Study of the order-disorder transition in yttria-stabilised zirconia by neutron diffraction Iain R. Gibson" and John T. S. Irvineb "Departmentof Chemistry, University of Aberdeen, Meston Walk, Aberdeen, UK AB9 2UE bSchool of Chemistry, University of St. Andrews, St. Andrews, Fife, UK K Y16 9ST A comprehensive study of 8 mol% yttria-stabilised zirconia has been made between 150 and 1000 "C, using ac impedance spectroscopy and high-temperature neutron powder diffraction. It has been demonstrated that the conductivity anomaly, which occurs at ca. 650 "C, is structural in origin. A sharp decrease in the activation energy for conduction of ca. 0.2 eV was observed at ca. 650 "C.Additional broad, diffuse scattering peaks were observed below 600 "C in the neutron diffraction patterns; above 650 "C, the diffuse scattering peaks disappeared. A deviation from linearity was observed at a similar temperature in the plots of both Y/Zr and 0 isotropic temperature factors us.temperature. The low-temperature behaviour can be explained in terms of ordering of oxygen vacancy-(dopant) cation clusters to form microdomains, which are evidenced by the presence of diffuse scattering peaks. At high temperature, the association of vacancies with defects breaks down, or at least becomes randomised, allowing vacancies to move more freely as indicated by the decrease in activation energy for conduction. A discontinuity in thermal expansion coefficient (from neutron diffraction data) confirms the second-order nature of the transition. Cubic yttria-stabilised zirconia (YSZ) exhibits high oxide-ion conductivity at 1000°C and is used as the electrolyte in solid oxide fuel cell (SOFC) devices.It is generally agreed that the maximum conductivity occurs with an yttria content of around 8 mol%.',2 Increasing the amount of yttria dopant above 8 mol% produces a decrease in conductivity, which is thought to arise from an increase in oxygen vacancy-(dopant) cation associations, reducing the number of 'free' oxygen vacancies available to migrate.3 From conductivity measurements of 8 mol% YSZ between 600 and lOOO"C, a linear Arrhenius plot (log aTvs. 1/T) can be obtained and the activation energy for conduction calcu- lated, although curvature has been observed in Arrhenius plots at lower temperature^.^ In a more recent study, Badwal observed a change in the slope of the Arrhenius plot at approximately 550 OC,' indicating a higher activation energy at low temperatures.We generally observe similar behaviour, e.g. Fig. 1. An increase in activation energy at low temperatures is in accord with the behaviour expected for some form of vacancy- dopant cation association. A number of such models for oxide- ion conduction have been developed to account for results from studies of ceria and perovskite Kilner and Steele8 stated that the low-temperature activation energy com- prised an oxygen vacancy migration enthalpy (AmH),and an association enthalpy (AaH),due to the vacancy4opant com- 2 ---a 1 -. c .2 o--z .E -1 -2 *\ '*, plex.At higher temperatures, the vacancy-dopant complexes dissociate, allowing oxygen vacancies to migrate freely. The activation energy now contains only the migration enthalpy term, AmH.Atomistic calculations predict that the additional association enthalpy, equal to half the binding energy, is of the order of 0.1-0.2eV,' in accord with our recent studies, Table 1." As the amount of dopant increases, the additional association enthalpy at low temperatures also increases (Table 1) indicating an increase in the degree of association. The nature of the vacancy-dopant interaction in YSZ at low temperatures has been reviewed extensively." Although it is widely considered that oxygen vacancies are associated with the aliovalent dopant cation, Y3+, an EXAFS study of 10 mol% YSZ by Catlow et al.indicated that oxygen vacancies were preferentially sited adjacent to the Zr4 cation, and not + the dopant Y3+ cation.12 This results in seven-fold oxygen coordination of Zr4+, similar to its structural environment in monoclinic zirconia; Y3+ ions are in sites of eight-fold coordi- nation. Similar results were obtained in a series of EXAFS studies by Li et a1.13-1s Many structural studies have been made on cubic stabilised zirconias. Room-temperature neutron diffraction experiments on single crystals of YSZ by Steele and Fender,16 and CSZ (calcia-stabilised zirconia) by Moringa et and Cohen et ~1.'