Karrooite green pigments doped with Co and Zn: synthesis, color properties and stability in ceramic glazes

The solid-state synthesis and stabilization of Co doped (Mg1-xCoxTi2O5), Zn doped (Mg1xZnxTi2O5) and Coand Zn-codoped karrooite solid solutions (Mg0.8-xZn0.2CoxTi2O5 and (Mg0.5Zn0.5)1-xCoxTi2O5) were investigated. In addition, the optical spectra, color properties and technological performance of (Co,Zn)-karrooite compositions as new green ceramic pigments were also analyzed. XRD characterization revealed for the first time the high solid solubility of Zn 2+ in MgTi2O5 karrooite at 1200 oC (between 60 and 80 mol% per Mg or karrooite formula unit). In contrast, the reactivity and stabilization of karrooite phase decreased in the case of Co 2+ doping. Interestingly, codoping with Zn 2+ ions at high molar ratios (Zn:Mg ratio equal to 1:1) enhanced the reactivity and enabled the stabilization of (Co,Zn)-MgTi2O5 karrooite solid solutions, even with high Co 2+ loadings (20 mol% per karrooite formula unit). The (Co,Zn)MgTi2O5 pigments exhibited yellowish-green colors associated to Co 2+ ions allocated in octahedral M1 and M2 sites of karrooite lattice, and becoming more intense and less yellow the higher the Co content. However, Zn 2+ codoping produced less saturated green colors with similar green but lower yellowish hues. The obtained pigments were not stable enough within the tested ceramic glazes, giving rise to turquoise colorations due to cobalt leaching and incorporation into tetrahedral sites of the glassy phase. The stability of Co-karrooite green

3 are generally associated to cobalt in tetrahedral coordination [4][5][6]. Today a common source of green color for the ceramic industry is a mixture of Pr-zircon yellow and V-zircon blue pigments, since other alternative green pigments contain large amounts of toxic and pollutant elements such as Cr, Co or Ni ("green of Sèvres" based on Cr 2 O 3 ; Uvarovite green garnet, Ca 3 Cr 2 Si 3 O 12 , with CPMA number 04-07-3 [7]; cobalt chromite green spinel, CoCr 2 O 4 , with CPMA number 13-30-3; cobalt titanate green spinel, Co 2 TiO 4 ,with CPMA number 13-31-3; and nickel silicate green olivine, Ni 2 SiO 4 , with CPMA number 05- . In this respect, the employment of allochromatic solid solutions doped with a low or minimized amount of Co (Ni or Cr) may be an interesting, less-pollutant and less-expensive alternative for green ceramic pigments [8][9][10][11], instead of idiochromatic (non-substituted) pigments of CoCr 2 O 4 , Co 2 TiO 4 and Ni 2 SiO 4 .
With the above precedents, in the present investigation we scrutinized the formation by ceramic route of MgTi 2 O 5 -CoTi 2 O 5 greenish solid solutions (Mg 1-x Co x Ti 2 O 5 ) in a more extended compositional range (up to 40 mol% of Co per karrooite unit formula, due to the environmental concerns previously mentioned). CoTi 2 O 5 crystallizes in the same pseudobrookite structure as MgTi 2 O 5 karrooite, and it is also an entropy-stabilized phase [12][13][14]; it becomes more stable the higher the temperature, and below 1142 ºC it decomposes in a mixture of the more stable CoTiO 3 ilmenite and TiO 2 rutile phases [15][16][17]. Given the isomorphism between Co-Ti and Mg-Ti pseudobrookites as well as the quite similar effective ionic radii for Co 2+ (74.5 pm) and Mg 2+ (72 pm) ions in octahedral coordination [18], we could expect a whole range of solid solubility of Co ions in MgTi 2 O 5 karrooite.
