An ceru So.8713465-72(20 urna Tetragonal-to-Monoclinic Transformation in Mg-PSZ Studied by in Situ Neutron Diffraction Yuxiang ma and Erich H. Kisi Department of Mechanical Engineering. The University of Newcastle. Callaghan, NSW 2308, Australia Shane J. Kennedy and Andrew J Studer Neutron Scattering Group, Australian Nuclear Science and Technology Organization, Private Mail Bag I Menai NSW 2234, Australia The deformation of 9.4 mol% magnesia-partially-stabilized reviews. 0-1 The dominant mechanism responsible for the high zirconia under compressive loads up to 1225 MPa was studied toughness of Mg-PSZ is widely held to be transformation tough- using mechanical testing with in situ neutron diffraction. The ening from the r-m transformation. -Transformation toughen- material shows obvious plastic deformation at applied stresses ng was first proposed by Garvie et al. in Ca-PSZ. Since then, in excess of an estimated critical stress of 925 t 20 MPa. Most significant new toughening mechanisms, such as micro-crack of the accumulated strain occurred by transient room. shielding and crack deflection have been proposed to make an temperature creep. Plastic deformation was associated with important contribution to the process. The bulk of the experimental considerable stress-induced tetragonal-to-monoclinie transfor work has been conducted by examining samples using TEM and mation. The volume change calculated from the strain gauges XRD after they have been mechanically tested. -Stress- correlates well with the amount of t- m transformation induced martensitic transformations, such as 1-m, are highly observed. Unlike previous studies of Ce-TZP and Y-TZP, dependent on the local stress-strain state of the sample. It is well ferroelasticity was not observed, nor was the t-o transfor- mation observed. Minor microstructural changes were noted, known that different surface preparation techniques give quite cluding an increase in the root mean square internal strain of 0.05%, commensurate with an increase in internal stress of tion, and so it can be used to give the bulk-phase proportions in the -100 MPa. It would appear that transformation selectivity ceramic. Er situ neutron diffraction has bee en success ul in solving was exercised with the transformation occurring first in tet- ragonal crystallites favorably oriented to the applied stress. such problems as the phase composition, low temperature he st strong preferred orientation. Comparison is made with the Mg-PSZs. Room-temperature creep has been observed using strain-gauge techniques in samples of 9. 4Mg-PSZ subjected to other commercially interesting zirconia ceramics, Ce-TZP and tensile stresses. An overall volume increase in the sample Y-TZP, which have been studied using the same techniques. fi.e, negative Poisson's ratio) suggested that the creep and duced t-m transformation were related. This was lat L. Introduction confirmed by ex situ neutron diffraction measurements on regions taken from the gauge volume and the(low stress) gripped region AGNESIA-PARTIALLY-STABILIZED ZIRCONIA (Mg-PSZ) has re- of the samples. When a long period of time had elapsed between ceived considerable attention in the literature for its high the tensile testing and neutron diffraction, discrepancies were fracture toughness. The material is a complex mixture of the noted between the amount of m phase observed and the measured ambient pressure cubic (c), tetragonal (0), and monoclinic (m volume increase in the samp 28.79As the strain gauge and phases of pure zirconia. The active microstructural component is neutron observations were conducted separately, it was suggested lenticular precipitates of the t phase, which is present in a largely that stored elastic stresses in the microstructure had triggered the c phase matrix. Under an applied stress, the I phase is able to reverse m-I transformation on relaxation of the applied stress undergo a martensitic transformation to the m phase, accompanied Hence, in addition to the surface sensitivity highlighted above, the by a volume increase of.9%. In its most refined form, results from ex situ experiments may not reflect the real transient controlled cooling from the solution treatment temperature at behavior of the material. In situ neutron diffraction has the 1700C followed by an aging treatment leads to precipitation of capacity to overcome both problems at once and hence to capture dynamic changes in the phase proportions, structural parameter the anion-ordered 8-phase(Mg2Zr, O 2), which replaces a major and internal stress distribution of each phase at different stresses proportion of the c matrix. The 8-phase precipitates at the the interface, leading to localized strain, metastability, and destabili- Recently, in situ neutron diffraction studies have been con- zation of the I phase. . 