LiF ceramics by hot-pressing: Host–additive interaction mechanisms issue revisited

Journal of the European Ceramic Society 36 (2016) 1731–1742

Contents lists available at www.sciencedirect.com

Journal of the European Ceramic Society journal homepage: www.elsevier.com/locate/jeurceramsoc

Transparent MgAl2 O4 /LiF ceramics by hot-pressing: Host–additive interaction mechanisms issue revisited Adrian Goldstein a,∗ , Jan Raethel b , Michael Katz a , Mila Berlin a , Ehud Galun c a b c

Israel Ceramic and Silicate Institute, Technion City, Haifa 3200003, Israel Fraunhofer-IKTS, Winterbergstrasse 28, 01277 Dresden, Germany Elbit Systems, El-Op, Adv. Tech. Park, Rehovot 7611101, Israel

a r t i c l e

i n f o

Article history: Received 27 September 2015 Received in revised form 1 February 2016 Accepted 1 February 2016 Available online 16 February 2016 Keywords: Hot-pressing Spinel Sintering Optical properties Armor

a b s t r a c t Hot-pressing schedules – able to ensure a proper balance between positive and negative effects of LiF – allow fabrication of highly transparent MgAl2 O4 parts under low-pressure (≤50 MPa); peak temperatures in excess of 1550 ◦ C are needed, with best results obtained at ∼1650 ◦ C. At the concentration-levels and conditions present during hot-pressing, LiF does not react with spinel; it is inert also toward graphite. The lubricant and solvent abilities of liquid LiF make possible achievement of densification levels >90%TD under 1200 ◦ C. All LiF leaves the specimens as such at temperatures ≤1550 ◦ C. LiF’s ability to close porosity at temperatures lower than those (>1350 ◦ C) – which allow significant carbon penetration into spinel – is essential in achievement of transparency. Light absorption – occurring when carbon is present – is thus prevented. The main light-scattering defect produced by HPing, in the presence of LiF, is micro and macro-cracking; opaque spots, due to un-complete densification, are rare. © 2016 Elsevier Ltd. All rights reserved.

1. Introduction Highly transparent MgAl2 O4 (spinel) ceramic can be produced by unidirectional hot-pressing (HP), sometimes followed by hot-isostatic-pressing (HIP), by sinter-HIP, or by spark-plasmasintering (SPS) [1–8]. LiF was proposed (during the ‘70s) as a sintering aid for spinel, when densification is done by HPing – by W. Rhodes, D. Roy and R. Rice – after the salt was found to be effective in the hot-pressing of transparent MgO [1–3,8–10]; the early work also deciphered, partially, the mechanisms involved. Hot-pressing spinel, in the presence of LiF, allows fabrication of plates having surface areas of up to 72 × 74 cm2 , and large (base diameter 180 mm) missile noses [11,12]. The additive allows the achievement, for pressure ≤50 MPa, of transmission levels higher than 80% (theoretical limit for most of the spectral range of interest – 0.25–5.5 ␮m – is ∼87%) [11–14]. For 10 × 10 cm2 plates, sometimes even larger, the transmission level is uniform over all the major surface; very few micronssize residual scattering centers remain. However, for commercially successful manufacturing, further refinements of the method are needed, so as to be able to significantly reduce fabrication costs, improve mechanical properties and maintain the characteristics’

∗ Corresponding author. Fax: +972 4 8325339. E-mail address: [email protected] (A. Goldstein). http://dx.doi.org/10.1016/j.jeurceramsoc.2016.02.001 0955-2219/© 2016 Elsevier Ltd. All rights reserved.

reproducibility also in the case of large (L ≥ 0.3 m) size and/or complex shape items. Simultaneous fulfillment of these tasks (the main condition for achieving commercially relevant items), yet a goal for the future rather than an accomplished mission, may be facilitated by a comprehensive understanding of the physical-chemistry of the hot-pressing operation. Therefore the topic was, and is yet, the object of a quite extensive research effort. Several scenarios – purporting to explain the series of events occurring during a hotpressing cycle – have been put forward. Most of the recent progress in this domain was owed to the work of a few teams led, or inspired, by Reimanis and co-workers [15–20]. While, as a consequence of these studies, important aspects are well understood now, a model able to satisfactorily explain all the observed facts has not yet been, it seems to us, configured. This work’s intention is to provide some new insights allowing further advance toward the goal of a full understanding of the spinel hot-pressing operation in the presence of LiF or similar sintering aids. First, individual processes – like the spinel/LiF and spinel/carbon interactions, lubricant action and vaporization of LiF, and LiF replacement by MgF2 – were studied. The interaction pattern of the individual processes was then examined. Explanations proposed in the previous publications devoted to this topic are rediscussed, where necessary, in the light of the new data obtained here. Guidelines for the design of optimal HPing schedules are given.

1732

A. Goldstein et al. / Journal of the European Ceramic Society 36 (2016) 1731–1742

2. Experimental 2.1. Raw materials The grade SCR-30 spinel powder (Baikowski, La Balme de Sillingy, France) was used. This material is derived from Al and Mg sulphates, calcined at ∼900 ◦ C and jet-milled. Its most deleterious impurity are the SO2− 4 ions (100–600 ppm of sulphur). It also includes 30 ppm of Si, 10 ppm of Fe, and 10 ppm of Na; the specific surface area is A ≈ 30 m2 /g. The sintering additive (also labelled as “dopant” below) LiF was a reagent grade material supplied by Alfa–Aesar (Ward Hill, MA), having a 99.99% purity. MgF2 (99.99%) was supplied by an American company (American Elements, LA).

2.2. Processing

been heated to respectively 750, 950 and 1100 ◦ C, with a 2 h dwell being kept at these temperatures. For the study of the interaction between LiF and carbon, weighed graphite grains were seated in an alumina crucible into which LiF powder was poured in so as to submerge the graphite grog. The crucible was inserted in a large, loosely lidded, graphite crucible. This “reactor” was heated to 1350 ◦ C, under ∼1 atm Ar, for 1 h. The grains were weighed again in order to determine if mass loss, owed to interaction with LiF, occurred. For the examination of MgAl2 O4 solubility in liquid LiF, a spinel disc [pre-sintered at 1250 ◦ C to prevent permeation by LiF (l)] was covered with LiF grog and heated to 1050 ◦ C/2 h in air (in a lidded alumina crucible). The Mg and Al content in the LiF layer, formed atop the spinel base, were measured after cooling the specimens by quenching, in air. Each of the experiments concerning chemical reaction were repeated three times. 2.3. Characterization

Two methods were used for the introduction of LiF. In the first (used in most cases) solid-LiF and the spinel powder were dispersed, under the action of strong ultrasonication, in isopropyl alcohol. The dried powder was pelletized by sieving (70 mesh) and the granules were spheroidized by tumbling in a plastic cylinder (see [21]). The resulting material constituted the ready-to-press (RTP) feed for the hot-press. In the second approach, based on [22,23], LiF was dissolved in very slightly acidulated (HNO3 ) large volumes of water (LiF solubility in such conditions is of ∼0.2%) and then the spinel powder was suspended in this liquid by the aid of high intensity short (3 min) ultrasonication. Slow vaporization of the water (leading to precipitation of LiF on the spinel particles) was effectuated by heating, under stirring. Granulation was effectuated like above. The amount of LiF added varied between 0.7 and 5 %. Hotpressing was effectuated under vacuum or Ar in graphite dies with an internal diameter, ϕ, varying between 25 and 75 mm (most of the specimens had ϕ = 58 or 30 mm). The specimens were separated from the graphite punches by graphoil spacers. The thermal insulation system is made of graphite felt (carbon atom clusters are released by these graphite made items). The larger size discs were hot-pressed by the aid of a 50 tons machine (model 1-2300, Vacuum Ind. Corp. G.C.H., Somerville, MA; Ar). The smaller specimens were processed within a press of 18 ton (model HPW 200/250-2200-180 of FCT Group, Rauenstein, Germany; vacuum). Various processing regimes (schedules) were examined; some of them will be presented in section 3. The specimens (discs having a ϕ = 56 mm) used for air-sintering experiments were formed by cold-isostatic-pressing (CIP), at 80 MPa, of the feedstocks used for the HPing experiments. After heating with 500 ◦ C/h to peak temperature, a 2 h dwell was kept. For the study of the interaction between Al2 O3 , MgO, MgAl2 O4 and LiF – in the case of high salt concentrations – the powders were thoroughly mixed in a 1:2 molar ratio, by the aid of mortar and pestle. The mixtures, placed in lidded alumina crucibles, have

