The Influence of Boron on the Crystal Structure and Properties of Mullite

Home , Mullite

THE INFLUENCE OF BORON

ON THE CRYSTAL STRUCTURE AND

PROPERTIES OF MULLITE

Investigations at Ambient, High-Pressure, and High-Temperature Conditions

Dissertation zur Erlangung des Doktorgrades der Naturwissenschaften (Dr. rer. nat.)

am Fachbereich Geowissenschaften

der Universität Bremen

vorgelegt von

Hanna Lührs

Bremen, September 2013

Reviewer: Prof. Dr. Reinhard X. Fischer Prof. Dr. Thorsten M. Gesing

Date of public colloquium: 21 November 2013

This cumulative thesis is the outcome of the ZF project 05/104/08 of the Central Research Development Fund of the University of Bremen. The studies compiled in this thesis were carried out from October 2009 until September 2013 at the Faculty of Geosciences, University of Bremen. Additional experiments were carried out at the Forschungs-Neutronenquelle Heinz Maier-Leibnitz (FRM II, Garching) and within the Materials Solid State NMR group in the department of Physics at the University of Warwick. High-pressure synchrotron X-ray diffraction data was provided by K. Lipinska and P.E. Kalita from the University of Nevada, Las Vegas (UNLV).

Contents

Contents ...... V

Abstractt ...... 1

Kurzfassung ...... 3

1 Introduction ...... 5

1.1 Synthesis and applications of mullite Al4+2xSi2-2xO10-x (x = 0.2 … 0.9) ...... 7 1.2 Crystal structure of mullite and definiittion of mullite-type compounds ...... 9 1.2.1 Crystal structure of mullite ...... 9 1.2.2 Definition of mullite-type compounds ...... 10 1.2.3 Foreign cation incorporation in mullite ...... 11 1.3 Boron-mullites ...... 12 1.4 Boron ...... 16 1.4.1 Geological occurrence and applications of boron ...... 16 1.4.2 Crystal chemistry of boron ...... 17 1.4.3 Chemical analyses of boron ...... 18

2 Scope and objectives, thesis outline ...... 19

2.1 Scope and objectives ...... 19 2.2 Thesis outline ...... 20

3 Material and methods ...... 23

3.1 Syntheses in the system Al2O3 - SiO2 - B2O3 ...... 23 3.2 Analytical methods ...... 24 3.2.1 Powder diffraction experiments ...... 24 3.2.2 Differential thermal analyses (DTA) ...... 25 3.2.3 Infrared spectroscopy (IR) ...... 25 3.2.4 Magic angle spinning nuclear magnetic resonance spectroscopy (MAS NMR) ...... 25 3.2.5 Chemical analyses ...... 25 3.2.6 Computer programs used ...... 26

4 Boron mullite: Formation and basic characterization ...... 27

4.1 Introduction ...... 28 4.2 Synthesis, sample preparation ...... 29 4.3 Experimentaal ...... 30 4.3.1 Scanning electron microscopy and energy dispersive X-ray spectroscopy ...... 30 4.3.2 Powder X-ray diffraction at ambient temperature ...... 30 4.3.3 In situ high temperature X-ray diffraction ...... 31

V Contents

4.3.4 Rietveld refinements...... 31 4.3.5 X-ray fluorescence ...... 31 4.3.6 Infrared spectroscopy ...... 31 4.3.7 Thermal analyses ...... 32 4.4 Results and discussion ...... 32 4.4.1 Boron-mullite formation from sol-gel derived precursors ...... 32 4.4.2 Phase formation ...... 35 4.4.3 Lattice parameters, chemical commposition ...... 37 4.4.4 Thermal stability of B-mullites ...... 41 4.4.5 Thermal expansion ...... 42 4.5 Conclusion ...... 44

5 Neutron diffraction and ¹¹B solid state NMR studies of the crystal structure of B-doped mullite ..... 47

5.1 Introduction ...... 48 5.2 Experimentaal ...... 49 5.2.1 Synthesis of B-doped mullite ...... 49 5.2.2 Neutron powder diffraction and structure refinements, difference Fourier synthesis, grid search 50 5.2.3 NMR spectroscopy ...... 52 5.3 Results ...... 52 5.3.1 11B MAS NMR spectroscopy ...... 52 5.3.2 Neutron diffraction ...... 54 5.4 Discussion ...... 59 5.5 Conclusion ...... 63

6 Ambient and high-pressure synthesis, composition, and crystal structure of B-mullites ...... 65

6.1 Introduction ...... 66 6.2 Experimentaal ...... 67 6.2.1 Sample preparation ...... 67 6.2.2 Powder X-ray diffraction (XRD) ...... 68 6.2.3 Neutron diffraction ...... 68 6.2.4 Prompt gamma activation analyses (PGAA) ...... 68 6.2.5 11B MAS NMR ...... 68 6.2.6 Distance least squares ...... 69 6.3 Results and discussion ...... 69 6.3.1 Chemical analyses of B-mullites – B-mullites with increasing B-content ...... 69 6.3.2 B-rich high pressure B-mullite ...... 72 6.4 Conclusion ...... 77

7 Chemical composition of B-mullites ...... 79

7.1 Materials and methods ...... 79 7.2 Results and discussion ...... 79

8 Crystal chemistry of mullite and B-mullite aat high pressure ...... 81

8.1 Introduction ...... 81

VI Contents

8.2 Material and methods ...... 82 8.2.1 Syntheses and samples ...... 82 8.2.2 Laboratory powder X-ray diffraction ...... 82 8.2.3 High-pressure synchrotron X-ray diffraction using a Diamond Anvil Cell (DAC) ...... 83 8.2.4 Rietveld refinements...... 84 8.3 Results and discussion ...... 84 8.3.1 Pressure dependent lattice parameters...... 85 8.3.2 Crystal chemical changes in mullite and B-mullite upon pressure ...... 89 8.4 Summary, conclusion, and outlook ...... 93

9 Conclusion and future perspectives ...... 95

9.1 Objectives of this thesis ...... 95 9.2 Additional investigations related to thhis thesis ...... 97 9.3 Future perspectives ...... 98

10References ...... 99

Acknowlledgements ...... 107

Appendix ...... 109

Appendix A ...... 110 Appendix B ...... 111 Appendix C ...... 112 Supplementary CD ...... 115

Erklärunng ...... 117

VII

Abstract

Mullite is one of the most important synthetic compounds for advanced structural and functional ceramic materials. The crystal structure of mullite with the composition Al2[Al2+2xSi2-2x]O10-x can incorporate a large variety of foreign cations, including (amongst others) significant amounts of boron. However, no chemical or crystal structure analyses of boron-mullites (B-mullites) were available prior to this work, thus representing the key aspects of this thesis. Furthermore, the influence of boron on selected properties of mullite under ambient, high-temperature, and high-pressure conditions are addressed. Starting from a 3:2 mullite composition (Al4.5Si1.5O9.75), the initial hypothesis for this study was a 1:1 isomorphous replacement of silicon by boron according to the coupled substitution mechanism: 2 Si4+ + O2- o 2 B3+ + Ƒ. Based on a series of compounds synthesized from sol-gel derived precursors at ambient pressure and 1200°C, the formation conditions and physical properties of B-mullites were investigated. The formation temperature for B-mullites decreases with increasing boron-content, as revealed by thermal analyses. An anisotropic development of lattice parameters is observed: Whereas lattice parameters a and b only exhibit minor changes, a linear relationship between lattice parameter c and the amount of boron in the crystal structure was established, on the basis of prompt gamma activation analyses (PGAA) and Rietveld refinements. According to this relationship about 15% of the silicon in mullite can be replaced by boron yielding single-phase B-mullite. B-mullites with significantly higher (~ factor 3) boron-contents in the mullite structure were also observed but the respective samples contain alumina impurities. Fundamental new details regarding the response of B-mullite to high-temperature and high- pressure are presented in this thesis. On the one hand, long-term thermal stability at 800°C was proved for B-mullite, whereas on the other hand, complete decomposition into boron-free mullite and corundum is observed at 1400°C. Furthermore, the incorporation of boron into the crystal structure reduces the mean metric thermal expansion coefficient by 15% in comparison to boron-free mullite. Such a reduction by chemical substitution makes B-mullites a potential candidate for technical applications in the temperature range below 1000°C. Boron incorporation is associated with the formation of additional oxygen vacancies which reduces the mechanical stability of the mullite structure at high-pressure. Moreover, a slight increase of the overall (volume) compressibility of B-mullite compared to boron-free mullite is observed. The compressibility in mullite is anisotropic with the a-axis being the most and the c-axis being the least compressible one. The increasing

1 Abstract divergence with pressure between the compressibilities in a- and b-direction can be explained by a rotation of the octahedra and the increasing inclination angle Ȧ. One major outcome of this thesis is the crystal structure of B-mullite, synthesized at 1200°C and ambient pressure. The refinements in space group Pbam based on neutron diffraction and 11B MAS NMR data clearly confirm the suggested silicon–boron substitution mechanism and yield a composition of Al4.64Si1.16B0.2O9.58. Boron resides in planar BO3 groups crosslinking the mullite-type

AlO4 octahedral chains perpendicular to the c-axis. The position and the intrinsic rigidity of the BO3 group imposes local distortion of the AlO6 octahedra. As a consequence split positions of the oxygen atoms are required in the first coordination sphere of boron, which in turn lead to significantly shortened oxygen-oxygen distances in c-direction and only minor shortenings in the a- and b-directions. Herewith, the crystallographic model provides an explanation for the anisotropic behavior of lattice parameters upon boron-incorporation described above. Single-phase B-mullite with 40% replacement of silicon by boron was synthesized at 10 kbar and 875°C representing a marked increase in boron-content compared to the B-mullites synthesized at ambient pressure and 1200°C. The composition Al4.5Si0.9B0.6O9.4 was derived from refinements based on X-ray diffraction data in combination with the established silicon-boron substitution mechanism. Besides the three-coordinated boron, the chemical shifts in the 11B MAS NMR spectra clearly resolve additional replacement of some aluminum in the AlO4 tetrahedra by boron which is in good agreement with the PGAA results.

Kurzfassung

Synthetischer Mullit mit der Zusammensetzung Al2[Al2+2xSi2-2x]O10-x ist einer der wichtigsten Bestandteile von modernen Funktions- und Strukturkeramiken. Mullit hat die Eigenschaft, eine Vielzahl verschiedener Fremdkationen in die Kristallstruktur einbauen zu können, unter anderem Bor. Da bisher weder chemische Analysen noch ein Kristallstrukturmodell für Bor-Mullit vorlagen, bilden diese beiden Aspekte den Schwerpunkt dieser Arbeit. Des Weiteren wurde der Einfluss von Bor auf ausgewählte Eigenschaften von Mullit unter Umgebungsbedingungen, bei hoher Temperatur sowie unter hohem Druck untersucht. Die Hypothese für diese Arbeit war der 1:1 Ersatz von Silizium durch

Bor ausgehend von 3:2 Mullit (Al4.5Si1.5O9.75), entsprechend der gekoppelten Substitution: 2 Si4+ + O2- o 2 B3+ + Ƒ. Die Bildungsbedingungen und die physikalischen Eigenschaften von Bor-Mulliten wurden anhand von einer Serie mittels sol-gel Synthese bei 1200°C und Umgebungsdruck hergestellten Proben untersucht. Mit zunehmendem Borgehalt nimmt die Bildungstemperatur für Bor-Mullit ab, was durch thermische Analysen nachgewiesen wurde. Die Gitterparameter zeigen ein anisotropes Verhalten: Während die Gitterparameter a und b lediglich geringfügige Änderungen aufweisen, wurde ein linearer Zusammenhang zwischen Gitterparameter c und dem Borgehalt mittels Prompter Gamma Aktivierungs-Analyse (PGAA) und Rietveldverfeinerung nachgewiesen. Gemäß dieser linearen Beziehung kann einphasiger Bor-Mullit mit einem Ersatz von bis zu ca. 15% des Siliziums im Mullit durch Bor hergestellt werden. Bor-Mullite mit wesentlich höheren (~ Faktor 3) Borgehalten in der Mullitstruktur wurden ebenfalls beobachtet, diese Proben enthalten allerdings Aluminiumoxid als Nebenphase. In dieser Arbeit werden erstmals grundlegende Details zum Verhalten von Bor-Mulliten unter erhöhten Druck- und Temperaturbedingungen vorgestellt. Zwar zersetzt sich Bor-Mullit durch längeres Erhitzen bei 1400°C in borfreien Mullit und Korund, bei 800°C jedoch konnte die Langzeitstabilität von borhaltigem Mullit nachgewiesen werden. Weiterhin wird durch den Einbau von Bor in die Mullitstruktur der mittlere thermische Ausdehnungskoeffizient um 15% im Vergleich zu borfreiem Mullit reduziert. Eine so starke Abnahme der thermischen Ausdehnung aufgrund von chemischer Substitution macht Bor-Mullite sehr interessant für technische Anwendungen im Temperaturbereich unterhalb von 1000°C. Durch den Einbau von Bor und der damit verbundenen Zunahme der Sauerstoffleerstellen wird die mechanische Stabilität von Mullit bei hohem Druck verringert und die Kompressibilität im Vergleich zu borfreiem Mullit leicht erniedrigt. Die Kompressibilität im Mullit ist anisotrop, wobei die a-Achse die größte und die c-Achse die geringste

3 Kurzfassung

Kompressibilität aufweist. Der zunehmende Unterschied zwischen den Kompressibilitäten in a- und b- Richtung kann durch eine Drehung der Oktaeder und die Zunahme des Inklinationswinkels Ȧ erklärt werden. Ein weiteres wichtiges Ergebnis dieser Arbeit ist die Kristallstruktur von Bor-Mullit, hergestellt bei 1200°C unter Umgebungsdruck, in der Raumgruppe Pbam. Basierend auf Neutronenbeugungs- und 11B MAS NMR Daten wurde der angenommene Substitutionsmechanismus bestätigt und die folgende chemische Zusammensetzung bestimmt: Al4.64Si1.16B0.2O9.58. Bor befindet sich in planaren

BO3-Gruppen, welche die Mullit-typischen AlO4 Oktaederketten senkrecht zur c-Achse verbinden.

Aufgrund ihrer großen Steifigkeit führt der Einbau von BO3 Gruppen zu lokalen Verzerrungen der

AlO6 Oktaeder, infolge dessen eine Aufspaltung der Sauerstoffpositionen in der ersten Koordinationssphäre von Bor stattfindet. Diese Aufspaltung wiederum hat eine wesentliche Verkürzung der Sauerstoff-Sauerstoff-Abstände in c-Richtung zur Folge; im Vergleich dazu ist die Abnahme der Abstände in a- und b-Richtung gering. Hiermit liefert das Kristallstrukturmodell eine Erklärung für das oben erwähnte anisotrope Verhalten der Gitterparameter mit zunehmendem Boreinbau. Einphasiger Bor-Mullit mit einem Ersatz von 40% des Siliziums durch Bor wurde bei 875°C und 10 kbar hergestellt, dies stellt eine deutliche Zunahme des Borgehaltes gegenüber den Synthesen bei Umgebungsdruck und 1200°C dar. Die Zusammensetzung von Al4.5Si0.9B0.6O9.4 ergibt sich aus der Strukturverfeinerung auf Basis von Röntgenbeugungsdaten in Kombination mit dem entwickelten 11 Substitutionsmodell. Neben den BO3 Gruppen bestätigt die chemische Verschiebung im B MAS NMR Spektrum den zusätzlichen Ersatz von tetraedrisch koordiniertem Aluminium durch Bor.

1 Introduction

The aluminosilicate mullite with the chemical composition of Al4+2xSi2-2xO10-x (x = 0.2 …0.9) is one of the most prominent ceramic materials and probably the most frequent phase in “conventional” ceramics like pottery, porcelains, sanitary ceramics, refractories, building bricks, pipes, and tiles. In china, ceramic manufacturing looks back on more than 11000 years of history (Kerr and Wood, 2004). The first porcelain produced from kaolin at a firing temperature of about 1300°C is dated back to about 620 AD. After the industrial revolution in the nineteenth century mullite-bearing refractories and technical porcelains with a wide variety of applications have gained worldwide importance (Schneider, 2005a and references therein). Due to its outstanding properties such as low thermal expansion, low thermal conductivity, excellent creep resistance, high temperature strength and good resistance against chemical attacks, mullite has become a very important material for advanced structural and functional ceramics (see chapter 1.1). This is emphasized by the constantly increasing number of scientific studies that were published on mullite-related topics since the 1960’s (Fig. 1-1).

Fig. 1-1: Number of publications on mullite-related topics between 1960 and 2012 (Web of Science, topic=mullite, July 2013).

The annual world production of synthetic mullite as of 2004 is approximately 235000 t (Kogel et al., 2006). In contrast to its great industrial importance the natural mineral mullite is very rare. The type locality is the Isle of Mull in NW Scotland where mullite occurs in fused argillaceous inclusions in tertiary eruptive rocks in association with corundum and feldspar. In other localities mullite is known to be associated with sillimanite and kyanite (Val Sissone, Italy) or at Sithean Sluaigh

5 1 | Introduction

(Scotland) with magnetite, spinel, pseudobrookite, sanidine, and cordierite (Anthony et al., 2003). Mullite crystals form prismatic needles in contact metamorphosed clay-sandstones of the sanidinite facies (Tröger, 1982; Matthes, 2001). An example of natural mullite in a druse of a volcanic rock from the Bellerberg (Eifel district, Germany) is given in Fig. 1-2.

Fig. 1-2: Natural mullite from Bellerberg, Ettringen, Eifel district, Germany.

Mullite was first mentioned as a new mineral in the “New Minerals” section of the American Mineralogist by Foshag (1924) referring to Bowen et al. (1924). In fact, mullite has been observed as early as 1847 in the glass phase of porcelains but was mistaken for sillimanite (Al2O3 · SiO2); in the th late 19 century compounds with compositions close to 3Al2O3 · 2SiO2 were reported but also designated as sillimanite (Pask, 1990 and references therein). Until 1924 sillimanite was believed to be the only stable compound within the Al2O3 - SiO2 system. Bowen and Greig (1924) revised the

Al2O3 - SiO2 phase diagram pointing out that 3Al2O3 · 2SiO2 is the only stable compound and that crystals of this phase are common constituents of all alumina-silica refractories. In a footnote they propose the name “mullite” for the natural analogous of this phase that was discovered in rocks from the Island of Mull, Scotland. In their review on the “Structure and properties of mullite” Schneider et al. (2008) give four reasons for the outstanding scientific and technical importance of mullite: 1. Its favorable properties like the high thermal stability with a melting point at 1890°C (Klug et al., 1987), the low thermal expansion coefficient (~4.5*10-6 °C-1), low thermal conductivity, high creep resistance, and corrosion stability in harsh chemical environments, combined with suitable strength and fracture toughness. 2. The starting materials (see chapter 1.1) are available in big quantities on earth.

3. The ability of mullite to form solid solutions in a wide Al2O3/SiO2 range and to incorporate a large variety of foreign cations (Schneider, 2005b).

6 1 | Introduction

4. The fact that the structural principles of mullite can be extended to a large number of phases belonging to the family of mullite-type structures (chapter 1.2).

Within this introduction a brief summary of the synthesis and applications of mullite will be given followed by a description of the crystal structure of mullite and the definition of mullite-type compounds. Subsequently an outline of the state of the art in the field of boron-mullites will be given, concluding with an overview on the crystal chemistry of boron. Following the introduction, the scope and objectives of this thesis will be presented followed by a brief description of the materials and methods used during this work. The main part of this cumulative thesis consists of three manuscripts (chapters 4-6) published in peer-reviewed scientific journals and two chapters (7, 8) with contributions for manuscripts in preparation for publication. The thesis concludes with a summary of the major findings and specification of future perspectives.

1.1 Synthesis and applications of mullite Al4+2xSi2-2xO10-x (x = 0.2 … 0.9)

For mullites the resulting chemical composition and properties are controlled by the synthesis process itself more than by the bulk chemical composition. A detailed summary on the different ways of producing mullite is given by Komarneni et al. (2005). Apart from some special methods such as spray-pyrolysis, chemical vapor deposition and hydrothermally produced mullite, the authors present three categories of synthesis routes: 1. Solid-state derived mullite (sinter mullite, stoichiometric mullite, 3:2 mullite, x = 0.25) is essentially produced by solid-state reactions of natural aluminosilicate minerals below the melting point. Common starting materials are kaolinite, sillimanite, andalusite, kyanite, many types of oxides, oxyhydroxides, hydroxides, inorganic salts, and metal organics as well as alumina and silica precursors. Depending on the purity of the raw materials, sinter-mullites contain considerable amounts of impurities and usually an additional alumina source has to be added to avoid the formation of free silica. Refractory and furnace materials are typically produced by reaction sintering of silica and alumina. 2. Liquid-state derived mullite (fused mullite, 2:1 mullite, x ~ 0.4) is prepared by crystallization of melted alumina and silica mixtures. This involves temperatures of up to 2000°C for commercial fused-mullite ceramic products. Single crystals of up to 20 by 60 mm in size (Fig. 1-3) were grown with the Czochralski method (Guse and Mateika, 1974). 3. Solution sol-gel derived mullites (chemical mullites, variable composition) are produced from organic and inorganic precursors by polymerization and ceramization. The advantage of this method is the (fast) mullitization at relatively low temperatures (900 - 1300°C). This is achieved by increasing the reaction rates due to atomic, molecular, or nanoscale mixing of the components. The chemical composition of the product depends on the starting materials and the

7 1 | Introduction

formation temperature, with lower temperatures yielding Al-richer mulllites. In this study xerogels were prepared by the solution-plus-solution process using slow hydrolysis of an aluminum nitrate and TEOS (tetraethoxysilan) solution at 60°C as first descriibed by Hofffman et al. (1984), more details are given in chapter 3.1.

Fig. 1-3: Czochralski-grown mullite single crystals of 2:1-composition (§77 wt.% Al2O3) from Schneiider et al. (2008).

Thee outstanding properties of mullite (paage 6) give rise to a wide field of technical applications for mullite ceramics. Generally three groups of applications for mullite ceramics caan be distinguished (Schneider et al., 2008): 1. Monolithic mullite ceramics (Fig. 1-4 a) cover both traditional and advanced applications, e.g. porcelain, construction and engineering ceramics, refractories, substrates for catalytic convertors, electronic and optic devices (Okada and Schneider, 2005). Recent investigations focus on acicular mullite as a leading candidate material for diesel particulaate filters (Pyzik et al., 2011; Hsiung et al., 2013a, 2013b). 2. Mullite surface coatings have been successfully applied as environmenttal barrier coatings (EBC) to make materials resistant against harsh environmental conditions. IIn contrast to other silicon based ceramics, mullite has superior corrosion resistance, creep reesistance as well as high-temperature strength and toughness (Basu and Sarin, 2005). One prominent example for the use of EBCs is the heat shield of re-entry space vehicles (Fig. 1-4 b). 3. Mullite matrix composites cover the fields of composite materials with mullite matrices and mulllite fibers. The aim is to reduce the iinherent brittleness of systems by improvement of their toughness. As a break-through has not been achieved for the matrix compounds yet, the recent research focusses on continuous fiber-reinforced mullite matrix composites using alumina and mulllite fibers. Mullite matrix composites are used e.g. in components for gas turbine engines and heat shields for re-entry space vehicles. For more details refere to Schneider (2005c, 2005d).

8 1 | Introduction

(a) (b)

Fig. 1-4: (a) Sinter-mullite-based conveyor belt for continuous charging of annealing furnaces. (b) Panel for a re-entry space vehicle (mullite-coated C/C–SiC composite) from Schneider et al. (2008).

1.2 Crystal structure of mullite and definition of mullite-type compounds

The aluminosilicate mullite in the strict sense has the composition 3Al2O3 · 2SiO2 and is one of the two main compounds within the mullite solid-solution series Al2[Al2+2xSi2-22x]O10-x, with x ranging from 0.18 to 0.88, corresponding to about 57-92 mol% Al2O3 (Fischer et al., 1996). The second main compound is 2:1 mullite (fused mullite, x = 0.4). The chemical composition of mullite can be derived from lattice parameter a with a linear relatioonship given by Fischer et al. (1996)).

1.2.1 Crystal structure of mullite

A detailed systematic description of the crystal structure of mullite and the definition for mullite-type compounds was provided by Fischer and Schneider (2005) and is summarized in the following paragraphs. The average crystal structure of mullite is closely relatted to the more simple structure of sillimanite (Al2SiO5, Al2O3·SiO2, Pnma). In sillimanite (Fig. 1-5 a) the octahedral AlO6 chains are cross-linked by double chains of ordered SiO4 and AlO4 tetrahedra running parallel to the crystallographic c-axis. In contrast to that mullite (Fig. 1-5 b) is characteerized by a disordered arrangement of (Al,Si)O4 tetrahedra and therefore the c lattice parameter iis halved compared to sillimanite. Theoretically mullite can be derived from sillimanite by the coupled substitution: 2Si4+ + O2- o 2Al3+ + vacancy (Ƒ) (1) The replacement of 2Si4+ by 2Al3++ is accompanied by the formation of one vacancy in the oxygen position bridging the two tetrahedra in the double chain (O3 or O(C) position). This results in a displacement of the two tetrahedral sites (T) to positions designated as T* and so called triclusters

(T3O) of tetrahedra are build. As a consequence the O3 atoms are pulled towards the cations on T* and are therefore slightly displaced from their special position and have been desiignated O4 (or O(C*)). According to Angel and Prewitt (1986) thee occupancy of T* with Si is very small or even zero and the x value of the mullite solid solution series Al2[Al2+2xSi2-2x]O10-x corresponds too the number of oxygen vacancies.

9 1 | Introduction

Fig. 1-5: Crystal structure of mullite derived from sillimanite according to the substitution mechanism: 2Si4+ + O2- o 2Al3+ + vacancy (Ƒ). The big and thin arrows indicate the migration direction of the T to T* position and the displacement of O3 off the special position, respectively. (a) In sillimanite octahedral chains (blue) are linked by T2O7 groups (green). (b) In mullite the octahedral chains are linked by T2O7 groups or by T3O10 (so-called triclussters of tetrahedra). The oxygen vacancy is indicated by the square. After Fischer and Schneider (2005).

Thee local aluminum silicon distribution and the local order of oxygen vacancies deviates from the average structure described here and has been the aim of numerous studies.. A comprehensive summary on the real structure of mullite is giveen by Rahman and Freimann (2005).. According to their videographic 3D simulations, mullite is the first non-metallic mineral in which thhe diffuse scattering can be completely explained by short-range ordering of the oxygen vacancies. Investigations on the tetrahedral silicon and aluminum distribution using 29Si NMR data in comparison to simulated NMR data yielded a moderate degree of aluminum to silicon ordering in mullite (Schmücker et al., 2005a).

1.2.2 Definition of mullite-type compounds

Fischer and Schneider (2005, p. 1-2) define specific requirements for a mullite-type structural arrangement: 1. “The space group of a mullite-type structure must be a subgroup of the aristotype in space group P4/mbm.”

2. “The chains of edge-sharing MO6 octahedra (M = octahedral coordinated cation) must be lineaar representing single Einer-chains in their highest topological symmetry in space group P4/mbm.” (Fig. 1-6 a). 3. “The axis through the terminating atoms (non-edge-sharing atoms) of the octahedra must point towards the edges (parallel to the chaiin direction) of adjacent octahedra with 30° Ȧ 90” (Fig. 1-6 b). 4. “The chain structure should resemble the orthogonal metric of the aristotype perpendicular to the chain direction as closely as possible (Ȗ’ = 90 ± 5°).” (Fig. 1-6 b).

10 1 | Introduction

Point 4 of this classification scheme was extended by Fischer and Schneiider (2008, p. 919-920) by the two measures Qa and Qr. Qa should be close to 1 and is “defined as the ratio of the absolute values of the vectors enclosing Ȗ’.” Furtheermore the spacing between the octahedral chains should be as close as the mullite-type linkages, this is defined by Qr (in %), “representing the ratio of the ionic radius (Shannon, 1976) of the octahedrallly coordinated atom in the chains divided by the distance between neighboring chains” (Fig. 1-6 b, dotted line). For mullites the angle Ȧ between the octahedral axes is typically is around 60° and Ȗ’ = 90°, for 2:1 mullite Qa = 0.986 and Qr = 9.92 %. Additionally, “the linkage between the octahedral chaains must not entirely consist of octtahedrally coordinated atoms of the same kind as the chain-forminng cation” (Fischer et al., 2012, p. 406). A comprehensive overview on members of the mullite-type family of crystal structures is given by Fischer and Schneider (2005). According to their subgroup symmetry the individual structtures are sorted into 14 groups.

Fig. 1-6: Octahedral chains in mullite-type crystal structures. (a) Octahedral chains viewed perrppendicular to the c-axis. (b) Viewed parallel c and illustrating the inclination angle Ȧ between the octahedral chains as well as the angle Ȗ’. The distance between the octahedrally coordinated atoms for the calculation of Qr is represented by the dotted lline.

1.2.3 Foreign cation incorporation in mullite

In natural mullites the occurrence of iron and titanium (Agrell and Smitth, 1960) and occasion- ally chromium (Cameron, 1976) has been reported. In contrast to that a large variety of cations (mostly transition metals) can be incorporated in the mullite structure depending on the synthesis conditions and attmosphere as summarized in chapter 5.1 of this work. Detailed reviews on the foreign cation incorporation into mullite are available (SSchneider, 1990, 2005b; Schneider et al., 2008). Besides transition metals significant amounts of Ga3+ (Schneider, 1986a) and B3+ (Griesser et al., 2008) have been reported to enter the crystal structure of mullite.

11 1 | Introduction

1.3 Boron-mullites

Based on the observation of systematic changes in the refractive indices and the reduction of lattice parameter c, the existence of a solid-solution series between 3:2 mullite (Al4.5Si1.5O9.75) and

Al18B4O33 by substitution of boron for silicon was suggested in the 1950s (Dietzel and Scholze, 1955; Scholze, 1956). The term “B-mullite” or “boron-mullite” was introduced by Werding and Schreyer (1984) and later on extended (Werding and Schreyer, 1996) to a compositional range (gray area in Fig.

1-7) between the mullite solid solution series and the Al-borates Al5BO9 with mullite type structure

(Sokolova et al., 1978) and AlBO3 with calcite-type structure (Capponi et al., 1972). In contrast to the proposed miscibility between mullite and Al-borates, this work (chapters 4 and 5) as well as the investigations by Griesser et al. (2008) show that there is no complete solid solution between mullite and mullite-type aluminumborates. Stable phases in the boron-mullite compositional field within the ternary system Al2O3 – SiO2 – B2O3 as well as possible solid solution paths from literature are given in Fig. 1-7 and will be described in the following. In Fig. 1-7 phases with known mullite-type crystal structure are given as filled circles and will be discussed more detailed within this chapter. Compounds with unknown crystal structure are represented by open circles and triangles represent phases with known crystal structures different from mullite-type. Theoretic 1:1 Si-B substitution paths are indicated by the dotted lines in Fig. 1-7 with starting compositions of sillimanite, 3:2 mullite, and 2:1 mullite. The end-members of these series are

Al4B2O9, Al3BO6, and Al8B2O15, respectively. Only Al4B2O9 has a mullite-type crystal structure whereas the crystal structures of the other two compounds are unknown in the case of Al8B2O15,

(Reynaud, 1977) or different from mullite-type in the case of Al3BO6 (Capponi et al., 1972). Possible solid solution paths or polysomatic series reported in literature were reviewed by Fischer and Schneider (2008) and are given as solid gray lines in Fig. 1-7. Line ‘a’ represents a series of natural boromullites (Buick et al., 2008) with compositions plotting close to the joint between Al2SiO5

(sillimanite) and mullite-type Al5BO9. The compositions of the synthetic “boron-mullites” described by Dietzel and Scholze (1955) and Gelsdorf et al. (Gelsdorf et al., 1958) plot on the same line. However, a compositional gap between boromullite and sillimanite is proposed (Buick et al., 2008). Line ‘b’ (Fig. 1-7) represents compositions synthesized from gels (Grew et al., 2008), including a compound mentioned by Letort (1952). Compositions of natural and synthetic boralsilite (Grew et al.,

2008) plot on line ‘c’ (Fig. 1-7), with a minor solid solution range between boralsilite and Al8Si2B2O19 on the theoretic substitution path from sillimanite to mullite-type Al4B2O9. The compound Al8Si2B2O19 was first described by Werding and Schreyer (1992) as an orthorhombic sillimanite derivative but later on found to be a “boron-mullite” with impurities of disordered boralsilite (Grew et al., 2008). However, the question of the crystal structure and whether the phase is orthorhombic or monoclinic like boralsilite and Al4B2O9, remained open (Fischer and Schneider, 2008) and will be answered within this work (chapter 6).

12 1 | Introduction

Fig. 1-7: The ternary system Al2O3 – SiO2 – B2O3 after Fischer and Schneider (2008). The boron-mullite compositional field (Werding and Schreyer, 1996) is highlighted in gray. Dotted lines represent theoretical compositions of compounds with constant Al/(Si+B) ratio. Solid lines (gray) represent possible solid solution paths or polysomatic series with observed members of (a) natural and synthetic boromullites (Buick et al., 2008), including compounds from Dietzel and Scholze (1955), and Gelsdorf et al. (1958) (b) synthetic “boron-mullites” (Grew et al., 2008), including a compound from Letort (1952) (c) natural and synthetic boralsilite (Grew et al., 2008). Compounds with known mullite-type crystal structure are represented by filled circles, unfilled circles refer to compounds without structural investigations, compounds with crystal structures different from mullite-type are represented by triangles. Dashed blue lines I and II refer to initial bulk compositions of investigations from Zhang et al. (2010) and Griesser et al. (2008), respectively. [III, IV]: Boralsilite (Grew et al., 1998; Peacor et al., 1999). [V]: Al8Si2B2O19 (Werding and Schreyer, 1992), [VI]: Projected compositions of the minerals werdingite and grandidierite are given as squares (Anovitz and Grew, 1996). For details regarding the other binary aluminumborate compounds, refer to Fischer and Schneider (2008) and references therein.

