Molecular H2O in armenite, BaCa2Al6Si9O30�2H2O, andepididymite, Na2Be2Si6O15�H2O: Heat capacity, entropy

and local-bonding behavior of confined H2O in microporoussilicates

Charles A. Geiger a,b,*, Edgar Dachs b, Maria Chiara Dalconi c, Gilberto Artioli c

a Institut fur Geowissenschaften, Abteilung Mineralogie, Christian-Albrechts-Universitat Kiel, D-24098 Kiel, GermanybFachbereich Materialforschung und Physik, Abteilung Mineralogie, Universitat Salzburg, Hellbrunnerstrasse 34, A-5020 Salzburg, Austria

cDipartimento di Geoscienze, Universita degli Studi di Padova, Via Giotto 1, I-35137 Padova, Italy

Received 1 February 2010; accepted in revised form 20 May 2010; available online 9 June 2010

Abstract

Armenite, ideal formula BaCa2Al6Si9O30�2H2O, and its dehydrated analog BaCa2Al6Si9O30 and epididymite, ideal formulaNa2Be2Si6O15�H2O, and its dehydrated analog Na2Be2Si6O15 were studied by low-temperature relaxation calorimetry between5 and 300 K to determine the heat capacity, Cp, behavior of their confined H2O. Differential thermal analysis and thermo-gravimetry measurements, FTIR spectroscopy, electron microprobe analysis and powder Rietveld refinements were under-taken to characterize the phases and the local environment around the H2O molecule.

The determined structural formula for armenite is Ba0.88(0.01)Ca1.99(0.02)Na0.04(0.01)Al5.89(0.03)Si9.12(0.02)O30�2H2O and forepididymite Na1.88(0.03)K0.05(0.004)Na0.01(0.004)Be2.02(0.008)Si6.00(0.01)O15�H2O. The infrared (IR) spectra give information onthe nature of the H2O molecules in the natural phases via their H2O stretching and bending vibrations, which in the caseof epididymite only could be assigned. The powder X-ray diffraction data show that armenite and its dehydrated analog havesimilar structures, whereas in the case of epididymite there are structural differences between the natural and dehydratedphases. This is also reflected in the lattice IR mode behavior, as observed for the natural phases and the H2O-free phases.The standard entropy at 298 K for armenite is S� = 795.7 ± 6.2 J/mol K and its dehydrated analog is S� = 737.0 ± 6.2 J/mol K. For epididymite S� = 425.7 ± 4.1 J/mol K was obtained and its dehydrated analog has S� = 372.5 ± 5.0 J/mol K.The heat capacity and entropy of dehydration at 298 K are DCrxn

p = 3.4 J/mol K and DSrxn = 319.1 J/mol K andDCrxn

p = �14.3 J/mol K and DSrxn = 135.7 J/mol K for armenite and epididymite, respectively. The H2O molecules in bothphases appear to be ordered. They are held in place via an ion–dipole interaction between the H2O molecule and a Ca cationin the case of armenite and a Na cation in epididymite and through hydrogen-bonding between the H2O molecule and oxygenatoms of the respective silicate frameworks. Of the three different H2O phases ice, liquid water and steam, the Cp behavior ofconfined H2O in both armenite and epididymite is most similar to that of ice, but there are differences between the two silicatesand from the Cp behavior of ice. Hydrogen-bonding behavior and its relation to the entropy of confined H2O at 298 K isanalyzed for various microporous silicates.

The entropy of confined H2O at 298 K in various silicates increases approximately linearly with increasing average wave-number of the OH-stretching vibrations. The interpretation is that decreased hydrogen-bonding strength between a H2O mol-ecule and the silicate framework, as well as weak ion–dipole interactions, results in increased entropy of H2O. This results in

0016-7037/$ - see front matter � 2010 Elsevier Ltd. All rights reserved.

doi:10.1016/j.gca.2010.05.033

* Corresponding author at: Fachbereich Materialforschung und Physik, Abteilung Mineralogie, Universita t Salzburg, Hellbrunnerstrasse

34, A-5020 Salzburg, Austria. Tel.: +43 662 8044 5407.E-mail address: cageiger@sbg.ac.at (C.A. Geiger).

www.elsevier.com/locate/gca

Available online at www.sciencedirect.com

Geochimica et Cosmochimica Acta 74 (2010) 5202–5215

increased amplitudes of external H2O vibrations, especially translations of the molecule, and they contribute strongly to theentropy of confined H2O at T < 298 K.� 2010 Elsevier Ltd. All rights reserved.

1. INTRODUCTION

The H2O molecule, in its various phases, plays a centralrole in many natural Earth processes and it is essential forlife on the planet. Although H2O is a “simple” three-atommolecule, it interacts in complex ways with a variety ofEarth materials and it can have profound effects on theirchemical and physical behavior. This, in turn, affects vari-ous geologic processes. The H2O system is notable, forexample, for its unusual properties such as its high boilingand freezing points and by the large thermodynamic heatcapacity of liquid H2O. These physical properties are the re-sult of intermolecular hydrogen-bonding between the mol-ecules. Hydrogen-bonding also plays an important role inmany H2O-containing systems (Eisenberg and Kauzmann,1969; Jeffrey, 1997; Marechal, 2007). However, the role ofhydrogen-bonding and associated thermodynamic behaviorin natural H2O-containing systems has received relativelylittle mineralogical and geochemical study. Much workneeds to be done in this direction before a good level of sci-entific understanding is reached.

H2O-bearing minerals provide an excellent startingpoint to investigate how the H2O molecule interacts withnatural crystalline materials. Many types of minerals con-tain H2O (Hawthorne, 1992) and they interact with it invarious ways. One general group of minerals, namelymicroporous silicates or silicates with microporous frame-works (see McCusker et al., 2001 for a definition of microp-ores in crystals), offer the possibility to investigate thenature of hydrogen-bonding and thermodynamic behavior,especially heat capacity, Cp, of confined H2O at a relativelysimple, yet fundamental level. The most studied phases inthis regard are the zeolites with their confined, channelH2O molecules that are hydrogen bonded to oxygen atomsof the aluminosilicate framework and/or each other (Kvick,1986). Much and varied types of research have been doneon them (Breck, 1974; Bish and Ming, 2001). In terms ofthermodynamic behavior, because hydrogen-bonding asso-ciated with confined H2O is generally a relatively weakforce, it expresses itself predominantly in Cp behavior atlow temperatures – at T < 300 K. There has been calorimet-ric Cp work done in this direction on a number of naturalzeolites (e.g., Johnson et al., 1982, 1983, 1985, 1992; Paukovet al., 2002). However, Cp investigations at T < 300 K onthe behavior of H2O in other types of silicates have only re-cently started (Paukov et al., 2007; Geiger and Dachs,2009). More study is required before a certain quantitativelevel of understanding is reached.

Herein, we investigate two H2O-bearing phases havingmicroporous frameworks, namely armenite, BaCa2Al6-Si9O30�2H2O, and epididymite, Na2Be2Si6O15�H2O, withan emphasis on the thermodynamic and bonding behaviorof the confined H2O molecules that occur in structuralmicropores. Armenite is a seldom-occurring silicate that is

typically found in hydrothermal veins and fissures and itwas first described from the Armenmine in Kongsberg, Nor-way (Neumann, 1941). It has since been reported from otherlocalities worldwide (e.g., Pouliot et al., 1984; Mason, 1987;Semenenko et al., 1987; Zak andObst, 1989; Balassone et al.,1989; Senn, 1990; Fortey et al., 1991). It is a double-ring sil-icate belonging to the milarite group of minerals (Haw-thorne et al., 1991). Its structure and crystal-chemicalproperties were investigated by Armbruster and Czank(1992) and Armbruster (1999). Epididymite is also a ratherrare silicate and it occurs with various late-stage mineralsin alkaline pegmatites. It has been reported from variouslocalities in a number of works (e.g., Brøgger, 1890; Flink,1894; Levy, 1961; Petersen, 1966; Cerny, 1963; Shilin andSemenov, 1957; Mandarino and Harris, 1968). Its crystalstructure was studied and recently clarified by Gatta et al.(2008). We are not aware of any calorimetric or thermody-namic studies that have been made on either phase.

We focus our study here, therefore, on the nature of theconfined H2O molecules in these two silicates. Importantquestions include the nature of hydrogen-bonding and heatcapacity and entropy, S, behavior of the H2O and the rela-tionship between bonding and thermodynamic behavior.The results give further insight on H2O behavior and itsproperties in different types of structural micropores in sil-icates. This allows better understanding of the physical andchemical nature of H2O–silicate interactions.

