Enhanced aging and thermal shock performance of Mn1.95-xCo0.21Ni0.84SrxO4 NTC ceramics

: Herein, the Mn 1.95- x Co 0.21 Ni 0.84 Sr x O 4 (MCNS) (0 ≤ x ≤ 0.15) based negative temperature coefficient (NTC) materials are prepared by co-precipitation method. The replacement of Mn by Sr plays a critical role in controlling the lattice parameter, relative density, microstructure and electrical properties. The lattice parameter and relative density increase with the increase of Sr content. A small amount of Sr restrains the grain growth and increases the bulk density. Moreover, the ρ 25 , B 25/50 , E a and α values of MCNS ceramics are influenced by the Sr content and ranged from 1535.0-2053.6 Ω ·cm, 3654-3709 K, 0.3149-0.3197 eV and -4.111--4.173%, respectively. The XPS results explain the transformation of MCNS ceramics from n- to p-type semiconductors. The conduction could arise from the hopping polaron between Mn 3+ /Mn 4+ and Co 2+ /Co 3+ in the octahedral sites. The impedance data analysis also discussed the conduction mechanism of the MCNS ceramic, while grain resistance dominates the whole resistance of the samples. Furthermore, the aging coefficient ( △ R / R ) of MCNS ceramics is found to be < 0.2%, which indicates the stable distribution of cations in the spinel. Finally, the MCNS ceramics demonstrate excellent thermal durability with <1.3% of resistance shift after 100 thermal shock cycles.

agent, into the solution mentioned above (1 M) at 50 ℃ under continuous stirring. The whole reaction was carried out for 10 h. Finally, the precipitate was filtered, washed and dried at 80 ℃ for 12 h. The as-obtained precursor was ball-milled for 12 h and calcined in a muffle furnace at 800 ℃ for 4 h. The calcined powder was ground in an agate mortar for 12 h and pressed into pellets with a diameter of 10 mm and thickness of 2 mm. The pellets were sintered at 1175 ℃ for 2 h. Finally, the silver paste was coated on both surfaces of the sintered pellets and heated at 850 ℃ for 20 min.
The A0, A1, A2 and A3 powders were analyzed by thermogravimetric and differential thermal analysis (TG/DTA, NETZSCH STA 449C) in the temperature range of 30 to 1200 ℃ in air. The temperature was increased at the heating rate of 10 ℃/min. The crystalline structure was determined by X-ray diffraction (XRD, Bruker D8 Advance). The X-ray diffractometer is equipped with Cu K α (λ = 1.5418 Å) radiations and a nickel filter. The XRD patterns were recorded in the angular range of 10° to 80° (2θ). The microstructure of the samples was observed by using a scanning electron microscope (SEM, ZEISS SUPRA55VP), equipped with energy dispersive spectroscopy (EDS, Bruker Nano GmbH Berlin, Germany) for elemental analysis. The valence state of cations was investigated by X-ray photoelectron spectroscopy (XPS, Thermo Scientific Escalab 250xi), equipped with Al Kα radiations. The complex impedance spectrum was performed in the frequency range from 20 Hz to 120 MHz, using an impedance analyzer (E4990A, KEYSIGHT Technologies) controlled by a computer. The resistance was measured in the temperature range of -30 ℃ to 70 ℃ by Agilent 34970A digital multimeter. The sample was immersed in absolute alcohol to minimize the temperature fluctuations (< ±0.05 ℃). The aging performance was assessed at 125 ℃ in a temperature-controlled furnace. Briefly, the sample was subjected to 100 thermal shock cycles and the low and high temperatures were fixed at -30 ℃ and 70 ℃, respectively. The dwelling time at each temperature was 10 minutes, whereas the dwelling time at ambient temperature was 15 minutes. The testing equipment and temperature-cycling profile are shown in Fig. 1.   Fig. 2 presents the TG and DTA curves of the A2 hydroxide precursor in the temperature range of room temperature (RT) to 1200 ℃. The DTA curve exhibits an endothermic peak at 125 ℃, and an initial weight loss of 5.07% in the temperature range of RT to 200 ℃, which can be ascribed to the removal of water adsorbed by hydroxide and crystal lattice. In the temperature range of 200 ℃ to 800 ℃, A2 ceramic rendered a weight loss of 15.44% due to the decomposition of precursor and formation of the MCN phase. Moreover, the exothermic peak in the DTA curve indicates the decomposition of the precursor. In the next step, the further increase in temperature (>800 ℃) resulted in a weight loss of 1.82% due to the reduction of Mn 4+ into Mn 3+ [11,20]. Also, a broad endothermic peak at ~1100 ℃ can be attributed to the stable formation of MCNS ceramics due to cation redistribution. The influence of calcination temperature on the phase composition of A2 precursor is shown in Fig. 3. Apparently, the calcination at 700 ℃ resulted in a mixture of cubic Fd3m spinel-type (JCPDS no. 01-1110) and hexagonal P63/mmc perovskite-type phases (JCPDS no. 72-0197). In the temperature range of 800 to 1175 ℃, the A2 precursor exhibited a single-phase cubic spinel structure, which is mainly characterized by distinct diffraction peaks at 2θ = 18.3° (111), 30.1° (220), 35.5° (311), 37.1° (222), 43.2° (400), 53.5° (422), 57.1° (511), 62.7° (440) and 74.1° (533). One should note that the solubility of Sr improved with increasing calcination temperature, resulting in the formation of a single-phase cubic structure. Therefore, 800 ℃ is chosen as the calcination temperature. The XRD patterns of single-phase MCNS ceramics, obtained with different amounts of Sr-doping, are shown in Fig. 4, confirming that the cubic structure is maintained after doping with Sr. Moreover, the typical diffraction peak of MCNS ceramics shifted towards lower diffraction angles with increasing concentration of Sr, which can be ascribed to the larger ionic radius Sr comparing to Mn or Co. Furthermore, the (311) peak of A3 ceramic shifted towards a higher diffraction angle, which is beyond the expectation. However, the observed phenomena is likely to be associated with the conversion of Co 2+ to Co 3+ , where the ionic radius is decreased from 0.745 Å to 0.61 Å [21,22]. The lattice parameter (a) is fitted using a Powdercell software with Least Square Method [23,24], as shown in Fig. 4b and Table 1, which increased from 8.3821 Å to 8.3896 Å with increasing the Sr content from 0 to 0.1, respectively. On the other hand, the lattice parameter decreased to 8.3833 Å at x = 0.15, indicating the solubility limit of Sr (x = 0.1). The X-ray density (ρ x ) can be calculated from the given relationship:

Structure and density analysis
where Z refers to the number of molecules per unit cell (herein, Z = 8), M represents the molecular weight, N denotes the Avogadro constant and V corresponds to the unit cell volume (V = a 3 , cubic phase). The X-ray density of A0, A1, A2 and A3 ceramics increased from 5.251 to 5.276, 5.31 and 5.359 g/cm 3 , respectively. The difference in X-ray density can be explained by the difference in atomic weight of different ions (M Sr ＞M Mn ). The lattice parameter, crystal phase, and X-ray density of A0, A1, A2 and A3 ceramic samples are presented in Table 1.  The bulk density and relative density of A0, A1, A2 and A3 NTC ceramics. The relative density (ρ rel ) can be given as: where ρ b refers to the bulk density, which is determined by the Archimedes method. The change in bulk density (ρ b ) with respect to Sr 2+ content is illustrated in Fig. 5. Obviously, the bulk density of A0 to A3 ceramics linearly increased with increasing Sr concentration from 5.057 to 5.1774 g/cm 3 , respectively. Fig. 5 shows that the relative density of A0, A1 and A2 ceramics linearly increased from 96.3% to 96.8% with increasing Sr content from 0 to 0.1 et. %, respectively, and decreased to 96.4% at x = 0.15.  Fig. 6a and 6b show the SEM images of A2 ceramic after sintering at 1150 ℃ and 1175 ℃, respectively. Herein, the density and grain size increased with increasing sintering temperature. Secondary electron image (SEI) and elemental mapping were used to determine the distribution of different elements, as shown in Fig. 6c. The element maps show that Mn, Co, Ni, Sr and O are homogeneously distributed in A2 ceramic after sintering at 1175 ℃, which is selected as the sintering temperature in subsequent experiments.