Systematic investigations of boron-doped mullites with constant Al/Si = 3 ratio up to 7.4 mol%

B2O3 (line I in Fig. 1-7) were performed by Zhang et al. (2010). The results of sol-gel synthesis from bulk compositions with constant Al2O3 content starting from 70 and 60 mol% Al2O3 (lines II in Fig.

1-7) were published by Griesser et al.(2008). According to the authors 20 mol% B2O3 can be incorporated into 3:2 mullite. Fisch (2011) describes some preliminary investigations of samples with initial compositions on the lines sillimanite – Al5BO9 (line ‘a’ in Fig. 1-7) and sillimanite – Al4B2O9 (including the boralsilite composition) but without any final conclusions about the incorporation mechanism or solid solution behavior. However, all three approaches do not correspond to a 1:1 isomorphous substitution of Si3+ by B4+ in mullite. Furthermore no quantitative chemical analyses of the products are given. Both issues fall within the scope of this thesis and will be discussed in chapters 4 to 7.

13 1 | Introduction

In the following section mullite-type compounds within the B-mullite compositional field will be described according to the systematic introduced by Fischer and Schneider (2005). The information given here is summarized from Fischer and Schneider (2008), with a special focus on the aluminoborosilicate structures that are subject to this study and the linkage of the mullite-type octahedral chains. The primary references for crystal structure determinations are given, usually not corresponding to the first mention of the compound. The description of the linkage of the octahedral chains always refers to the systematic descriptions given in Fischer and Schneider (2008).

MUL-IV.4, I4/m: Al6B4Cu2O17 group

In Al6B4Cu2O17 the octahedral AlO4 chains are cross-linked by BO3 groups and trigonal bipyramids of (Cu,Al)O5 forming (Cu,Al)4O13 clusters with one common oxygen in the center of the cluster

(Kaduk et al., 1999). The bipyramids in the Li compound Al7B4LiO17 (Åhman et al., 1997) are only occupied by Al whereas Li is found in the channels not occupied by Al4O13 clusters.

MUL-VIII.2, Pbnm: Grandidierite group (Mg,Fe)Al3SiBO9

In grandidierite the octahedral chains are linked by AlO5 and MgO5 trigonal bipyramids, SiO4 tetrahedra and BO3 triangles (Stephenson and Moore, 1968).

MUL-II.3, Pbam: Mullite group The first systematic attempts for substitution of silicon by boron in 3:2 and 2:1 mullite (Griesser et al., 2008) indicate an incorporation of significant amounts of boron into the mullite crystal structure. However, no chemical or crystal structural analyses were available prior to this work and therefore are one of the key aspects of this thesis.

Mazza et al. (1992) describe crystal structures of the two metastable phases Al5BO9 and

Al4B2O9 in space group Pbam with BO3 groups crosslinking the octahedral chains. In the low alumina compound additionally some of the tetrahedral Al3+ is replaced by B3+. A re-examination of the crystal structure of Al4B2O9 (Fischer et al., 2008) yielded a monoclinic symmetry closely related to boralsilite with all lattice parameters doubled compared to mullite (see Boralsilite and Al4B2O9 group).

Two compounds Al8+xP1-xB1+xO16-x/2 with x = 0 and x = 0.5 were described to contain BO3 groups but with unreasonably large B-O distances (Mazza et al., 2001).

MUL-IV.32, Pbnm: Sillimanite group

Sillimanite is reported to contain up to 0.43 wt% B2O3 (Grew, 1996) which is not enough for structural diffraction studies. However, the sillimanite group contains several PbMBO4 phases (M=Al3+, Ga3+, Fe3+, Cr3+, Mn3+) having Pbnm symmetry but the structural details being quite different from sillimanite. The MO6 octahedral chains are cross-linked by BO3 groups and asymmetric pyramidal four coordinated Pb2+ having one lone pair electron. For these structures interesting magnetic properties have been reported (Park and Barbier, 2001; Park et al., 2003a, 2003b).

14 1 | Introduction

MUL-VII.33, Bb21m: A9B2 (Al18B4O33) group

In A9B2 mullite-type octahedral chains are cross-linked by edge-sharing AlO5 bipyramids alter- nating with isolated AlO4 tetrahedra and BO3 triangles. The crystal structure of the stable compound known as Al18B4O33 was refined with a composition of Al19.4B4.6O36 being rounded to

Al20B4O36 = Al5BO9 (Sokolova et al., 1978). There has been a long debate whether the A9B2-type compound actually has Al18B4O33 or Al5BO9 composition (e.g. Garsche et al., 1991). The result of recent crystal chemical investigations with multiple methods (Fisch et al., 2011) yielded a stoichiometry very close to Al5BO9, assuming that the Al18B4O33 composition might be the result of inaccurate chemical analyses. However, PGAA analyses of the commercial A9B2 compounds Alborex and Alborite (Shikoku Chemical Co., Marugame, Japan) yield a composition of Al18.0(4)B4.0(1)O33, resulting from investigations loosely related to this project (Söllradl et al., 2013, see abstract in Appendix B).

The crystal structure of the natural mineral boromullite Al9BSi2O19 (Buick et al., 2008) is closely related to A9B2 but with half of its structure consisting of sillimanite-type modules. It therefore represents a 1:1 polysome of Al5BO9 and Al2SiO5.

MUL-XVI.351, B112/m: Boralsilite and Al4B2O9 group

The monoclinic crystal structure of boralsilite Al16B6Si2O37 has an eightfold superstructure with all lattice parameters doubled in comparison to mullite. The octahedral chains are cross-linked by

Si2O7 groups, BO4 tetrahedra, BO3 triangles, and AlO5 bipyramids (Peacor et al., 1999). The crystal structure of Al4B2O9 can be derived from boralsilite by replacing 2 Si + 1 O by 2 B and was recently shown to have monoclinic symmetry (Fischer et al., 2008) rather than orthorhombic as described before (Mazza et al., 1992).

MUL-XXXII.352, P1¯ : Werdingite group

The natural mineral werdingite Mg2Al14Si4B4O37 is described having a triclinic structure with

AlO6 octahedral chains cross-linked by Si2O7 groups, (Fe,Mg)O5 bipyramids, (Al,Fe)O4, AlO5 and

BO3 groups (Niven et al., 1991).

15 1 | Introduction

1.4 Boron

1.4.1 Geological occurrence and applications of boron

A comprehensive overview of the “Mineralogy, Petrology and Geochemistry of Boron” is given by Anovitz and Grew (1996) and summarized in the following paragraph. The chemical element boron is the 27th abundant element (15 ppm) in the Earth’s upper continental crust where iit is highly enriched compared to the primitive mantle (0.6 ppm). Boron therefore is a common constituent of crustal rocks. The geological environments for the formation of boron minerals range from suubblimates formed in volcanic fumaroles to soluble salt deposits and boric acid lagoons to highly refractory materials formed under granulite-facies (high-pressure hiigh temperature digenesis zone) conditions. Borates and borosilicates in plutonic systems and metamorphic rocks are characterized by the absence or only small amounts of hydroxyl, whereas in saline deposits borates with substantial water and hydroxyl content are formed. With the exception of three fluorides, all of the over 200 boron minerals approved by the International Mineralogical Association (IMA) are oxygen compounds. In contrast to other elements oof similar abundance boron is sussceptible to fractionation processes and therefore can become concentrated in minerals (Hawthorne et al., 1996). The four boron minnerals important for mining are borax (Na2B4O5(OH)4·8H2O), colemanite (CaB3O4(OH)2·H2O), ulexite

(NaCaB5O6(OH)6*5H2O), and kernite (Na2B4O6(OH)2*3H2O). Together they account for 90 % of the world production of boron (Lorenz and Gwosdz, 2003). The worldwide production of boron minerals is constantly increasing since 1975 (Fig. 1-8). Boron ore is predominantly exported from Turkey, USA, Russia, Argentina, Chile and China ((Pohl, 2005). Boron has a wide range of industrial applications in ceramics, glasses, metallurgy and other fields (Anovitz and Grew, 1996).

Fig. 1-8: Worldwide production of boron ore in thousand metric tons. From 2006 on (red symbols)) the U.S. production is excluded (Buckingham et al., 2012).

16 1 | Introduction

1.4.2 Crystal chemistry of boron

If no specific references are given, the following chapter refers to the review on “The crystal chemistry of boron” by Hawthorne et al. (1996). Boron is the only non-metal eelement of group III of the periodic table and has the ground-state electronic structure of [He]2s22p1. Due to its high ionization potential bond formation involves rather covalent than ionic mechanisms. For covalent bonds involving the four orbitals, boron only contributes three electrons and therefore is a strong electron-pair acceptor with very high affinity for oxygen. In general boron haas many similarities to carbon and silicon and also the structural chemistry of B and Si, when associated with oxygen, is quite similar. All groups, BO3, BO4, and SiO4, have a marked tendency to polymerize in the solid state, resulting in a great structural complexity.

In crystal structures boron occurs in triangular (B-ࢥ3) and tetrahedral (B ࢥ4) coordination to oxygen and hydroxyl groups (ࢥ: O2-, OH-). Hawthorne et al. (1996) reviewed the B-ࢥ distances and angles in 80 refined mineral structures, giving a B-ࢥ3 grand mean distance of 1.370(17) Å with minimum and maximum observed distances of 1.322 and 1.428 Å, respectively. The B-ࢥ4 grand mean distance is 1.476(25) Å with minimum and maximum observed distances of 1.397 and 1.512 Å, respecttively. These values are in excellent agreement with the respective sums oof ionic radii according to Shannon (1976). In borates under ambient conditions boron polyhedra are usually isolated or share vertices with one another and groups of 3-6 boron-oxygen polyhedra are formed. In borosillicates boron polyhedra share vertices with BO4 or SiO4 tetrahedraa, and, rarely, with AlO4 or BeO4 tetrahedra (Anovitz and Grew, 1996). In boron-bearing mullite-type compounds, boron-oxygen polyheedra connect the typical octahedral chains. For the systematization of the different polyhedral arrangements in borates refer to Hawthorne et al. (1996) and the review of Filatov and Bubnova (2000). An alternative approach for the classification borates and borosilicates involves a description based on the definition for mullite- type structures. The authors recommend “… to designate all boron compoundss with the characteristic mullite-type MO4 chains of MO6 octahedra as ‘mullite-type’ boron compounds and to use the term ‘boron-mullite’ or ‘B-mullite’, initially introduced by (Werding and Schreyer, 11984) for the subgroup of Al borates and Al borosilicates with mullite-type structures” (Fischer and Schneider, 2008, p. 917). Boron-bearing mullite-type compounds and boron-mullites have been systeemmatically described in chapter 1.3.

In borates the BO3 and BO4 polyhhedra practically do not change upoon heating, similarly to tetrahedra in silicates (Filatov and Bubnovva, 2008) which leads to highly anisootropic or even negative thermal expansion in many borates (Filatov and Bubnova, 2000). Applicaation of the pressure/ coordination rule would suggest that minerals with BO3 groups are stable preffeerably at relatively low pressurres, while those with BO4 tetrahedra are stable at higher pressures (Werding and Schreyer, 1996).

17 1 | Introduction

1.4.3 Chemical analyses of boron

As the precise chemical analysis of boron has been an issue throughout this thesis, a short overview on the different instrumental techniques is presented here, summarized frfrom Robertson and Dyar (1996) and Anovitz and Grew (1996). Reliable wet chemical analyses of boron in minerals were not possible before the end of the 19th century. Especially borosilicates remain a challenge until today as boron is not detected quantitatively with X-ray fluorescence and the chemical digestion of borosilicates is a very complex and vulnerable process. Today, technological advances in the wavelengtth-dispersive electron microprobe analyses (EMPA) as well as secondary ion microprobe (SIMS) analyses enable precise quantification of boron. However, the analysis of boron with the EMPA requires special spectrometer crystals and instrumental setup. Additionally both methods, EMPA annd SIMS, require considerable care with respect to matrix effects and sample preparation. Furthermore standard materials are necessary that match as closely as possible the major-element chemistry of the sample. Robertson and Dyar (1996) summarize the four nuclear methods used for borron (and other light element) analyses. There are two fundamental advantages of the nuclear methods over EMPA and SIMS: All methods are less vulnerable to surffaace texture effects and are nearly free of matrix effects. However, only particle-induced gamma-ray emmission (PIGE) and particle-induced particle emission (NRA) caan be used as microprobe techniques whereas prompt-gamma neutron activation analysis (PGNAA) and fast neutron activation analysis (FNAA) are restricted to analyses of the bulk sample. The major disadvantage applicable to all nuclear methods is the limited access to faaccilities. Thee usage of laser-ablation ICP-MS for the analyses of boron as a major element as done by Fisch et al. (2011) requires an elaborate cleaanning procedure of the spectrometer after the measurements. Otherwise memory effects will considderably reduce the detection limit for subsequent boron analyses.

2 Scope and objectives, thesis outline

2.1 Scope and objectives

The unique properties of mullite such as low thermal expansion, low thermal conductivity, excellent creep resistance, high temperature stability and very good chemical resistance are the basis for the great importance of mullite ceramics in technical applications (chapter 1). Aluminum borates

(e.g. A9B2, Al18B4O33) represent a second class of industrially important materials with mullite-type crystal structure. The most cited applications are the reinforcement of aluminum alloys by incorporation of aluminum borate whiskers and, due to the corrosion resistance of aluminum borates against molten glasses containing boron, the use in refractory linings (Garsche et al., 1991 and references therein). The combination of the two systems promises a great potential to design high- performance materials. Solid solution between 3:2 mullite and Al18B4O33 was proposed in the 1950’s (Dietzel and Scholze, 1955; Scholze, 1956) as described in chapter 1.3. More recent systematic studies (Griesser et al., 2008; Zhang et al., 2010; Fisch, 2011) showed that there is no complete solid solution between mullites and aluminum borates. However, boron-doping of mullite results in significant changes of lattice parameters b and c. In contrast to that no significant changes are observed for lattice parameter a, which is linearly correlated with the Al/Si ratio in mullite. From infrared spectroscopic data the presence of BO3 rather than BO4 in B-mullites was proposed (Griesser et al., 2008). Despite their great potential for high-performance materials the information on the solid solution behavior, the boron incorporation mechanism, the crystal structure, and properties of boron-doped mullites is very sparse. A central part of this thesis is the determination of the crystal structure of B-mullite; this includes the formulation of a substitution mechanism and will lead to a better understanding of the physical properties of B-mullite. Furthermore the determination of some basic properties of B-mullites falls within the scope of this thesis as well as the determination of the actual chemical compositions and the incorporation limit for boron in the crystal structure of mullite. Thus, the following questions and tasks regarding mullite-type compounds in the ternary system

Al2O3 - SiO2 - B2O3 will be addressed within this thesis:

Objective 1: Syntheses and properties of B-mullites Synthesize B-mullites and study their in-situ formation conditions as well as their behavior and stability at high temperature and high pressure (thermal expansion, thermal stability, response to pressure).

19 2 | Scope and objectives, thesis outline

Objective 2: Crystal structure of B-mullite Develop a crystal structure model for B-mullite with focus on the coordination of boron- polyhedra and the linkage between the octahedral chains. This includes the formulation of a substitution mechanism as well as the determination of the chemical composition and the incorporation limit for boron in mullite. Furthermore the relationships between crystal structure and physical properties are addressed.

Objective 3: Synthesis and crystal structure of Al8Si2B2O19

The question whether the compound Al8Si2B2O19 has orthorhombic or monoclinic symmetry (chapter 1.3) is addressed.

2.2 Thesis outline

Following the introduction and methodic chapters (1-3), the central part of this thesis is presented in five chapters made-up of three manuscripts (chapters 4-6), which are published in peer- reviewed scientific journals, and two chapters (7, 8) with contributions for manuscripts in preparation for publication. Chapters 4-6 correspond to the original manuscripts, compared to the published versions some cross references were added and figures were colored. All manuscripts were written by Hanna Lührs, the individual contributions from the co-authors are pointed out separately in the following.

The first manuscript [chapter 4] “Boron mullite: Formation and basic characterization” Hanna Lührs, Reinhard X. Fischer, and Hartmut Schneider Materials Research Bulletin 47 (2012) 4031–4042 DOI 10.1016/j.materresbull.2012.08.064 addresses the formation conditions and properties of a series of B-doped mullites synthesized from sol-gel precursors. A detailed description of the synthesis protocol is given and the in-situ phase formation of B-mullites is described including a preliminary estimation of the incorporation-limit for boron, mainly based on the qualitative phase composition and lattice parameters. The characteristic development of lattice parameters with increasing B-content is discussed, as well as the thermal expansion behavior and the thermal stability of B-mullite. A first idea of the incorporation mechanism for boron is developed from calculations based on the amount of alumina impurities and confirmed by IR spectroscopic data. All syntheses, experiments, and data evaluation were done by Hanna Lührs with support from Petra Witte at the scanning electron microscope and Ute Jarzak for the infrared spectroscopy. The X-ray fluorescence analyses were performed by Bernhard Schnetger (Universität Oldenburg).

20 2 | Scope and objectives, thesis outline

In the second manuscript [chapter 5] “Neutron diffraction and 11B solid state NMR studies of the crystal structure of B-doped mullite” Hanna Lührs, Anatoliy Senyshyn, Scott P. King, John V. Hanna, Hartmut Schneider, and Reinhard X. Fischer Zeitschrift für Kristallographie 228 (2013) 457-466 DOI: 10.1524/zkri.2013.1595 the crystal structure of B-mullite is presented based on the results of neutron diffraction data and 11B MAS NMR data from a series of B-doped mullites. The combination of Rietveld refinements, difference Fourier calculations, distance least squares refinements, and grid search methods was necessary in order to develop a reliable structural model. A special focus lies on the substitution mechanism, the coordination of B-polyhedra, and the consequences of the substitution for the local and average crystal structure. Furthermore this crystallographic model provides an explanation for the anisotropic behavior of the lattice parameters upon B-incorporation. The syntheses were done by Hanna Lührs with general lab support from Malik Šehoviü. The neutron diffraction experiments were performed by Hanna Lührs with support from Reinhard X. Fischer and technical assistance of Anatoliy Senyshyn (FRM II, Garching). The MAS NMR measurements and data fitting were done by Scott P. King and John V. Hanna (University of Warwick). All other data evaluation and processing was done by Hanna Lührs.

The third manuscript [chapter 6] “Ambient and high-pressure synthesis, composition, and crystal structure of B-mullites” Hanna Lührs, Stefan Söllradl, Scott P. King, John V. Hanna, Jürgen Konzett, Reinhard X. Fischer, and Hartmut Schneider Available online in Crystal Research and Technology (2013) DOI 10.1002/crat201300210 deals with the chemical analyses of single phase B-mullite samples using prompt gamma activation analyses (PGAA) and the comparison to the refinement-results based on neutron diffraction data. The second key aspect of this manuscript covers the high-pressure synthesis and crystal structure of the 11 Al8Si2B2O19 compound based on X-ray diffraction, PGAA, and B MAS NMR data. The model is compared to the crystal structure of B-mullite from the second manuscript (chapter 5). The syntheses at ambient pressure were done by Hanna Lührs with general lab support from Malik Šehoviü. For the high-pressure syntheses the precursors were sent to Jürgen Konzett (Universität Innsbruck). The PGAA experiments and data evaluation were done by Stefan Söllradl (FRM II). All other measurements were performed by Hanna Lührs, with support from Reinhard X. Fischer and technical assistance of Anatoliy Senyshyn (FRM II) at the neutron powder diffractometer and assistance from Scott P. King (University of Warwick) for the MAS NMR measurements.

21 2 | Scope and objectives, thesis outline

In chapter 7 “Chemical composition of B-mullites” the correlation between lattice parameters and boron-content of different B-mullites is addressed. Compared to chapters 4 and 6, an enhanced relationship is presented, that allows for the calculation of the B2O3 content in B-mullites from lattice parameter c.

Chapter 8 addresses the “Crystal chemistry of mullite and B-mullite under high-pressure”. Here the first results of an in-situ high-pressure Synchrotron X-ray diffraction study of B-mullite up to 28 GPa are presented and compared to different boron-free mullites with respect to the crystal chemical changes in the structure. The in-situ high-pressure Synchrotron X-ray diffraction data were collected by P.E. Kalita and K. Lipinska (University of Nevada Las Vegas), the Rietveld refinements were done by Hanna Lührs. These results are in preparation for publication1.

1 The results of chapter 8 have not been submitted to a scientific journal as the copyright on the experimental data is at the University of Nevada, Las Vegas, who reserves the right of primary publication of the data from samples sent to them three years ago (2010).

3 Material and methods

Within this section the synthesis protocols and analytical methods used within this work are described briefly. More detailed information on the syntheses, the instruments and their configuration is given in the respective chapters (4 to 8).

3.1 Syntheses in the system Al2O3 - SiO2 - B2O3

All samples reported on in this work were synthesized by the sol-gel procedure and follow the nitrate decomposition method using aluminum-nitrate nonahydrate, tetraethoxysilane and boric acid as reactants (Table 3-1). The chemicals were dissolved in pure ethanol and heated at 60°C in a water bath to form transparent sols followed by gelation at 60°C and subsequent drying at 150°C. The resulting yellowish, spongy glass was ground and calcined at 350°C with subsequent mullitization in corundum or platinum crucibles at temperatures between 900 and 1400°C. This method is similar to the solution- plus-solution method described by Hoffman et al. (1984) (chapter 1.1) and was already successfully used for the B-mullite synthesis in the past (Griesser et al., 2008). Preliminary experiments not reported here in detail showed that careful and slow preparation of the gel and a mullitization temperature of 1200°C yield the best results with respect to the amount of alumina impurities and crystallinity. The precursor for the high-pressure syntheses was prepared by the same procedure only after calcination at 350°C the precursor was heated to 600°C in order to remove residual nitrates and organic compounds. Platinum or gold capsules and a piston cylinder apparatus were used to synthesize the samples at 800-875°C and pressures between 7 and 10 kbar. A detailed description of the high- pressure syntheses is given in chapter 6. A list of all samples with their initial gel composition, synthesis conditions, and applied analytical methods is available in Appendix C. The raw data of all measurements can be found on the supplementary CD.

Table 3-1: List of chemicals used for the syntheses of B-mullites.

chemical elemental formula purity producer aluminum-nitrate nonahydrate Al(NO3)3•9H2O > 98% Fluka Chemicals tetraethoxysilane (TEOS) C8H20O4Si > 99% Sigma-Aldrich boric acid H3BO3 p.a. Merck 11 11 99% B boric acid H3BO3 99 atom % B Sigma-Aldrich ethanol C2H6O 98% Sigma-Aldrich

23 3 | Material and methods

3.2 Analytical methods

Numerous analytical methods were applied in order to achieve the objectives and answer the questions mentioned in the previous chapter (2.1). All methods and instrumentss used will be very briefly introduced in the sections below. Details regarding the configuration of tthe instruments are given in the respective chapters (chapters 4 to 8). If not stated otherwise the experiiments were carried out in the group of Crystallography, Department of Geosciences, University of Bremen.

3.2.1 Powder diffraction experiments

Thee methodological focus of this crystal chemical thesis lies on the Rietveld refinements based on diffraction data obtained from laboratory X-ray powder difff ractometerss, neutron powder diffraction, and in-situ high-pressure synchrotron X-ray diffraction using a diamond anvil cell.

3.2.1.1 Powder X-ray diffraction (XRD) under ambient and high temperature connditions Powder X-ray diffraction experiments were performed under ambient and high-temperature conditions up to 1200°C (chapters 4-8). In most cases the Bragg-Brentano PANalytical X’Pert MPD

PRO diffraction system was used, equipped with Cu-KĮ radiation (Ȝ = 1.5418Å) and X’Celerator detector system. For the temperature-dependent measurements the system was exttended by the high- temperature chamber HTK 1200N (Anton Paar Co.). Individual measurements were run on the Philips PW1800 Bragg--Brentano powder diffractometer using Cu-KĮ radiation (Ȝ = 1.5418 Å) and a PW1711 proportioonal detector. The instrument is equipped with an automatic divergence slit, primary and secondary Sooller slits with 0.04 rad aperture, as well as a secondary PW1801/29 monochromator crystal. Only for samples with very small amounts of material transmission geometrry using glass capillaries was applied. Experiments were carriied out on the Bruker AXS D8 Advance powder diffractometer, equipped with a position sensitiive LynxEye detector and a Johansson Monochromator, accomplishing the usage of pure Cu-KĮ1 (Ȝ = 1.540598 Å) or Mo-KĮ1 (Ȝ = 0.70932 Å) radiation.

3.2.1.2 Neutron diffraction Due to the very low scattering factor of boron using X-rays and its relatively low concentration in the samples, detailed crystal structural investigations including the boron position cannot be accomplished using X-ray diffraction data. In contrast to X-rays, boron exhibbits a high neutron scattering length permitting the location and refinement of boron positions with relatively low occupancy. Furthermore oxygen positions can be refined more reliably based on neutron diffraction data. Neutron diffraction experiments (chapter 5 and 6) at room temperature were carried out in vanadium vessels on the high-resolution powder diffractometer SPODI at FRM-II (Garching, Germany) using a wavelength of Ȝ = 1.54812 Å (Hoelzel et al., 2012).

24 3 | Material and methods

3.2.1.3 In-situ high pressure synchrotron X-ray diffraction Diamond anvil cells and angle-dispersive synchrotron X-ray diffraction (ADXRD) was used for in-situu high-pressure experiments on boron-free mullites and B-mullitte. The high-pressure measurrements were carried out by Patricia E. Kalita and Kristina Lipinska at the 16-IDB beamline of the High Pressure Collaborative Access Team (HCCAT), Advances Photon Source, Argonne National Laboratory, USA. Details regarding the experimental conditions are given in chhapter 8.

3.2.2 Differential thermal analyses (DTA)

The in-situ formation and thermal stability of B-mullites was studied usiing differential thermal analyses (DTA). The experiments were performed on a Netzsch 449 F3 Jupiiter STA (simultaneous thermal analyses) apparatus under synthetic air atmosphere (chapter 4).

3.2.3 Infrared spectroscopy (IR)

A first characterization of the boron incorporation mechanism (chapter 4) was done by Fourier transform infrared spectroscopy (FTIR). The experiments were carried out in the Department of Chemistry, University of Bremen, using the KBr method on the Avatar 370 FTIR Thermo Nicolet spectrometer under nitrogen atmosphere.

3.2.4 Magic angle spinning nuclear magnetic resonance spectroscopy (MAS NMR)

Crucial information on the local boron environment in B-mullites was gathered from the 11B MAS NMR spectroscopic investigations (chapters 5 and 6). All experiments were performed in the Materials Solid State NMR group of the Department of Physics, University of Warwick. The 11B solid state MAS NMR single pulse measurements were performed at a B0 field of 144.1 T or 14.5 T using a Bruker Avance II+ 600 or Varian 600 spectrometer, respectively.

3.2.5 Chemical analyses

In chapter 1.4.3 the difficulties generally involved in the chemical analyses of borosilicates are described including a summary of methods that are applicable for the determination of the boron content in minerals. As the actual chemical composition of B-mullites is a crucial aspect within this thesis, several methods were applied while other methods such as wet chemical analyses or laser ablation ICP-MS as used by Fisch et al. (2011) were not possible due to restrictions mentioned in chapter 1.4.3.

3.2.5.1 X-ray fluorescence (XRF), scanning electron microscope (SEM), annd energy dispersive X- ray spectroscopy (EDX) Although the element boron cannot be analyzed using XRF and EDX analyses, these methods were applied in order to determine the Si/Al ratio in selected samples (chapter 44). Parallel to the EDX analyses the particle morphology of B-doped mullites was investigated using the Supra 40 (Zeiss) field

25 3 | Material and methods emission SEM equipped with an INCA PentaFET-x3 EDX detector (Oxford). For X-ray fluorescence the Philips PW2400 XRF spectrometer at the University of Oldenburg, was used (chapter 4).

3.2.5.2 Prompt gamma activation analyses (PGAA) PGAA is a non-destructive multi-element method that can be used for samples in any aggregation state and is especially useful for the analyses of boron at a trace level. For the analysis of boron as a major element special requirements are necessary that were set up at the PGAA spectrometter at the FRM II within a project loosely related to this thesis (Söllradl et al., 2013, abstract in Appendix B). In detail this setup involves the attenuation of the signal using a 10 mm lead foil and finally led to a precise determination of the chemical composition of several B-mullite samples (chapters 6 and 7).

3.2.6 Computer proograms used

Thee Rietveld refinements were performed with the programs BRASS (Birkeennstock et al., 2012) and Topas (Bruker, 2009). Indexing and space group determination were done by tthe programs XFIT (Cheary and Coelho, 1996), crysfire (Visser, 1969; Shirley, 2002), and checkcell (Laugier and Bochu, 2004). Distance least squares refinements were performed using the DLS program by Baerlocher et al. (1978). For the crystal structure drawings the program Struplo (Fischer and Messneer, 2013) was used. The temperature and pressure dependent diffraction patterns were visualized uusing the program multisurfaace (Messner, 2013). For the evaluation of the thermal analyses the Nettzsch software was used (NETZSCH, 2010). The NMR spectra were simulated using the program DMfit (Massiot et al., 2002). For the evaluation of the PGAA spectra the programs Hypermet-PC (Fazekkas et al., 1996) and ProSpeRo (Revay, 2009) were used by S. Söllradl. The integration of the in-situ high-pressure Synchrotron X-ray diffraction data was done with the FIT2D software (Hammersley, 2005) by P.E. Kalita and K. Lipinska.

4 Boron mullite: Formation and basic characterization

Hanna Lührsa,*, Reinhard X. Fischera, Hartmut Schneidera,b

a Universität Bremen, FB 5 Geowissenschaften, Klagenfurter Straße, D-28359 Bremen, Germany b Universität Köln, Institut für Kristallographie, Greinstraße 6, D-50939 Köln, Germany

Published in: Materials Research Bulletin 47 (2012) 4031-4042 DOI: 10.1016/j.materresbull.2012.08.064

submitted: 26 May 2012 | revised: 24 July 2012 | accepted: 23 August 2012 | online: 1 September 2012

Keywords: inorganic compounds | ceramics | chemical synthesis | X-ray diffraction | thermal expansion

A series of boron doped mullites (B-mullite) was prepared from single-phase gels with initial compositions based on a 1:1 isomorphous substitution of Si by B, starting from a 3:2 mullite composition (Al4.5Si1.5O9.75). A high amount of boron (>10 mol.%) can be incorporated into the crystal structure of mullite where it most likely replaces Si. In-situ phase formation of B-mullites was studied with high temperature X-ray diffraction and thermal analysis. A decrease of the formation temperature for B-mullite with increasing boron content was observed. With increasing boron content lattice parameters b and c significantly decrease, while no systematic evolution of a is observed. Long annealing at 1400°C results in decomposition of B-mullite to boron free mullite and Į-alumina. At 800°C B-mullite appears to be stable over a period of at least 12 days. The mean thermal expansion coefficient was reduced by 15% upon incorporation of boron which makes the material technologically interesting.

* Corresponding author: E-mail address: [email protected] Telephone: +49-421-218 65181, Telefax: +49-421-218-65189

27 4 | Boron mullite: Formation and basic characterization

4.1 Introduction

Mullite is rare in nature but one of the most important synthetic materials for advanced structural and functional ceramics. This is due to its outstanding properties such as low thermal expansion, low thermal conductivity, excellent creep resistance, high temperature stability, and very good chemical resistance (Schneider, 2005a). Mullite in the strict sense has the composition 3Al2O3·2SiO2. However, chemical mullites formed from organic and inorganic precursors by polymerization and ceramization have a variable aluminum to silicon ratio according to Al2[Al2+2xSi2-2x]O10-x with 0.18 x 0.88, where x can be derived from lattice parameter a according to a linear relationship (Fischer et al., 1996).

Aluminum borates represent another very important class of materials, especially Al18B4O33, is of great industrial importance. The most cited application is the reinforcement of aluminum alloys by incorporation of Al borate whiskers to obtain high mechanical strength. Furtheron it is used as a material for refractory linings due to its favorable corrosion resistance against molten glass containing boron (Garsche et al., 1991 and references therein). The Al2O3-rich aluminoborate is reported having two slightly different chemical compositions: Al5BO9 and Al18B4O33. The result of recent crystal chemical investigations by Fisch et al. (2011) yielded a stoichiometry close to Al5BO9, assuming that the Al18B4O33 composition might be the result of inaccurate chemical analyses. The combination of the two systems, mullite and Al borate, has a great potential to design high- performance materials. A solid solution between 3:2 mullite and the aluminum borate Al18B4O33 was proposed due to the similarities between the physical properties of mullite and Al18B4O33 (Dietzel and Scholze, 1955; Scholze, 1956). The term ‘boron-mullite’ was introduced by Werding and Schreyer

(1996) for compounds represented by the gray field in Fig. 4-1 of the ternary system Al2O3-SiO2-

B2O3with vertices formed by 3:2 and 2:1 mullite, Al5BO9 with mullite-type structure (Sokolova et al.,

1978), and AlBO3 with a calcite-type structure (Capponi et al., 1972). Several compositions within this field are known to have a mullite-type structure and are represented by filled circles in Fig. 4-1. For further details and criteria for the classification of mullite-type structures see Fischer and Schneider (2008). Unfilled circles represent compounds that have not been structurally investigated yet, triangles refer to compounds with crystal structures different from mullite-type (Fischer and Schneider, 2008; Fischer et al., 2012). First studies within the compositional range of boron-mullites were performed by Griesser et al. (2008) and Zhang et al. (2010). They investigated starting compositions that are represented by black crosses and squares in Fig. 4-1, respectively. Griesser et al. (2008) investigated two series with a constant Al2O3 content of 70 and 60 mol.% respectively. According to these authors, 20 mol.% B2O3 can be incorporated in the series with 3:2 mullite as end-member (60 mol.% Al2O3). However, their approach does not correspond to a 1:1 isomorphous substitution of Si4+ by B3+. Zhang et al. (2010) investigated boron-doped mullites at a constant Al/Si=3 ratio with up to 7.4 mol.% B2O3. In both

28 4 | Boron mullite: Formation and basic characterization studies (Griesser et al., 2008; Zhang et al., 2010) quantitative chemical analyses were not performed and only the initial gel composition is reported. Griesser et al. (2008) do not give detailed information on phase composition and the existence of amorphous components. Hong et al. (1996) investigated the anisotropic grain growth in B2O3-doped diphasic mullite gels with Al/Si ~ 3. The aim of this work is to produce B-doped mullites which are likely to have improved properties such as reduced thermal expansion compared to boron-free mullite. The syntheses were performed by a sol-gel process with starting compositions where B3+ replaces Si4+ in equal amounts. We present qualitative results of the incorporation of boron into the mullite structure based on powder X-ray diffraction data at ambient and high temperatures, thermal analyses, infrared spectroscopy and XRF bulk analyses.