2. EXPERIMENTAL METHODS

2.1. Sample characterization

For this study, single-crystal armenite from Wasenalp,Valais, Switzerland and a cm-sized piece of epididymitematerial from the Zomba-Malosa pluton in southern Mala-wi were used. More information on the armenite samplecan be found in Senn (1990) and Armbruster (1999) andthat for the epididymite sample in Gatta et al. (2008).

To begin, several hundred milligrams of single-phaseclear and transparent armenite and white-looking epididy-mite, as based on examination under a binocular micro-scope, were collected and ground in an agate mortar.Preliminary X-ray powder diffraction measurements onground material using a Siemens D-5000 diffractometerwith 0.1� step scans every 10 s between 10� and 60� 2hshowed only reflections that could be indexed to armeniteand epididymite, respectively. Both samples were subse-quently measured using an X’Pert diffractometer in high-resolution mode (see below). Full-profile Rietveld structurerefinements confirmed both the purity of the samples andconformity to published structure models.

The chemical composition of armenite and epididymitewas determined using a JEOL JXA-8900R electron probe.A number of point analyses were made on different parts

H2O in Microporous Silicates 5203

of single-crystal fragments taken from the bulk sample.Analyses were made for the elements Si, Ti, Al, Cr, Fe,Mn, Mg, Ca, Ba, K and Na using an accelerating voltageof 15 kV and a beam current of 15 nA with a beam diame-ter of 10 lm. The PRZ method (modified ZAF) was usedfor data correction. For armenite 30 oxygen atoms were as-sumed and for epididymite 15 oxygen atoms in the calcula-tion of their crystal-chemical formulae.

2.2. Differential thermal analysis and thermogravimetry

(DTA/TG) measurements

DTA/TG measurements were carried out in order todetermine the dehydration behavior of both armenite andepididymite. The samples were placed in Pt capsules andheated at 10 �C/min, starting at room temperature, usinga SHIMADZU DTG-60H instrument under a flow of N2

gas.

2.3. Dehydration experiments

A roughly 100 mg fraction of each ground sample wastaken and used for heating experiments to expel the con-fined H2O. In the case of armenite, the sample was placedin an open Au-capsule and heated at 50 �C/h to 600 �Cand held there for 12 h. The epididymite sample was heatedin two steps, first at 550 �C for 24 h and then at 725 �C for24 h. Heating was done using a four-zone tube furnace witha temperature control of better than ±2 �C, which was fur-ther checked by use of an external Cr–CrNi thermocouple.Following heating, the Au-capsule with the sample was re-moved from the furnace and cooled and the sample wascarefully removed from the capsule. The heated materialwas characterized by X-ray powder diffraction and IR spec-troscopy (see below).

2.4. Rietveld refinements

The natural and heated phases were placed in 0.4 mmdiameter boron-glass capillaries and measured using astate-of-the-art PANalytical X’Pert PRO powder diffrac-tometer in the focusing mode. The instrumental configura-tion encompasses a fine-focus Cu X-ray tube, a focusingmirror, antiscatter slits, 0.002 rad Soller slits, and a PIXceldetector. Data collection conditions were: 3–100� 2h range,0.013� steps, time per step 200 s, and 10 repetitions for eachsample to ensure adequate statistics. The powder scans wererefined using the GSAS system (Larson and von Dreele,1988).

2.5. Fourier transform infrared spectroscopy

The FTIR spectra of armenite and epididymite and theirdehydrated analogs were measured with a BRUKER IFSS/66v FTIR spectrometer, using the KBr pellet technique, inthe wavenumber region between 4000 and 350 cm�1. Mea-surements on the dehydrated samples were made severalweeks after the samples were removed from the furnaceand then stored in plastic vials. A pellet of 13 mm diameterwas prepared by pressing a fine mixture consisting of one

mg of sample and 200 mg of dried KBr under vacuum.The pellet was kept in a vacuum desiccator for 24 h at150 �C to minimize adsorbed water. Spectra were recordedunder vacuum with a resolution of 2 cm�1 by combining512 scans.

2.6. Low-temperature calorimetry measurements

The low-temperature heat capacity of armenite and epi-didymite and their dehydrated analogs was measured with acommercially designed heat-pulse calorimeter (i.e., heatcapacity option of the Physical Properties MeasurementSystem (PPMS) manufactured by Quantum Design). Mea-surements were made at temperatures between 5 and 300 Kon mg-sized samples. The armenite sample had a mass of22.02 mg with a formula weight of 1132.60 g/mol and itsdehydrated analog had a mass of 20.11 mg with a formulaweight of 1096.57 g/mol (see below for sample composi-tions). The epididymite sample had a mass of 17.40 mg witha formula weight of 493.52 g/mol and the dehydrated ana-log a mass of 12.76 mg with a formula weight of 475.51g/mol. For measurement, the samples were placed in anAl-container covered by a lid. Cp data were collected at60 different temperatures on cooling from 300 to 5 K witha logarithmic spacing, with Cp determined three separatetimes at each temperature. Details behind the relaxationcalorimetric method, as well as a description of the dataacquisition and evaluation procedures, are given in Dachsand Bertoldi (2005) and Dachs and Geiger (2006).

3. EXPERIMENTAL RESULTS

3.1. Chemical composition

Both the armenite and epididymite crystals werehomogeneous in terms of composition. The determinedstructural formula for armenite is Ba0.88(0.01)Ca1.99(0.02)Na0.04(0.01)Al5.89(0.03)Si9.12(0.02)O30�2H2O and for epididy-mite Na1.88(0.03)K0.05(0.004)Na0.01(0.004)Be2.02(0.01)Si6.00(0.01)O15�H2O assuming 10.20 wt.% BeO. Stoichiometric H2Ocontents were assumed for both phases.

3.2. DTA/TG measurements

The DTA/TG data are shown in Figs. 1 and 2. The TGresults show that armenite begins to lose water roughlyaround 300 �C and at the end of the measurement at800 �C the recorded weight loss was 3.46 wt.%. The theoret-ical weight loss of H2O from BaCa2Al6Si9O30�2H2O is3.13%. This difference is about 10%, but is considered tobe within the overall experimental uncertainty. Weight losswas continuous and appears to occur in a single dehydra-tion process. The DTA trace does not show any evidencefor a phase transition or other major structural changeother than that related to the loss of H2O. Epididymite be-gins to show weight loss in the TG trace roughly around700 �C and at the end of the measurement at 1000 �C thesample decreased 3.75% in weight. The theoretical weightloss of H2O from Na2Be2Si6O15�H2O is 3.67%. Weight lossoccurred between about 700 and 900 �C. The DTA signal

5204 C.A. Geiger et al. /Geochimica et Cosmochimica Acta 74 (2010) 5202–5215

showed no evidence for any notable structural modifica-tions until about 850 �C.

3.3. FTIR spectroscopy

The IRpowder spectrumof armenite in theOH-stretchingregion between 4000 and 3000 cm�1 and in the H2O bendingregion between 1800 and 1500 cm�1 is shown in Fig. 3. Twomajor OH-stretching bands are observed at 3469 and3410 cm�1. In addition, a small shoulder on the high wave-number side of the strong doublet is observed at roughly3600 cm�1, as is a very weak band at 3256 cm�1. A broadasymmetric H2O bending mode is observed at 1645 cm�1.The dehydrated sample shows, in comparison, only a weakbroad band at roughly 3450 cm�1 and in the bending regionthe spectrum is nearly featureless (Fig. 3). The IR powderspectra of natural and dehydrated armenite in the latticevibration region between 1500 and 350 cm�1 are shown inFig. 4. The spectra are complex in terms of full interpretationand no specific modes assignments can be made. A compar-ison of the two spectra shows that there are no major differ-ences between the spectrum of natural armenite and itsdehydrated analog.

The IR powder spectrum of epididymite in the OH-stretching region between 4000 and 3000 cm�1 and in theH2O bending region between 1800 and 1500 cm�1 is shownin Fig. 3. Two sharp OH-stretching bands are observed at3523 and 3479 cm�1. A couple of very weak bands atapproximately 3392 and 3318 cm�1 are also observed. A

single H2O bending mode is observed at 1660 cm�1. Thedehydrated sample shows, in comparison, only a weakbroad band at roughly 3475 cm�1 and in the bending regionthe spectrum is also nearly featureless (Fig. 3). The IR pow-der spectra of natural and dehydrated epididymite in the re-gion between 1500 and 350 cm�1 are shown in Fig. 4. Thespectra are, once again, complex and no specific modesassignments can be made. There are notable differences inthe 1200 and 900 cm�1 region between the spectrum of nat-ural epididymite and heated epididymite, but at lowerwavenumbers differences between the two spectra are less.

