Available online at www.sciencedirect.com
Acta Materialia 58 (2010) 5057–5062 www.elsevier.com/locate/actamat
In situ synchrotron diﬀraction of the electrochemical reduction pathway of TiO2 q R. Bhagat a,*, D. Dye b, S.L. Raghunathan b, R.J. Talling b, D. Inman b, B.K. Jackson b, K.K. Rao c, R.J. Dashwood a a
University of Warwick, Coventry CV4 7AL, UK b Imperial College, London SW7 2AZ, UK c Metalysis, Rotherham S63 5DB, UK
Received 2 November 2009; received in revised form 17 May 2010; accepted 20 May 2010 Available online 16 June 2010
Abstract Despite over ten years of work into the low-cost electrowinning of Ti direct from the oxide, the reduction sequence of TiO2 pellets in molten CaCl2 has been the subject of debate, particularly as the reduction pathway has been inferred from ex situ studies. Here, for the ﬁrst time white beam synchrotron X-ray diﬀraction is used to characterize the phases that form in situ during reduction and with 100 lm resolution. It is found that TiO2 becomes sub-stoichiometric very early in reduction, facilitating the ionic conduction of O ions, that CaTiO3 persists to nearly the end of the process and that, ﬁnally, CaO forms just before completion of the process. The method is quite generally applicable to the in situ study of industrial chemical processes. Implications for the industrial scale-up of this method for the low-cost production of Ti are drawn. Ó 2010 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. Keywords: Electrochemistry; Titanium; Synchrotron radiation; XRD; Phase transformation kinetics
1. Introduction In the FFC Cambridge process, metal oxides are electrochemically reduced to metal using a molten chloride ﬂux . Currently the Kroll process is used to produce Ti from rutile ; potentially, the FFC process may lead to a step change in the cost of extraction of this and other alloy systems. The process involves the progressive reduction and deoxidation of a porous TiO2 pellet cathode in a molten halide salt . At the cathode TiO2 is reduced to Ti. The oxide ions dissolve into CaCl2 and then migrate to a C anode, forming
CO2 and CO. This reduction pathway has been studied using pellet reductions , metal-cavity electrode [4–6] and thinﬁlm electrode experimentation [7,8]. Using cyclic voltammetry and X-ray diﬀraction (XRD) analysis of thin ﬁlms after 0 each event, the electrochemical events C4, C3, C2 , C2 and C1, corresponding to reactions (1)–(6), respectively, have been identiﬁed during reduction. Despite being observed in XRD, the formation of CaTiO3 was not observed on the voltammograms and was therefore believed to form via a chemical reaction . 2TiO2 þ 2e ! Ti3 O5 þ O2
R.J.D. and D.D. conceived and designed the study. B.K.J., R.J.D. and D.D. designed the apparatus, which B.K.J. built and commissioned. R.B., R.J.D., D.D., S.L.R., R.J.T. and K.K.R. conducted the experiments. R.B. and S.L.R. analysed the data with support from R.J.T., R.J.D. and D.D. The paper was written by R.B. with contributions and editing from R.J.D., D.D., D.I., R.J.T. and S.L.R. * Corresponding author. E-mail address: [email protected]
2Ti3 O5 þ 2e ! 3Ti2 O3 þ O 2þ
CaTiO3 þ 2e ! Ca
þ TiO þ 2O 2
Ti2 O3 þ 2e ! 2TiO þ O
TiO þ 2e ! 2TiO þ O
TiO2 þ Ca
1359-6454/$36.00 Ó 2010 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.actamat.2010.05.041
R. Bhagat et al. / Acta Materialia 58 (2010) 5057–5062
The reduction pathway of TiO2 can be rationalized using an electrochemical predominance diagram  (Fig. 1a), which indicates the thermodynamically stable species at different potential/oxide activities. The electrolyte oxide content after pre-electrolysis is found to be 3–4 pO2 (log(aO2) . Fig. 1a shows that the evolution of phases at a particular electrode potential is dependent on the electrolyte oxide content. For example, reducing TiO2 in an oxide-saturated electrolyte (pO2 = 0) will lead to a chemical reaction between CaO and TiO2, forming CaTiO3, which would subsequently reduce to Ti. If the electrolyte contained lower levels of oxide (pO2 = 6), then TiO2 would reduce through the lower Ti oxides. Thus, to achieve a consistent reduction route and product, the fundamentals of the process must be understood. In previous FFC papers phase identiﬁcation was performed ex situ, once the samples were cooled and washed. This information was then correlated to in situ voltammetric or amperometric data. However, upon cooling any number of crystallographic changes or chemical reactions may occur. Further, washing may remove water-soluble species. This has resulted in conﬂicting reduction pathways being proposed. By performing in situ synchrotron XRD,
Fig. 1. (a) Electrochemical predominance diagram for Ca–Ti–O–Cl at 1173 K. (b) Phase evolution near the pellet surface (1.2 mm from the centre).