Moreover, we also investigated the effect of Zn 2+ codoping on the stabilization of Co-doped karrooite solid solutions (Mg 1-x-y Co x Zn y Ti 2 O 5 ), the color and optical properties of the resulting pigments, and its industrial performance either as ceramic pigments or dyes within ceramic glazes [19][20][21][22][23]. In spite of the existence of some precedent investigations analyzing the effect of Co-and Zn-codoping on related MgTiO 3 ilmenites (in order to improve its dielectric properties) [24,25], as far as we know there are not preliminary reports in the literature concerning Co-and 4 Zn-codoped MgTi 2 O 5 karrooite. In addition, there is also no experimental evidence of the synthesis of ZnTi 2 O 5 with a parent structure to MgTi 2 O 5 and CoTi 2 O 5 pseudobrookites. Indeed, although some TiO 2 ·xZnO compositions including ZnTi 2 O 5 stoichiometry have been prepared as high frequency dielectric materials [26] or as photoactive ZnTi 2 O 5 thin films for solar cells [27], the supposed crystallization of ZnTi 2 O 5 was not demonstrated.
In this respect, different ternary compounds with interesting technological applications are already known as stable phases in the Zn-Ti-O ternary system [28][29][30][31], such as Zn 2 TiO 4 inverse cubic spinel, ZnTiO 3 rhombohedral ilmenite and Zn 2 Ti 3 O 8 cubic defect spinel. Interestingly, other unreported or missing phases in this system have been very recently predicted to be borderline stable, such as ZnTi 2 O 4 normal spinel, ZnTi 2 O 5 pseudobrookite (chemically similar or equivalent to MgTi 2 O 5 karrooite), and ZnTi 5 O 10 in the high-symmetry Ti 3 O 5 pseudobrookite structure [32]. As experimental evidence, the related Zn x Ti 3-x O 5 phase (with x  0.6) was successfully synthesized in this recent study, adopting essentially the same structure as ZnTi 5 O 10 . This compound was found as much stable as the other unreported ternary Zn-Ti-O phases, highlighting the interest to perform further explorative investigations on these missing materials of the Zn-Ti-O system, including ZnTi 2 O 5 pseudobrookite.
Accordingly, in addition to the use of Zn as codoping agent in Co-MgTi 2 O 5 pigments, in this study we firstly investigated the MgTi 2 O 5 :ZnTi 2 O 5 system to elucidate the solid solubility limit of Zn 2+ ions in karrooite lattice. In the case of Co-karrooite pigments, the addition of Zn 2+ as codoping agent or structure modifier could induce a considerable color change due to modification of Co 2+ octahedral crystal field [33] and the covalent character of Co-O bonds [34] in karrooite lattice. On the other hand, the choice of Zn 2+ ions could also serve to improve the stability of Co-karrooite compounds as true ceramic pigments within Zn-Ca-borosilicate ceramic glazes, as it has been recently found in other Zn-containing pigments such as Co-doped Ca 2 (Zn,Co)Si 2 O 7 hardystonite [35].

Samples preparation (Ceramic route)
A set of four different (Mg 1-x-y Co x Zn y )Ti 2 O 5 karrooite-based solid solutions (series) were prepared by the conventional ceramic route, using Mg 5 (CO 3 ) 4 (OH) 2 .xH 2 O (42% as Mg, Aldrich), Co 3 O 4 (70% as Co, Panreac), ZnO (98%, Panreac) and TiO 2 (anatase, 99-100.5%, Panreac) as precursors. The location of the four series compositions on a ternary Mg-Zn-Co diagram is shown in Figure 1. First of all, a whole series of Zn:karrooite solid solutions (Kar- In a typical sample preparation, the corresponding stoichiometric mixture of precursors was homogenized in planetary mills (20 min) using acetone as dispersant. The dried powders were then directly calcined in an electrical furnace up to 1200 ºC with heating rate of 5ºC/min, a soaking time of 3 hours and with free cooling to room temperature. Kar-Zn samples were also subsequently fired at 1400 ºC (after due homogenization and micronization of previously fired compositions). After the firing treatment, all the fired powders were homogenized and micronized (mortar and pestle) and sieved below 63 m before further characterization (section 2.2).