3.4. X-ray powder measurements and Rietveld refinements

The X-ray powder patterns for natural and heat-treatedarmenite and natural and heat-treated epididymite areshown in Figs. EA1 and EA2 (see Electronic Annex),

Fig. 1. DTA/TGA results on armenite. The DTA trace is shown by

the thin line.

Fig. 2. DTA/TGA results on epididymite. The DTA trace is shown

by the thin line.

Fig. 3. IR powder spectra of natural (solid line) and dehydrated

armenite (dashed line) and natural (solid line) and dehydrated

epididymite (dashed line) in the OH-stretching region between 4000

and 2800 cm�1 and in the H2O bending region from 1800 and

1500 cm�1.

Fig. 4. IR powder spectra of natural (solid line) and dehydrated

armenite (dot-dashed line) and natural (soild line) and dehydrated

epididymite (dot-dashed line) in the lattice vibration region

between 1500 and 350 cm�1.

H2O in Microporous Silicates 5205

respectively. Differences between the diffraction patterns ofnatural and heat-treated armenite are small and involveminor peak shifts due to cell contraction (see Fig. EA1),whereas there are notable differences between the patternsof natural and heat-treated epididymite (Fig. EA2).

The results of the full-profile Rietveld refinements forboth natural phases (Figs. 5 and 6) were checked againstthe structure models available in the literature (Armbruster,1999; Gatta et al., 2008). The residual factors (i.e.,Rp = 0.055 and 0.064 and wRp = 0.078 and 0.079 forarmenite and epididymite, respectively) and the observedvs. calculated patterns confirm the purity of the phases.The R(F2) values for both natural armenite and epididymiteare in the range 0.08–0.10, depending on whether the atom-ic displacement factors are fixed, as reported in the single-crystal studies, or tentatively refined. These values indicatethat both powder refinements are consistent with the single-crystal results. The residuals for the Rietveld refinement ofarmenite show some misfit at low and high angles. This ismainly caused by an unsatisfactory treatment of theabsorption correction. Despite the fact that several modelsfor the absorption correction were attempted by fixing thestructure model, the absorption coefficient is sufficientlyhigh, due to the presence of Ba, to cause biases in the profilerefinement. The refined unit-cell parameters for thenatural phases are: armenite (Pnc2) a = 18.6504(3), b =10.7007(1) and c = 13.8826(2) A giving a unit-cell volumeof 2770.57(8) A3 and epididymite (Pnma) with a =12.7383(4), b = 13.6403(3) and c = 7.3526(2) A giving aunit-cell volume of 1277.55(7) A3.

Attempts were made to refine the structures of the twoheat-treated samples, with little success in both cases,although for different reasons. In the case of armenite, thestructure model of the natural phase can be used to fit thediffraction pattern of the dehydrated phase. Thus, there islittle doubt that the space group and framework geometry,

including the positions of the Ba and Ca cations, differ littlebetween the two phases. However, because the Ba atomsdominate the total scattering, the powder data are inade-quate to characterize any small structural changes due tothe loss of H2O molecules and local geometry changesaround the Ca cation in the dehydrated phase. DifferenceFourier syntheses are featureless, with the major peaks re-lated to the anisotropic residual electron density associatedwith the Ba atoms. The refined unit-cell parameters area = 18.4181(3), b = 10.5999(2) and c = 14.0327(2) A givinga cell volume of 2739.61(9) A3. They indicate net minorcontraction in the structure along a and b, and minorexpansion along c and a decrease in the cell volume of30.96 A3 upon dehydration.

In the case of epididymite, the powder diffraction pat-tern of the heat-treated sample cannot be fit with a structuremodel derived from the natural phase. Compared to theX-ray pattern of the natural phase, a number of additionalBragg peaks are observed (for example at 14.01�, 14.12�,etc. 2h), which can only be explained by a change in spacegroup. A number of possibilities can be envisaged, includ-ing several orthorhombic subgroups violating the system-atic extinctions of the original Pnma space group, or amonoclinic subgroup caused by cell distortion during dehy-dration. Several refinement attempts using orthorhombicsubgroups and a doubling of the c-axis proved to be unsuc-cessful. It is expected that single-crystal study will be re-quired to solve the structure of H2O-free epididymite.

3.5. Low-temperature calorimetry

The raw experimental Cp data for natural armenite andepididymite and their dehydrated analogs are given in Ta-bles EA1 and EA2 (see Electronic Annex). The raw Cp datawere fitted to an equation of the form (Boerio-Goates et al.,2002):

Fig. 5. Rietveld refined powder diffraction pattern of natural armenite. The graph shows the observed (points), calculated (solid line) and the

difference curve (lower part).

5206 C.A. Geiger et al. /Geochimica et Cosmochimica Acta 74 (2010) 5202–5215

Cp ¼ 3Rðn � DðhDÞ þ m � EðhEÞ þ ns � SðhSÞÞ; ð1Þ

where D(hD), E(hE) and S(hS) are Debye, Einstein andSchottky functions (see Table 1 for their definitions), R isthe gas constant and n, m, ns, hD, hE and hS are adjustableparameters (Table 1). A good fit of the data could be madeby dividing the Cp data into two segments, as given by thetwo sets of fit parameters in Table 1. The segments are di-vided at the temperature, Tswitch, where both fits yield iden-tical Cp values. Using (1), the experimental Cp values couldbe reproduced well within 0.2–0.3 ± 0.2% at T > 50 K,within 0.6 ± 0.4% at T < 10 < 50 K, and within a few %

at T < 10 K. The fitted Cp results for natural armeniteand epididymite and their dehydrated analogs at 5 and10 K intervals are given in Tables 2 and 3.

The standard third-law entropy, S�, of the differentphases was determined by numerically solving the integral:

S�

¼ ST¼298:15 K � S

T¼0 K ¼

Z 298:15

0

Cp

TdT : ð2Þ

The uncertainty in S� was calculated as described in BenisekandDachs (2008) and the entropy atT = 0 Kwas assumed tobe zero. We obtained entropy values for natural armenite of

Fig. 6. Rietveld refined powder diffraction pattern of natural epididymite. The graph shows the observed (points), calculated (solid line) and

the difference curve (lower part).

Table 1

Parameters for calculating smoothed values of the heat capacity of armenite, dehydrated armenite, epididymite and dehydrated epididymite

from Eq. (1).

Armenite Dehydrated armenite Epididymite Dehydrated epididymite

Low-temperature

n (mol) 12.60046 12.01341 6.62327 6.37887

hD (K) 317.4413 349.4247 299.2115 339.1062

M (mol) 3.07216 3.69974 1.29481 1.45534

hE (K) 144.1993 142.5476 135.0973 148.0983

ns (mol) 0.561623 0.669348 0.204085 0.307865

hS (K) 79.6369 77.4202 79.7119 89.1265

Tswitch (K) 36.31 37.39 34.81 41.76

High-temperature

n (mol) 27.89796 28.41690 12.75018 6.36962

hD (K) 476.7712 545.2335 441.3839 262.5823

m (mol) 21.35167 16.55131 11.77776 18.59712

hE (K) 979.1070 1113.4587 919.7816 1188.4978

ns (mol) 5.880165 7.373312 2.909293 9.489310

hS (K) 147.3338 156.1916 142.7899 554.6221

The following functions (Gopal, 1966) were used to fit the data. (A) Debye function: CV ¼ 3 ð ThDÞ3R

hD=T0

x4ex

ðex�1Þ2dx, where hmD

k¼ hD and x ¼ hm

kT.

(B) Einstein function: CV ¼hE

Tð Þ2e

hET

e

hET �1

� �2. (C) Schottky function: CV ¼ðhS

TÞ2e

hS

T

1þe�hS

T

� �2. The corresponding Mathematica� programs are available on

request.

H2O in Microporous Silicates 5207

S� = 795.7 ± 6.2 J/mol K and for its dehydrated analogS� = 737.0 ± 6.2 J/mol K. For natural epididymite we ob-tain S� = 425.7 ± 4.1 J/mol K and for the dehydrated sam-ple one has S� = 372.5 ± 5.0 J/mol K. Results for both Cp

and S� at 298 K are summarized in Table 4.