~concluded that oxygen atoms were displaced along the (100) direction.In contrast, Carter and Roth" and Horiuchi et studying CSZ and YSZ respectively, reported a dis- placement along the (1 11) direction. Diffuse scattering, or a modulated diffuse background, was a significant feature of the Table 1 Activation energies and association enthalpy of 3, 8-11 molo/o yttiria-stabilised zirconia (in eV) mol% yttria EA(300-500 "C) E, (600-800 "C) A,H 3" 0.94 0.89 0.05 8 1.08 0.92 0.16 9 1.11 0.93 0.18 10 1.14 0.95 0.19 11 1.18 0.98 0.20 " 3 mol% yttria-stabilised zirconia is single-phase tetragonal, the other compositions were single-phase cubic (4-7 mol% is a two-phase mixture). J. Muter. Chem., 1996, 6(S), 895-898 895 neutron diffraction pattern obtained by Steele and Fender,16 and possible short-range ordering by vacancy association was considered.Further work on YSZ and CSZ18*21-25 supports the concept of ordering of vacancy-cation complexes, or clusters, visible as diffuse scattering. A detailed study on the effect of high temperatures on the diffuse background scattering has not been made. High-temperature neutron diffraction experiments on YSZZ6 and CSZ27 have recently been reported. Both studies examined the change in temperature factors of the cations and anions with increasing temperature; however, only a few data points were obtained in the temperature region 300-10OO0C, where the deviation in the activation energy of conduction is observed. Martin et studying CSZ, observed a change in the increase in B,,,(oxygen) at ca.lOOO"C, which is close to the order-disorder tran~formation.'~ Proffen et again studying CSZ, observed a decrease in the diffuse scat- tering as this temperature was approached. Our conductivity data, and the neutron diffraction work on CSZ and YSZ previously discussed, suggest that a similar order-disorder transition may be observed in YSZ at ca. 650°C. Here, we make a detailed study of 8 mol% YSZ using time-of-flight neutron diffraction over the temperature region 25-1000 "C. The effect of increasing temperature on the change in diffuse scattering, isotropic temperature factor, and lattice parameter is discussed. Experimental The powder used in this study, TZ-8Y, was produced by Tosoh Corporation (Japan).For conductivity measurements, the powder was uniaxially pressed at 80 MPa in a 13 mm die. Samples were heated at 10°C min-', sintered at 1500°C for 2 h, and cooled at 10 "C min-l. Platinum electrodes were applied to both faces of the sample. Two terminal ac impedance measurements were performed using a HP4192A impedance analyser, over the frequency range 100 Hz-13 MHz. Measurements were made in air, over the temperature range 300-1000 "C. For neutron diffraction experiments, the powder was sintered at 1500 "C for 20 h. Powder neutron diffraction data were collected on the Polaris diffractometer at the UK spallation neutron source ISIS, Rutherford Appleton Laboratory. Samples were contained in vanadium cans, and a vanadium wound furnace was used between 25 and 1000°C.A thermo- couple was attached next to the sample can, ensuring accurate temperature control of f1 "C. The crystal structures were refined by the Rietveld method with the program TF14LS29730 using data collected over the time-of-flight range 2500-19500 ps. Refinements were carried out using a similar approach to that of Argyriou.26 No attempt was made to model any displacement of the anions in either the (1 11) or the (100) direction. All refinements were carried out in the space group Fm3m, with the occupancy and positions of Zr/Y and 0 fixed (Table 2). Background, lattice parameters and isotropic temperature factors for Zr/Y and 0 were refined for all temperatures. Table 2 Initial parameters for 8 mol% YSZ ZrV) 0 X 0.0 0.25 Y 0.0 0.25 Z 0.0 0.25 occupancy 0.8515 (0.1485) 0.9634 scattering lengths" 0.716 (0.775) 0.5805 " Ref.3 1. 