Fig. 7
The surface morphology of (a) A0, (b) A1, (c) A2 and (d) A3 ceramics after sintering at 1175 ℃. Fig. 7 presents surface morphology of MCNS ceramics after sintering at 1175 ℃, showing high density and regular morphology. Nonetheless, it can be seen that the grain size of ceramics decreased from 10-12 μm (A0) to 3-4 μm (A2) after doping with Sr, indicating that a small amount of Sr dopant hinders the grain growth during sintering due to the dragging effect [25]. On the other hand, the surplus Sr in A3 ceramic may cluster at the grain boundaries and could improve the transport rate for grain coarsening [26]. Therefore, the A3 ceramic exhibited distinct grain boundaries with some undesirable pores. It is worth emphasizing that an optimal amount of dopant is required to achieve the desired densification and avoid grain coarsening [27,28]. Fig. 8 The relationship between (a) resistivity (ρ) and absolute temperature (T), and (b) ln (ρ) and reciprocal of absolute temperature (1000/T) of A0, A1, A2 and A3 ceramics. The relationship between resistivity and absolute temperature of MCNS ceramics is shown in Fig. 8(a). The resistivity of MCNS ceramics exponentially decreased with increasing temperature, which can be given by the Arrhenius equation:

Electrical characterization, XPS and complex impedance analysis
where ρ 0 represents the resistivity at an infinite temperature (T→∞), T corresponds to the absolute temperature, and B denotes the material constant, as given by Eq. (3), where E a refers to the activation energy and k represents the Boltzmann constant. The B 25/50 constant can be given as: 50 where R 25 and R 50 represent the resistance of ceramic at 25 ℃ and 50 ℃, respectively. Fig. 8(b) shows the relationship between ln(ρ) and reciprocal of absolute temperature (1000/T) of MCNS ceramics. The room-temperature resistivity, thermal constant, activation energy and temperature coefficient of MCNS ceramics are summarized in Table 2. Table 2 The room-temperature resistivity (ρ 25 ), B constant, activation energy (E a ) and temperature coefficient (α) of MCNS ceramics.
As shown in Table 2 The linear relationship between ln(ρ) and 1000/T shows the characteristics of small-polaron hopping transport, which is often observed in Mn-based spinel structures (Fig. 8b). One should note that the MCNS spinel phase contains Mn element with different valence states, which facilitates electron jumping between Mn 3+ and Mn 4+ ions at octahedral sites [2,11]. As mentioned above, the NTC thermistor characteristics are described by the Nernst-Einstein relationship: where N oct refers to the concentration of octahedral sites per cubic centimeter, d represents the jumping distance for charge carriers, υ 0 denotes the lattice vibrational frequency (which is associated with conduction), k refers the Boltzmann constant, e represents the electronic charge, N denotes the concentration of per formula unit of sites, which are available for the charge carriers, NC (1-C) is related to the number of Mn 3+ and Mn 4+ cations at octahedral sites and E a refers to the hopping energy.
In the next step, XPS analysis was carried out to quantify the Mn 3+ /Mn 4+ and Co 3+ /Co 2+ couple and to clarify the mechanism of resistivity variation.   The chemical components and elemental valence of the A0, A1, A2, A3 samples are characterized by XPS spectra (Fig. 9 and Fig.10). In the XPS spectrum of sample A0 from Fig. 9(a), the Ni 2p 3/2 peak and its satellite peak are located at the binding energies of 855.0 eV and 861.0 eV, respectively. While the Ni 2p 1/2 peak and its satellite peak are located at the binding energies of 872.7 eV and 879.0 eV, respectively. These peaks are in agreement with the one reported by Michalska et al [29], implying the only one kind of Ni ion such as Ni 2+ exist in A0 ceramic. In the Sr 2+ doped ceramics, the Ni 2+ ion can also be detected, although the related peaks have about 0.3 eV shift towards lower binding energies compared with sample A0 [30]. The area of each peaks have changed minutly, indicating the constant Ni ion content. From Fig. 9(b) of Co 2p spectra, the 2p 3/2 and 2p 1/2 peaks are fitted by two peaks, which are according to Co in octahedral (B) and tetrahedral (A) sites. The peaks at 780.1 and 795.7 eV are corresponding to the Co in octahedral sites, the peaks at 782.4 and 798.3 eV are assigned to the Co in tetrahedral sites, while the peaks at 786.4 and 802.6 eV are their satellite peaks [31,32]. The respective satellite peaks of 2p 3/2 and 2p 1/2 are fitted using one peak each. The ratios of the respective area under the curves are tabulated in Table 3, and the ratio of Co 3+ and Co 2+ in the ceramic samples can be obtained by using of the area of the satellite peaks. Finally, the ratio of Co in A site and B site (Co A /Co B ) for A0, A1, A2 and A3 ceramics was calculated to be 0.37, 0.41, 0.34 and 0.27 respectively. Hence, the Co 3+ /Co 2+ ratio in A0, A1, A2 and A3 ceramics was found to be 0.15, 0.14, 0.12 and 0.11, respectively. It suggested that the redistribution of Co in the octahedral and tetrahedral sites