The bulk density (BD) was measured by the Archimedes technique, while phase composition determinations were made by the aid of XRD patterns recorded with an X-ray diffractometer (model APD 1000, Italstructures, Riva del Garda, Italy). Microstructure examination was conducted with a SEM (model Quanta 200, FEI, Eindhoven, NL) and an optical microscope (model Panphot-1 of Ernest Leitz, Wetzlar, Germany). Electronic chemical analysis (Mg, Al content in quenched LiF) was performed with the EMPA (EDS) attachment of a SEM microscope (Oxford instruments, Tubney Woods, Abingdon, UK). Transmission measurements were effectuated on polished specimens using a spectrophotometer (model Lambda 950, PerkinElmer, Shelton, CT). The scattered fraction of the transmitted light was cut by the aid of a 4 × 4 mm slit located between the specimen and detector. Therefore the optical transmission values given (T%) have an in-line transmission significance. The Li content (not measurable by EMPA) was determined by wet-chemistry. For this, the powder (−200 mesh) was dissolved in hot, concentrated solution, of HNO3 , and filtered. Measurement was made by atomic absorption spectroscopy (model 403 of Perkin Elmer, Waltham, MA; sensitivity ∼0.02%). For a number of the specimens for which the residual Li concentration was first analyzed—the concentration of F was also determined by the aid of an integrated capillary HPIC (conductive detection)(model 4000 ICS, Thermo-Scientific Dionex Div., Waltham, MA). Samples of 0.2 g were melted with Na2 CO3 at 950 ◦ C in Pt crucible. After leaching with water and filtering, the chromatographic measurement was effectuated. 3. Results and discussion 3.1. Densification level, light-scattering defects morphology and optical spectra of LiF-containing hot-pressed spinel In Table 1 the bulk density (BDHP ), residual LiF content, open porosity (OP), and visual aspect of HPed spinel parts – pure or

Table 1 Characteristics of HPed spinel discs (ϕ = 58 mm; t = 2 mm) as a function of LiF content and hot-pressing conditions. Additives (%)

Hot pressing conditions (◦ C/h/MPa)

BDHP (g/cm3 )

OP (%)

Visual aspect

Residual LiF (%)

– – 1.0% LiF 1.0% LiF 0.5% LiF + 1.5% MgF2 a 2.5% MgF2 b

1650/3/50 1580/3/50 1580/3/50 1650/3/50 1650/3/50 1650/3/50

3.573 3.540 3.570 3.577 3.577 3.576

0 0 0 0 0 0

Slightly translucent, black hue Slightly translucent, gray hue Colorless, significant transparency Colorless, highly transparent Colorless, significant transparency Colorless, significant transparency

– – ∼0.01 ∼0.01 <0.01 –

a b

Transmission spectrum of such a specimen is shown in Fig. 1 as curve “d”. Transmission spectrum of such a specimen is shown in Fig. 1 as curve “e”.

A. Goldstein et al. / Journal of the European Ceramic Society 36 (2016) 1731–1742

1733

Fig. 1. Transmission spectra of specimens HPed according to various schedules (dopants: LiF, MgF2 ). Peaks at  < 350 nm produced by Fe3+ [26]. a – Top quality commercial plate (t = 3 mm) produced by sinter-HIP; benchmark. b – Disc (ϕ = 30 mm, t = 3 mm) fabricated by the aid of the best HPing schedule developed here (2% LiF, 950 ◦ C/1 h dwell, t◦ max 1650 ◦ C, Pmax = 50 MPa; detailed schedule not disclosed; remnant defects are of the type shown in Fig. 3d). c – Disc (ϕ = 58 mm, t = 1.7 mm; 1% LiF) fabricated by the aid of the HPing schedule shown in Fig. 8 (shown in Fig. 2, disc “b”). d – Disc (ϕ = 58 mm, t = 2 mm) derived from material including 0.5% LiF + 1.5% MgF2 . e – Disc (ϕ = 58 mm, t = 2 mm) derived from material including 2.5% MgF2 .

including fluoride additives – is given. The LiF content indicated is based on the results given by Li concentration measurement; in all cases where the F amount was also checked (vide supra) it came out as being roughly in a 1:1 (molar) proportion to Li. In Fig. 1 the transmission of some specimens produced during this work is presented, together with that of one, top quality, plate, made by sinter-HIP (has a benchmark role here). In Fig. 2 some typical transparent specimens are imaged. The shape and size of the defects, able to scatter light, is illustrated in Fig. 3. Table 1 first indicates that near theoretical density can be achieved both with and in the absence of fluoride type (LiF, MgF2 or LiF + MgF2 ) sintering aids. The difference from specimen to specimen as far as BDHP is concerned (and consequently the residual porosity PO ), is barely sensed by the Archimedes technique based measurement. It is significant enough, however, to massively affect light transmission [14–20,24]. While the well processed LiF- (or MgF2 ) containing specimens are colorless (or exhibiting a faint grayish hue) and highly transparent, un-doped ones are only slightly translucent or opaque; all exhibit a deep gray to black tint. This is not surprising if one recalls the fact that calculations show how raising the pores content (uniform volume distribution; size in the 0.2–2 ␮m range) from ∼200 to not more than 400 ppm increases scattering intensity to a level which makes a transparent ceramic become almost opaque [24]. Fig. 1 shows that LiF-doped specimens can exhibit, after hot-pressing, high (T > 60 % at t ≥ 2.0 mm) levels of transparency in the visible domain of wavelengths. Fig. 1 also shows that while requiring different HP schedules, both LiF and MgF2 are able to prevent black coloration (i.e. massive uniform light absorption) and reduce porosity to a level consistent with high optical transmission. The LiF doped discs exhibit higher transparency than those based on MgF2 . As of now, however, the effectiveness of the two additives cannot be properly ranked based

on our data. The HPing schedules for spinel/LiF mixtures underwent a long optimization program; this remains as a future task in the case of MgF2 . Microscopic examination (optical and SEM) shows (Fig. 3) that two types of defect act as light-scatterers: opaque spots and different shape cracks. The defects of the type illustrated in Fig. 3a represent not fully densified opaque spots (with a size running between a few tens of micrometer up to 2 mm), including internal pores; individual pores are very rare. Those imaged in Fig. 3b and c (the most ubiquitous) are fine cracks running along grain boundaries; either localized systems, formed from a few intersecting cracks, or networks encompassing the whole volume of the specimens, were detected. Gilde et al. consider defects they identified by optical microscopy as being LiF decorated grain boundaries; it is more likely – considering their shape – that those features are cracks too (as detailed below the transient LiF is absent in the hot-pressed items) [25]. Radial crack systems originating from a common point toward the surface of a sphere (ϕ = 4–20 ␮m; not illustrated), and others shaped like a Vickers indenter’s mark (see Fig. 3 image “d”), could also be observed. The micro-cracks are present in almost all LiF-containing specimens. Even in the highly transparent one – for which the transmission is shown in Fig. 1 as curve “b” – a few defects are present, mostly indent shaped cavities (see Fig. 3 image “d”). Opaque white spots, however, are absent in well pressed LiF-containing parts, as long as ϕ ≤ 40 mm); such spots are observed in some of the specimens when ϕ > 58 mm. White opaque spots dominate in LiF-free specimens; cracking, on the other hand, was not observed in such parts. The transmission of what we label here as “transparent” fluoride-containing parts varied between 50 and 84% (at  = 650 nm; t = 2.0 mm) as a function of the pattern and surface

1734

A. Goldstein et al. / Journal of the European Ceramic Society 36 (2016) 1731–1742

Fig. 2. Illustration of transparent spinel parts fabricated by hot-pressing. a – Largest diameter disc (ϕ = 75 mm; t = 2 mm; 1.5% LiF) produced during this study (large crack developed during polishing), fabricated according to the HPing schedule shown in Fig. 8. b – Disc (ϕ = 58 mm; t = 1.7 mm; 1% LiF) fabricated according to the HPing schedule shown in Fig. 8. Its transmission spectrum appears in Fig. 1 as curve “c”. c – Disc (ϕ = 40 mm; 1% LiF) produced according to the HPing schedule shown in Fig. 8. d, e, f – Discs (ϕ = 58 mm; t = 3 mm; 1% LiF) produced according to the best HP schedule developed; (e = disc held at 2 cm above lettering.)

density of the micro-cracks. When T < 75%, a network type pattern is, usually, present. While the opaque spots are a type of defect frequently present also in spinel parts produced by AS + HIP approach, the types of crack described above are not observed there [5,21,26]. In the next sections the main individual processes – the interaction of which leads to specimens exhibiting the macroscopic characteristics described in this section – are analyzed.

