Influence of neutron and gamma irradiation on the electrocaloric properties of Mn-doped 0.9Pb(Mg1/3Nb2/3)O3–0.1PbTiO3 ceramics

The influence of neutron and gamma irradiation on the low- and high-field dielectric and electrocaloric (EC) properties of Mn-doped 0.9Pb(Mg1/3Nb2/3)O3–0.1PbTiO3 (PMN–10PT) ceramic is studied. Upon exposure to neutron fluences of up to 1017 cm−2 and gamma-ray doses of up to 1200 kGy the Mn-doped PMN–10PT exhibits a lower saturated polarization, increased internal bias field and reduced EC temperature change. In comparison, the respective properties of the undoped PMN–10PT remain almost unchanged upon exposure to neutrons and gamma rays. In Mn-doped PMN–10PT, the acceptor-oxygen vacancy defect complexes, introduced via doping, contribute to the lowering of the threshold radiation dose that the material survives without noticeable changes in properties. Radiation-induced degradation of the EC response of Mn-doped PMN–10PT can be partially healed by annealing at 450 °C. The study provides guidance for designing EC ceramic materials for solid-state cooling applications in environments of high ionizing radiation, such as the medical field or space technologies.


Introduction
Harsh conditions, such as high-intensity ionizing radiation, imply additional requirements for the stability of the operation of electronic components. Application areas include medical diagnostics and therapy, such as linear proton accelerators, space technologies, satellites, particle colliders, nuclear fusion and fission reactors [1]. The versatile functionality of ceramics and their inherent chemical inertness make them excellent candidates for use in such harsh conditions. In terms of size and shape, up-and-down-scaling possibilities are enabled by a broad range of ceramic technologies resulting in bulk, thick and thin films or multilayer elements.
Electrocaloric (EC) cooling, driven by a change in entropy in a dipolar material under the application of electric field [2], is an attractive approach to mitigating problematic heat generation in electronic components [3]. Relaxor ferroelectrics are among the strongest candidates for EC cooling as they exhibit EC temperature changes of a few K across temperature intervals of a few 10 K. The prototypical relaxor Pb(Mg 1/3 Nb 2/3 )O 3 (PMN) or PMN-rich PMN-PbTiO 3 (PMN-PT) solid solutions, exemplified by 0.9Pb(Mg 1/3 Nb 2/3 )O 3 -0.1PbTiO 3 (PMN-10PT), represent a material system that has been considered for EC cooling [4,5].
The EC temperature change increases with the increasing electric field, shifting the peak values to higher temperatures. For example, the dielectric permittivity maximum for PMN-10PT ceramic is at ∼50 • C at 1 kHz and at zero bias field, and the maximum EC temperature change (3.5 • C) is at 127 • C, and at 160 kV·cm −1 [4]. A classical approach to tailoring operational temperature range or, in other words, phase transitional behavior and the dielectric, ferro-and piezoelectric properties of ferroelectric ceramics has been by chemical modification, best exemplified by donor or acceptor doping of ferroelectric Pb(Zr,Ti)O 3 (PZT) ceramic which leads to softening or hardening of the electrical, mechanical and electromechanical response, respectively [6,7]. Li et al found that donor doping of morphotropic phase boundary compositions of relaxor-ferroelectric PMN-PT ceramics by selected rare earth ions, most pronouncedly by samarium, resulted in an exceptionally high piezoelectric response, at least in part related to the enhanced chemical and structural disorder at the nanoscale [8]. Studies of the acceptor (manganese) doping of relaxor-ferroelectrics revealed that the dopant introduced similar 'hard' properties as in PZT: lowering the remanent polarization, increasing the coercive field, and lowering the losses, which was related to the presence of acceptor-oxygen vacancy defect complexes [9][10][11]. A recent study of manganese doping of PMN-10PT showed that the acceptor-oxygen vacancy defect complexes contributed to the dynamics of polar nano regions resulting in strongly suppressed frequency dispersion of dielectric permittivity and reduced freezing temperature compared to the undoped material. Consequently, the EC temperature change in Mn-doped PMN-10PT was less temperature dependent than in the undoped material, meaning that the temperature interval over which the EC response was maximum was broadened [12].