occurs, as well as conversion of equivalent amount of Co 2+ to Co 3+ , indicating that more Co 2+ are converted to Co 3+ . The experimental data were normalized by the least square procedure and peak intensities of Mn 2+ (641.2 eV), Mn 3+ (642.4 eV) and Mn 4+ (643.8 eV) were calculated [33,34], as shown in Fig. 10. It can be concluded that the ratio of Mn 2+ , Mn 3+ and Mn 4+ ions in different ceramics is equal to the ratio of peak areas, respectively. Hence, the Mn 3+ /Mn 4+ ratio in A0, A1, A2 and A3 ceramics was found to be 1.24, 1.66, 1.67 and 1.40, respectively ( Table 3). The Co 3+ /Co 2+ ratio gradually decreased with increasing Sr, while the Mn 3+ /Mn 4+ ratio initially increased with increasing Sr content, followed by a decrease at x = 0.15. One should note that Sr 2+ can occupy the octahedral or tetrahedral sites in MCNS as the interstitial occupancy is not allowed due to the large ionic size. To maintain the electrical neutrality, the incorporation of Sr leads to the following reaction [15, 35,]: 2SrO → 2Sr Mn , + 2O + V O * * (6) Hence, Sr-doping facilitates the formation of oxygen vacancies. It has already been reported that more oxygen vacancies exist in grains of 0.05% Sr sample than MCN. Moreover, the re-oxidation reaction eliminates oxygen vacancies and produces holes in the grain boundaries during the cooling process, as given below [36,37]: In order to keeping the charge neutrality, the resulting holes could combine with Mn 3+ (Mn 3+ + h * → Mn 4+ ) or Co 2+ (Co 2+ + h* → Co 3+ ) and influences the resistivity according to the Nernst-Einstein relationship. Hence, Sr-doping lowered the resistivity of Sr-doped MCN ceramics for A1 sample, because the Mn 3+ increased significantly while Co 2+ changed negligibly. With increasing the content of Sr, migration of some Co ions from A sites to B sites has occurred. For A2, A3 samples, the Co 2+ in B site increased, combining with an increase Co 2+ /Co 3+ ions in octahedral, and it is well supported also by XRD analysis (decrease in the lattice parameter). To preserve the overall electrical neutrality, some Mn 4+ convert to Mn 3+ , and induces a decrease in the amount of Mn 3+ /Mn 4+ , which increases the resistivity of A3 ceramic [38]. These results are in good agreement with XPS analysis. Moreover, it can be seen that the electrons migrate through the shortest contact point between the particles. Therefore, the pores located at the grain boundaries increase the electron transportation distance between Mn 3+ and Mn 4+ ions according to Eq. 5. Consequently, the long transportation distance in hopping of carriers increases the resistivity and E a , as shown by the A3 ceramic, which is an agreement with the results of the impedance spectroscopy of A3 sample, as follows [39].
For NTC ceramic, the "choice" between n type or p type is largely depended on the small polaron carrier. Yokoyama et al. synthesized the Mn 2-x Co 2x Ni 1-x O 4 spinel type oxide, and concluded that the oxides with x = 0 and x = 0.25 were n-type conductors [40]. The Mn 1.95 Co 0.21 Ni 0.84 O 4 is similar with this composition, and the major carrier is also electron. However, Sr 2+ ions doped into the MCNS may act as the semiconducting acceptors and lead to the formation of conduction holes, followed by the p-type conductor. The transition process can be explained by Eq. 6 and 7. It is worth to notice that the carrier-type can also be calculated using the ratio of Mn 3+ /Mn 4+ ratio through the thermopower Q value, which can be defined as:   Impedance spectroscopy (IS) is a powerful tool for investigation of electrical behavior of ceramic samples, as a successive of semicircle represents the grain effect, grain boundary effect and electrode interfacial phenomena. The complex impedance spectra of A0, A1, A2, A3 ceramic samples obtained at room temperature is shown in Fig. 11. With different Sr-doped content the spectrum displays depressed semicircles arches [42,43]. An equivalent circuit (inset in Fig.11) modelled by ZSimDemo software was used to fit the IS data. Herein, R g and R gb represent the grain resistance and grain boundary resistance, C is capacitance and CPE is constant phase element from grain boundary, which can be defined as: [44]. Herein, j is the imaginary factors and 1 j =−, and ω is the angular frequency ( 2 f  = , f being the frequency in Hz). The parameter n represents the rotation angle according to the impedance response of ideal capacitor, and the value usually ranges between 0.5 and 1. For this reason, CPE is applied as flexible parameter to fit the impedance data. From Fig. 11, the simulated curve fitted the IS data well, reflecting that each impedance spectra consists two parts reflecting the grain and grain boundary effect. Table 4 shows the fitted grain resistance, grain boundary resistance and parameter n for A0, A1, A2, A3 ceramics at RT. The R g and R gb both decrease from A0 to A2, then increase for A3 sample. One can see that the R gb is higher than R g for each samples at RT, suggests the grain boundary resistance R gb dominates the total of MCNS ceramics [45].

