betaFeedback

WorksPhysics and Astronomy

Eminuscent phase in frustrated magnets: a challenge to quantum spin liquids

Abstract

A geometrically frustrated (GF) magnet consists of localised magnetic moments, spins, whose orientation cannot be arranged to simultaneously minimise their interaction energies. Such materials may host novel fascinating phases of matter, such as fluid-like states called quantum spin liquids. GF magnets have, like all solid-state systems, randomly located impurities whose magnetic moments may freeze'' at low temperatures, making the system enter a spin-glass state. We analyse the available data for spin-glass transitions in GF materials and find a surprising trend: the glass-transition temperature grows with decreasing impurity concentration and reaches a finite value in the impurity-free limit at a previously unidentified, hidden'', energy scale. We propose a scenario in which the interplay of interactions and entropy leads to a crossover in the permeability of the medium that assists glass freezing at low temperatures. This low-temperature, ``eminuscent'', phase may obscure or even destroy the widely-sought spin-liquid states in rather clean systems.

GFHE NC

Introduction

In magnetic systems, frustration [1, 2] is essential to avoid long-range order and realise theoretically predicted quantum-spin-liquid states, of potential use for quantum computation. Frustration may come from tuned interactions, as in potential Kitaev materials [3], or from the geometry of the lattice, as in a broad class of materials called geometrically frustrated (GF) materials, the focus of this work. As a result, spin liquids are actively sought in materials with antiferromagnetic (AF) interactions on geometrically frustrating lattices [4], exemplified by the pyrochlore, Kagome and triangular structures.

Whereas spin liquid theories assume perfect lattices, real materials possess quenched disorder, such as vacancies or impurities, in varying degrees. Such disorder may fundamentally alter the nature of the ground state in geometrically frustrated systems by inducing a disordered state, possibly precluding the formation of a spin liquid.

Here, we first address the disorder-induced spin-glass-freezing in GF materials and compare it with conventional spin-glass (SG) transitions in non-GF materials. The nature of disorder and the medium between impurities or vacancies are quite different in GF and non-GF SGs. The medium between impurities in non-GF systems is non-magnetic, while the medium in GF systems consists of atomic spins. This usually leads, for example, to a large quadratic in temperature specific heat C(T)C(T) in GF SGs [5, 6], in contrast with the impurity-dominated linear-in-temperature C(T)C(T) found in non-GF SGs [7]. Defects in GF systems are normally represented by magnetic atom vacancies, while impurities in non-GF SGs are magnetic atoms.

Despite these differences, spin glasses in GF and non-GF systems share a number of common properties. Both classes of systems are believed to display spin-glass transitions in the same universality class [8] and both: i) display a cusp in magnetic susceptibility χ(T)\chi(T) at the glass-transition temperature TgT_g; ii) display hysteresis in χ(T)\chi(T) below TgT_g [5, 9]; iii) show a divergence in the non-linear susceptibility at TgT_g [5, 10]; iv) show no specific heat anomaly at TgT_g. Also, increasing the amount of quenched disorder (defects) in both GF and non-GF systems is commonly believed, and confirmed by numerous models, to favour spin freezing and increase the glass-transition temperature, regardless of the nature of disorder, (see, for example, Refs. [8, 11, 12, 13]).

In this paper, we are questioning conventional understanding of the response of GF systems to defects. We analyse the available experimental data for χ(T)\chi(T) and TgT_g as a function of the density of randomly located defects and demonstrate that the phenomenology of the glass transition in GF and non-GF systems is dramatically different. In both classes of materials, χ(T)\chi(T) increases with increasing the density of defects, magnetic impurities in non-GF systems and vacancies in GF systems. In GF systems, however, this trend is accompanied by a decrease of the critical temperature TgT_g, in contrast with non-GF SGs as well as with common intuition. Furthermore, extrapolating TgT_g in GF materials to the limit of zero vacancies gives a finite value, T∗T^*, far below the Weiss constant θW\theta_W. Since vacancies represent the main source of disorder in GF materials, this suggests a transition at T∗T^* in pure GF systems.

This observation calls into question the observability of quantum spin liquids in GF materials since all solids possess some disorder and an SG state may be anathema to a spin liquid. In this paper, we propose a scenario of the transition, consistent with the available experimental data, in which the interplay of interactions and the entropy of the impurity spins leads to a crossover in the effective magnetic permeability of the medium, which in turn drives the transition.