9 ducted on the microstructurally simpler tetragonal zirconia poly The phase-equilibrium, microstructure, and mechanical proper- crystals 12Ce-TZP 0.3 and 3Y-TZP. 2 There it was observed that the stress-induced I- m transformation is both time-dependent and reversible in 12Ce-TZP and that both materials show a degree of ferroelasticity under compressive loads. 3Y-TZP did not un- dergo any observable t-m transformation under compressive R. Hannink--contnbuting editor loads up to 2.3 GPa. The results of similar in situ loading experiments have not previously been reported for Mg-PSZ. It is important to comment briefly on the creep observed in Mg-PsZ by ex situ measurements and that observed in 12Ce-TZP 87072 Received March 19, 2002: approved December 30, 202. and 3Y-TZP by in situ diffraction measurements. The high scien y d e hunterian Research Couneil and the Australian temperature creep of ceramics is not a new phenomenon, and it is an Ceramic society well understood using the theoretical framework developed for
466 Journal of the American Ceramic Society-Ma et al. Vol 87. No. 3 creep in metallic materials. The mechanisms involved are not active at room temperature and cannot explain the observed creep in structural zirconia ceramics. The behavior ously observed in Mg-PSZ was transient creep and the strains(both longitudinal and transverse)approximated a power law similar to that proposed by Andrade ∈=Br It was observed that the stress also approximated a power b 38≥ leading to a composite equation of the form III II III In I! III IlIl El Il The creep parameters were quite variabl 2000 It I lI1 1101011IIa sensitively on the processing and thermal history of the samples. Although creep in 12Ce-TZP was also observed to follow a 2e(degree er law such as Eq (1). the observations were made during a stepwise loading experiment and it was accompanied by large Fig. I. Rietveld refinement result for 9.4Mg-PSz(sample 2).The bursts of transformation that prevented a systematic study of the observed data are indicated by (+)and the calculated pattern by the solid stress dependence. Very much smaller creep strains were observed in 3Y-TZP 32.37 and the time constant for reversal was much he observed and calculated patterns on ame scale. The rows of shorter. The only analysis that could be conducted was in the final markers below the pattern show(from the top down) the positions of peaks unloading step from 500 MPa to 0 MPa, during which the strains from cubic. tetragonal, and monoclinic zirconia and the 6 phase decayed to zero in good accord with a creep-relaxation model (Mg2 Zr,O1) This paper presents the results of an in situ neutron diffraction study of time-dependent stress-induced microstructural changes in 9.4 mol% Mg-PSZ during uniaxial compression testing shown in Fig. 1. Observed data are shown as(+) and the calculated profile as a solid line, The curve below is a difference plot between the observed data and the calculation, Reflection IL. Experimental Methods markers are indicated with small bars for the cubic, tetragonal monoclinic, and 8-rhombohedral phases(from the top down). Samples used for this study were Ts grade 9. 4 mol%o Mg-PSZ ICI Advanced Ceramics, Melbourne, Australia). The ceramic preparation is a proprietary process, the precise details of which IlL. Experimental Results are commercially sensitive. The process involves firing at -1700 C followed by controlled cooling and an aging treatment at (1) Stress-Strain Curves 100C. Sample I was cut to a height of 15 10 mm from a The stress-strain curves of the second 9. 4Mg-PSZ are shown in 90-mm-diameter rod with a diamond-impregnated saw Sample Fig. 2. Here, the strains are those at the end of the constant-stress 2 was cut into a prism having a cross section of 9. 70 mm X 9.80 holds and have been corrected for gauge off-set and an apparently mm at both ends. After cutting, the samples were polished with erroneous value in the longitudinal gauge at low stress. In both 1200-grit emery paper to remove surface micro-cracks, which may samples, when the stress was smaller than 800 MPa, the longitu- become the potential source of collapse. In addition the edges and dinal and transverse strains were linear functions of the applied corners were ground with fine SiC paper to have smooth 0.5 mm stress Sample I gave a Youngs modulus of 205+8 GPa, and radius rounded edges. Poissons ratio 0.33+ 0.06 in the range -100-1000 MPa. For The experiments were conducted with the medium-resolution Sample 2, Youngs modulus was determined to be 217+3 GPa powder diffractometer (MRPD) at HIFAR(high-flux Australian and Poissons ratio is 0.35=0.01, measured in the range 200-800 reactor of the Australian Nuclear Science and Technology Orga MPa. When the applied stress was >1000 MPa, nonlinearity nization). The angle range for neutron diffraction was 4-104 2 indicates that the sample had begun to plastically deform. Some with a step size of 0. 1. Neutron wavelengths were 1.6663A for nonlinearity in the transverse strain data were already apparent Sample I and 1. 