Corundum-related polar oxides exhibit a lot of appealing physical characters in transition-metal enriched systems [1-3]. However, the number of double corundum-related polar magnets is very limited compared to chemical-space predictions [4], and thus worthy of further exploration. There a key issue is to lower the synthesis pressure of these materials and reduce the cost in a scaled-up manner, to meet the ultimate goal for practical applications. Stabilization of the high pressure (HP) phase in an isostructural or structurally related matrix by chemical pressure has been proven to be feasible [5-9]. Recently, the LiNbO₃ (LN)-type Mn₂FeNbO₆ has been stabilized in LN matrix by forming a solid-solution [9]. Mn₂FeNbO₆ prepared at high-pressure and high temperature (HPHT, 7 GPa and 1573 K) crystallizes in LN-type R3c with disordered arrangement of Fe and Nb, namely Mn(Fe_0.5Nb_0.5)O₃ [10]. It is antiferromagnetic (AFM) below 90 K with short-range magnetic ordering up to 200 K, and demonstrates pyroelectric response at lower temperature. However, the Mn₂FeNbO₆ product is only around 20 mg/batch at 7 GPa in our multi-anvil press. Chemical pressure can dissolve 18% of the HP-Mn(Fe_0.5Nb_0.5)O₃ in a LN matrix at ambient pressure (AP) in our previous studies [9]. While at 5 GPa, the HP-phase dominated (57%) LN-Mn₂FeNbO₆ solid solution can be obtained in gram level, which indicates that the assistance of external physical pressure favors stabilization of more HP-phase [9]. The interception of Mn₂FeNbO₆ in isostructural LN-host significantly modifies the local structure, resulting in enhanced magnetic interactions. The solid-solution exhibits magnetostriction-induced irreversible lattice evolution around the magnetic transition temperatures, showing ferromagnetic (FM) transitions between 516 K and 554 K [9]. These findings show that chemical pressure and the integration of chemical and physical pressure between isostructural compounds are an effective method to decrease the synthetic pressure for metastable phase prepared at HPHT [9, 11, 12], and further enhance the desired material function by local structure modulation.

Previous research indicates that Mn₂MnMoO₆ (R3, prepared at 8 GPa), a new transitional-metal-only polar corundum oxide, shows interesting magnetism with two magnetic transitions, a large spontaneous polarization, and semiconductor behavior [13]. Unlike Mn₂MnMoO₆, Mn₄Ta₂O₉ [Mn₂(Mn_2/3Ta_1/3)TaO₆] displays different local structural connection style and magnetodielectric coupling in the AP phase (P3c1), and diversity of structure and characters under additional pressure, partially owing to the larger ionic size of 5d⁰ Ta⁵⁺(0.64 Å) than that of 4d⁰ Mo⁶⁺ (0.59 Å) [14-16]. Mn₂MnTaO₆, a predicted but unreported HP phase [4], is expected to adopt the LN-structure similar to Mn₂FeTaO₆ (R3c) [10], but demonstrates different local structure and valence states compared to Mn₄Ta₂O₉. Clearly, the crystal structures and physical properties of these Mn₂BB'O₆ are strongly related to the coordinated environment (atomic-scale local structure) and valance of magnetic ions (electronic structure), providing an ideal platform for chemical-pressure induced phase interception and modification. In this work, chemical pressure was applied to stabilize the HP forms of Mn₂MnMoO₆, Mn₄Ta₂O₉, and Mn₂MnTaO₆ at AP to reveal the competitions between cationic order-disorder degree and chemical and physical pressures, and investigate the diversity of characters with different interaction between ions. Hence, a suitable host phase is crucial.

