Universal spin-glass behaviour in bulk LaNiO₂, PrNiO₂ and NdNiO₂

The nickelates are widely believed to be promising candidates for unconventional superconductors because of their very similar structures and possibly related electronic states to the cuprates. Superconductivity was realized in nickelate thin films. Nd_1−x Sr_x NiO₂ films grown on SrTiO₃ were found to be superconducting with T_c = 9–15 K [1], and corresponding Pr_1−x Sr_x NiO₂ films with T_c = 7–12 K [2]. Very recently, Sr- and Ca-doped LaNiO₂ thin films [3, 4] have also been shown to be superconducting. These discoveries have aroused intensive research and discussions, but several open questions remain.

First, superconductivity is absent in bulk samples, e.g. in bulk Nd_1−x Sr_x NiO₂ [5, 6] and Sm_1−x Sr_x NiO₂ [7], and no signs of long-range magnetic orders have been observed. This discrepancy between non-superconducting bulk samples and superconducting thin films has not been well understood so far. Off-stoichiometry defects or chemical inhomogeneity in bulk samples are likely suspects to hamper superconductivity, while the role of the SrTiO₃ substrate has also been considered [8–11]. Strain effects have been frequently invoked to explain the occurrence of superconductivity in thin films [9, 12]. However, the 3% larger lattice constant along the c direction in thin films that must affect the Ni ${d}_{{z}^{2}}$ orbital and is considered to be crucial for nickelate superconductivity [13], may be alternatively ascribed to the details of the film-growth process, in a similar way to MgTi₂O₄ films where a change of the out-of-plane lattice parameter alone can induce superconductivity [14]. Second, the parent phase RNiO₂ shows several differences from the isostructural parent compound CaCuO₂, which can be a high-temperature superconductor upon doping [15, 16]. Although NdNiO₂ and CaCuO₂ have the same infinite-layer structure and the same 3d⁹ configuration, NdNiO₂ is a conductor [17] (or a weak insulator [1, 5, 6, 17–19]) without any long-range magnetic order, whereas CaCuO₂ is an antiferromagnetic (AFM) Mott insulator [20]. The superconductivity in the nickelate may therefore be far away from the well-known scenario in high-temperature superconductors, where AFM spin fluctuations have been suggested to act as the medium to promote superconductivity. More distinctions, such as a larger charge transfer gap and smaller AFM exchange energy have been found in experiments [21, 22] and theoretical calculations [23–27]. Moreover, the role of the 4f electrons of the R ions in RNiO₂ is still unknown. Theoretical calculations propose that the 4f electrons have an important influence on the electronic structure through 4f–5d interactions [28–30], although the doped LaNiO₂ thin films have recently also been shown to superconduct [3, 4].

A good starting point to clarify these questions would be to study the magnetism of the parent phases. The difficulty we face is to obtain bulk samples of RNiO₂, especially PrNiO₂, and to distinguish the intrinsic magnetism therein. In this work, we synthesized bulk LaNiO₂, PrNiO₂ and NdNiO₂ samples and systematically investigated their magnetic properties. We show that magnetic memory effects and frequency-dependent spin freezing peaks in alternating current (AC) magnetization reveal a universal spin-glass behaviour in RNiO₂.

The synthesis of RNiO₂ (R = La, Pr, Nd) samples was conducted in two stages, according to the same procedure as in previous reports [5–7, 17, 31]. Here, we put an emphasis on bulk PrNiO₂, which is obtained for the first time up to our best knowledge. In the first stage, high-quality PrNiO₃ samples were grown under a high-pressure oxygen atmosphere [32]. In the second stage, the precursor PrNiO₃ samples were topochemically reduced by applying the following route, mixed with CaH₂, sealed in quartz tubes and heated at 270 °C for 40 h. Afterwards, the mixtures were washed by NH₄Cl in ethanol to remove the extra CaH₂ and Ca(OH)₂. The resulting black powders were PrNiO₂. LaNiO₂ and NdNiO₂ were synthesized in the same way. The pictures taken by a scanning electron microscope (SEM) indicate that all these samples consist of 1–5 μm sized particles, as shown in the insets in figure 1. Last but not least, the samples were stored in sealed centrifuge tubes and handled in the N₂ atmosphere in order to avoid degradation caused by water or oxygen in the air.