Irradiation introduces additional defects in materials; for example, energetic photons such as gamma rays can induce ionization, contribute to the generation of electron-hole pairs, and/or induce atom displacements. Neutrons or protons interact with the crystal lattice and introduce interstitials, vacancies, and/or clusters of defects as well as ionize the matter [1,13,14]. Sternberg et al studied the effects of neutron (with fluences up to 2 × 10 18 neutrons·cm −2 ) and/or gamma-radiation (up to 5 × 10 7 Gy) on dielectric properties of lanthanum-doped Pb(Zr 0.65 Ti 0.35 )O 3 (PLZT) ceramics with 4-10.5 mol% La. High doses of neutrons were more detrimental for ferroelectric PLZT with a low La-content than for the relaxor PLZT formulations with high La contents. Neutron irradiation reduced the dielectric permittivity and (high-field) polarization and increased the coercive and internal bias fields. The original polarization state of the materials could be restored by annealing to a few 100 • C (depending on the dose), which was related to the relaxation of radiation-induced defects [15][16][17][18]. Brewer et al noted that manganese doping of PZT films that resulted in a decreased polarization response compared to the undoped films enhanced their gamma-radiation tolerance for doses up to 10 5 Gy [19]. A phenomenological model has been proposed to relate measurable degradation of physical properties of functional oxides, including ferroelectrics, depending on the total ionization dose, notwithstanding the type of ionization [20]. Microstructural features, such as grain and related domain size and porosity, were found to influence the radiation hardness of polycrystalline ferroelectrics, bulk and thin films [21][22][23].
Limited work has been reported on radiation effects on EC properties. A theoretical study on the effects of proton irradiation on the EC response of BaTiO 3 thin films suggested that proton irradiation would contribute to an increased oxygen vacancy density and radiation-induced internal electric field, resulting in a decreased EC temperature change [24].
The present work explores how a relaxor ferroelectric material with intentionally introduced charged defects, acceptor (Mn) doped PMN-10PT, responds to neutron and gamma radiation. PMN-10PT ceramics, doped with 0.5 and 1 mol% Mn, together with the undoped PMN-10PT, used as a reference, were exposed to a neutron fluence of 10 16 cm −2 and 10 17 cm −2 and gamma-ray doses of 145 kGy and 1200 kGy in the Jožef Stefan Institute (JSI) TRIGA Mark II research reactor [25]. The present results using a new batch of undoped PMN-10PT samples confirm that neutron and gamma radiation have no significant effect on low-and high-field dielectric and EC properties, consistent with our previous report [26]. Further, we show here that the presence of intentionally introduced charged defects arising from acceptor doping makes samples more susceptible to modification by irradiation-induced defects, but the radiation-induced degradation in ferroelectric properties can be healed, at least in part, by annealing.

Experimental procedure
For the synthesis of the stoichiometric PMN-10PT, 0.5 and 1 mol% Mn doped PMN-10PT assuming substitutional doping on B-sites of the perovskite lattice, PbO (99.9%, Sigma Aldrich), MgO (99.95%, Sigma Aldrich), Nb 2 O 5 (99.9%, Sigma Aldrich), TiO 2 (99.8%, Sigma Aldrich), and MnO 2 (99.9%, Sigma Aldrich) powders were used as reagents. The samples were prepared by the mechanochemical reaction described in Table 1. The samples were irradiated in a mixed field of neutrons and gamma-rays and exposed to two different irradiation levels with specific values of neutron flux and gamma-ray dose. The unirradiated level is referred to as level 0, while the highest irradiation is referred to as level 2. In this paper, the unirradiated samples are referred to as the 'pristine' . detail in [27]. First, PbTiO 3 seeds were prepared by solid-state synthesis at 750 • C for 2 h. In the next step, stoichiometric amounts of MnO 2 , PbO, MgO, and Nb 2 O 5 and PbTiO 3 were homogenized in a planetary mill (Retsch, PM400, Germany) at 200 min −1 for 2 h. The powder mixture was transferred to a tungsten carbide milling vial (V = 250 cm 3 ) filled with 12 tungsten carbide milling balls (2r = 20 mm) and mechanochemically activated at 300 min −1 for 80 h. Afterwards, the synthesized powder was milled in an attrition mill (Netzsch, PE075/PR01, Germany) with yttrium-stabilized zirconia beads at 800 min −1 for 4 h. The powders were compacted by uniaxial (50 MPa) and isostatic pressing at 300 MPa. Sintering took place at 1200 • C for 2 h in air in double alumina crucibles with lids in the presence of a packing powder of the same chemical composition as the pellet. The heating and cooling rates were 2 • C·min −1 . Some 1 mol% Mn-doped PMN-10PT samples were additionally annealed at 450 • C for 1 h and cooled with 2 • C·min −1 . From here on, the PMN-10PT, 0.5, and 1 mol% Mn-doped samples are denoted as undoped, Mn-0.5 and Mn-1, respectively.