whilst reducing TiO2, here we are able to present the ﬁrst unequivocal description of the reduction pathway using the crystallographic information gathered (Fig. 1b). 2. Method TiO2 precursors were formed using reagent-grade TiO2 (99.5% Alfa Aesar), ball milled with ethanol and zirconia spheres for 24 h. The slurry was dried and the powder ground in a mortar and pestle with a small quantity of distilled water, which acted as a binding agent. 0.4 g of powder was uniaxially compacted at 100 MPa in a 13 mm diameter die. The precursors were then drilled to accept the commercial-purity (CP) Ti current collector and sintered in Ar (BOC pureshield – 150 mL min1 at 1373 K for 3 h at a rate of 3 K min 1. The pellets were then ground to approximately 7 mm and 0.2 g. The electrochemical cell was designed to minimize attenuation of the white X-ray beam and avoid additional sources of diﬀraction. Reductions were performed in a resistance furnace-heated quartz glass vessel (Fig. 2), as in Ref. . Openings were incorporated into the furnace for the synchrotron X-ray beam. A glassy C crucible was used to contain the salt and act as a consumable anode. CaCl2 granules (Fluka) were dried and prepared  within the C crucibles, which were then vacuum sealed prior to use. The sintered TiO2 precursor was attached to the Ti current collector and suspended above the salt prior to preelectrolysis. The apparatus was heated at a rate of
Fig. 2. Electrochemical reduction cell used, showing the sampling volumes employed.
1 K min1 under Ar (BOC pureshield – 50 mL min1) to the electrolysis temperature (1173 K). The electrolyte was then thermally equilibrated for 3 h. Prior to experiments the salt was pre-electrolysed for 1–3 h to remove electroactive impurities by polarizing a grade 2 CP-Ti rod (3 mm) vs. the graphite crucible. Following the pre-electrolysis, the precursor was lowered into the molten salt and polarized at a rate of 0.5 mV s1 from open circuit to 3100 mV vs. the C anode. The ID15A beamline at the ESRF was used to produce an unmonochromated (white) X-ray beam in energy dispersive mode. A 50 lm 100 lm incident beam and 5° scattering angle were deﬁned by slits yielding a gauge length of 1000 lm (Fig. 2). The assembly was mounted on a translation stage, allowing nine locations at diﬀerent heights in the pellet to be studied. Acquisition times were 60 s per pattern. The diﬀraction patterns were analysed by Rietveld reﬁnement using GSAS , allowing phases to be identiﬁed and atomic fractions obtained. The source spectrum was normalized using both an Al powder and the powder TiO2 pattern obtained from the pellet prior to insertion into the cell. Beam hardening within the cell was accounted for by comparison with the initial patterns obtained from the pellet. 