The XRD patterns of polycrystalline Ca₃CoNb₂O₉ samples are presented in Fig.  2. Refinement of the observed diffraction peaks using the structure parameters obtained from previous neutron diffraction^[17] gives a final weighted residual error R_wp = 2.64, which means that our XRD data match well with the calculated results. The obtained structure parameters a = 9.5988(9)  Å , b = 5.4576(3)  Å , and c = 16.8643(9)  Å are also consistent with previous results.^[17] No additional impurity peak is found in the whole diffraction range from 5° to 120° indicating that there is no obvious impurity phase in our samples.

Fig.  2.

	Figure Option ViewDownloadNew Window
	Fig.  2. The room-temperature XRD pattern of polycrystalline Ca3CoNb2O9 samples. The red solid cycles are our experimental data. The green solid line is the calculated XRD pattern for powder Ca3CoNb2O9. The vertical lines indicate the reflection positions for Ca3CoNb2O9. The difference curve is also presented as the black solid line.

The dc susceptibility and its inverse above 2  K under an external field of 0.1  T for Ca₃CoNb₂O₉ are shown in Fig.  3.

Fig.  3.

	Figure Option ViewDownloadNew Window
	Fig.  3. The dc susceptibility (χ = M/H) and its inverse measured at 0.1  T. The solid line is a linear Curie– Weiss fitting of χ in a temperature range 100  K– 300  K.

It is shown from Fig.  3 that the 1/χ deviates from the Curie– Weiss curve at around 50  K. The susceptibility data above 100  K can be fit by the Curie– Weiss formula: χ _M = C/(T − θ _CW)+ χ ₀, where C is the Curie constant, θ _CW is the Weiss constant, and χ ₀ is the temperature-independent paramagnetic susceptibility. The fitting gives an effective moment of 5.1  μ _B and a Weiss constant of θ _CW∼ − 55  K. Considering the contribution from the orbital angular momentum, the value of the moment is reasonable for the Co²⁺ ion in the highspin state.^[18] The negative Weiss constant reflects dominant antiferromagnetic interactions between Co²⁺ ions. The large Weiss temperature and the absence of magnetic ordering down to 2  K imply the presence of strong frustration in this system.

To study the ground state magnetism of Ca₃CoNb₂O₉, we performed susceptibility measurements down to 0.46  K with ZFC and FC methods. As shown in Fig.  4(a), under a 0.01-T external field, both ZFC and FC susceptibility increase with cooling. With further cooling, a cusp occurs at around 1.45  K, which indicates a transition from the paramagnetic phase to an ordered state. The susceptibility is less than 0.2- emu/mol Co²⁺ in the ordered state, which suggests a longrange order antiferromagnetism (LROAFM). Combined with the large Weiss constant obtained above, the frustration factor f (= | θ _CW | /T_N) is about 38, which is comparable to that of strong frustrated regular TLHAF Ba₃CoNb₂O₉ (θ _CW= − 51  K and T_N1 = 1.36  K).^[14] In contrast to the regular triangular lattice antiferromagnet Ba₃CoNb₂O₉, Ca₃CoNb₂O₉ has DM interactions and spatially anisotropic exchange interactions. Theoretical studies suggested that both interactions release frustration and enhance the LROAFM in general.^{[19, 20]} For Ca₃CoNb₂O₉, however, the anisotropy does not lift up magnetic ordering temperature significantly, which suggests that the DM interactions and spatial anisotropy are weak.

Fig.  4.

	Figure Option ViewDownloadNew Window
	Fig.  4. (a)– (f) The dc susceptibility down to 0.46  K after ZFC and FC measured at external magnetic fields: 0.01  T, 0.1  T, 0.2  T, 0.5  T, 0.65  T, and 1  T, respectively.

The magnetically ordered ground states were expected for the insulating triangular lattice antiferromagnets, including isotropic Heisenberg, ^[4] XXZ, ^[21] and distorted^[11] models. Experimentally, almost all triangular antiferromagnets have been found with LROAFM at low temperatures, except κ -(BEDT-TTF)₂Cu₂(CN)₃, which is probably closer to a Hubbard model.^[22] Therefore, the LROAFM state is natural for our triangular lattice antiferromagnet.

At T ∼ 1.2  K, which is slightly below T_N, deviation appears between the ZFC and FC data under 0.01-T field [Fig.  4(a)]. We further measure dc susceptibility after FC and ZFC at higher fields [see Figs.  4(b)– 4(f)]. Under a 0.1-T field, the FC susceptibility is suppressed slightly below T_N, while the suppression effect is much stronger for the ZFC data, resulting in an enhancement of the deviation between the FC and ZFC data. With further increasing fields, the discrepancy is gradually suppressed monotonously and completely removed by a field of 1  T, while the position of T_N remains at 1.45  K.