Results

We analyse and compare experimental data on the glass transition for GF and non-GF systems with varying degrees of quenched disorder. For the former, we consider only strongly frustrated GF systems, i.e. f=θW/Tg≳10f = \theta_W/T_g \gtrsim10 [14].

Disorder in GF systems. Unlike in non-GF systems, where disorder scales with atomic spin density, in GF materials disorder can arise from different sources, the most common of which are spin vacancies and structural distortions. For most of the materials considered here, disorder was varied by substituting a non-magnetic atom for a magnetic atom. Such a spin vacancy is generally considered to be the degree of freedom that freezes at TgT_g – the vacancy induces a shielding effect among its neighbouring spins, thus creating a composite spin-like variable, a “quasispin” or a “ghost spin” (see, e.g. [15, 16]). In a few systems, the structure itself creates disorder, an extreme example of which is Y2_2Mo2_2O7_7, where a dynamic Jahn-Teller transition, driven by the Mo4+^{4+} ion, induces a (disordered) orbital dimer phase at its TgT_g, thus leading to bond disorder among the spins [17]. When this native disorder is increased by substitution of non-magnetic Ti4+^{4+} [18], the behaviour, to be discussed below, is similar to spin vacancy introduction in other GF systems. Another perhaps less extreme example is Na4_4Ir3_3O8_8, where incomplete occupancy of Na1+^{1+} on the AA-site [19] likely leads to a random strain field that may induce randomness in the Ir4+^{4+} spin exchange interactions. The other systems considered here do not possess such obvious defects but may, in synthesis, acquire disorder that exists below typical X-ray diffraction detection limits.

Dichotomy between GF and non-GF systems. To reveal universal trends in GF and non-GF systems, which are independent of the nature of disorder, we plot in Figure 1 χ(T)\chi(T) as a function of TT for various amounts of disorder (upper panel) and χ(Tg)\chi(T_g) as a function of TgT_g (lower panels). The upper panels demonstrate a dramatic difference between representative non-GF and GF systems – while χ(Tg)\chi(T_g) increases with TgT_g for the non-GF example, χ(Tg)\chi(T_g) decreases with TgT_g for the GF example. In the lower panels of Figure 1 we show the dimensionless χ(Tg)\chi(T_g) versus TgT_g, using all available experimental data for families of compounds with variable disorder [5, 9, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30]. (Details on methodology are provided in the SI.) We emphasise that, since different reviewed compounds have different microscopic sources of quenched disorder (to which impurity atoms contribute but of which they may not be the only source), the observed trends in the behaviour of χ(Tg)\chi(T_g) vs. TgT_g are universal.

Behaviour of susceptibilities in conventional and geometrically frustrated spin glasses

Figure 1. Behaviour of susceptibilities, χ(T)\chi(T) and χ(Tg)\chi(T_g) in conventional [(a) and (c)] and geometrically frustrated [(b) and (d)] systems. (a) and (b): The susceptibilities χ(T)\chi(T) of (Eu,Sr)S [20] and NiGa2_2S4_4 [21] as a function of temperature for various vacancy concentrations. The lines have been rendered from digitized data in the referenced publications and the TgT_g values are plotted in the lower panels. (c) and (d): The susceptibilities χ(Tg)\chi(T_g) at TgT_g vs. TgT_g for conventional [22, 23] and frustrated spin glasses. The reference sources for the GF data are provided in Table 1.

Hidden energy. The second representation of the data (Figure 2a) considers, for GF systems, the behaviour of TgT_g not in relation to χ(Tg)\chi(T_g) but as a function of a known amount of disorder, specifically the nominal spin vacancy concentration, xx. Although spin vacancies may not be the only source of disorder, as in Y2_2Mo2_2O7_7, decreasing xx results in decreasing the total disorder strength, at least for sufficiently small xx. In this linear xx vs. log⁡Tg\log T_g plot, we see that the glass-transition temperature increases with decreasing disorder, strongly implying that TgT_g reaches a finite value, T∗T^*, in the pure system. We note that since no other substantial sources of disorder distinct from spin vacancies are known in the reviewed materials, except for Y2_2Mo2_2O7_7 as discussed, the extrapolation to x→0x \to0 gives the perceived transition temperature, T∗T^*, in pure materials. Since T∗T^* is an apparent property of the pure compound, and since it is at least an order of magnitude smaller than θW\theta_W, and has not been described via a microscopic theory, we refer to this as a “hidden” energy scale. Values for T∗T^*, θW\theta_W, and other materials characteristics are compiled in Table 1.