4. DISCUSSION

4.1. The crystal chemistry of armenite and epididymite and

their dehydrated analogs

The crystal structure of armenite was investigated by X-ray single-crystal diffraction methods by Armbruster

(1999). The orthorhombic structure and framework of sym-metry Pnc2 is thus known, but the positions of the H atomsof the H2O molecule could not be determined. The struc-ture is shown in Fig. 7a. Belonging to the milarite groupof silicates, armenite contains six-membered double-ringsof corner-sharing Td-type SiO4 and AlO4 tetrahedra. Therings are cross-linked together by Tc tetrahedra and theresulting aluminosilicate “framework” contains smallmicropores next to Tc. In these micropores a Ca cation oran H2O molecule are found and they alternate with eachother along the c-axis. Six framework oxygen atoms andthe O anion of an H2O molecule coordinate every Ca cationproducing an irregular coordination polyhedron consisting

Table 2

Fitted Cp values for armenite and dehydrated armenite and the

difference between the two (i.e., H2O in armenite).

T

(K)

Cp

(armenite)

(J/mol K)

Cp (dehydrated

armenite)

(J/mol K)

Cp (one H2O in

armenite)

(J/mol K)

5 0.09188 0.0681 0.01189

6 0.1621 0.1232 0.01945

7 0.2709 0.2166 0.02715

8 0.4381 0.3733 0.0324

9 0.686 0.62 0.033

10 1.035 0.9797 0.02765

11 1.502 1.469 0.0165

12 2.1 2.1 0

13 2.838 2.88 �0.021

14 3.726 3.819 �0.0465

15 4.771 4.921 �0.075

20 12.49 12.97 �0.24

25 24.23 24.9 �0.335

30 39.25 39.67 �0.21

35 56.6 56.27 0.165

40 75.68 74.99 0.345

45 95.11 94.27 0.42

50 114.8 112.7 1.05

60 155 147.9 3.55

70 195.8 182.4 6.7

80 236.2 217.1 9.55

90 275.3 252 11.65

100 312.8 286.6 13.1

110 348.8 320.3 14.25

120 383.2 352.7 15.25

130 416.4 383.9 16.25

140 448.5 413.6 17.45

150 479.5 442 18.75

160 509.5 469.2 20.15

170 538.5 495.3 21.6

180 566.5 520.4 23.05

190 593.4 544.5 24.45

200 619.1 567.6 25.75

210 643.8 589.8 27

220 667.3 611.2 28.05

230 689.6 631.7 28.95

240 710.9 651.3 29.8

250 731 670.2 30.4

260 750 688.2 30.9

270 768 705.4 31.3

280 785.1 721.8 31.65

290 801.1 737.6 31.75

298.15 813.5 749.8 31.85

Table 3

Fitted Cp values for epididymite and dehydrated epididymite and

the difference between the two (i.e., H2O in epididymite).

T

(K)

Cp

(epididymite)

(J/mol K)

Cp (dehydrated

epididymite)

(J/mol K)

Cp (H2O in

epididymite)

(J/mol K)

5 0.06 0.04 0.02

6 0.11 0.06 0.04

7 0.17 0.10 0.07

8 0.27 0.17 0.10

9 0.41 0.26 0.15

10 0.60 0.39 0.21

11 0.86 0.57 0.29

12 1.18 0.80 0.38

13 1.59 1.09 0.50

14 2.08 1.45 0.63

15 2.66 1.88 0.78

20 7.05 5.26 1.79

25 13.79 10.67 3.12

30 22.37 17.67 4.70

35 32.23 25.77 6.46

40 43.19 34.71 8.48

45 54.20 44.20 10.00

50 65.05 53.59 11.46

60 86.44 72.16 14.28

70 107.60 90.76 16.84

80 128.40 109.40 19.00

90 148.50 127.90 20.60

100 167.80 145.90 21.90

110 186.20 163.10 23.10

120 203.80 179.70 24.10

130 220.60 195.50 25.10

140 236.90 210.60 26.30

150 252.70 225.10 27.60

160 268.10 239.10 29.00

170 283.10 252.60 30.50

180 297.70 265.60 32.10

190 311.90 278.20 33.70

200 325.60 290.40 35.20

210 339.00 302.10 36.90

220 351.80 313.50 38.30

230 364.20 324.40 39.80

240 376.10 334.80 41.30

250 387.50 344.90 42.60

260 398.30 354.50 43.80

270 408.70 363.70 45.00

280 418.60 372.50 46.10

290 428.10 380.90 47.20

298.15 435.40 387.50 47.90

5208 C.A. Geiger et al. /Geochimica et Cosmochimica Acta 74 (2010) 5202–5215

of CaO7-like clusters (Fig. 7b). The large Ba cations are lo-cated above and below the double-rings of tetrahedra.

Armbruster (1999) discusses the local structural environ-ment of the CaO7-like clusters and the placement of theH2O molecules (W) in armenite and the following descrip-tion (i.e., atom lables and bond lengths) follows from his

work. The two crystallographically independent Ca1 andCa2 cations are bonded to six O framework anions with dis-tances between 2.30 and 2.55 A and with one oxygen froman H2O molecule with a distance of 2.471(9) A for Ca1 or2.44(1) A for Ca2. The O14 and O11 anions of the frame-work are slightly underbonded and are thus likely acceptorsof hydrogen bonds from the W1 molecule. They are located2.759 and 2.941 A from W1, respectively. In terms of thesecond H2O molecule, W2, the distances of O16 and O13to W2 are 2.793 and 2.821 A, respectively and they proba-bly act as H-bond acceptors.

The powder IR spectrum of armenite is characterized inthe OH-stretching region by two main bands at approxi-mately 3469 and 3410 cm�1 and in the H2O bending regionby a band at 1645 cm�1. Armbruster (1999) presented sin-gle-crystal polarized IR spectra of armenite and the powderresults herein are in agreement with those of his study,although the wavenumbers for the two main OH bandsare somewhat different in energy than his. Assignmentsfor and an interpretation of the IR bands in the OH-stretch-ing and bending region are not simple matters. Becausethere are two crystallographically distinct H2O molecules,there should be two H2O bending modes and four OH-stretching vibrations. However, only a single H2O bendingmode is observed and two main OH bands. Full modeassignments will probably require IR work at low tempera-tures (e.g., Kolesov and Geiger, 2006) in order to obtain aninterpretable IR spectrum, as well as neutron diffractionstudy to identify the H-atom positions. At low tempera-tures two H2O bending modes may possibly be observedin the IR spectrum and the envelope centered around3450 cm�1 may consist of several closely overlapping OH-stretching bands.

The IR spectra of both armenite and its dehydrated ana-log in the wavenumber region 1200 to 350 cm�1 are similar.The loss of H2O does not lead to major changes in the IR-active phonon spectrum in this energy range. The modesoccurring in the wavenumber region from 1200 to900 cm�1 can be best assigned to internal stretching vibra-tions of SiO4 and AlO4 tetrahedra and they show little var-iation upon loss of H2O. Thus, the framework probablychanges little with dehydration. This interpretation is con-sistent with the X-ray powder results that show little struc-tural difference between the two phases.

The crystal structure of epididymite was most recentlyinvestigated by neutron diffraction by Gatta et al. (2008),who gives all atomic positions including those for the Hatoms. The structure of orthorhombic symmetry Pnma con-tains [Si6O15]1 double chains of corner-linked SiO4 tetrahe-dra that run parallel to [0 0 1]. The chains are linked byBeO4 tetrahedra that are edge-shared with one another.These tetrahedral units build the epididymite “framework”in which micropores are located (Fig. 7c). Here, the “extra-framework” Na cations are found as well as the single crys-tallographic H2O molecule (Gatta et al., 2008, whose atomlabels and bond lengths and angles are cited below). Theoxygen atom of the H2O molecule is bonded to a Na atom,which is further bonded to six oxygen atoms of the “frame-work”. This produces an irregular coordination polyhedronof seven oxygens around Na (Fig. 9d). The H atoms of the

Table 4

Heat capacity, Cp, calorimetric standard third-law entropy, S�, of

armenite, dehydrated armenite, epididymite and dehydrated epi-

didymite at 298.15 K.

Phase Cp (J/mol K) S� (J/mol K)

Armenite 813.5 ± 8.1 795.7 ± 6.2

Dehydrated armenite 749.8 ± 6.1 737.0 ± 6.2

Epididymite 435.4 ± 3.1 425.7 ± 4.1

Dehydrated epididymite 387.5 ± 4.3 372.5 ± 5.0

Fig. 7. Polyhedral models of the armenite and epididymite

structures at 300 K. (a) Armenite, where the red tetrahedra

represent the SiO4 groups and the blue tetrahedra AlO4 groups.

