Faulted defect aggregates in neutron-irradiated MgAl2O4 spinel

Single crystals of stoichiometric MgA1204 spine1 were neutron-irradiated to fluences (> 0.1 MeV) of 2.8 x loz5 n m-2 (1.5 dpa) at 1015 K and 1.9 x loz6 n m-2 (10 dpa) at 925 K and 1 100 K. The resulting point defect'aggregates were investigated using bright-field and weak-beam dark-field transmission electron microscopy. For the higher dose, faulted interstitial Frank dislocation loops were found on { 110 } and { 111 } planes with b = 114 < 110 ) (cation fault) and b = 116 < 111 ) (anion + cation fault) respectively. The latter preserves stoichiometry only for a randomized cation distribution. 1 . Introduction. -The nuclear industry has current and projected need for refractory insulating solids characterized by low swelling and structural integrity in severe radiation environments [I-31. Density measurements [2] of fission-neutron irradiated material have indicated that certain complex ceramics, notably MgA1204, Y3Al5Ol2, Si2N20 and SiA10N7s exhibit low swelling to doses as high as 10 displacements per atom (dpa), in contrast to simpler ceramics such as Be0 and A1203 which swell catastrophically, and often anisotropically. Void swelling in ceramic solids typically occurs over the temperature range 0.2-0.5 T, (T,=melting point) [3]. MgA1204 spinel has the advantage of being cubic, and thus isotropic, and has a large and complex unit cell (a = 0.808 nm, 56 atomslunit cell). Single crystal MgA1204 additionally exhibits negligible change in thermal diffusivity even after 10 dpa [4], indicating that a large isolated point defect concentration does not remain after irradiation and that Frenkel defects either recombine efficiently or accumulate in aggregates which somehow do not lead to void swelling. We have sought evidence from TEM to distinguish these possibilities and thus explain the radiation resistance of MgA1204. 2. Irradiation and specimen preparation. Single crystals of Czochralski-grown stoichiometric MgA1,04, obtained from Linde Division of Union (*) Work performed under the auspices of the U.S. Department of Energy. Carbide Corporation and containing 100 ppm Si, 20 ppm Fe, 5 ppm Ca and 8 ppm B impurity, were sawn to dimensions 10 mm x 10 mm x 0.5 mm and neutron irradiated in EBR-I1 to fast neutron (> 0.1 MeV) fluences of 2.8 x loz5 n m-2 at 1015 1 4 0 K (0.42 T,) and 1.9 x n m-2 at 925 t20 K (0.38 T,) and 1 100 + 20 K (0.46 T,). These two fluences correspond to approximately 1.5 dpa and 10 dpa, based on an oxygen displacement energy of 130 eV 151. The resulting primary Frenkel defect spectrum is unlikely to have been stoichiometric, however [6]. TEN specimens were sectioned parallel to ( 111 }, thinned to electron transparency using argon-ion bombardment [7] and coated with -20 nm evaporated carbon to preclude charge acquisition in the electron beam [8]. 3. Transmission electron microscopy. Thinned discs (foil normal n = [I 111) were examined in conventional transmission [9] at 200 kV using bright-field and weak-beam dark-field [lo] imaging methods. Little or no aggregate damage was observable in the lower dose (1.5 dpa, 1 015 K) sample, but in those subjected to the higher dose (10 dpa ; 925 K, 1 100 K) two large defect aggregate species were observed (A, B in figure 1) which are described separately below. 3.1 ( 110 ) LOOP ROSETTES. Clusters of faulted dislocation loops on all six equivalent ( 110 } planes were observed to intersect centrally, forming rosettes ; three of the loops are inclined by 350 to the (1 11) Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphyscol:1980659 FAULTED DEFECT AGGREGATES IN NEUTRON-IRRADIATED MgAI,O, SPINEL C6-233 Fig. 1. { 110 ) loops in (111) foil irradiated at 925 K and imaged in g = 2%. a) Bright-field. b) Weak-beam dark-field showing fault fringes for { 110 } loops at A and inclined { 111 } loops at B. foil plane at A in figure 1, the other three normal to the foil plane with their traces along ( 112 ). At 925 K, the mean loop size was -200 nm. At 1 100 K, the loops had grown to such large dimensions (> 1 pm) that they intersected the foil surfaces (Fig. 2). Contrast from the { 110 } loops