The densities of the sintered ceramic pellets were obtained using a gas-displacement density analyzer (Micromeritics, AccuPyc II1340 Pycnometer, USA) at room temperature (RT). The theoretical densities were calculated from the Rietveld refinement analysis, as shown in the supplementary material table S4. The pellets were cut into disks, thinned, and polished to thicknesses from 100 µm to 220 µm with RMS roughness values between 150 nm to 350 nm measured by a Stylus Profilometer (Bruker DektakXT Advanced System). In order to relax internal stress generated during mechanical processing, the disks were heated at 600 • C for 1 h with a heating rate of 5 • C·min −1 and then cooled at 1 • C·min −1 to RT.
The irradiation of ceramic samples was performed in a mixed field of neutrons and gamma-rays in a TRIGA Mark II research reactor [25]. The samples were irradiated in the so-called central channel with a neutron fluence of 10 16 neutrons·cm −2 to 10 17 neutrons·cm −2 in terms of 1 MeV equivalent neutron fluence for silicon [28]. The gamma-rays were not blocked during the experiment; therefore, the samples were also exposed to the cumulative doses (air kerma equivalent) of 145-1200 kGy of gamma-irradiation. The total doses of individual irradiation are summarized in table 1. The temperature was maintained at approximately 100 • C during irradiation.
X-ray diffraction (XRD) was performed using a high-resolution diffractometer (X'Pert PRO, PANalytical, Netherlands) with Cu-Kα 1 radiation source. XRD data were collected in the Bragg-Brentano geometry using a 100-channel X'Celerator detector in the 2θ range of 10 • -120 • . The XRD measurements were carried out with a step angle of 0.017 • and an integration time of 100 s per step. Furthermore, the XRD patterns were simulated using the Topas R package (version 2.1, Bruker AXS GmbH, Germany) through the Rietveld refinement method, where the fundamental parameters approach was employed for the line-profile fitting of all the samples [29]. Several parameters were chronologically refined during the refinement process, such as the background (7th order Chebyshev), unit cell, crystallite size, scale factors, sample displacement, atomic coordinates, and thermal parameters. The goodness-of-fit parameters, e.g. R wp , R p , R exp , R B , and G.O.F correspond to the best-fitted XRD patterns.
The microstructure of fracture surfaces of pristine and irradiated specimens was investigated using a field-emission scanning electron microscope (FE-SEM, JSM-7600 F, Jeol Ltd, Japan).
For electrical measurements, Au electrodes with a radius of 5 mm were deposited on the surfaces of the pellets (2r = 6 mm) using RF-magnetron sputtering equipment (5Pascal, Italy). The measurements of relative dielectric permittivity (ε) and dielectric loss (tan δ) as a function of temperature and polarization-electric field (P-E) hysteresis loops were performed by an Aixacct TF analyzer 2000 (aixACCT Systems GmbH, Germany) setup equipped with an HP 4284 A Precision LCR impedance meter (Hewlett-Packard, California, USA). The temperature range for ε and tan δ measurements was −40 • C to 200 • C in the frequency range of 0.1-50 kHz. For the P-E loop measurements, the electric field up to 30 kV·cm −1 in sinusoidal waveform with a frequency of 1 Hz was applied. Direct EC measurements were performed on the pellets with sputtered Au electrodes in a modified high-resolution differential scanning calorimeter (Netzsch DSC 204 F1, Germany), which allowed to maintain a precise temperature stabilization (±0.001 • C). The electric field was applied on the samples in the form of a square waveform for a period of 200 s which implies 100 s on-field and 100 s off-field in the range of 5-30 kV·cm −1 using Keithley High Voltage Source-Measure Unit 237. The EC temperature change (∆T EC ) was recorded directly by a small-bead thermistor attached to the sample surface. The thermistor is connected to a Keithley 2100 digital multimeter (Keithley Instruments, Solon, Ohio, USA). The deviation in ∆T EC values between individual samples of the same composition exposed to the same irradiation dose was ∼10%, see supplementary note 1 and figure S1. The ∆T EC values reported in this work are the experimental values multiplied by a correction factor of 1.4-1.6 as described in [12,30].