Fig. 4-1: The ternary system Al2O3 – B2O3 – SiO2. The gray field represents the compositional range for ‘boron-mullites’ as defined by Werding and Schreyer (1996). Verified compositions with mullite-type structure are given as filled circles (black), unfilled circles correspond to poorly characterized compounds. Triangles represent phases other than B-mullites, crosses and squares represent the compositions investigated by Griesser et al. (2008) and Zhang et al. (2010). Red filled circles are the compositions investigated in this work. Three phases with distinct compositions that contain Si have been investigated so far: Boromullite (Buick et al., 2007, 2008), boralsilite and Al8Si2B2O19 (Grew et al., 2008). 4.2 Synthesis, sample preparation

All sol-gel syntheses followed the procedure given by Griesser et al. (2008) with modified gel compositions to account for the Si-B substitution. Appropriate amounts of aluminum-nitrate nonahydrate (>98%, Fluka), tetraethoxysilane (>99%, Aldrich) and boric acid (Merck, p.a.) were suspended in pure ethanol and stirred in an oil bath at 60°C to become transparent sols. These were stored in an oven at 60°C for 4 days for gelation. The transparent gels were dried at 150°C resulting in a yellowish, dry, spongy glass which was ground and calcined at 350°C for 5 h in a corundum crucible. Mullite forms during the 5 h annealing at 1200°C, which turned out to be the best

29 4 | Boron mullite: Formation and basic characterization temperature to generate crystalline mullite powder. In addition a series of samples wwas prepared using the compositions investigated by Griesser et al. (2008) replacing Si by 2B. The iinitial gel compositions of all samples are given in Table 4-1 as well as the respective syntheses temmperatures and the analytic methods applied to each sample. Wheenever a chemical composition is giveen within this work it refers to the initial gel composition wit Al2O3 + SiO2 + B2O3 = 100 %.

Table 4-1: Initial gel compositions, synthesis temperatures and analytic methods for all samples.

product

Al2O3 B2O3 SiO2 5 h 5 h 5 h 90 h [mol.%] [mol.%] [mol.%] 1200°C 1300°C 1400°C 1400°C XRPD XRPD HT-XRPD STA precurso STA STA product XRF FTIR 3:2 mullite 60.0 0.0 40.0 x x x x x x x B-mullite 60.7 1.0 38.3 x x x x x x x x x x 61.4 2.0 36.6 x x x 61.7 3.1 35.3 x x x x x 61.9 4.1 34.0 x x x x x x x 63.0 5.2 31.8 x x x x x x 63.9 6.4 29.8 x x x x x x x 64.3 7.5 28.2 x x x x x 65.7 9.9 24.4 x x x x x 67.2 12.3 20.5 x x x x x x 68.1 13.6 18.3 x x 69.1 14.9 15.9 x x x x x x 70.4 17.6 12.1 x x x x x x 72.0 20.4 7.5 x x x x x 73.9 23.4 2.8 x x x x x 75.0 25.0 0.0 x x x x x

Al18B4O33 81.8 18.2 0.0 x x

4.3 Experimental

4.3.1 Scanning electron microscopy and energy dispersive X-ray spectroscopy

Particle morphology was recorded by a Supra 40 (Zeiss) field emissionn scanning electron microscope (SEM). The chemical compossition was determined by energy dispersive X-ray spectroscopy (Oxford with INCA PentaFET-x3 EDX detector), attached to the SEM. Therefore the powdered samples were sputtered with a thin film of gold and EDX spectra were taken using an excitation voltage of 15 kV.

4.3.2 Powder X-ray diffraction at ambient temperature

X-ray powder diffraction data were collected at room temperature on a Bragg-Brentano PANalytical X’Pert MPD PRO diffraction system, equipped with Cu-KĮ radiation (Ȝ = 1.5418 Å), ¼° fixed divergence, primary and secondary Solller slits with 0.04 rad aperture, secoondary Ni-filter and X’Celerator detector system (127 channels, channel width 0.01671°2ș). Samples were prepared with

30 4 | Boron mullite: Formation and basic characterization the standardized PANalytical backloading system using circular sample holders with 16 mm diameter. Scans were performed in the range from 3° to 140°2ș, step width 0.0167°2ș; tthe measuring time per step was 25 seconds. For samples with small amounts of material (after thermal analyyses) measurements in transmission geometry were performed using glass capillaries on a Bruker AXS D8 Advance powder diffractometer with Cu-KĮ1(Ȝ = 1.540598 Å) or Mo-KĮ1 (Ȝ = 0.70932 Å) radiation. A Johansson

Monochromator accomplishes the usage of pure KĮ1 radiation. The instrument is equipped with a primary divergence slit, primary (4°), and ssecondary Soller (2.5°) slits, a secondary iris aperture (6.42 mm) and a position sensitive LynxEye detector. Scans were performed with a step width of 0.01°2ș, for Cu radiation a measuring time of 3 s//step from 5° to 120°2ș was applied and for Mo radiation scans were performed in the range of 5° to 60°2ș with a measuring time of 15 s//step.

4.3.3 In situ highh temperature X-ray diffraction

For the temperature-dependent measurements the same configuration of tthe PANalytical X’Pert MPD PRO diffraction system as descriibed in section 3.2 was used, equipped with the high- temperature chamber HTK 1200N (Anton Paar Co.). Measurements were perffoormed in steps of 50°C (25°C, 50, 100, …, 1200°C) during heating as well as during cooling. The heatiing rate was 50 °C/min, at each temperature the sample was equilibrated for 5 min before starting the scan from 3° to 140°2ș with a step width of 0.0167°2ș and 25 seconds measuring time per step. For sample preparation the powders were dispersed in alcohol and the resulting slurry pipetted into the corundum sample holder, thus producing porous samples to allow for a thermal expansion of the sample during heating.

4.3.4 Rietveld refinements

Rietveld refinements were carried out using the program Topas (Bruker, 2009). For the determination of lattice parameters Pawley fits were applied with starting parameters taken from the 3:2 mullite structural model of Saalfeld and Guse (1981).

4.3.5 X-ray fluorescence

For X-ray fluorescence analysis (XRF), 600 mg of the sampm le powder were mixed with 3600 mg lithium tetraborate, pre-oxidized at 500°C with NH4NO3 and fused to glass beads. The beads were analyzed by a spectrometer (Philips PW 2400), calibrated with 29 carefully selected geostandards. Analytical precision is better than 1% for Si and Al (Schnetger et al., 2000).

4.3.6 Infrared spectroscopy

For the Fourier transform infrared spectroscopy (FTIR) KBr pellets were prepared using a sample/KBr ratio of 0.0025. Spectra were taken from 4000 to 400 cm-1 usingg the Avatar 370 FT-IR

31 4 | Boron mullite: Formation and basic characterization

Thermo Nicolet spectrometer under nitrogen atmosphere. Background and sample spectra were obtained from 100 spectra each.

4.3.7 Thermal analyses

Differential thermal analyses (DTA) were performed on a Netzsch 449 F3 Jupiter STA (simultaneous thermal analyses) apparatus for two types of samples: the uncalcined precursor in order to investigate phase formation, and the product to investigate thermal stability. Samples were heated from ambient temperature to 1540°C (precursor) or 1500°C (product) with a heatinng rate of 10 °C/min (precursor) and 5 °C/min (product) and coolled down at 10 °C/min. For the product samples an isothermal segment of 2h at 1500°C was used. One long-term experiment was performed in order to investigate long-term stability of B-mullite at 800°C. Therefore the sample was kept at 800°C for 12 days, a small amount of the sample was withdrawn after 7 days in order to determine lattice parameters. All experiments were carried out in synthetic air atmosphere with a controlled gas flow of 20 or 50 mL/min. For each run 20 mg (80 mg for long-time experiment) of sample were heated in corundum crucibles, an empty corundum crucible was used as reference. All measurements were blank-curve corrected and evaluated with the NETZSCH software (NETZSCH, 20110).

4.4 Results and discussion

4.4.1 Boron-mullite formation from sol-gel derived precursors

Thee results of the DTA measurements of the precursor materials with difffferent initial boron contents are given in Fig. 4-2. There is one prominent exothermic DSC signall between 965 and 1000°C, corresponding to the formation of mullite. The formation temperature linearly decreases for samples with increasing amounts of boron (Fig. 4-3). By doping the gel with 177.6 mol.% B2O3 the mullite-formation-temperature was reduced by 32°C. A reduction of 41°C for thee mullite-formation- temperature was also reported by Zhang et al. (2010) for boron-doped single phase mullite precursors with constant Si/Al ratio, which is in good accordance with the results of this worrkk. The reduction of the formation-temperature is explained by a redduction of glass phase viscosity and therefore enhanced atomic diffusion. Hong et al. (1996) investigated the formation-temperature of diphasic mullite precursors and observed a reduction from 13550°C to 1200°C for mullite formatiion by doping with

5 wt.% B2O3 (Si/Al~3). They explain the reduction by increased alumina solubiliity. However, these results cannot be directly compared with this work or the work of Zhang et al. (2010) which are based on single phase mullite precursors that transforrm to mullite below 1000°C. The formation of B-mullite from the precursor was studied here by in situ high--temperature X-ray diffraction for a sample containing 5.3 mol.% B2O3. Prior to the experiment the sammple was heated to 600°C in order to remove organic componennts and nitrates. In Fig. 4-4 the diffraction patterns are given for temperatures between 600 and 1200°C. The material is amorphous to X-rays up to about

32 4 | Boron mullite: Formation and basic characterization

925-950°C. In the diffraction pattern taken at 950°C first broad diffraction peaks for mullite and Ȗ- alumina can be observed which is in good accordance with the DTA results. The broad diffraction peaks for Ȗ-alumina start to become smaller above 1100°C and almost disappears around 1200°C, complete disappearance can be expected after 5h at 1200°C, as observed for samples with the same composition and 5h annealing.

Fig. 4-2: DTA signals of STA measurements as a function of Fig. 4-3: Linear dependence of mullite formation temperature temperature for a series of precursor samples with different (exothermic DTA signal) from initial boron content. R² from B2O3 contents. linear fit is 0.97.

Fig. 4-4: X-ray diffraction patterns recorded between 600 and 1200°C of a sample containing 5.3 mol.% B2O3. Diffracted intensities are given in gray scales and were interpolated between single measurements (horizontal lines) using the program multi surface (Messner, 2013). For specific temperatures original diffraction patterns (red) are superimposed. The DTA signal is given on the right.

33 4 | Boron mullite: Formation and basic characterization

The mullite diffraction peaks get narrower with increasing temperature, additional sharpening of mullite diffraction peaks was also observed after heating samples for 5h at 1300°C and 1400°C (Fig. 4-5, bottom) and especially after heating the samples at 1400°C for 90h (Fig. 4-5, top) Rietveld refinements confirmed an increase of crystallite size upon long-term heating. The average crystal size for samples synthesized at 1200°C from Rietveld refinements is between 40 and 100 nm which is still very small and leads to significantly broadened diffraction peaks. Several experiments regarding the experimental conditions were performed but no significant influence on the resulting crystallite size was observed. Also longer annealing at 1200°C (48h) does not lead to the formation of larger crystals, which is in good accordance with the results of Schmücker et al. (2005b) who state that mullite grain sizes show little change up to 1500°C.

Fig. 4-5: Diffraction peaks (120) and (210) of B-mullite (1 mol.% B2O3) synthesized at 1200°C, subsequently heat treated for 5h at 1300°C, 1400°C (top), pressed to pellets, and heated for 90h at 1400°C (bottom).

SEM images taken from a boron-free mullite sample and a sample containing 7.5 mol.% B2O3 are given in Fig. 4-6. Both samples consist of particles with sharp edges; the particle shape of the amorphous precursor material is retained for the mullite particles. The individual particles measure up to ~50 μm and consist of mullite crystals measuring about 50 nm. In the boron-free sample some elongated mullite crystals were observed on the particle surface, measuring several hundreds of nm.

34 4 | Boron mullite: Formation and basic characterization

Fig. 4-6: SEM images. Top: boron-free mullite. Botttom: sample containing 7.5 mol.% B2O3 in the gel. Both samples were synthesized at 1200°C followed by 5h heat treatment at 1300°C and 1400°C.

4.4.2 Phase formation

The qualitative phase composition was determined by carefuul visual investigation of the X-ray powder diffraction patterns. Phases were identified according to characteristic diffraction peaks of five phases: A (B-) mullite phase (space group Pbam), three different alumina phases (Į, Ȗ, ș), and the aluminum borate Al18B4O33. According to Chakraborty (2008) and Chakraboorty and Das (2003) an Al-Si spinel phase can be formed during sol-gel syntheses of mullite that cannoot be distinguished from

Ȗ-Al2O3 by means of X-ray diffraction. Therefore we refer to Ȗ-Al2O3 although the existence of a spinel phase containing Si cannot be excluded. Furthermr ore the distinction bettween ș-alumina and Ȗ- alumina is not always clear, most Rietveldd refinements turned out to yield a better fit using ș-alumina although the most characteristic diffractiion peak was closer to the Ȗ-alumina position. Several diffraction patterns show a broad diffraction peak at about 20.2 °2ș (d = 4.40 Å) which is assigned to poorly crystalline Al18B4O33. As the presence of a silica phase was reported (Griesser et al., 2008) we carefully checked the possible occurrence of silica in our samples. Therefore a sample from the pure silica source (TEOS) was prepared at 12000°C and the presence of a crystallinne silica phase could be excluded.

35 4 | Boron mullite: Formation and basic characterization

Diffraction patterns of representative samples can be found in Fig. 4-7, characteristic diffraction peaks for phase identification are labeled. The three alumina phases can be distinguished using the reflections between 43° and 46°2ș (d = 2.10-1.97 Å), Al18B4O33 can be identified by the presence of the 20.3°2ș, 23.1°2ș, and 23.7°2ș (d = 4.37, 3.87, and 3.75 Å) reflections and the splitting of the 16.4°2ș (d = 5.40 Å) mullite-reflection into two reflections at 16.4° and 16.7°2ș (d = 5.40 and 5.31 Å). The qualitative phase compositions for a series of samples with initial boron contents up to

25 mol.% B2O3 are summarized in Fig. 4-8. For the syntheses at 1200°C the major phase can be well described with the metric of the mullite unit cell (Pbam). (B-)Mullite is the only abundant phase up to about 10 mol.% B2O3. Samples containing 10 mol.% B2O3 or more contain increasing amounts of Ȗ- alumina. Only diffraction patterns of samples with more than 23 mol.% B2O3 show the characteristic reflections of Al18B4O44 (Fig. 4-7). Upon heating the samples for 5 h at 1300°C and 5h at 1400°C transformation of the alumina phase from Ȗ-alumina to ș-alumina and Į-alumina is observed. After heating the uniaxially pressed samples for 90 h at 1400°C mullite and Į-alumina become the only abundant phases. The amount of Į-alumina increases with increasing boron content and for the silicon free sample Į-alumina becomes the only present phase, thus indicating the decomposition of B-mullite and Al-borate to boron-free mullite and Į-alumina.

Fig. 4-7: Diffraction patterns of B-doped mullite samples containing 15 and 23 mol. % B2O3, syntheses were performed at 1200°C, additional heating at 1300 and 1400°C.: m = (B-)mullite, Į = Į-alumina, ș = ș-alumina, Ȗ = Ȗ-alumina A=Al18B4O33.

This means: Below 10 mol.% B2O3 all initial material is used to form a B-mullite phase, if more

B2O3 is present alumina occurs as an impurity. Consequently, 10 mol.% can be interpreted as a limit

36 4 | Boron mullite: Formation and basic characterization for the incorporation of boron. Long heat treatment at 1400°C leads to decomposition of B-mullite to boron-free mullite and the formation of Į-alumina. A series of samples with initial compositions used by Griesser et al. (2008) was synthesized at

1200°C. In contrast to the authors we could not find a (crystalline) SiO2 phase. NNo details are given by the authors regarding the SiO2 modification and no diffraction patterns are presented.

Fig. 4-8: Qualitative phase composition for a seriess of B-doped mullite samples. B2O3 refers to the initial gel composition. Mullite= (B-)mullite.

4.4.3 Lattice parameters, chemical composition

Lattice parameters for all samples were refined by the Pawley method using the Rietveld software TOPAS (Bruker, 2009). The results are given as a function of the iinitial boron content in Fig. 9. For boron-free mullite lattice parameter a ranges between 7.540 and 7.5557 Å, depending on the annealing temperature (Fig. 4-9) and corresponding well with that of 3:2 mullitte. For lattice parameter a the observed variation with increasing B2O3 reveals no systematic changes and the changes are within the range that can be expected for different Al/Si ratios in mullite (gray bars in Fig. 4-9, according to Fischer et al , 1996). The values for a are significantly smaller than expected for Al-rich

2:1 mullite, therefore formation of Al-rich mullite due to loss of B2O3 can be excluded. Lattice parameters b and c significantly decrease upon higher boron contents and clearly reside outside the expected range for Al/Si variation of undoped mullite. Considering the qualitaattive phase composition (see top of Fig. 4-9 and Fig. 4-8), the limit for boron incorporation was assumed to be about 10 mol.%

B2O3. But as the decrease of lattice parameters b and c proceeds between 10 and 18 mol.% B2O3, this limit is certainly higher for the B-mullite coexisting with Al2O3. However, a definite number cannot be given yet as an unknown part of the initial boron is likely to escape during the synthesis procedure. The treend of decreasing lattice parameter c upon higher boron contents was observed by all authors (Griesser et al., 2008; Zhang et al., 2010) and it can be used as a relative measure for the boron content. For the samples additionally annealed for 5h at 1300°C and 1400°C latttice parameter c gradu- ally increases (Fig. 4-9). The samples heated for an extended period (90 h at 1400°C) yield lattice

37 4 | Boron mullite: Formation and basic characterization parameters b and c significantly larger than for the same samples synthesized at 1200°C. In Fig. 4-9 b and c for these samples plot within the gray bars and thus indicate the presence of boron-free mullite which is in good accordance with the increasing amount of Į-alumina.

Fig. 4-9: Lattice parameters a, b (top) and c (bottom) of Fig. 4-10: Amount of Į-alumina calculated assuming that all mullite as a function of the initial B2O3 content. Black B2O3 evaporated and boron-free mullite and Į-alumina are triangles refer to 1200°C synthesis, red filled circles to heat present (line). The actual amount of Į-alumina was derived treatment at 1400°C for 90 hours. Unfilled symbols (only for from Rietveld quantifications (circles). c) represent samples annealed for 5h at 1300°C (blue triangles) and 1400°C (orange circles). Gray bars represent the variation in lattice parameters expected due to Al2O3 variation in mullite according to Fischer et al. (1996). Qualitative phase composition for both series (1200°C and 90h 1400°C) is given at the top.

The amount of alumina could be a measure of the amount of escaped boron and therefore a measure for the amount of boron incorporated into mullite. For Į-alumina and boron-free mullite a reliable Rietveld quantification was possible. In Fig. 4-10 the amount of Į-alumina is given as a function of the initial B2O3 content. The filled circles represent the results of Rietveld quantification for the samples annealed at 1400°C for 90h. Assuming that all boron escaped under these conditions, the theoretical Į-alumina amount was calculated (line in Fig. 4-10), resulting in a good correlation with the results of Rietveld quantification. Thus quantification of alumina phases represents a convenient tool to estimate the amount of boron in the sample. Unfortunately for the 1200°C samples a reliable quantification of Ȗ-alumina and ș-alumina is difficult and an additional mistake would be made by using the structural model of boron free mullite for the refinement of the B-mullite phase. Theoretical calculations were done based on a series of initial bulk compositions where Si is replaced by B (corresponding to the samples synthesized in this work). The amount of impurity

38 4 | Boron mullite: Formation and basic characterization alumina was calculated for different bulk compositions assuming that varying parts of the boron escape during synthesis and no amorphous phase is present. Calculations were performed for two different assumptions: (1) B replacing Si and (2) B replacing Al, both based on the same initial bulk composition. If only 50% of the initial boron is left in a sample originally containing 5.5 mol.% B2O3 the resulting amount of impurity alumina is 10 wt.% if B replaces Si (case 1). If B replaces Al (case 2), 24 wt.% impurity alumina has to be present. The above calculations show that the amount of alumina for case 1 increases with the amount of volatile boron. But the impurity alumina amount is much higher if B replaces Al. As the observed amount of impurity alumina for the 1200°C samples is significantly lower than expected for case 2, we conclude that B replaces Si rather than Al. The Rietveld quantification using structural models of boron-free mullite and ș-alumina reveals significant uncertainties. Nevertheless, based on the Rietveld quantifications the amount of boron escaped during

1200°C syntheses was calculated for samples initially containing 10-15 mol.% B2O3. According to these calculations 28-35% of the initial boron escapes during 1200°C syntheses, yielding B-mullite compositions with y = 0.28…0.41 corresponding to the following stoichiometry: Al4.5Si1.5-2yB2yO9.75-y. For the calculations it is assumed that all Si + B is incorporated in the mullite phase and that Si + B = 1.5. However, these calculations yield rather large errors and therefore can only be used for a very rough estimation of the actual chemical composition. Bulk chemical analyses were performed on four single-phase samples by means of XRF; the results are given in Table 4-2. Al2O3 and SiO2 can be directly analyzed; the amount of B2O3 was calculated from the weight loss during preparation of the glass beads. The analyzed Si/Al ratio is reproduced with less than 1% deviation compared to the initial weights. However, the calculated amount of B2O3 is distinctly smaller than expected. Although the EDX spectra for boron-rich samples yielded small peaks around the spectral line for boron, no reliable quantification of boron is possible. However, the analyzed Si/Al ratio is in very good accordance with XRF results and the gel composition.

Table 4-2: Initial compositions in the gel and the compositions calculated from XRF results are given. The difference to 100% from XRF was calculated as B2O3.

initial gel composition XRF

Al2O3 [mol.%] B2O3 [mol.%] SiO2 [mol.%] Al/Si Al2O3 [mol.%] B2O3 [mol.%] SiO2 [mol.%] Al/Si 60.2 0.0 39.8 3.0 60.1 0.0 39.9 3.01 60.8 1.0 38.2 3.2 61.3 0.2 38.4 3.19 61.9 4.1 34.0 3.6 64.0 1.0 35.0 3.65 63.9 6.4 29.8 4.3 66.7 2.0 31.3 4.27

FT-IR spectra were taken from four (B)-mullite samples with 0 - 6.4 mol.% B2O3. The Si-O and Al-O absorption bands were assigned according to Voll et al. (2002). Griesser et al. (2008) assigned the bands between 1407 and 1278 cm-1 and 695-670 cm-1 to the B-O stretching and bending vibrations in the BO3 unit. Fig. 11 shows the FTIR spectra with clearly increasing intensities for the BO3 band at

39 4 | Boron mullite: Formation and basic characterization

1325 cm-1 for samples with increasing boron content. According to Griesser et al. (2008) a band for

B-O stretching vibration in the BO4 unit would be expected between 1095 and 1025 and cannot be found in the data presented in Fig. 4-11. The BO3 vibrations were also reported for boron-doped mullites (Griesser et al., 2008; Zhang et al., 2010) but in contrast to this work and the work of Griesser et al. (2008), Zhang et al. (2010) describe an IR absorption band at 1090 cm-1 which they attribute to the stretching vibration of a BO4 group. It can be concluded from the IR data of this work that boron is present in samples with increasing amounts for increasing initial boron contents and that boron occurs in threefold rather than fourfold coordination.

Fig. 4-11: FT-IR spectra of four B-doped mullite samples containing between 0 and 6.4 mol.% B2O3 in the gel. The absorption bands were assigned according to Voll et al. (2002) and Griesser et al. (2008).

Fig. 4-12: Thermogravimetric signals for three B-mullite samples containing between 1.0 (black) and 6.4 (red) mole % B2O3. Grey bar: isothermal segment 2h at 1500°C. Diffraction patterns 1-3: from the three samples after STA analyses, patterns 4+5: from a sample containing 5.2 mole % B2O3 heated for 5h and 90h at 1400°C.

40 4 | Boron mullite: Formation and basic characterization

4.4.4 Thermal stability of B-mullites

The phase composition (Fig. 4-8) and the development of lattice parameters (Fig. 4-9) for the B- mullite samples heated for an extended perriod (90 h 1400°C) indicated the decomposition of B-mullite and formation of Į-alumina and boron-free mullite. Thermal analyses of three B-mullites with different compositions (synthesized at 1200°C) were performed between room temperature and 1500°C with a 2h isothermal segment at 1500°C (Fig. 4-12). The TG signal iindicates a weight loss slowly starting at 700°C with a steep sloppe between 1300°C and 1500°C. The weight loss is highest for the sample with the highest amount of boron. But the weight loss is not commppleted yet, indicated by the break in the slope at the end of the isothermal segment and an additional weight loss in subsequent thermal analyses at 1500°C. This also explaains that the weight loss does not equal the expected weight loss due to the release of boron. Longer issothermal segments at 1500°C were nnot performed to avoid any damage of the instrument. The diffraction patterns in Fig. 4-12 support the theory of incomplete decomposition of B-mullite. Pattern 4 and 5 represent samples after heat treatment for 5h and 90h at 1400°C, Į-alumina is only present in the sample heated 90h whereas the diiffraction peaks of the mullite phase shift to values typical for boron-free mullite. Thus complete decomposition of B-mullite has taken place only after 90h at 1400°C. TThe diffraction patterns recorded aftteer the thermal analyses (2h at 1500°C) show a significant amount of Į-alumina for the boron richest saample only (6.4 mol.%

B2O3, pattern 3) but still being less than would be expected due to complete decomposition of

B-mullite. In pattern 2 (4.1 mol.% B2O3) only a very small peak indicates the presence of a very small amount of Į-alumina and no Į-alumina is present in pattern 1 (1.0 mol.% B2O3). If decomposition during thermal analyses had been completed, lattice parameter c would be expected to yield values typical for boron-free mullite (gray bars in Fig. 4-99) as it shows the most distinct changes upon boron incorporation. For the samples after the DTA measurements (2h at 1500°C) lattice parameter c yields values distinctly higher than for the syntheses at 1200°C but still significantly lower than expected for boron-free mullite. Therefore it can be concluded from DTA and XRD measurements that the decomposition has not been completed yet. Although B-mullite is not stable at 1400°C for an extended period, the changes in phase composition and lattice parameters are rather small for samples exposed to high temperatures (1400°C, 1500°C) for a short period (5 h and 2 h) only. Richards et al. (1991) investigated the microstructure development in mullite fibers and found that 35% of the B2O3 was retained after 60 h at 1400°C. According to Hong ett al. (1996) 60 wt.% of the B2O3 remained in the 5 wt.% boria-doped sample after sintering the pellets at 1650°C for 5 h. The authors explain the higher amount of boron retained in the sample despite higher sintering temperature by the differences in sample dimension and geometry. This is in good accordannce with the observation that complete decomposition of the powdered or loosely pressed samples of this work takes place during 90 h at 1400°C.

41 4 | Boron mullite: Formation and basic characterization

Thee long-term stability of B-mullites at 800°C was investigated for a sample synthesized at

1200°C initially containing 6.4 mol.% B2O3. Therefore the weight loss at 800°C was recorded for a period of 12 days using the STA. X-ray diffraction patterns were recorded before and after the long- term experiment, in addition a pattern was recorded from a small amount of sample that was withdrawn from the DTA after 7 days at 800°C. For both runs (1. run: 7 days, 2. rruun: another 5 days) the weight loss was only 0.1 mg which is less than the maximum drift of the sccales (<5μg/h). No significant changes in the lattice parameters were observed comparing the diffraction patterns recorded before, during, and after the long-term experiment. It can therefore be concluded that B-mullite is stable for a long period at 800°C and no boron is lost under these condittiions (synthetic air atmosphere). No significant differences in lattice parameter c were observed for two samples with the same composition (5.3 mol.% B2O3 in the gel) synthesized at 1200°C for 5h and 48h, respectively. It can therefore be concluded that B-mullites do not decompose if exposed at 1200°C for a moderate period.

4.4.5 Thermal expansion

Thee development of lattice parameters between ambient temperature and 12200°C was investigated for three single-phase samples with difffeerent boron contents up to 6 mol.% B2O3. In addition a sample with impurity alumina was investigated, initially containing 13.6 mol.% B2O3. All diffraction patterns from the HT-XRD experiments were Pawley refined using the Rietveld software TOPAS (Bruker, 2009). The temperature dependent lattice parameters between ambient temperature and 1200°C are given in Fig. 4-13. For better comparison to literature data, thermal expansion coefficients were calculated from the gradient between 300 and 1000°C, the results are given in Table 4-3. The thermal expansion is largest in the crystallographic b direction, followed by the c-axis and is smallest in a direction, which is in agreement with literature data for undoped and Cr-doped mullites (Schneider and Eberhard, 1990; Brunauer et al., 2001). With increasing amounts of boron the thermal expansion coefficients become smaller. The mean thermal expansion coefficient iis 5.9(1) × 106 °C-1 6 -1 for pure mullite and 5.0(1) × 10 °C for the sample initially containing 13.6 mol.% B2O3, which means a reduction in thermal expansion of aboout 15% and a reduction of the expanssion of the unit-cell volume of 17%. The generally lower thermal expansion of B-doped mullite in comparison to undoped mullite can be explained by the substitution of Si by B, producing short and stronng B-O bonds. For single axes the reduction is 4% (a), 21% (b), and 17% (c) respectively. In comparisoon to literature data for undoped mullite (Schneider and Eberhard, 1990; Brunauer et al., 2001) thee expansion for the undoped mullite of this work in a direction is laarger than in literature, whereas in b and c direction it is about the same. However, it should be noticed that the variation of the absolute thermal expansion coefficients in literature is quite high, varying between 6.0 × 10-6 °C-1 and 7.0 × 10-6 °C-1 for Į(b) (Table 4-3). The relatively strong reduction of the thermal expansion of B-doped mullite with respect to undoped mullite further improves the excelllent thermal shock behavior of mulllite. This makes the

42 4 | Boron mullite: Formation and basic characterization material very interesting for a potential use as electron packaging devices and for catalyst supports in the temperature stability range of B-doped mullite, i.e. up to about 800°C (see e.g. Okada and Schneider, 2005). In Fig. 4-13 the volume of the unit cell is given as a function of temperature. The volume of the unit cell is smallest for the boron richest sample. Interestingly the boron free sample has a slightly smaller volume than the sample containing a small amount of boron. The thermally induced relative volume expansion of the unit cell is given as the slope of the linear fit between 300°C and 1000°C (Table 4-3). A reduction of 16% of the volume expansion was achieved by incorporation of boron into the mullite.

Fig. 4-13: Temperature dependent lattice parameters and unit-cell volumes for mullite samples with different boron contents. Linear regression fits are given between 300°C and 1000°C, the respective errors and R² can be found in Table 4-3.

43 4 | Boron mullite: Formation and basic characterization

Table 4-3: Thermal expansion coefficients and coefficient of correlation (R²) for linear regression fits between 300 and 1000°C. Sample with initially 13.6 mol.% B2O3 contains alumina impurities. Volume expansion coefficient is given as the slope of linear fit. Estimated standard deviations (e.s.d.) are given in parenthesis. Literature data for undoped and Cr-doped mullite is given (Schneider and Eberhard, 1990; Brunauer et al., 2001).

-6 -1 B2O3 in gel [mol.%] Thermal expansion coefficients [× 10 °C ] Volume expansion Į (a) Į (b) Į (c) Į¯ * coefficient [× 10-3 Å3 °C-1] 0.00 4.8(1) 6.8(1) 6.0(1) 5.9(1) 2.96(3) R² 0.998 0.999 0.999 0.999 1.0 4.7(1) 6.01(4) 5.7(1) 5.5(1) 2.78(2) R² 0.998 1.000 0.999 0.999 6.4 4.5(1) 5.8(1) 5.3(1) 5.2(1) 2.62(3) R² 0.992 0.999 0.998 0.998 13.6 (alumina!) 4.6(1) 5.3(1) 5.0(1) 5.0(1) 2.50(2) R² 0.996 0.999 0.999 0.999 Mullite, undoped 4.1 6.0 5.7 5.3 (Brunauer et al., 2001) 3.9 7.0 5.8 5.6 (Schneider and Eberhard, 1990) Cr-doped mullite 3.6 5.9 5.2 4.9 (Brunauer et al., 2001) 3.1 6.2 5.6 5.0 (Schneider and Eberhard, 1990) * Mean thermal expansion coefficients are given by the equation: Į¯ = (Į (a) + Į (b) + Į (c))/3

4.5 Conclusion

This work is based on the 1:1 substitution of Si by B and shows that more than 10 mol.% B2O3 can be incorporated into mullite. B-mullite is formed from the gel phase between 965°C and 1000°C. The mullite-formation temperature decreases linearly with the boron content in the precursor. The lattice parameter c strongly depends on the initial boron content of the precursor and can be used as a relative measure for the amount of boron in mullite. Up to 10 mol.% initial B2O3, pure B-mullite is present but the limit for boron incorporation must be higher as the downward trend of c with increasing boron content still proceeds beyond this limit. Above 10 mol.% B2O3, different alumina phases are present as impurities and above 23 mol.% B2O3 Al18B4O33 forms as well. The amount of alumina present in the sample can be used as a measure for estimating the amount of boron in the sample. According to such calculations 28-35% of the initial boron evaporates during the syntheses at 1200°C. This indirect method is necessary as a direct quantification of the boron content in mullite was not possible yet. The presence of boron in threefold rather than in fourfold coordination was verified by FTIR spectroscopy. Heating B-mullite at temperatures higher than the synthesis temperature (e.g. 1300°C and 1400°C) leads to slow decomposition and formation of boron-free mullite and Į-alumina. According to the results of DTA and XRD the decomposition is neither completed after 5h at 1400°C nor after 2h at 1500°C, but no boron is left after 90h at 1400°C. However, long-term stability at 800°C was proved by DTA and XRD and no significant changes in the lattice parameters occur after 48 h at 1200°C. Combined with the 15% reduction of the thermal expansion in comparison to boron-free mullite these facts make B-mullite a promising candidate for industrial applications like catalyst supports and

44 4 | Boron mullite: Formation and basic characterization electron packaging devices. With respect to possible applications the sintering behavior of boron- doped mullites was studied by Zhang et al. (2010). They found that the densification temperature is reduced as well as the bulk density with increasing amount of boron. The latter is explained by the light element boron and the increasing sample volume which is due to the higher amount of glass phase in the samples. According to Zhang et al. (2010) the high amount of glass phase is also the reason for reduced IR transparency of boron-doped mullites. For mullite and mullite-type materials the thermal expansion behavior is of great importance. Fisch and Armbruster (2012) recently investigated the structural evolution of synthetic mullite-type

Al5BO9 and Al4B2O9 with increasing temperature. The rigid body behavior of the BO3 units agrees well with the results of Gatta et al. (2010) who investigated the pressure-induced structural evolution of Al5BO9 and also give a comparative analysis of the elastic behavior of mullite-type materials. Details on the temperature-induced structural changes in B-doped mullite cannot be given yet as it was not possible to identify the B-position from powder X-ray diffraction data. Hence, more work will be necessary to study the crystal structure (neutron diffraction, NMR) of B-mullite and to investigate the mechanisms of decomposition and transformation of boron mullite. This includes the direct chemical analyses (PGAA), mass or infrared spectroscopy attached to a STA, and crystal structural investigations. The latter will also help to explain the anisotropic behavior of lattice parameters upon boron incorporation as well as their thermally induced anisotropic behavior. In addition a comparative study is planned on the temperature- and pressure-induced structural evolution of mullite doped with various cations. Acknowledgements We would like to thank the University of Bremen for financial support from the Central Research Development Fund. Bernhard Schnetger (Universität Oldenburg) for the XRF analyses, Petra Witte (Universität Bremen) for support with REM and EDX, and Ute Jarzak for the IR spectra.