As shown in Fig. 1, the connection modes of LiTaO₃, Mn₂MnMoO₆, AP-Mn₄Ta₂O₉, and predicted Mn₂MnTaO₆ show certain similarities, where the octahedra are edge- and face-sharing between intralayer and interlayers, respectively, and form honeycomb layered arrangements in the ab-plane. However, the cationic ordering degree is in great difference for those compounds. For example, unlike the two crystallographically cationic positions ordered into four and three Wyckoff sites, respectively. The B-sites for theoretical Mn₂MnTaO₆ are expected to be disordered in LN-type polymorph due to the second order Jahn-Teller effect of d⁰-Ta⁵⁺ [10]. Compared with three candidate compounds, the simple-corundum LiTaO₃ not only has similar octahedral connection pattern, but also owns smaller cell dimensions (a = 5.1546(4) Å, c = 13.7799(13) Å, V = 317.08(6) ų) due to the ionic radius difference between octahedrally coordinated Mn²⁺ (0.83 Å, high spin, HS) and Li⁺ (0.726 Å) [16]. Thus, the volumetrically compressive strain/stress effect on these materials inside a LiTaO₃ matrix, which is believed to be equivalent to exerted physical (mechanical) pressure [9, 11], is expected to stabilize Mn₂MnMoO₆, Mn₄Ta₂O₉ and Mn₂MnTaO₆ (hereafter redefined as Mn(Mn_0.5Mo_0.5)O₃, Mn(Mn_0.33Ta_0.67)O₃ and Mn(Mn_0.5Ta_0.5)O₃ in ABO₃ form for better comparison), and tune their physical characteristics by regulating the cationic order degree in solid solutions. The solid state solutions of (LiTaO₃)_1-x-[Mn(Mn_0.5Mo_0.5)O₃]_x, (LiTaO₃)_1-y-[Mn(Mn_0.33Ta_0.67)O₃]_y and (LiTaO₃)_1-z-[Mn(Mn_0.5Ta_0.5)O₃]_z are presented by LM1, LM2 and LM3, respectively, for succinct description. Fig. S1 (Supporting information) presents the powder x-ray diffraction (PXD) patterns of LiTaO₃-ABO₃ synthesized at AP, where the main peaks of all samples can be well indexed with R3c symmetry, suggesting the absence of cationic ordering at B/B′-site2, otherwise, a double-corundum R3 structure would be formed [4]. Mn²⁺(Mn_0.33²⁺Ta_0.67⁵⁺)O₃ and Mn²⁺(Mn_0.5³⁺Ta_0.5⁵⁺)O₃ can be stabilized/dissolved by LiTaO₃ up to 50%, higher than that of Mn²⁺(Mn_0.5²⁺Mo_0.5⁶⁺)O₃ at 30%, which can be attributed to the average B-site ionic radius deviation (ΔB): ΔB[Mn(Mn_0.5Ta_0.5)O₃] = 0.0025 Å, ΔB[Mn(Mn_0.33Ta_0.67)O₃] = 0.063 Å, and ΔB[Mn(Mn_0.5Mo_0.5)O₃] = 0.12 Å. Here ΔB is defined as:

where r(G) and r(Q) are the B-site radii of guest and matrix material, respectively, n represents the number of B-site atoms. Here, r(G) denotes r(Ta⁵⁺) = 0.64 Å, r(Mn³⁺, HS) = 0.645 Å, r(Mn²⁺, HS) = 0.83 Å and r(Mo⁶⁺) = 0.59 Å; r(Q) = r(Ta⁵⁺) = 0.64 Å at octahedral surrounding [16].