6676A for Sample 2, as determined from a between 800 and 1000 MPa. We conclude that the critical stress is standard rutile specimen marginally <1000 MPa. Unlike in the elastic region, where the Compression testing of sample I began with a holding stress of transverse strain was much smaller than the longitudinal strain after the onset of plasticity, the transverse strains were much 2. 8 MPa, then 100 MPa, followed by 50 MPa increments up to larger. The plastic strains were estimated by extrapolating the 000 MPa For sample 2, the applied stress sequence was 5.2. 200 400,600.800.1000,100.1200.1225,1200.1100.1000.900 800.700,600.530.480.420.380,250.180.100,0MPa, followed by post-loading neutron diffraction patterns to detect microstruc- tural changes immediately after the load was released Transverse Two strain gauges were glued to the surfaces of the two Horzontal amples. One, placed vertically, measured the longitudinal strain 4000 and the other, placed horizontally, measured the transverse strain. Strain data were recorded every 30 s(Datataker DT50, Data 0 Electronics, Victoria, Australia) and downloaded to a computer. Rietveld analysis was used to extract information from leutron diffraction data as outlined in previous publications. 3/5 Four phases were found in these les, cubic zirconia (c) tetragonal zirconia(n), monoclinic zirconia (m), and the rhomb- hedral 8-phase(Mg, Zr, O,2). The refined parameters were global 1200 parameters and scale factors of all the four phases. In addition, for Applied stress(MPa the major phases, lattice parameters, the breadth of the internal strain distribution, and the March coefficient for preferred orien Fig. 2. Stress-strain curves of a 9. 4Mg-PSZ sample during uniaxial tation were also refined. An example refinement for sample 2 is compression testing
March 2004 Tetragonal-to-Monoclinic Transformation in Mg-PSZ Studied by in Situ Neutron Diffraction 467 elastic part of the stress-strain curves to the maximum stress. In Table 1, wt%o monoclinic phase estimated sample 2, the longitudinal strain deviated from the straight line by from Macroscopic Volume Change a maximum of-1331(He), whereas the transverse strain deviated from the straight line by-5887(He tress(MPa) Hold time (min) (2) Diffraction Patterns 1000 4.66 The diffraction patterns recorded during compression testing of sample 2 are shown in Fig. 3. On close scrutiny, in the loading 1150 1200 154 2336 process, the(I 11) reflection at% 26, grew gradually as the 1225 109 applied stress increased. The reflection appears to increase in size Total 989920.2 ven at relatively low loads. An obvious increase occurred when the applied stress was >1000 MPa. There is a corresponding decrease in the(013)and (121) peaks of tetragonal phase near 65 20 especially after o exceeded 1 100 MPa. At higher stresses, the To provide an independent measure of the real phase comp (I1I)m peak continued to increase as the load increased, until the tion In 9. 4Mg-PSZ under each applied stress, quantitative phase maximum stress, 1225 MPa. During the unloading process, no change in the size of the (11 1)m peak was observed In addition to analysis was conducted using the Rietveld refinement scale fac- peak intensity changes, some additional line broadening can be tors where the wt% of phase p is given by observed, e.g., in the double peak near 6520, where there is a (SZMV progressive loss of resolution y (SZMV 3)t→ m Transformation where Wp is the weight fraction of phase p, S is the scale factor, Z From the strain data, it is clear that the sample volume increased after the applied stress exceeded 800 MPa. It is well known that the one unit cell, and V is the unit cell volume of each phase. In total tetragonal-to-monoclinic phase transformation will give rise to a four phases were considered: the cubic, tetragonal, monoclinic 4.9%e volume increase. If it is presumed that the volume increase and 8-rhombohedral phases. The phase composition for the start in this experiment came totally from the f-m phase transforma- ing material is listed in Table II with error estimates. tion, the volume increase can be used to estimate the percentage of The calculated weight percentage of tetragonal phase and new monoclinic phase, The estimated fraction of monoclinic phase corresponding weight percentage of monoclinic phase during the produced during each hold at constant load was calculated and is whole mechanical testing process are shown in Figs. 4(a) and(b) shown in Table L. These results neglect a small amount of The arrows indicate the sequence in which the external stress was transformation that occurred essentially instantaneously during applied. When the load was between 0 and 800 MPa. the fraction application of the load increment In the loading half-cycle. a total of monoclinic phase was constant to within the estimated errors. of -20 wt% of new monoclinic phase was predicted by this As expected from the raw neutron data( Fig 3), during loading, the calculation greatest increase in the monoclinic phase was observed at stresses 20000 MPa 15000 420 058 4567 120 1200 5000}1150 1000 400 20(degree) Fig 3. A portion of the diffraction patterns recorded during compression testing of the 94Mg- PSZ. Note the growth of the monoclinic-phase peaks, e.g. at31°20