3. Results and discussion The scans at the pellet surface (1.2 mm) are shown in Fig. 1b. Rietveld reﬁnements at key intervals during reduction for position 1.2 mm are shown in Fig. 3. In some cases, due to the small gauge volume, insuﬃcient grains were sampled to produce a powder pattern, resulting in some disparity between observed and calculated peak intensities. Molar fractions of the phases present at each position are shown in Fig. 4 (e.g. after 4 h at the 1.2 mm position, approximately 55% CaTiO3 and 45% TiO2 is present; after 10 h approximately 75% Ti, 20% CaTiO3 and 5% TiO is present). The reﬁned lattice parameters can be used to estimate the interstitial O content in hexagonal close-packed (hcp) a-Ti. The thermal expansion data for the lattice parameters , from room temperature to 700 °C, were combined with the room temperature dependence of O content on the lattice parameter . The data used is shown in Table 1 and the combined thermal and compositional eﬀects are given in Eqs. (7) and (8). The hexagonal a lattice parameter is largely independent of O content, so only the c parameter data was analysed. It was assumed that the Ti that ﬁrst forms was saturated with O (14 wt.% O at 900 °C). The variation in c lattice parameter and inferred interstitial O content in the a-Ti phase are shown in Fig. 5a. Given the previously discussed assumption made, the inferred interstitial oxygen content (shown in Fig. 5a) will have an uncertainty of ±0.1 at.% O arising from the data showing the dependence of O content on the lattice parameter . As such, error bars are not included in Fig. 5a as they would not be visible on the scale
Intensity (arb. units)
R. Bhagat et al. / Acta Materialia 58 (2010) 5057–5062
CaTiO TiO Ti O Ti O
α-Ti β-TiO CaTiO
CaO α-Ti β-TiO CaTiO
18 h Fit Data
d-spacing (Å) Fig. 3. Diﬀraction spectra and Rietveld reﬁnements (with key peaks) obtained near the pellet surface (1.2 mm from the centre).
R. Bhagat et al. / Acta Materialia 58 (2010) 5057–5062
Table 1 a-Ti single-crystal thermal expansion coeﬃcients and O-dependence of the lattice parameters. ˚) a0 (A 2.9511  ˚) c0 (A 4.6826 
Ti CaO a
aa 106 (K1) ac 106 (K1) ˚ at.% O1) da/dC(O) 104 (A ˚ at.% O1) dc/dC(O) 104 (A
1.2 2 mm
CaO O TiO
of the diagram as presented. By combining this data with the phase fraction information, the total O content at diﬀerent locations has been estimated (Fig. 5b). For this diagram it is assumed that phases other then a-Ti are stoichiometric; however, given the low oxygen environment within the apparatus and the reducing conditions, the phases present will probably be sub-stoichiometric (e.g TiO2, Ti4O7 Ti2O3 and TiO may have approximately 1.0, 2.0, 0.2 and 5.5 at.% less oxygen than assumed , respectively). Assuming each phase has the most reduced stoichiometry, the data points in Fig. 5b may be overestimated by up to 0.5–3.0 at.% O.
Ti O a CaO
11.03  13.37  7.8459  47.112 
CaTi O CaO
0.6 TiO T
0.6 CaO O
0.4 TiO T
Fig. 4. Rietveld-reﬁned phase evolution throughout the pellet. Centre 0 mm to near-surface, 1.2 mm.
Fig. 5. (a) Evolution of the a-Ti c-lattice parameter and inferred interstitial O content and (b) inferred oxygen content of phases present during reduction through the pellet.