Fig.12
Temperature dependence of impedances from R g and R gb for A2 sample. Moreover, Fig. 12 shows the temperature dependence of the R g and R gb obtained by the equivalent circuit fitting for the MCNS ceramic. All of lnR gb and lnR g present obvious NTC characteristics. Besides, the activation energy of grain and grain boundary for ceramic sample can be obtained, and their values are 0.3064 eV and 0.3359 eV, suggesting that the grain boundary need higher activation energy for hopping than the grain [46]. The conductance of the grains and grain boundaries is caused by hole carries hopping. It could be explained by the activation energy, because the activation energy of the grains and grain boundaries is lower than the activation energy of oxygen vacancies [47,48]. Combining with the IS analysis, the conductive mechanism for NTC ceramic would improve the application for industry thermistors.  Furthermore, the aging performance of MCNS ceramics, at 125 ℃, has been assessed by resistance variation and results are shown in Fig. 13. We have observed that the variation of aging coefficient (

Aging performance and thermal shock cycling
) ranged from 0.16 to 1.31% during 1000 h. However, the relative resistance drift is found to be smaller in Sr-doped MCNS at x = 0.1. Fig. 14 shows the relative resistance variation of MCNS ceramics with different numbers of thermal shock cycles. The relative resistance shifts were found to be 1.3-8.9% after 100 thermal shock cycles. The resistance of MCNS ceramics significantly changed with the increasing number of thermal shock cycles but attained an almost constant value after 50 cycles. Hence, a small amount of Sr-doping can significantly improve the aging and cyclic thermal performance of MCNS ceramics. One should note that the single-phase Sr-doped MCN ceramics do not exhibit aging [49]. However, the large amount of dopant results in a deviation from the dense microstructure and compromises the aging performance. Moreover, the cationic vacancies diffuse into B-sites within the grains and increase the resistance. The addition of Sr decreases the concentration of cationic vacancies at the grain boundaries. Consequently, the inhibition of cationic vacancies diffusion significantly improves aging performance [50,51]. XRD patterns of A0 ceramic have been recorded during aging after 0 h, 1000 h and 100 thermal shock cycles. The results revealed that the A0 ceramic maintained the spinel structure after aging for 1000 h (Fig. 15a and 15b), whereas a rock salt phase appeared after 100 thermal shock cycles due to the segregation of NiO phase (Fig. 15c). In the case of A2 ceramic, the cubic spinel structure is maintained after aging for 1000 h and 100 thermal shock cycles. However, the diffraction peak at 2θ = 35.5°, corresponding to (311) planes of the spinal phase, has been shifted towards a higher diffraction angle after 100 thermal shock cycles, which is likely to be associated with the conversion of Co 2+ to Co 3+ . The A2 ceramic surface does not exhibit any noticeable change after long time aging, as shown in Fig. 16b and 16c. Meanwhile, the SEM images of the A2 ceramic were taken after 30 and 50 thermal shock cycles ( Fig. 16d and 16e). However, the pores started to appear after 30 cycles and increased after 50 cycles, as shown in Fig. 16d and 16e. Hence, the thermal shock testing resulted in pore formation and decreased the densification of A2 ceramic. The increment porosity caused by thermal shock testing was reported in the literature 17. Moreover, the grain size of A2 ceramic