Experimental realization of Lieb-Mattis plateau in a quantum spin chain

Topological properties in material science have recently received tremendous attention, as exemplified by topological insulators. As for quantum spin systems, the Haldane chain with integer spins is the best known example1. The realization of a topological quantum phase in the Haldane chain is an innovative research development related to the 2016 Nobel Prize in Physics. For mixed spin chains composed of two different spins, the appearance of a topologically quantized magnetization plateau is predicted by the Lieb-Mattis theorem2 in combination with the Oshikawa-Yamanaka-Affleck criterion3. However, the actual magnetization plateau in the mixed spin chain has not yet been observed. Here, we present a model compound forming a mixed spin-(1/2, 5/2) chain. We observe a clear Lieb-Mattis plateau and well explain it quantitatively. The present results demonstrate a quantum many-body effect based on quantum topology and provide a new stage in the search for topological properties in condensed matter physics.

strong AF couplings in the candidate compounds, it is difficult to observe the entire predicted LM plateau using actual high-field measurements.
In this study, we realize a new mixed spin-(1/2, 5/2) chain in a charge-transfer salt ( , where s and S are the spin-1/2 and 5/2 operators, respectively. The J and J′ have very close values and from a mixed spin-(1/2, 5/2) chain with a slight alternation (J′/J = 0.85), as shown in Fig. 1c. We note that intermolecular interactions between organic radical and 3d transition metal anions tend to be underestimated. From our previous study 20 , the actual values in verdazyl-based salts are expected to be approximately three times larger than those of the MO evaluation, i.e., J/k B ≃ 12 K and J′/k B ≃ 10 K. The MO calculations also indicated two types of interchain interactions (see Supplementary Information). However, the effective interchain couplings are sufficiently small to be ignored in the current discussion of the mixed spin chain, as described in the following analysis of magnetization. Figure 2a shows the magnetization curve at 1.5 K. We observe a clear 2/3 magnetization plateau between approximately 2 and 30 T and subsequent gapless behavior. The magnetization curve reaches its saturation at approximately 40 T. Considering the isotropic g value of ~2.0 resulting from radical and Fe 3+ ion in this compound, the saturation value of 5.7μB/f.u. indicates that the purity of the radicals and anions is approximately 95%. We consider this purity in the following analysis. The field derivative of the magnetization curve (dM/dB) exhibits a double peak structure accompanied by a phase transition towards the gapless LL phase.