Results and discussion
The room-temperature XRD patterns of pristine and irradiated Mn-doped PMN-10PT ceramics are collected in figure 1. The ceramics crystallize in a single-phase cubic structure with space group Pm3m. There is no visible effect on the crystal structure due to irradiation that could be observed in the XRD patterns. The structural refinements of the XRD data of the pristine and irradiated samples were carried out and are listed in supplementary tables S1 and S2. The Rietveld refinement of respective XRD patterns is demonstrated in figure S2. No noticeable effect of irradiation on structural parameters of Mn-doped PMN-10PT is observed. As a reference, we also analyzed pristine and irradiated undoped PMN-10PT ceramic samples and obtained results consistent with our earlier study of undoped PMN-10PT [26]. The structural data and XRD patterns of undoped PMN-10PT are collected in supplementary table S3 and figures S3(a), (b). (For completeness, figure S3(c) shows the change in lattice parameters upon incorporating manganese ions into the perovskite lattice in agreement with [12]). Similar to the structural analysis results, we observed no differences in the microstructure of pristine and irradiated samples on the level of FE-SEM (supplementary figure S4). In all cases, the microstructures consisting of micron-sized grains were dense, in agreement with the relative densities of investigated samples of 94% or more (supplementary table S4). Figure 2 depicts the temperature-dependent relative dielectric permittivity ε ′ and the dielectric loss (tan δ) of Mn-doped PMN-10PT ceramics before and after irradiation experiments measured at 1 kHz. The peak permittivity values of pristine Mn-0.5 and Mn-1 at 1 kHz are 12 600 and 8300, respectively. No noticeable changes in the dielectric response between the pristine and irradiated Mn-doped samples are observed up to the irradiation dose 1, i.e. 10 16 neutrons·cm −2 and 145 kGy. Upon exposure of the samples to irradiation dose 2, i.e. 10 17 neutrons·cm −2 and 1200 kGy, the peak permittivity values exhibit variations of up to ∼10% compared to pristine samples. Please note that different samples (pellets of different cuts) were measured in individual experiments; thus, it is impossible to distinguish between a possible spread of peak permittivity values between individual pellets of the same chemical composition and the irradiation effect. Prah et al reported that a difference of ∼13% in peak permittivity values between different pristine samples of PMN-35PT exceeded by about half the difference in the peak permittivity values between a pristine specimen and a specimen exposed to the dose of 10 17 neutrons·cm −2 and 1200 kGy [26]. We observed no noticeable difference in the suppressed frequency dispersion of Mn-doped PMN-10PT samples that could be related to irradiation. The frequency dispersion results, collected in supplementary figure S5, agree with previous studies of acceptor doping in relaxor ferroelectrics [11,12]. By comparing the influences of Mn-doping and irradiation on the dielectric response of doped and undoped formulations, it is evident that the effect of doping prevails.
RT high-field P-E hysteresis loops measured at 1 Hz and the corresponding current density-electric field (J-E) responses of the pristine and irradiated Mn-0.5 and Mn-1 samples are collected in figure 3. Both pristine samples exhibit pinched P-E loops. The double current-density peaks additionally support the shape of the loops. Bradeško et al considered various possible origins of pinching P-E loops in acceptor-doped PMN-10PT. They concluded that it was due to the pinning of domain walls by the acceptor (Mn)-oxygen vacancy defect complexes [12]. By inspecting more closely the P-E loops of the Mn-1 sample (figure 3(c) inset), an increasing internal bias field E ib (i.e. a shift of the loop along the horizontal axis, in this case, towards negative fields) becomes evident in irradiated samples (E ib ∼ 0.01 kV·cm −1 , 0.05 kV·cm −1 and 0.10 kV·cm −1 for 0, 1 and 2 dose levels, respectively). Since E ib is commonly attributed to the preferred alignment of acceptor-oxygen-vacancy defect complexes [10], it is reasonable to assume that the irradiation has an effect on the defect states, likely altering the polarization-pinning behavior as inferred from changes in the bias field. Irradiation-induced bias field was reported in PLZT for doses of/above 5 × 10 17 neutrons·cm −2 [17,18].