Samples characterization
Crystal chemical characterization of karrooite-based (Mg,Zn,Co)Ti 2 O 5 calcined samples was performed by X-ray powder diffraction (XRD) in a D4 Endeavor (Bruker-AXS) powder diffractometer with Cu-K  radiation (from 10 to 70º 2, with steps of 0.05º 2 and a counting time of 2s per step). The diffractometer was equipped with a graphite secondary monochromator to eliminate K  and fluorescence signals. The cell parameters were also determined by indexing and least-squares refinement of XRD patterns of selected powder compositions mixed (50 wt%) with Al 2 O 3 corundum as internal standard and measured under slower conditions (steps of 0.02º 2 and a counting time of 4 s per step), using the POWCAL and LSQC programs (Department of Chemistry, University of Aberdeen, UK).
On the other hand, the microstructure and morphology of (Mg,Zn,Co)Ti 2 O 5 fired samples was examined by scanning electron microscopy (SEM) with a JEOL 7001F electron microscope (following conventional preparation and imaging techniques). The chemical composition and homogeneity of the samples was determined by semi-quantitative elemental analysis with an EDX analyzer (supplied by Oxford University) attached to the microscope.
In order to analyze the stability and optical or coloring properties (pigmenting performance) Enameled samples were fired following fast-firing schedules (52 min of duration from cool to cool at a maximum temperature of 980, 1050 or 980ºC for glazes A, B and C respectively). The optical properties of fired powders and glazed samples were then analyzed by diffuse reflectance spectroscopy (UV-vis-NIR) performed with a Jasco V670 spectrophotometer. The color parameters (L*a*b*) were also measured following the CIE-L*a*b* colorimetric method recommended by the CIE (Commission Internationale de l'Eclairage) [36], using an 8/d geometry (diffused illumination of 8º), with the observer at 10º and a standard lighting D65. On this method, L* is the lightness axis (black (0)  white (100)), a* is the green (-) red (+) axis, and b* is the blue (-)  yellow (+) axis.

Solid solubility of Zn in MgTi 2 O 5 (Mg 1-x Zn x Ti 2 O 5 solid solutions)
A whole set of Zn-karrooite solid solutions (Kar-Zn, Mg 1-x Zn x Ti 2 O 5 ) was first prepared in order to elucidate the solid solubility limit of Zn 2+ ions in karrooite. Figure 2a  Interestingly, a subsequent thermal treatment at a higher firing temperature (1400 ºC) did not succeed in increasing the range of solid solubility above x = 0.6, although reactivity increased considerably. As it may be observed in Figure 2b  On the other hand, the XRD patterns of Mg 1-x Zn x Ti 2 O 5 powders fired at 1400 ºC (up to x = 0.6, and mixed with Al 2 O 3 corundum as internal standard) were also refined and indexed, according to orthorhombic symmetry and Cmcm (63) space group, in order to measure the cell parameters of karrooite lattice and analyze the effect of Zn doping. As it is well-known, it is not a simple task in doped karrooite compositions to separate the contributions on unit cell dimensions of both ionic radii variations and order-disorder effects; for instance, an increase of unit cell dimensions may arise not only from the introduction of a larger cation but also from an increase of cationic disorder (since this involves a higher ratio of the larger Mg ions on the smaller M2 octahedral sites of pseudobrookite lattice) [2,37,38]. The coexistence of both effects usually leads to irregular trends in the variation of unit cell dimensions in doped karrooite samples. In our case, Figure 3 shows the evolution of cell parameters and cell volume with Zn content (x) in Zn-doped karrooite solid solutions (Mg 1-x Zn x Ti 2 O 5 ). As it may be observed, Zndoping produces an anisotropic variation of cell parameters, with a slight decrease of a parameter, an increase of c parameter and without any clear trend for b parameter. As a result, the unit cell volume decreases for the first doping stage (from x = 0 to 0.2), which could be explained by an increase of cationic ordering [12][13][14], while further Zn-doping (x = 0.4 and 0.6) causes a slight increase of unit cell volume, which would be in agreement with the slightly higher ionic radius of Zn 2+ (74 pm) with respect to Mg 2+ (72 pm) [18].

XRD characterization of (Co,Zn)-karrooite pigment powders.
The XRD patterns of (Co,Zn)-Karrooite solid solution pigments prepared by the ceramic route and fired at 1200 ºC are shown in x Co x Ti 2 O 5 ; 40-50 mol% of Zn). This result evidences a favourable effect of Zn codoping on the stabilization of Co-karrooite structure at higher molar loadings. This enhanced stabilization of (Co,Zn)-doped karrooite may arise either from an increase of configurational entropy (cationic disorder) with Zn codoping (above 20 mol%), or it could simply be due to kinetics reasons (increased solid-state reactivity), given the higher proportion of ZnO in raw mixtures, which would be therefore more reactive than Mg hydroxy-carbonate.