Journal of the American Ceramic Sociery-Ma er al Vol. 87. No. 3 Table I. Phase Content of 9, 4Mg-/ at 5.2 MPa (4) Structural and Microstructural Effects Determined by Neutron Diffraction In addition to phase quantification, the in situ neutron diffrac z M(amu) V(A') wt ErTwt tion and Rietveld analysis gives data on changes to the structure and microstructure of the ceramic during loading and un 4I1.61131.520.5 9 particular, changes were noted in the axial ratio, c/a, of the 2123.2267.0412.162.61 tetragonal phase, preferred orientation in the tetragonal and mon m-Zr( 4123.22139.660.36.50.7 aclinic phases, and the internal strain distribution in the tetragonal Mg2Zr3O123696.72679060.0492180.9 phase. A composite diagram summarizing these results is shown in (A) Tetragonality: The c/a ratio of the tetragonal phase >1000 MPa(Fig 4(b). Likewise, when the load was between 0 adopting the pseudocubic unit cell, is shown in Fig. 5(a). Overall c/a increases as a function of applied stress and the slopes of the within one standard deviation(Fig. 4(a)). in good correspondence curves during loading and unloading are the same at stresses with the weight fraction of monoclinic phase. The amount of <1100 MPa. The lattice parameters observed were the average phase began to decrease after the load reached 800 MPa, and the alues of those crystallites oriented so as to satisfy Braggs law. So rate accelerated after the stress exceeded 1 100 MPa In agreement from-0-1 100 MPa, tetragonal particles with c axes perpendicu- with the strain gauge data, the critical stress for the stress-induced lar to the applied stress expanded(Poisson strains) faster than transformation would appear to be-1000 MPa in this material those with the a axes perpendicular to the applied stress or more These results confirm that the f-m phase transformation is responsible for the bulk of the strains observed In the unloading half of the stress cycle, the fraction of monoclinic phase did not stop increasing, but rose quickly until 1.0215 1000 MPa. Between 1000 MPa and 0 MPa. the fraction of monoclinic phase continued to increase but at a much smaller rate. The apparent tetragonal phase content too continued to decrease ith a stable rate during unloading 1.0210 From the observations shown in Figs. 4 (a) and(b). it was estimated that in the loading half cycle (i.e, <1225 MPa). the tetragonal phase decreased 10.9=1, 2 wt% and the monoclinic the tetragonal phase was 190+1.5 wt% and total increase in the o10205 hase increased 12.2+ lI wt%. The total estimated decrease in monoclinic phase was 16.9+ 1. I wt%. 1.0200 0.35 0.20 1.04 Applied stress(MPa) 1200 Applied stress(MPa) a)Tetragonal-phase content in 9.4Mg-PSZ during a compression loading cycle calculated from Rietveld refinement scale factors: Fig. 5. (a) Tetragonality of the I phase: (b) internal strain distribution in rved monoclinic-phase content during the same the loading. I phase; (c) March coefficient of t phase, during loading and unloading of cycle. Error bars indicate one estimated standard deviation. 9.4Mg-PSZ. Error bars indicate one estimated standard deviation
March 2004 Tetragonal-to-Monoclinic Transformation in Mg-PSZ Studied by in Situ Neutron Diffraction simply, for the tetragonal phase s13 >$12. This behavior contrasts load, were averaged over every 30 s. The averaged strain data were with 3Y-TZP and 12Ce-TZP37 with simpler phase compositions. hen plotted as a function of time at each applied stress. An where c/a was unchanged below the critical stress At >1100 MPa example of such curves is given in Fig. 6(a), showing th the c/a ratio in Mg-PSZ decreased by approximately three standard accumulation of vertical, horizontal, and volume strains at 800 deviations. Reasons for the decrease may be: (i) removal from the MPa, below the gross critical stress. Creep curves for samples diffraction pattern by transformation of the most highly stressed I loaded >800 MPa were fitted to functions using the model given articles, causing the average to shift to a lower value: (i)removal by Finlayson et al. 20 rom the diffraction pattern of the least-stabilized t particles, so Unlike the strain from the tensile creep samples. the strain in again the average shifts to lower values: or(ii) residual strain this compression test was not a direct function of applied stress imposed on I-phase particles by the transformed monoclinic phas from the step-wise loading. The fitting was to a modified Andrade If the latter effect is responsible, the residual stresses can be function estimated to be of order I 10 MPa (B) Internal Strain Distribution in the t Phas ∈=∈n+Brn uare(rms) width of the internal strain dis lculated from the tane component of the The fitted creep curves are shown in Fig. 6(b). Uobs. using the following equation: <A The volume strains caused by creep have been used to estimate amount of m phase formed from creep at each load on the right-hand scale in Fig. 6(b). The resemblance of the terminal em=m-bm×18042n2 values from Fig. 6(b)to those in Table I suggest that an overwhelming majority of the observed volume change in 9. 