In the obtained LiTaO₃-ABO₃ solid solutions, the B-site Mn, Ta and Mo are randomly distributed as Mn(Mn_0.33Ta_0.67)O_3, Mn(Mn_0.5Ta_0.5)O₃ and Mn(Mn_0.5Mo_0.5)O₃ in the LiTaO₃ matrix. The peak shifts toward lower angle compared with those of LiTaO₃ with the increase of x, y and z as observed in Fig. S1, which illustrate the expansion of cell volume and successful formation of solid solution, originated from the larger ionic radius of Mn²⁺ (0.83 Å, HS) than that of Li⁺ (0.76 Å) at the A-site, and comparable ionic size between Ta⁵⁺ and Q^~4+. Fig. 2, Figs. S2 and S3 (Supporting information) display the Rietveld refinement plots of a series of AP-made LiTaO₃-ABO₃. Detailed crystallographic parameters and structural information are listed in Tables S1–S3 (Supporting information). Li_1.6Mn_2.2Mo₃O₁₂ (Pnma) related impurity [17] appears to be around 8.90(1) and 6.23(2)% for LM1 at x = 0.3, 0.4, respectively. A few unidentified impurities in LM2 and LM3 appear at y = 0.5 and z = 0.5, respectively. HPHT synthesis was thus applied to reduce the impurity of LM1 with x = 0.3 and 0.4 from 1 to 3 GPa, and further raise the solid solution limit of the HP component as applied in LiNbO₃-Mn₂FeNbO₆ system [9]. Surprisingly, the additional physical pressure did not improve the purity for LM1 (Fig. S4 in Supporting information) as expected, implying that the cooperation of chemical and applied physical pressure is not enough to overcome the size/charge effect, namely the difference of ionic radius and charge between cations to cause ionic ordering. Figs. S5a–c (Supporting information), respectively shows the x, y and z-dependent cell evolution for LiTaO₃-ABO₃, where the cell parameters show different variation tendency for different B-site ions. For LM1, the c-axis gradually increases with incremental x. In contrast, a almost remains constant when x increases from 0.2 to 0.3, which suggests that the crystal structure is more elastic along the c-axis [18]. As for LM2, although the overall cell evolution trend of V is similar to that in Fig. S5a, the a and c show abnormal change at 0.4 and 0.2, respectively, and deviate from the increasing linearity. This is probably due to the suppression from chemical pressure like physical pressure, and change of the spin state of magnetic ion [19]. The variation of cell parameters for LM3 bears different trend compared with the other two series, in that the a, c and V are nearly linear increasing with enlarged z, and roughly follow the Vegard's law [20]. The smallest size difference between Mn³⁺ and Ta⁵⁺ at M-site in LM3, in which similar cell distortion tendency is generated with increasing substitution, is responsible for the different variation trend of lattice parameters to the other two series. The two kinds (three short and three long) of A-O distances for LiTaO₃-ABO₃ are listed in Tables S1–S3 (Supporting information). The AO₆ octahedra are very distorted in LM1, LM2 and LM3 with 10% guest phase, the short and long A-O bond lengths are 1.950(20) and 2.232(18) Å, 2.052(14) and 2.363(16) Å, and 2.028(6) and 2.396(11) Å, respectively, giving large octahedral distortion parameters [21]. Δ_Li/Mn1 of 45.47 × 10⁻⁴, 49.6 × 10⁻⁴, and 69.2 × 10⁻⁴ due to the low Mn content (10%) at the A-site, which is, however, insufficient to affect the overall structural distortion to generate an highly ordered structure than the host phase LiTaO₃, as previously reported for (Li_1-xMn_x)(Fe_x/2Nb_1-x/2)O₃ [9]. As the Mn content increases to x = 0.4 and y = 0.5 in LM1 and LM2, the large discrepancy (0.275 and 0.254 Å) between the two kinds of < Li/Mn-O > bond lengths (2.018(19)−2.293(17) Å and 2.003(13)−2.257(14) Å) brings secondary large Δ_Li/Mn1 of 40.67 × 10⁻⁴ and 35.6 × 10⁻⁴, which can be attributed to the fact that, the chemical pressure stabilized crystal structure has reached its solid-solution limit, and the competition between thermodynamic stability and chemical pressure has reached an equilibrium. The Δ_Li/Mn1 of LM3 regularly decreases with the increase of A-site Mn²⁺ ions, since that the distortion will move with the hopping of e_g electrons between Mn²⁺ and Mn³⁺, which makes the maximum Δ_Mn2/Ta (53.4 × 10⁻⁴) and minimum Δ_Li/Mn1 (0.9 × 10⁻⁴). In addition, it is noticeable that the BO₆ octahedra in LM1 are more regular than those in LM2 and LM3. The difference value between shorter and longer B-O bonds gradually decreases from 0.088 Å to 0.033 Å with incremental x in LM1.