5 Neutron diffraction and ¹¹B solid state NMR studies of the crystal structure of B-doped mullite

Hanna Lührs *;I, Anatoliy SenyshynII, Scott P. KingIII, John V. HannaIII, Hartmut SchneiderI,IV, Reinhard X. FischerI

I Fachbereich Geowissenschaften, Universität Bremen, Klagenfurter Straße, D-28359 Bremen, Germany II FRM II Technische Universität München, Forschungsneutronenquelle Heinz-Maier-Leibnitz (FRM II), D-85747 Garching, Germany III Department of Physics, University of Warwick, Coventry, CV4 7AL, UK IV Institute of Crystallography, Universität Köln, Greinstr. 6, D-50939 Köln, Germany

Published in: Zeitschrift für Kristallographie 228 (2013) 457-466 DOI: 10.1524/zkri.2013.1595

submitted: 3 December 2012 | accepted: 27 February 2013 | online: 21 May 2013

Keywords: Mullite | Crystal structure | Boron | Neutron diffraction | MAS NMR

The crystal structure of boron-doped mullite (B-mullite) was studied by means of neutron powder diffraction and 11B solid state, magic-angle-spinning (MAS) NMR spectroscopy. The samples were prepared from single-phase gels consisting of aluminum nitrate nonahydrate, tetraethoxysilane, and boric acid, annealed at 1200° C. Boron resides in a planar, trigonally coordinated position crosslinking the octahedral chains perpendicular to the c-axis. The B position was derived from a combination of Rietveld refinements with difference Fourier calculations and gridsearch analyses applied to a series of samples with varying B contents. Based on the sample with the highest B content the local BO3 configuration was determined by distance least squares (DLS) refinements resulting in split positions for two O atoms in the AlO6 octahedron which leads to a local distortion of the octahedral geometry. This model was verified by subsequent Rietveld refinements of several samples, details of the crystal structure are given for the sample with the highest B content. The model for B incorporation is corroborated by the MAS NMR results which clearly show that the 11B isotropic chemical shift (įiso) and quadrupolar coupling constant and asymmetry parameter (CQ, Ș) describe that of three-coordinate B in a near-trigonal planar environment. Furthermore this crystallographic model provides an explanation for the anisotropic behavior of the lattice parameters upon B-incorporation.

* Correspondence author (e-mail: [email protected])

47 5 | Neutron diffraction and ¹¹B solid state NMR studies of the crystal structure of B-doped mullite

5.1 Introduction

Mullite as a mineral is rare in nature but it is one of the most important phases in both traditional and advanced ceramics. The great technical importance of mullite ceramics can be explained by the outstanding properties of mullite: high thermal stability, low thermal expansion and conductivity, high creep resistance and corrosion stability. A comprehensive review on the structure and properties of mullite is given in Schneider and Komarneni (2005) and Schneider et al. (2008). The chemical composition of mullite is given by the solid-solution series Al2Al2+2xSi2-2xO10-x, with x ranging between

0.2 and 0.9, corresponding to about 58-94 mol% Al2O3 (Fischer et al., 1996). Main compounds are 3:2 mullite (x = 0.25) and 2:1 mullite (x = 0.40). A large variety of cations can be incorporated into the crystal structure of mullite depending on the synthesis temperature and atmosphere. There have been intensive investigations on the incorporation of transition metal cations into mullite which are summarized in two reviews (Schneider, 2005b; Schneider et al., 2008 and references therein ). It is known that transition-metal ions preferably enter the octahedral positions in mullite replacing Al. Maximum incorporations are observed for M3+ cations replacing Al3+ in the structure. The replacement of octahedrally coordinated Al3+ by cations with other oxidation states, e.g., Ti4+ and V4+ is less favorable, since it requires simultaneous substitution of Si4+ by Al3+. The incorporation of Ti3+, Ti4+, V3+, V4+, Cr3+, Mn2+, Fe2+, Fe3+, and Co2+ has been reported (Schneider, 1990). In small quantities, Eu2+, Eu3+, Zr4+, and Mo3+are also present in mullite structures (Schneider, 1986b; Tomsia et al., 1998; Kutty and Nayak, 2000). Besides transition metals the incorporation of significant amounts of Ga3+ (favorably at octahedral position, see Schneider, 1986a) and B3+ (Griesser et al., 2008) have been reported. Also Si4+ can be replaced by Ge4+, which opens the wide field of Ge-mullite-type compounds (Schneider and Werner, 1982). Caballero and Ocana (2002) mention that small amounts of Sn4+ can be incorporated in mullite. For Na+ and Mg2+ interstitial and octahedral incorporation is suggested, respectively (Schneider, 1984, 1985).

An existence of a solid-solution series between 3:2 mullite (Al4.5Si1.5O9.75) and Al18B4O33 by substitution of B for Si in mullite was suggested (Dietzel and Scholze, 1955) due to similarities in the physical properties. Recent investigations show that mullite can incorporate a large amount of boron but that there is no complete solid solution series between mullite and mullite-type Al borates (Griesser et al., 2008). However, B-doping of mullite results in significant changes of lattice parameters b and c as reported by several authors (Griesser et al., 2008; Zhang et al., 2010; Lührs et al., 2012, chapter 4 in this thesis). While the lattice parameter a is linearly correlated with the Al/Si ratio in pure aluminosilicate mullite (Fischer et al., 1996), no significant variation in a is observed for B-doped mullites. Furthermore the incorporation of B results in a strong reduction of the mean thermal expansion coefficient of 15% (Lührs et al., 2012, chapter 4). The anisotropic behavior of lattice parameters with increasing boron content as well as with increasing temperature have not been explained crystallographically so far as the structure model for B-mullite was not known. The low

48 5 | Neutron diffraction and ¹¹B solid state NMR studies of the crystal structture of B-doped mullite scattering factor for X-rays and the low weight of boron aggravate crystal-structure investigations and chemical analyses of B-mullites. However, the use of 11B solid state NMR does not suffer from the same pproblems; 11B is a very sensitivve nucleus with a well-characterized chemical shift range, and the quadrupole parameters from this spin I=3/2 nucleus are well understood. As such, it is a routinely studied NMR active nucleus in the solid state (MacKenzie et al., 2007)). Typically, observable differences in the 11B isotropic chemical shift (įiso) are critical indicators of the coordination number of the B position in nearest- neighbor oxo environments. Further structural insights can be elucidated from the electric field gradient tensor of this spin I=3/2 quadruppolar nucleus, making it possible to analyze the local point symmetry of the sites occupied by B. The substitution process is assumed to occur according to the following equationn: 2 Si4+ + O2- l 2 B3+ + Ƒ (2) Here we present the results of neutron powder diffraction experiments oon a series of B-doped mullites including Rietveld refinements, ddifference Fourier and gridsearch analyses, distance least squares calculations and 11B solid state MAS NMR.

5.2 Experimental

5.2.1 Synthesis of B-doped mullite

The samples were prepared from single-phase gels consisting of aluminum nitrate nonahydrate, tetraethoxysilane, and boric acid, followwing the procedure given by Griesser et al. (2008) with modified gel compositions to account forr the Si-B substitution as described by Lührs et al. (2012, chapter 4). Transparent sols were prepared at 60°C by suspending the chemicalls in pure ethanol. After gelation for 4 days at 60°C the transparent gels were dried at 150°C and ground to fine powders which were calcined for 5h at 350°C before annealing for 5h at 1200°C in air. The neutron diffraction samples were synthesized using 99 atom% 11B boric acid (Aldrich) in order tto account for the very high absorption cross section of the isotope 10B. Selected neutron diffraction samples were additionally heated for 5h at 1400°C in order to increase crystallinity and avoid broadening of diffraction peaks. As this leads to the release of B and the formation of additiional alumina (Lührs et al., 20012, chapter 4) only the B-free samplle heated at 1400°C is reported in this work. For the NMR measurrements samples with the same initial compositions as samples I (NMR-I), II (NMR-II), and V (NMR-IV) were synthesized as well as an intermediate composition between samples II and IV (NMR-III). Whenever a chemical composition in mol% is given within this worrk it refers to the initial gel composition with Al2O3 + SiO2 + B2O3 = 100 %. Due to a partial loss of B during the synthesis and the associated formation of alumina impurities the actual composition of the mullite phase can be different (Lührs et al., 2012, chapter 4). In Table 5-1 the initial compositions and synthesis temperatures of all investigated samples are listed.

49 5 | Neutron diffraction and ¹¹B solid state NMR studies of the crystal structure of B-doped mullite

Table 5-1: Initial chemical composition of all investigated samples.

synthesis initial gel composition temperature Al2O3 B2O3 SiO2 [mol%] [mol%] [mol%] Neutron diffraction samples I 1400°C 60.00 0.00 40.00 II 1200°C 60.58 1.01 38.41 III 1200°C 62.16 3.11 34.73 IVa 1200°C 63.26 5.27 31.47 Va 1200°C 63.91 6.39 29.70 VIa 1200°C 65.02 7.59 27.39

MAS-NMR samples NMR-I 1200°C 60.01 0.00 39.99 NMR-II 1200°C 60.76 1.01 38.23 NMR-III 1200°C 61.89 4.12 33.99 NMR-IV 1200°C 63.86 6.38 29.76 a Samples contain alumina impurities

5.2.2 Neutron powder diffraction and structure refinements, difference Fourier synnthesis, grid search

5.2.2.1 Neutron powder diffractometer SPODDI Thee neutron diffraction experiments at room temperature were carried out in vanadium vessels on the high-resolution powder diffractometer SPODI at FRM-II, Germany (Hoelzel et al., 2012). A collimated beam of thermal neutrons (O=1.54812 Å) has been obtained at a take-off angle of 155° using the (551) reflection of a vertically focuusing composite germanium monochhrromator. Details of data collection and crystal data are listed in Table 5-2.

Table 5-2: Data collection parameters for all samples, refinement parameters for B-mullite sample VI.

Crystal data a [Å] 7.564902(1) b [Å] 7.681806(1) c [Å] 2.875613(4) V [Å3] 167.113425 Space group Pbam (No. 55) Z 1 a Chemical formula Al4.64Si1.16B0.2O9.58 Diffractometer SPODI at FRM-II Wavelength [Å] 1.54812 Monochromator Ge(551) Data collection temperature room temperature 2ș range [°] 1-152 Step size in 2ș [°] 0.05 Refinement parameters number of refined parameters 24 Rp´b [%] 9.50 Rpc [%] 3.55 Rwpd [%] 4.52 RB(B-mullite)e [%] 1.97 a chemical formula of B-mullite phase from Rietveld refinements b Rp‘ = Ȉi|yio - yic| / Ȉi|yio - yib| yio/yic/yib = observed/ calculated/ c Rp = Ȉi|yio - yic| / Ȉi yio background step-intensity d 2 2 1//2 Rwp = (Ȉiwi(yio - yic) / Ȉiwiyio ) Iko/Ikc = observed/calculated integrated e RB = Ȉk|Iko - Ikc| / Ȉk Iko intensity of reflection k w = weighting factor

50 5 | Neutron diffraction and ¹¹B solid state NMR studies of the crystal structure of B-doped mullite

5.2.2.2 Rietveld refinements and difference Fourier synthesis All Rietveld refinements and difference Fourier calculations were performed using the Rietveld software BRASS (Birkenstock et al., 2012). Instrumental profile parameters were derived by refinement of a Si-standard and used for all subsequent refinements. The crystal structure of the boron- free sample was refined using a 3:2 mullite structure model (Saalfeld and Guse, 1981, unit cell transformed to the setting used by Angel and Prewitt, 1987) and varying the following parameters: zero point, scale factor, lattice parameters, atomic positions, occupancy of T* (and T, O3 and O4 as dependent parameters), and iPG for the Gaussian peak width. The resulting model was then used as a first approximation for the refinement of the B-doped samples with subsequent difference Fourier analysis in order to find missing scattering matter. Alumina impurities in samples IV to VIII were simulated and refined using structure models of theta and alpha alumina, respectively.

5.2.2.3 Grid search analysis As a second method to locate missing scattering matter the gridsearch approach was applied. This method was introduced by Baur and Fischer (1986) as an alternative tool to the Fourier methods biased by termination effects. The gridsearch method is implemented in the BRASS program package. To locate scattering matter a dummy atom is shifted on a grid within the asymmetric unit, at each position in the grid the occupancy of a dummy atom is refined and recorded together with the respective residuals (Rwp and RB). A maximum in the site occupancy factor (SOF) and minimum in the residuals of the grid search analysis indicates a missing atom in the structural model at this position. The method has already been successfully applied to Cr-mullite, where the Cr atom could be located in the octahedral site by both Fourier and grid search analyses (Fischer and Schneider, 2000) and was also successfully used to locate missing O atoms in Al4B2O9 (Fischer et al., 2008). For all neutron diffraction patterns a dummy O atom was shifted in steps of 0.02 in x, y, and z direction, yielding 35937 data points with refined occupancies and residuals.

5.2.2.4 Distance least squares (DLS) refinements Due to enlarged B-O distances after the Rietveld refinements and the lack of a threefold coordinated position within the mullite structure with appropriate B-O distances, DLS refinements were necessary. Therefore the program DLS-76 (Baerlocher et al., 1978) was used to refine the local environment of B in the B-mullite crystal structure with geometric restraints for the interatomic distances. Lattice parameters were fixed to the refined values from Rietveld analyses as well as the atomic coordinates for Al, B, and the O positions of the octahedra not belonging to the first two coordination spheres of B. The prescribed distances and respective weights are listed in Table 5-3. The

Al-O distance for AlO6 of 1.91 Å is the mean distance in 3:2 mullite (Saalfeld and Guse, 1981), 2.70 Å is the O-O distance assuming a regular octahedron. Taking the ideal B-O distance (1.37 Å ) from

Hawthorne et al. (1996), recently confirmed by Gatta et al. (2010) for mullite-type Al5BO9 and

51 5 | Neutron diffraction and ¹¹B solid state NMR studies of the crystal structure of B-doped mullite

hambergite (Gatta et al., 2012), the O-O distances in the regular BO3 trianglle are 2.37 Å. The refinement yielded a low residual of 0.014 and B-O distances of B-O11 = 1.37 Å and B-O21 = 1.42 Å.

Table 5-3: Prescribed atomic distances and respective weights used for DLS refinements.

prescribbed distance d0 [Å] weight w Al-O (AlO6) 1.91 0.5 O-O (AlO6) 2.70 0.3 a B-O (BO3) 1.37 1.5 O-O (BO3) 2.37 0.3 Rb 0.014 a Hawthorne et al. (1996) b 2 2 1/2 R = (Ȉ(w(d0-d)) /Ȉ(w d0) ) with d0 and d = observed and calculated distances, respectively

5.2.3 NMR spectroscopy

11 All B solid state MAS NMR measurements were performed at a B0 field of 14.1 T using a Bruker Advance II+ 600 spectrometer operatiing at a Larmor frequency of 192.3 MHz. Single pulse experiments were facilitated using a Bruker 4 mm HX MAS probe which enabled MAS frequencies of 12 kHz to be achieved to remove broadenings from quadrupolar and dipolar interactions. Pulse time calibrations were performed on NaBH4 and a ‘nnon-selective’ (solution-like) ʌ/2 pullse time of 6 ȝs was obtained which corresponded to a ‘selective’ (solids) pulse time of 3 ȝs; all experiiments used a pulse time of 1 ȝs corresponding to a tip angle of ʌ/6. A recycle delay of 6 s was common to all experiments. An additional experiment with an empty MAS NMR rotor was performed in order to subtract the background signal arising from the boron nitride stator material in the probe. The acquired 11 B MAS NMR data were referenced to the IUPAC primary standard BF3. Et2O aat įiso 0 ppm, via a secondary solid reference of NaBH4 at įiso -42.06 ppm (Hayashi and Hayamizu, 1989). Experimental results were simulated using the DmFit software in order to extract information on the quadrupole and isotropic chemical shift parameters (Massiot ett al., 2002). 2D Triple-quantum, magic-angle-spinning (3QMAS) data were obtained using the amplitude modulated Z-filter experiment (p1 - t1(3Q evolution) - p2 - Ĳ - p3 - t2(acquire)) (Amoureux et al., 1996).

The optimized pulse lengths of the triple quantum excitation (p1) and reconversion (p2) pulses were p1

= 3.6 ȝs and p2 = 1 ȝs, respectively, implementted with an RF power of 125 kHz, while the soft ʌ/2 Z- filter (p3) pulse was set to 22 ȝs which was delivered with an RF power of 11 kHz. In all 3QMAS measuremments ȣr = 12 kHz, the Z-filter Ĳ period was set to 20 ȝs and the recycle delay was 6 s.

5.3 Results

5.3.1 11B MAS NMR spectroscopy

Thee presence of threefold coordinated B was initially proposed by FT-IR speectroscopy (Lührs et al., 2012, chapter 4) and is corroborated by the 11B MAS NMR study in this work. From the 11B MAS NMR data of Fig. 5-1 (a)-(c) and the summary of measured NMR parameters in Table 5-4 acquired

52 5 | Neutron diffraction and ¹¹B solid state NMR studies of the crystal structure of B-doped mullite from B-substituted mullites of variable composition, it can be observed that isotropic chemical shift nd 11 values of įiso 21 - 22 ppm consistently characterize these 2 order quadrupole broadened B resonances. This įiso value can be unambiguously assigned to BO3 moieties in the structure. The simulation of these data (as in Table 5-4) consistently shows that these lineshapes represent a quadrupole coupling constant (CQ) of 2.59 MHz, which is also indicative of a threefold coordinated B environment. In addition, a low ȘQ value of 0.15 suggests that the B site possesses a near-three fold axis of rotation, as expected for a slightly distorted trigonal planar BO3 environment. This contrasts markedly with BO4 units which are typically characterized by resonances located further upfield at įiso

~0 - 10 ppm and much smaller CQ values of ~0 – 0.5 MHz.

Fig. 5-1: 11B single pulse MAS NMR measurements of (a) NMR-IV, (b) NMR-III, (c) NMR-II mullite samples carried out at 14.1 T with ȣr = 12 kHz. The experimental data are displayed with the simulation of the each spectrum (above). (d) 2D 3QMAS result of the NMR-IV sample also acquired at 14.1 T, ȣr = 12 kHz.

Table 5-4: Summary of 11B solid state NMR experimental parameters from simulated fits of experimental data.

įiso [ppm] CQ [MHz] ȘQ Environment NMR-II 22.1(5) 2.6(1) 0.15(1) BO3 NMR-III 21.9(5) 2.6(1) 0.15(1) BO3 NMR-IV 21.8(5) 2.6(1) 0.16(1) BO3

It is interesting to note that as the nominal compositions of the samples varies from 1.0 to

6.4 mole% B2O3, no change to the immediate B environment can be detected through the measured

įiso, CQ and ȘQ values, thus suggesting that the local (short range) BO3 configuration is largely invariant to the degree of B incorporation. This agrees well with MacKenzie et al. (2007) who found that the three different BO3 environments in Al6-xBxO9 (x=1-4) are not distinguishable by means of

53 5 | Neutron diffraction and ¹¹B solid state NMR studies of the crystal structure of B-doped mullite

NMR. Furthermore, from the 2D 3QMAS data of Fig. 5-1 (d) it can be observed that some disorder is associatedd with the local BO3 environment. The contours describing this BO3 resonance are not aligned parallel with the F2 (or MAS/ppm) axis, and the quadrupole lineshape singularities cast an unaligned projection spectrum on the F1 axis. This result suggests that a very smalll distribution of B- O bond lengths and bond angles exists, which manifest itself as corresponding small dispersion of the

įiso, CQ and ȘQ parameters. These geometric variations will be discussed in more detail in the Rietveld and DLS Refinement section below. 29Si and 27Al MAS NMR measurements revealed results very similar to the pprevious reports on B-free mullites (Merwin et al., 1991; Rehak et al., 1998; Schmücker et al., 2005a). Furthermore, these 29Si and 27Al MAS studies showed no significaannt variation with B content and were thus not diagnostic of the B incorporation mechanism. This can be explained by the low amount of B in the structure of less than 3 mol% B2O3, corresponding to onnly one B atom per five unit cells. In contrast to that,

MacKenzie et al. (2007) investigated mullite-type aluminum borates with up to 66 mol% B2O3 and observed significant changes of the 27Al NMR signal with B content.

5.3.2 Neutron diffraction

In thhe neutron diffraction patterns no major changes due to increasing borron content can be observed (Fig. 5-2). The most distinct changes in the diffraction patterns are duee to the presence of alumina impurities resulting from the loss of B during the synthesis (Lührs et al., 2012, chapter 4).

Fig. 5-2: Neutron diffraction patterns of all mullite samples doped with different amounts of boron. Characteristic diffraction peaks for Al2O3 phases that occur as impurities are marked; samples containing impurities are marked with an asterisk.

54 5 | Neutron diffraction and ¹¹B solid state NMR studies of the crystal structure of B-doped mullite

5.3.2.1 Rietveld refinements For all refinements of B-doped samples the refined model of the B-free mullite sample was used as a starting model, already resulting in good fits. A decrease of lattice parameter c compared to boron-free mullite was observed for all B-doped samples. As discussed in our previous work (Lührs et al., 2012, chapter 4) this indicates the incorporation of boron into the crystal structure of mullite. In order to determine the B position within the mullite structure difference Fourier and gridsearch calculations had to be applied to all samples.

5.3.2.2 Difference Fourier synthesis For all samples difference Fourier calculations were performed after the best possible profile fit had been achieved with the B-free model. The positions of the maxima and minima in the difference scattering density map were carefully checked for correspondence with atomic positions of the B-free mullite. The highest maximum in all calculations corresponds to the O1 position, for samples II-IV the O1 maximum is followed by maxima for the O2 and Al (not IV) position. For all B-doped samples the first maximum that could not be assigned to an atomic position of the B-free mullite model was found at about x = 0.21, y = 0.27, z = 0.50. Difference Fourier maps for samples I (B-free) and VI (B-doped) are given in Fig. 5-4 a, b for z = 0.5. From the contour maps it clearly emerges that there is a distinct maximum around x = 0.2, y = 0.27, z = 0.5 for the B-doped sample, indicated by the red contour lines. For the B-free sample no residual neutron scattering matter is observed near this position. Additionally it is observed that the maximum in samples II-VI shows increasing intensity with increasing B-content (Fig. 5-3). Therefore this position is likely to be occupied by B. For samples V and VI at least one of the positions in the list of minima can be assigned to the O3 position and for sample VI a minimum is also observed near the T position (minima = blue contour lines in Fig. 5-4).

Fig. 5-3: Relative residual neutron scattering density RHO (highest maximum = 999) as a function of the initial B2O3 content. Numbers of samples are given according to Table 5-1.

The following model is derived from the difference Fourier calculations: B enters a threefold coordinated position between the octahedral chains and tetrahedral Si is removed in equal amounts. To allow for charge compensation the Si atoms of two neighboring tetrahedra are replaced by trigonally coordinated B atoms and at the connecting O3 position between the tetrahedra an oxygen vacancy forms. This model is in good accordance with the hypothesis given in equation (2). The maxima in scattering density near the O1 and O2 positions can possibly be interpreted as indication for a rotation or distortion of the octahedra.

55 5 | Neutron diffraction and ¹¹B solid state NMR studies of the crystal structure of B-doped mullite

Fig. 5-4: a, b) Difference Fourier maps for the x-y-plane with z=0.5 for boron free sample I (a) and boron-doped sample VI (b). A projection of the polyhedral framework of mullite is given with octahedra in blue and tetrahedra in green. c) SOF contour map of the grid search analysis for sample VI (B-doped). d) SOF (red), RB (gray) and Rwp (gray) as a function of x, along the red line given in the contour map (y = 0.24, z = 0.5).

5.3.2.3 Grid search analysis For all B-doped samples grid search calculations were performed based on the same model as the difference Fourier calculations. The occupancy (SOF) of the dummy atom reveals a local maximum around x = 0.22, y = 0.24, z = 0.5, simultaneously minima for the residuals (RB and Rwp) are observed. All three effects become more pronounced with increasing B-content, and similarly to the difference Fourier results for the B-free sample no effects can be observed at this position. In Fig. 5-4 (c) a contour map of the occupancy (SOF) is given for sample VI. The red contour lines clearly indicate the maximum at the above mentioned position. The correlation of the maximum in SOF and the corresponding minima in Rwp and RB can be seen in an x-profile with y = 0.24 and z = 0.5 shown in Fig. 5-4 (d). As well as from the difference Fourier calculations it emerges from the gridsearch analyses that there is an increasing amount of observed scattering matter with increasing B-content on

56 5 | Neutron diffraction and ¹¹B solid state NMR studies of the crystal structure of B-doped mullite the threefold coordinated position connecting the octahedral chains. This can only be explained by B in this position.

5.3.2.4 Rietveld and DLS Refinements of the B-mullite model Up to this point all Rietveld refinements were performed using a B-free mullite model. The sample with the highest B-content and a relatively small amount of alumina impurity (sample VI) was chosen for the final structure refinements. A B position was added to the structural model at the position derived from difference Fourier and gridsearch analyses (x = 0.2, y = 0.27, z = 0.5). The refinement with a fixed occupancy for the B position of SOF = 0.02 corresponding to 0.16 B atoms per unit cell yielded the best RB value (2.51 %) which was further improved by reducing the occupancy of the Si position by 0.02 (RB = 2.42 %). Independent refinement of occupancies for B, Si and O3 resulted in a RB of 2.24 % and yielded occupancies very close to the values expected for the substitution model given in equation (2). Further improvement of the residuals (RB = 2.02 %) was achieved by refining the isotropic displacement parameters after fixing the occupancy for B to 0.025 (0.2 B atoms per unit cell) and adjusting the occupancies of Si and O3 according to equation (2).

Within the expected errors this corresponds to a B-mullite composition of: Al4.64Si1.16B0.2O9.58. The refinement resulted in strongly enlarged B-O distances of 1.64 Å and 1.65 Å, whereas a distance of

1.37 Å is expected for a triangular BO3 unit (Hawthorne et al., 1996; Gatta et al., 2010, 2012). The local environment as refined by DLS and Rietveld methods is shown in Fig. 5-5. The refinement yielded a low DLS reliability index of 0.014 and B-O distances of 1.37 Å and 1.42 Å. For the O1 and O2 positions the refinement resulted in significant shifts for the oxygen atoms directly connected to B (O11, O211, O212), smaller shifts were observed for the O positions in the second coordination sphere. Finally, Rietveld refinements were performed using the results of the DLS calculations. Therefore, split positions of O1 and O2 were introduced (O11, O211, O212) and the occupancies fixed according to the following substitution scheme: for each B introduced into the structure one O1 and two O2 atoms were replaced by O11, O211, and O212 with the same occupancy as B. The refinement resulted in B-O distances close to the ideal value of 1.37 Å. Compared to the model without split positions the RB value decreased from 2.02 % to 1.97 %. The results of the refinements are listed in Table 5-5, selected interatomic distances and angles are given in Table 5-6. Rietveld refinements for samples III to V revealed the same substitution mechanism, yielding increasing site occupancies for the B position and decreasing site occupancies for the Si and O3 position compared to the B-free model. With respect to the loss of B during synthesis the amount of B refined for samples III to V correlates well with the initial compositions of the samples (Table 5-1) yielding a smooth series together with sample VI.

57 5 | Neutron diffraction and ¹¹B solid state NMR studies of the crystal structure of B-doped mullite

Fig. 5-5: Section of B-mullite crystal structure with atomic positions derived from 3:2 mullite (red, black) and after DLS calculations with subsequent Rietveld refinement of split positions for O (green). The latter yielded almost ideal B-O distances and confirmed the local distortion of octahedra due to B incorporation. * red positions for 3:2 mullite from Saalfeld and Guse (1981).

Table 5-5: Atomic coordinates, Wyckoff positions, and site occupancies (occ.) as refined for sample VI with the refined composition of Al4.64Si1.16B0.2O9.58. Pbam: a= 7.564902(1), b= 7.681806(1), c = 2.875613(4). Rp`=9.50, Rwp=4.52, RB=1.97.

atom Wyck. x y z occ. B_iso [Å²] 0 0 0 Al 2a 1 0.56(4) 0.1516(4) 0.3391(3) 0.5 T(Al) 4h 0.5 0.81(4) 0.1516(4) 0.3391(3) 0.5 T(Si) 4h 0.291(2) B(T(Al)) 0.2654(21) 0.2022(20) 0.5 T*(Al) 4h 0.160 1.36 0.3564(2) 0.4234(1) 0.5 O1 4h 0.95 0.94(2) 0.1249(2) 0.2178(2) 0 O2 4g 0.90 0.89(2) 0 0.5 0.5 O3 2d 0.463(1) 1.76(8) 0.4458(10) 0.0530(11) 0.5 O4 4h 0.160 1 0.2306(25) 0.2547(26) 0.5 B 4h 0.050(2) 0.2(6) 0.3652 0.3746 0.5 O11 4h 0.05 B(O1) 0.1432 0.2074 0.9075 O211 8i 0.025 B(O2) 0.1439 0.2063 0.0925 O212 8i 0.025 B(O2)

58 5 | Neutron diffraction and ¹¹B solid state NMR studies of the crystal structure of B-doped mullite

Table 5-6: Selected interatomic distances and angles as refined for sample VI with the refined composition of Al4.64Si1.16B0.2O9.58. Values are given for 3:2 mullite, the Rietveld refinement, and the Rietveld refinement after DLS refinements. Distances and angles including split positions of O are given in italics.

Distances [Å] Angles [°] 3/2 B- After 3/2 B- After mullitea mullite DLS mullite a mullite DLS M1-site M1 Al-O1 4x (2x) 1.896 1.898(1) 1.896(1) O1-Al-O1 2x 180 180 180 Al-O11 2x 2.009 O2-Al-O2 1x 180 180 180 Al-O2 2x 1.943 1.920(1) 1.922(1) O1-Al-O1 2x 99.17 98.49(5) 98.66(5) Al-O211 1x 1.945 O1-Al-O1 2x 80.83 81.51(5) 81.34(5) Al-O212 1x 1.941 O1-Al-O2 4x 89.80 89.53(5) 89.34(5) B-free mean 1.912 1.905 1.905(1) O1-Al-O2 4x 90.20 90.47(5) 90.66(5) B cont mean 1.949 angles involving split positions O11, O211, O212 O-Al-O 80.12(4) – 99.88(4)

T-site T T-O1 1x 1.700 1.680(3) 1.679(3) O1-T-O3 1x 111.08 110.6(1) 110.2(1) T-O2 2x 1.725 1.728(1) 1.725(1) O1-T-O2 2x 106.69 107.9(1) 108.4(1) T-O211 1.982(1) O2-T-O2 1x 113.56 112.6(1) 112.9(1) T-O212 1.986(1) O2-T-O3 2x 109.39 108.9(1) 108.5(1) T-O3 1x 1.658 1.684(3) 1.686(2) mean 1.702 1.705(2) 1.704(2)

T*-site T* Al*-O1 1x 1.814 1.79(2) 1.83(2) O1-T*-O4 1x 106.19 109.1(9) 108.0(8) Al*-O2 2x 1.773 1.80(1) 1.79(1) O1-T*-O2 2x 99.97 100.0(6) 99.2(6) Al*-O4 1x 1.852 1.77(2) 1.78(2) O2-T*-O4 2x 118.96 119.3(6) 119.8(5) mean 1.803 1.79(1) 1.80(1) O2-T*-O2 1x 108.96 105.8(8) 106.7(8)

B-site B B-O1 1x - 1.64(2) 1.61(2) O1-B-O2 2x 113.7(7) 114.8(7) B-O11 1.37(2) O11-B-O211 121.8(7) B-O2 2x - 1.65(1) 1.67(1) O11-B-O212 121.9(7) B-O211 1.39(1) O2-B-O2 1x 121.6(1.2) 118.9(1.1) B-O212 1.39(1) O211-B-O212 114.5(1.3) a 3/2 mullite data: (Saalfeld and Guse, 1981)

5.4 Discussion

Only from the combination of a series of samples with different B contents using various methods were we able to collect sufficient information to develop a model for the incorporation of B into the crystal structure of mullite. The comparison with a B-free sample was especially crucial with respect to the results of difference Fourier and gridsearch calculations. The substitution model developed from difference Fourier and gridsearch calculations suggesting that B resides in a trigonally coordinated position connecting the octahedral chains perpendicular to the crystallographic c-axis agrees well with the results of NMR spectroscopy and confirmed the hypothesis regarding the substitution mechanism given in equation (2). Rietveld refinements using this model resulted in unacceptably large B-O distances. As there does not exist a threefold coordinated position within the crystal structure of mullite with B-O distances close to the expected value of 1.37 Å, a distortion and/or rotation of the octahedra has to be considered. This is also indicated by maxima in the difference Fourier maps close to the O1 and O2 position and the increasing inclination angle Ȧ with

59 5 | Neutron diffraction and ¹¹B solid state NMR studies of the crystal structure of B-doped mullite increasing boron content (Fig. 5-6). The inclination angle Ȧ represents the angle between the octahedral chains and was calculated from the results of Rietveld refinements based on a B-free model. All values for Ȧ are within the range that is expected for mullites (Fischer and Schneider, 2005). By allowing local shifts of the O1 and O2 positions a model with reasonable B-O distances and split positions for O1 and O2 was derived from geometric refinements using the distance least squares (DLS) method (Baerlocher et al., 1978). The subsequent Rietveld refinement yielded B-O distances of 1.37 Å (B-O11) and 1.39 Å (B-O211/2), the complete results are summarized in Table 5-5 and 5-6 All distances and angles (Table 5-6) for the average structure are well within the range that can be expected for mullite. Considering the split positions of O1 and O2 the only distances significantly larger than expected are the Al-O11 distance of 2.009 Å and the T-O211/O212 distances of 1.982/1.986 Å. With respect to the disorder in the local environment the error might be larger than expected from the esd’s. Nevertheless these values indicate a significant distortion of the polyhedra in the coordination sphere of the BO3 group. Generally the octahedra and tetrahedra become more regular due to B incorporation compared to B-free 3:2 mullite as shown by the distortion indices calculated according to Baur (1974) given in Table 5-7. Especially the distortion index for the octahedral distances (MO) shows a significant decrease upon B-incorporation and a small decrease is also observed for the tetrahedral angle index (OTO). The distortion indices for the octahedral angle (OMO) and the tetrahedral distances (TO) show no significant variation. As no increase in distortion is observed upon B-incorporation in the average structure, it can be concluded that the increasing Ȧ (Fig. 5-6) indicates a rotation of the octahedra.