Dependencies of the glass transition temperatures on the amount of disorder in geometrically frustrated systems

Figure 2. The dependencies of the glass transition temperatures TgT_g on the amount of disorder in geometrically frustrated systems. (a) The dependencies for real materials following from the available experimental data analysed in this work. The non-vanishing of TgT_g in the limit of vanishing disorder for NiGa2_2S4_4, Na4_4Ir3_3O8_8 and Y2_2Mo2_2O7_7 suggests the existence of a hidden energy scale T∗T^*. (b) The dependence proposed here for a broader range of disorder. The upper part of the curve (“experimentally accessed range of disorder”) illustrates the experimental data analysed in this work. The lower part, for weaker disorder, is conjectured.

Table 1. Table 1. Strongly geometrically frustrated magnets and their magnetic lattice, atomic spin (SS), AF Weiss constant θW\theta_{\mathrm{W}}, glass transition temperature TgT_{\mathrm{g}}, frustration parameter f=θW/Tgf = \theta_{\mathrm{W}}/T_{\mathrm{g}} and the hidden energy scale T∗T^*, for compounds pure enough to allow an estimate. Compounds for which T∗T^* is not estimated due to insufficient disorder range are indicated by n.e. Entries not having an error estimate can be considered accurate to one digit in the least significant place. The compounds listed are the lowest-disorder members of their respective dilution series, data for which can be found in the references.

Other experimental signatures of the hidden energy scale. While susceptibility yields information about the q=0q = 0 response, other measurements, such as specific heat, muon relaxation, and neutron scattering, shed light on the spin configuration in the vicinity of T∗T^*. Neutron scattering is especially informative since it shows, for the systems that approach x=0x = 0, i.e. those that most closely approach “pure”, a rapid rise in the intensity of short-range elastic scattering on approaching T∗T^* from above, and little temperature dependence below T∗T^* [31, 32, 33]. Thus, T∗T^* coincides with the development of short-range order among atomic spins on cooling. This short-range order, also revealed as a broad peak in specific heat, is clearly a property of the pure system. Moreover, unlike systems of reduced dimensionality where the temperature at which short-range order occurs reflects the magnitude of the dominant exchange interaction, no such microscopic origin for T∗T^* is presently known for GF systems. Absent a microscopic model, however, it is possible to create a framework phenomenology of this short-range ordered state, as described later.

Effective degrees of freedom. In order to fully appreciate the implications of the GF/non-GF dichotomy shown in Figure 1, one needs to consider the degrees of freedom that govern the glass freezing and the magnetic susceptibility in these classes of materials. For the non-GF systems there can be no doubt that the degrees of freedom are atomic spins, e.g. Mn atoms in a Cu host, as alluded to above [11, 12, 20, 22, 23]. For GF systems, however, there are two options for magnetic degrees of freedom: 1) “quasispins” [15, 16] forming around vacancies and acting as free magnetic moments or 2) the bulk spins away from the vacancies. The growth of the susceptibility χ(T)\chi(T) in GF systems with increasing the density of vacancies, similar to the growth of χ(T)\chi(T) with increasing the density of magnetic moments in non-GF systems, is evidence in favour of the quasispin scenario, which, however, does not explain the behaviour of TgT_g. The decrease of TgT_g with increasing the vacancy concentration, on the contrary, suggests an essential role of the bulk degrees of freedom: as vacancies dilute the bulk spins, they lower the glass-transition temperature, provided the transition is driven by the bulk spins. At the same time, vacancies would decrease the contribution of the bulk spins to χ(T)\chi(T), which is observed to grow with vacancy density. Therefore, any explanation, including microscopic theories, of the behaviour of GF SGs must involve both the quasispin and bulk degrees of freedom and their interplay and be constrained by the equation