The variation of d_A-B distance between face-sharing AO₆-BO₆ octahedral pairs along the c-axis for LM1 is shown in Fig. 3, which firstly shows an increasing and then decreasing with x from 0.1 to 0.4. As shown in Figs. S6 and S7 (Supporting information), for LiTaO₃-ABO₃ (B = Mn_0.5Ta_0.5, Mn_1/3Ta_2/3), the B-site is occupied by Ta/Mn, the variation of d_A-B exhibits remarkable difference from that in LM1. The d_A-B evolution is more regular, decreasing from 3.03 Å to 2.87 Å and 3.05 to 2.91 Å with the substituted ions increase in B = Mn_0.5Ta_0.5, Mn_1/3Ta_2/3, respectively. The lattice parameters for LM1 and LM2 show nonlinear evolution in some region, disobeying the Vegard's law [20], due to the larger ionic radius difference at the mixed B-site than that in LM3. Thereby, it is becoming more difficult to trap ABO₃-phase within LiTaO₃ and reaches the solid-solubility limit with incremental guest phase, in that the limited chemical pressure is not enough to overcome the size-difference effect to further boost the contents of guest phase.

Although the reported Mn(Mn_0.33Ta_0.67)O₃ [14, 15] and Mn(Mn_0.5Mo_0.5)O₃ [13] are in different space groups from LiTaO₃ (R3c), their local octahedral connection patterns are similar, and only differ in the cationic ordering degree (Fig. 1). Our findings demonstrate that the experimental or theoretical high-pressure phase can be intercepted at AP by solid fusion in host matrix with similar arrangement of local structure motifs rather than strictly isostructural lattices. Here the volumetric difference between LiTaO₃ (317.08(2) ų) and Mn(Mn_0.5Mo_0.5)O₃ (R3–343.25(1) ų), and Mn(Mn_0.33Ta_0.67)O₃ (P3c-353.55(1) ų) reaches 26.16 and 36.47 ų, respectively, which, together with the mono-B site inLiTaO₃, generates compressive effect (positive chemical pressure), and induces lattice stress to break the energy barrier between ionic order and disorder and thus stabilize the phases to R3c at AP. It is noticeable that the chemical pressure, which derives from the volumetric difference between LiTaO₃ with one crystallographic B site and desired candidate phases with ordered B/B' sites, prefers to stabilize the phase with short-range random order than long-range order, which is the preference of physical pressure [22]. For Mn(Mn_0.5Ta_0.5)O₃, the comparable ionic radius difference at A-site and close ionic radius between Mn³⁺ and Ta⁵⁺ at B-site offer an effective chemical pressure to stabilize Mn(Mn_0.5Ta_0.5)O₃ in LiTaO₃ up to mole ratio of 50%. Understandably, the positive (compressive) chemical pressure exerted on the high-pressure polymorphs equals to the negative (tensile) chemical pressure experienced by LiTaO₃. So, in order to evaluate the chemical pressure, which stems from the volumetric difference between LiTaO₃ and the solid solutions, the P-V curves were calculated for LiTaO₃ by the Murnaghan equation of state [23] and the detailed results are shown in Fig. S8 (Supporting information), which indicate equivalent pressure up to 3.9 GPa, for example to stabilize LM3 (z = 0.5), which further proved the feasible trapping for metastable phase with chemical pressure at AP. The measurements of magnetic and dielectric response were executed to further study the properties with the change of ionic ordered degree modulated under chemical pressure.