The deviation between ZFC and FC susceptibility data at low measurement fields was reported in some spin-canted antiferromagnets. For example, in DTLAFM Cu₂(OH)₃(C_mH_{2m+ 1}COO), m = 7, 9, 11, ^[23] the DM interactions were suggested to play an essential role to the spin canting. A similar situation may occur in Ca₃CoNb₂O₉, where the DM interactions assist the canting of antiferromagnetic moments.

The bifurcation of ZFC and FC susceptibility was also seen in spin glass system with spin-freezing. Earlier theoretical studies suggested that the spin glass phase usually occurs in frustrated systems with quenched disorder.^{[24, 25]} However, the spin glass can also be realized in the system with competing ferro- and antiferro-magnetic bonds distributed in a certain rule, rather than a random way.^[26] Experimentally, the spin freezing transition has been found recently in several nominally disorder-free frustrated systems such as Y₂Mo₂O₇, ^{[27, 28]} Gd₃Ga₅O₁₂, ^[29] ZnCr₂O₄, ^[30] CsVCl₃, ^[31] Cu₂(OH)₃(C_mH_{2m+ 1}COO), m = 7, 9, 11.^[23] The spin glass and LROAFM phases coexist in the latter three materials. Whether disorder is essential to the generation of the spin glass in a strongly frustrated magnet remains to be established. Structurely, Ca₃CoNb₂O₉ is a site-ordered system which has been confirmed by neutron diffraction experiment.^[17] However, we cannot exclude the possibility of mixing tiny amounts of Nb⁵⁺ /Co²⁺ on the perovskite B-sites. It has been expected that the spin glass can be stabilized in geometrically frustrated antiferromagnets even with smallest degree of disorder.^[32] After all, based on our current dc susceptibility data, we cannot exclusively determine the origin of observed deviation between the ZFC and FC susceptibility. Specific heat and ac susceptibility experiments down to ³He temperatures may help to resolve this issue.

To investigate the evolution of magnetism with external field in Ca₃CoNb₂O₉, the dc magnetization was measured at 0.46  K [Fig.  5]. The magnetization increases linearly with field and then changes slope around 6  T, which corresponds to a crossover from a long-range ordered state to a fully polarized state, as in Ba₃CoNb₂O₉.^{[14, 33]} Above 6  T, though the magnetization still increases, it has a tendency to saturate around 2  μ _B/Co²⁺ . Due to spin– orbital couplings and the crystal field, the ground state of Co²⁺ ion coordinated in octahedral environment is a Kramers doublet with an effective spin of 1/2.^{[34, 35]} The low spin state of Co²⁺ has also been observed in many other TLAFM with octahedral Co sites, such as Ba₃CoNb₂O₉, ^[14] Ba₃CoSb₂O₉, ^[12] and CsCoCl₃.^[2] We further plot the first derivative of the magnetization curve as a function of field. Below 6  T, two valleys are observed in the derivative curve, which is reproducible on different samples. The first one located around 0.7  T, which is consistent with the field where the bifurcation between ZFC and FC susceptibility disappears [Fig.  4(b)– 4(f)]. The second one occurs at around 2.5  T with the magnetization of 0.75  μ _B/Co²⁺ , which is close to 1/3 of the saturation magnetization (M_s). The possible appearance of a 1/3 magnetization plateau is consistent with the fact of strong quantum fluctuations from frustration and small effective spins in this triangular lattice.^{[7, 8]} However, we are aware that the magnetization plateau is not strong, which may be because our samples are randomly aligned powders. Single crystal samples are needed to confirm this observation.

Fig.  5.

	Figure Option ViewDownloadNew Window
	Fig.  5. The magnetization curve of Ca3CoNb2O9 at a fixed temperature 0.46  K. The solid squares are M– H curve and the empty ones are the first deviation of magnetization curve as a function of field. The vertical arrows indicate the positions of the valleys.

Our study revealed that the ground state for Ca₃CoNb₂O₉ is a LROAFM coexisting with spin canting or spin glassiness. However, for a regular triangular lattice model, the ground state is coplanar 120° antiferromagnetic order, which has been confirmed in Ba₃CoSb₂O₉, ^[12] Ba₃NiNb₂O₉, ^[13] and Ba₃CoNb₂O₉.^[14] The different ground state of Ca₃CoNb₂O₉ may be a result of the interplay of strong frustration and anisotropic interactions, and provides possibility of searching for new novel quantum states in this geometrically frustrated magnet. Moreover, the 1/3 plateau in magnetization for a DTLAF is a macroscopic manifestation of quantum effect predicted by theoretical studies.^[11] To be more unique, this novel magnetic state can be stabilized by a very small field (2.5  T) which is easily accessible by experiments, and provides a promising material for further investigation of magnetization plateaus which are very interesting theoretically.