3.2. Processes occurring during hot-pressing 3.2.1. Chemical reactions in the MgO, Al2 O3 , MgAl2 O4 -LiF systems; high (»5%) LiF content The tests run, in air, on specimens containing large LiF amounts (see Experimental) did not show reaction between any of the oxides considered and LiF (solid at this t◦ ) after 3 h at 750 ◦ C. MgO remained inert even at 950 or 1100 ◦ C. As opposed to that, reaction was detected in the case of Al2 O3 at the higher temperatures. LiF and Al2 O3 may form by their reaction both oxyfluorides like LiAl4 FO6 or double oxides like Li2 Al2 O4 , Li2 Al4 O7 or Li2 Al10 O16 (some of these compounds exhibit various polymorphs). In the conditions examined here, 2 h at 1100 ◦ C (air), a fraction of the LiF is consumed (another, as shown by data given in Table 3, is lost by vaporization, despite the presence of a lid) in reaction with Al2 O3 , leading to the formation of Li2 Al2 O7 (possibly also some Li2 Al10 O16 ); the same phase composition is seen at 950 ◦ C. The spinel/LiF mixture, after heating 2 h at 1100 ◦ C, gives the XRD pattern shown in Fig. 4, which includes spinel LiF and MgF2 peaks; the same pattern, but with less intense MgF2 peaks and somewhat stronger LiF peak (at 2 = 37.5◦ ) was obtained at 950 ◦ C. MgF2 ’s presence means that reaction occurred. No separate peaks corresponding to Li-containing aluminate(s) – which have to be present in proportion to the MgF2 formed – are seen, however. This is possible only if the residual spinel XRD peaks cover those of the formed Li-aluminate(s); this may happen only for Li2 Al10 O16 and/or Li2 Al4 O7 [the phase detected after LiF + Al2 O3 reaction]. Chemical reaction – when mixtures rich in LiF are heated – has been also

observed by Muller and Kleebe (5% LiF) and Villalobos et al. (50% LiF), albeit Li2 Al2 O4 was the aluminate detected [19,27]. During hot-pressing – when the LiF amount is always ≤5%, usually lower than 2% and the reaction conditions differ from those used above – host/additive (liquid or vapor phase) reactions seem (curiously, but luckily) to be absent. Were such reactions to occur, there is no way transparent spinel could result at the end of the hotpressing process. Li-containing phase layers, formed in the grain boundaries region, would consume all the sintering aid (owing to the large oxides supply available), under 1000 ◦ C, preventing it from assisting (as liquid) the densification process, in its low temperature stage. Rozenburg et al. suggest that only part of LiF is consumed in the reaction [their Eq. (1)] [15]. No reason for such a peculiar course (taking into account the low concentration of LiF and the very large one of Sp) of events is given, however. Moreover, at high temperatures the Li-aluminate layers (not removable by events taking place at t◦ > 1100 ◦ C [15]), between which the grain boundary would be located, are transported by ionic diffusion to the pores adjacent to the grain boundaries. That may help densification, but also create spots large enough to scatter light. In fact, further work by Reimanis et al., and others including the present work, did not find any evidence for Li-aluminates formation during hot-pressing [18]. In order to explain the absence of Li-containing phases in the HPed parts – shown by the experimental data to exist outside the hot-press – Rozenburg et al. suggested the occurrence, over 950 ◦ C, of the backward reaction (1) [15]. 2LiAlO2 (s) (and/or 2Li2 Al7 O4 ) + MgF2 (g) → MgAl2 O4 (s) + LiF (g)

(1)

The authors gave no thermodynamic reasons for the occurrence of (1). The fact that LiF (g) exits the system favorably affects the kinetics of the process, but is not a motivation for its occurrence. The theoretical arguments against the process (1) are supported by the experimental data, in Fig. 4, which shows that at 1100 ◦ C the process taking place (in conditions differing from those existent

A. Goldstein et al. / Journal of the European Ceramic Society 36 (2016) 1731–1742

1735

Fig. 3. Types of defect, causing light scattering, produced by HPing in the presence of LiF. a – Variable size, opaque (white) regions (spherical in most cases), not fully sintered, which appear as black discs in the figure (rare in all HPed specimens). b – Cracks as appearing on a specimen’s surface. c – Cracks on plane located inside the specimen. d – Indent shaped defects, seen especially in highest transparent specimens (cracks are short, rare or absent, in such discs).

Table 2 Characteristics of spinel/LiF (0 → 3.5%) discs (ϕ = 58 mm) after hot-pressing (50 MPa; Ar) up to various temperatures ≤1400 ◦ C (see text for schedules). LiF concentration (initial) (%) 1.0 1.0 1.0 1.0 1.0 1.5 3.5 3.5 1.0 3.0 1.0b 1.0b 1.0 – – – – – a b

Peak t◦ (◦ C) RT 700 860 860 1050 1050 1050 1270 1350 1350 1300 1350 1400 RT 1100 1270 1350 1400

Dwell time (min) a

10 30 30 120 40 120 120 20 40 40 40 40 60 10 120 30 30 60

BDHP (g/cm3 )

OP (%)

Residual LiF (%)

Visual aspect

1.48 2.10 2.70 3.10 3.30 3.44 3.47 3.20 3.00 2.90 3.50 3.52 3.55 1.70 2.05 2.25 3.25 3.30

47 38.0 18.0 4.2 2.0 1.0 0.9 7.0 4.0 8.0 0.4 0 0 – 28.0 22.0 3.5 2.5

1.00 1.00 0.89 0.85 0.49 0.60 2.90 0.28 0.05 0.20 0.08 0.08 0.05 – – – – –

White, opaque

White, translucent

White opaque

Gray, opaque

Cold isostatic pressing. Schedule different from that used for the rest of the specimens (see text Section 3.2.2).

in the hot-press) is the reaction also occurring at 950 ◦ C, i.e. the process (1) proceeds from right to left at all temperatures. In our view the fast (as long as open porosity exists) continuous movement of the gaseous fraction of LiF is the factor which prevents – in the conditions existent in the hot-press – a good liquid/solid contact, necessary for chemical reaction. Thus, the reaction process, despite being favored thermodynamically, is obstructed by a concurrent process: LiF intense vaporization. This is speculative; a definitive answer requires further study. 3.2.2. LiF (l) as lubricant and facilitator of liquid-assisted-sintering (LAS) (850–1300 ◦ C range) LiF melts at ∼850 ◦ C and – owing to its significant ability to wet spinel and its relatively low viscosity – spreads rapidly through the

powder compact’s void channels, so that a quite uniform distribution, at the micro and macro scales (does not necessarily imply penetration along grain boundaries), is achieved by the aid of the external pressure, in a matter of minutes [18]. As noted by Reimanis et al. the pattern of LiF (l) distribution in the 850–1250 ◦ C range is not known [18]. A low value of the static contact angle,  c,s – as measured by Meir et al. – while favorable for, does not automatically ensure penetration along grain boundaries [20]. In fact, Reimanis et al. did detect (specimens quenched from the 800–1000 ◦ C region) LiF only in pores. These authors attribute this to de-wetting, on cooling. The fact that even under quenching conditions residual LiF in the grain boundary region was not found, suggests a rather limited ability of LiF (l) to penetrate the inter-grains region. Now, even if LiF (l) does not massively penetrate the grain boundary regions

1736

A. Goldstein et al. / Journal of the European Ceramic Society 36 (2016) 1731–1742

Fig. 4. XRD pattern of material resulting from reaction between MgAl2 O4 and LiF (2:1 molar ratio) at 1100 ◦ C/2 h, air. S – Peaks related to MgAl2 O4 and, probably, Li2 Al2 O7 ; M – Peaks related to MgF2 ; L – Peaks related to LiF. Table 3 Characteristics of spinel discs including zero or 4% LiF after pressureless heating (air) at various temperatures. Initial LiF concentration (%)

t◦ (◦ C) WL (%) Residual LiF (%) BDa (g/cm3 ) OP (%)

0

900 1350 900 1350

4 a

0.6 0.8 1.8 5.2

– – 2.80 0.15

1.72 2.59 1.85 2.15

– 27 47 38

TD spinel = 3.578 g/cm3 .