Fig. 5-6: Inclination angle Ȧ bet-ween octahedra as a function of the initial content of B2O3. Numbers of samples are given according to Table 5-1.

In Fig. 5-7 the crystal structure of B-mullite is given as refined for sample VI including the Si-B substitution mechanisms developed in this work. Only one out of many possible distributions for oxygen vacancies, T3O groups, and BO3 groups is represented (Fischer et al., 2012). The substitution mechanism was shown to be the same independent of the B-content. Compared to the initial composition of sample VI the refined B2O3 content is strongly reduced (2.8 mol% compared to 7.6 mol%). This agrees very well with the fact that some B is lost during synthesis and therefore additional alumina forms (Lührs et al., 2012, chapter 4). As far as reasonable refinements of the B positions were possible for the samples II to V they yielded a reduction of the B-content in the same order of magnitude as for sample VI.

60 5 | Neutron diffraction and ¹¹B solid state NMR studies of the crystal structure of B-doped mullite

Table 5-7: Polyhedral distortion indices (DI) calculated according to Baur (1974)

Fig. 5-7: Crystal structure of B-mullite as refined for sample VI showing the Si-B substitution mechanism developed in this work. Oxygen vacancies are indicated by squares. Only one out of many possible distributions (see Fischer et al., 2012) of oxygen vacancies, T3O groups and BO3 groups is shown.

61 5 | Neutron diffraction and ¹¹B solid state NMR studies of the crystal structure of B-doped mullite

For B-doped mullite an anisotropic behavior of the lattice parameters due to the incorporation of B is described (Griesser et al., 2008; Zhang et al., 2010; Lührs et al., 2012, chapter 4). Changes are reported to be highest along the crystallographic c-axis, followed by significant changes in b direction. The reduction of lattice parameter c can directly be derived from Fig. 5-5 as the split O positions (O211 and O212) are closer to each other than O2-O2. Comparing the crystal structure of 3:2 mullite with B-doped mullite (Table 5-5), changes in the oxygen positions including split positions for O1 and O2 are observed (Fig. 5-5). This leads to significant changes in the O-O distances for the unit cells with B instead of Si. The changes for each O-O distance were calculated for the three crystallographic directions and are given in Table 5-8. In direction of c a local reduction of the O2-O2 distance of 0.53 Å is observed for B-mullite compared to B-free mullite, this corresponds to 18.5 % of lattice parameter c. The change in b-direction is significantly lower (3.7 %) and for a it is only 1 %, this agrees well with Lührs et al. (2012 chapter 4) who did not find systematic changes in lattice parameter a upon B-incorporation but significant changes for b and c. With this crystallographic model the anisotropic behavior of lattice parameters upon the incorporation of B can be explained and it is shown that the lattice parameters are strongly affected by the O-O distances.

Table 5-8: Shift of O1 and O2 atoms from 3:2 mullite to the model of B-mullite (Table 5-5) given for the different crystallographic directions.

Shift in direction of lattice vector a b c O1-O11 0.07 Å -0.37 Å 0.00 Å O2-O211 0.14 Å -0.09 Å 0.27 Å O2-O212 0.14 Å -0.08 Å -0.27 Å Change of O-O distance -0.08 Å -0.29 Å -0.53 Å Change of O-O distance in parts of lattice parameter -1.0 % -3.7 % -18.5 %

62 5 | Neutron diffraction and ¹¹B solid state NMR studies of the crystal structure of B-doped mullite

5.5 Conclusion

The crystal structure of B-mullite was solved by means of neutron diffraction and 11B MAS NMR spectroscopy. The precise B position could only be derived from a combination of Rietveld refinements with difference Fourier calculations and gridsearch analyses applied to a series of samples with varying B contents. The Si atoms in two neighboring tetrahedra are replaced by B and the connecting oxygen atom is removed from the structure. Hence, the incorporated B resides in a near-trigonal planar position crosslinking the octahedral chains perpendicular to the c-axis. The local BO3 configuration was determined unequivocally by 11B MAS NMR, and also confirmed by distance least squares refinements resulting in split positions for two O atoms in the adjacent AlO6 octahedron which leads to a local distortion of the octahedral geometry. Furthermore the split positions lead to significantly shortened O-O distances in the c-direction compared to the a- and b-direction. Therefore the crystallographic model provides an explanation for the anisotropic behavior of the lattice parameters upon B-incorporation. In order to explain the anisotropic thermal expansion behavior of B-mullite, the temperature dependent development of distances and angles has to be considered. Acknowledgements: We would like to thank the University of Bremen for financial support from the Central Research Development Fund. The authors gratefully acknowledge the financial support provided by FRM II to perform the neutron scattering measurements at the Forschungs-Neutronenquelle Heinz Maier- Leibnitz (FRM II), Garching, Germany. JVH thanks EPSRC and the University of Warwick for partial funding of the solid state NMR infrastructure at Warwick, and acknowledges additional support for this infrastructure obtained through Birmingham Science City: Innovative Uses for Advanced Materials in the Modern World (West Midlands Centre for Advanced Materials Project 1), with support from Advantage West Midlands (AWM) and partial funding by the European Regional Development Fund (ERDF).

6 Ambient and high-pressure synthesis, composition, and crystal structure of B-mullites

H. Lührs*1, S. Soellradl2,3,4, S. P. King5, J. V. Hanna5, J. Konzett6, R.X. Fischer1, and H. Schneider1,7

1 Universität Bremen, Fachbereich Geowissenschaften, Klagenfurter Str., 28359 Bremen, Germany 2 FRM II Technische Universität München, Forschungsneutronenquelle Heinz-Maier-Leibnitz (FRM II), 85747 Garching, Germany 3 Paul Scherrer Institute, Laboratory for Radiochemistry and Environmental Chemistry, 5232 Villigen PSI, Switzerland 4 University of Berne, Department of Chemistry & Biochemistry, Freiestrasse 3, 3012 Berne, Switzerland 5 University of Warwick, Department of Physics, Coventry, CV4 7AL, UK 6 Universität Innsbruck, Institut für Mineralogie & Petrographie, Innrain 52, A-6020 Innsbruck, Austria 7 Universität zu Köln, Institut für Kristallographie, Greinstr. 6, 50939, Köln, Germany

Published in: Crystal Research and Technology (2013) DOI: 10.1002/crat.201300210

submitted: 4 July 2013 | revised: 22 July 2013 | accepted: 26 July 2013 | online: 21 August 2013

Keywords: Boron mullite | high pressure synthesis | PGAA | XRD | 11B MAS NMR

The chemical compositions of several B-doped mullite samples were analyzed using prompt gamma activation analyses (PGAA) indicating that 15% of the Si in the crystal structure of mullite can be replaced by B during sol-gel synthesis at ambient pressure and 1200°C without the formation of impurities. Furthermore the PGAA results agree very well with the chemical compositions derived from Rietveld refinements based on neutron diffraction data. High-pressure and high-temperature synthesis yielded a B-mullite with significantly higher B-content than observed before (close to composition Al8Si2B2O19). The results of PGAA, XRD, and neutron diffraction experiments show linear behavior of lattice parameters b and c as well as of the inclination angle Ȧ of the AlO6 octahedra in the ab-plane with increasing B-content. The Rietveld refinements support the substitution mechanism known for B-mullites, involving the formation of oxygen vacancies and the replacement of

40% of the tetrahedral Si by BO3 units during the synthesis at 875°C and 10 kbar. However, the refined chemical composition as well as a very low lattice parameter a suggest a second mechanism for the incorporation of B into mullite. This is clearly supported by the 11B MAS NMR experiment indicating the presence of B in BO4 configuration but BO3 being dominant. Here for the first time a B- mullite crystal structure is presented yielding BO3 and BO4 units in space group Pbam.

* Corresponding author: e-mail [email protected], Phone: +49 (0) 421 218 65181, Fax: +49(0) 421 218 65189

65 6 | Ambient and high-pressure synthesis, composition, and crystal structure of B-mullites

6.1 Introduction

Mullite is a rare mineral in nature but it represents one of the most prominent phases in both traditional and advanced ceramic materials. The outstanding properties of mullite such as high thermal stability, low thermal expansion and conductivity, high creep resistance and corrosion stability make mullite ceramics technologically important. The chemical composition of sol-gel derived mullites is given by the solid-solution series Al2(Al2+2xSi2-2x)O10-x, 0.2 < x < 0.9, corresponding to about

58-94 mol% Al2O3 (Fischer et al., 1996) with the thermodynamically stable compounds 3:2 mullite (sinter-mullite, x = 0.25) and 2:1 mullite (fused-mullite, x = 0.4). Depending on cation charge and diameter, synthesis temperature, and oxygen fugacity, the crystal structure of mullite is able to incorporate a large variety of cations (Schneider, 2005b; Schneider et al., 2008 and references therein).

Dietzel and Scholze (1955) investigated the system B2O3-Al2O3-SiO2. From systematic changes of the refractive indices and the reduction of lattice parameter c with increasing B2O3 content, they proposed a solid-solution series between 3:2 mullite (Al4.5Si1.5O9.75) and Al18B4O33 by substitution of B for Si. Recent investigations (Zhang et al., 2010; Lührs et al., 2012, 2013a, chapters 4 and 5) showed that there is no complete solid solution series between mullite and mullite-type aluminumborates. In comparison to undoped 3:2 mullite, B-doping of mullite results in a significant reduction of lattice parameters b from 7.6944(2) to 7.6703(6) and c from 2.8844(1) to 2.8393(3) whereas lattice parameter a does not alter appreciably (7.5572(3) for B-free mullite and 7.5431(7) for B-mullite), both structures are described in space group Pbam. Furthermore, the incorporation of B results in a strong reduction of the mean thermal expansion coefficient of 15%, which makes the material interesting, e.g., for substrates being thermally shock resistant up to 1000°C (Lührs et al., 2012, chapter 4). The crystal structure of B-mullite has recently been solved from neutron diffraction data and 11B MAS NMR spectroscopy of a series of sol-gel derived B-mullites with various compositions synthesized at 1200°C (Lührs et al., 2013a, chapter 5). Here, we report the results of the chemical analyses of B- mullites using prompt gamma activation analyses (PGAA) compared with the results of the Rietveld refinements based on neutron diffraction data.

Another borosilicate phase in the B2O3-Al2O3-SiO2 system with a composition given as

Al8Si2B2O19 was synthesized by Werding and Schreyer (1992) at high pressure and high temperature being described as a boron-bearing sillimanite derivative. Later on it was described as a “boron- mullite” with traces of disordered boralsilite (Grew et al., 2008). However, no details of the crystal structure and the crystal chemistry of this phase are known so far. Therefore, high-pressure syntheses were performed and the resulting products were analyzed by XRD, PGAA, and 11B MAS NMR in order to develop a structural model for this phase.

66 6 | Ambient and high-pressure synthesis, composition, and crystall structure of B-mullites

6.2 Experimental

6.2.1 Sample preparation

All B-mullite gels were prepared from aluminum nitrate nonahydrate (Al(NO3)3*9H2O), tetraethoxysilan (C8H20O4Si), and boric acid (H3BO3). Stoichiometric amounts (for high-pressure synthesis of Al8Si2B2O19, 50% H3BO3 in excess was used) were dissolved in pure ethanoll at 60°C, resulting in a transparent sol. The sol was kept in a drying oven at 60°C for 4-5 days to allow for gelation. The resulting gel was then dried for 5h at 150°C, ground and calcined as follows and given in Table 6-1.

For the high-pressure synthesis of the Al8Si2B2O19 phase, two runs were performed, HP1 and HP2. The precursor of the same batch was calcined for 2h at 600°C (HP1), and for HP2 additionally for 2h at 1000°C. Both starting materials were thhen filled into a gold (HP1) or plattinum (HP2) tube with 3.0 mm outer diameter and heated to 500°C for 3h before welding shut. TThis heating step was performed in order to minimize the amount of water from air humidity adsorbed to the starting materials and, thus, to generate nominally dry experimental conditions as mentioned in Werding and Schreyer (1992). For the high-P-T runs, NaCl-pyrophyllite assemblies were used with the capsules embedded in a boron nitride cylinder. The temperature was measured with a K-type thermocouple and both pressure and temperature were computer controlled during the entire run. P-T conditions and run durations for runs HP1 and HP2 are listed in Table 6-1. For the syntheses att ambient pressure, the dried gel was ground and then calcined at 350°C for 5h before it was heated to 1200°C for 5h to form B-mullite.

Table 6-1: synthesis conditions for ambient (AP) and high-pressure (HP) syntheses of B-mullites. Cor=corundum crucible.

initial gel tube calcination pre- duration T P composition /crucible heating [h] [°C] [bbar] B) 11 B NMR G 3 XRD PGAA SPODI (99% 11 HP1 Al4.39Si0.93B1.38O9.46 Au, 3mm 3h 600°C 3h 500°C 168 800 7*10 x 3 HP2 Al4.39Si0.93B1.38O9.46 Pt, 3mm 3h 600°C + 3h 500°C 260 875 10*10 x x x 1h 1000°C AP1 Al4.51Si1.42B0.08O9.71 cor 5h 350°C 5 1200 1 x x x AP2 Al4.47Si1.23B0.30O9.61 cor 5h 350°C 5 1200 1 x x x AP3 Al4.50Si1.05B0.45O9.52 cor 5h 350°C 5 1200 1 x x x AP4 Al4.50Si1.50B0.00O9.75 cor 5h 350°C 5 1200 1 x x AP5 Al4.50Si1.43B0.7O9.71 cor 5h 350°C 5 1200 1 x x AP6 Al4.51Si1.26B0.23O9.63 cor 5h 350°C 5 1200 1 x x AP7 Al4.50Si1.12B0.38O9.56 cor 5h 350°C 5 1200 1 x x AP8 Al4.50Si1.05B0.45O9.52 cor 5h 350°C 5 1200 1 x x AP8* Al4.52Si0.95B0.53O9.48 cor 5h 350°C 5 1200 1 x x * used for structure refinement in Lührs et al. (2013a, chapter 5).

67 6 | Ambient and high-pressure synthesis, compoosition, and crystal structure of B-mullites

6.2.2 Powder X-ray diffraction (XRD)

Powder diffraction patterns of the HP samples were collected on a Bruker AXS D8 Advance powder diffractometer with Cu-KĮ1 (Ȝ=1.5405598 Å) in transmission geometry using glass capillaries.

A Johansson monochromator accomplished the usage of pure KĮ1 radiation. The instrument is equipped with a primary divergence aperture, primary (4°) and secondary (2.5°) Soller slits, a secondary iris aperture (6.42 mm) and a position sensitive LynxEye detector. Scans were performed with a step width of 0.0198°2ș and a measuring time of 10 (HP1) or 3 (HP2) s/step from 5 to 140°2ș.

6.2.3 Neutron diffraction

Thee neutron diffraction experiments at room temperature were carried out in vanadium vessels on the high-resolution powder diffractometer SPODI at FRM-II, Germany (Hoelzel et al., 2012). A collimated beam of thermal neutrons (Ȝ = 1.5482 Å) has been obtained at a take-off angle of 155° using the (551) reflection of a vertically focusing composite germanium monochroomator. Diffraction data were collected from 1 to 152°2ș with a step size of 0.05°2ș.

6.2.4 Prompt gamma activation analyses (PGAA)

Thee chemical compositions of four samples (Table 6-1) were determined ussing prompt gamma activation analyses (PGAA) at the FRM-II, Germany (Canella et al., 2011). For the HP2 sample only 4 mg material were available for the PGAA measurement which is close to the minimum limit of material needed for precise quantification. The HP2 sample was irradiated with a cold neutron beam with a thermal flux equivalent of 4*1010 n cm-2 s-1 using the standard PGAA settup (Canella et al., 2011). For the measurement of the three ambient pressure (AP) samples 600 mg sample mass were used. Due to the high neutron-capture cross section of B (716 barn, Molnár, 20044) compared to the cross sections of Al (231 mbarn, Molnár, 2004) and Si (172 mbarn, Molnár, 2004),, the boron signal at 478 keV and its introduced background dominated the spectrum. Therefore 10 mm of lead were introduced between the sample and the high-purity germanium detector in ordder to attenuate the contribution of boron in the spectrum and achieve a better statistical uncertainty (SSöllradl et al., 2013, abstract in Appendix B) in case of the ambient pressure (AP) samples. The PGAA spectra were then evaluated with Hypermet-PC (Fazekas et al., 1996) and the composition of the samples was determined with the ProSpeRo program (Revay, 2009).

6.2.5 11B MAS NMR

11 Thee B solid state MAS NMR measurement was performed at a B0 fieldd of 14.5 T using a Varian 600 spectrometer operating at a Larmor frequency of 193.23 MHz. Single-pulse experiments were facilitated using a Varian T3 HXY MAS probe and a MAS frequency of 12.5 kHz in order to remove broadenings from quadrupolar and dipolar interactions. A ‘non selective’ (solution like) pulse time of 12 μs was obtained with a selective (solids) pulse time of 2 μs (corresponndding to ~ ʌ/6) used

68 6 | Ambient and high-pressure synthesis, composition, and crystall structure of B-mullites with a recycle delay of 5 s. The acquired 11B MAS NMR data were referencedd to the IUPAC primary standarrd BF3. Et2O at įiso 0 ppm, via a secondary solid reference of NaBH4 at įiso -42.06 ppm (Hayashi and Hayamizu, 1989). Experimeental results were simulated using tthe DmFit software in order to extract information on the quadrupole and isotropic chemical shift parameters (Massiot et al., 2002).

6.2.6 Distance least squares

The program DLS-76 (Baerlocher et al., 1978) was used to refine a part oof the B-mullite crystal structure with geometric restraints for the interatomic distances. This was necessary because of enlarged B-O distances after the Rietveld refinement. Lattice parameters weerre fixed to the refined values from Rietveld analyses as well as the atomic coordinates for Al and the O positions of the octahedra not belonging to the first two coordination spheres of B. The prrescribed distances and respecttive weights were taken from Lührs et al. (2013a, chapter 5, Table 5-3). The refinement resulted 2 2 1/2 in a low reliability index of R = 0.018 (R = (Ȉ(w(d0-d)) /Ȉ(w d0) ) with d0 and d = observed and calculated distances, respectively).

6.3 Results and discussion

6.3.1 Chemical analyses of B-mullites – B-mullites with increasing B-content

The crystal structure of B-mullite was recently solved by Lührs et al. (2013a, chapter 5) and the following substitution mechanism was developed with respect to undoped mullite: 2Si4+ + O2- o 2B3+ + Ƒ (Ƒ = oxygen vacanccy). The Rietveld refinement of the crystal structure based on neutron diffraction data and 11B MAS NMR measurements revealed a threefold coordinated B position linking the octahedral chains perpendicular to the c-axis. Furthermore, a split position for two oxygen atoms was introduced explaining the anisotropic behavior of latticce parameters upon B- incorporation. Reasonable occupancies for B were refined indicating that not alll of the initial B enters the crystal structure but leads to the formmaation of an Al-richer B-mullite compared to the initial gel compositions. The extent of B-incorporation could not be determined by standard methods and even for the PGA analyses a special instrumental setup using a lead attenuator waas necessary in order to measurre these samples (Söllradl et al., 2013, abstract in Appendix B). The chemical compositions of three single-phase samples from the ambient pressure (AP) series (1 bar, 1200°C) are given in Table 6-2. It clearly emerges from Fig. 6-1 that there is a linear correlation between the initial B-content and the results from PGAA, indicating that about 50(1) % of the initial B enters tthe crystal structure of mullite during 1200°C synthesis at ambient pressure. For the single phase sammpple with the highest B- content (AP3) this corresponds to a replacement of 15 % of the tetrahedral Si by BO3 units compared to B-free mullite.

69 6 | Ambient and high-pressure synthesis, composition, and crystal structure of B-mullites

Table 6-2: chemical compositions of B-mullite samples: initial gel composition, PGAA, and Rietveld refinement.

initial gel composition PGAA Rietveld refinement AP1 Al4.51Si1.42B0.08O9.71 Al4.58(3)Si1.39(3)B0.032(1)O9.69 AP2 Al4.47Si1.23B0.30O9.61 Al4.62(3)Si1.23(3)B0.147(1)O9.62 AP3 Al4.50Si1.05B0.45O9.52 Al4.71(3)Si1.06(3)B0.228(3)O9.69 AP8 Al4.50Si1.05B0.45O9.52 Al4.7Si1.1B0.2O9.6 (neutron) HP2 Al4.39Si0.93B1.38O9.46 Al4.19(7)Si0.91(6)B0.90(2)O9.45 Al4.5Si0.9B0.6O9.4 (XRD)* * Additionally some of the tetrahedral Al is replaced by B

Fig. 6-1: Comparison of initial B-content in atoms per formula unit (a.p.f.u.) in the gel with PGAA results of three selected B-mullite samples (see Table 6-1). The uncertainties for PGA analyses are within the symbol size.

In the Rietveld refinement of sample AP8 (neutron diffraction data), the known Si-B and Si-Al substitution mechanisms were applied. The refinement resulted in a composition of Al4.7Si1.1B0.2O9.6 only deviating from the PGA analysis of sample III (with the same initial gel composition) in the second position after the decimal point. For neutron diffraction experiments, samples enriched in 11B are required which is rather disadvantageous for PGAA. Therefore, it was not possible to analyze the same sample with both methods. Compared to the initial gel composition the Rietveld refinement and the PGA analyses resulted in Al richer mullites. According to Fischer et al. (1996) an increase in lattice parameter a is expected for Al rich mullite. This was not the case for the B-mullites where no significant variation for a was observed for increasing B-content (and Al-content) (Lührs et al., 2012, chapter 4) up to 5 mol% B2O3. Most likely, the effect of increasing cell dimension in a direction is compensated by the reduction in the same direction due to B-incorporation. The Rietveld refinement of samples with higher initial B-content yielded B-mullite compositions containing less than 50(1)% of the initial B which can be ascribed to the presence of alumina impurities. It was shown by Lührs et al. (2012, chapter 4) that the lattice parameter c is most indicative for the amount of boron in the crystal structure of mullite. The same behavior is observed in Fig. 6-2 where the lattice parameters for mullites with different amounts of B are given as well as some reference values from literature. For lattice parameters b and c, the refined values for the high-pressure sample HP2 (star in Fig. 6-2) fit very well into the trend of decreasing lattice parameter with increasing B-content. A linear relationship can be applied between the B-content and lattice parameters b and c, resulting in coefficients of correlation of 99% and 96%, respectively. The reported c lattice parameters of Al8Si2B2O19 (Werding and Schreyer, 1992), boromullite Al9BSi2O19 (Buick et

70 6 | Ambient and high-pressure synthesis, composition, and crystal structure of B-mullites

al., 2008), and boralsilite Al16B6Si2O37 (Grew et al., 2008) are very close to the extrapolated linear line for lattice parameter c in Fig. 6-2. For lattice parameters a and b, a significant decrease is observed at higher B-contents. For b there is an abrupt decrease for boralsilite only (> 15 mol% B2O3). Whereas for lattice parameter a no significant variation was observed below 5 mol% B2O3 a significant decrease is observed between 5 and 25 mole %, including the high-pressure sample HP2. Especially this low value for lattice parameter a for HP2 suggests that there is a second mechanism for B-incorporation, this could be tetrahedral BO4 as observed for boralsilite (see 6.3.2). Rietveld refinements of a B-free mullite model were performed for a series of samples with different B-content (neutron diffraction data) and sample HP2 (XRD data). The inclination angle between adjacent octahedral chains was calculated as the angle between the octahedral chains in the a-b-plane (Fig. 6-2). The inclination angle systematically increases with increasing boron content as already reported by Lührs et al. (2013a, chapter 5, Fig. 5-6). The HP2 sample yields a significantly higher inclination angle than observed for B-mullites before and fits very well into a linear trend including the other B-mullite data (R² = 98%).

Fig. 6-2: Lattice parameters and inclination angle Ȧ (between the octahedral axes of neighboring AlO6 octahedra in the ab-plane) as a function of the B-content (estimated from PGAA or relationship given in Fig. 6-1). Filled blue circles (ambient pressure AP, neutron diffraction) and red star (HP2, XRD) represent parameters of this work whereas squares refer to literature values (a: Al8Si2B2O19 (Werding and Schreyer, 1992), b: boromullite Al9BSi2O19 (Buick et al., 2008), and c: boralsilite Al16B6Si2O37 (Grew et al., 2008)). Linear relationships are given for lattice parameters b and c and Ȧ with the respective coefficients of correlation R2.

71 6 | Ambient and high-pressure synthesis, compoosition, and crystal structure of B-mullites

6.3.2 B-rich high pressure B-mullite

Two high-pressure syntheses (HP1, HP22) were performed in order to synthesize a phase with the composition Al8Si2B2O9 reported by Werding and Schreyer (1992). Compaared to the homo- geneously gray HP2 sample, sample HP1 consisted of transparent particles up to about 30 μm size and very small dark particles. According to the powder diffrraction pattern, the firstt experiment (HP1, 7 kbar at 800°C for 168 h) yielded dumortiorite and disordered boralsilite. This aggrees well with the pressure-temperature conditions given by Grew et al. (2008). According to the same authors, the P-T conditions (10 kbar at 875°C for 260 h) of the second experiment (HP2) are withiin the stability field for B-mullite, quartz, and dravite. The XRD pattern (Fig. 6-3) agrees very well wiith the d-values and relative intensities given by Werding and Schreyer (1992) for the phase Al8Si2B2O19. The only reflection mmissing in the HP2 dataset compared with the data given in Werding and Schreyer (1992) is the reflection at 20.3°2ș with relative intensity of 5%, which was according to Grew et al. (2008) the only reason for Werding and Schreyer (1992) to index the phase with two lattice parameters doubled compared to mullite. According to Grew et al. (2008), the 20.3°2ș diffraction peak indicates the incipient development of a boralsilite-like phase obviously not present in the HP2 sample presented here (Fig. 6-3). The broad bump between 16 and 26°2ș indicating the presence of an amorphous phase can entirely be explained by the glass capillary used for the XRD measurement as was shown by a measuremment of an empty capillary (Fig. 6-3). The HP2 sample was therefore conssidered to consist of a single phase, the chemical composition determined by PGAA is Al4.19(7)Si0.91(66)B0.90(2)O9.45 (Table 6-2).

Fig. 6-3: Rietveld plot of sample HP2 (top) and the corresponding measurement of an empty capillarry (bottom). Tick marks are shown for B-mullite, * represent peaks not indexed in Pbam but explained by gamma alumina.

72 6 | Ambient and high-pressure synthesis, composition, and crystal structure of B-mullites

Indexing and space group determination with the programs XFIT (Cheary and Coelho, 1996), crysfire (Visser, 1969; Shirley, 2002), and checkcell (Laugier and Bochu, 2004) yielded different orthorhombic unit cell suggestions with either mullite-like lattice parameters or one of them doubled. Suggested space groups were Pbam (mullite) or subgroups of Pbam. Rietveld refinements of the structural model of mullite in Pbam and transformed to Pbn21 (subgroup of Pbnm – sillimanite) were performed. For the latter, the calculation yields small diffraction peaks at 19.5, 22.9, 23.7, and 30.6°2ș which are not observed in the measurement. Therefore the mullite space group Pbam was chosen although the reflections at 46.3 and 67.5°2ș with a relative intensity of 2% each cannot be indexed using this space group. Both reflections could be explained by the introduction of 1 w% gamma alumina as an impurity.

6.3.2.1 11B MAS NMR

So far in B-mullites, boron was only observed in BO3 groups (Lührs et al., 2013a, chapter 5). From the data presented here there are several indications for the presence of a second incorporation mechanism. Furthermore the presence of four coordinated B is clearly supported by the 11B MAS NMR spectrum (Fig. 6-4). The spectrum is dominated by the broad second-order quadrupolar lineshape that was simulated to yield an isotropic chemical shift of įiso = 16.9(1) ppm and a nuclear quadrupole coupling constant CQ = 2.61(1) MHz, both values being clearly indicative for a B- environment with nearly trigonal symmetry (MacKenzie and Smith, 2002). The low asymmetry parameter Ș of 0.12(2) suggests a slight distortion of the trigonal planar BO3 environment as already observed by Lührs et al. (2013a, chapter 5). In contrast to that, BO4 units are typically characterized by resonances at more negative chemical shifts (-4 to 2 ppm). Thus, the second signal of the 11B MAS

NMR spectrum was fitted by a single Gaussian peak at įiso = 0.0(1) ppm and clearly confirms the presence of four coordinated B in the HP2 sample. Similar values for the chemical shifts and CQ were observed for BO3 and BO4 groups in the aluminum borate phase Al4B2O9 by Fischer et al. (2008) and

Fisch et al. (2011). Integration of the two resonances yields about 7% of BO4. However, the uncertainty of this value might be relatively high as despite a very long measuring time (~40 h) the signal is not very good due to the very small amount of sample (less than 10 mg were used).

Fig. 6-4: 11B single pulse MAS NMR measurement of sample HP2 carried out at 14.5 T, with 12.5 kHz MAS (black) and fitted profile (red).

73 6 | Ambient and high-pressure synthesis, composition, and crystal structure of B-mullites

6.3.2.2 Crystal structure refinement For the structure refinement, the crystal structure of 3:2 mullite (Saalfeld and Guse, 1981) was used as a starting model but without T* and O4 positions that result from Si-Al substitution.

Furthermore, the occupancies of T(Si) and O3 were adjusted to a composition of Al8Si2O19 corresponding to the desired composition without B. After a good profile fit was achieved, subsequent difference Fourier calculations yielded maxima near the T* and O4 position as well as a minimum close to the O3 position. This clearly indicates the presence of a mullite like Si-Al substitution mechanism (2 T(Si4+) + 3 O32- o 2 T*(Al3+) + 2 O42- + Ƒ ). The T* and O4 positions were introduced into the structure model, and positional parameters of all atoms were refined, the atomic positions of O4 were fixed after this refinement step. Subsequently, the site occupancy factors (SOF) for T* (= O4), T, and O3 were refined independently. The occupancies were fixed at the refined values rounded to the second decimal digit before displacement parameters were refined in groups of M, T, and O atoms. The composition and B-content derived from this refinement will be discussed in the following. The refinement resulted in a composition of Al4.5Si0.9O9.4. Charge balance can be achieved by introducing 0.59 B atoms which is close to 0.60 B atoms expected from the sum of cations (Al + Si + B) assumed to be 6 resulting from a 1:1 substitution scheme of B replacing Si in mullite.

Thus, the resulting composition is Al4.5Si0.9B0.6O9.4. The occupancy for Al in the T* position (0.6 atoms per unit cell, equivalent with the occupancy of O4) corresponds to twice the number of oxygen vacancies in a pure aluminosilicate mullite, yielding x = 0.3 vacancies per unit cell and consequently

Al4.6Si1.4O9.7 (Al4+2xSi2-2xO10-x) as an initial composition for a pure aluminosilicate mullite before B [3] substitution. Substituting 0.6 Si by B and introducing 0.3 oxygen vacancies yields Al4.6Si0.8B0.6O9.4 which is in good agreement with the experimentally determined amount of Si (0.91(6) a.p.f.u., PGAA) and the 0.6 B atoms needed for charge balance. Residual B not yet accounted for might replace Al on the tetrahedral sites in 4-coordination as indicated by the NMR spectroscopy. Introducing 25% B in

BO4 configuration means replacement of 0.2 Al atoms by B, leading to a composition of

Al4.4Si0.8B0.8O9.4 which would be in good agreement with the PGAA results (Al4.19(7)Si0.91(6)B0.90(2)O9.45).

However, 25% BO4 is not consistent with the NMR analysis yielding only about 7% of the B atoms in

BO4 configuration. Replacement of 0.1 Al by B yields 14% BO4 and a composition of

Al4.5Si0.8B0.7O9.4. Considering the difficulties in refining occupancies of oxygen atoms based on powder diffraction data and to distinguish between Si and Al with similar scattering factors the results of the refinements have higher uncertainties than those obtained from neutron diffraction data. Therefore, the structure model presented here is a result of combined Rietveld and DLS refinement, structure modeling by crystal chemical considerations, PGAA and NMR analyses.