Zoom In Zoom Out Reset image size

We made an elemental analysis with both energy-diffraction spectra (EDX) and thermal gravity (TG). The EDX measurements as well as the SEM pictures were performed on a Zeiss GeminiSEM 450 with a beam energy of 10 keV. The obtained atomic ratios were R:Ni = 1.01/1.04/1.02 for La/Pr/NdNiO₂, respectively. The TG analysis was done to determine the oxygen content in the respective oxides. Considering that the RNiO₂ samples would be totally reduced to R₂O₃ and Ni by H₂ in 700 °C, and by comparing precisely the mass before and after the reduction, the oxygen contents were deduced to 2.05/1.95/1.99 ± 0.04 for La/Pr/NdNiO₂, respectively. Altogether, the chemical composition of RNiO₂ we obtained is very close to the nominal compounds RNiO₂ within experimental resolution.

The x-ray diffraction (XRD) measurements were performed on a Bruker AXS D8 Advance diffractometer (Bruker AXS GmbH, Karlsruhe, Germany), equipped with a Ni-filtered Cu Kα radiation and a 1D LynxEye PSD detector. The Rietveld refinement analysis [33] of the diffraction patterns was performed with the package FULLPROF SUITE (version March-2019). All the magnetic measurements including the direct current (DC), AC magnetic susceptibility and memory effect, were performed in a magnetic properties measurement system (quantum design MPMS-3).

We performed XRD measurements to characterize the crystal structures of the as-synthesized RNiO₂ samples. The corresponding data are shown in figure 1, together with refinements via the Rietveld method and the calculated Bragg peaks. As illustrated in figure 1(a), after extracting the apical oxygen in RNiO₃, the NiO₂-planes in RNiO₂ are separated by rare-earth metal ions layer by layer. The reduction might in principle be incomplete or introduce crystalline disorder or even destroy the NiO₂-planes into elemental Ni. However, the calculated curves fit well with the experimental ones, and all RNiO₂ compounds can be successfully refined as tetragonal phases with the same space group P4/mmm. These XRD results indicate that no significant distortions were introduced in the NiO₂-plane by the reduction, and the compounds can indeed be considered as the infinite-layer phase. The crystal parameters obtained from the Rietveld refinements are presented in table 1. The c-axis parameters of LaNiO₂, PrNiO₂ and NdNiO₂ are 3.363 $\mathring{\text{A}}$, 3.285 $\mathring{\text{A}}$ and 3.279 $\mathring{\text{A}}$, respectively. They are close to the reported values for bulk samples [5, 17, 31, 34, 35], but 3% smaller than in thin films [1, 2]. The different c values in our samples can be simply attributed to the different radii of rare-earth metal ion for coordination number N = 8 that r(La³⁺) = 1.15 $\mathring{\text{A}}$, r(Pr³⁺) = 1.11 $\mathring{\text{A}}$ and r(Nd³⁺) = 1.10 $\mathring{\text{A}}$ [36].

Table 1. Systematic comparison between RNiO₂ (R = La, Pr, Nd) and Pr₄Ni₃O₈ [39].

Compound	LaNiO2	PrNiO2	NdNiO2	Pr4Ni3O8
R3+ configuration	4f  0	4f  2	4f  3
a at 300 K ($\mathring{\text{A}}$)	3.9582(1)	3.9403(3)	3.9279(4)
c at 300 K ($\mathring{\text{A}}$)	3.3632(3)	3.2845(8)	3.2793(11)
a/c at 300 K	1.18	1.20	1.20
Rp (%)	2.1	2.1	1.9
Rwp (%)	3.0	3.4	2.6
FWHM⟨100⟩ (°)	0.21	0.23	0.29
FWHM⟨110⟩ (°)	0.28	0.31	0.38
FWHM⟨101⟩ (°)	0.41	0.87	1.03
Tp at 0.02 T (K)	11.8	6.7	4.8
δ	0.077	0.049	0.064	0.057
τ* (s)	2.8 × 10−5	2.7 × 10−6	1.9 × 10−6	∼10−6
Ts (K)	11.4	8.7	5.1	68.3
zν	3.7	4.2	5.2	3.8
τ0 (s)	1.7 × 10−5	6.7 × 10−6	1.3 × 10−6	∼10−6
T0 (K)	9.7	7.8	4.4	60.1
Ea /kB (K)	26.1	15.9	15.2	130
τr (s)	1.4 ×103	1.4 ×103	1.6 ×103	1.8 ×103
β	0.49	0.47	0.45	0.49