(nano scale uniformity) but is uniformly distributed (at least at the micro scale) and present in the channels separating the green body’s agglomerates, the additive still may act as a good lubricant. The rheological and wetting characteristics of the liquid ensure the efficient correction of non-uniformities that may exist in LiF’s distribution, within the green-specimen. Therefore methods like the mixing of spinel with LiF in suspension can be used despite their inability to provide perfect mixing homogeneity. That is true, however, for specimens not too large in size. For very large plates (say 0.5–1.0 m) the more tedious solution method (see Section 2.2) – but which offers best LiF distribution uniformity – is preferred [22,23]. The BD, OP, residual LiF content and visual aspect of spinel specimens – pure or including fluoride additives – hot-pressed in different conditions (t◦ ≤ 1400 ◦ C; initial LiF content: 0–3.5%) are given in Table 2 for discs having a ϕ = 58 mm. For most of these experiments the pressure was raised to an initial level of 20 MPa at RT and then the temperature was raised in ∼1.5 h to the peak value. The pressure was further raised, simultaneously with the temperature, so as to reach 50 MPa when the peak temperature was attained. After the dwell time, pressure was released, and the system was brought back to RT by natural cooling. Further work indicated that such a schedule was not optimal for full densification. Even so, the BD values of Table 2 clearly show that the LiF (l) distribution pattern, whatever its actual configuration, facilitates advanced densification (92–99.5% TD) at low temperatures (1050–1400 ◦ C). For two specimens (marked “*” in Table 2) including 1% LiF a better schedule was used. For these specimens a 45 min dwell was kept at 950 ◦ C and then the temperature was raised in 8 h to 1300 or

1350 ◦ C, where a 40 min dwell was kept. The RT pressure of 20 MPa was kept till the end of the dwell at 950 ◦ C and then raised with the temperature, so as to attain its final 50 MPa value at 1300 or 1350 ◦ C. A first conclusion Table 2 suggests is that LiF (s) is not an efficient lubricant at RT. At 700 ◦ C, because it undergoes plastic deformation under quite low stress, its lubricating effect improves somewhat [28]. It is also seen that the moment it becomes liquid, its abilities in this sense are massively augmented. At temperatures as low as 860 ◦ C bulk density values of ∼87%TD can be attained, and this may be further improved to ∼97%TD at no more than 1050 ◦ C; dwell times at peak temperature of 40–120 min are necessary for achieving such BDHP levels. Various temperature-pressure-dwell times-LiF content combinations can be used, but ∼1% LiF is enough to trigger massive specimen consolidation. In Fig. 5 the shrinkage (represented by the punch displacement) under a certain HP schedule is shown, up to the end of the pressing cycle, for both a spinel containing 1% LiF and an un-doped one. The shrinkage is by and large finished around 1400 ◦ C for both cases; in other HP program designs it can stop already around 1100 ◦ C. Solid-state ionic diffusion in spinel is not fast enough, at such temperatures, to support shrinkage values at the level shown in Fig. 5. Particles rearrangement is responsible for at least the major part of this densification. In the absence of LiF (l) (curve “a1”) an initial fast, but limited scope, rearrangement is quickly slowed down, so that the level of shrinkage attained at 1250 ◦ C, in the presence of LiF, is reached at a temperature 150 ◦ C higher in un-doped spinel. In the LiF (l)-containing specimen (curve “a2”) the rearrangement is faster and more radical. The high ultimate densification levels attainable (95–97% TD) under 1400 ◦ C (Table 2), suggest that besides particles rearrangement, additional densification mechanism(s) operate. Plastic deformation or creep are unlikely to contribute at such low temperature and in the presence of a liquid lubricator, but the conditions required for liquid-assisted-sintering (LAS) are met. While not proving LAS occurrence, the fact that some dissolution of spinel in LiF (l) occurs shows that one of the essential conditions for LAS is met [others like good wetting and low viscosity of the LiF (l) also exist]. The existence of a limited level of spinel dissolution is substantiated by the results of the LiF coating chemical

A. Goldstein et al. / Journal of the European Ceramic Society 36 (2016) 1731–1742

1737

Fig. 5. Hot-pressing schedule and shrinkage thermal profile for a doped (1% LiF) and un-doped specimen. a1 – Punch displacement curve of un-doped disc (a measure of shrinkage along the pressing direction). a2 – Punch displacement curve of doped disc (a measure of shrinkage along the pressing direction). b – Temperature = f(time) profile. c – Applied force = f(time) profile.

analysis (LiF coated spinel disc, exposed to 1050 ◦ C/2 h and then quenched to RT). The average of measurements gave: 98.3% F, 1.18% Al, 0.52% Mg (Li not analyzed). Dissolution of spinel in LiF (l) was also observed by Reimanis and Kleebe [18]. Action of pressure assisted LAS, in the case of MgO hot-pressing, was proposed previously by Hart et al. [29]. According to these authors MgO sintering, under HP conditions, follows the general LAS model defined by Kingery; a good illustration of the way this mechanism works was provided in [30]. LAS, as one of the sintering mechanisms operating during MgAl2 O4 hot-pressing, is introduced also in one of the densification models proposed by Reimanis and co-workers [16,18]. The low apparent sintering activation energies reported by Rozenburg et al. for the case of LiF-containing specimens hotpressing, do not reflect the solid-state sintering stage alone but also the contributions of the processes facilitated by liquid LiF [31]. In the case of the schedule used for the majority of specimens in Table 2, the raising of the peak temperature over 1270 ◦ C, to 1350 ◦ C, reduces somewhat – instead of further increasing – the densification level; the larger the LiF concentration, the more pronounced this reduction is. This reduction does not take place, however, if another schedule (see specimens “*” also in Table 2) – which after ensuring densification under 1000 ◦ C includes a much slower further raise of the temperature to a peak value – is used. While occasionally light-gray hues may appear in them also, the LiF-containing specimens are white opaque under 1270 ◦ C, and white translucent when processed in the 1270–1400 ◦ C range. The additive-free specimens are, in most cases, white opaque under 1350 ◦ C, but exhibit various hues of gray over this temperature. The data in Table 2 also shows that most of the densification process is completed at 1400 ◦ C, for both doped and additive-free specimens (lowest BD = 92%TD, for the un-doped discs). Nonetheless the densification temperature profiles of doped and un-doped specimens differ to a certain extent. The LiF-free specimens lag behind the doped ones at any given temperature. The par becomes smaller

starting from around 1350 ◦ C. Over 1550 ◦ C the difference is further reduced, so that over 1600 ◦ C it is barely detectable; this evolution is illustrated by the curves a1 and a2 of Fig. 5. The consequences of this difference – by no means un-important – will be examined in Section 3.2.4. 3.2.3. LiF dissolution in spinel Li+ with an r4 = 0.59 Å may easily replace Mg2+ located in tetrahedral symmetry spinel sites, which has an r4 = 0.57 Å. The r6 = 0.76 Å of Li+ makes substitution of Al3+ (octahedral sites and an r6 = 0.54 Å) much more difficult [32]. The radius difference between F− (1.33 Å) and O2− (1.40 Å) is also moderate (∼5%), allowing replacement of the latter by the former. LiF entrance into a spinel lattice – according to the mechanism in which both ions are involved – may be thus represented as:

LiF + MgMg + OO ↔ LiMg + FO• + MgO

(2)

if the most likely replacement Li+ ↔ Mg2+ (documented for MgO) is considered [10,33]. Reimanis and Kleebe propose a different entrance process, leading to VO•• formation [16]: 3LiF

