Randomness-induced quantum spin liquid on honeycomb lattice

Quantum entanglement in magnetic materials is expected to yield a quantum spin liquid (QSL), in which strong quantum fluctuations prevent magnetic ordering even at zero temperature. This topic has been one of the primary focuses of condensed-matter science since Anderson first proposed the resonating valence bond state in a certain spin-1/2 frustrated magnet in 1973. Since then, several candidate materials featuring frustration, such as triangular and kagome lattices, have been reported to exhibit liquid-like behavior. However, the mechanisms that stabilize the liquid-like states have remained elusive. Here, we present a QSL state in a spin-1/2 honeycomb lattice with randomness in the exchange interaction. That is, we successfully introduce randomness into the organic radial-based complex and realize a random-singlet (RS) state (or valence bond glass). All magnetic and thermodynamic experimental results indicate the liquid-like behaviors, which are consistent with those expected in the RS state. Our results suggest that the randomness or inhomogeneity in the actual systems stabilize the RS state and yield liquid-like behavior.

even in the case of very weak frustration. Although some honeycomb-lattice-based compounds have recently been reported to exhibit liquid-like behavior, their exact lattice systems have not been clarified 22,23 . In this work, we present a new S = 1/2 Heisenberg AF honeycomb lattice composed of three dominant interactions, as shown in Fig. 1a. Those interactions are designed to have bond-randomness with a weak additional AF interaction J 4 inducing frustration in the lattice. Our experimental results regarding the magnetic and thermodynamic properties indicate the realization of the RS state, as schematically shown in Fig. 1b.
Recently, we developed verdazyl radical systems with flexible molecular orbitals (MOs) that enable tuning of the intermolecular magnetic interactions through molecular design [24][25][26] . In this study, we utilized a new verdazyl-based complex Zn(hfac) 2 (A x B 1−x ), where hfac represents 1,1,1,5,5,5-hexafluoroacetylacetonate, and A and B equivalent to regioisomers of verdazyl radical. It is essential for introduction of randomness that the rotational degrees of freedom of verdazyl radical disappear owing to the coordination to Zn(hfac) 2 , as shown in Fig. 2a. Accordingly, two different regioisomers, labeled A-type (x) and B-type (1-x), arise and randomly align in the crystal, yielding randomness of the intermolecular exchange interactions.
The randomness effect reaches maximum at x = 0.5, where the numbers of A-and B-type molecules are identical. Some numerical inequalities exist depending on the conditions of the solution used in the complex-forming reaction (see Method section), and the actual crystals have slightly large x values. Here, we successfully synthesized two different single crystals with x = 0.64 and 0.79. The crystallographic parameters were determined at room temperature and 25 K for both crystals (Supplementary Table S1). Only slight differences were observed between the results for x = 0.64 and 0.79. Note that, because this investigation focused on the low-temperature magnetic properties, the crystallographic data at 25 K are discussed hereafter. The crystallographic parameters at 25 K for x = 0.64 were as follows: monoclinic, space group P2 1 /n, a = 9.
denotes the sum over the corresponding spin pairs. Note that J 4 /k B is almost independent of the pair formation and evaluated to be 0.08 K. Figure 3a shows the temperature dependence of the magnetic susceptibilities (χ = M/H) for various magnetic fields. Although we performed field-cooled and zero-field-cooled measurements to examine spin-freezing for x = 0.64, no distinguishable differences were found (Fig. 3a). Below 0.25 T, a shoulder is apparent at approximately 1 K, along with gapless behavior with a Curie tail in the lower temperature region, for both x = 0.64 and 0.79. This shoulder indicates development of AF correlations forming spin-singlet dimers, and the Curie-like diverging components indicate a small fraction of free spins owing to some unpaired gorphan h spins, as illustrated in Fig. 1b. The appearance of the small free-spin fraction generating Curie-like low-temperature χ is indeed expected in the RS picture 12,20,21 .
The magnetization curves at the lowest temperature of 0.08 K also indicate gapless behaviors, as shown in Fig. 3b. The entire magnetization curve for x = 0.64 exhibits the near-linear behavior expected for the RS state 12 , whereas that for x = 0.79 exhibits slight bending at approximately 3.5 T. Such bending was also observed in the magnetization curve calculated for the RS state near the saturation field 12,20 . In general, there is a sharp change in Entangled spinsinglet dimers are indicated by ovals that cover two lattice sites. The dimers can be formed in a spatially random manner, not only between neighboring sites, but also between distant sites through higher-order interactions. The arrow indicates an unpaired gorphan hspin.
the magnetization curve at the saturation field for non-randomness phase. By introducing bond-randomness, the magnetization curve near the saturation field becomes gradual, originating from the widely distributed binding energy of the singlet dimers. That is, bending appears when the randomness is small. When the randomness is increased, the bending eventually becomes almost linear. As the randomness increases when x approaches 0.5, the near-linear behavior in the magnetization curve for x = 0.64 is consistent with the theoretical prediction for RS state.  The temperature dependence of the specific heat, C p , for x = 0.64 is shown in Fig. 4. In the low-temperature regions below 0.5 K, a clear gapless T-linear behavior was observed, C p γ  T . This T-linear behavior is robust against an applied magnetic field and appears even under a high-magnetic field near the saturation field. Such T-linear behavior of the specific heat is consistent with the specific heat expected in the RS picture 21 , which originates from the widely distributed binding energy. We also found a broad hump structure in the temperature dependence of C p /T (Fig. 4, inset). Note that similar broad hump structures have also been observed for the C p /T/ of organic triangular salts, in which the broad hump is considered to be a crossover to QSL state 3,6 . We roughly evaluated γ from the C p /T values at the lowest temperatures and obtained 1.20 and 0.72 J/mol K 2 at 0 and 3 T, respectively. From numerical analysis of the RS state, we thus deduced that the T-linear term of C p is strongly dependent on the fundamental ground state without randomness and, also, on the degree of introduced randomness 12,20,21 . The obtained γ-values are somewhat large, but do not differ significantly from the calculations.
In the honeycomb lattice, a relatively small lattice distortion can induce a disordered gapped phase (even in the non-frustrated case) owing to strong quantum fluctuation 27,28 . From the theoretical analysis, it is deduced that the introduction of bond-randomness into the gapped phases is more effective for RS state formation than introduction into the gapless ordered phase 12,20,21 . Therefore, in the present model, the lattice distortion as well as the weak frustrated interaction should enhance the bond-randomness effect, inducing formation of the RS state.
In summary, we have succeeded in synthesizing single crystals of the verdazyl-based complex Zn(hfac) 2 (A x B 1−x ). Two different regioisomers, A-type (x) and B-type (1-x), arise and randomly align in the crystal, yielding randomness of the intermolecular exchange interactions. Ab inotio MO calculations indicate the formation of the S = 1/2 Heisenberg AF honeycomb lattice composed of three dominant interactions, and there is a weak additional AF interaction inducing frustration in the lattice. All magnetic and thermodynamic experimental results indicate the liquid-like behaviors, which are consistent with those expected in the RS state, These results demonstrate that the randomness or inhomogeneity in the actual systems stabilize the RS state and yield liquid-like behavior. Furthermore, our method to introduce a bond-randomness into spin lattices enable further investigations on the randomness-induced QSL in other lattice systems.