dχ(Tg)dTg<0\frac{d\chi(T_g)}{dT_g}<0
GF compoundlatticeSSθW\theta_W (K)TgT_g (K)ffT∗T^*Ref.
ZnCr1.6_{1.6}Ga0.4_{0.4}O4_4spinel3/21152.448n.e.[16, 24]
Zn0.95_{0.95}Cd0.05_{0.05}Cr2_2O4_4spinel3/2500±20500 \pm201050n.e.[27, 34]
SrCr8_8Ga4_4O19_{19}Kagome (layered)3/25153.5147n.e.[5, 25]
BaCr8.1_{8.1}Ga3.9_{3.9}O19_{19}Kagome (layered)3/26954.9142n.e.[35]
NiGa2_2S4_4triangular (layered)1808.0108±0.28 \pm0.2[6, 21]
Mn0.53_{0.53}Cd0.47_{0.47}Tefcc5/229215±315 \pm319n.e.[26]
Na4_4Ir3(1−x)_{3(1-x)}Ti3x_{3x}O8_8hyperkagome1/25706.01006.0[19, 29]
Y2_2Mo2_2O7_7pyrochlore120023.58.525±125 \pm1[9, 18]
(Fe,Ga)2_2TiO5_5pseudobrookite5/2900±10900 \pm105516n.e.[28]

Table 2.

Discussion

As discussed above, the observed increase of TgT_{\mathrm{g}} with decreasing disorder, to which vacancies contribute but of which they may not be the only source, is in sharp contrast with numerous models of spin glasses, including GF glasses [8, 11, 12] (we confirm, by a microscopic calculation, in [13] that TgT_{\mathrm{g}} vanishes in the limit of vanishing disorder strength for quenched disorder and scales as Tg∝nimp1/2T_{\mathrm{g}} \propto n_{\mathrm{imp}}^{1/2} if determined by sparsely located vacancies with the average density nimpn_{\mathrm{imp}}). Furthermore, this trend implies the existence of a glass transition in a disorder-free system, at a temperature given by the hidden energy scale T∗T^*.

A naïve explanation for the hidden energy scale in GF systems is the presence of non-vacancy sources of quenched disorder, such as random strain in Y2Mo2O7\mathrm{Y}_2\mathrm{Mo}_2\mathrm{O}_7, that could drive the glass transition in the zero-vacancy limit. However, the hidden energy scale has the same order of magnitude, T∗∼10KT^* \sim10\mathrm{K}, in all GF compounds (with the exception of Fe3−3xGa3x−1TiO5\mathrm{Fe}_{3-3x}\mathrm{Ga}_{3x-1}\mathrm{TiO}_5, in which T∗∼100KT^* \sim100\mathrm{K} and the transition is apparently dominated by clusters of ∼30\sim30 Fe3+\mathrm{Fe}^{3+} spins rather than by single-vacancy spins [30]). Since the other compounds do not have significant known sources of disorder distinct from vacancies, this magnitude of the hidden energy scale requires an explanation.

Another possibility is a glass-freezing transition in the absence of quenched disorder [36]. Such a possibility has never been demonstrated rigorously but would be consistent with the saturation of the glass-transition temperature to a finite value T∗T^* in the disorder free limit.

It is also possible that the GF medium undergoes a sharp crossover at a scale T∗T^*, determined only by the properties of the disorder-free GF medium, which drives the glass transition in the presence of a small residual number of vacancies (possibly undetected in experiments) or non-vacancy disorder. Below, we consider in more detail the latter scenario, which we believe to be the most likely, and which is consistent with the neutron-scattering data [31, 32, 33]. We assume that the transition temperature will eventually vanish in the limit of zero disorder and that GF systems studied to date have residual disorder (in the form of defects or other sources). A sharp crossover in the medium, as a function of temperature, can trigger a transition to the glass state with the temperature T∗T^* determined by the properties of the medium and weakly dependent on the strength of the residual disorder for experimentally available systems.

The above discussion raises the question of what the transition at T∗T^* in the pure system represents. At T∗T^*, the system transitions or crosses over from a high-temperature state with suppressed correlations between impurity atoms to a low-temperature phase in which the long-range entropic interactions are essential and assist the freezing of the impurity degrees of freedom. We refer to the latter phase as an “eminuscient”, derived from the Latin “ex” (away from) and “manus” (hand), with “eminusque” being Latin for “long range”.

The existence of the hidden energy scale suggests that a spin glass phase may be inescapable in real materials, even if they can be made significantly purer, which raises the important question of whether a quantum spin liquid can coexist with a spin glass – a question that is beyond the scope of the present work – and if not, whether spin liquids are achievable in GF systems.