R. Bhagat et al. / Acta Materialia 58 (2010) 5057–5062
da dCðOÞ dc ¼ c0 ð1 þ DT aa Þ þ CðOÞ dCðOÞ
aaTi ¼ a0 ð1 þ DT aa Þ þ CðOÞ
At the beginning of the reduction, TiO2, Ti4O7 and CaTiO3 were present at all locations (Fig. 4). The presence of Ti4O7 is consistent with the general electronic theory of TiO2 [15–17], which states that TiO2 will equilibriate in a low oxygen atmosphere. This is achieved via the creation of crystal defects, resulting in TiO2 being reduced. It should be noted that in addition to Ti4O7, other Magne`li phases (e.g Ti5O9) ﬁt relatively well. Ti4O7 is the Magne`li phase most often detected during the FFC reduction of TiO2 [3,18]. The transformation to substoichiometric TiO2 phases and the consequent release of O ions into the electrolyte results in the formation of CaTiO3. It has been suggested that the slight reduction of TiO2 facilitates the transformation to the CaTiO3 perovskite structure . Interestingly Ti3O5 was not detected during reduction, despite its formation being predicted by Fig. 1 and by the Ti–O phase diagram . Signiﬁcant quantities were observed previously [3,7] during ex situ analysis of thin ﬁlms. The narrow homogeneity range of Ti3O5 (Fig. 1)  indicates that signiﬁcant quantities would only form if the electrode potential were held within the Ti3O5 phase ﬁeld. It is hypothesized that the signiﬁcant quantities of Ti3O5 observed  were a result of a disproportionation reaction of Magne`li phases during cooling. Consequently, either the Ti3O5 phase ﬁeld is traversed rapidly or kinetic considerations prevent its formation. Ti2O3 seems to have a signiﬁcant impact on reduction as it is present throughout the sample. Fig. 2 shows that Ti2O3 reduces to b-TiO. The monoclinic variants, a-TiO and Ti3O2, were not detected, which is inconsistent with the Ti–O phase diagram . The b-TiO then gradually reduces to a-Ti. CaTiO3 and CaTi2O4 (in the pellet inner regions) were present for most of the reduction. It is believed that the oxygen content of CaTiO3 does not signiﬁcantly change until the point of reduction to TiO or Ti. CaTiO3 is reported to have a compositional range ; however, the derived lattice parameters remain quite consistent throughout reduction. The detected CaTiO3 is most likely non-stoichiometric, i.e Ca/Ti ratio – 1 and oxygen sub-stoichiometric (thus CaTiO3 could be more accurately referred to as CaxTiyO3z, where x/y – 1 and z is related to the number of oxygen vacancies present). It should be noted that between the 1.1 h and 3.3 h pattern in Fig. 3, small shifts in the three main CaTiO3 peaks to lower d-spacing values can be observed. Such changes suggest limited removal of oxygen ions from the CaTiO3, which distort the crystal structure. Such ﬁndings are consistent with the work of Bak et al. , who observed the removal of oxygen ions from CaTiO3 to create doubly ionized oxygen vacancies. We previously speculated that CaTiO3 may act as a solid electrolyte, allowing the transmission of oxide ions , which may account for the
consistent stoichiometry. At the surface layers it would appear that CaTiO3 is reduced directly to Ti despite the fact that thermodynamically this would only occur at very low pO2. In the inner layers, CaTi2O4 forms at the expense of CaTiO3 and TiO. Previously, it was theorized that the formation of CaTi2O4 was a result of either the electrochemical reduction of CaTiO3 or a comproportionation reaction between CaTiO3 and TiO. Fig. 4 shows that the formation of CaTi2O4 results in the equimolar consumption of TiO and CaTiO3, indicating that a comproportionation reaction is indeed responsible for its formation (Reaction (9)) [3,20]. CaTiO3 þ TiO ! CaTi2 O4