The EC temperature changes (∆T EC ) of the pristine and irradiated Mn-doped PMN-10PT ceramics were obtained through direct measurements at electric field amplitudes of 5-30 kV·cm −1 . Selected results are included in figure 4, while all measured ∆T EC can be found in supplementary figure S6.
We note that indirect measurements support direct measurements. The polarization-electric field data measured between 30 • C and 110 • C and the extracted ∆T EC values are shown in supplementary figures S7 and S8. Supplementary figure S9 includes ∆T EC values obtained by direct and indirect EC measurements for a clear comparison. The data from the indirect EC measurements adequately follow the trend of direct EC measurements. Nevertheless, the discrepancies become apparent, especially at high electric fields. The indirect measurements are based on Maxwell relations, and their validity in the case of relaxor ferroelectric materials is still debatable [31]. In addition, for the calculation of ∆T EC , constant specific heat capacity is assumed, although it is temperature and electric field dependent. As a reference, the direct and indirect EC measurements of undoped PMN-10PT are included in supplementary figures S6-S9.
∆T EC values of pristine Mn-doped PMN-10PT ceramics agree well with reported data: compared to the undoped PMN-10PT; the incorporation of the Mn-dopant progressively reduces the ∆T EC and also flattens its temperature dependence [12]. For example, at 30 kV·cm −1 and at about 60 • C the ∆T EC values for undoped PMN-10PT, Mn-0.5 and Mn-1 are 0.89 K, 0.46 K, and 0.51 K, respectively (supplementary figure S6).  lies within the range of measurement uncertainty (±10%), which was previously reported for different pristine samples of the same composition [26], meaning that possible irradiation effects are within those uncertainty limits. Figure 4(b) summarizes the influence of irradiation on the ∆T EC of the Mn-1 sample. While at 5 kV·cm −1 the influence of the irradiation is negligible, it becomes evident at 15 kV·cm −1 : the values of the irradiated samples are lower by 22%-28% within the whole temperature interval. At the field of 30 kV·cm −1 , there is an even more substantial drop in ∆T EC for the irradiated samples, which extends across the whole temperature interval, and the drop is more significant for the higher irradiation dose. For example, at ∼60 • C the ∆T EC values of the pristine, dose-1, and dose-2 irradiated Mn-1 samples are 0.52 K, 0.47 K and 0.41 K. It is worthwhile to note that the ∆T EC in the undoped PMN-10PT obtained by direct measurements remain practically unchanged after exposure to the same doses of irradiation, i.e. 10 17 neutrons·cm −2 and 1200 kGy [26] (see also the results for the reference material in supplementary figure S6). A decrease in the EC response of BaTiO 3 film upon proton irradiation was predicted in a modeling study and attributed to an increased concentration of oxygen vacancies [24].
To inspect possible reversible changes induced by irradiation on the level of defects, the samples were de-aged by annealing at elevated temperatures, i.e. above the temperature of the maximum permittivity. An annealing step could eventually contribute to the relaxation of either irradiation-induced defects or irradiation-altered Mn-induced defect complexes [17]. The pristine and level-2 irradiated Mn-1 samples were annealed at 450 • C for one 1 h and slowly cooled to test this hypothesis. Figure 5 shows the room-temperature polarization vs electric field loops of pristine and irradiated Mn-1 samples before and after annealing. The annealing has very little effect on the P-E dependence of the pristine Mn-1 sample; the values of polarization at zero field and saturated polarization are 1.8 µC·cm −2 and 27 µC·cm −2 at 30 kV·cm −1 . But the P-E loop of the irradiated sample is noticeably influenced by the annealing step. Namely, the saturated polarization increases after the annealing step from 25.7 µC·cm −2 to 27.4 µC·cm −2 at 30 kV·cm −1 . We note that the pinched shape of the P-E loop related to the Mn-dopant-induced defect complexes is, to a lesser extent, influenced by the annealing step, showing a permanent effect of the irradiation on the bias field (see also the inset of figure 3(c)). The slanting of the loop is less pronounced upon annealing, and the polarization at zero field slightly increases from 2.3 µC·cm −2 to 3.2 µC·cm −2 .