MgAl2 O4

→





LiMg + 2LiAl + 3FO• + VO••

(3)

There are reasons to consider a process like (3) unlikely. For instance, large r makes replacement of Al3+ by Li+ very difficult. Moreover, the two electrons freed by the formation of a VO•• are not accounted for in (3); if they remained trapped at the VO•• , the vacancy cannot function as electrostatic compensator for the negatively charged defects produced by the replacement of host cation(s) with Li. The displaced host ions whereabouts are not indicated. What is important to stress is that even if process (2), or similar, work, the spinel lattice – quite reluctant in accommodating foreign ions – is able to accept only a minute amount of LiF, mainly owing to an entropic drive. Such a view is consistent with the data of Table 2. It can be seen that the residual LiF content is low over 1250 ◦ C and

1738

A. Goldstein et al. / Journal of the European Ceramic Society 36 (2016) 1731–1742

Fig. 6. Hot-pressing schedule and shrinkage thermal profile for a doped (2.5% MgF2 ) specimen. a – Punch displacement curve (a measure of shrinkage along the pressing direction). b – Temperature = f(time) profile. c – Applied force = f(time) profile.

very low for specimens heated to 1400–1450 ◦ C. Chemical analysis of specimens heated over 1550 ◦ C – even ones including initial LiF concentrations in the 2–4% range – did not find residual LiF or, in a few cases, gave values at the detection limit of the method. Table 1, for instance, shows that from an initial amount 1%, the residual amount of LiF in the lattice – after HPing at temperatures ≥1580 ◦ C – is of ∼0.01% or less. If the entrance is according to (2) [it was shown

above that (3) is not correct] the LiMg is compensated by FO• not VO•• . But even if (3) were correct, the amount of VO•• , proportional to that





of LiMg (LiAl ), at peak temperatures, would be too low to significantly affect oxide ions diffusion and thus densification. The dopant’s influence on the high temperature stage of the sintering process is an observed fact, even if not via VO•• . An alternative mechanism of influence – the prevention of carbon penetration (and consequences on densification) – are discussed below (Sections 3.2.5 and 3.2.6). That presence of VO•• , created by LiF, is not a necessary condition for achieving full densification of spinel, is substantiated also by the fact that other fluorides have an effect similar to that of LiF on densification. For instance, as shown in Table 1 and Fig. 1 (curves “e” and “d”), MgF2 (or MgF2 + LiF) – if the HP schedule is properly adapted – is also able to assist in the process of transparent spinel fabrication. This shows that it is possible to achieve the level of

spinel densification required by transparency in the absence of LiMg . The fact that MgF2 operates by the same mechanism as LiF [namely generating a liquid able to facilitate low temperature particles rearrangement, into very compact configurations, and offering protection against carbon penetration (its vapor also contributes to the latter task)] is illustrated in Fig. 6. It can be seen that under the effect of external pressure the specimen undergoes a very fast shrinkage (curve “a”) over the melting point of MgF2 (∼1250 ◦ C). The shrinkage profile is similar to those of Fig. 5 (curve “a2”) and Fig. 8 – which refer to spinel/LiF HPing – while different from that produced in the absence of liquid (Fig. 5, curve “a1”). This shows that liquid MgF2 , under pressure, facilitates spinel powder densification in the same manner as liquid LiF does. “Same manner” means that first the liquid MgF2 facilitates densification to BDHP levels >95%TD at temperatures below 1400 ◦ C. For instance under 50 MPa – if the HPing schedule ends with a 1 h dwell at 1300 ◦ C – discs with a BDHP = 3.47 g/cm3 , OP = 0, translucent, white – are obtained. Pressing with a peak temperature of 1050 ◦ C (MgF2 still solid) gives a BDHP = 2.30 g/cm3 . The final (98–99.995% TD) densification, over 1500 ◦ C (based on ionic diffusion), is not associated with further shrinkage (being thus similar to the case of LiF also in

Fig. 7. Weight loss (in air, un-lidded crucible) of pure LiF after 2 h dwells at various temperatures in the 700–1200 ◦ C range.

this portion of the HPing schedule) because it sets in after the liquid assisted stage already effectuated most of the densification work. The fact that the HPing cycle of Fig. 6 brings about full densification is demonstrated by the transparency (see Fig. 1 curve “e”) of specimens thus obtained. 3.2.4. LiF vaporization dynamics When significant, vaporization of components is always exerting a negative influence on ceramic densification. In the case of un-doped MgAl2 O4 , for instance, vaporization of MgO may have such an effect [34,35]; the process is significant at t◦ > 1650 ◦ C, especially when oxygen-lacking atmospheres are used. In the case of LiF-containing specimens, the vaporization of the dopant has a number of important and contradictory effects. The vapor pressure produced by sublimation of solid lithium fluoride [LiF (s)] is low, being of only ∼3 × 10−3 Pa at 575 ◦ C in vacuum [36]. Larger values are measured at temperatures higher than the melting point, i.e. ∼850 ◦ C. This is illustrated by the data in Fig. 7, where the weight loss (WL) of LiF – located in an alumina un-lidded crucible, after 2 h dwell at various temperatures (1 atm, air) – is given. As it can be seen, the vaporization rate of LiF (l) steeply increases over 1180 ◦ C. The effect of LiF vaporization (air; spinel/LiF mixtures) in the case where no external pressure is active, is illustrated in Table 3, where the WL, residual LiF content, the BD and OP are given for the case of LiF-free spinel and MgAl2 O4 + 4% LiF specimens subjected to pressureless heating in air at 900 and 1350 ◦ C. Table 3 shows that a LiF-free powder compact – heated within a temperature range that leaves a significant fraction of the pores in an open-state – shows a mass loss consistent with a situation in which adsorbates (OH− , H2 O, SO2− 4 , possibly

some CO2− 3 ) are released. In the case of LiF-containing specimens, only a moderate fraction of it vaporizes as long as t◦ ≤ 1000 ◦ C, but almost all the additive is lost over 1300 ◦ C (together with adsorbates = 1.2%). It seems that adsorbates loss and LiF exit are parallel processes, like found in the case of MgO [37]. The exit of almost all the LiF initially present, at 1350 ◦ C, also implies that none of its ions are retained in reaction products. This means that at low concentration – even in an air type atmosphere – LiF and spinel do not react, as opposed to what happens for LiF amounts close to 50% (see Section 3.2.1). The data in Table 3 corroborates with that in Tables 1 and 2 (hot-pressing conditions) which also indicates that over 1250 ◦ C the residual Li content is very low. The reason for reaction blocking – in the specimens subjected to hot-pressing – may be the fact that in a totally or partially open system, the turbulence created by the LiF vapor’s movement within the internal channels (over 850 ◦ C) prevents an efficient contact between the liquid LiF and the solid. As a result, from the two concurrent possible pro-

A. Goldstein et al. / Journal of the European Ceramic Society 36 (2016) 1731–1742 Table 4 Characteristics of spinel + LiF discs after pressureless sintering (air or Ar) at 1520 ◦ C/2 h. LiF concentration Initial (%)

Residual (%)

0 0.7 3.0 0.7a

0 <0.02 0.06 <0.02

a

BDA (g/cm3 )

OP (%)

3.52 2.35 2.06 2.20

0.02 36.0 41.5 39.0

1 atm of Ar.

cesses (LiF vaporization and LiF/spinel reaction) only the former actually occurs. In the thermodynamic analysis effectuated by Esposito et al. it was found that the presence of carbon is also a factor which reduces the drive toward Li-aluminates formation [38]. If this latter effect is relevant is not clear, because, as will be seen below, carbon enters the system only at high (over 1300 ◦ C) temperatures. In Table 4 characteristics of spinel discs, subjected to sintering under air or Ar at 1520 ◦ C, are given. As it can be seen, most of the LiF is lost, no matter its initial amount, if temperatures over 1500 ◦ C are reached. The larger the initial LiF amount, the lower the BD reached by sintering. Table 4 also shows that in the absence of external pressure liquid LiF does not act as a sintering aid, but on the contrary, drastically hampers the process. In this context it is also worth noting that a specimen hot-pressed (50 MPa) at 700 ◦ C (BD = 2.1 g/cm3 ) and then slowly (60 ◦ C/h) heated to 1250 ◦ C/1 h for LiF elimination, maintained its bulk density, but its average pores size increased from ∼45 nm to ∼2 ␮m. Further heating, in air, to 1450 ◦ C did not improve the densification level. It is seen that during pressureless heating at temperatures ≥1450 ◦ C, where massive shrinkage of un-doped spinel greenbodies occurs, the densification level decreases, and size of the pores increases when the LiF concentration is raised. Correlation of the data of Tables 3 and 4 strongly suggests that the deleterious influence on densification stems from the massive internal gas pressure increase, at temperatures higher than ∼1250 ◦ C, caused