Methods
We synthesized Zn(hfac) 2 (A x B 1−x ) using a conventional procedure similar to that used to prepare the typical verdazyl radical 1,3,5-triphenylverdazyl 29 . A solution of p-Cl-o-Py-V [1-(4-chlorophenyl)-3-(2-pyridyl)-5-p henylverdazyl] (119 mg, 0.34 mmol) in 10 ml of CH 2 Cl 2 was slowly added to a solution of [Zn(hfac) 2 ]·2H 2 O (175 mg, 0.34 mmol) in 20 ml of heptane at 80 °C, and stirred for 1 h. After the mixed solution cooled to room temperature, a dark-green crystalline solid of Zn(hfac) 2 (A x B 1−x ) was separated by filtration and washed with pentane. The dark-green residue was recrystallized using CH 2 Cl 2 in an acetonitrile atmosphere. The crystal structure was determined on the basis of intensity data collected using a Rigaku AFC-8R Mercury CCD RA-Micro7 diffractometer with a Japan Thermal Engineering XR-HR10K. The magnetizations were measured down to approximately 80 mK using a commercial SQUID magnetometer (MPMS-XL, Quantum Design) and a capacitive Faraday magnetometer. The experimental results were corrected for diamagnetic contributions (−4.2 × 10 −4 emu mol −1 for x = 0.64 and −4.0 × 10 −4 emu mol −1 for x = 0.79), which were determined to become almost χT = const. above approximately 200 K, and close to the value calculated using Pascal's method. The specific heat was measured using a hand-made apparatus and a standard adiabatic heat-pulse method down to ∼0.1 K. Considering the isotropic nature of organic radical systems, all experiments were performed using small randomly oriented single crystals. The ab initio MO calculations were performed using the UB3LYP method as broken-symmetry hybrid-density functional theory calculations. All the calculations were performed using the Gaussian 09 program package, with the basis functions being 6-31G. To estimate the intermolecular magnetic interaction of the molecular pairs, we applied our previously presented evaluation scheme 30 .