896 J. Muter. Chem., 1996, 6(5), 895-898 Results and Discussion A clear change in the slope of the Arrhenius plot for conduction is observed at ca. 650°C (Fig. 1). Below this temperature the activation energy is higher (1.07 eV), and as the temperature is increased to above 650°C, the activation energy decreases (0.92eV). The transition between the two regions is quite sharp, certainly for an order-disorder process in an ionic conductor, although there is a limited region of slight curvature joining the two linear portions extending over less than 100 "C. Examples of observed and calculated neutron diffraction profiles and their difference curves, at room temperature and lOOO"C, are given in Fig.2. At room temperature, a heavily modulated diffuse background is visible in the difference curve. At 1000 "C, the background appears virtually flat, and the only features in the difference curve greater than 2 esd are associated with 'allowed reflections'. The diffuse scattering peaks observed at room temperature do not correspond to expected reflections from the space group Fm3m, although they can approximately be indexed on a primitive cubic unit cell with the same unit- cell edge. Previous reports indicate that this is due to some type of short-range ordering within the cubic lattice.16 As the temperature was increased, the diffuse background decreased, and above 600 "C it had effectively disappeared (Fig.3). The temperature region where the diffuse peaks are observed corresponds to the region where the activation energy for conduction is larger. This suggests that above 650"C, when the activation energy could be interpreted as arising solely from the enthalpy of migrati~n,~ the cubic lattice no longer shows signs of local ordering. A decrease in the weighted profile residual, R,,, with increasing temperature is another indication of the change occurring at ca. 650°C. It is normal to see a small increase in R,, with increasing temperature as the background noise increases; however, in Table 3, R,, is shown to decrease with increasing temperature, indicating an " 0.5 1 1.5 2 2.5 3 d-spacing/A $400 1 0.5 1 1.5 d-spacingA 2.52 3 4 -10 Fig. 2 Observed and calculated neutron diffraction profiles for 8 mol% yttria-stabilised zirconia at 20 "C (a) and 1000 "C(b) Table 3 Zr/Y and 0 istropic temperature factors (ITF) and R,, T/"C Zr/Y ITF 0 ITF RWP ~~ ~ 150 0.71(1) 2.19( 2) 3.83 225 0.76( 1) 2.28(2) 3.72 300 0.82( 1) 2.37( 2) 3.48 400 0.89( 1) 2.49(2) 3.23 500 0.96( 1) 2.60(2) 2.98 600 1.04( 1) 2.72(2) 2.81 700 1.12( 1) 2.84(2) 2.61 800 1.20(1) 2.98(2) 2.41 900 1.29(1) 3.14(2) 2.22 1000 1.38(1) 3.30(2) 2.24 707-----7 y60 t3850 Cg 40 30 20 11 I b.75 0.80 0.85 0.90 0.95 d-spacinglA 1:OO 1:05 Fig.3 Comparison of neutron diffraction profiles at 150, 300, 500, 700 and 900 "C, locations of low-temperature diffuse peaks are arrowed 3.50 3 00 2.50 2.00 1SO 1.oo LLk 1.40 1.20 1.oo 0.80 060 040 I 0 200 400 600 800 loo0 1200 1400 TIK Fig.4 Isotropic temperature factors (ITF) us. temperature in Kelvin for 0 (a) and Y/Zr (b).Guidelines show fitting to high temperature and low temperature regimes, intercepts with T=O K indicate static contributions to ITFs. improvement in the refinement due to the disappearance of the diffuse scattering. The change in isotropic temperature factor (ITF) of Zr/Y and 0 with temperature is shown in Fig. 4, and listed in Table3. The ITF for Zr/Y and 0 both appear to increase linearly between 150 and 600°C. Above 600°C a change in slope occurs, with a greater increase in ITF with temperature.The temperature at which this deviation is observed also 5.19 5.17 5 rp 5.15 5.13 I 0 200 400 600 800 lo00 T/"C Fig. 5 Unit cell parameter, a, us. temperature corresponds to the temperature at which the activation energy for conduction changes. The observed linear dependence of the temperature factor upon temperature is in accord with a simple Debye-type model, although strictly this model is only applicable to monatomic solids with the atoms centred on their ideal crystallographic sites. An approximation developed by Martin et for YSZ is to split the temperature factor into two components, giving a static term which is independent of temperature, and a temperature-dependent term which varies in accord with the Debye approximation [eqn.