As the annealing influences the saturated polarization of the irradiated Mn-1 sample, it was reasonable to expect that a related enhancement would also be observed in the EC effect. Figure 6 shows the EC temperature changes of the pristine and level-2 irradiated Mn-1 samples. Indeed, the ∆T EC of the irradiated sample directly measured at 60 • C and 30 kV·cm −1 increases from 0.41 K to 0.57 K upon annealing. We note that the annealing step almost does not influence the ∆T EC of the pristine Mn-1 sample, the respective ∆T EC at 60 • C are 0.52 and 0.51 K.
Compared to the undoped material, which exhibits no noticeable changes in the measured properties upon exposure to neutron and gamma irradiation, the lower response of the Mn-doped PMN-10PT indicates that the presence of acceptor-oxygen vacancy defect complexes contributes to the lowering of the threshold radiation dose that the material survives without degradation. The increased bias field likely indicates additional domain wall pinning due to the radiation-induced defects upon exposure to the highest fluence of neutrons and gamma rays, 10 17 neutrons·cm −2 and 1200 kGy, respectively.
Sternberg et al showed that annealing introduces the relaxation of radiation-induced defects. The annealing temperature depends on the ionization source and dose; temperatures between 400 • C and 500 • C were needed for fluences of up to 10 18 neutrons·cm −2 , while for gamma radiation, about 100 • C lower temperatures were sufficient to recover the pre-irradiation polarization state of PLZT fully [17]. In our case of Mn-doped PMN-10PT, irradiated with 10 17 neutrons·cm −2 and 1200 kGy, annealing at 450 • C healed the radiation-induced degradation of the EC response. However, the annealing did not influence its bias field, which remained larger than in the pristine material. Whether this is due to the pinning of domain walls by irradiation-induced defects that cannot be healed by annealing or, possibly, to a too-low thermal budget of the annealing step remains to be explored.

Summary and conclusions
We studied the role of manganese-doping in the amounts of 0.5 and 1 mol% in the EC response of relaxor ferroelectric PMN-10PT ceramics upon exposure to high doses of neutron and gamma radiation, up to 10 17 neutrons·cm −2 and 1200 kGy, respectively. When considering EC cooling in environments which may include exposure to sources of ionizing radiation, such as medical diagnostics and therapy or space technologies, the radiation hardness of the material is one of the criteria that need to be fulfilled. Manganese doping of PMN-10PT has been previously shown to broaden and flatten the temperature interval of the peak EC compared to undoped PMN-10PT [12], making the doped material a suitable candidate for cooling applications operating in a wide temperature interval in the vicinity of RT.
The exposure to neutron and gamma radiation had no noticeable effect on the phase composition, microstructure and low-field dielectric properties of Mn-doped PMN-10PT ceramics. Mn-doping contributed to the suppression of the dielectric permittivity and frequency dispersion of the dielectric permittivity peak both in pristine and irradiated specimens. The influence of neutron and gamma radiation was evident in the high-field polarization-electric field loops of Mn-doped PMN-10PT: with a higher Mn content and a higher radiation dose, the loops slightly broadened, and the internal bias field increased, indicating a progressive change in the switching behavior. Such changes were further reflected in the EC response of the Mn-doped PMN-10PT. A decrease in the EC temperature change was evident, especially for the higher doping level at the highest applied field and at the higher irradiation dose. To verify whether the irradiation-induced defects and/or the defects related to manganese doping could be relaxed, an annealing step was introduced. Indeed, annealing restored the saturated polarization, and, consequently, the EC temperature change reached almost the same values as in the pristine Mn-doped PMN-10PT. It should be noted, however, that high-intensity ionizing radiation left an imprint in the high-field polarization of Mn-doped PMN-10PT, evidenced by a slight broadening of the hysteresis loop and an increase of the bias-field, which could not be fully relaxed by annealing.
The study reveals that the interactions of charged defects introduced via aliovalent doping with the defects arising from high-intensity ionizing radiation sources and their influence on the high-field response of relaxor ferroelectrics should be considered in the development of cooling components for use in harsh environments.

Data availability statements
All data that support the findings of this study are included within the article (and any supplementary files). G L B gratefully acknowledges financial support for this research by the Fulbright U.S. Scholar Program, which is sponsored by the U.S. Department of State.