162) 0.9842 0.20 ⫺51581 (1812) ⫺96633 (1597) ⫺151.18 (2.87) 42160 (1691) ⫺27117 (1349)

0.0403 ⫺16 (1) 17873 (745) ⫺14543 (1449) 9352 (1130) 0.9036 0.35 ⫺108928 (6656) ⫺148593 (6196) ⫺133.11 (8.17) 120907 (12046) ⫺77756 (9394)

Note: numbers in parentheses represent uncertainties propagated from the standard errors (1␴) of the regression coefficients.

all the samples containing 0 to 3 mol of H2O pfu were pretreated above 175°C. Barrer and Cram (1971) used a much milder heat treatment (up to 190°C, for 2 to 12 hours) for their partially dehydrated samples but used a heat-treatment of 360°C to completely outgas the sample.

Measurements of the enthalpy of dehydration using transposed temperature drop (TTD) calorimetry (e.g., Shim et al., 1999; Ogorodova et al., 2002) may have complications related to these irreversible structural modifications (Shim et al., 1999; Ogorodova et al., 2002). TTD measurements involve dropping

Fig. 12. Predicted fugacity of water (in Log fH2O, with f in bars) as a function of temperature and H2O-content of chabazites chemically similar to the Wasson’s Bluff chabazite. Symbols represent the liquid-vapor line. Under this line, chabazite is in equilibrium with vapor. Above this line, it is in equilibrium with water.

2306

C. I. Fialips, J. W. Carey, and D. L. Bish

Fig. 13. Predicted H2O-content (in wt.%) of chabazites chemically similar to the Wasson’s Bluff chabazite as a function of temperature and water fugacity (in Log fH2O, with f in bars).

a hydrated sample from room temperature (RT) into an empty Pt crucible at high temperature (i.e., 700°C). The measured enthalpy contains three components: the enthalpy of dehydration at 700°C, the enthalpy change between RT and 700°C, and the enthalpy of any structural transition occurring between RT and 700°C. The enthalpy of dehydration is obtained by comparing these measurements to a TTD measurement of a fully dehydrated sample or, if this is not available, an estimated value for the enthalpy change of the anhydrous material. Both Shim et al. (1999) and Ogorodova et al. (2002) had to use the estimation method. The estimation introduces uncertainty of ⬃4 kJ/mol according to Shim et al. (1999), which is a significant fraction of the 5 kJ/mol difference between our results. In addition, a structural transition will effect the measured enthalpy in a way that depends on whether the transition is endothermic (the apparent enthalpy of dehydration would be greater) or exothermic (the apparent enthalpy of dehydration would be smaller). If the structural transition does explain some of the difference between our phase-equilibria results and the TTD data, then the transition must be exothermic in nature. Shim et al. (1999) also made TTD measurements of partially dehydrated samples that were obtained by heat treatment between 300 and 700°C. They extracted partial molar enthalpy of hydration values by the difference in enthalpy of samples with differing H2O contents (Fig. 10). This comparison allowed Shim et al. (1999) to eliminate the requirement for estimating the enthalpy of anhydrous chabazite in their calculation of partial molar enthalpy. The effect of a structural transition on these measurements is difficult to predict and depends on

whether the two samples used in the comparison had different levels of structural modification. It is likely, however, that the potential effects of a structural transition would be progressively more pronounced for decreasing values of n (Fig. 10). Enthalpy of immersion (Himm) measurements require dehydrating chabazite to varying amounts before immersing the sample in water and recording the enthalpy change (Barrer and Cram, 1971; Valueva and Gorianov, 1992). As shown in Figure 4, the H2O sorption capacity of our chabazite decreases after exposure to temperatures above 175°C. Thus, during immersion experiments, the less hydrated samples (n ⬍3) probably did not rehydrate fully and the measured Himm would be too small (not as exothermic; Fig. 11). The extent of the undermeasurement of the Himm would increase with the temperature of pretreatment and thus with the decrease in n. Such an under-measurement of Himm is supported by the apparent change in the slope of the data of Valueva and Goryainov ៮ hydr be(1992) in Figure 11. For most if not all zeolites, ⌬H comes more negative during dehydration and thus the slope of the Himm data should steepen with decreasing n (Carey and Bish, 1996). The observation that the slope becomes shallower ៮ hydr and likely reflects the structural implies a less negative ⌬H transition. The Barrer and Cram (1971) Himm data are perhaps even more instructive (Fig. 11). Their partially dehydrated chabazite samples at n ⬎1.5 were obtained at lower temperature (from 0 to 190°C, for 2 to 12 hours) and probably are not significantly affected by loss of H2O sorption capacity. Their Himm values in this region are similar to Valueva and Gorianov (1992) and in