1739

by LiF vaporization. It is remarkable (Table 4) that despite the copious open porosity, LiF (g) escape is not fast enough to prevent internal pressure build-up; the high pressure cracks and/or breaks pore walls, and counters the effect on densification progress mass transport by solid-state diffusion would otherwise cause. In the case of hot-pressing, the deleterious effect of internal gas-pressure is successfully counterbalanced by the action of the external force applied. Let us recall that over 1550 ◦ C there is practically no LiF left no matter its initial content or hot-pressing schedule. LiF is leaving the specimens whether they retain open porosity or not (gas diffusion along grain boundaries in the latter case) in a way similar to that observed during transparent MgO fabrication [33,37]. The low amount of residual LiF (up to ∼0.2%) yet present up to ∼1550 ◦ C (from ∼1350 ◦ C open porosity is available no more), is not harmless. Its gas pressure may still lead to internal micro-cracking; the more difficult exit way and the high temperatures lead to high pressure despite the progressively lower amount of LiF present inside the specimen. 3.2.5. Carbon penetration into spinel. Mechanism of protection provided by LiF The experiments performed here indicate, in direct or indirect form, that the presence of LiF drastically reduces the ability of carbon clusters to settle inside the densifying spinel porous powder compact. For instance, Table 1 indicates that no blackening, during HPing, of properly LiF-doped specimens occurs (in the worst case some light gray hues may appear), while un-doped discs hotpressed at t◦ ≥ 1350 ◦ C exhibit a coloration consistent with carbon atoms penetration. Table 2 shows that even un-doped discs remain white as long as HPing temperature remains ≤1350 ◦ C. At higher temperatures carbon penetration in such specimens does occurs. It was speculated that the defense mechanism, introduced by LiF’s presence, is a chemical reaction between the dopant and C, at temperatures lower than 1000 ◦ C, according to [20]: nLiF(g) + nC → nLi+ + (CF)n

Fig. 8. Hot-pressing schedule, with dwell at 1250 ◦ C, allowing obtainment of highly transparent specimens (see curve “c” of Fig. 1; some microcracks yet present). a1 – Punch displacement curve (data corrected considering graphite parts volume variation); (a measure of shrinkage along the pressing direction). a2 – Punch displacement curve (un-corrected data). b – Applied force = t(time) profile. c – Temperature = f(time) profile.

(4)

1740

A. Goldstein et al. / Journal of the European Ceramic Society 36 (2016) 1731–1742

with the Li+ left behind entering the host lattice according to the process (3). As noted above, we think that Li+ enters (only in minute amounts, however) the spinel lattice together with F− , and not according to (3). This view is supported, among other things, by comparison of (3) and (4) regarding kinetics. Even supposing (3) and (4) as possible, the kinetics of (3) cannot match that of (4); without (3) consuming fast all the Li+ produced by (4), the latter process cannot occur. There are also other difficulties with the supposed process (4) [20]. For instance, (CF)n type compounds appear only when molecular F is the reactant (F2 + C reactions) [39]. In the presence of LiF the dominant products are Li/(CF)n type intercalation compounds; they decompose, however, over 800 ◦ C, into LiF + C. Carbon is not present, below 1300–1350 ◦ C, in amounts large enough to engage in reactions like (4) (see Table 2). It also has to be noted – in the discussion of protection mechanisms against carbon – that were carbon (present in quite large amounts) be able to chemically engage LiF, under 1000 ◦ C, it will consume all of it, preventing it from acting as a sintering aid. Moreover, simulation of the reaction conditions within a hot-press also suggests absence of chemical interaction between C and LiF. As described in the Experimental section, graphite grains were exposed to LiF vapor at 1350 ◦ C for 1 h. While the alumina container was damaged (Li2 Al2 O4 and/or Li2 Al10 O16 formed) the graphite particles, as gravimetry showed, remained untouched. Owing to the above, one has to consider other mechanisms for explaining the protection against carbon LiF is able to offer. We think, as suggested by Esposito et al., that physical processes explain these abilities of LiF [38]. Esposito et al. point out that the liquid LiF is a barrier making spinel/C contact difficult. We add the fact that LiF gas pressure, within the void channels (porosity) of the spinel specimen, is also a factor (probably the main one) able to prevent carbon atoms clusters penetration. Related to the issue of impurification with graphite, it is critical to add that visual observation shows that, in the conditions prevalent in the machines used here, carbon does not enter the specimens at temperatures lower than ∼1350 ◦ C; this holds for both LiF-containing and un-doped parts. However, un-doped specimens processed at temperatures higher than that invariably show various shades of gray, the coloration intensity increasing with temperature. On the other hand, most of the LiF-containing specimens remain white even at temperatures in which carbon penetration abilities become high. The main reason for the different behavior, regarding carbon penetration, seems to be the influence of LiF on the densification dynamics. Within most of the temperatures range where carbon becomes aggressive, the LiF-containing specimens, subjected to suitable pressing protocols (specimens “*” of Table 2 are such examples; not the best), exhibit only closed porosity, as opposed to the un-doped one. Carbon penetration in parts lacking open pores, while not necessarily fully stopped, proceeds with much more difficulty. Owing to the external pressure, the carbon penetration process may continue also for specimens lacking open pores, when the temperature is high enough. The higher the residual porosity (even if closed), the deeper the carbon penetration. For this segment of the pressing cycle, it is important that a slight positive LiF (g) pressure persists as long as possible. LiF out-diffusion along the grain boundaries may hamper carbon penetration also along this track. While keeping suitable amounts of LiF (g) inside the specimen is necessary, its pressure (high at these temperatures despite the low amounts of LiF) and pressure increase rates need to be controlled so as to avoid cracking (see Section 3.4). In the 1300–1350 ◦ C range a small amount of open porosity, together with closed porosity, may sometimes (depending on the specific profile of the HPing regime used) survive even in the presence of LiF. In such cases, for specimens including <1% LiF, the

probability of some (light gray color appears) carbon finding its way inside increases. Regarding the modalities of carbon penetration, some unclear aspects persist for the temperatures range where no more open pores survive. Particles originating from the various graphite sources can easily enter the spinel specimens as long as some open porosity exists; most of the carbon seen in un-doped specimens enters them in this stage. At higher temperatures, carbon atoms may still enter the closed specimens by diffusing along grain boundaries [26]. Another possibility is that CO – formed at the spinel/graphite interface – diffuses inside the part. 3.2.6. Effects of carbon penetration on densification and light absorption It is known that carbon in the form of single atoms or oligoatomic clusters may slow-down densification of oxide ceramics. It was shown, for instance, that in alumina very low amounts of carbon lead to formation of CO and CO2 , i.e. a gas pressure opposing densification [40]. Carbon interaction with spinel, leading to gaseous species, has been also discussed [38]. We propose that if located at grain boundaries it can also disturb ions diffusion toward the pore or across the boundaries (the latter effect reducing grains growth). Both effects get stronger when temperature increases. The presence of carbon, able to strongly absorb many frequencies of the visible domain, is also inducing, of course, gray to black tints into the host [41]. 3.3. Interaction of the individual processes during a full HP cycle Let us now briefly describe the sequence of events and their interaction during a typical HP run, as suggested by the data given in the prior sections. Vaporization of LiF begins to be noticeable around 700 ◦ C, with its rate increasing from 850 ◦ C (on melting of the salt), and on another inflexion on the LiF vapor pressure = f(t◦ ), around 1180 ◦ C. LiF does not engage in chemical reactions (under the conditions prevalent in the hot-press) with spinel, graphite or adsorbates on the particles surface, and exits the specimen as such (gaseous state). Up to ∼1550 ◦ C practically all LiF leaves the specimens, no matter the densification level achieved. The way this exit occurs has critical consequences on the configuration of the cracks system, with which the part leaves the HP machine; it also influences the way the high-temperature, final, stage of sintering plays-out; these, because the gas pressure inside the densifying spinel specimens may rapidly attain, if LiF evolvement is not fast enough over 850 ◦ C, values high enough to disrupt the microstructure, leading to pores enlargement and building-up of mechanical stress fields having a random volume distribution of their intensity. From 850 ◦ C, under the effect of pressing, consolidation and densification of the powder takes place by the aid of liquid LiF, in parallel to the gas pressure buildup. The rate and level of densification are markedly influenced by the way external pressure and temperature correlate over 850 ◦ C; this correlation controls the level of internal gas pressure. BDHP values of up to 97%TD can be attained under 1100 ◦ C, but excessive premature densification brings the LiF gas pressure to dangerous levels. Around 1350 ◦ C carbon atom clusters concentration and mobility reaches values which facilitate its penetration into the specimens (this temperature may vary from machine to machine, as a function of graphite release rates); the resistance to penetration, on the other hand, increases with the densification level of the specimen. The minute amount of residual LiF (g) present at temperatures higher than 1300 ◦ C is still able to contribute to the carbon penetration blocking. However, if its pressure is not controlled, it also may cause pores enlargement and cracking. At temperature over 1550 ◦ C only ionic diffusion based solid-state sintering mechanisms (including