To illustrate the anisotropic widths of the Bragg reflections in RNiO₂, the three main reflection classes ⟨100⟩, ⟨110⟩, ⟨101⟩ are chosen and the corresponding full-widths at half-maximum (FWHM) of the corresponding XRD peaks are presented in table 1. It can be seen that the sharper reflections are those from diffraction planes ⟨hk0⟩, while the broad peaks belong to the class where l has a non-zero value. In other words, the coherent domains along the [001] direction are smaller than along [100]. This anisotropic shrinking of the crystalline size after topotactic reduction has already been observed in previous studies on nickelates [5, 6, 17, 31, 34, 37] and cobaltates [38], and are characteristic for systems with anisotropic particle size, stacking disorder, or uniaxial strain in layered compounds. In the RNiO₂ system, this anisotropy may be due to the stronger intra-layer than inter-layer bonding, making the structure to more likely break up along the [001] direction. On the other hand, the crystalline sizes become smaller from LaNiO₂, PrNiO₂ to NdNiO₂, especially for ⟨101⟩. This indicates an increasing instability of the RNiO₂ structure from R = La, Pr to Nd, which could be caused by the smaller lattice parameters and consequently the stronger inter-layer Ni–Ni interactions. This would be consistent with previous research on Nd_1−x Sr_x NiO₂ where a higher Sr-doping level with a larger c-axis parameter exhibits a sharper and more intense ⟨101⟩ peak [5, 6].

It is worth noting that there are no observable nickel impurities in these samples. The nickel was one of the prime suspects responsible for the FM-like transition at low temperature as well as the weak magnetic hysteresis in bulk Nd_0.8Sr_0.2NiO₂ and Sm_0.8Sr_0.2NiO₂ [5–7]. To further exclude the possibility of nickel impurities, we re-oxidized the samples back to RNiO₃ and did not observe any NiO. In addition, we measured the nickel powders synthesized using the same method and did not find any frequency-dependent behaviour as we describe below. Combining all of this evidence, we may claim to have no observable amount of crystalline nickel impurities in the as-synthesized LaNiO₂, PrNiO₂ and NdNiO₂ samples, and the magnetic measurements to be presented below reveal intrinsic properties of these poly-crystalline samples. The broad feature around 2θ = 28° for PrNiO₂ and NdNiO₂ of unknown origin which also occurs in the corresponding data of other works [5, 17, 21] is not likely to affect this result, as the LaNiO₂ without this feature still show a similar magnetic behaviour.

The DC magnetic susceptibilities χ(T) under a magnetic field of 0.02 T for the RNiO₂ samples (R = La, Pr, Nd) are shown in figure 2. It is obvious that there are peak-like structures at 11.8 K for LaNiO₂, 6.7 K for PrNiO₂ and 4.8 K for NdNiO₂, respectively. Taking into account of the huge difference between ZFC and FC data above the peak temperature, we interpret this peak-like temperature as a spin-glass freezing process, and the corresponding peak temperature is in fact a spin freezing temperature T_f. The weak hysteresis shown in figure 3 in bulk RNiO₂ samples can also be considered as a feature of a spin-glass state.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The AC susceptibility technique can be used as a probe of magnetic dynamics to distinguish spin-glass systems from other, long-range ordered systems [40]. The results of systematic AC susceptibility measurements on RNiO₂ (R = La, Pr, Nd) samples for AC frequencies ranging from 1 Hz to 300 Hz are plotted in figure 4. The applied DC magnetic field is 0, and the AC magnetic fields have the same amplitude of 0.7 mT. In all cases, we observe distinct peaks in the real parts χ' of the magnetic susceptibility, and an upward shift of the peak temperatures with increasing frequency f. At the same time, the imaginary parts χ'' indicate strong dissipation starting already below ∼100 K, with a maximum around the peak temperatures of the respective χ' data, with the exception of PrNiO₂ where the peak structure in χ'' is less clear. Such a behaviour is characteristic of spin-glass systems and distinguishes them from disordered AFM systems, in which χ'' remains zero even below the transition temperature [41–43]. Apart from these, however, there is an intriguing deviation from the typical spin-glass behaviour: the frequency-dependence disappears only at temperatures as high as 70–100 K, nearly 10 times higher than the peak temperature. This may indicate a complex magnetism in RNiO₂, similar to that we have reported for the square planar nickelate Pr₄Ni₃O₈ [39]. Multiple magnetic disorders may participate in the frustration of magnetic interactions, one that is dominating the freezing process at T_p, accompanied by another, more gradual one at a higher temperature. Interestingly, this temperature region is close to the metal–insulator transition temperature observed in the electric resistivities of RNiO₂ thin films [1, 2, 44]. This may imply an underlying relationship between the magnetic disorder and insulating behaviour.