The brown spheres represent the H2O molecules, the yellow spheres

the Ca cations and blue-green spheres the Ba cations. (b) Local

coordination around a Ca cation in armenite. (c) Epididymite,

where the light blue tetrahedra represent the BeO4 groups and the

red tetrahedra SiO4 groups and the yellow spheres are Na cations.

The brown spheres represent oxygen atoms of the H2O molecules

and the small black spheres are hydrogen atoms. (d) Local

coordination around a Na cation in epididymite.

H2O in Microporous Silicates 5209

H2O molecule are hydrogen bonded to framework oxygenatoms as H1. . .O6 = 1.973 A and H2. . .O8 = 2.147 A. TheH2O molecule has bond lengths of Ow–H1 = 0.987 A andOw–H2 = 0.993 A and the H1–Ow–H2 angle is 101.7(3)�.These data allow an assignment of the two OH modes ob-served in the IR spectrum. The higher wavenumber OHmode at 3523 cm�1 is assigned to the Ow–H1 stretchingvibration and the lower wavenumber mode at 3479 cm�1

is assigned to the Ow–H2 stretching vibration. Theseassignments are different than those given in Stavitskayaet al. (1967), which were based on older crystal structuredeterminations that are not fully correct (see Gatta et al.,2008, for a discussion of previous structure determinations).The wavenumber difference between the two OH modes is44 cm�1. This is much less than 99 cm�1, which is the casefor a free H2O molecule (Eisenberg and Kauzmann, 1969).

The IR spectra of epididymite and its dehydrated analogshow differences most notably between 1200 and 900 cm�1,where the internal stretching vibrations of the SiO4 groupsshould occur. The situation is different from that in armen-ite. The reason for the difference in the IR spectra may bedue to a change in space group symmetry as suggested bythe X-ray powder results.

The Ca- and Na-centered clusters in armenite and epi-didymite, respectively, share some similarities with cation-H2O clusters that are present in certain natural zeolites.In this respect, the two small-pore zeolites natrolite,Na16[Al16Si24O80]�16H2O, with extra-framework Na cat-ions and H2O and structurally related scolecite, Ca8[Al16Si

24O80]�24H2O, with extra-framework Ca cations and H2Ocome to mind. The clusters and bonding arrangements inboth zeolites are shown in Kolesov and Geiger (2006). Inepididymite and armenite, as well as in both zeolites, con-fined H2O interacts with either Na or Ca via an ion–dipolearrangement and it is also hydrogen bonded to the respec-tive frameworks. H-bonding in natrolite and scolecite ap-pears to be approximately similar in strength to that inarmenite and stronger than that in epididymite, based onthe average O–H wavenumber obtained from the totalnumber of observed OH bands in the spectra of both struc-tures (Kolesov and Geiger, 2006, see below).

The IR and X-ray measurements made on dehydratedarmenite and epididymite indicate that both phases donot rehydrate on exposure to the atmosphere over severalweeks time. This behavior is different to the general situa-tion for most zeolites and is consistent with the crystalchemistry of armenite and epididymite with their smallmicropores. The change in volume upon dehydration inthe case of armenite is approximately 1%. This value is low-er that the volume change associated with many naturalzeolites upon dehydration (Bish and Carey, 2001).

4.2. Heat capacity and entropy behavior

The Cp data of the various phases can be described as“normal” with Debye T3-like behavior at the lowest tem-peratures (e.g., Gopal, 1966). No phase transitions appearto be present. The heat capacity behavior of confinedH2O in both armenite and epididymite between 0 and300 K, which is obtained by subtracting the fitted Cp values

of the heat-treated samples from the fitted Cp values of thenatural crystals, is shown in Fig. 8. The behavior of theH2O molecules is illustrated and compared, in more detail,in Fig. 9 along with the Cp behavior for ice and super-cooled liquid water and steam. Of the three different H2Ophases, the Cp behavior of H2O in both armenite and epi-didymite is most similar to that of ice. The H2O moleculesin both phases appear to be ordered and they are held inplace via an ion–dipole interaction between the Ca cation

Fig. 8. Heat capacity of natural and heat-treated armenite, filled

and open squares, respectively, and natural and heat-treated

epididymite, filled and open circles, respectively, from 0 to 300 K

using fitted Cp data (Table 2). The � and + symbols show the

difference in Cp between the natural armenite and epididymite and

their heat-treated analogs, respectively. The symbols are larger

than the experimental uncertainty.

Fig. 9. Heat capacity behavior of confined H2O in armenite and

epididymite as well as for hemimorphite (Geiger and Dachs, 2009)

and analcime (Johnson et al., 1982) at 0 K < T < 300 K. The

squares with the + symbol are the Cp of ice (Giauque and Stout,

1936), the squares with the � symbol the Cp of super-cooled liquid

water (Angell et al., 1982) and the circles with the � symbol the Cp

of ideal H2O gas (Eisenberg and Kauzmann, 1969).

5210 C.A. Geiger et al. /Geochimica et Cosmochimica Acta 74 (2010) 5202–5215

and the O atoms of the H2O molecule in the case of armen-ite and Na cation in the case of epididymite and throughhydrogen-bonding between the H2O molecule and oxygenatoms of the respective frameworks. Therefore, the thermo-dynamic behavior of the confined H2O molecules in bothsilicates should be most similar to that of the H2O mole-cules in ice. This is reflected in their respective macroscopicCp behavior.

The heat capacity and entropy of the dehydration reac-tions, DCrxn

p and DSrxn, at 298 K, shown here using the idealmineralogical formulas:

BaCa2Al6Si9O30 � 2H2O ! BaCa2Al6Si9O30

þ 2H2O ðgasÞ ð3Þ

and

Na2Be2Si6O15 �H2O ! Na2Be2Si6O15 þH2O ðgasÞ ð4Þ

can be calculated using the determined heat capacity andentropy values for the natural crystals and their H2O-freeanalogs, along with thermodynamic properties for idealgaseous H2O (Table 5). We obtain DCrxn

p = 3.42 J/mol Kand DSrxn = 319.1 J/mol K for armenite and DCrxn

p =�14.3 J/mol K and DSrxn = 135.7 J/mol K for epididymitedehydration reactions. The entropies of sublimation andevaporation at 298 K for H2O, i.e.,

H2O ðiceÞ ! H2O ðgasÞ ð5Þ

and

H2O ðliquid waterÞ ! H2O ðgasÞ; ð6Þ

are DSsub. = 147.0 J/mol K and DSevap. = 118.94 J/mol K,respectively (using data in Table 5). The entropy associated

with the breaking of the H-bonds for one mole of H2O inarmenite is most similar to that in (5) even being somewhatgreater in value. In the case of epididymite, the entropy ofdehydration is lower with a value between that in (5) and (6).

4.3. The system H2O and confined H2O in various

microporous silicates

In terms of analyzing the thermodynamic, bonding andvibrational behavior of confined molecular H2O in silicates,the pure H2O system can be used as a reference state. Eisen-berg and Kauzmann (1969) discuss the structures, vibra-tional states and thermodynamic properties of gaseous,liquid and solid H2O. A review of the complexities inherentto the low-temperature thermodynamic behavior of theH2O system is given by Mishima and Stanley (1998). Animportant difference between liquid and solid H2O and gas-eous H2O (steam) is the presence of hydrogen-bonding inthe first two phases. Hydrogen-bonding plays a central rolein determining the physical and chemical properties of theH2O system (Eisenberg and Kauzmann, 1969; Jeffrey,1997; Marechal, 2007) and it must be considered in H2Omolecule–silicate interactions. It should be noted that thehydrogen-bond is generally relatively weak with energiesroughly around 20 J/mol compared to typical covalentchemical bonds with energies roughly around 400 J/mol.

4.3.1. Macroscopic-thermodynamic behavior

Geiger and Dachs (2009) analyzed in simple terms thethermodynamic behavior of confined H2O in various micro-porous silicates. They showed that the heat capacity and en-tropy values at 298 K for H2O varies between the different

Table 5

Heat capacity and entropy values for various H2O phases and for one mole of confined H2O in different microporous silicates and silicate

frameworks at 298 K and their H2O stretching wavenumber behavior. Zeolites are given in italics.