A. Goldstein et al. / Journal of the European Ceramic Society 36 (2016) 1731–1742

creep and possibly some plastic deformation) operate. In this stage less than 1% of the total porosity is eliminated. A uniformly narrow size distribution of the pores (as ensured by LiF presence around 1000 ◦ C) facilitates full densification. Besides its effect on pores size distribution, LiF is not able to assist densification in this stage by other means, like [VO•• ] increase, as proposed by Reimanis and coworkers [15,18,19]. While only a small fraction of the porosity is dealt with in this stage, it is the one which makes the difference between translucent and transparent final states [24]. The tasks of LiF (l) can be accomplished also by liquids derived from MgF2 or LiF + MgF2 mixtures. In the absence of LiF (or similar), when low pressure sintering is used (P ≤ 50 MPa), carbon penetration cannot be fully prevented. Absorbed carbon is deleterious to densification, as far as rate and uniformity are concerned; blackening is also facilitated (carbon). As a result the HPed parts are colored and not transparent (include randomly distributed, variable size spots, of not fully densified regions and carbon atoms). If sufficient pressure is applied (>150 MPa, but best close to 400 MPa) however, full densification can be achieved even in the absence of LiF. Under such pressures densification may be completed at temperatures where carbon penetration is yet a low intensity process, and transparent spinel parts can be obtained [6,42]. Under 400 MPa, for instance, the peak temperature needed is of only ∼1200 ◦ C [6]. In the presence of LiF significant transparency can be attained – as shown in Fig. 1 – under much lower (≤50 MPa) pressure, the level depending on how adequate a HP schedule was used; however maximal temperatures in the 1600–1650 ◦ C range are necessary. There is one dominant defect which controls the transparency of LiF-doped parts: various shape and size cracks; such cracks have been reported also by Dericioglu et al. [43]. In our view the defects considered by Gilde et al. as LiF decorated grainboundaries, are in fact cracks too (see Fig. 5 of [25]). The cracks appear when LiF gas pressure and the densification rate are not well correlated. 3.4. Hot-pressing schedule optimization The previous Sections (3.2 and 3.3) indicate that a well designed HP schedule must ensure avoidance of carbon penetration and micro-cracks development. The experiments done during this work show that schedules leading to transparent specimens can be designed for LiF content in the 1.0–2.0% range. The essential aspect is to use temperature and pressure time profiles correlated with the amount of LiF in a suitable manner. Transparent specimens can be produced also by using MgF2 based doping packages. The transmission levels achieved here, when using MgF2 as dopant, are lower than those exhibited by parts including only LiF. For the former, however, no sustained effort to pressing schedule optimization was made. A practically important feature is the fact that various, quite different, temperature profiles can offer similar results. The schedule given in Fig. 8 (corresponding to the specimen the spectrum of which is represented by curve “c” in Fig. 1 and the specimens illustrated in Fig. 2a–c) and that of Fig. 5 have a dwell located at 1250 ◦ C. That (not given in detail here) used for obtaining the most transparent discs (T700 m = 84%; curve “b” in Fig. 1 and photos “d” to “g” in Fig. 2) has a 1 h dwell at 950 ◦ C. Dwells at low (<1500 ◦ C) temperature, while helpful, are not a necessary feature of the temperature time profile, but the 850–1400 ◦ C range has to be covered slowly enough to ensure pores closure without excessive internal gas pressure build-up. The slower the heating rate the lower the risk of micro-cracking. The effect on transparency of heating rate reduction can be appreciated by comparing the spectra of specimens “b” and “c” in Fig. 1; the heating rate used for specimen “b” was much lower than that used in the schedule, shown in Fig. 8, for specimen “c”.

1741

No schedule devised here was able to fully eliminate defects forming. Even the most transparent specimens, like that represented by the curve “b” of Fig. 1 (its actual HP schedule is not given owing to patenting intentions), has a few indent form defects (see Fig. 3d). 4. Conclusions The presence of LiF – or other additives with similar liquid generating abilities (e.g. MgF2 ) – is necessary for the fabrication of transparent spinel by hot-pressing, if low pressure (P ≤ 50 MPa) is used. Under such low pressures high sintering temperatures (≥1600 ◦ C) are required. The use of LiF can be avoided only if very high pressure (best toward 400 MPa) are brought to bear; such pressure allows the reduction of the peak temperatures under 1400 ◦ C [6]. LiF, when present in concentrations <5% and in the conditions existent in the hot-press, does not chemically interact with spinel. Carbon, if entering spinel, retards densification and confers black coloration. Carbon atoms clusters concentration within the volume of the die, and their mobility, are not high enough to lead to their entrance into the specimens, despite the presence of open pores in the latter, as long as temperature is kept under ∼1350 ◦ C. A main positive effect of LiF is the sealing of the specimens at temperatures lower than 1350 ◦ C; it contributes to carbon penetration blocking also after open porosity disappearance. In carbon-free machines it is possible that transparent spinel could be obtained even at low (<50 MPa) pressures, at temperatures around 1650 ◦ C, if MgO vaporization does not prevent this (using Ar instead of vacuum may mitigate this problem). The gas pressure of LiF also has an important negative effect, leading to pores size increase and cracking, if not kept at suitable levels. Other additives, like MgF2 , may work in a way similar to LiF. The main defects (able to act as light-scatterers) excessive LiF gas pressure causes are cracks of different size and pattern. Efficient hot-pressing schedules – able to keep a proper balance between the negative and positive effects of LiF – can be designed; the best of those we were able to design are quite long (i.e. >8 h till power switch-off), but allow one to achieve T values of up to 84% for t = 3 mm. References [1] D.W., Roy, F.J., Stermole, Method for Manufacturing a Transparent Ceramic Body; U.S. Patent 3,974,249, 1976. [2] D.W., Roy, Development of Hot-Pressed Spinel for Multispectral Windows and Domes, Final Research report AFTWAL-TR-81-4005, 1981. [3] D.W. Roy, J.L. Hastert, Polycrystalline MgAl2 O4 spinel for high temperature windows, Ceram. Eng. Sci. Proc. 4 (7/8) (1983) 502–509. [4] A. Krell, J. Klimke, T. Hutzler, Advanced spinel and sub-(m Al2 O3 for transparent armor applications, J. Eur. Ceram. Soc. 29 (2) (2009) 275–281. [5] A. Goldstein, A. Goldenberg, M. Hefetz, Transparent polycrystalline MgAl2 O4 spinel with submicron grains by low temperature sintering, J. Ceram. Soc. Jap. 117 (11) (2009) 1281–1283. [6] M. Sokol, S. Kalabukhov, M.P. Dariel, N. Frage, High-pressure spark plasma sintering (SPS) of transparent polycrystalline magnesium aluminate spinel (PMAS), J. Eur. Ceram. Soc. 34 (16) (2014) 4305–4310. [7] G. Bonnefont, S. Trombert, L. Bonneau, ‘Fine-grained transparent MgAl2 O4 spinel obtained by spark plasma sintering of commercially available nanopowders’, Ceram. Int. 38 (1) (2012) 131–140. [8] D.W. Roy, J.L. Hastert, L.E. Coubrough, K.E. Green, A. Trujillo, Transparent Polycrystalline Body with High Ultraviolet Transmittance, Process for Making, and Applications Thereof EP 0447390B1, 1994. [9] W. Rhodes, D. Sellers, T. Vasilos, Development and Evaluation of Transparent MgO. Report of Avco to U.S. Army, AMRA CR 67–01 (F), 1967. [10] R.W. Rice, Ceramic Fabrication Technology, Marcel Dekker, NY, 2003, pp. 166. [11] J. Spilman, J. Voyles, J. Nick, L. Shaffer, Manufacturing process scale-up of optical grade transparent spinel ceramic at armorline corporation, Proc. SPIE 8708 (870807) (2013). [12] J.L. Sepulveda, R.O. Loutfy, S. Ibrahim, S. Bilodeau, Large-Size Spinel Windows and Domes, Pro SPIE, 8708, Window and Dome Technologies and Materials XIII, 870806 15 pages (2013).