CaTi2O4 has been produced by mixing equimolar or Ti-rich mixtures of CaTiO3 and Ti powder with a CaCl2 ﬂux at 1000 °C . When the Ti powder was oxidized to TiO, CaTi2O4 was not formed, suggesting the comproportionation reaction has a small positive DG . Fig. 4 shows that the formation of CaTi2O4 at the inner layers coincides with the occurrence of Ti in the outer layers (0.6, 0.9 and 1.2 mm). It is speculated that the presence of titanium in close proximity, during reduction, makes the comproportionation between TiO and CaTiO3 more favourable. Fig. 5 shows that signiﬁcant removal of O occurs at events 1, 2 and 3. Event 1 can be identiﬁed as the formation of TiO (Fig. 4). For the surface layers (1.2, 0.9 and 0.6 mm), event 2 was identiﬁed as the reduction of TiO to Ti. For the inner layers (0.3 and 0 mm), event 2 is retarded due to the formation of CaTi2O4. Event 3 involves the removal of CaO, which occurs at all locations except the pellet centre (0 mm). Following the dissolution of CaO, the removal of interstitial O from Ti commences. CaO is detected throughout the pellet at the later stages of reduction and forms when the electrolyte in the pores becomes saturated with oxide ions (22 mol.%). The onset of CaO formation correlates with the reduction of CaTi2O4 in the core. This results in a rapid evolution of oxide ions that have to be transported through the pellet pores to the bulk electrolyte. During reduction, the porosity of the pellet (originally 60% open porosity) is reduced through a combination of sintering of Ti and formation of titanates. This signiﬁcantly reduces the pellet permeability, allowing oxide ion saturation to occur. After event 2 at the inner layers the ﬂux of oxide ions passing through the pellet diminishes. At this point the surface CaO rapidly dissolves, closely followed by the core. We have proposed two mechanisms for the surface formation of TiC : (i) A chemical reaction with C dust ﬂoating on the electrolyte surface, occurring as the sample is removed. (ii) Electrochemical deposition of C and subsequent reaction at the cathodic Ti pellet surface. This occurs via the chemical formation of CO2 3 via a reaction between CO2 formed at the anode and O2 in the electrolyte (Reaction (10)). This complex is then reduced at the cathode forming CaO and C, which chemically reacts with a-Ti to form TiC (Reaction (11)). This is supported by the present in situ data.
R. Bhagat et al. / Acta Materialia 58 (2010) 5057–5062
CO2 þ O2 ! CO2 3 CO2 3
þ Ti þ 4e ! 3O
TiC formation is restricted to the pellet surface and only forms during the later stages of reduction. It is proposed that TiC only forms once the cathode is suﬃciently negatively polarized for Reaction (11) to occur. The potential at which this occurs is dependent on the activity of both the CO2 3 and more importantly the oxide ions. The oxide ion concentration in the region of the pellet will only reduce once reduction is nearly complete, meaning Reaction (11) becomes more favourable. There is an increase in c lattice parameter (Fig. 5) at the later stages of reduction at positions 0.6, 0.9, 1.2 mm, coinciding with the formation of TiC. It is suggested that this eﬀect is due to the solubility of interstitial C in a-Ti (approximately 0.5 at.% at 1373 K), leading to an increase in the c parameter. Once the solubility of interstitial C is exceeded, the formation of TiC occurs.