Phase Heat

capacity

(J/mol K)

Entropy

(J/

mol K)

H2O stretching

wavenumber(s)

(cm�1)

Avg. H2O stretching

wavenumber

(cm�1)

Reference

Ice (hexagonal)a 40.96 41.89 �3200 (m1/3) �3200 Eisenberg and Kauzmann

(1969)

Liquid waterb 75.35 69.95 3280 (m1) and 3490 (m3) 3385 Eisenberg and Kauzmann

(1969)

Steam (ideal gas)c 33.56 188.89 3657 (m1) and 3756 (m3) 3707 Eisenberg and Kauzmann

(1969)

Armenite 31.9 29.4 3410 and 3469 �3440 This work

Natrolited – 32 �3320 and 3540 �3431 Kolesov and Geiger (2006)

Scolecited – 32 �3230 to 3590 �3408 Kolesov and Geiger (2006)

Paranatrolitee 55.8 45.8 – �3450 Belitsky et al. (1992)

Epididymite 46.9 53.2 3479 and 3523 3504 This work

Mordenitee 54.3 54.1 �3600 to 3400 �3500 Breck (1974)

Hemimorphiteg 38.4 55.2 �3340 to 3650 �3492 (4 K) Kolesov (2006)

Analcimef 47.9 55.0 �3560 to 3675 �3618 Velde and Besson (1981)

Mg-cordieritee 42.2 80.5 3595 (m1) and 3689 (m3) 3642 Paukov et al. (2007)

Cp and S data:a Extrapolated values using Giauque and Stout (1936) data.b Chase (1998).c Eisenberg and Kauzmann (1969).d Johnson et al. (1983) – from calorimetry and model calculations (Cp of the H2O-free analogs was not measured).e Calculated from data in Paukov et al. (2007).f Johnson et al. (1982).g Geiger and Dachs (2009).

H2O in Microporous Silicates 5211

phases. Table 5 lists Cp and S values for the three mostcommon H2O phases and for confined H2O in a number ofmicroporous silicates that have been studied using low-tem-perature calorimetry methods. It also includes the results ofthis study on armenite and epididymite. Cp and S values at298 K vary between 32 and 56 J/mol K and 29 and81 J/mol K, respectively. In terms of the pure H2O phases,the entropy of steam is the greatest and the least for ice.Armenite shows the smallest averaged entropy value ofapproximately 29 J/mol K for its confined H2O molecules.Each H2O molecule in the two small-pore zeolites natroliteand scolecite was estimated to have an entropy value ofapproximately 32 J/mol K (Johnson et al., 1983). These arethe lowest entropy values for confined molecular H2O re-ported to date in microporous zeolites. The entropy of H2Oin paranatrolite is more, namely 45.8 J/mol K, being a littlelarger than that of ice. ConfinedH2O in hemimorphite, anal-cime andmordenite have similar entropy values around 55 J/mol K, which lie between the entropy value for ice and liquidH2O. The entropy of H2O in epididymite is similar with a va-lue of 53.2 J/mol K. Hydrous Mg-cordierite is considerablydifferent in nature from all of these other microporous sili-cates, because its S value of 80.5 J/mol K for H2O is greatereven than that of liquid H2O.

What about Cp behavior as a function of T? Fig. 9 showsCp behavior for confinedH2O in armenite and epididymite at0 K < T < 300 K. Also shown is that for hemimorphite(Geiger and Dachs, 2009) and analcime (Johnson et al.,1982) as well as Cp behavior for ice, liquid water and steam.H2O in armenite and epididymite behaves thermodynami-cally more like ice, as does H2O in hemimorphite and anal-cime, compared to liquid water or steam. Cp behavior forconfined H2O in the various microporous phases does show,of course, differences.At the lowest temperatures,Cp forH2Oin armenite remains very small and then increases signifi-cantly in value at about T > 50 K but remains less than theCp of ice to 300 K. The Cp behavior of H2O in epididymiteis roughly similar to that of ice until aboutT = 30 Kand thenit becomes greater. Confined H2O in all these microporoussilicates behaves considerably differently than H2O in cordi-erite, whose Cp behavior is more like ideal steam between100 K < T < 300 K (see Paukov et al., 2007).

4.3.2. Crystal chemistry and pore size, hydrogen-bonding and

vibrational behavior of the H2O molecule

All the silicate phases in Table 5 can be considered asmicroporous. There are, however, considerable differencesin crystal chemistry and porosity between them and thisleads to variations in the S values for confined H2O at298 K. The low entropy value for H2O of 29 J/mol K inarmenite is consistent with its crystal structure, which isnot microporous in the “classic sense”. Armenite’s H2O re-sides in small pores and is held in place by relatively strongbonds. In the case of epididymite, the entropy value of H2Oof 53.2 J/mol K is considerably larger. This is in line with itslarger pores and the weaker hydrogen-bonding involvingthe H2O molecules.

With the exception of cordierite and hemimorphite, allthe silicate phases in Table 5 contain extra-framework cat-ions in the micropores to which the H2O molecules are

bonded. This complicates quantitative comparisons be-tween the thermodynamic behavior of confined H2O be-tween the various phases. In the case of thosemicroporous silicates having extra-framework cations,upon loss of H2O, there must be some structural relaxation.The extra-framework cations and their surrounding anionsmust adjust locally and will obtain a slightly differentarrangement after loss of H2O. The Ca–O and Na–O bondsare long and relatively weak in armenite and epididymitecompared to those bonds operating in the frameworks.Thus, changes in phonon behavior, beyond that related tothe loss of H2O due to local structural relaxation associatedwith loss of the H2O molecule, should be most appreciableat low energies. This, in turn, will affect Cp behavior atroughly T < 100 K. Some of the “fine structure” observedin the Cp behavior for H2O in the various phases (Fig. 9)may partly reflect small structural relaxations and is not re-lated purely to the behavior of confined H2O. There is, atthis point, no way to analyze for these different effectsand to separate out quantitatively the various contributionsto Cp behavior. In order to better interpret and understandthe experimental Cp data fully, and the thermodynamicbehavior of the confined H2O, lattice dynamic calculationswill be required. At any rate, we think the contributionfrom the confined H2O plays the largest role in affectingCp behavior in Fig. 9.

The entropy of confined H2O at 298 K in microporoussilicates will be mainly a function of the energies of theexternal (i.e., libration and translation) H2O modes. They,in turn, will be controlled by the nature of the bonding ofthe H2O molecules in the micropores. Raman and IR spec-troscopic investigations have been made on most of thephases in Table 5 and they can be analyzed to first orderto describe the hydrogen-bonding of the H2O molecules.Hydrogen-bond strength can be described by the wavenum-ber of the internal stretching vibrations of the H2O mole-cule (Jeffrey, 1997; Marechal, 2007). Studies show, thehigher the wavenumber of the internal stretching vibrations(symmetric stretch – m1 and asymmetric stretch – m3) of theH2O molecule, the weaker the hydrogen-bonding. For afree H2O molecule as in steam, for example, they lie at3657 (m1) and 3756 cm�1 (m3), while in liquid water theyare about 3280 (m1) and 3490 (m3) and they are in ice cen-tered around 3200 cm�1 (Eisenberg and Kauzmann,1969). Geiger and Dachs (2009) analyzed the H-bondingbehavior of the H2O molecules for a number of phases inTable 5 and its relationship to the entropy value of confinedH2O at 298 K. They proposed that the strength of the inter-actions (e.g., H-bonding) between a H2O molecule and itssurroundings increases approximately as:

steam > cordierite > analcime > mordenite

P hemimorphite > natroliteðparanatroliteÞ

� scolecite > liquid H2O > ice: ð7Þ

They suggested, moreover, that the strength of the bondingholding the H2O in place is inversely proportional to the en-tropy value for confined H2O at 298 K.

We carry this analysis further and consider three differ-ent groups of phases in terms of their local-bonding and

5212 C.A. Geiger et al. /Geochimica et Cosmochimica Acta 74 (2010) 5202–5215

macroscopic-thermodynamic properties. They are: (i) thepure H2O phases ice, liquid water and steam, (ii) cordieriteand hemimorphite that contain only H2O in microporesand no extra-framework cation and (iii) the zeolites, as wellas armenite and epididymite, that have H2O that is bondedto extra-framework alkali or alkaline-earth cations inmicropores. In Fig. 10 are plotted the S values for H2Oat 298 K for the different phases as a function of the aver-age wavenumber obtained from the OH-stretching vibra-tions of the H2O molecules, as taken from published IRand Raman measurements (values and references given inTable 5). A couple of observations are in order. For allthree groups of phases, the entropy of H2O increasesapproximately linearly with increasing average wavenum-ber of the internal OH-stretching vibrations. The interpre-tation is that decreased hydrogen-bonding strengthbetween a confined H2O molecule in a pore and the silicateframework (or between H2O molecules in the case of thepure H2O phases) results in increased entropy for H2O.The data can be interpreted further. There appears to bean increase in entropy for H2O at constant hydrogen-bondstrength in going from type (iii) microporous silicates totype (ii) microporous silicates to (i) the pure H2O phases.This observation is, once again, consistent with the ex-pected vibrational behavior of the H2O molecules in thethree groups of phases, whereby external molecular vibra-tional freedom increases going from groups (iii) to (ii) to(i). This arises because increased H-bonding strength andthe ion–dipole interaction results in a dampening of theexternal molecular H2O vibrations (i.e., hindered transla-tions and librations).