Difference Fourier calculations based on the B-free model (Al4.5Si0.9O9.4) yield a maximum near the B position reported earlier (Lührs et al., 2013a, chapter 5). The local environment of this B position was refined using the DLS software as described in Lührs et al. (2013a, chapter 5), resulting in a low DLS reliability index of 0.018, reasonable B-O (1.3944 Å, 1.3943 Å, 1.4067 Å) distances,

74 6 | Ambient and high-pressure synthesis, composition, and crystal structure of B-mullites and split positions for O1 and O2. Subsequent Rietveld refinement with 0.6 B atoms per unit cell in

BO3 configuration and the corresponding number of split positions for O1 and O2 resulted in an improvement of the R-values. Compared to 3:2 mullite, in this model 40% of the tetrahedral Si is replaced by B in BO3 configuration. The resulting structure model, considering B in BO3 configuration and replacing some of the tetrahedral Al by B, is given in Table 6-3 and Fig. 6-5, the resulting distances and angles are listed in Table 6-4. All distances and angles are well within the expected range for (B-)mullites. Compared to the B-mullite structure of Lührs et al. (2013a, chapter 5) with a composition of Al4.64Si1.16B0.2O9.58, the structure presented here contains three times the amount of B in

BO3 configuration. However, the basic local configuration is similar in both structures as shown in Fig. 5-5 and in Fig. 5-7 of Lührs et al. (2013a, chapter 5). Replacing two neighboring Si4+ atoms by 3+ B yields an oxygen vacancy with adjacent BO3 groups dislocating the octahedral O-atoms. Due to the low amount of B in BO4 configuration in combination with the very low scattering factor of B it is virtually impossible to introduce this position into the structural model for refinements based on X-ray diffraction data. It is therefore presumed that B replaces Al in the tetrahedron, leading to a strong distortion of the tetrahedron itself as well as its next neighbors.

The refined composition of Al4.5Si0.9B0.6O9.4 can be written more clearly with the tetrahedral atoms given in brackets and the occupancies of the different oxygen sites: # Al2 (Al1.84 Si0.92 Al*0.64) B0.6 O13.4 O110.6

O22.8 O2110.6 O2120.6

O30.8

O40.6

Ƒ0.6 Some of the tetrahedral Al (#) is replaced by B.

75 6 | Ambient and high-pressure synthesis, composition, and crystal structure of B-mullites

Fig. 6-5: Crystal structure of B-mullite with the refined composition Al4.5Si0.9B0.6O9.4. Oxygen vacancies are indicated by squares. Only one out of many possible distributions (Fischer et al., 2012) of oxygen vacancies, T3O groups, and BO3 groups is shown. The presence of BO3 groups (red) leads to a distortion of the neighboring octahedra (blue) as indicated by the split positions for O1 (O11) and O2 (O211, O212). Additionally, some of the tetrahedral Al (T) is assumed to be replaced by B.

Table 6-3: Atomic coordinates, site symmetries, Wyckoff positions, site occupancies (occ.), and isotropic displacement parameters for the refinement of sample HP2 with the refined composition of Al4.5Si0.9B0.6O9.4. Pbam: a = 7.508466(1), b = 7.651508(1), c = 2.832082(7), Rp` = 16.69%, Rwp = 4.74%, RB = 4.05%.

atom Wyck. x y z occ. B_iso [Å²] Al 2a 0 0 0 1 0.67(4) T(Al) # 4h 0.1504(3) 0.3365(3) 0.5 0.46 1.26(4) T(Si) 4h 0.1504(3) 0.3365(3) 0.5 0.23 B(T(Al)) T*(Al) 4h 0.265(1) 0.196(1) 0.5 0.16 B(T(Al)) O1 4h 0.3548(4) 0.4178(4) 0.5 0.85 1.53(5) O2 4g 0.1344(5) 0.2166(3) 0 0.7 B(O1) O3 2d 0 0.5 0.5 0.4 B(O1) O4 4h 0.4153 0.0318 0.5 0.16 B(O1) B 4h 0.2022 0.2821 0.5 0.15 0.6 O11 4h 0.3424 0.4041 0.5 0.15 B(O1) O211 8i 0.1401 0.2098 0.9220 0.075 B(O1) O212 8i 0.1403 0.2093 0.0808 0.075 B(O1) # some of the tetrahedral Al is replaced by B

76 6 | Ambient and high-pressure synthesis, composition, and crystal structure of B-mullites

Table 6-4: Selected interatomic distances and angles for the refinement of sample HP2 with the refined composition of Al4.5Si0.9B0.6O9.4. For comparison values are given for 3:2 Mullite (Saalfeld and Guse, 1981) and B-mullite with the composition Al4.64Si1.16B0.2O9.58 (Lührs et al., 2013a, chapter 5). Distances and angles including split positions of O are given in italics.

Distances [Å] Angles [°] 3/2 B- HP2 3:2 B- HP2 mullitea mulliteb mullitea mulliteb M1-site M1 Al-O1 4x (2x) 1.896 1.896(1) 1.895(2) O1-Al-O1 2x 180 180 180 Al-O11 2x 2.009 1.986 O2-Al-O2 1x 180 180 180 Al-O2 2x 1.943 1.922(1) 1.941(3) O1-Al-O1 2x 99.17 98.66(5) 96.7(1) Al-O211 1x 1.945 1.932 O1-Al-O1 2x 80.83 81.34(5) 83.3(1) Al-O212 1x 1.941 1.930 O1-Al-O2 4x 89.80 89.34(5) 90.9(1) B-free mean 1.912 1.905(1) 1.910 O1-Al-O2 4x 90.20 90.66(5) 89.1(1) B cont mean 1.949 1.939 T-site T T-O1 1x 1.700 1.679(3) 1.656(4) O1-T-O3 1x 111.08 110.16(13) 110.0(2) T-O2 2x 1.725 1.725(1) 1.692(2) O1-T-O2 2x 106.69 108.42(12) 105.6(2) T-O211 1.982(1) 1.904(1) O2-T-O2 1x 113.56 112.92(13) 113.7(2) T-O212 1.986(1) 1.913(1) O2-T-O3 2x 109.39 108.45(11) 110.8(1) T-O3 1x 1.658 1.686(2) 1.685(2) mean 1.702 1.704(2) 1.681

T*-site T* Al*-O1 1x 1.814 1.834(15) 1.829(9) O1-T*-O4 1x 106.19 107.95(84) 116.3(5) Al*-O2 2x 1.773 1.792(10) 1.728(6) O1-T*-O2 2x 99.97 99.24(59) 97.1(3) Al*-O4 1x 1.852 1.782(17) 1.689(8) O2-T*-O4 2x 118.96 119.82(52) 116.6(3) mean 1.803 1.800(13) 1.744 O2-T*-O2 1x 108.96 106.74(80) 110.0(5)

B-site B B-O1 1x - 1.61(2) 1.546(3) O1-B-O2 2x 114.8(7) B-O11 1.37(2) 1.407 O11-B-O211 121.77(71) 120.8 B-O2 2x - 1.67(1) 1.586(1) O11-B-O212 121.86(71) 121.0 B-O211 1.39(1) 1.3977 O2-B-O2 1x 118.9(1.1) B-O212 1.39(1) 1.3917 O211-B-O212 114.5(1.3) 117.4 a (Saalfeld and Guse, 1981) b (Lührs et al., 2013a, chapter 5)

6.4 Conclusion

The chemical composition of different B-mullite samples was investigated using PGAA. For single-phase products, 50(1)% of the initial B from the gel enters the mullite structure. The PGAA results are in excellent agreement with the Rietveld refinement based on neutron diffraction data.

The phase described as Al8Si2B2O19 by (Werding and Schreyer, 1992) was synthesized and found to have a composition of Al4.19(7)Si0.91(6)B0.90(2)O9.45 (PGAA). In agreement with Grew et al. (2008), it was found to have mullite-like lattice parameters and no impurities of (disordered) boralsilite. Based on a series of B-mullites with different B-content, linear trends for lattice parameters b and c as well as for the inclination angle Ȧ were observed up to ~15 mol% B2O3. For lattice parameter c the linear trend can be extrapolated towards boralsilite. The strong decrease of lattice parameter a above 5 mol% B2O3 for the high pressure phase (HP2) suggests the presence of a second mechanism for B-incorporation which is also indicated by the number of oxygen vacancies determined by the Rietveld refinement. The existence of about 7% of the B in BO4 coordination was confirmed by the 11B MAS NMR experiment.

77 6 | Ambient and high-pressure synthesis, composition, and crystal structure of B-mullites

By combination of Rietveld refinement (XRD data), DLS refinement, and 11B MAS NMR the composition was determined to be Al4.5Si0.8B0.7O9.4. Considering the relatively big uncertainties in the refinement of the occupancies based on X-ray diffraction data, this composition is in good agreement with the PGAA analysis.

Due to the low amount of BO4 and the low scattering factor of B using XRD, no explicit positions for the four coordinated B can be given here. In comparison to boralsilite, where BO4 tetrahedra are connected to trigonal bipyramids of AlO5, it is expected that in B-mullite some of the B replaces Al in the tetrahedra, resulting in additional distortion of the neighboring AlO6 polyhedra. Acknowledgements We would like to thank the University of Bremen for financial support from the Central Research Development Fund. The authors gratefully acknowledge the financial support provided by FRM II to perform the neutron scattering and PGAA measurements at the Forschungs-Neutronenquelle Heinz Maier-Leibnitz (FRM II), Garching, Germany. At the FRM II we specially would like to thank Anatoliy Senyshyn for the technical support during beam time at SPODI.

7 Chemical composition of B-mullites

This chapter is an enhancement to chapter 6 (Lührs et al., 2013b) and presents a more precise relationship between lattice parameter c and the chemical composition derived from PGAA measurements of B-mullites. It has been pointed out earlier that lattice parameter c is indicative for the amount of boron incorporated into the mullite crystal structure (Lührs et al., 2012, 2013b, chapters 4 and 6). However, the relationships presented before are based on the initial boron content (Fig. 4-9) or estimations of the boron content and neutron diffraction data (Fig. 6-2). In this section a more precisely defined relationship is presented between lattice parameter c, calculated from X-ray diffraction data, and the boron-content determined by PGA analyses.

7.1 Materials and methods

Details regarding the syntheses of the samples presented here were given in the previous chapters of this thesis (3.1, 6.2.1). All precursors were prepared by the sol-gel method, followed by mullitization at 1200°C and ambient pressure (AP0-AP3) or at 875°C and 10 kbar (HP2). X-ray diffraction patterns were acquired on the X’Pert MDP Pro diffractometer using the same configuration as described in section 4.3.2. The lattice parameters were calculated by means of LeBail fits using the Rietveld software package BRASS (Birkenstock et al., 2012) and starting from a 3:2 mullite model (Saalfeld and Guse, 1981). The boron content of the samples was taken from the PGAA analyses presented in chapter 6 (Lührs et al., 2013b) and is given in Table 7-1 together with the initial gel composition and the refined lattice parameters.

Table 7-1: Initial gel composition, PGAA composition, and refined lattice parameters.

initial gel composition PGAA composition* lattice parameters from LeBail fits a [Å] b [Å] c [Å] AP0 Al4.50Si1.45B0.00O9.75 7.5631(3) 7.6909(3) 2.88361(8) AP1 Al4.51Si1.42B0.08O9.71 Al4.58(3)Si1.39(3)B0.032(1)O9.69 7.5580(3) 7.6854(2) 2.88049(8) AP2 Al4.47Si1.23B0.30O9.61 Al4.62(3)Si1.23(3)B0.147(1)O9.62 7.5511(3) 7.6796(3) 2.87161(9) AP3 Al4.50Si1.05B0.45O9.52 Al4.71(3)Si1.06(3)B0.228(3)O9.69 7.5524(3) 7.6784(3) 2.86672(9) HP2 Al4.39Si0.93B1.38O9.46 Al4.19(7)Si0.91(6)B0.90(2)O9.45 7.508466(1) 7.651508(1) 2.832082(7) * PGAA compositions from chapter 6 (Lührs et al., 2013b, Table 6-2).

7.2 Results and discussion

The relationship between lattice parameters and boron-content is given in Fig. 7-1 for different

B-mullites containing up to 13 mol% B2O3. With respect to the whole compositional range up to

13 mol% B2O3, linear fits for all three lattice parameters appear to be reasonable, with coefficients of

79 7 | Chemical composition of B-mullites correlation (R2) above 98%. However, the variations in lattice parameters a and b at low boron- contents are rather large and do not strictly follow linearity, especially in the case of a (c.f. Fig. 4-9, Fig. 6-2). Excluding the sample with the highest boron content (HP2) from the linear fits in Fig. 7-1 yields significantly lower coefficients of correlation for lattice parameters a (R² = 0.7678) and b (R² = 0.8677) whereas for lattice parameter c there is no significant change (R² = 0.9928). Furthermore the relative changes in the lattice parameters (d/d0, Fig. 7-1d) are significantly higher for c (1.8%) compared to a (0.7%) and b (0.5%). Considering the strong dependence of lattice parameter a on the

Al2O3/SiO2 ratio (Fischer et al., 1996) and the very small relative variation in b, it is recommended to use the following equation in order to calculate the boron content mB2O3 in mol% of B-mullites from lattice parameter c in Å:

mB2O3 = (c - 2.881(1)) / - 0.0038(2) (3)

For several pure B-mullites the B2O3 content was calculated based on this relationship and found to be in good agreement with the composition calculated from the correlation between the initial

B2O3 content and the PGA analyses (Fig. 6-1). Single phase B-mullite with 3.2 mol% B2O3 was synthesized at 1200°C (AP3), corresponding to a replacement of ~15 % of the silicon in 3:2 mullite by boron. The B-mullite with the highest B-content synthesized at 1200°C contains alumina impurities (Fig. 4-9) and yields a lattice parameter c = 2.8393(3) Å. According to (3) it consists of 11.0(9) mol%

B2O3, theoretically corresponding to a replacement of ~50 % of the silicon by boron. However, in this case the replacement of tetrahedral aluminum by boron has also to be considered as, in contrast to the single-phase B-mullite, no 11B MAS NMR data exists for B-mullite samples containing alumina impurities.

Fig. 7-1: (a)-(c) Lattice parameters and B2O3 content of B-mullite samples synthesized under different temperature and pressure conditions. (d) Fractional lattice parameters are given as a function of the B2O3content (a/a0, b/b0, c/c0 with index 0 referring to the lattice parameter of boron-free mullite).

8 Crystal chemistry of mullite and B-mullite at high pressure

In the following section a comparative high pressure study is presented of B-mullite and boron- free mullite. The intensive study first represents the change of the lattice parameters in response to pressure followed by crystal chemical interpretations. This chapter is in preparation for publication in a scientific journal in cooperation with the colleagues P. E. Kalita, K.E. Lipinska2, University of Nevada Las Vegas, who performed the in-situ high-pressure experiments.

8.1 Introduction

A number of investigations on the bulk compressibility of sillimanite (e.g., Brace et al., 1969) and 3:2 mullite (e.g., Balzar and Ledbetter, 1993) was reported. The crystal structure of sillimanite at high pressure was studied by Burt et al. (2006) and Yang et al. (1997) using high-pressure single crystal X-ray diffraction up to 8.5 GPa and 5.29 GPa, respectively. Both authors observed anisotropic axial compression of sillimanite in the order b > a > c, as b is the most compressible. The relatively small compressibility along the c-axis is usually explained by the rigidity of the polyhedral chains along this direction. In contrast to that changes in the inter-polyhedral angles or relative orientations of the polyhedra in the ab-plane (out-of-plane tilting, rotation of octahedra) can readily be achieved and lead to a higher compressibility perpendicular to the c-axis. The major compression mechanism is reported to be the shortening of bond lengths within the AlO6 octahedra, with the longest bond (Al- O2) being the most compressible one. Thus, explaining the stronger compressibility in b-direction compared to the a-direction, as the long Al-O2 bond encloses a small angle of 30° with the b-axis. In contrast to the significant compression of the octahedra (7.4 % of volume reduction up to 7.6 GPa), almost no compression was observed in the tetrahedra in the investigated pressure range (Burt et al., 2006). The response of sillimanite to pressures up to 46 GPa was investigated by Friedrich et al. (2004) using powder synchrotron X-ray diffraction and a diamond anvil cell (DAC). The c-axis was found to be the least compressible, and a is slightly more compressible than b. The same trend for the pressure- dependent lattice parameters a and b was observed by Kalita et al. (2013) who demonstrated the first in-situ static-pressure study of 3:2 mullite, 2:1 mullite and sillimanite using powder synchrotron X-ray diffraction. According to them the oxygen vacancies play an important role in the compression

2 Not yet published as the copyright on the experimental data is at the University of Nevada, Las Vegas, who reserves the right of primary publication of the data in a broader context.

81 8 | Crystal chemistry of mullite and B-mullite at high pressure mechanisms of the mullites studied; the meechanical stability was found to bee a function of the concentration of the oxygen vacancies. Pressures above ~20 GPa for 2:1 mullite and ~22 GPa for 3:2 mullite lead to irreversible amorphization, whereas for sillimanite only partial amorphization was observed above 30 GPa. Pressure dependent unit cell parameters and bulk modulli are presented for sillimanite, 3:2 mullite and 2:1 mullite and some crystal chemical hypotheses are giiven by Kalita et al. (2013). In this study the aim is to compare thee response to pressure of B-mullite to boron-free mullite with an emphasis on the concerned crystal chemical changes. Additionally, more crystal chemical information of 3:2 mullite is extracted to justify the hypotheses (Kalita et al., 2013)..

8.2 Material and methods

8.2.1 Synntheses and samples

Thee in-situ high-pressure synchrotron X-ray diffraction data of three mullite samples was investigated. New refinements of the 3:2 mulliite of Kalita et al. (2013) allow for more detailed crystal chemical interpretations, additionally a boron-free mullite and a B-mullite were investigated. The detailed synthesis of B-mullite and boron-freee mullite are available elsewhere (chapters 3.1 and 4.2).

The precursor gels were prepared from aluminum nitrate nonahydrate (Al(NO3)3•9H2O), tetraethoxysilan (C8H20O4Si), and boric acid (H3BO3). Stoichiometric amounts were dissolved in pure ethanol at 60°C, resulting in a transparent sol that was then kept in a drying oven att 60°C for 4 days to allow for gelation. The resulting gel was then dried for 5h at 150°C, ground and calcined at 350°C for 5 h before annealing at 1200°C for another 5 h. Baseed on in-house X-ray diffraction (performed at ambient condition) data Rietveld refinement, the chemical compositions of the two samples were determined. For the boron-free sample the linear relationship between lattice parameter a and the Al2O3 content (Fischer et al., 1996) was used. The sample was found to consist of 63.4 mol % Al2O3, corresponding to a Al2O3:SiO2 ratio of 7:4, which is a composition between 3:2 and 2:1 mullite, this sample is therefore designated as “7:4 mullite”. For the “B-mullite” the boron-content was estimated to be 3.5(4) mol % B2O3 based on an established relationship (chapter 7, Fig 7-1). The 3:2 mulliite from Kalita et al. (2013) was synthesized at 1600°C and contains 61.6 mol% Al2O3.

8.2.2 Laboratory powder X-raya diffraction

A Bragg-Brentano PANalytical X’Pert MPD PRO diffractometer was used for the measurements at ambient conditions. The instrument is equipped with Cu-KĮ radiation (Ȝ = 1.5418 Å), ¼° fixed divergence, primary and secondary Solller slits with 0.04 rad aperture, secoondary Ni-filter and X’Celerator detector system (127 channels, channel width 0.01671°2ș). The sammpples were prepared with the standardized PANalytical backloading system using circular sample holders with 16 mm

82 8 | Crystal chemistry of mullite and B-mullite at high pressure diameter. Scans were performed in the range from 3-140°2ș, step width 0.0167°2ș; the measuring time peer step was 25 seconds.

8.2.3 High-pressure synchrotron X-raay diffraction using a Diamond AAnvil Cell (DAC)

The instrumental setup for the in-situ high-pressure diffraction experiments corresponds to that described by Kalita et al. (2013) and illustrated in Fig. 8-1. Pressure dependent in-situ, angle- dispersive, synchrotron X-ray diffraction (ADXRD) measurements were performed by Patricia Kalita and Kristina Lipinska (UNLV) at the 16-IDB beam-line of the High Pressure Collaborative Access Team (HPCAT), Advanced Photon Source, Argonne National Laboratory. The 320 x 300 μm² monochromatic X-ray beam, with a wavelength Ȝ = 0.398160 Å (3:2 mullite) oor Ȝ = 0.373790 Å (7:4 mullite, B-mullite) was focused down to a ~7 x 5 μm spot using Kirkpatrick-Baez 200 mm mirrors. Measuring times between 30 s and 6 min were applied and the intensities recorded using a MAR345 imaging plate detector. The intensities were integrated and corrected for distortion using the FIT2D software (Hammersley, 2005) after maskiing overexposed spots on the image plate due to sample texture. Calibration of the sample to detector distance and geometric parameters was done at the beginning of each high-pressure run, using a CeO2 standard refef rence material from the National Instituute of Standards and Technology. The samples “were compressed in a symmetric type diamond anvil cell (DAC), at ambient temperature. Rhenium gaskets were pre-indented to a thickness of ~40 μm,, using diamonds with 300 μm diameter culets. The sample chammber consisted of a 120 μm hole, drillled in the pre-indented rhenium gasket. For accurate pressure reeadings during compression, fragments of gold foil were mixed in with the samples and used to measure pressure using the equation of state of gold by Anderson et al. (1989). For the gas-loadiing procedure only, a ruby microsphere was placed in the sample chamber and used to measure pressure using the pressure scale of Mao et al. (1978). To insure quasi-hydrostatic pressure conditions, all compression runs were carriieed out with … “ neon (7:4 mullite, B-mullite) or helium (3:2 mullite) “… as a quasi-hydrostatic pressure-transmitting medium (Takemura, 2001; Dewaele and Loubeyre, 2007), which was loaded into the sample chamber at about 1.38 x 108 Pa (20000 psi), using the gas-loading setup of Sector 13 of APS, ANL (Rivers et al., 2008). The interval between a pressure increase and the X-ray measurement was kept at several minutes, to allow for the pressure inside tthe sample chamber to equilibrate. Pressure increase was controlled remotely using a motorized mechanical pressure control device off ‘gearbox’.” (Kalita et al., 2013, p. 1636). The Rietveld refinements of diffraction patterns were performed using the program Topas (Bruker, 2009). For B-mullite 27 pressure points from 0.65 to 28.8 GPaa were analyzed, for 7:4 mullite 17 pressure points from 1.0 to 28.0 GPa, and for the 3:2 mullite sample of Kalita et al. (2013) 23 pressure points between 1.3 and 27.3 GPa were investigated.

83 8 | Crystal chemistry of mullite and B-mullite at high pressure

Fig. 8-1: Schematic (not to scale) setup of the in situ high-pressure synchrotron X-ray diffraction experiments in a DAC. The synchrotron X-ray beam is focused down to a few micrommeters, passes one diamond anvil, and reaches the sample chamber. The image plate detector records the three-dimensional Debeye diffraction rings which are then integrated to yield a two dimensional diffraction pattern. From Kalita et al. (2013).

8.2.4 Rietveld refinements

As a starting model for the Rietveld refinements the 3:2 mullite crystal structure (Saalfeld and Guse, 1981) was used. In the case of 7:4 mullite the site occupancies in the sttarting model were modified to account for the 7:4 composition. The peak profiles were only refined for the measurement at lowest pressure using the PV_TCHZ function in Topas (Bruker, 2009) and keptt at these values for subsequent refinements. This allows the determination of a relative strain parameteer for measurements at elevated pressures. Lattice parameters, atomic positions (dependently for T(Si) and T(Al)), Lorentzian strain, and crystallite size parameters were refined. The displacement parameters were only refined for the measurement at the lowest pressure for octahedral Al1 and in ttwo groups for the tetrahedral (T) and oxygen (O) positions. All displacement parameters were kept at these values for subsequent refinements. Only scale factor, lattice parameters, strain, and crystallitte size were refined for the pressure indicator gold. Diffraction peaks of the pressure transmitting medium neon occur at pressures above 12.4 GPa for 7:4 mullite and above 8.2 GPa for B-mullite. In 3:2 mullite about 1.5 w% corundum is present as an impurity, only the scale factor of corundum was refined whereas the lattice parameters were calculated from pressure dependent lattice parameters given by Hart and Drickamer (1965).

8.3 Results and discussion

Thee in-situ high-pressure diffraction patterns of 7:4 mullite and B-mullite from 1 to 28.0 GPa and from 0.65 to 28.8 GPa, respectively, are given in Fig. 8-2. A shift of the diffraction peaks towards higher 2ș values is observed as well as a significant broadening of the difffrraction peaks with increasing pressure. Starting from 12.4 GPa for 7:4 mullite and 8.2 GPa for B-mullite, diffraction peaks of the pressure transmitting medium neon are present. Complete amorphization of mullite is

84 8 | Crystal chemistry of mullite and B-mullite at high pressure observed above 25 GPa for 7:4 mullite and above 23 GPa for B-mullite and only diffraction peaks of neon and gold remain.

Fig. 8-2: Pressure dependent diffraction patterns of 7:4 mullite and B-mullite between 1 and 28 GPa, the diffracted intensities are given in scales of gray and are interpolated between the individual measurements (horizontal lines). Representative diffraction patterns for individual pressure points are given in red, the diffraction peaks/lines of gold (Au) and neon (Ne) are marked.

8.3.1 Pressure dependent lattice parameters

All three investigated samples (3:2 mullite, 7:4 mullite, and B-mullite) show a similar pressure evolution of the relative strain parameter that takes place in three steps (Fig. 8-3). After an initial increase of strain (step 1) the second stepp (gray in Fig. 8-3) is characterized by a range of constant strain followed by a second rise of the straain parameter in step 3. The range off constant strain (step 2) is distinctly different for the three mullites, ranging from about 9-22 GPa for the 3:2 mullite of Kalita et al. (2013), from ~7.5 to 13 GPa for the 7:4 mullite, and from ~5-11 GPa for tthe B-mullite. In step 3 the increasing strain is accompanied by a significant broadening of diffracttion peaks and loss of intensity, indicating gradual structural deggrradation of the samples (Fig. 8-2). Reasonable refinements were possible up to about 27 GPa, 25 GPPa, and 23 GPa for 3:2 mullite, 7:4 mmullite, and B-mullite, respecttively. However, due to the low quality of the data at higher pressures (~ step 3 in Fig. 8-3), the refined atomic parameters, and all crystal chemical information calculated from the latter, have to be interpreted cautiously. The decreasing structural stability at high pressures from 3:2 mullite to B-mullite can be explained by the increasing number of oxygen vacancies, whiich were shown to have a significant influence on the resistance of the structure against pressure (Kalita et al., 2013). The estimated numbers of oxygen vacancies arre x = 0.25, 0.33, and 0.4 for 3:2 mullite, 7:4 mullite, and B-mullite, respectively.

85 8 | Crystal chemistry of mullite and B-mullite at high pressure

Fig. 8-3: Pressure evolution of the relative Lorentzian Fig. 8-4: Pressure dependent lattice parameters c for 3:2 strain (strain = 0 for measurement at lowest pressure) in mullite (circles), 7:4 mullite (squares), and B-mullite three steps: Step 1 initial increase in strain, step 2 range of (triangles). At ~6, 8, and 10 GPa, respectively, the slope constant strain (gray), step 3 increase of strain indicating changes, indicated by the red and blue symbols and slopes. incipient amorphization.

Similar to the pressure evolution of the strain (Fig. 8-3), a three-step compression mechanism of lattice parameter c was described for sillimanite, 3:2 mullite, and 2:1 mullite (Kalita et al., 2013). Lattice parameter c decreases almost linearly up to 20 GPa with a small reduction of the slope at 10 GPa (mullite) or 20 GPa (sillimanite), followed by a dramatic decrease above 20 GPa and 30 GPa for mullite and sillimanite, respectively. In this work the discontinuity is observed at 10 GPa (3:2 mullite), ~8 GPa (7:4 mullite), and ~ 6 GPa (B-mullite), corresponding to the onset of step 2 in the evolution of the strain parameter (Fig. 8-3). The change in the slope is highest for B-mullite and lowest for 3:2 mullite and therefore increasing with the number of oxygen vacancies in the structure. It emerges from Fig. 8-5 that in a- and b-direction the mullite structure is distinctly more compressible than in c-direction. At low pressures, lattice parameters a and b display nearly identical linear pressure evolution, however, at higher pressures a is more compressible than b. The divergent behavior of compressibility in a- and b-direction starts at 13 GPa (3:2 mullite), 9 GPa (7:4 mullite), and 7 GPa (B-mullite) within step 2 of the strain evolution and at slightly higher pressures than the changes in the slope of lattice parameter c.

86 8 | Crystal chemistry of mullite and B-mullite at high pressure

Fig. 8-5: Fractional lattice parameters a/a0, b/b0, c/c0 for 3:2 mullite (black, bottom), 7:4 mullite (red, middle), and B-mullite (blue, top). The index 0 in the fractional lattice parameters indicates the lattice parameter at the lowest pressure.

This anisotropic compression behavior of mullite-type aluminosilicate structures is commonly explained by the firmly bound edge-sharing AlO6 octahedra and corner-sharing double chains of SiO4 and AlO4 tetrahedra along the c-axis. In contrast to that compression in the a-b plane can be more easily achieved by out-of-plane tilting of the tetrahedra or rotation of octahedra as described for sillimanite (Yang et al., 1997; Burt et al., 2006) and hypothesized for mullite (Kalita et al., 2013). In the closely related sillimanite structure the main compression mechanism below 8 GPa was found to be the shortening of octahedral bonds, whereas the tetrahedra behave as rigid units (Yang et al., 1997; Burt et al., 2006). Furthermore the compressibilities of the individual Al-O distances in sillimanite differ significantly. The slightly stronger compression in b-direction is explained by the highest compressibility of the longest Al-O bond that encloses a small angle (30°) with the b-axis (Yang et al., 1997; Burt et al., 2006). In contrast to that, this work yields the highest compressibility in a-direction for all samples (Fig. 8-5) at high pressures (> 10GPa), which is in good agreement with the results of Kalita et al. (2013) and Friedrich et al. (2004), who also observed the a-axis of sillimanite to be more compressible than the b-axis for pressures up to 40 GPa. However, below 10 GPa in all three studies based on powder diffraction data for mullite and sillimanite (Friedrich et al., 2004; Kalita et al., 2013; this work) the compressibilities of the a- and b-axis are very close to each other. A very close look at the sillimanite data below 10 GPa of Kalita et al. (2013) suggests a slightly higher compressibility in b-direction. Thus the difference to the observations based on single-crystal investigations up to 8 GPa (Yang et al., 1997; Burt et al., 2006) is negligible. However, no crystal chemical explanation for the higher compressibility in a-direction is available so far.

87 8 | Crystal chemistry of mullite and B-mullite at high pressure

In Fig. 8-6 the pressure dependent fractional lattice parameters a/a0, b/b0, and c/c0 are given for all three mullite samples. In all cases the pressure evolution of lattice parameters a and b is almost identical up to 20 GPa. The compression at 20 GPa of the a-axis is 4 - 4.5 % and the compression of the b-axis is between 3.4 and 3.9 %, which is in good agreement with the observations of Kalita et al. (2013) for sillimanite, 3:2 mullite, and 2:1 mullite. However, lattice parameter c for the samples of this work (7:4 mullite and B-mullite) is significantly more compressible (2.5 %) than lattice parameter c of 3:2 mullite (1.5 %, Fig. 8-6c). It is therefore also different from 2:1 mullite and sillimanite as these are reported to have a similar compressibility as 3:2 mullite up to 20 GPa (Kalita et al., 2013). The differences in pressure evolution of the c lattice parameter is also obvious in Fig. 8-4, where the two boron-free mullites have similar values at the lowest pressure but the difference is constantly increasing with pressure. Similar volume compressibilities are observed for all three samples with a slight increase in compressibility with an increasing number of oxygen vacancies in the structure (3:2 mullite < 7:4 mullite < B-mullite).

Fig. 8-6: Pressure dependent fractional lattice parameters a/a0 (a), b/b0 (b), and c/c0 (c) and fractional change in volume V/V0 (d) for 3:2 mullite (black), 7:4 mullite (red), and B-mullite (blue). The index 0 in the fractional lattice parameters indicates the lattice parameter at the lowest pressure.

There is no simple explanation for the distinctly different behavior of lattice parameter c for the samples of this work compared to 3:2 mullite, 2:1 mullite and sillimanite (Kalita et al., 2013). The difference on hand is the synthesis process that involved temperature of 2000°C and 1600°C for 2:1 mullite and 3:2 mullite, respectively. Although the chemical composition (Al/Si ratio) is a function of the annealing temperature in mullite solid solution, the influence on lattice parameter c is rather small

88 8 | Crystal chemistry of mullite and B-mullite at high pressure compared to the changes in lattice parameter a. Possibly the relatively low synthesis temperature of 7:4 mullite and B-mullite causes some disorder in c-direction which in turn leads to a higher comprressibility of the c-axis. The fact that both samples prepared at 1200°C in this work show similar differences compared to the mullites from Kalita et al. (2013) once more emphasizes that the synthesis process itself has a big influence on the physical properties of the mullite product (Komarneni et al., 2005).

8.3.2 Crystal chemical changes in mullite and B-mullite upon pressure

The following section addresses thee crystal chemical interpretation of the response to pressure of the different mullites. Some general trends can be given but due to the system-inherent relatively low quality of the data, no further interpretations are possible. The polyhedrall distances, angles, and volummes were calculated from the refined structures as well as the polyhedraal indices of distortion according to Baur (1974). Furthermore, the inclination angle Ȧ and the outt of-plane tilting angle (OPTA) is presented. As illustrated in Fig. 8-7, the inclination angle Ȧ is defined as the angle between two adjacent octahedra in the a-b plane and the OPTA as the angle between tthe (100) plane and the tetrahedral face defined by the O2 and O3 position. For 7:4 mullite at 12.4 GPa several refined values are not within the general trend observed, this pressure point is therefore considered an outlier and not considered in Fig. 8-8 to Fig. 8-11.