Zoom In Zoom Out Reset image size

The relative variation of the freezing temperature T_f per frequency decade $\delta ={\Delta}{T}_{\mathrm{f}}/\left(\right.{T}_{\mathrm{f}}{\Delta}$log₁₀  f), is often employed as a parameter to compare different kinds of spin-glass systems. It reflects the strength of magnetic interactions between the underlying entities. Here, the peak temperature T_p is identified as the spin glass freezing temperature T_f. The values of δ obtained in this way are all very close, namely 0.077, 0.049, and 0.064 for LaNiO₂, PrNiO₂ and NdNiO₂, respectively. These values are intermediate between canonical spin-glass systems such as CuMn [45] (δ = 0.005) and superparamagnetic systems such as α-${\left[\right.\mathrm{H}\mathrm{o}}_{2}{\mathrm{O}}_{3}$(B₂O₃)] [48] (δ = 0.28). They fall within the order of magnitude usually observed for cluster spin glasses, such as Ni-doped La_1.85Sr_0.15CuO₄ [46] (δ = 0.012), PrRhSn₃ [47] (δ = 0.086), and therefore RNiO₂ resembles the case of cluster spin glasses.

To obtain a more detailed insight in the spin-glass behaviour, we have studied the relation between T_p and the fluctuation time τ = (2πf)⁻¹, as presented in figure 5(a). The CDS theory predicts a power law:

$$\begin{equation}\tau ={\tau }_{\ast }{\left(\frac{{T}_{\mathrm{p}}}{{T}_{\mathrm{s}}}-1\right)}^{-z\nu }.\end{equation} \tag{ 1 }$$

In this equation, τ describes the fluctuation time scale, z the dynamic critical exponent reflecting the correlated dynamics diverging as τ ∝ ξ^z , and ν the static critical exponent describing the divergence of the correlation length $\xi \propto {(T-{T}_{\mathrm{s}})}^{-\nu }$. The τ_* is the relaxation time of a fluctuating entity flip, and T_s < T_p is the static freezing temperature as f tends to zero. From table 1, the resulting τ_* of RNiO₂ is 10⁻⁵–10⁻⁶ s, far larger than the typical values for a canonical spin glass, 10⁻¹²–10⁻¹⁴ s [48], and still larger than the values of the reported cluster spin-glasses 10⁻⁷–10⁻¹⁰ s [46, 48]. Such a high value of τ_* implies a slow dynamics of the fluctuating entities of RNiO₂. The resulting zν of RNiO₂ lies around 4–5, at the lower bound found in typical spin glass systems 4–12 [48].

Zoom In Zoom Out Reset image size

Alternatively, the spin glass state can also be described by the VF law. By introducing an empirical 'true transition point or anomaly' at T₀ into the conventional Arrhenius law, the VF law can be written as

$$\begin{equation}\tau ={\tau }_{0}\enspace \mathrm{exp}\left(\frac{{E}_{a}/{k}_{\mathrm{B}}}{{T}_{\mathrm{p}}-{T}_{0}}\right).\end{equation} \tag{ 2 }$$

In this equation, τ₀ is essentially equivalent to τ_* (see table 1), and E_a is an energy barrier. Considering E_a in general to be correlated with the transition temperature for different spin-glass systems, E_a /k_B T₀ is a more meaningful measure of the coupling between the interacting entities. An E_a /k_B T₀ ≫ 1 would indicate a weak coupling, and E_a /k_B T₀ ≪ 1 a corresponding strong coupling. The value we obtain for RNiO₂ is 2–3, in the intermediate regime. The scaling plots according to the CDS theory and the VF law are shown in figures 5(b) and (c). As neither type of representation spanning nearly three frequency decades yields a strictly linear behaviour, we cannot conclusively distinguish between the two models, however. The corresponding fitting parameters are summarized in table 1.

Furthermore, the DC magnetic field dependence of AC susceptibility χ' around the freezing temperature is shown in figure 6 for (a) LaNiO₂, (b) PrNiO₂ and (c) NdNiO₂, respectively. The applied DC magnetic fields range from 0 to 0.08 T. The amplitude of AC magnetic field is 0.7 mT and the frequency is 20 Hz. One can see that the peaks in χ' are significantly suppressed with increasing DC magnetic field and may even vanish in moderately high DC magnetic fields of the order of 1 T or even less. The corresponding peak temperatures T_p, and along with them most likely also the spin freezing temperatures T_f < T_p, decrease rapidly as it is usually observed in spin glasses (see figure 6(d)) [47, 49–52]. This may explain why a recent NMR study [55] performed in μ₀ H = 12 T did not reveal any signs for spin-freezing in LaNiO₂ down to T = 0.24 K.