Results and Discussion
Here, we discuss the ground state of the present mixed spin-(1/2, 5/2) chain. We consider the uniform (J = J′) and alternating (J > J′) types. The uniform type is a typical model forming the LM ferrimagnet. Its ground state has the total spin S tot = 2 per unit cell, which are degenerate at zero-field. The degeneracy is lifted by the finite magnetic field, and the eigenstate with S z tot = 2 becomes the lowest in the plateau region, as described in Fig. 2b. The LM theorem is also applied to the alternating type, and thus the ground state is expected to become a similar ferrimagnetic state with S z tot = 2. We can understand the situation by the valence bond picture in Fig. 2b, where the two spin-1/2 particles on the different sites form a nonmagnetic singlet dimer through J. We calculated the magnetization curves of the mixed spin-(1/2, 5/2) chain using the quantum Monte Carlo (QMC) method. The calculated result for the uniform type is confirmed to be the same as that of the previous works 5, 6 , and is shown in Fig. 2a. We obtained good agreement between the experiment and calculation by using J/k B = 9.0 K. The calculated result reproduces not only the 2/3 magnetization plateau but also the field-induced LL phase. Further, the calculated dM/dB reproduces the double peak structure qualitatively. The slight differences in the low-field thermal paramagnetic 21 and LL phases are considered to originate from the interchain interactions. Below the plateau phase, the effective spin-2 state fully polarizes at only approximately 2 T, which indicates the interchain couplings are much weaker than the intrachain ones. These weak interchain couplings are understood by the valence bond picture of the alternating type in Fig. 2b. The MO calculations indicate that the dominate interchain interactions are between s = 1/2 radicals (see Supplementary Information). However, the s = 1/2 spins can be regarded as forming a nonmagnetic singlet dimer. The effective interactions between the remaining spins are caused through the triplet excited states of the singlet dimers and become much weaker than the MO evaluations [22][23][24] . The magnetization curve increases relatively rapidly as the α decreases. The corresponding differences appear in dM/dB, where the shape of the peak dramatically changes with decreasing α. The decrease of α corresponds to a reduction of one-dimensionality towards a complete dimer without an intermediate LL phase for α = 0. We confirm that the uniform type for α = 1 is closest to the experimental result, and the α dependence becomes almost indistinguishable for 0.7 <α < 1. Accordingly, the actual value of α in the present model is considered to be between 0.7 and 1, which is consistent with the MO evaluation. Figure 4a shows the temperature dependence of the magnetic susceptibility (χ = M/H) at 0.1 T. We observe an anomalous change at 5.0 K, which indicates that a phase transition to the long-range order (LRO) occurs at T N = 5.0 K owing to the small but finite interchain interactions. In the higher-temperature region, the value of χT exhibits a rounded minimum at approximately 24 K, as shown in the inset of Fig. 4a. This behavior is a typical characteristic of mixed spin chains [13][14][15][16] . We then calculated χT for the uniform mixed spin-(1/2, 5/2) chain with J/k B = 9.0 K by using the QMC method, obtaining a good agreement in the temperature region above T N . The difference below T N originates from the interchain interactions causing the AF LRO as in the case of the low-field part of the magnetization curve.
The experimental result for the specific heat at zero-field clearly exhibits a λ-type sharp peak associated with the phase transition to the LRO, which is consistent with the discontinuous change in the magnetic susceptibility, as  www.nature.com/scientificreports www.nature.com/scientificreports/ shown in Fig. 4b. We have not observed the expected double peak and √T behavior in the low-temperature region 6 due to the phase transition behavior. Because the experimental behavior below T N exhibits almost T-linear dependence, the gapless ferromagnetic dispersion is considered to change to an AF linear dispersion owing to the effect of the AF interchain interactions. In magnetic fields, the phase transition temperature decreases with increasing fields and almost disappears above 2 T, which is consistent with the formation of the plateau phase with a spin gap.

conclusion
In conclusion, we have succeeded in synthesizing a new model compound forming a mixed spin-(1/2, 5/2) chain. We observe the entire magnetization curve up to saturation, which exhibits a clear Lieb-Mattis magnetization plateau and subsequent quantum phase transition towards the gapless LL phase. The present results demonstrate a quantum many-body effect based on the Lieb-Mattis theorem and will stimulate studies on topological properties in mixed spin chains.

Methods
We synthesized 4-Br-o-MePy-V using a conventional procedure 25 and prepared an iodide salt of the radical cation (4-Br-o-MePy-V)I using a reported procedure for salts with similar chemical structures 26 . 1-butyl-3-methylimidazolium tetrachloroferrate (239 mg, 0.71 mmol) was slowly added to a solution of (4-Br-o-MePy-V)I (149 mg, 0.28 mmol) in 10 ml of ethanol and stirred for 40 min. The dark-green solid (4-Br-o-MePy-V)FeCl 4 was then separated by filtration. The dark-green residue was recrystallized using mixed solvent of acetonitrile and methanol. The crystal structure was determined on the basis of intensity data collected using a Rigaku AFC-8R Mercury CCD RA-Micro7 diffractometer. High-field magnetization measurement in pulsed magnetic fields of up to approximately 53 T was conducted using a nondestructive pulse magnet. The magnetic susceptibility was measured using a commercial SQUID magnetometer (MPMS-XL, Quantum Design). The experimental result was corrected for the diamagnetic contributions calculated by Pascal's method. The specific heat was measured using a commercial calorimeter (PPMS, Quantum Design) by using a thermal relaxation method. Considering the isotropic nature of organic radical systems, all experiments were performed using small randomly oriented single crystals. Ab initio MO calculations were performed using the UB3LYP method with the basis set 6-31 G(d,p) in the Gaussian 09 program package. For the estimation of intermolecular magnetic interaction, we applied our evaluation scheme that have been studied previously 27 . The QMC code is based on the directed loop algorithm in the stochastic series expansion representation 28 . The calculations for the mixed spin-(1/2, 5/2) Heisenberg chain were performed for N = 128 under the periodic boundary condition, where N denotes the system size. It was confirmed that there is no significant size-dependent effect. All calculations were carried out using the ALPS application 29,30 .

Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.