The eminuscent-phase scenario seems, at first glance, to contradict experiments in some s=1/2s=1/2 GF systems [4, 37, 38], such as such as Herbertsmithite [39, 40], H3LiIr2O6\mathrm{H}_{3}\mathrm{LiIr}_{2}\mathrm{O}_{6} [41] and Cu3Zn(OH)6FBr\mathrm{Cu}_{3}\mathrm{Zn}(\mathrm{OH})_{6}\mathrm{FBr} [42] that show no signs of spin freezing at temperatures down to 50 mK, and possess features not inconsistent with a spin liquid interpretation. This may be attributable to their effectively 2D nature, which precludes a glass transition at a finite temperature (see Supplemental Information): a system may already be in a glass state for the entire range of accessed temperatures or disallow for the glass phase (e.g., 2D Ising models with bond disorder [43]).

To sum up, in this paper, we have analysed known SG transitions in GF magnets and show that they differ qualitatively from those in conventional glasses. The freezing temperature of such SG transitions grows with decreasing the concentration of vacancies, the main source of disorder in GF magnets, and reaches a finite value, the ``hidden energy scale’’ in the perceived disorder-free limit. This observation calls into question the achievability of quantum spin liquids in GF magnets. Indeed, if a glass transition occurs at a temperature determined by the properties of the medium and not by the amount of disorder, a system may always be entering the spin-glass state instead of becoming a spin liquid. We emphasise that such a possibility would be peculiar to materials with GF lattices and does not concern systems with bipartite lattices, such as the Shastry-Sutherland [44] and Kitaev’s honeycomb [3] models, in which frustration is achieved via tuning of the interactions and not via the lattice structure.

The results in this work call for a thorough theoretical investigation of the role of quenched disorder in GF systems. While weak disorder is perturbatively irrelevant in a GF material [45], non-perturbative fluctuations of the locations of the impurities and inter-spin couplings may still result in a glass state even for very small amounts of disorder.

We have also proposed a phenomenological scenario for the hidden energy scale. We conjecture that the SG transition temperature vanishes at strictly zero disorder. However, for available impurity concentrations to date, glass freezing is driven by the crossover in the magnetic permeability rather than by the amount disorder. As a result, an extrapolation of the SG transition temperature in existing materials to the limit of zero disorder (vacancy density) leads to a finite value. The phenomenological picture of the SG transition developed here calls for a thorough investigation of the critical features of the transition and for synthesis of cleaner GF materials. Another question which deserves a thorough investigation is the possibility of a spin-freezing transition in a GF material in the absence of quenched disorder. We leave the investigation of such a possibility to future studies.

Methods

To illustrate how the discussed crossover in the GF medium may occur, we consider a model of a spin ice or a system with vector spins in the ``Coulomb phase’’ [4, 46, 47, 48]. In such a description, the medium is mapped to classical Coulomb fields B\mathbf{B} and a system of charges (monopoles), topological defects that may be activated thermally or created by impurities. The spin medium between the defects determines the interaction energy Eint=1μ(T)∑i,j<iQiQj∣ri−rj∣E_{\mathrm{int}} = \frac{1}{\mu(T)} \sum_{i,j<i} \frac{Q_i Q_j}{\lvert\mathbf{r}_i-\mathbf{r}_j\rvert} between monopole charges QiQ_i via the effective permeability μ(T)\mu(T).

By introducing the characteristic energy T~\tilde{T} of the static field B\mathbf{B} per spin (arising, e.g., from the dipole-dipole interactions or further-neighbour interactions), we arrive at the permeability (see [13] for details)

μ(T)=T~T~+T(1)\mu(T) = \frac{\tilde{T}}{\tilde{T}+T} \tag*{(1)}

that exhibits a crossover from a constant value μ=1\mu=1 at T≪T~T \ll\tilde{T} to a strongly suppressed value at T≫T~T \gg\tilde{T}. The crossover in the permeability, μ(T)\mu(T), leads to a crossover in the ratio δE/T∝[μ(T)T]−1=T+T~TT~\delta E/T \propto[\mu(T)T]^{-1} = \frac{T+\tilde{T}}{T\tilde{T}} of the fluctuation of the interaction energy of impurity spins to the temperature.

In weakly disordered samples, this ratio is small at high temperatures T≫T~T \gg\tilde{T} and grows rapidly for T≪T~T \ll\tilde{T}, thus favouring the glass state. In the presence of quenched disorder, the crossover in the magnetic permeability (1) will drive a glass transition at a temperature T∗T^* on the order of T~\tilde{T}. While we considered a model of a spin glass here, we expect that such crossovers exist generically in GF media, with the details of the function μ(T)\mu(T) depending on the microscopic details of the Hamiltonian.