Similarly to Kar-Zn series, we measured also the unit cell parameters and volume of Kar-Co and Kar-CoZnH samples from XRD patterns, in order to evaluate both the effect of Co doping and of Zn codoping on karrooite unit cell. According to the representation shown in  Regarding to Kar-Co ceramic pigments (Mg 1-x Co x Ti 2 O 5 ), the fired powders contained rather compact and sintered particle aggregates (see e.g. Figure 6a, corresponding to non-sieved powder with x = 0) with variable sizes ranging from a few to ca. 50 m. These aggregates consisted of smaller grain-like particles, with sizes between 1 and 4 m and presenting more or less rounded morphologies, as it may be observed in SEM images of Figures 6b-6f. Noteworthy, the morphology or microstructure of pigment powders was very similar in all samples (irrespective of the amount of Co or Zn doping), apart from the slightly larger particle sizes Semiquantitative EDX analyses were performed in different regions of samples aiming to determine the chemical composition and homogeneity of Co-and Zn-codoped Karrooite solid solutions. In general terms, the average sample compositions determined by EDX analyses (see Table 1) were not far from nominal formulations, although they showed a considerable deficiency of Mg and an excess of Co (lower Mg:Co ratio).

Color characterization of (Co,Zn)-karrooite pigment powders.
The effect of Co and Zn doping on the color intensity and chromaticity of the obtained (Co,Zn)-karrooite fired powders was also investigated, in order to evaluate their potential use as ceramic pigments. As it may be seen in Figure 7 (which also includes the measured CIE-L*/a*/b* color parameters), all the obtained compositions exhibited a varied gamut of yellowish-green colors, with a considerable variation of color hue and intensity with Co and Zn doping. In order to facilitate the analysis of color variations in the different pigment series,

UV-VIS-NIR spectroscopy of (Co,Zn)-karrooite pigment powders.
The UV-vis-NIR absorption spectra (Kubelka Munk absorbance function, K/S) of Cokarrooite pigment powders (Kar-Co, Mg 1-x Co x Ti 2 O 5 ) prepared by ceramic route and fired at 1200 ºC/3h are shown in Figure 9 (above). The absorption profiles of Kar-Co pigments presented similar features to those previously described by Matteucci et al. for cobalt-doped karrooite [2]. These spectra fit perfectly with the spectral features of Co 2+ (a 3d 7 ion) in octahedral coordination [33,39]. Noteworthy, the absorption bands are considerably red-shifted (resulting in green colors) with respect to those exhibited by octahedral-coordinated Co 2+ ions in other pigment structures such as Co-olivines [40] or Ca,Co-pyroxenes [4], which develop the more usual violet or pink colorations.
As it may be observed, the spectra present the typical three spin-allowed transitions from Co 2+ ground state ( 4 T 1g ( 4 F)) to the different exited terms. The highest energy (and more intense) 12 band associated to  3 transition ( 4 T 1g ( 4 F)  4 T 1g ( 4 P)) is centered in the visible region at ca. 590 nm (yellow-orange region) leaving a deep reflection window around 500 nm which is responsible for the yellowish green colors of the pigments. Then, a much weaker or less intense band is observed extending from 700 to 900 nm, which is associated to the less-intense (twoelectron)  2 transition ( 4 T 1g ( 4 F)  4 A 2g ( 4 F)). Finally, a very broad and double-peak band may be observed in the near-IR region extending from 1050 to 1650 nm, with the two peaks centered at around 1190 and 1430 nm, which is attributed to the lowest energy spin-allowed transition  1 ( 4 T 1g ( 4 F)  4 T 2g ( 4 F)). This clearly visible splitting of  1 transition could be attributed to the different local environments found by Co 2+ ions at the non-equivalent M1 and M2 octahedral sites of karrooite, which causes the corresponding optical absorption bands to occur at different energies for both sites. However, this "splitting effect" was not so evident for  2 and  3 transitions. To this respect, a perfect deconvolution of the spectra (not shown) was not feasible without considering this doubling of all the spin-allowed transitions associated to both M1 and M2 sites (as it was also reported by Matteucci et al. [2]). Moreover, the small absorption contribution (shoulders) associated to other spin-forbidden transitions to 2 E g ( 2 G), 2 T 1g + 2 T 2g ( 2 G) and 2 A 1g ( 2 G) excited terms (from lower to higher energy) must be also considered. The energy positions of these forbidden transitions are also indicated by arrows in Figure 9.   2). Therefore, and due to the competitive bond effect, replacement of Mg 2+ by Zn 2+ in our compositions is expected to reduce the covalence (or increase the ionicity) of competing Co-O bonds in octahedral sites, thus also modifying the crystal field applied on Co 2+ chromophore ions and the observed color.