Fig. 8-7: Projection along c of the mullite crystal structure showing the sillimanite-like linkagge of mullite-type octahedral (blue) annd tetrahedral (green) chains. The inclination angle Ȧ and the out of-plane tilting angle (OPTA) are indicated as well as three (I to III) additional angles between polyhedrral planes.

89 8 | Crystal chemistry of mullite and B-mullite at high pressure

In sillimanite the compressibility of the octahedral bonds is a function of their lengths, with the longer bonds being more compressible (Yang et al., 1997). This is not unambiguously the case for the data presented here (Fig. 8-8). Whereas for 7:4 mullite and B-mullite the longer Al-O2 bond seems to be slightly more compressible than the shorter Al-O1 bond, for 3:2 mullite no difference in compressibility is observed up to 10 GPa. The different behavior of mullite compared to sillimanite is probably due to the fact that the octahedral distances are closer to each other in 3:2 mullite, with 1.896 Å and 1.943 Å (Saalfeld and Guse, 1981) than they are in sillimanite where the octahedral distances are between 1.863 Å and 1.953 Å (Burt et al., 2006). This also results in lattice parameters closer to each other in mullite compared to sillimanite: For 3:2 mullite the lattice parameter a = 7.553(1) and b = 7.686(1) (Saalfeld and Guse, 1981) whereas for sillimanite a = 7.48388(17) and b = 7.6726(3) (Burt et al., 2006). Taking a closer look at the polyhedral volumes Fig. 8-9, a comparable decrease in the volumes of octahedra and tetrahedra is observed for 3:2 mullite and 7:4 mullite. In contrast, the tetrahedral volume in B-mullite is almost constant over the whole pressure range whereas the octahedral volume decreases continuously. This is in good agreement with the almost constant tetrahedral indices of distortion between 7 and 14 GPa (Fig. 8-10). Thus, the incorporation of boron into the crystal structure of mullite seems to stabilize the tetrahedra.

Fig. 8-8: Fractional change of octahedral distances Al-O1 Fig. 8-9: Pressure-dependent fractional volume of the (black line) and Al-O2 (red line). octahedra (gray squares) and tetrahedra (red triangles), the same evolution is observed for the mean distances in the octahedra and tetrahedra.

90 8 | Crystal chemistry of mullite and B-mullite at high pressure

A similar pressure evolution of the indices of distortion for the polyhedral distances and angles is observed for B-mullite and 7:4 mullite (Fig. 8-10). The distortion of the octahedral distances is rather constant up to 17 GPa. On the contrary, the distortion of the octahedral angles increases above 10 GPa and 7 GPa for 7:4 mullite and B-mullite, respectively. These pressures correspond to the observed discontinuity for the compression in c-direction (Fig. 8-4). For 3:2 mullite both octahedral distortion indices are more or less constant. Slightly higher octahedral distortions are observed for the boron-containing mullite compared to the 7:4 mullite. This is in good agreement with the fact that in the mullite structure the BO3 groups lead to a significant distortion of the octahedra in the first coordination sphere of boron (Lührs et al., 2013a, 2013b, chapters 5 and 6). The tetrahedral angle distortions show an increasing trend upon pressure, whereas the distortion of the tetrahedral distances remains more or less constant up to 15 GPa. Generally the polyhedra are relatively rigid with respect to their distances whereas the angles have a stronger tendency for distortion. This is in good agreement with Yang et al. (1997), who report the quadratic elongation for the polyhedra to be rather stable whereas the angle variance increases with pressure. The quadratic elongation is a measure for the distortion of the distances in the polyhedra and the angle variance measures the angular distortion.

Fig. 8-10: Octahedral (a) and tetrahedral (b) distortion indices according to Baur (1974) for distances (filled symbols) and angles (open symbols) for 3:2 mullite (black, circles), 7:4 mullite (red, squares) and B-mullite (blue, triangles).

For the slowdown of the compression in c-direction (Fig. 8-4) Kalita et al. (2013) suggest that the rigid polyhedral chains reach a lower limit of cation-oxygen bond lengths. This is supported by the increase in the index of distortion for the octahedral angles (Fig. 8-10) and the fact that the decrease in the Al-O2 distance slows down slightly at higher pressures (Fig. 8-8). All studies of sillimanite and different mullites at pressures above 10 GPa yield a stronger compression of the a-axis compared to the b-axis (Friedrich et al., 2004; Kalita et al., 2013; this

91 8 | Crystal chemistry of mullite and B-mullite at high pressure work). This is in contrast to the observations on single crystals of sillimanite (Yang et al., 1997; Burt et al., 2006), where the b-axis was reported to be more compressible than the a-axis up to 8 GPa. The stronger compressibility of the b-axis is usually explained by the more compressible Al-O2 bond that encloses an angle of about 30° with the b-axis. It was suggested before (Kalita et al., 2013) that the stronger compressibility of the a-axis at higher pressures can be explained by a rotation of the octahedra leading to an increasing angle between the b-axis and the Al-O2 bond and therefore reduces the compressibility in b-direction. The increasing inclination angle Ȧ (Fig. 8-11 a) indicates a rotation of the octahedra and an increase of the angle between the Al-O2 bond and the b-axis. This is in good agreement with the increasing difference between the compressibility in a- and b-direction but does not sufficiently explain the higher compressibility along the a-axis. The second frequently used inter- polyhedral angle (OPTA) is the out of-plane tilting angle that describes the change in relative orientations of the tetrahedra (Fig. 8-7). Yang et al. (1997) compare and explain the structural changes with pressure and temperature of andalusite, kyanite, and sillimanite based on the OPTA and the corresponding clockwise or anticlockwise rotations of the different tetrahedra. This works well for sillimanite where SiO4 and AlO4 tetrahedra can be distinguished and rotate independently. However, for mullite there is only one OPTA that is calculated based on the coordinates of the O2 atomic position and therefore directly correlated to the inclination angle Ȧ. A similar increase is observed for OPTA and Ȧ, thus, not yielding any additional information (Fig. 8-11 b).

Fig. 8-11 (a): Pressure evolution of the inclination angle Ȧ, calculated as the angle between adjacent octahedra in the a-b plane. (b): Out of-plane tilting angle (OPTA), calculated as the angle between the (100) plane and the tetrahedral face through O2, O2, and O3.

As the highest compressibility in a-direction was also observed for sillimanite above 10 GPa, the oxygen vacancies cannot play a significant role in this process. The polyhedra are expected to be rather rigid and the normally used inter-polyhedral angles Ȧ and OPTA cannot sufficiently explain the observed effect. In Fig. 8-7 additional angles between polyhedral planes in the mullite structure are indicated as angles I to III and a first hypothesis is given if the respective angles are expected to increase (Ĺ) or decrease (Ļ) upon compression along the a-axis and b-axis (Table 8-1). The calculation and interpretation of these angles could be subject to further investigations based on this data.

92 8 | Crystal chemistry of mullite and B-mullite at high pressure

Table 8-1: Angles between polyhedral planes in the mullite structure according to Fig. 8-7 with a first hypothesis if the respective angles are expected to increase (Ĺ) or decrease (Ļ) upon compression along the a-axis and b-axis.

angle compression along a compression along b I Ĺ Ļ II Ļ Ĺ III Ĺ Ļ

8.4 Summary, conclusion, and outlook

The response to pressure of different mullites and B-mullite compared to sillimanite is highly interesting and has the potential to lead to a better understanding of the basic physical properties of aluminosilicates. Due to the fact that the high-pressure data of B-mullite were provided less than four weeks prior to submission of this thesis, some questions had to remain unanswered. Oxygen vacancies were previously demonstrated to reduce the mechanical stability of the mullite-type structure (Kalita et al., 2013). The incorporation of boron involves the formation of additional vacancies and therefore the incipient amorphization is shifted to lower pressures compared to boron-free mullite. Furthermore B-mullite yields a slightly increased overall (volume) compressibility and increased distortion of the octahedral distances compared to boron-free mullite. Generally the response to pressure is similar as observed for the closely related sillimanite and boron- free mullite with the c-axis being the least compressible. A three-step compression mechanism in the c-direction, similar to the findings of Kalita et al. (2013), was observed for all samples. The slowdown of compression in c-direction increases the difference in compressibility between the c-direction and the a-b-plane and leads to an increasing angular distortion of the octahedra. With an increasing number of oxygen vacancies in the structure, the change in the slope of the linear compression increases and the discontinuity is shifted to lower pressures. Despite the higher number of oxygen vacancies and reduced stability of the structure, the tetrahedra seem to be stabilized/non-responsive upon boron incorporation, as the tetrahedral volume and the tetrahedral distortion indices remain constant over a wide pressure range. To date the stronger compressibility in b-direction than in a- direction for sillimanite below 10 GPa was explained by the small angle (30°) between the most compressible octahedral bond and the b-axis (Burt et al., 2006). However, for sillimanite above 10 GPa and different mullites the compressibility in a-direction is higher than in b-direction (Friedrich et al., 2004; Kalita et al., 2013; this work). In contrast to sillimanite (Yang et al., 1997) no strong dependence of the compressibility of the octahedral bonds on their length was observed, which can be partially explained by the smaller difference between the two octahedral bonds compared to sillimanite. The increasing inclination angle Ȧ can explain the increasing divergence between the compressibility in a- and b-direction but does not justify the stronger compressibility in a- than in b- direction. Here, a detailed analysis of the angles between the polyhedral planes might yield valuable information. A first hypothesis about the expected relative changes in the angles upon compression in

93 8 | Crystal chemistry of mullite and B-mullite at high pressure a- and b-direction was given and can be used as a basis for future investigations. In combination with the data presented in chapter 4, a comparative study between the response to pressure and temperature of mullite and B-mullite would yield interesting crystal chemical information. One significant difference compared to earlier high-pressure studies (Kalita et al., 2013) on mullites and sillimanite is the distinctly higher compressibility in c-direction of 7:4 mullite and B- mullite compared to 3:2 mullite, 2:1 mullite, and sillimanite. Due to the relatively low synthesis temperature (1200°C) of 7:4 mullite and B-mullite these phases may be distorted along to the crystallographic c-axis which may increase their compressibilities. It would be highly interesting to directly compare the pressure response of the different mullites to the crystal chemical changes in sillimanite at higher pressure. Sillimanite powder diffraction data up to 38 GPa from Kalita et al. (2013) was provided. However, no stable refinement of the three octahedral oxygen positions was possible due to the system-inherent low quality of the high-pressure data. Additionally the number of refineable parameters is significantly higher in sillimanite (13 atomic coordinates) compared to mullite with only 6 atomic coordinates (without tricluster positions T* and O4). In-situ high pressure diffraction experiments on mullite single crystals would also yield more reliable crystal chemical data for the response to pressure. Acknowledgements In situ high-pressure x-ray diffraction measurements and diamond anvil cells preparation were carried out by Patricia Kalita, Andrew Cornelius (Univ. of Nevada Las Vegas, Dept. of Physics and Astronomy) and Kris Lipinska (Univ. of Nevada Las Vegas, Dept. of Mechanical Engineering, FAME-Tech Labs)and Stanislav Sinogeikin (Carnegie Institution of Washington, Geophysical Lab). A portion of high-pressure x-ray diffraction data analyses were carried out by Patricia Kalita, (Univ. of Nevada Las Vegas, Dept. of Physics and Astronomy) and Kris Lipinska (Univ. of Nevada Las Vegas, Dept. of Mechanical Engineering, FAME-Tech Labs). High-pressure x-ray diffraction experiments were done at HPCAT (Sector 16), Advanced Photon Source (APS), Argonne National Laboratory. HPCAT operations are supported by DOE-NNSA under Award No. DE-NA0001974 and DOE-BES under Award No. DE-FG02-99ER45775, with partial instrumentation funding by NSF. APS is supported by DOE-BES, under Contract No. DE-AC02-06CH11357. Work at UNLV is supported by DOE award No. DEFG36-05GO08502. The UNLV High Pressure Science and Engineering Center is supported by the US Department of Energy, National Nuclear Security Administration, under Cooperative Agreement DE-FC08-01NV14049. Patricia Kalita acknowledges the use of the gas loading facility for diamond anvil cells at GSECARS-COMPRESS (Sector 13) of the Advanced Photon Source, Argonne National Laboratory). Diamond anvil cells preparation to carry out high-pres sure x-ray diffraction measurements was done using experimental facilities of FAME-Tech Labs (Univ. of Nevada Las Vegas, Dept. of Mechanical Engineering). FAME-Tech Labs are supported by DOE-EERE under Award No. DE-FG 3606GO8636 and Award No. EE-0000269.

9 Conclusion and future perspectives

The major findings in this thesis regarding the crystal structure and properties of B-mullites, synthesized at different pressure and temperature conditions, are summarized with respect to the three previously defined objectives (chapter 2.1). Concluding this thesis a short outlook on experiments in progress and future perspectives is given.

9.1 Objectives of this thesis

Objective 1: Syntheses and properties of B-mullites [chapters 4 and 8] In contrast to other authors working on boron-doped mullites (Griesser et al., 2008; Zhang et al., 2010; Fisch, 2011), the hypothesis for this study was a 1:1 isomorphous substitution of silicon by boron starting from a 3:2 mullite composition (Al4.5Si1.5O9.75). The formation conditions and basic properties of sol-gel derived B-mullites were studied. With increasing boron-content the formation temperature for B-mullites significantly decreases and a significant decrement in lattice parameters b and c is observed. On the contrary lattice parameter a seems to be virtually unaffected by the incorporation of boron. This anisotropic behavior of lattice parameters was also reported by other authors (Griesser et al., 2008; Zhang et al., 2010) and is explained based on crystal chemical considerations in this work (c.f. objective 2). All samples of this work synthesized at 1200°C and 1300°C yield B-mullite in space group Pbam accompanied by alumina impurities at higher boron- contents. A mullite-type aluminum borate phase is only observed for samples with more than 15 mol%

B2O3 in the precursor. Fundamentally new details on the thermo-elastic properties of B-mullites are presented in this thesis. On the one hand long-term stability of B-mullite at 800°C is demonstrated whereas on the other hand long-term annealing at 1400°C results in complete decomposition of B-mullite to boron-free mullite and corundum. An important finding is the reduction of the mean thermal expansion coefficient by maximally 15%, from 5.9(1) × 10-6°C-1 for boron-free mullite to 5.0(1) for B-mullite. The significantly reduced thermal expansion makes B-mullites technologically interesting for applications at moderate temperatures below 1000°C. For the first time a comparative study of the response to pressure of B-mullite and mullites with different Si/Al ratios is presented including first crystal chemical interpretations. The incorporation of boron reduces the mechanical stability of mullite due to the higher number of oxygen vacancies and complete amorphization of the investigated B-mullite is observed at 23 GPa, in comparison to 7:4 mullite at 25 GPa. Due to the rigidity of the polyhedral chains parallel to the c-axis, this direction

95 9 | Conclusion and future perspectives is the least compressible one. In contrast to earlier investigations on sillimanite up to 8 GPa (Yang et al., 1997; Burt et al., 2006), the compressibility in a-direction is higher than in b-direction for sillimanite above 10 GPa and for different mullites (Friedrich et al., 2004; Kalita et al., 2013; this work). The increasing inclination angle Ȧ can explain the increasing divergence between the compressibility in a- and b-direction with pressure but does not justify the stronger compressibility in a- than in b-direction. Here, a detailed analysis of the angles between the polyhedral planes might yield valuable information. A first hypothesis on the expected relative changes in the angles upon compression in a- and b-direction is given and represents a good basis for future investigations.

Objective 2: Crystal structure of B-mullite, substitution mechanism, chemical composition, incorporation limit [chapters 5 to 7] One major outcome of this thesis is the crystal structure of B-mullite in space group Pbam. During the 1200°C synthesis at ambient pressure 13 % of the tetrahedral silicon was replaced by boron, yielding a refined composition of Al4.64Si1.16B0.2O9.58 which is in good agreement with the PGA analysis. In the mullite crystal structure boron replaces silicon according to the coupled substitution as 4+ 2- 3+ 2 Si + O o 2 B + Ƒ. Boron resides in planar BO3 groups crosslinking the mullite-type AlO4 octahedral chains perpendicular to the c-axis. The rigidity of the BO3 group imposes local distortion of the AlO6 octahedra and consequently leads to split positions of the oxygen atoms in the first coordination sphere of boron. The split positions lead to significantly shortened O-O distances in c-direction compared to the a- and b-direction. Therefore, the crystallographic model provides an explanation for the strong reduction of lattice parameter c upon boron incorporation compared to the relatively small changes in lattice parameters a and b. The lattice parameter c is an indicative parameter representing the amount of boron in the B- mullite structure. PGA analyses and Rietveld refinements of B-mullites yield a linear relationship between lattice parameter c and the boron-content. In pure B-mullite up to about 15 % of the silicon can be replaced by boron during sol-gels synthesis at ambient pressure and 1200°C, corresponding to

3.2 mol % B2O3. B-mullites with as much as 11 mol % B2O3 were also observed, but contain alumina impurities. Using high-pressure and high-temperature syntheses, pure B-mullite with significantly higher boron-contents can be produced (c.f. objective 3).

Objective 3: Synthesis and crystal structure of Al8Si2B2O19 [chapter 6] Another important result of this thesis is the crystal structure of a phase close to the composition

Al8Si2B2O19 in space group Pbam. Whether this phase is orthorhombic like mullite or monoclinic like boralsilite and Al4B2O9 has been a matter of debate (Fischer and Schneider, 2008) since its first description (Werding and Schreyer, 1992). Careful Rietveld refinements and application of the established silicon-boron substitution mechanism yielded a refined composition of Al4.5Si0.9B0.6O9.4, corresponding to a replacement of 40% of the SiO4 tetrahedra replaced by BO3 units. In good 11 agreement with the PGA analysis (13 mol % B2O3), the B MAS NMR indicates the additional

96 9 | Conclusion and future perspectives

replacement of some aluminum in the AlO4 tetrahedra by boron but BO3 being dominant. Hence this is the first time a B-mullite crystal structure in space group Pbam is described containing BO3 and BO4 groups. The presence of BO4 in B-mullite synthesized at elevated pressure is in good agreement with the pressure/coordination rule suggesting that minerals with BO3 groups are stable preferably at relatively low pressures, while those with BO4 tetrahedra are stable at higher pressures (Werding and Schreyer, 1996).

9.2 Additional investigations related to this thesis

Parts of the experimental results within this thesis originate from collaboration with different partners. In three cases the partners have performed additional experiments on topics related to this thesis and partially based on samples of this work and/or additional samples provided. As a consequence of the difficulties arising from the PGA analyses of B-mullites with boron as a major element, a special technique was developed in order to increase the dynamic range for the analysis of boron in PGAA (Söllradl et al., 2013, abstract in Appendix B). In detail this method involves the attenuation of the signal using a 10 mm lead foil and was established at the PGAA spectrometer at the FRM II (Garching, Germany). As a co-product of the PGAA study the commercial

A9B2 compounds Alborex and Alborite (Shikoku Chemical Co., Marugame, Japan) were analyzed to yield a composition of Al18.0(4)B4.0(1)O33. This is in contrast to recent crystal chemical investigations on

A9B2 compounds with multiple methods (Fisch et al., 2011) that yielded a stoichiometry very close to

Al5BO9, assuming that the Al18B4O33 composition might be the result of inaccurate chemical analyses. First results on the high-pressure behavior of B-mullite compared to boron-free mullite are presented in this thesis, with a focus on the crystal chemical changes upon pressure (chapter 8). A more detailed study in collaboration with P.E. Kalita and K. Lipinska (UNLV), who performed the high-pressure Synchrotron X-ray diffraction experiments, is in preparation. This study will include details on the compressibility, bulk modulus, equation of state, and will also include the crystal chemical considerations. Furthermore a detailed comparison of the response to temperature and pressure of mullite and B-mullite based on the data of chapters 4 and 8 would yield valuable crystal chemical information. Especially in comparison with crystal chemical considerations on sillimanite at high-pressure and high-temperature this would lead to a better understanding of the physical properties of alumino-silicates with mullite-type structure. 11B MAS NMR experiments yielded valuable information on the local arrangement of boron in different B-mullites studied in this work (chapters 5 and 6). However, using one-dimensional MAS NMR, no significant variation was observed between samples with different boron-content and the 29Si and 27Al MAS NMR experiments yielded results similar to those for boron-free mullites. Additional MAS NMR experiments have been carried out by S.P. King and J.V. Hanna (University of Warwick) and will be published as a part of S.P. King’s PhD thesis (presumably in 2014). The analyses aim at the distribution of tricluster positions and oxygen vacancies as well as the connectivity

97 9 | Conclusion and future perspectives

and ordering of SiO4 and AlO4 tetrahedra. For this purpose a 2:1 single crystal was provided and special 29Si isotopically enriched 3:2 mullite and B-mullite samples were prepared. The normal one pulse 29Si MAS NMR spectra of the 29Si enriched samples exhibit significantly better quantitation of sites present and yield at least two additional 29Si resonances compared to the un-enriched samples and older NMR data of mullites. For the first time 29Si refocused INADEQUATE measurements were performed on mullite, yielding information on the SiO4 - SiO4 connections. Furthermore advances in the analytical technique enable different two dimensional experiments. In the 27Al 3QMAS spectrum the 2nd order quadrupolar broadened line is viewed in a second dimension free of second order broadening. Separation of overlapping resonances distinctly resolves interactions between the different aluminum environments. The preliminary results yield additional information on the local aluminum arrangement and distribution of tricluster positions. 27Al - 29Si correlations experiments (J-HQMC) were also performed and yield promising results with respect to the silicon-aluminum ordering. Direct investigations of the distribution of oxygen vacancies require analyses of 17O enriched samples and would be highly interesting. However, the special requirements regarding the syntheses conditions (high temperature, inert atmosphere) are quite challenging.

9.3 Future perspectives

This work was focused on the 1:1 isomorphous substitution of silicon by boron starting from a 3:2 mullite composition. Additional investigations are necessary for a better description of the dimensions of the B-mullite stability field. This should include experiments regarding the incorporation of higher amounts of boron into the sillimanite structure using high-pressure syntheses. Experiments on the incorporation of boron into 2:1 mullite, complementary to the work of Griesser et al. (2008), might also yield valuable information. Difficulties are expected due to the relatively low syntheses temperatures yielding poorly crystallized samples and high amounts of impurities. As part of the boron disappears during the syntheses, additional experiments with excess boron and/or using noble metal tubes could lead to higher B-contents in the structure. The chemical composition of samples yielding impurities can be estimated from the established relationship between lattice parameter c and the B2O3 content (chapter 7). Exact chemical analyses of single phase B-mullite samples are now routinely possible using PGAA. With respect to the application potential of B- mullites ceramics, more information regarding the sinter behavior, mechanical properties, thermal properties, as well as optical and electrical properties of B-mullite ceramics is necessary, although some information on the sintering behavior, microstructure, optical-, and dielectric properties of boron-doped mullite ceramics is already available (Zhang et al., 2010).

10 References

Agrell, S.O. and Smith, J.V. (1960) Cell dimensions, solid solution, polymorphism, and identification of mullite and sillimanite. Journal of the American Ceramic Society, 43(2), 69-76. Åhman, J., Svensson, G., and Grins, J. (1997) Lithium aluminium borate, LiAl7B4O17. Acta Chemica Scandinavica, 51(11), 1045-1050. Amoureux, J.P., Fernandez, C., and Steuernagel, S. (1996) Z filtering in MQMAS NMR. Journal of Magnetic Resonance Series A, 123(1), 116-118. Anderson, O.L., Isaak, D.G., and Yamamoto, S. (1989) Anharmonicity and the Equation of State for gold. Journal of Applied Physics, 65(4), 1534-1543. Angel, R.J. and Prewitt, C.T. (1986) Crystal structure of mullite: A re-examination of the average structure. American Mineralogist, 71, 1476-1482. Angel, R.J. and Prewitt, C.T. (1987) The incommensurate structure of mullite by Patterson synthesis. Acta Crystallographica Section B-Structural Science, 43, 116-126. Anovitz, L.M. and Grew, E.S. (1996) Mineralogy, petrology and geochemistry of boron: An introduction. In Grew, E.S., and Anovitz, L.M., Eds. "Boron. Mineralogy, petrology and geochemistry ". Reviews in Mineralogy 33, p. 1-40. Mineralogical Society of America, Washington, D.C. Anthony, J.W., Bideaux, R.A., Bladh, K.W., and Nichols, M.C. (2003) Handbook of Mineralogy. Mineralogical Society of America, Chantilly, VA 20151-1110, USA. http://www.handbookofmineralogy.org/ access date: July 2013. Baerlocher, C., Hepp, A., and Meier, W.M. (1978) DLS-76 - A program for the simulation of crystal structures by geometric refinement. ETH Zurich, Zurich. Balzar, D. and Ledbetter, H. (1993) Crystal-structure and compressibility of 3:2 Mullite. American Mineralogist, 78(11-12), 1192-1196. Basu, S. and Sarin, V. (2005) Mullite Coatings. In Schneider, H., and Komarneni, S., Eds. "Mullite". p. 349-376. WILEY-VCH, Weinheim. Baur, W.H. (1974) The geometry of polyhedral distortions. Predictive relationships for the phosphate group. Acta Crystallographica Section B-Structural Science, 30, 1195-1215. Baur, W.H. and Fischer, R.X. (1986) Recognition and treatment of background problems in neutron powder diffraction refinements. In Barrett, C.S., Cohen, J.B., Faber, J., Jenkins, R., Leyden, D.E., Russ, J.C., and Predecki, P.K., Eds. "Advances in X-ray Analysis". 29, p. 131–142. Plenum Publishing Corporation, New York. Birkenstock, J., Fischer, R.X., and Messner, T. (2012) BRASS - The Bremen Rietveld Analysis and Structure Suite. Fachbereich Geowissenschaften, Universität Bremen. Bowen, N.L. and Greig, J.W. (1924) The system: Al2O3-SiO2. Journal of the American Ceramic Society, 7(4), 238-254. Bowen, N.L., Greig, J.W., and Zies, E.G. (1924) Mullite, a new silicate of alumina. Journal of the Washington Academy of Sciences, 14(9), 183-191. Brace, W.F., Scholz, C.H., and Lamori, P.N. (1969) Isothermal compressibility of kyanite andalusite and sillimanite from synthetic aggregates. Journal of Geophysical Research, 74(8), 2089-2098. Bruker, A. (2009) Diffracplus Topas 4.2. Bruker AXS, Karlsruhe. Brunauer, G., Frey, F., Boysen, H., and Schneider, H. (2001) High temperature thermal expansion of mullite: an in situ neutron diffraction study up to 1600 degrees C. Journal of the European Ceramic Society, 21(14), 2563-2567. Buckingham, D.A., Lyday, P.A., Kostick, D.S., Polyak, D.E., Angulo, M.A., and R.D. Crangle, J. (2012) Mineral commodity boron statistics. In Kelly, T.D., and Matos, G.R., Ed. Historical statistics for mineral and material commodities in the United States: U.S. Geological Survey Data Series 140. U.S. Geological Survey.

99 10 |References Buick, I., Grew, E.S., Yates, M.G., Medenbach, O., Bebout, G.E., and Clarke, G.L. (2007) A natural analog of "boron-mullite" in granulite-facies metapelites from Mount Stafford, central Australia. Geochimica et Cosmochimica Acta, 71, A130. Buick, I.S., Grew, E.S., Armbruster, T., Medenbach, O., Yates, M.G., Bebout, G.E., and Clarke, G.L. (2008) Boromullite, Al9BSi2O19, a new mineral from granulite-facies metapelites, Mount Stafford, central Australia: a natural analogue of a synthetic "boron-mullite". European Journal of Mineralogy, 20(5), 935-950. Burt, J.B., Ross, N.L., Angel, R.J., and Koch, M. (2006) Equations of state and structures of andalusite to 9.8 GPa and sillimanite to 8.5 GPa. American Mineralogist, 91(2-3), 319-326. Caballero, A. and Ocana, M. (2002) Synthesis and structural characterization by X-ray absorption spectroscopy of tin-doped mullite solid solutions. Journal of the American Ceramic Society, 85(8), 1910-1914. Cameron, W.E. (1976) Coexisting sillimanite and mullite. Geological Magazine, 113(6), 497-513. Canella, L., Kudejova, P., Schulze, R., Turler, A., and Jolie, J. (2011) Characterisation and optimisation of the new Prompt Gamma-ray Activation Analysis (PGAA) facility at FRM II. Nuclear Instruments & Methods in Physics Research Section a-Accelerators Spectrometers Detectors and Associated Equipment, 636(1), 108-113. Capponi, J.J., Chenavas, J., and Joubert, J.C. (1972) New aluminum and gallium borates obtained by hydrothermal high-pressure synthesis. Bulletin de la Société Française Minéralogie et de Cristallographie, 95(3), 412-417 (in French). Chakraborty, A.k. and Das, S. (2003) Al-Si spinel phase formation in diphasic mullite gels. Ceramics International, 29, 27-33. Chakraborty, A.K. (2008) Reaction study of various mixtures of tetra ethyl ortho silicate and aluminum nitrate. Journal of Material Science, 43, 5376-5384. Cheary, R.W. and Coelho, A.A. (1996) XFIT. deposited in CCP14 Powder Diffraction Library, Engineering and Physical Sciences Research Council, Daresbury Laboratory, Warrington, England. Dewaele, A. and Loubeyre, P. (2007) Pressurizing conditions in helium-pressure-transmitting medium. High Pressure Research, 27(4), 419-429. Dietzel, A. and Scholze, H. (1955) Untersuchungen im System B2O3-Al2O3-SiO2. Glastechnische Berichte, 28, 47-51 (in German). Fazekas, B., Belgya, T., Dabolczi, L., Molnár, G., and Simonits, A. (1996) Hypermet-Pc: Program for automatic analysis of complex gamma-ray spectra. Journal of Trace and Microprobe Techniques, 14, 167-172. Filatov, S.K. and Bubnova, R.S. (2000) Borate crystal chemistry. Physics and Chemistry of Glasses, 41(5), 216-224. Filatov, S.K. and Bubnova, R.S. (2008) Structural mineralogy of borates as perspective materials for technological applications. In Krivovichev, S.V., Ed. "Minerals as Advances Materials I". Springer-Verlag, Berlin Heidelberg. Fisch, M., Armbruster, T., Rentsch, D., Libowitzky, E., and Pettke, T. (2011) Crystal-chemistry of mullite-type aluminoborates Al18B4O33 and Al5BO9: A stoichiometry puzzle. Journal of Solid State Chemistry, 184(1), 70-80. Fisch, M. (2011) Crystal chemistry of boron-bearing mulilte-type compounds. Doctoral thesis, Dr. rer. nat., Philosophisch-naturwissenschaftliche Fakultät, Universität Bern, CH. Fisch, M. and Armbruster, T. (2012) Thermal expansion of aluminoborates. In Krivovichev, S.V., Ed. "Minerals as Advanced Materials II". p. 255-268. Springer-Verlag, Berlin, Heidelberg. Fischer, R.X., Schneider, H., and Voll, D. (1996) Formation of aluminum rich 9:1 mullite and its transformation to low alumina mullite upon heating. Journal of the European Ceramic Society, 16(2), 109-113. Fischer, R.X. and Schneider, H. (2000) Crystal structure of Cr-mullite. American Mineralogist, 85(9), 1175-1179. Fischer, R.X. and Schneider, H. (2005) The mullite-type family of crystal structures. In Schneider, H., and Komarneni, S., Eds. "Mullite". p. 1-46. WILEY-VCH, Weinheim. Fischer, R.X. and Schneider, H. (2008) Crystal chemistry of borates and borosilicates with mullite- type structures: a review. European Journal of Mineralogy, 20(5), 917-933.

100 10 |References Fischer, R.X., Kahlenberg, V., Voll, D., MacKenzie, K.J.D., Smith, M.E., Schnetger, B., Brumsack, H.J., and Schneider, H. (2008) Crystal structure of synthetic Al4B2O9: A member of the mullite family closely related to boralsilite. American Mineralogist, 93(5-6), 918-927. Fischer, R.X., Gaede-Kohler, A., Birkenstock, J., and Schneider, H. (2012) Mullite and mullite-type crystal structures. International Journal of Materials Research, 103(4), 402-407. Fischer, R.X. and Messner, T. (2013) STRUPLO, a new version of the structure drawing program, Fachbereich Geowissenschaften Universität Bremen. Foshag, W.F. (1924) Mullite. The American Mineralogist, 9(10), 211-212. Friedrich, A., Kunz, M., Winkler, B., and Bihan, T.L. (2004) High-pressure behavior of sillimanite and kyanite: Compressibility, decomposition and indications of a new high-pressure phase. Zeitschrift für Kristallographie - Crystalline Materials, 219(6), 324-329. Garsche, M., Tillmanns, E., Almen, H., Schneider, H., and Kupcik, V. (1991) Incorporation of chromium into aluminum borate 9Al2O3·2B2O3 (A9B2). European Journal of Mineralogy, 3(5), 793-808. Gatta, G.D., Rotiroti, N., Fisch, M., and Armbruster, T. (2010) Stability at high pressure, elastic behavior and pressure-induced structural evolution of Al5BO9, a mullite-type ceramic material. Physics and Chemistry of Minerals, 37(4), 227-236. Gatta, G.D., McIntyre, G.J., Bromiley, G., Guastoni, A., and Nestola, F. (2012) A single-crystal neutron diffraction study of hambergite, Be2BO3(OH,F). American Mineralogist, 97(11-12), 1891-1897. Gelsdorf, G., Müller-Hesse, H., and Schwiete, H.E. (1958) Einlagerungsversuche an synthetischem Mullit und Substitutionsversuche mit Alliumoxyd und Germaniumoxyd. Teil II. Archiv für das Eisenhüttenwesen, 29, 513-519 (in German). Grew, E.S. (1996) Borosilicates (exclusive of tourmaline) and boron in rock-forming minerals in metamorphic environments. In Grew, E.S., and Anovitz, L.M., Eds. "Boron. Mineralogy, petrology and geochemistry". Reviews in Mineralogy 33, p. 387-502. Mineralogical Society of America, Washington, D.C. Grew, E.S., McGee, J.J., Yates, M.G., Peacor, D.R., Rouse, R.C., Huijsmans, J.P.P., Shearer, C.K., Wiedenbeck, M., Thost, D.E., and Su, S.C. (1998) Boralsilite (Al16B6Si2O37): A new mineral related to sillimanite from pegmatites in granulite-facies rocks. American Mineralogist, 83(5- 6), 638-651. Grew, E.S., Graetsch, H.A., Poter, B., Yates, M.G., Buick, I., Bernhardt, H.J., Schreyer, W., Werding, G., Carson, C.J., and Clarke, G.L. (2008) Boralsilite, Al16B6Si2O37, and "boron-mullite": Compositional variations and associated phases in experiment and nature. American Mineralogist, 93(2-3), 283-299. Griesser, K.J., Beran, A., Voll, D., and Schneider, H. (2008) Boron incorporation into mullite. Mineralogy and Petrology, 92(3-4), 309-320. Guse, W. and Mateika, D. (1974) Growth of mullite single crystals (2Al2O3 · SiO2) by the Czochralski method. Journal of Crystal Growth, 22(3), 237-240. Hammersley, A. (2005) computer program FIT2D V12.077. ESRF, Grenoble, France. Hart, H.V. and Drickamer, H.G. (1965) Effect of high pressure on the lattice parameters of Al2O3. The Journal of Chemical Physics, 43(7), 2265-2266. Hawthorne, F.C., Burns, P.C., and Grice, J.D. (1996) The crystal chemistry of boron. In Grew, E.S., and Anovitz, L.M., Eds. "Boron. Mineralogy, petrology and geochemistry". Reviews in Mineralogy 33, p. 41-115. Mineralogical Society of America, Washington, D.C. Hayashi, S. and Hayamizu, K. (1989) Shift references in high-resolution solid-state NMR. Bulletin of the Chemical Society of Japan, 62(7), 2429-2430. Hoelzel, M., Senyshyn, A., Juenke, N., Boysen, H., Schmahl, W., and Fuess, H. (2012) High- resolution neutron powder diffractometer SPODI at research reactor FRM II. Nuclear Instruments & Methods in Physics Research Section A: Accelerators Spectrometers Detectors and Associated Equipment, 667, 32-37. Hoffman, D.W., Roy, R., and Komarneni, S. (1984) Diphasic xerogels, a new class of materials - phases in the system Al2O3-SiO2. Journal of the American Ceramic Society, 67(7), 468-471. Hong, S.H., Cermignani, W., and Messing, G.L. (1996) Anisotropic grain growth in seeded and B2O3- doped diphasic mullite gels. Journal of the European Ceramic Society, 16(2), 133-141.