increased by the thermal shock cycling. The high-resolution Mn 2p 3/2 XPS spectra from A2 samples after aging and thermal shock testing are shown in Fig. 17. As mentioned earlier, the Mn 3+ / Mn 4+ ratio decreased to 1.45 and 1.42 after aging for 200 and 600 h, respectively. Similarly, the Mn 3+ / Mn 4+ ratio reduced to 1.14 and 1.11 after 30 and 50 thermal shock cycles, respectively. Based on Fig. 13 and Fig. 14, the significant resistance shift between aging and thermal shock cycling can be interpreted as: (1) in the case of A0 ceramic under thermal shock cycling, the precipitation of low-conductivity NiO phase resulted in a significant increase in ceramic resistance (8.9%); (2) compared with the aging performance, the thermal shock testing of A2 ceramic resulted in pore formation due to material deterioration, thus influenced the drift resistivity of ceramics; (3) [12,52,53]; and (4) the presence of a large number of pores at grain boundaries induced a high energy barrier and reduced the time between electron scattering events of charge carriers. Therefore, it can be concluded that the investigations based on the XRD, SEM and XPS analysis explained the resistance shift under aging and thermal shock cycling.

Conclusions
In summary, Sr-doped Mn-Co-Ni ceramics have been successfully synthesized by using the co-precipitation method. The MCNS ceramics exhibited a single-phase cubic spinel structure after sintering. Furthermore, the addition of a small amount of Sr improved the density and decreased the resistivity, on the other hand, the excessive amount of Sr (x > 0.1) resulted in higher resistivity due to grain coarsening. Moreover, the Sr-doped MCNS ceramics demonstrated a transition from n-to p-type semiconducting behavior, which can be ascribed to Sr-doping. Impedance analysis revealed that the whole resistance of the samples was dominated by grain resistance, and conductance mechanism is hole carries hopping. The aging and thermal shock rendered a similar effect on resistivity shift according to the Sr content. However, the thermal aging of MCNS ceramics exhibited a rapid resistance shift under thermal shock than the constant temperature. Besides, the thermal shock aging resulted in microstructural evolution from highly dense to the porous structure under thermal shock testing. These results demonstrate that Sr-doping is an efficient strategy to improve the thermal stability of MCNS ceramics. Figure 1 The schematic illustration of thermal shock equipment (a) and temperature-cycling pro le (b).    The bulk density and relative density of A0, A1, A2 and A3 NTC ceramics.  The surface morphology of (a) A0, (b) A1, (c) A2 and (d) A3 ceramics after sintering at 1175 .

Figure 11
Impedance spectroscopy of MCNS ceramics, a) A0 (with the equivalent circuit for tting the impedance data in inset), b) A1, c) A2 and d) A3 ceramics at RT.

Figure 12
Temperature dependence of impedances from Rg and Rgb for A2 sample.

Figure 13
The relative resistance drift of MCNS ceramics at 125 of 1000 h.

Figure 14
The relative resistance drift of MCNS ceramics under thermal shock cycles.

Figure 15
XRD patterns of A0 ceramic after the aging process after (a) 0 h, (b) 1000 h, and (c) 100 thermal shock cycles, and A2 ceramic during aging after (d) 0 h, (e) 1000 h and (f) 100 thermal shock cycles.  The high-resolution Mn 2p3/2 XPS spectra from A2 ceramic during aging after (a) 200 h and (b) 600h, and thermal shock testing after (c) 30 cycles and (d) 50 cycles.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. supplementarymaterials.pdf