Technological performance of (Co,Zn)-karrooite powders as ceramic pigments or dyes
Finally, the potential application of these greenish (Co,Zn)-karrooite solid solutions as ceramic pigments or dyes for the coloration of low-temperature ceramic glazes was also tested.
In order to evaluate their stability and coloring performance, fired pigments were enameled (5 wt% of pigment) within three conventional transparent ceramic glazes (A, B and C) of relatively low firing temperature compressed between 980 and 1050 ºC (the composition details and firing schedules were given in section 2.2). The main research interests were to ascertain whether these Co-doped karrooite pigments were or not sensitive to ceramic glazes rich in alkalis (CaO and ZnO) [2], and also to analyze the effect of Zn-codoping on the stability of these Co-karrooite pigments. Indeed, Zn-containing pigments have been found in some cases to be more stable against chromophore (Co 2+ ) dissolution in the glaze in these more aggressive environments of Zn-Ca-borosilicate melts [35].
The color aspect of glazed samples including the measured color parameters (L*a*b*) may be seen in Figure 10 for the three pigment series (Kar-Co, Kar-CoZnL and Kar-CoZnH). A very simple and direct indication of the higher or lower stability of these (Co,Zn)-karrooite pigments may be inferred just by visually observing the color change produced or not after enamel firing (980-1050 ºC) in glazed samples. As it may be observed, glazed samples presented rather bluish or turquoise (greenish blue) colorations instead of the original greenish color of pigment powders, and this color change indicates that the pigments did not withstand the chemical attack within the ceramic glazes. Surprisingly, the greenish hue (negative a* parameter), which can be unambiguously associated to the color of stabilized (not dissolved) Co-karrooite pigment particles, was better preserved in Ca-and Zn-rich ceramic glaze (B) fired at 1050 ºC. In contrast, pigments enameled in glazes with a low Ca and Zn content (A and C) and fired at a lower temperature (980 ºC) presented more turquoise or bluish colorations (negative b* 15 parameter), especially in glaze A (free of Pb). This fact is indicative of much lower pigment stability in these glazes. Indeed, this bluish coloration is well-known in the literature to be associated to Co 2+ ions dissolved and incorporated mainly in tetrahedral environments of the glassy phase, the commonly named "cobalt leaching" or "cobalt bleeding" [4,6,9,10,35,[40][41][42][43].
In our compositions, the bluish hue of glazed samples was observed to increase with the nominal Co-content in the pigment, which would be therefore indicating a higher incorporation of Co 2+ within the glassy matrix.
Also noteworthy, codoping with Zn 2+ ions even at higher levels (Kar-CoZnH samples) was not observed to ameliorate significantly the stability of Co-karrooite pigments, which would be followed by a better preservation of the original green color of the pigments. Effectively, it is evident that glazed samples with Kar-CoZnH pigments exhibited rather similar turquoise (glaze B) or bluish (glaze C) colorations; however, it must be highlighted that in glaze C the bluish coloration was considerably reduced (having even a yellowish hue with positive b* value for x ≤ 0.1) with respect to Zn-free samples.