The vibrational entropy of an H2O molecule can beapproximated as the sum of various contributions as:

SH2O ¼ STðH2OÞ þ SRðH2OÞ þ SIðH2OÞ; ð8Þ

where T = translational, R = rotational/librational andI = internal vibrations. At T < 300 K, SIðH2OÞ can be ne-glected and the T(H2O) modes and R(H2O) modes willdetermine SH2O behavior. Little is known about the vibra-tional behavior of confined H2O in microporous silicatesin terms of its hindered translations and librations. It isknown, however, in general, that large amplitude vibrationsare associated with smaller energy quantum excitations,whereas smaller amplitude vibrations have larger energies.Thus, because the amplitude of external H2O vibrationsshould increase in going from groups (iii) to (ii) to (i), morequantum vibrations will be populated at a given T and heatcapacity and entropy increases accordingly. This is the linkbetween molecular behavior and macroscopic-thermody-namic properties.

The bonding situation in ice is not so different from thatacting in many microporous silicates. In the case where ex-tra-framework cations are present they act like H-bond do-nors, because they attract the lone-pair electrons of theH2O molecules in a similar manner to the H atoms in iceor liquid water. With increasing charge on the extra-frame-work cations, deshielding of the H atoms of the H2O mol-ecule will increase and the H-bonds between the frameworkand the H2O molecule will become shorter (see Jeffrey,1997). Thereby, the amplitudes of translation and libra-tional motions of the H2O molecule should decrease, asshould the entropy.

There are differences, however, between the vibrationalbehavior of H2O in ice and liquid water compared to thatof H2O in microporous silicates. This is clearly shown inthe IR spectra of the various phases. Ice, and especially li-quid water, shows broad bands for both internal and exter-nal vibrations (Eisenberg and Kauzmann, 1969). Internaland external modes of H2Omolecules in small-pore silicates(i.e., armenite, epididymite, cordierite, natrolite, etc.) ap-pear to have narrower bandwidths. This probably reflects,in part, the more harmonic character of H2O vibrationsin the case of microporous silicates. This is probably relatedto their more ordered local environments, especially com-pared to the situation in liquid water.

Consider the external mode behavior of confined H2O inthose few microporous phases that have been investigatedexperimentally. There is inelastic neutron scattering workon cordierite (Winkler and Hennion, 1994) and several zeo-lites, including natrolite and scolecite (Line and Kearley,2000), in this direction. In addition, IR/Raman spectrahave been used to determine the mode energies of externalH2O vibrations in scolecite and natrolite (Kolesov and Gei-ger, 2006), hemimorphite (Kolesov, 2006) and hydrous ber-yl (Kolesov and Geiger, 2000), the latter of which should besimilar to the situation in cordierite. In the case of cordier-ite, energies of the external modes have also been estimatedby applying Einstein theory to low-temperature Cp data(Paukov et al., 2007). For cordierite, T(H2O) modes wereobserved at 6 meV (48 cm�1) and 10 meV (81 cm�1) and alibrational/rotational mode occurs around 27 meV

Fig. 10. Relationship between the entropy of one mole of H2O at

298 K and the average wavenumber of the OH/H2O stretching

vibrations for various H2O-containing phases and groups (see text).

The circles with a dot represent data from microporous silicates in

which the H2O is hydrogen bonded to the framework and also

bonded to an extra-framework cation (ion–dipole interaction). The

squares with crosses are for the three pure H2O phases (liquid

water, ice and steam). The crosses are for hemimorphite and

cordierite in which the H2O molecule is not bonded to an extra-

framework cation. The various lines represent linear fits to the data

of the three groups of phases.

H2O in Microporous Silicates 5213

(218 cm�1) according to Winkler and Hennion (1994). TheCp data can be fit with three translational modes occurringat 45, 70 and 116 K (31, 49 and 81 cm�1) and two libra-tional modes at 255 and 313 K (177 and 218 cm�1) follow-ing Paukov et al. (2007). Line and Kearley (2000) reportcation-H2O stretch and H-bond stretching vibrations fornatrolite between 104 and 210 cm�1 and H2O librations be-tween 516 and 701 cm�1. In scolecite the T(H2O) modes areestimated to lie between 45 and 178 cm�1 and one T(H2O)mode in natrolite lies around 145 cm�1 based on Ramanspectra (Kolesov and Geiger, 2006). In hemimorphite theyappear roughly similar in energy as they occur between 30and 148 cm�1 (Kolesov, 2006). In the case of beryl (cordier-ite) H2O librations lie approximately around 200 cm�1

(Kolesov and Geiger, 2000) and for natrolite/scolecite libra-tional modes may lie between 430 and 670 cm�1 (Kolesovand Geiger, 2006). Following this, one can expect that theT(H2O) and R(H2O) modes for armenite and epididymiteare more similar to those of the zeolites than those of cor-dierite or beryl. These modes are estimated to lie roughlybetween 30 and 200 cm�1 and 400 and 700 cm�1, respec-tively. Those of epididymite will be at lower energies thanthose of armenite. Considerable work needs to be done tomeasure directly the vibrational properties of the externalH2O modes and to determine quantitatively how they arerelated to macroscopic Cp behavior.

ACKNOWLEDGMENTS

The armenite sample was provided by T. Armbruster (Bern,

Switzerland), obtained through Andres Martin who collected it,

and the epididymite sample by A. Guastoni (Padova, Italy).

N.-S. Rahmoun (Kiel, Germany) helped record the IR spectra. Re-

views by E. Libowitzky, L. Glasser and E.H. Oelkers helped im-

proved the style of presentation. This work was financed by the

Austrian Science Fund project number P20210-N10, which is grate-

fully acknowledged. The authors thank them all.

APPENDIX A. SUPPLEMENTARY DATA

Supplementary data associated with this article can befound, in the online version, at doi:10.1016/j.gca.2010.05.033.

REFERENCES

Angell C. A., Oguni M. and Sichina W. J. (1982) Heat capacity of

water at extremes of supercooling and superheating. J. Phys.

Chem. 86, 998–1002.

Armbruster T. (1999) Si, Al ordering in the double-ring silicate

armenite, BaCa2Al6Si9O30�2H2O: a single-crystal X-ray and 29Si

MAS NMR study. Am. Mineral. 84, 92–101.

Armbruster T. and Czank M. (1992) H2O ordering and super-

structures in armenite BaCa2Al6Si9O30�2H2O: a single-crystal

X-ray and TEM study. Am. Mineral. 77, 422–430.

Balassone G., Boni M., Di Maio G. and Franco E. (1989) Armenite

in southwest Sardinia: first recorded occurrence in Italy. Neues

Jahrb. Mineral. Monatsh. 1989, 49–58.

Belitsky I. A., Fursenko B. A., Gabuda S. P., Kholdeev O. V. and

Seryotkin Yu. V. (1992) Structural transformations in natrolite

and edingtonite. Phys. Chem. Miner. 18, 497–505.

Benisek A. and Dachs E. (2008) The uncertainty in determining the

third law entropy by the heat-pulse calorimetric technique.

Cryogen 48, 527–529.

Bish D. L. and Carey J. W. (2001) Thermal behavior of natural

zeolites. In Natural Zeolites: Occurrence, Properties, Applica-

tions, vol. 45 (eds. D. L. Bish and D. W. Ming). Mineralogical

Society of America, Reviews in Mineralogy and Geochemistry,

Washington, DC, pp. 403–452.

Bish D. L. and Ming D. W. (Eds.) (2001) Natural Zeolites:

Occurrence, Properties, Applications. Reviews in Mineralogy

and Geochemistry, vol. 45. Mineralogical Society of America

and the Geochemical Society, 662p.

Boerio-Goates J., Stevens R., Hom B. K., Woodfield B. F.,

Piccione P. M., Davis M. E. and Navrotsky A. (2002) Heat

capacities, third-law entropies and thermodynamic functions of

SiO2 molecular sieves from T = 0 to 400 K. J. Chem. Thermo-

dyn. 34, 205–227.

Brøgger W. C. (1890) Die Mineralien der Sienitpegmatitgange der

sudnorwegischen Augit- und Nephelinsyenite. Zeit. Kristall. 16,

586–597.