the likelihood of oxide saturation and CaO formation. Alternatively, engineering of the preform to maximize initial pore size and permeability could reduce titanate formation. The formation of TiC appears to be unavoidable with the current cell design. Its formation can only be prevented if C-free anodes are employed or a membrane used. Acknowledgements The authors gratefully acknowledge the Defence Advanced Research Project Agency (DARPA), the Oﬃce of Naval Research and the Engineering and Physical Sciences Research Council (EPSRC) for their support. The authors also thank T. Buslaps, J. Daniels and A. Steuer at the ESRF, together with M. Jackson for their assistance. The results described in this paper were collected in conjunction with data for the reduction of NiTiO3. This work is described elsewhere . References
4. Conclusions The key ﬁndings of this work can be summarized as follows. Upon entering the low O atmosphere, TiO2 quickly forms sub-stoichiometric phases, which signiﬁcantly improves ionic conductivity over stoichiometric TiO2. In contradiction to previous work , CaTiO3 forms chemically from the sub-stoichiometric TiO2 phases. CaTi2O4 formed from the comproportionation of TiO and CaTiO3 and was detrimental to the reduction process. The phases Ti3O5, a-TiO and Ti3O2 were not observed despite being predicted by the Ti–O phase diagram. The formation and eventual dissolution of CaO was related to the oxide ﬂux within the pellet. The deoxidation of Ti was found to occur only once CaO had been removed locally within the pellet. TiC formed on the reduced pellet surface through an in situ process, involving the electrochemical deposition of C, but only seemed to occur when the O content in Ti reached approximately 16 at.%. The relative ease with which low electrical conductivity TiO2 can be reduced via this process can be attributed to the low O2 partial pressure. This results in a nearly instantaneous formation of conductive Magne`li phases. The time required to reduce TiO2 pellets appears to be controlled by the formation of Ca compounds (CaTiO3, CaTi2O4 and CaO). Their formation reduces the permeability of the pores, eﬀectively causing passivation. Consequently, understanding the mechanism by which these phases form and how to prevent their formation is key to improving the eﬃciency of the process. The chemical formation of CaTiO3, and the subsequent formation of CaTi2O4 could be avoided if lower valence metal oxides (i.e. Ti2O3 and TiO) were used as the starting product. Preventing the formation of the titanates will improve permeability and hence decrease
   
                  
Chen G, Fray DJ, Farthing TW. Nature 2000;407:361. Kroll W. Seventy-Eighth General Meeting 1940:35. Schwandt C, Fray D. J Electrochim Acta 2005;51(1):66. Rao K, Brett D, Inman D, Dashwood RJ. Investigation of the reduction electrochemistry of titanium oxides in molten calcium chloride using cavity electrode. In Niinomi M, et al., editors. Titanium-2007 Science and Technology. Sendai, Japan: The Japan Institute of Metals; 2007. p. 3. Cha CS, Li CM, Yang HX, Liu PF. J Electroanal Chem 1994;368:47. Qiu G, Ma M, Wang D, Jin X, Hu X, Chen GZ. J Electrochem Soc 2005;152:E328. Dring K, Dashwood RJ, Inman D. J Electrochem Soc 2005;152(3):E104. Chen G, Fray DJ. J Electrochem Soc 2002;149:E455. Dring K, Dashwood RJ, Inman D. J Electrochem Soc 2005;152(10):D184. Bhagat R. PhD Thesis. London: Imperial College; 2008. Larson AC, Von Dreele RB. Los Alamos National Laboratory Report; 2000. Berry RLP, Raynor GV. Research 1953;6:21S. David D, Garcia EA, Lucas X, Beranger G. J Less-Common Met 1979;65:51. Murray JL, Wriedt HA. Bull Alloy Phase Diagr 1987;8:148. Bak T, Nowotny J, Rekas M, Sorrell CC. J Phys Chem Solids 2003;64(7):1057. Nowotny J, Radecka M, Rekas M. J Phys Chem Solids 1997;58(6):927. Nowotny J, Radecka M, Rekas M, Sugihara S, Vance ER, Weppner W. Ceram Int 1998;24(8):571. Alexander DTL, Schwandt C, Fray DJ. Acta Mater 2006;54(11):2933. Jiang K, Hu X, Ma M, Wang D, Qiu G, Jin X, et al. Angew Chem Int Ed 2006;45(3):428. Bhagat R, Jackson M, Inman D, Dashwood RJ. J Electrochem Soc 2009;156(1):E1. Bak T, Nowotny J, Sorrell CC. J Phys Chem Solids 2004;65:1229. Rogge MP, Caldwell JH, Ingram DR, Green CE, Geselbracht MJ, Siegrist T. J Solution Chem 1998;141:338. Jackson BK, Dye D, Inman D, Bhagat R, Talling RJ, Raghunathan SL, et al. J Electrochem Soc 2010;157(4):E57.