Hydration-dehydration behavior of chabazite

reasonable agreement with our calculated enthalpies of hydration (Fig. 11). However, to obtain a completely dehydrated sample, Barrer and Cram (1971) used a pretreatment of 360°C for 24 hours. This fully dehydrated sample has apparently lost significant H2O sorption capacity as shown by a Himm value that is not even as negative as their value at n ⬃1.5. The observation that chabazite undergoes irreversible loss of sorption capacity and structural modifications before complete dehydration creates significant complications in the measurement of the total enthalpy of hydration. Our use of phase equilibria at high temperature to calculate the integral molar enthalpy of hydration has significant uncertainty because of the loss of sorption capacity at elevated temperature. Similarly, the interpretation of both the TTD and immersion calorimetry data are complicated by the potential for irreversible structural transitions and loss of sorption capacity. Figures 10 and 11 show that all methods are in substantial agreement in the measurement of the partial molar enthalpy of hydration at n ⬎2. Given the uncertainties, the true values of the partial molar enthalpy for n ⬍2 are unknown at present. However, we note that the fact that chabazite does experience irreversible changes upon dehydration to small H2O contents is consistent with significantly more energetic H2O at low H2O content. Our results allow calculation of the detailed evolution of the partial molar enthalpy of hydration and reveal distinct transitions in the energetics of H2O (Fig. 10). These transitions reflect three different energetic types of H2O, consistent with the calorimetry and Raman spectra observations of Valueva and Gorianov (1992). For the present, we lack sufficient crystallographic information to associate the three types of H2O (S-1, S-2, and S-3) with specific crystallographic sites in the chabazite structure. As pointed out by Shim et al. (1999) and shown by crystal structural studies (e.g., Mortier et al., 1977; Grey et al., 1999; Grey et al., 2000), the site occupancy of extraframework cations, such as Ca2⫹, may differ slightly with the hydration state of chabazite. One could argue that possible variations in cation location and thus binding of H2O may be responsible for the distinct H2O hydration energetics observed in the present study. However, Grey et al. (1999) concluded that minimal site-to-site cation diffusion would be expected for chabazite even at elevated temperature (up to 370°C) and that no significant diffusion would occur for low water content. Likewise, the reversible behavior of chabazite hydration we observed suggests that, although cation migration may occur during hydration and dehydration, the effect of such structural modifications on the energetics is probably minor. 7. APPLICATION

The thermodynamic model and calculated data (Eqn. 19 and Table 5) allow prediction of the hydration state of chemically similar chabazites at any T and PH2O. Such calculations may be useful in the analysis of the stability of chabazite in the natural environment and in the interpretation of experimental investigations of the stability of chabazite. Figures 12 and 13 show our predicted H2O content of chabazite in plots of H2O-wt.% vs. T contoured at constant water-fugacity (fH2O) and LogfH2O vs. T contoured at constant H2O content for T between 25 and 400°C and fH2O between 10⫺5 and 104 bars. For example, Figure 12 could be used to predict the equilibrium fH2O for a

2307

chabazite containing 18 wt.% H2O at 175°C (slightly less than 1 bar). Similarly, Figure 13 could be used to predict the H2O content of a chabazite equilibrated with a fH2O of 0.001 bars at 225°C (⬃4 wt.%). In the natural environment and in most phase equilibria studies, zeolites generally form under water-vapor saturated conditions. From Figures 12 and 13, chabazite will crystallize in most environments with an H2O content between 20 and 22 wt.% (␪ ⬎0.9). Among diagenetic/metamorphic sequences of zeolites, chabazite forms at relatively low temperature, generally ⬍100°C (e.g., Chipera and Apps, 2000), and the likely H2O content of chabazite at low temperatures is between 21 and 22 wt.%. Samples of chabazite stored under relatively dry conditions will dehydrate slightly according to Figures 12 and 13. For example, a chabazite sample stored at 25°C and 5%RH (fH2O ⬃1.34 mbar) would equilibrate from an H2O content of nearly 22 wt.% to ⬃20 wt.% (LogfH2O ⬃⫺3). Acknowledgments—We wish to thank Marjorie G. Snow of LANL for performing the microprobe analyses of the chabazite sample. This work was supported by Los Alamos National Laboratory-Directed Research and Development funding and by NASA MFRP02-0000-0085. We express our thanks to F. Podosek, R. Wogelius, P. Neuhoff, and K. Brodie for their suggestions and comments which improved the manuscript. Associate editor: R. Wogelius REFERENCES Alberti A., Galli E. and Vezzalini G. (1982) Position of cations and water molecules in hydrated chabazite. Natural and Na-, Ca-, Srand K-exchanged chabazites. Zeolites 2, 303–309. Armbruster T. and Gunter M. E. (1991) Stepwise dehydration of heulandite-clinoptilolite from Succor Creek, Oregon, U.S.A.: A single-crystal X-ray study at 100 K. Am. Mineral. 76, 1872–1883. Barrer R. M. and Cram P. J. (1971) Heats of immersion of outgassed and ion-exchanged zeolites. In Molecular Sieve Zeolites—II (eds. E. M. Flanigen and L. B. Sand), pp. 105–131. Am. Chem. Soc., Washington, DC. Bish D. L. (1984) Effects of exchangeable cation composition on the thermal expansion/contraction of clinoptilolite. Clays Clay Miner. 32, 444 – 452. Bish D. L. and Carey J. W. (2001) Thermal behavior of natural zeolites. In Natural Zeolites: Occurrence, Properties, Applications (ed. D. L. Bish and D. W. Ming), Vol. 45 pp. 403– 452. Rev. Mineral. Geochem, Mineral. Soc. Am. Bish D. L. and Ming D. W. (2001) Natural Zeolites: Occurrence, Properties, Applications. Vol. 45. Rev. in Mineralogy and Geochemistry, Mineral. Soc. Am., 654p. Calligaris M., Nardin G., Randaccio L. and Chiaramonti P. C. (1982) Cation-site location in a natural chabazite. Acta Cryst. B38, 602– 605. Carey J. W. (1993) The heat capacity of hydrous cordierite above 295K. Phys. Chem. Miner. 19, 578 –583. Carey J. W. and Bish D. L. (1996) Equilibrium in the clinoptiloliteH2O system. Am. Mineral. 81, 952–962. Chipera S. J. and Apps J. A. (2000) Geochemical stability of natural zeolites. In Natural Zeolites: Occurrence, Properties, Applications (eds. D. L. Bish and D. W. Ming), Vol. 45, pp. 403– 452. Rev. Mineral. Geochem, Mineral. Soc. Am. Drebushchak V. A. (1999) Measurements of heat of zeolite dehydration by scanning heating. J. Therm. Anal. Calorim. 58, 653– 662. Fridriksson T., Bish D. L. and Bird D. K. (2003a) Hydrogen-bonded water in laumontite I: X-ray powder diffraction study of water site occupancy and structural changes in laumontite during room-temperature isothermal hydration/dehydration. Am. Mineral. 88, 277– 287.

