Dynamics of nanoscale polar regions and critical behavior of the uniaxial relaxor Sr 0.61 Ba 0.39 Nb 2 O 6 :Co J. Banys, 1 J. Macutkevic, 1 R. Grigalaitis, 1 and W. Kleemann 2, * 1 Faculty of Physics, Vilnius University, Sauletekio 9, 10222 Vilnius, Lithuania 2 Angewandte Physik, Universität Duisburg-Essen, D-47048 Duisburg, Germany Received 23 January 2005; published 12 July 2005 The complex dielectric susceptibility of Sr 0.61 Ba 0.39 Nb 2 O 6 : Co has been measured at frequencies 10 1 f 10 9 Hz and temperatures 300 T 400 K before and after poling. The high-frequency relaxation peak observed above the ferroelectric phase transition temperature, T c 348 K, is analyzed in terms of distribution functions of the relaxation time, g, using Tikhonov regularization. It reveals the criticality of the largest relaxation time in agreement with activated dynamic scaling of the three-dimensional random-field Ising model. Differences of zero-field- and field-cooled gare discussed. DOI: 10.1103/PhysRevB.72.024106 PACS numbers: 77.22.Ch, 64.70.Pf, 77.80.-e, 77.84.Dy I. INTRODUCTION Solving the relaxor enigma is still one of the most chal- lenging problems in the physics of ferroelectrics. Ever since the discovery of the archetypical relaxor material PbMg 1/3 Nb 2/3 O 3 PMN, 1 the origin of its key features—i huge frequency dispersion of the dielectric susceptibility in the Curie range above the transition temperature T c , iinon- ergodic behavior at low temperatures T T c , and iiivirtu- ally no symmetry breaking after zero-field cooling ZFCto below T c —has been discussed controversially. Very probably and in accordance with long-standing propositions of Cross and co-workers 2,3 the appearance of fluctuating polar precur- sor clusters polar nanoregionsat temperatures T T c has to be considered as the primary signature of relaxor behavior. 4 Indeed, it appears plausible that polar nanoregions with a wide size distribution are at the heart of the observed giant polydispersivity, 5 which is observed in quite different sys- tems like disordered solid solutions like PMN Ref. 1or PLZT Ref. 2, or doped quantum paraelectrics like SrTiO 3 :Bi Ref. 6. Hence, finding out the microscopic origin of the polar precursor clusters is the key to understanding relaxor prop- erties. Very probably, different mechanisms with different degrees of complexity have to be envisaged. They may range from purely statistical aggregation of polar centers in the sense of percolation theory probably realized in the case of doped quantum paraelectrics like SrTiO 3 :Ca Refs. 7 and 8 to the appearance of weak “exponentially rare”singulari- ties within a dilution-induced Griffiths phase 9 possibly real- ized in the case of solid solutions like PLZT, which reveal anomalies of the refractive index below the phase transition temperature T c of the parent PZT phase irrespective of the amount of La doping. 10 . Apart from such very special mechanisms, however, the random-field RFmechanism 11 is supposed to be a widely applicable concept. Since spatial fluctuations of the RF’s give rise to correlations between the fluctuating dipole moments, the formation of precursor clus- ters of mesoscopic size is expected in systems with charge disorder. 12–14 Starting from the existence of polar nanoregions and as- suming random interactions between them, a spherical random-bond random-field SRBRFtheory 15 has success- fully been introduced to describe the phase transition into a disordered low-T cluster glass phase. By assuming fully frus- trated superspin interactions and dominance of these random bonds RB’sover the RF’s, it models the glassy freezing of mesoscopic nanoregions in cubic relaxors like PMN Ref. 15and PLZT Ref. 16. Indeed, quite weak RF’s are gener- ally expected to act on the polar clusters, since only the fluctuations of the microscopic RF’s are effective on a nano- scale. While cubic relaxors with nearly continuous order pa- rameter have no chance to experience ferroelectric long- range order, 11 this must not be the case for random-field sys- tems with a discrete order parameter. This is the reason why the well-known family of uniaxial relaxor crystals related to strontium-barium niobate, Sr x Ba 1-x Nb 2 O 6 SBN, 17 is con- sidered as a prototypical three-dimensional 3Drandom- field Ising-model RFIMsystem since recently. 18 In contrast to the cubic perovskite family related to PMN, 1 the polarization of the SBN family is a single- component vector directed along the tetragonal c direction, which drives the point group from 4 / mmm to 4mm at the phase transition into the low-T polar phase. Only at low tem- peratures is a slight tilt of the polarization out of the 001 direction observed, giving rise to a monoclinic structure monoclinic point group mbelow T 0 70 K. 19 At T T 0 , however, SBN is tetragonal on the average and thus belongs to an Ising model universality class rather than to a Heisen- berg one as suggested for the PMN family. 12 Assuming the presence of quenched RF’s, theory in this case predicts the existence of a phase transition into long-range order LRO within the RFIM universality class 11 preceded by a giant critical slowing-down above T c . Hence, when approaching the ferroelectric state, the polar nanoregions, which are cor- related by RF’s in the paraelectric phase, should exhibit ac- tivated dynamics. 20 Below T c the polar nanoregions are expected to transform into polar LRO, provided that cooling is performed at a suf- ficiently low rate. 12 Recent studies of SBN and SBN:Ce have, indeed, evidenced that ithe dielectric susceptibility probes clusters above and domains below T c , respectively, 21 while iithe linear birefringence is a measure of polar pre- PHYSICAL REVIEW B 72, 024106 2005 1098-0121/2005/722/0241066/$23.00 ©2005 The American Physical Society 024106-1