Fig. 4 illustrates the variation tendency of temperature-dependent magnetization for LiTaO₃-ABO₃ series. The zero-field cooled (ZFC) and field cooled (FC) curves for all samples show the like-λ form, which implies the possibility of spin-glass or magnetic competition between FM and AFM phases. The negative Curie-Weiss temperature θ, determined by Curie-Weiss fitting from the 200-300 K data, demonstrates that all samples are AFM dominated. The detailed magnetic parameters are listed in Tables S4–S6 (Supporting information). It is noticeable a nonlinear change of T_N between 33.4–37.7 K, 39.2–41.1 K and 38.2–41.1 K for B = Mn_0.5Mo_0.5, Mn_0.33Ta_0.67, and Mn_0.5Ta_0.5, respectively. For LM1 and LM2, compared with the reported guest phases Mn(Mn_0.5Mo_0.5)O₃ and AP-Mn(Mn_0.33Ta_0.67)O₃, the change of cationic arrangement directly modifies the magnetic interactions between Mn²⁺-O²⁻-Mn²⁺. The increased disordered degree and lower content of magnetic ions induce local structural relaxation and weaken magnetic interaction between ions. Hence, the magnetic transition at 19 K rooted in the A-sublattice Mn²⁺ in Mn(Mn_0.5Mo_0.5)O₃ disappears in LM1. The magnetic transitions in LM1 and LM2 are lower than those in Mn(Mn_0.5Mo_0.5)O₃ and AP-Mn(Mn_0.33Ta_0.67)O₃ from 47 K and 102 K to 33.4–37.7 K and 39.2–41.1 K, respectively.

The nearest interlayer Mn distance, bond angle of Mn-O-Mn, and the change of MnO₆ octahedral distortion accompanied with anomalous lattice parameter variation, are the inducement of T_N change. The variation of θ of LM1 first increases and then decreases with x increasing. In contrast, the θ evolution behaviors of LM2 and LM3 behave oppositely. The competition between AFM and FM directly reflects the θ value, where the dominance of AFM interactions leads θ decreasing and vice versa. The growth of disordered degree between ions of unlike elements enhances the complexity of magnetic interaction to strengthen the frustration [24, 25]. LM2 owns a relatively lowest magnetic frustration f (defined as |θ|/T_N) than those of the other two systems, as the lowest content of Mn ions in B-site will weaken the strength of magnetic interactions between different atomic sites. The calculated μ_eff values via all site HS-Mn²⁺ and HS-Mn²⁺(A-site)/HS-Mn³⁺(B-site) of LM1 (HS-Mn²⁺), LM2 (HS-Mn²⁺) and LM3 [HS-Mn²⁺(A-site), HS-Mn³⁺(B-site)], respectively, are lower than the theoretical values because of the frustration between magnetic lattice and ionic disorder [26].

Isothermal magnetization curves for LiTaO₃-ABO₃ are illustrated in Figs. S9–S11 (Supporting information). The FM component obviously exists in all samples at low temperature region. The hysteresis loops of isothermal magnetization gradually emerge at 10 K, due to the enhancement of magnetic phase competition with enriched magnetic ions. Nevertheless, the AFM interactions are still dominant. Above T_N, the isothermal magnetization plots exhibit abnormal variations when the metastable guest phases reach ultimate values for LM1 and LM2. The isothermal curves of LM1 (x = 0.4) and LM2 (y = 0.5) deviate linearity and exhibit a hysteresis loop at 50 and 45 K, respectively, indicating weak interactions of short-range magnetic ordering or canting AFM.