4Mg- where U. is the standard width from the instrument(0.31 when PSZ was from creep MRPD is used in high-intensity mode at a wavelength of 1. 67A The results shown in Fig. 5(b) indicate that the internal strain IV. Discussion distribution is not constant, even during elastic loading and unloading at <1000 MPa. although the graph in these regions is Mg-PSZ loaded in uniaxial compression shows almost purely parallel, Elastic anisotropy is believed to be responsible for the elastic deformation <800 MPa and well-developed plastic defor slope in this region. An increase in internal strain at >1000 MPa mation >1000 MPa. The plastic deformation is accompanied by a accompanied the r-m phase transformation. The overall increase volume expansion. In situ neutron diffraction has shown that the in internal strain of 0.05% is equivalent to an increase in rms plastic deformation is accompanied by a substantial amount of internal stress of-lo0 MPa stress-induced t- m transformation within the ceramic. No (C) Preferred Orientation: Previous uniaxial compression evidence was found in this material for ferroelastic switching of t ests with in situ neutron diffraction 1.2 on 3Y-TZP and 12Ce- crystallites nor was the competing t-o transformation that has TZP have shown considerable changes to the orientation distribu- been previously reported in tensile room temperature creep sam convenient measure of the degree of preferred orientation and from the volume change (assuming a 4.9% local volume change on density function due to March"obtained during the Rietveld tron diffraction data. The two estimates for the total amour refinements, In both cases referred to above, the ferroelastically transformed agree to within two standard deviations. It can switched domains oriented to diffract from the shorter a crystal therefore be concluded that the plastic strains are primarily due to lographic axes out of the neutron beam and switched the longer c the t-m transformation. coefficient of the t phase above the critical stress. By contrast, for as The transition from elastic to plastic deformation in Fig. 2 is not axes into the beam. This resulted in a strong decrease in the March 9. 4Mg-PSZ, the March coefficient of the t phase began very close stress takes. A value in excess of 1000 MPa is indicated by the to unity, indicating an initially random orientation of t crystallites. longitudinal strain in Fig. 2. whereas the transverse strain gives a The influence of the compressive stress is shown in Fig. 5(c)to be slight increase in the March coefficient. It can be concluded from this that ferroelastic reorientation of the I phase does not occur in observable amounts in 9.4Mg-PSZ during uniaxial compression. It is believed that the slight increase in the March coefficient is due to slight transformation-induced texture. I42 The March coefficient for a monoclinic structure is often x nadequate, as there is no a priori reason to select a particular pole axis or preferred orientation vector. It does, however, often serve as a useful correction to the calculated intensities by use of a vector chosen based on the observed intensities In these experiments, it was noted that the stress-induced m phase has a more prominent 2500 (II reflection than a random orientation. The number of other phases present and the initially small amount of m phase present made the refinement of meaningful preferred orientation coeffi cients as a function of applied stress difficult. Instead, a March coefficient was refined at the maximum m phase content, and this >1000 was used as a constant in all other refinements above the critical stress and during unloading. The preferred orientation was con firmed by one of the post-loading neutron diffraction patterns that was taken after the sample was turned 90 to put the compression 120160 axis of the sample in the diffraction plane. This pattern showed a Time(min far smaller( 11 1) peak in agreement with the above observations (5) Time-Dependent Strain/Creep Fig. 6. (a) Transverse strain. 800 MPa of 9 4Mg-PSZ showi As noted in the introduction, strains in zirconia ceramics have fitting of the volume strain at been reported to have a strong time-dependence. .- To assess the monoclinic phase fraction this, the strain data, recorded while the sample was held at each by 0. 049% for each I% increase in volume