Breck D. W. (1974) Zeolite Molecular Sieves. Structure, Chemistry

and Use. Wiley-Interscience Publication.

Cerny P. (1963) Epididymite and milarite-alteration products of

beryl from Vezna, Czechoslovakia. Mineral. Mag. 33, 450–

457.

Chase, Jr., M.W. (1998) NIST-JANAF thermochemical tables, J.

Phys. Chem. Ref. Data. Monogr. No. 9, 4th ed. Parts I and II.

Dachs E. and Bertoldi C. (2005) Precision and accuracy of the heat-

pulse calorimetric technique: low-temperature heat capacities of

milligram-sized synthetic mineral samples. Eur. J. Mineral. 17,

251–261.

Dachs E. and Geiger C. A. (2006) Heat capacities and entropies of

mixing of pyrope-grossular (Mg3Al2Si3O12–Ca3Al2Si3O12) gar-

net solid solutions: a low-temperature calorimetric and a

thermodynamic investigation. Am. Mineral. 91, 894–906.

Eisenberg D. and Kauzmann W. (1969) The Structure and

Properties of Water. Oxford Univ. Press.

Flink G. (1894) Beschreibung eines neuen Mineralfundes aus

Gronland. Epididymit. Zeit. Kristall. 23, 353–358.

Fortey N. J., Nancarrow P. H. A. and Gallagher M. J. (1991)

Armenite from the Middle Dalradian of Scotland. Mineral.

Mag. 55, 135–138.

Geiger C. A. and Dachs E. (2009) Quasi-ice-like Cp behavior of

molecular H2O in hemimorphite Zn4Si2O7(OH)2�H2O: Cp and

entropy of confined H2O in microporous silicates. Am. Mineral.

94, 634–637.

Gatta G. D., Rotiroti N., McIntyre G. J., Guastoni A. and Nestola

F. (2008) New insights into the crystal chemistry of epididymite

and eudidymite from Malosa, Malawi: a single-crystal neutron

diffraction study. Am. Mineral. 93, 1158–1165.

Giauque W. F. and Stout J. W. (1936) The entropy of water and

the third law of thermodynamics: the heat capacity of ice from

15 to 273 K. J. Am. Chem. Soc. 58, 1144–1150.

Gopal E. S. R. (1966) Specific Heats at Low Temperatures. Plenum

Press.

Hawthorne F. C. (1992) The role of OH and H2O in oxide and

oxysalt minerals. Zeit. Kristall. 201, 183–206.

Hawthorne F. C., Kimata M., Cerny P., Ball N., Rossman G. and

Grice J. D. (1991) The crystal chemistry of the milarite-group

minerals. Am. Mineral. 76, 1836–1856.

Jeffrey G. A. (1997) An Introduction to Hydrogen Bonding. Oxford

Univ. Press.

Johnson G. K., Flotow H. E., O’Hare P. A. G. and Wise W. S.

(1982) Thermodynamic studies of zeolites: analcime and dehy-

drated analcime. Am. Mineral. 67, 736–748.

5214 C.A. Geiger et al. /Geochimica et Cosmochimica Acta 74 (2010) 5202–5215

Johnson G. K., Flotow H. E., O’Hare P. A. G. and Wise W. S.

(1983) Thermodynamic studies of zeolites: natrolite, mesolite

and scolecite. Am. Mineral. 68, 1134–1145.

Johnson G. K., Flotow H. E., O’Hare P. A. G. and Wise W. S.

(1985) Thermodynamic studies of zeolites: heulandite. Am.

Mineral. 70, 1065–1071.

Johnson G. K., Tasker I. R., Flotow H. E., O’Hare P. A. G. and

Wise W. S. (1992) Thermodynamic studies of mordenite,

dehydrated mordenite, and gibbsite. Am. Mineral. 77, 85–93.

Kolesov B. (2006) Raman investigation of H2O molecule and

hydroxyl groups in the channels of hemimorphite. Am. Mineral.

91, 1355–1362.

Kolesov B. A. and Geiger C. A. (2000) A single-crystal Raman

study of the orientation and vibrational states of molecular

water in synthetic beryl. Phys. Chem. Mineral. 27, 557–564.

Kolesov B. A. and Geiger C. A. (2006) Behavior of H2O molecules

in the channels of natrolite and scolecite: a Raman and IR

spectroscopic investigation of hydrous microporous silicates.

Am. Mineral. 91, 1039–1048.

Kvick A. (1986) Hydrogen bonding in zeolites. Trans. Am. Crystal.

Assoc. 22, 97–105.

Larson A. C. and Von Dreele R. B. (1988) General Structure

Analysis System (GSAS). Los Alamos National Laboratory

Report LAUR 86-748, Los Alamos, New Mexico.

Levy C. M. (1961) Remarques sur l’epididymite. Bull. Soc. Fr.

Mineral. 84, 320.

Line C. M. B. and Kearley G. J. (2000) An inelastic incoherent

neutron scattering study of water in small-pored zeolites and

other water-bearing minerals. J. Chem. Phys. 112, 9058–9067.

Mandarino J. A. and Harris D. C. (1968) Epididymite from Mont

St. Hilaire, Quebec. Can. Mineral. 9, 706–709.

Marechal Y. (2007) The Hydrogen Bond and the Water Molecule.

Elsevier.

Mason B. (1987) Armenite from Broken Hill, Australia, with

comments on calciocelsian and barium anorthite. Mineral.

Mag. 51, 317–318.

McCusker L. B., Liebau F. and Engelhardt G. (2001) Nomencla-

ture of structural and compositional characteristics of ordered

microporous and mesoporous materials with inorganic hosts.

Pure Appl. Chem. 73, 381–394.

Mishima O. and Stanley H. E. (1998) The relationship between

liquid, supercooled and glassy water. Nature 396, 329–335.

Neumann H. (1941) Armenite, a water-bearing barium–calcium–

alumosilicate. Nor. Geol. Tidsskr. 21, 19–24.

Paukov I. E., Moroz N. K., Kovalevskaya Yu. A. and Belitsky I.

A. (2002) Low-temperature thermodynamic properties of dis-

ordered zeolites of the natrolite group. Phy. Chem. Mineral. 29,

300–306.

Paukov I. E., Kovalevskaya Yu. A., Rahmoun N.-S. and Geiger C.

A. (2007) The heat capacity of hydrous cordierite at low

temperatures: the thermodynamic behavior of the H2O mole-

cule in selected hydrous micro and nanoporous silicates. Am.

Mineral. 92, 388–396.

Petersen O. V. (1966) Crossed axial plane dispersion in epididy-

mite. Am. Mineral. 51, 916–919.

Pouliot G., Trudel P., Valiquette G. and Samson P. (1984)

Armenite-thulite-albite veins at Remigny, Quebec: the second

occurrence of armenite. Can. Mineral. 22, 453–464.

Semenenko N. P., Litvin A. L., Sharkin O. P., Boiko V. L.,

Egorova L. N., Shurdin G. S., Terets G. Y., Savitskaya A. B.

and Ilovaiskaya S. V. (1987) Armenite from northern Dneiper

region. Mineral. Zh. 9, 83–90 (in Russian).

Senn T. (1990) Ba-Gneise and Ba-Kristallisation im Gebiet der

Wasenalp VS. Schweiz. Mineral. Petrogr. Mitt. 70, 167.

Shilin L. L. and Semenov E. I. (1957) Beryllium minerals,

epididymite and eudidymite, in alkali pegmatites of the Kola

Peninsula. Dokl. Akad. Nauk SSSR 112, 325–328.

Stavitskaya G. P., Ryskin Y. I., Kol’tsov A. I., Shul’gi E. I. and

Pobedonostseva A. A. (1967) Study of the nature of the

hydration of epididymite by means of the infrared absorption

spectrum and proton magnetic resonance. J. Struct. Chem. 8,

201–205.

Velde B. and Besson J. M. (1981) Raman spectra of analcime under

pressure. Phys. Chem. Mineral. 7, 96–99.

Winkler B. and Hennion B. (1994) Low temperature dynamics of

molecular H2O in bassanite, gypsum and cordierite investigated

by high resolution incoherent inelastic neutron scattering. Phys.

Chem. Mineral. 21, 539–545.

Zak L. and Obst P. (1989) Armenite-feldspar veins in basic volcanic

rocks from Chvaletice (Czechoslovakia). Casopsis Mineral.

Geolog. 34, 337–351.

Associate editor: Eric H. Oelkers

H2O in Microporous Silicates 5215