The temperature-dependent relative dielectric constants (ε_r) and dielectric loss (tanδ) at different frequencies without field for selected LM1 (x = 0.4) are illustrated in Fig. S12a (Supporting information). The ε_r increases from 17.5 to 18.5 from 10 K to 200 K, which almost remains constant at different frequencies and shows a weak dependence on temperature and frequencies. The tanδ plot also exhibits a weak temperature correlation at the same temperature range, where the electrodes induce a large loss at 1 MHz than at low frequencies. These phenomena derive from the strengthened restraint of cooperation conversion between local electric polarization as the A- and B-site ions are disorderedly aligned [27], which induce the absence of platform in the curves of ε_r dependent T as same as in Mn(Mn_0.5Mo_0.5)O₃ [13]. At the low disorder degree and relative weak restraint effect, the ε_r values of LM2 (y = 0.4) and LM3 (z = 0.4) show relatively clear differences, which are about 4 and 6 between 10 K and 200 K. However, the charge thermal motion is motivated and leads a significant increase for ε_r with a divergence in frequency with T increasing for all selected samples. Besides, the dielectric relaxation of LM1 (x = 0.4) can be observed near room temperature in Fig. S12a (bottom), which may be from the polarization difference because of the charge thermal motion. The dielectric relaxation phenomenon of LM2 (y = 0.4) and LM3 (z = 0.4) occur at T about 40 K, which is much lower than that in LM1, as the weak polarization-temperature dependency with frequencies changing. Figs. S12a and S13a (Supporting information) demonstrate the temperature-dependence of ε_r at different additional magnetic field for LM1 and LM2. No anomalous dielectric peaks are observed around the magnetic transition temperatures under additional field, and the permittivity hardly changes over the test ranges with and without field, indicating the absence of magneto-electric and dielectric coupling.

The real part permittivity ε′ and imaginary part of dielectric permittivity (ε″) with frequency at selected temperatures are measured for selected samples as illustrated in Figs. S12–S14 (Supporting information). The ε′ values of all samples show more obvious frequency dependence with temperature increase, and the ε′ values decrease with increasing frequency. The ε″ of LM1 (x = 0.4) appears two frequency dependent variation trends around 250 K as dividing line. Above 250 K, where dielectric relaxation occurs, the platform-like area appears and shifts to high frequency region. For LM2 (y = 0.4) and LM3 (z = 0.4), the ε″ exhibits similar variation from 10 K to relaxation peak of high-frequency about 90 K, then changes from negative to positive correlation with frequency. The tanδ vs. frequency curves are displayed in Figs. S15a–c (Supporting information). According to the relationship between tanδ and frequency, a linear fitting was performed with Arrhenius relation f = f₀exp(-E_a/k_BT) [28], where f₀ is the frequency factor, E_a is the activation energy and k_B is the Boltzmann constant. As shown in the insets, the fitting results for LM1 (x = 0.4), LM2 (y = 0.4) and LM3 (z = 0.4) are E_a = 0.44, 0.022, 0.027 eV and f₀ = 6.97 × 10¹¹, 3.2 × 10⁶, 6.9 × 10⁶ Hz, respectively. The higher E_a of LM1 than Mn(Mn_0.5Mo_0.5)O₃ [13] demonstrates the inducement of relaxation from ions hopping to the difference of polarization at boundaries and strengthened interaction between polarons. However, the relaxation for other selected samples derives from the electron transfer between magnetic ions.

In summary, the polar R3c-LiTaO₃ is used as a matrix to stabilize a series of high-pressure metastable polar and magnetic polymorphs of Mn(Mn_0.5Mo_0.5)O₃, Mn(Mn_0.33Ta_0.67)O₃ and Mn(Mn_0.5Ta_0.5)O₃ by chemical pressure in solid solution. The volumetric difference and likely local structural patterns between LiTaO₃ and targeted high-pressure phases are beneficial to modify the cationic ordering degree and introduce magnetic competition between ferromagnetic and antiferromagnetic interactions. The solid solution limitation performs a significant difference among three series due to the ionic size and electronic structure effect among different compounds. The structural modulation influences the magnetic interaction and change the magnetic ordering temperature and frustration. The dielectric behaviors indicate that both the A and B-site atomic substitutions, which contain strong spin orbit coupling ions, can significantly affect the dielectric relaxation mechanism and the interaction between polar clusters. These discoveries proposed a feasible way to trap metastable high-pressure phase in a like-matrix by chemical pressure, and thus tune the physical properties and soften the synthesis conditions in a cost-effective way to significantly scale up the metastable products.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

This work was financially supported by the National Natural Science Foundation of China (NSFC, Nos. 21875287, 22090041, 22105228 and 11804404), the China Postdoctoral Science Foundation (No. 2021M693603).

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.cclet.2022.03.078.