Zoom In Zoom Out Reset image size

A solid evidence for the spin glass behaviour in RNiO₂ is the memory effect in the magnetization. Memory-effect measurements were performed using two different protocols, an intermittent stop cooling process (ISC-process) and a continuous warming process (CW-process). In the first one, the sample is cooled in a magnetic field of 50 Oe from 300 K down to 1.8 K. Some intermittent stops are applied at certain temperatures. During these stops, the external field is switched off quickly and the temperature is kept constant for 5000 s while taking magnetization data simultaneously. Afterwards, the field is switched back to the original value and the sample further cooled down. Eventually at 1.8 K, this process is immediately followed by warming continuously from 1.8 K to 300 K. A typical feature of disordered spin glasses is a relatively long timescale to reach an equilibrium state when the magnetic field is changed. The magnetization in the glassy region would show decays in the ISC-process, and peculiar history dependent effects in the CW-process. As shown in figure 7, we can see a significant memory effect in all RNiO₂ compounds, in which the CW-process data (dashed lines) show distinct features at the previous stopping temperatures, and therefore reflect the history of the previous ISC-process. Another typical feature of spin glasses is that the magnetization below T_p decays slowly after the magnetic field is turned off. The insets of figures 7(b)–(d) show the magnetic moment measured at 5 K for 5000 s after setting the field to zero. The time decay of the magnetization can be fitted perfectly by a stretched exponential function,

$$\begin{equation}M(t)={M}_{0}+{M}_{\mathrm{g}}\enspace \mathrm{exp}\left[-{\left(\frac{t}{{\tau }_{\mathrm{r}}}\right)}^{\beta }\right],\end{equation} \tag{ 3 }$$

where M₀ is an intrinsic magnetization, M_g is related to a glassy component of magnetization, τ_r is the characteristic relaxation time constant, and β is the stretching exponent, which has values between 0 and 1 and is a function of temperature only. The fitting parameters τ_r and β for RNiO₂ and Pr₄Ni₃O₈ are presented in table 1, and the respective fitting results are shown as the green lines in figure 7. All these phenomena above are strongly indicative of a typical spin glass behaviour in RNiO₂.

Zoom In Zoom Out Reset image size

It is necessary to verify that the observed spin-glass behaviour in these nickelate samples is not caused by nickel impurities. On the one hand, the structural and elemental analysis showed no crystalline nickel in the samples. On the other hand, after a complete reduction to R₂O₃ and Ni as we mentioned above for the TG experiments, the resulting products show a complete absence of spin-glass behaviour (see supplementary material (https://stacks.iop.org/NJP/24/013022/mmedia)). Even if there had been elemental nickel impurities (crystalline or amorphous) induced by the initial reduction process and mimicking a spin-glass sate in RNiO₂, the same feature could also be expected in the fully reduced samples.

We state that bulk RNiO₂ samples show a universal spin glass behaviour, without any sign of long-range magnetic orders [17, 31]. This is distinct from the parent cuprates which host AFM long-range order. The fitting parameters from the spin-dynamics models are similar for all RNiO₂ as well as Pr₄Ni₃O₈ [39] (see table 1), implying the similar spin-glass nature in these nickelates. According to these results, the magnetic entities in RNiO₂ are expected to be cluster-like, rather than atomic single spins, and the interaction between the entities is not weak. Within the RNiO₂ system itself, we find that the relaxation times τ₀ and τ_* become shorter and the energy barrier E_a smaller from R = La, Pr to Nd. It is interesting to note that while the spin-freezing temperatures T_f decrease from La, Pr to Nd, the critical temperatures in the Sr-doped thin-film versions of RNiO₂ increase (no superconductivity for R = La [1], and T_c = 7–12 K for R = Pr [2], and T_c = 9–15 K for R = Nd [1]). This has been attributed to a possible essential role of the 4f-electrons of the R ions, and/or to stronger interactions between the Ni 3${d}_{{z}^{2}}$ orbitals owing to a smaller c-axis lattice parameter [13].