A complementary, more rigorous and precise confirmation of the relatively low stability of (Co,Zn)-karrooite pigments within the employed ceramic glazes (A, B and C) was also obtained by optical spectroscopy. The Kubelka-Munk (K/S) absorbance spectra corresponding to glazed samples with 1200ºC-fired pigments (Kar-Co, and Kar-CoZnH samples with a constant 20 mol% of Co-doping, x = 0.2) are shown in Figure 11. The spectra of the corresponding fired powders have been also included for comparison purposes. As an important remark, the obtained optical spectra of glazed samples may be interpreted as a the sum (convolution) of absorption bands arising from optical transitions associated to octahedral Co 2+ ions in karrooite pigment particles (remaining still undissolved in the ceramic glaze and imparting a greenish coloration) and those associated to tetrahedral Co 2+ ions dissolved and incorporated within the glaze (responsible for the bluish color).
Accordingly, the presence of stable karrooite particles dispersed in ceramic glazes may be followed in the spectra by the presence of the sharper, narrower and more intense absorption band around 550-620 nm (centered at 590 nm) associated to  3 transition of octahedral Co 2+ (named as  3 O in Figure 11 attributed to Jahn-Teller distortion of tetrahedral sites [44] and mainly to spin-orbit (L-S Russell-Saunders) coupling effects [43,45]. Then, the remaining  2 T and  1 T transitions of tetrahedral Co 2+ ions (to 4 T 1 ( 4 F) and 4 T 2 ( 4 F) excited terms, respectively) should appear as two broad and less intense bands extending in the near (around 1400 nm) and mid IR range (1600 nm) of the spectrum.
As it may be clearly observed in Figure 11 (Figure 11 below). Again, the typical absorption features of octahedral Co 2+ coming from karrooite pigment particles were more clearly preserved in glaze B (Ca-and Zn-rich), while in glaze C the absorbance spectrum was mostly dominated by the transition bands of tetrahedral Co 2+ ions incorporated in the glaze. Therefore, as it was previously stated, codoping with high amounts of Zn 2+ ions did not improve the stability of Co-karrooite pigments.
In summary, (Co,Zn)-karrooite green pigments proved to be more stable within the Ca-and Zn-enriched ceramic glaze (B) employed in this this study, and this result is in apparent contradiction with results obtained in previous investigations on the stability of Co-karrooite pigments [2]. Nevertheless, it must be highlighted that the complex oxide-based compositions of the ceramic glazes employed herein and in the previous investigations were not exactly the same.

Conclusions
In this study, the synthesis by conventional ceramic route of Co doped (Mg On the other hand, replacement of Mg 2+ by the larger Zn 2+ or Co 2+ ions in all the composition series was reflected in an anisotropic variation of cell parameters of karrooite lattice, with a general increase of cell volume. However, the microstructure and morphology of the grain-like aggregates forming the fired powders was not substantially affected by Co +2 and Zn +2 doping. Interestingly, the obtained (Co,Zn)-doped karrooite pigments exhibited a gamut of yellowish green colors, which became less yellowish (lower b* values) and more saturated or darker (lower L*) the higher the Co content. By UVvisNIR spectroscopy, the yellowish-green colors were associated with electronic transitions of Co 2+ located in the two non-equivalent M1 and M2 octahedral sites of karrooite pseudobrookite. Remarkably, codoping with Zn 2+ ions did not modify significantly the greenish component (similar -a* values) but diminished substantially the yellow hue (b*), and resulted also in lighter colors (higher L*). This color change was caused by the modification of crystal field exerted on Co 2+ ions and the lower covalence of Co-O bonds induced by Zn 2+ codoping. This was reflected in the optical spectra by a decreased absorption intensity (optical absorption density) associated with  3 ( 4 T 1g ( 4 F)  4 T 1g ( 4 P)) and also to Co-O charge transfer transitions.
Finally, (Co,Zn)-karrooite green pigments behaved rather as ceramic dyes, since glazed samples exhibited turquoise or blue colors associated with Co 2+ ions dissolved in tetrahedral environments of the glassy phase (the higher the Co content, the bluer the colors). Contrarily to previous reports, these pigments were more stable in a Ca-and Zn-enriched ceramic glaze (B) fired at a higher temperature (1050 ºC). This was evidenced by a better preservation of the green hue and optical absorption features of octahedral Co 2+ in glazed samples.    Counts (a.u.)