We address the unsolved question on the structure of relaxor ferroelectrics at the atomic level by characterizing lead-free piezoceramic solid solutions (100 − x)(Bi 1/2 Na 1/2 )TiO 3 -xBaTiO 3 (BNT-xBT) (for x = 1, 4, 6, and 15). Based on the relative intensity between spectral components in quadrupolar perturbed 23 Na nuclear magnetic resonance, we present direct evidence of the coexistence of cubic and polar local symmetries in these relaxor ferroelectrics. In addition, we demonstrate how the cubic phase vanishes whenever a ferroelectric state is induced, either by field cooling or changing the dopant amount, supporting the relation between this cubic phase and the relaxor state. Relaxor ferroelectrics have been intensively studied in the past 30 years because of their intriguing structural and dielectric features [1][2][3]. Despite that, the nature of their ground state remains an open question [4][5][6][7][8][9][10][11][12]. Several models have been proposed to describe this puzzling class of materials [1,[5][6][7][8]13], two of which are mainly concerned with the structure of the ground state of relaxors [4].The random field (RF) model proposes a single-phase structure broken up into ferroelectric nanodomains. These nanodomains are kept small under the constraint of quenched random electric fields which originate from chemical disorder [7,14]. The other model introduced the concept of "polar nanoregions" (PNRs) [13]. When relaxors are cooled below the Burns temperature, small and randomly oriented polarized regions (PNRs) appear within the otherwise nonpolar crystal structure. Upon further cooling, PNRs grow in size and number [11], but their percolation is prevented by structural disorder and random electric fields [8,10,12]. This behavior would imply the coexistence of PNRs and a nonpolar matrix, in contrast to the single-phase structure proposed by the RF model.An induced ferroelectric state can be established in relaxors when a strong electric field is applied. This state features macroscopic polarization and breaking of the cubic average symmetry [5,11,[15][16][17]. While the RF model suggests that this transformation and the formation of macroscopic ferroelectric domains are caused only by reorientation of previously nanometric ferroelectric domains, an additional mechanism is required when PNRs are considered. In that case, a long-range ferroelectric state can only be established if the nonpolar matrix becomes polarized, being incorporated by the growing PNRs.Two questions are raised by contrasting these two models: (1) Does the microscopic structure of relaxor ferroelectrics consist only of ferroelectric nanodomains or do regions of lower local symmetry (PNRs) coexist with a nonpolar matrix of * gerd.buntkowsky@chemie.tu-darmstadt.de † roedel@ceramics.tu-darmstadt.de undistorted structure? (2) How does the microscopic structure of relaxors evolve upon electric poling?In spite of their relevance, such questions have remained unsolved largely because the structural characterization of relaxors is a c...