One question to be clarified is whether the observed spin-glass behaviour in the bulk RNiO₂ systems is intrinsic or not. First, the spin-glass behaviour is not likely due to elemental Ni particles, as is stated above. Second, we may refer to the strong similarity to the spin-glass dynamics in Pr₄Ni₃O₈ [39] observed in very homogeneous single crystals. Given the fact that the electronic structure of the latter compound has been proposed to be very similar to that of NdNiO₂ [54], we may conclude that the magnetic frustration causing the universal spin-glass behaviour in bulk RNiO₂ and perhaps even in Pr₄Ni₃O₈ has a common cause, and is not due to weak crystallization. Nevertheless, the fact that T_s in the latter compound is almost one order of magnitude larger than in bulk RNiO₂, may hint to certain differences.

Considering that Pr₄Ni₃O₈ is prepared by the topotactic reduction method similar to the title compounds, we cannot entirely rule out that the spin-glass behaviour could result from local non-stoichiometries of oxygen caused by incomplete or inhomogeneous reduction. An off-stoichiometric oxygen content can alter the coordination environment, thereby lead to distributed high-valence-state Ni ions and consequently make a frustration between various magnetic interactions possible. However, taking the essential role played by the interface and strain effect in nickelate thin films [9] into account, the superconductivity along with the potential long-range magnetic order in thin films is not necessarily in contradiction to the occurrence of a disordered magnetism in bulk samples. With this respect, the spin-glass behaviour in bulk RNiO₂ can indeed be intrinsic. The spin glass state in RNiO₂ could arise from the self-doping effect of Nd 5d electron pockets [27, 56, 57]. Zhang et al [56], proposed NdNiO₂ to be a self-doped Mott insulator, where the low-density Nd 5d conduction electrons couple to the localized Ni 3${d}_{{x}^{2}-{y}^{2}}$ electrons to form Kondo spin singlets at low temperatures. This self-doping effect suppresses the AFM long-range order. Choi et al [58], proposed that the Ni 3${d}_{{z}^{2}}$ orbitals also participate strongly and forms a flat band pinned to the Fermi level in the nickelates. As a result, the magnetic orders are unstable and can be easily frustrated by the strong spin, charge, and lattice fluctuations, resulting in a spin disordered state but with AFM correlations [13]. Leonov et al [59], also proposed that an unanticipated frustration of magnetic interactions in Nd_1−x Sr_x NiO₂ suppresses magnetic order. Werner et al [60], argued that the optimally doped nickelate system is in a spin-freezing crossover region. Both overdoped and underdoped system are in a spin-freezing (or spin-glass-like) state, and the associated fluctuations of local moments may provide the pairing glue for the occurrence of superconductivity.

The spin-glass behaviour in the bulk infinite nickelates is partly reminiscent of that in the lightly-doped cuprates La_2−x Sr_x CuO₄ [53]. A spin-glass phase has been found at intermediate concentrations between the AFM phase and superconducting phases (0.02 < x < 0.05) [61]. This phase is formed by frozen AFM clusters, which have been argued to originate from the spatial charge segregation in the CuO₂ planes [61]. This spin-glass regime extends far into the SC state, indicating a microscopic coexistence of superconductivity and frozen AFM clusters at low temperatures [62]. As a comparison, our AC magnetic measurements reveal that the magnetic entities of nickelates are also clusters rather than single spins. Considering the fact that stripe phases have been widely observed in other layered nickelates [63–65], one may wonder if the magnetic clusters in bulk 112 nickelates indicate any internal charge segregation. Our presented data, however, do not allow to shed more light on this issue.

To conclude, we successfully synthesized bulk samples of the parent compounds RNiO₂ (R = La, Pr, Nd) and performed comprehensive magnetic measurements on them. The history-dependent DC magnetic susceptibilities, the magnetic hysteresis, the frequency dependence of the AC magnetizations, and more importantly, the memory effect reveal a universal spin-glass behaviour. According to phenomenological fitting results, we find that bulk RNiO₂ systems share a similarly slow spin dynamics as Pr₄Ni₃O₈, although with an almost one-order-of magnitude lower spin-freezing temperature. The magnetic entities are most probably cluster-like, and the interactions between these entities are not weak. Based on our investigations, we suggest that the universal spin glass behaviour in bulk RNiO₂ may be due to off-stoichiometric oxygen or even to an intrinsic magnetic frustration, rather than to impurities or weak crystallization.

This work was supported by the Swiss National Foundation under Grants Nos. 20-175554, 206021-150784.

All data that support the findings of this study are included within the article (and any supplementary files).