(l)]. BkexP =gkstatic + thermal (1) The plots in Fig. 4 show how the observed dependences of the temperature factor can be interpreted using this model. The behaviour of each atomic sublattice shows a low-tempera- ture and a high-temperature region; extrapolated values of the low- and high-temperature static values are presented in Fig. 4. For both atomic sublattices at low temperatures the static contribution is higher; however, the rates of change of the thermal contribution with temperature are lower at low tem- perature. This behaviour is consistent with microdomains containing ordered arrays of distorted subcells at low tempera- tures, with disordering of the microdomains at ca.650 "C. A plot of lattice parameter against temperature also shows a deviation at 600-700 "C (Fig. 5). Consequently, this deviation can be considered as a discontinuity in the thermal expansion coefficient, which is indicative of a second-order transition. Conclusions The change in activation energy of conduction with increasing temperature for 8 mol% yttria-stabilised zirconia was con-firmed as occurring at ca. 650 "C. Using high-temperature neutron diffraction, a modulated background was observed in all diffraction patterns below 650°C. Above 650°C, the diffuse scattering decreased significantly, indicating that short-range ordering only occurs at low temperatures. The deviation in both cation and anion isotropic temperature factors, again between 600 and 700 "C, suggested a larger static contribution to the ITFs at low temperatures, due to the displacement of ions in locally ordered domains.In the disordered regime at higher temperatures, the static contribution to the ITFs was much lower owing to randomisation or effective loss of local distortion. A deviation in thermal expansion coefficient between 600 and 700°C suggested a transition which was second order in nature. We gratefully acknowledge Tioxide Specialties plc for funding a research studentship and the CLRC for the award of neutron diffraction beam time at the Rutherford Appleton ISIS facility. J. Mater. Chem., 1996, 6(5),895-898 897 References 19 R E Carter and W L Roth, in Proc Electromotive Force Measurements in High Temperature Systems, ed C B Alcock, S P S Badwal, J Muter Sci, 1985,20,4593 S P S Badwal, Solid State Ionics, 1992,52,23 20 London Institution of Mining and Metallurgy, 1963, pp 125-144 H Horiuchi, A J Schultz, P C W Leung and J M Williams, D K Honke, Solid State Ionics, 1981,5, 531 J E Baurele and J Hrizo, J Phys Chem Solids, 1969,30, 565 21 Acta Crystallogr Sect B, 1984,40, 367 B Hudson and P T Moseley, J Solid State Chem , 1976,19,383 A S Nowick and D S Park, in Superionic Conductors, ed G Mahan and W Roth, Plenum Press, NY, 1976, pp 395-412 22 N H Andersen, K Claussen, M A Hackett, W Hayes, M T Hutchings, J E MacDonald and R Osborn, in Proc 6th 6 A S Nowick, D Y Wang, D S Park and J Griffith, in Fast Ion Ris0 Int Symp on Metallurgy and Materials Science, ed 7 Transport in Solids, ed P Vashishta, J N Mundy and G K Shenoy, North Holland, Amsterdam, 1979, pp 673-679 J A Kilner and C D Waters, Solid State Ionics, 1982,6,253 F W Poulsen, N Hessel Andersen, K Claussen, S Skaarup and 0 Toft Sarrensen, London Institution of Mining and Metallurgy, 1985, pp 279-284 8 J A Kilner and B C H Steele, in Non-stoichiometric Oxides, ed 23 R Osborn, N H Anderson, K Claussen, M A Hackett, 0 T Sarrensen, Academic Press, NY, 1981, pp 233-269 W Hayes, M T Hutchings and J E MacDonald, Muter Sci 9 C R A Catlow, Solid State Ionics, 1984, 12,67 Forum, 1985,7,55 10 I R Gibson and J T S Irvine, Solid State Ionzcs, submitted 24 S Hull, T W D Farley, M A Hackett, W Hayes, R Osborn, 11 B C H Steele, in High Conductivity Solid Ionic Conductors, ed N H Andersen, K Claussen, M T Hutchings and W G Stirling, T Takahashi, World Scientific, Singapore, 1989, pp 402-446 Solid State Ionics, 1988,28-30,488 12 C R A Catlow, A V Chadwick, C N Greaves and L M 25 R B Neder, F Frey and H Schulz, Acta Crystallogr, Sect A, Moroney, J Am Ceram Soc , 1986,69,272 1990,46,799 13 P Li, I-W Chen and J E Penner-Hahn, Phys Rev B, 1993, 26 D N Argyriou, J Appl Crystallogr , 1994,27, 155 48,10074 27 U Martin, H Boysen and F Frey, Acta Crystallogr , Sect B, 1993, 14 P Li, I -W Chen and J E Penner-Hahn, J Am Ceram SOC ,1994, 49,403 77,118 28 T Proffen, R B Neder, F Frey, D A Keen and C M E Zeyen, 15 P Li, I -W Chen and J E Penner-Hahn, J Am Ceram Soc ,1994, Acta Crystallogr , Sect B, 1993,49,605 77,1289 29 R I Smith and S Hull, Report RAL-94-115, Rutherford Appleton 16 17 D Steele and B E F Fender, J Phys C Solid State Phys, 1974, 7, 1 M Morinaga, J B Cohen and J Faber Jr, Acta Crystallogr, Sect A, 1979,35,789 30 31 Laboratory, 1994 W I F David, R M Ibberson and J C Mathewman, Report RAL-92-032, Rutherford Appleton Laboratory, 1992 V F Sears, Neutron News, 1992,3,26 18 J B Cohen, M Moringa and J Faber Jr, Solid State Ionics, 1981, 3/4,61 Paper 5/05805H, Received 1st September, 1995 898 J.Muter Chem., 1996, 6(5),895-898