2308

C. I. Fialips, J. W. Carey, and D. L. Bish

Fridriksson T., Carey J. W., Bish D. L., Neuhoff P. S. and Bird D. K. (2003b) Hydrogen-bonded water in laumontite II: Experimental determination of site-specific thermodynamic properties of hydration of the W1 and W5 sites. Am. Mineral. 88, 1060 –1072. Gottardi G. and Galli E. (1985) Natural Zeolites. Springer Verlag, Berlin, New York, 409p. Grey T., Gale J., Nicholson D., Artacho E. and Soler J. (2000) A computational exploration of cation locations in high-silica Cachabazite. Stud. Surf. Sci. Catal. 128, 89 –98. Grey T., Gale J., Nicholson D. and Peterson B. (1999) A computational study of calcium cation locations and diffusion in chabazite. Micropor. Mesopor. Mat. 31, 45–59. Haar L., Gallagher J. S. and Kell G. S. (1984) NBS/NRC Steam Tables: Thermodynamic and Transport Properties and Computer Programs for Vapor and Liquid States of Water in SI Units. Hemisphere Publishing, Washington, D.C., 320p. Howard C. J. (1982) The approximation of asymmetric neutron powder diffraction peaks by sums of Gaussians. J. Appl. Crystallogr. 15, 615– 620. Johnson G. K., Tasker I. R., Jurgens R. and O’Hare P. A. G. (1991) Thermodynamic studies of zeolites: Clinoptilolite. J. Chem. Thermodyn. 23, 475– 484. Larson A. C. and Von Dreele R. B. (1994) Generalized Structure Analysis System. Los Alamos Nat. Lab. Rept. LA-UR 86 –748, 161p.

Mortier W. J., Pluth J. J. and Smith J. V. (1977) Positions of cations and molecules in zeolites with the chabazite framework. I. Dehydrated Ca-exchanged chabazite. Mat. Res. Bull. 12, 97–102. Ogorodova L. P., Kiseleva I. A., Mel’chakova L. V. and Belitskii I. A. (2002) Thermodynamic properties of calcium and potassium chabazites. Geochem. Int. 40, 466 – 471. Sachs W. H. (1976) Implicit multifunctional nonlinear regression analysis. Technometrics 18, 161–173. Shim S. H., Navrotsky A., Gaffney T. R. and MacDougall J. E. (1999) Chabazite: energetics of hydration, enthalpy of formation and effects of cations on stability. Am. Mineral. 84, 1870 –1882. Toby B. H. (2001) EXPGUI, a graphical user interface for GSAS. J. Appl. Crystallogr. 34, 210 –213. Valueva G. P. and Goryainov S. V. (1992) Chabazite during dehydration (thermochemical and Raman spectroscopy study). Russian Geol. Geophys. 33, 68 –75. Woodcock D. A. and Lightfoot P. (1999) Negative thermal expansion in the siliceous zeolites chabazite and ITQ-4: a neutron powder diffraction study. Chem. Mater. 11, 2508 –2514.

ELECTRONIC ANNEX Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.gca.2004.11.007.