101 10 |References Hsiung, C.H.H., Pyzik, A.J., Gulsoy, E.B., De Carlo, F., Xiao, X., and Faber, K.T. (2013a) Impact of doping on the mechanical properties of acicular mullite. Journal of the European Ceramic Society, 33(10), 1955-1965. Hsiung, C.H.H., Pyzik, A.J., De Carlo, F., Xiao, X., Stock, S.R., and Faber, K.T. (2013b) Microstructure and mechanical properties of acicular mullite. Journal of the European Ceramic Society, 33(3), 503-513. Kaduk, J.A., Satek, L.C., and McKenna, S.T. (1999) Crystal structures of metal aluminum borates. Rigaku Journal, 16, 17-30. Kalita, P.E., Schneider, H., Lipinska, K., Sinogeikin, S., Hemmers, O.A., and Cornelius, A. (2013) High-pressure behavior of mullite: An X-ray diffraction investigation. Journal of the American Ceramic Society, 96(5), 1635-1642. Kerr, R. and Wood, N. (2004) Joseph Needham: Science and civilisation in China vol. 5: Chemistry and chemical technology, part 12: Ceramic technology. Cambridge University Press, Cambridge, UK. Klug, F.J., Prochazka, S., and Doremus, R.H. (1987) Alumina-Silica Phase-Diagram in the Mullite Region. Journal of the American Ceramic Society, 70(10), 750-759. Kogel, J.E., Trivedi, N.C., Barker, J.M., and Krukowski, S.T. (2006) Industrial minerals & rocks: Commodities, markets, and uses. Society for Mining, Metallurgy, and Exploration, Littleton, Colorado, USA. Komarneni, S., Schneider, H., and Okada, K. (2005) Mullite Synthesis and Processing. In Schneider, H., and Komarneni, S., Eds. "Mullite". WILEY-VCH, Weinheim. 2+ Kutty, T.R.N. and Nayak, M. (2000) Photoluminescence of Eu -doped mullite (xAl2O3·ySiO2; x/y=3/2 and 2/1) prepared by a hydrothermal method. Materials Chemistry and Physics, 65(2), 158-165. Laugier, J. and Bochu, B. (2004) LMGP-Suite: Suite of programs for the interpretation of X-ray experiments. ENSP/Laboratoire des Matériaux et du Génie Physique, BP 46. 38042 Saint Martin d'Hères, France. Letort, Y. (1952) Contribution a l'étude de la synthèse de la mullite. Transactions of the International Ceramic Congress, 19-32 (in French). Lorenz, W. and Gwosdz, W. (2003) Bewertungskriterien für Industrieminerale, Steine und Erden. Teil 7: Feldspäte und andere Flussmittel. Bundesanstalt für Geowissenschaften und Rohstoffe und den Staatlichen Geologischen Diensten in der Bundesrepublik Deutschland, Hannover (in German). Lührs, H., Fischer, R.X., and Schneider, H. (2012) Boron mullite: Formation and basic characterization. Materials Research Bulletin, 47(12), 4031-4042. Lührs, H., Senyshyn, A., King, S.P., Hanna, J.V., Schneider, H., and Fischer, R.X. (2013a) Neutron diffraction and 11B solid state NMR studies of the crystal structure of B-doped mullite. Zeitschrift für Kristallographie - Crystalline Materials, 228(9), 457-466. Lührs, H., Söllradl, S., King, P., Hanna, J.V., Konzett, J., Fischer, R.X., and Schneider, H. (2013b) Ambient and high-pressure synthesis, composition, and crystal structure of B-mullites. Crystal Research and Technology, in press, DOI: 10.1002/crat201300210. MacKenzie, K.J.D. and Smith, M.E. (2002) Multinuclear solid state NMR of inorganic materials. Pergamon Press, Oxford. MacKenzie, K.J.D., Smith, M.E., Kemp, T.F., and Voll, D. (2007) Crystalline aluminium borates with the mullite structure: A 11B and 27Al solid-state NMR study. Applied Magnetic Resonance, 32(4), 647-662. Mao, H.K., Bell, P.M., Shaner, J.W., and Steinberg, D.J. (1978) Specific volume measurements of Cu, Mo, Pd, and Ag and calibration of ruby R1 fluorescence pressure gauge from 0.06 to 1 Mbar. Journal of Applied Physics, 49(6), 3276-3283. Massiot, D., Fayon, F., Capron, M., King, I., Le Calve, S., Alonso, B., Durand, J.O., Bujoli, B., Gan, Z.H., and Hoatson, G. (2002) Modelling one- and two-dimensional solid-state NMR spectra. Magnetic Resonance in Chemistry, 40(1), 70-76. Matthes, S. (2001) MINERALOGIE - Eine Einführung in die spezielle Mineralogie, Petrologie und Lagerstättenkunde. Springer-Verlag, Berlin (in German). Mazza, D., Vallino, M., and Busca, G. (1992) Mullite-type structures in the systems Al2O3-Me2O (Me = Na,K) and Al2O3-B2O3. Journal of the American Ceramic Society, 75(7), 1929-1934.

102 10 |References Mazza, D., Ronchetti, S., Delmastro, A., Tribaudino, M., and Kockelmann, W. (2001) Silica-free mullite structures in the Al2O3-B2O3-P2O5 ternary system. Chemistry of Materials, 13(1), 103- 108. Merwin, L.H., Sebald, A., Rager, H., and Schneider, H. (1991) 29Si and 27Al MAS NMR spectroscopy of mullite. Physics and Chemistry of Minerals, 18(1), 47-52. Messner, T. (2013) Computer program multisurface. Fachbereich Geowissenschaften, Universität Bremen. Molnár, G.L. (2004) Handbook of Prompt Gamma Activation Analysis with Neutron Beams. Kluwer Academic Publishers, Dordrecht, The Netherlands. NETZSCH. (2010) NETZSCH Messung. NETZSCH-Gerätebau GmbH, Selb, Germany. Niven, M.L., Waters, D.J., and Moore, J.M. (1991) The crystal-structure of werdingite, (Mg,Fe)2Al12(Al,Fe)2Si4(B,Al)4O37, and its relationship to sillimanite, mullite, and grandidierite. American Mineralogist, 76(1-2), 246-256. Okada, K. and Schneider, H. (2005) Application of mullite ceramics. In Schneider, H., and Komarneni, S., Eds. "Mullite". p. 327-348. WILEY-VCH, Weinheim. Park, H. and Barbier, J. (2001) PbGaBO4, an orthoborate with a new structure-type. Acta Crystallographica Section E-Structure Reports Online, 57, i82-i84. Park, H., Lam, R., Greedan, J.E., and Barbier, J. (2003a) Synthesis, crystal structure, crystal chemistry, and magnetic properties of PbMBO4 (M = Cr, Mn, Fe): A new structure type exhibiting one-dimensional magnetism. Chemistry of Materials, 15(8), 1703-1712. Park, H., Barbier, J., and Hammond, R.P. (2003b) Crystal structure and polymorphism of PbAlBO4. Solid State Sciences, 5(4), 565-571. Pask, J.A. (1990) Critical review of phase-equilibria in the Al2O3-SiO2 system. Mullite and Mullite Matrix Composites, 6, 1-13. Peacor, D.R., Rouse, R.C., and Grew, E.S. (1999) Crystal structure of boralsilite and its relation to a family of boroaluminosilicates, sillimanite, and andalusite. American Mineralogist, 84(7-8), 1152-1161. Pohl, W.L. (2005) Mineralische und Energierohstoffe - Eine Einführung zur Entstehung und nachhaltigen Nutzung von Lagerstätten. E. Schweizerbart'sche Verlagsbuchhandlung (Nägele u. Obermiller), Stuttgart (in German). Pyzik, A., Ziebarth, R., Han, C., and Yang, K. (2011) High-porosity acicular mullite ceramics for multifunctional diesel particulate filters. International Journal of Applied Ceramic Technology, 8(5), 1059-1066. Rahman, S. and Freimann, S. (2005) The real structure of mullite. In Schneider, H., and Komarneni, S., Eds. "Mullite". p. 46-70. WILEY-VCH, Weinheim. Rehak, P., Kunath-Fandrei, G., Losso, P., Hildmann, B., Schneider, H., and Jäger, C. (1998) Study of the Al coordination in mullites with varying Al:Si ratio by 27Al NMR spectroscopy and X-ray diffraction. American Mineralogist, 83, 1266-1276. Revay, Z. (2009) Determining elemental composition using prompt gamma activation analysis. Analytical Chemistry, 81(16), 6851-6859. Reynaud, C. (1977) Synthesis, thermal-analysis and micrography of new aluminum borates. Bulletin de la Société Française Minéralogie et de Cristallographie, 100(1), 28-32 (in French). Richards, E.A., Goodbrake, C.J., and Sowman, H.G. (1991) Reactions and Microstructure Development in Mullite Fibers. Journal of the American Ceramic Society, 74(10), 2404-2409. Rivers, M., Prakapenka, V.B., Kubo, A., Pullins, C., Holl, C.M., and Jacobsen, S.D. (2008) The COMPRES/GSECARS gas-loading system for diamond anvil cells at the Advanced Photon Source. High Pressure Research, 28(3), 273-292. Robertson, J.D. and Dyar, M.D. (1996) Nuclear methods for analysis of boron in minerals. In Grew, E.S., and Anovitz, L.M., Eds. "Boron. Mineralogy, petrology and geochemistry ". Reviews in Mineralogy 33, p. 805-820. Mineralogical Society of America, Washington, D.C. Saalfeld, H. and Guse, W. (1981) Structure refinement of 3:2-mullite (3Al2O3 * 2SiO2). Neues Jahrbuch für Mineralogie-Monatshefte, 1981(4), 145-150. Schmücker, M., Schneider, H., MacKenzie, K.J.D., Smith, M.E., and Carroll, D.L. (2005a) AlO4/SiO4 distribution in tetrahedral double chains of mullite. Journal of the American Ceramic Society, 88(10), 2935-2937. Schmücker, M., Schneider, H., Mauer, T., and Clauȕ, B. (2005b) Temperature-dependent evolution of grain growth in mullite fibres. Journal of the European Ceramic Society, 25(14), 3249-3256.

103 10 |References Schneider, H. and Werner, H.D. (1982) Synthesis and crystal chemistry of andalusite- and mullite-type structures in the system Al2GeO5–Ga2GeO5. Neues Jahrbuch für Mineralogie-Abhandlungen, 143, 223-230. Schneider, H. (1984) Solid solubility of Na2O in mullite. Journal of the American Ceramic Society, 67(7), C130-C131. Schneider, H. (1985) Magnesium incorporation in mullite. Neues Jahrbuch für Mineralogie- Monatshefte, 1985(11), 491-496. Schneider, H. (1986a) Formation, properties and high-temperature behaviour of mullite. Habilitation Thesis, Universität, Münster, Germany. Schneider, H. (1986b) Zirconium incorporation in mullite. Neues Jahrbuch für Mineralogie- Monatshefte, 1986(4), 172–180. Schneider, H. and Eberhard, E. (1990) Thermal-expansion of mullite. Journal of the American Ceramic Society, 73(7), 2073-2076. Schneider, H. (1990) Transition-metal distribution in mullite. Mullite and Mullite Matrix Composites, 6, 135-157. Schneider, H. and Komarneni, S. (2005) Mullite. WILEY-VCH, Weinheim. Schneider, H. (2005a) Gerneral introduction. In Schneider, H., and Komarneni, S., Eds. "Mullite". p. XV-XIX. WILEY-VCH, Weinheim. Schneider, H. (2005b) Foreign cation incorporation in mullite. In Schneider, H., and Komarneni, S., Eds. "Mullite". p. 70-93. WILEY-VCH, Weinheim. Schneider, H. (2005c) Mullite fibers. In Schneider, H., and Komarneni, S., Eds. "Mullite". p. 377-397. WILEY-VCH, Weinheim. Schneider, H. (2005d) Mullite matrix composites. In Schneider, H., and Komarneni, S., Eds. "Mullite". p. 397-476. WILEY-VCH, Weinheim. Schneider, H., Schreuer, J., and Hildmann, B. (2008) Structure and properties of mullite - a review. Journal of the European Ceramic Society, 28(2), 329-344. Schnetger, B., Brumsack, H.-J., Schale, H., Hinrichs, J., and Dittert, L. (2000) Geochemical characteristics of deep-sea sediments from the Arabian Sea: A high-resolution study. Deep- Sea Research II, 47, 2735-2768. Scholze, H. (1956) Über Aluminiumborate. Zeitschrift für anorganische und allgemeine Chemie, 284, 272-277 (in German). Shannon, R.D. (1976) Revised effective ionic radii and systematic studies of interatomic distances in halides and chalcogenides. Acta Crystallographica A, 32, 751. Shirley, R. (2002) The Crysfire 2002 system for automatic powder indexing. Guildford, England. Sokolova, E.V., Azizov, A.V., Simonov, M.A., Leoniuk, N.I., and Belov, N.V. (1978) Crystal- structure of synthetic ortho-3-borate Al5(BO3)O6 Doklady Akademii Nauk SSSR 243(3) 655– 658 (in Russian). Söllradl, S., Lührs, H., Revay, Z., Kudejova, P., Canella, L., and Türler, A. (2013) Broadening the dynamic range of PGAA. Journal of Radioanalytical and Nuclear Chemistry, in press, DOI: 10.1007/s10967-013-2739-9. Stephenson, D.A. and Moore, P.B. (1968) Crystal structure of grandidierite (Mg,Fe)Al3SiBO9. Acta Crystallographica Section B-Structural Crystallography and Crystal Chemistry, B 24, 1518- 1522. Takemura, K. (2001) Evaluation of the hydrostaticity of a helium-pressure medium with powder x-ray diffraction techniques. Journal of Applied Physics, 89(1), 662-668. Tomsia, A.P., Saiz, E., Ishibashi, H., Diaz, M., Requena, J., and Moya, J.S. (1998) Powder processing of mullite/Mo functionally graded materials. Journal of the European Ceramic Society, 18(9), 1365-1371. Tröger, W.E. (1982) Optische Bestimmung der gesteinsbildenden Minerale. E. Schweizerbart'sche Verlagsbuchhandlung, Stuttgart (in German). Visser, J.W. (1969) A fully automatic program for finding the unit cell from powder data. Journal of Applied Crystallography, 2, 89-95. Voll, D., Angerer, P., Beran, A., and Schneider, H. (2002) A new assignment of IR vibrational modes in mullite. Vibrational Spectroscopy, 30(2), 237-243. Werding, G. and Schreyer, W. (1984) Alkali-free tourmaline in the system MgO-Al2O3-B2O3-SiO2- H2O. Geochimica et Cosmochimica Acta, 48(6), 1331-1344.

104 10 |References Werding, G. and Schreyer, W. (1992) Synthesis and stability of werdingite, a new phase in the system MgO- Al2O3 - B2O3 - SiO2 (MABS), and another new phase in the ABS-System. European Journal of Mineralogy, 4(1), 193-207. Werding, G. and Schreyer, W. (1996) Experimental studies on borosilicates and selected borates. In Grew, E.S., and Anovitz, L.M., Eds. "Boron. Mineralogy, petrology and geochemistry ". Reviews in Mineralogy 33, p. 117-163. Mineralogical Society of America, Washington, D.C. Yang, H., Hazen, R.M., Finger, L.W., Prewitt, C.T., and Downs, R.T. (1997) Compressibility and crystal structure of sillimanite, Al2SiO5, at high pressure. Physics and Chemistry of Minerals, 25(1), 39-47. Zhang, G.M., Fu, Z.Y., Wang, Y.C., Wang, H., Wang, W.M., Zhang, J.Y., Lee, S.W., and Niihara, K. (2010) Boron-doped mullite derived from single-phase gels. Journal of the European Ceramic Society, 30(12), 2435-2441.

105

Acknowledgements

I would like to thank Reinhard X. Fischer for the opportunity to carry out this work in his group and within this exciting field of science. I am grateful for his constant support, many helpful discussions, and last but not least his honest interest in my research. My thanks also go to my second supervisor, Hartmut Schneider, for his valuable ideas, many discussions and proof-reading of manuscripts. Thanks to Thorsten M. Gesing for helpful discussions and for agreeing to be the second reviewer of this thesis. All my colleagues in the crystallography group who supported me during the last years (not only regarding scientific or instrumental problems) deserve a big thank-you: Christoph Vogt, Gabriele Ebert, Iris Spiess, Johannes Birkenstock, Michael Wendschuh, Thomas Messner, my fellow PhD students Kristin Hoffmann and Li Wang as well as Malik Šehoviü for his support in the lab. I would also like to thank M. Mangir Murshed for his scientific advice and discussions. I would like to thank the following people for technical support at various instruments: P. Witte, L. Robben, J. Butzlaff (SEM, EDX, Universität Bremen), V. Kolb (microscope camera, Universität Bremen), B. Schnetger (XRF, Universität Oldenburg), J. Konzett (high pressure syntheses, Universität Innsbruck, Austria), A. Senyshyn (neutron diffraction, FRM II; Garching), J.V. Hanna and S. P. King (NMR, University of Warwick, UK), S. Söllradl (PGAA, FRM II, Garching, Germany), and P.E. Kalita and K. Lipinska (University Nevada, Las Vegas, USA) for providing the high-pressure synchrotron XRD data. I would like to thank the University of Bremen for financial support from the Central Research Development Fund for the whole project and additional travel support. Thanks to the DAAD for travel support. Furthermore I gratefully acknowledge the financial support provided by FRM II to perform the neutron scattering and PGAA measurements at the Forschungs-Neutronenquelle Heinz Maier- Leibnitz (FRM II), Garching, Germany. With all my heart I want to say thank you to my family, this piece of work would never have been possible without your backup and encouragement: My beloved husband and children who so often had to get along without me during the last weeks and months. My parents, my parents-in-law, and my sister, who spend so many hours of their time looking after the children and who were always around when we needed a helping hand. Thank you to all of my friends, who always believed in me and everyone I might have forgotten to mention.

107

Appendix

Appendix A contains a published manuscript with results from my master thesis that was compiled during my PhD studentship. Appendix B contains the abstract of a manuscript in press that is the outcome of further research activities related to this project. In Appendix C a list of all samples referred to in this thesis is given. An excel file containing this list can be found on the supplementary CD (supplementary_sample list.xlsx) together with the raw data of all measurements and *.cif files of the crystal structure presented in this thesis. The raw data of the in-situ high-pressure X-ray diffraction experiments are not on the CD as the copyright for this data is with the University of Nevada Las Vegas.

109 Appendix A Appendix A

K and Ca exchange behavior of zeolite A

Hanna Lührs,*, Jewgenij Derr, Reinhard X. Fischer

Fachbereich Geowissenschaften, Universität Bremen, Klagenfurter Straße 2, D-28359 Bremen, Germany

Published in: Microporous and Mesoporous Materials 151 (2012) 457-465

DOI: doi:10.1016/j.micromeso.2011.09.025

Submitted: 9 Jan. 2011 | revised: 20 Sept. 2011 | accepted: 26 Sept. 2011 | online: 1 Oct. 2011

Keywords: Zeolite A | LTA | Ion exchange | Lattice parameter | water content

Powdered samples of Zeolite-A ([Na11.8Al11.8Si12.2O48 • 26 H2O]8, LTA, Na-A) were completely Ca and K exchanged. Several compositions between the end members Na-, Ca-, and K-zeolite A were produced. Combined results from X-ray powder diffraction including lattice parameter refinements in space group F mC3 c, energy dispersive X-ray spectroscopy, and thermal gravimetric analyses are presented. Rietveld refinements are performed for the three Na-rich samples in the Ca exchange series. The linear dependence of the lattice parameter a on the calcium and potassium contents is described as well as the relationship between water, calcium and potassium content. The lattice parameter a decreases with increasing calcium and potassium content, ranging from 24.62 to 24.46 Å for the Ca- exchanged forms and from 24.63 to 24.58 Å for the K-exchanged forms. Higher calcium content results in a higher amount of water incorporated in the zeolite, ranging from 25 (Na-A) to 28 (Ca-A) molecules per formula unit (m.p.f.u.) whereas the water content in the K-exchanged forms varies about 23.5±1.0 m.p.f.u. not directly correlated with the K contents but significantly lower than the 26 m.p.f.u. of the Na-A starting material. The Ca and K contents can be estimated from lattice parameter a by linear regression.

* Corresponding author: E-mail address: [email protected] (H. Lührs) Telephone: +49-421-218 65181, Telefax: +49-421-281-6518

110 Appendix B Appendix B

Increasing the dynamic range for the analysis of boron in PGAA

Stefan Söllradl*1,2,3, Hanna Lührs4, Zsolt Révay3, Petra Kudejova3, Lea Canella3, Andreas Türler1,2

1 Paul Scherrer Institute, Laboratory for Radiochemistry and Environmental Chemistry, 5232 Villigen-PSI, Switzerland 2 University of Berne, Department of Chemistry & Biochemistry, Freiestrasse 3, 3012 Bern, Switzerland 3 Heinz Maier-Leibnitz Zentrum (MLZ), Technische Universität München, Lichtenbergstr. 1, 85748 Garching, Germany 4 Universität Bremen, Fachbereich Geowissenschaften, Klagenfurter Str., 28357 Bremen, Germany

In press: Journal of Radioanalytical and Nuclear Chemistry (Aug. 2013)

DOI: 10.1007/s10967-013-2739-9

Prompt gamma activation analysis (PGAA) is especially sensitive for elements with high neutron-capture cross sections, like boron, which can be detected down to a level of ng/g. However, if it is a major component, the high count rate from its signal will distort the spectra, making the evaluation difficult. A lead attenuator was introduced in front of the HPGe-detector to reduce low- energy gamma radiation and specifically the boron gamma rays reaching the detector, whose thickness was found to be optimal at 10 mm. Detection efficiencies with and without the lead attenuator were compared and it was shown that the dynamic range of the PGAA technique was significantly increased. The method was verified with the analyses of stoichiometric compounds: TiB2, NiB, PVC, Alborex, and Alborite.

* Corresponding author: E-mail address: [email protected]

111 Appendix C Appendix C

List of all samples referred to in this thesis. For each sample the sample ID and the respective name in the manuscripts/chapters is given as well as the initial gel composition, synthesis conditions and analytical methods. An excel file containing this list can be found on the supplementary CD (supplementary_sample list.xlsx) as well as the raw data of all measurements.

analytical methods manuscript number, synthesis chapter initial composition conditions powder diffraction

sample ID Al2O3 SiO2 B2O3 B NMR

raw data file 1, chapter 4 2, chapter 5 3, chapter 6 chapter 7 chapter 8 [mol%] [mol%] [mol%] h °C kbar (*.xrdml) X'Pert PW 1800(*.xrdml) (*.xrdml) HT-XRD capillaryD8 (*.raw) (*.xye) SPODI HP-XRD (*.chi) precursor STA product STA XRF FTIR PGAA 11 ABS3_B0000_21 x 56.56 43.44 0.00 5 1200 x ABS3_B0000_21_p1 x 56.56 43.44 0.00 0 0 x ABS3_B0000_21_p2 x 56.56 43.44 0.00 90 1400 x ABS3_B0000_21b x 56.56 43.44 0.00 0 1300 x ABS3_B0000_21c x 56.56 43.44 0.00 5 1400 x ABS3_B0000_22c x 56.35 43.65 0.00 5 1400 x x x ABS3_B0000_25 x NMR-I AP0 x 60.01 39.99 0.00 5 1200 x x x x ABS3_B0000_28 I AP4 60.00 40.00 0.00 5 1200 x ABS3_B0075_3 x 60.45 38.51 1.04 5 1200 x ABS3_B0075_3_p1 x 60.45 38.51 1.04 5 1200 x ABS3_B0075_3_p2 x 60.45 38.51 1.04 90 1400 x ABS3_B0075_3b x 60.45 38.51 1.04 5 1300 x ABS3_B0075_3c x 60.45 38.51 1.04 5 1400 x x ABS3_B0075_4 xNMR-IIAP1AP1 60.7638.231.01 51200 x x xxxxxx ABS3_B0075_4b_HTK x 60.76 38.23 1.01 5 1200 x ABS3_B0075_5 x 60.65 38.34 1.01 5 1200 x x ABS3_B0075_9 II AP5 60.58 38.41 1.01 5 1200 x ABS3_B0150_3 x 61.37 36.59 2.05 5 1200 x x ABS3_B0225_3 x 61.66 35.26 3.08 5 1200 x ABS3_B0225_3_p1 x 0.00 0.00 0.00 5 1400 x ABS3_B0225_3_p2 x 0.00 0.00 0.00 90 1400 x ABS3_B0225_3b x 61.66 35.26 3.08 5 1300 x ABS3_B0225_3c x 61.66 35.26 3.08 5 1400 x x ABS3_B0252_5 III AP6 62.16 34.73 3.11 5 1200 x ABS3_B0300_3 x 62.61 33.21 4.19 5 1200 x x ABS3_B0300_5 xNMR-IIIAP2AP2 61.8933.994.12 51200 x x xxxxxx ABS3_B0375_2 x 62.98 31.79 5.23 5 1200 x ABS3_B0375_2_p1 x 62.98 31.79 5.23 5 1400 x ABS3_B0375_2_p2 x 62.98 31.79 5.23 90 1400 x ABS3_B0375_2b x 62.98 31.79 5.23 5 1300 x ABS3_B0375_2c x 62.98 31.79 5.23 5 1400 x x ABS3_B0375_3d x 63.16 31.58 5.26 5 HTK x ABS3_B0375_4 IV AP7 63.26 31.47 5.27 5 1200 x ABS3_B0450_5 xNMR-IVAP3AP3 63.8629.766.38 51200 x x xxxxxx ABS3_B0450_5b_HTK x 63.86 29.76 6.38 5 1200 x ABS3_B0450_6 x 62.70 31.03 6.27 5 1200 x x x ABS3_B0450_7 V AP8 63.91 29.70 6.39 5 1200 x x ABS3_B0525_2 x 64.25 28.23 7.52 5 1200 x ABS3_B0525_2_p1 x 64.25 28.23 7.52 5 1400 x ABS3_B0525_2_p2 x 64.25 28.23 7.52 90 1400 x ABS3_B0525_2b x 64.25 28.23 7.52 5 1300 x ABS3_B0525_2c x 64.25 28.23 7.52 5 1400 x x ABS3_B0525_3 VI AP9 65.02 27.39 7.59 5 1200 x ABS3_B0675_2 x 65.70 24.45 9.86 5 1200 x ABS3_B0675_2_p1 x 65.70 24.45 9.86 5 1400 x ABS3_B0675_2_p2 x 65.70 24.45 9.86 90 1400 x ABS3_B0675_2b x 65.70 24.45 9.86 5 1300 x ABS3_B0675_2c x 65.70 24.45 9.86 5 1400 x x ABS3_B0825_2 x 67.21 20.46 12.33 5 1200 x ABS3_B0825_2_p1 x 67.21 20.46 12.33 5 1400 x ABS3_B0825_2_p2 x 67.21 20.46 12.33 90 1400 x ABS3_B0825_2b x 67.21 20.46 12.33 5 1300 x ABS3_B0825_2c x 67.21 20.46 12.33 5 1400 x x ABS3_B0825_3 x 67.09 20.61 12.30 5 1400 x x

112 Appendix C Appendix C - continued

analytical methods manuscript number, synthesis chapter initial composition conditions powder diffraction

sample ID Al2O3 SiO2 B2O3 B NMR

raw data file 1, chapter 4 2, chapter 5 3, chapter 6 chapter 7 chapter 8 [mol%] [mol%] [mol%] h °C kbar X'Pert (*.xrdml) (*.xrdml) 1800 PW HT-XRD (*.xrdml) (*.raw) capillary D8 (*.xye) SPODI (*.chi) HP-XRD precursor STA product STA XRF FTIR PGAA 11 ABS3_B0900_4 x 68.14 18.24 13.62 5 1200 x x ABS3_B0900_6 x 68.09 18.30 13.61 5 1000 x ABS3_B0975_2 x 69.11 15.94 14.94 5 1200 x ABS3_B0975_2_p1 x 69.11 15.94 14.94 5 1400 x ABS3_B0975_2_p2 x 69.11 15.94 14.94 90 1400 x ABS3_B0975_2b x 69.11 15.94 14.94 5 1300 x ABS3_B0975_2c x 69.11 15.94 14.94 5 1400 x x ABS3_B0975_3 x 69.15 15.88 14.97 5 1200 x x ABS3_B1125_2 x 70.35 12.08 17.57 5 1200 x ABS3_B1125_2_p1 x 70.35 12.08 17.57 5 1400 x ABS3_B1125_2_p2 x 70.35 12.08 17.57 90 1400 x ABS3_B1125_2b x 70.35 12.08 17.57 5 1300 x ABS3_B1125_2c x 70.35 12.08 17.57 5 1400 x x ABS3_B1125_3 x 70.37 12.05 17.59 5 1200 x x ABS3_B1275_3 x 72.04 7.54 20.42 5 1200 x ABS3_B1275_3_p1 x 72.04 7.54 20.42 5 1400 x ABS3_B1275_3_p2 x 72.04 7.54 20.42 90 1400 x ABS3_B1275_3b x 0.00 0.00 0.00 5 1300 x ABS3_B1425_2 x 73.87 2.75 23.37 5 1200 x ABS3_B1425_2_p1 x 73.87 2.75 23.37 5 1400 x ABS3_B1425_2_p2 x 73.87 2.75 23.37 90 1400 x ABS3_B1425_2b x 73.87 2.75 23.37 5 1300 x ABS3_B1425_2c x 73.87 2.75 23.37 5 1400 x x ABS3_B1500_3 x 75.00 0.00 25.00 5 1200 x ABS3_B1500_3_p1 x 75.00 0.00 25.00 5 1400 x ABS3_B1500_3_p2 x 75.00 0.00 25.00 90 1400 x ABS3_B1500_3b x 75.00 0.00 25.00 5 1300 x Al18B4O33_1 x 81.79 0.00 18.21 5 1200 x x x HP1 HP1 53.29 26.72 19.99 168 800 7 x HP2 HP2 HP2 53.29 26.72 19.99 260 875 10 x x x ABSG_B00_2 x 60.08 39.92 0.00 5 1200 x ABSG_B05_2 x 60.05 33.28 6.67 5 1200 x ABSG_B10_2 x 60.02 26.64 13.34 5 1200 x ABSG_B15_2 x 60.07 19.90 20.03 5 1200 x ABSG_B20_2 x 60.03 13.28 26.68 5 1200 x ABSG_B25_2 x 60.04 6.60 33.35 5 1200 x ABSG_B30_2 x 60.00 0.00 40.00 5 1200 x

113

Supplementary CD Supplementary CD Supplementary CD containing:

- Digital version of the thesis - A list of all samples referred to in this thesis: supplementary_sample list.xlsx (Appendix C) - *.cif files of the two B-mullite crystal structures presented in this thesis o B-mullite_Al4.64 B0.2 Si1.16 O9.58.cif (chapter 5, CSD number: 426638) o B-mullite_Al4.5 B0.6 Si0.9 O9.4.cif (chapter 6) - Diffraction data o D8_capillary (5 files *.raw) o HT-PXRD (High-temperature XRD of 5 samples, total: 201 files *.xrdml) o PW1800 (33 files *.xrdml) o SPODI (6 files *.xye, wavelength.txt) o X’Pert (57 files *.xrdml) - FTIR (4 files *.csv) - NMR (3 ASCII files) - PGAA (4 *.pdf files) - STA (2 *.xls files, 1 *.txt file)

115

Erklärung

Name: Hanna Lührs Adresse: Fachbereich Geowissenschaften, Universität Bremen

Hiermit versichere ich, dass ich x die Arbeit ohne unerlaubte fremde Hilfe angefertigt habe, x keine anderen als die von mir angegebenen Quellen und Hilfsmittel benutzt habe und x die den benutzen Werken wörtlich oder inhaltlich entnommenen Stellen als solche kenntlich gemacht habe.

Bremen, 06.09.2013

(Unterschrift)

117