using g = 2%-, 117or 511type reflections indicated a loop Burgers vector b along the ( 110 ) loop normals. The { 110 ) loops are therefore Frank [1 11 loops and represent addition or removal of material on { 110 } planes. The magnitude of the Burgers vector b (and thus of the fault vector R) was deduced from the presence of fault contrast for g = 220 and its absence for g = 440 (Fig. 3) where g . b = g.R is an integer and the phase shift across the fault 2 n. The fault is therefore a .n fault [12] with Burgers vector b = 114 ( 110 ). It is evident from figure 2 that this fault is not removed by an internal shear even when the loops become very large (> 1 pn). However, double layer loops at C, corresponding to nucleation of a second Frank loop on planes adjacent to an existing faulted loop, were observed which locally remove the faulted stacking sequence. The character of the { 110 } loops was deduced using the method of Groves and Kelly 1131 incorporating the precautions cited by Maher and Eyre [14]. The loops were determined to be of interstitial character and the faults extrinsic. 3.2 { 11 1 } LOOPS. Figure 1 reveals as well loops (at B) inclined by 70° to the (111) specimen plane which exhibited fault fringes for a g = 220type reflection. Equivalent loops were found parallel to the (111) specimen plane for which g. b = 0 but g.b x u # 0 (u = local tangent to the loop core) ; these loops exhibit residual contrast except for a line normal to g where g . b x u = 0. This behavior indicates a loop Burgers vector b along [Ill]. These { 111 } loops are thus Frank loops also. They were, however, consistently smaller (< 100 nm diameter) and found in lower density than the { 110 } Frank loops formed at 925 K, and were absent in the 1 100 K samples. The magnitude and character of the associated fault were deduced using image contrast for dark-field weak-beam diffraction conditions. For this determination, it was assumed that the fault vector R = 6 ( 111 ) was a simple 6 = 116, 114, 113 or 112 submultiple of the 12-layer { 11 1 ) stacking repeat. Fault contrast was observed for g = 220,311 and 131, consistent with a b = 116 ( 111 ) loop fault and Burgers vector (see discussion, 5 4.3). A loop character determination, carried out in parallel with the { 110 ) loop analysis, showed the { 111 } loops to be of interstitial type as well and the faults extrinsic. C6-234 L. W. HOBBS AND F. W. CLINARD, j~ 4. Discussion. 4.1 INTERSTITIAL CONDENSATION. The fact that the observed loops have interstitial character (extrinsic faults) requires that the loops correspond to condensation of displaced interstitial ions in a way which does not preserve the normal stacking sequence of { 110 } or { I l l ) planes in the spinel structure. The number of interstitials condensed (assuming stoichiometric condensation) was calculated from the measured total interstitial loop area and the magnitude of the Burgers vector and amounted to a stabilized interstitial concentration c 2: 5 x for the 925 K sample (0.005 % of the total displacements) and c = 9 x (0.009 % of the total displacements) for the 1 100 K sample. These numbers do not differ greatly and confirm the suspected absence of a swelling peak in the 0.3-0.5 T, temperature range for spinel. 4.2 {~~O}LOOPS.-The1/4(110){110}loops can be explained as the condensation of two extra layers aa (or a@) in the idealized { 110 } planar stacking sequence of normal spinel where the a layers contain 0 and A1 ions in the ratio AlO2, and the a and f i layers contain 0 , A1 and Mg ions in the ratio MgA102 (assuming no cation inversion). The 114 ( 110 ) { 110 } extrinsic fault is configurationally equivalent to the pure shear 114 ( 172 ) { 110 ) and has been associated previously with both F I ~ . 2 (1 11) foil of MgAI,O, neutron-irradiated to glide and climb dissociation of dislocations in stoichiometric and non-stoichiometric MgA1204 [15-191 and 1.9 x loz6 n m-2 (10 dpa) with growth faults in MgA1204 [20], Fe304 [21] and at 1 100 K showing large faulted loops on { 110 ) planes mtersectLiFesOs [22]. ing the foil surfaces and double-layer loops at C. Insertion of the two layers aa (or ap) preserves