1742

A. Goldstein et al. / Journal of the European Ceramic Society 36 (2016) 1731–1742

[13] D.C. Harris, History of Development of Polycrystalline Optical Spinel in the U.S., Proceedings SPIE, 5786. Window and Dome Technologies and Materials IX, 22 pages, 2005. [14] A. Krell, J. Klimke, T. Hutzler, ‘Transparent compact ceramics: inherent physical issues’, Opt. Mater. 31 (8) (2009) 1144–1150. [15] K. Rozenburg, I.E. Reimanis, H.J. Kleebe, R.L. Cook, Chemical interaction between LiF and MgAl2 O4 spinel during sintering, J. Am. Ceram. Soc. 90 (7) (2007) 2038–2042. [16] I.E. Reimanis, H.J. Kleebe, Reactions in the sintering of MgAl2 O4 spinel doped with LiF, Int. J. Mater. Res. 98 (12) (2007) 1273–1278. [17] M. Rubat du Merac, I.E. Reimanis, C. Smith, H.J. Kleebe, M.M. Muller, Effect of impurities and LiF additive in hot-pressed transparent magnesium aluminate spinel, Int. J. Appl. Ceram. Technol. 10 (Suppl. (s1)) (2013) E33–E48. [18] I. Reimanis, H.J. Kleebe, A review on the sintering and microstructure development of transparent spinel (MgAl2 O4 ), J. Am. Ceram. Soc. 92 (7) (2009) 1472–1480. [19] M.M. Muller, H.J. Kleebe, ‘Sintering mechanisms of LiF-doped Mg-Al-spinel’, J. Am. Ceram. Soc. 95 (10) (2012) 3022–3024. [20] S. Meir, S. Kalabukhov, N. Froumin, M.P. Dariel, N. Frage, Synthesis and densification of transparent magnesium aluminate spinel by SPS processing, J. Am. Ceram. Soc. 92 (2) (2009) 358–364. [21] A. Goldstein, A. Goldenberg, M. Vulfson, Technology for the obtainment of transparent MgAl2 O4 spinel parts, J. Ceram. Sci. Technol. 2 (1) (2011) 1–8. [22] R.O. Loutfy, J.L. Sepulveda, S. Chang, Ready-to-Sinter Spinel Nanomixture and Method for Preparing Same; U.S. Patent 8,313,725 B2, 2012. [23] G.E. Villalobos, J.S. Sanghera, S. Bayya, I. D. Aggrawal, Magnesium Aluminate Transparent Ceramic Having Low Scattering and Absorption Loss; U.S. Patent 7, 528,086 B2, 2009. [24] R. Apetz, M.P.B. van Bruggen, Transparent alumina: a light-scattering model, J. Am. Ceram. Soc. 86 (3) (2003) 480–486. [25] G. Gilde, P. Patel, P. Patterson, D. Blodgett, D. Duncan, D. Hahn, Evaluation of hot pressing and hot isostatic pressing parameters on the optical properties of spinel, J. Am. Ceram. Soc. 88 (10) (2005) 2747–2751. [26] A. Goldstein, Correlation between MgAl2 O4 -spinel structure, processing factors and functional properties of transparent parts (progress review), J. Eur. Ceram. Soc. 32 (11) (2012) 2869–2886. [27] G.R. Villalobos, J.S. Sanghera, I.D. Aggarwal, ‘Degradation of magnesium aluminum spinel by lithium fluoride sintering aid’, J. Am. Ceram. Soc. 88 (5) (2005) 1321–1322.

[28] R. Marder, C. Estournès, G. Chevallier, R. Chaim, Plasma in spark plasma sintering of ceramic particle compacts, Scripta Mater. 82 (2014) 57–60. [29] P.E. Hart, R.B. Atkin, J.A. Pask, Densification mechanisms in hot-pressing of magnesia with a fugitive liquid, J. Am. Ceram. Soc. 53 (2) (1970) 83–86. [30] L.S. Sigl, H.-J. Kleebe, Core/rim structure of liquid-phase-sintered silicon carbdie, J. Am. Ceram. Soc. 76 (3) (1993) 773–776. [31] K. Rozenburg, I.E. Reimanis, H.J. Kleebe, R.L. Cook, Sintering kinetics of a MgAl2 O4 spinel doped with LiF, J. Am. Ceram. Soc. 91 (2) (2008) 444–450. [32] R.D. Shannon, C.T. Prewitt, Effective ionic radii in oxides and fluorides, Acta Crystallogr. B 25 (5) (1969) 925–946. [33] E. Carnall Jr., The densification of MgO in the presence of a liquid phase, Mater. Res. Bull. 2 (12) (1967) 1075–1086. [34] Y.M. Chiang, Grain Boundary Mobility and Segregation in Nonstoichiometric Solid Solution of MgAl2 O4 Spinel, Ph.D. Thesis, MIT, Cambridge, MA, 1985. [35] C.J. Ting, H.Y. Lu, Deterioration in the final-stage sintering of magnesium aluminate spinel, J. Am. Ceram. Soc. 83 (7) (2000) 1592–1598. [36] A.M. Evseev, C.V. Pozharskaya, A.N. Nesmeyanov, Y.I. Gerasimov, Vapor pressure of LiF, Zh. Neorg. Khim. 4 (10) (1959) 2189–2191. [37] R.W. Rice, The effect of gaseous impurities on the hot pressing and behavior of MgO CaO and Al2 O3 , Proc. Brit. Ceram. Soc. 12 (1969) 99–123. [38] L. Esposito, A. Piancastelli, P. Miceli, S. Martelli, A thermodynamic approach to obtaining transparent spinel (MgAl2 O4 ) by hot pressing, J. Eur. Ceram. Soc. 35 (2) (2015) 651–661. [39] N. Watanabe, T. Nakajima, M. Kawaguchi, Ternary Intercalation Compound of a Graphite with an Alkali Metal Fluoride and Fluorine A Process for Producing the Same, and an Electrically Conductive Material Comprising the Ternary Intercalation Compound; U.S. Patent 4,515,709, 1985. [40] S.J. Bennison, M.P. Harmer, Swelling of hot-pressed Al2 O3 , J. Am. Ceram. Soc. 68 (11) (1985) 591–597. [41] T. Surovell, Atomic Spectra-Graph, J. Chem. Educ. 71 (5) (1994) 422. [42] L.V. Udalova, M.V. Maltzev, V.I. Petrik, Transparent Spinel Ceramic, in: P. Duran, J.F. Fernandez (Eds.), Proceedings of the 3rd Euro-Ceramics, vol. 1, Faenza Editrice Iberica, Madrid, 1993, pp. 709–714. [43] A.F. Dericioglu, A.R. Boccaccini, I. Dlouhy, Y. Kagawa, Effect of chemical composition on the optical properties and fracture toughness of transparent magnesium aluminate spinel ceramics, Mater. Trans. 46 (5) (2005) 996–1003.