We suggest, therefore, that the SG transitions in GF materials observed to date are driven by the GF medium, with the transition temperature T∗=T~T^*=\tilde{T} weakly dependent on the amount of disorder for the lowest achieved concentrations of impurities. The dependence of the critical temperature on disorder implies a region of higher purity, depicted in Figure 2b, that has not yet been probed. A careful verification of the scenario of the glass transition proposed here will require a thorough experimental investigation of the details of the transition and advances in the synthesis of clean GF systems.

Data Availability

The Data used to generate Figure 1, Figure 2 and Table 1 are available from the cited articles published by the Physical Society of Japan, American Physical Society and Institute of Physics.

Acknowledgements

We have benefited from useful discussion with C. Batista, S. Brown, J.T. Chalker, S. Kivelson, S.-H. Lee, R. Moessner, S. Profumo, A.W. Sandvik and A.P. Young. APR would like to acknowledge support by the U.S. Department of Energy Office of Basic Energy Science, Division of Condensed Matter Physics grant DE-SC0017862.

This version of the article has been accepted for publication, after peer review, but is not the Version of Record and does not reflect post-acceptance improvements, or any corrections. The Version of Record is available online at: https://doi.org/10.1038/s41467-022-30739-0

Author contributions

Both authors contributed significantly to the analysis of the data, calculations presented in this work and writing the manuscript.

Competing Interests Statement

The authors declare no competing interests.

Supplemental Material for “Eminescent phase in frustrated magnets: a challenge to quantum spin liquids”

Sergey Syzranov; Arthur Ramirez

SIhiddenEnergy

Supplemental Material for ``How frustrated magnets hide a phase transition and challenge quantum spin liquids''

Details of published data presentation

The data presented in Figure 1 and Figure 2 were obtained by digitising data from published work, referenced in the text. Any error introduced in digitising is less than the size of the symbols in our plots. Converting susceptibility units from the conventional emu/mole to dimensionless units requires knowledge of the mass density. For the end-members of the series, we used either published density values or obtained values from the Materials Project (https://materialsproject.org/). For compounds that are solid-solutions of two end-members, we assumed a Vegard’s law (linear) interpolation of the lattice constant and thus density. Values of TgT_g were taken from the position of the peak in zero-field-cooled susceptibility.

In Figure 2 are plotted TgT_g vs. disorder, xx, for those materials where disorder can be related to a chemical constituent. For most of the compound series, xx represents the concentration of spin vacancies due to the introduction of non-magnetic impurity atoms[1, 2, 3, 4], but it can also indicate substitution on a non-magnetic site[5, 6, 7]. In these cases, if the size of the impurity atoms differs significantly from that of the host atoms, as exemplified by La3+\mathrm{La}^{3+} in Y2Mo2O7\mathrm{Y}_{2}\mathrm{Mo}_{2}\mathrm{O}_{7}, chemical substitution can lead to large values of random strain, leading to a large modulation of the exchange interactions. Thus, the series Y2(1−x)La2xMo2O7\mathrm{Y}_{2(1-x)}\mathrm{La}_{2x}\mathrm{Mo}_{2}\mathrm{O}_{7} is not included in our Figure 2a survey because θW\theta_W changes sign from antiferromagnetic to ferromagnetic as xx increases from 0 to 0.3, indicating that introduction of the La3+\mathrm{La}^{3+} ion for Y3+\mathrm{Y}^{3+} significantly alters the character of the exchange interactions that govern the behaviour of the background spins[8]. The compound Zn1−xCdxCrO4\mathrm{Zn}_{1-x}\mathrm{Cd}_{x}\mathrm{CrO}_{4}, where Cd2+\mathrm{Cd}^{2+} replaces Zn2+\mathrm{Zn}^{2+}, is not included in Figure 2a due to the absence of vacancy variation. At the same time, it is included in Figure 1, as the disorder resulting from the replacement has no effect on frustration[5, 9].

As alluded to in the main text, some of the substitution series shown in Figure 2a do not approach the pure limit. For the (Mn,Cd)Te(\mathrm{Mn},\mathrm{Cd})\mathrm{Te} and Zn(Cr,Ga)2O4\mathrm{Zn}(\mathrm{Cr},\mathrm{Ga})_{2}\mathrm{O}_{4} series, the SG phase is replaced by an antiferromagnetic phase, accompanied by a change in the characteristic 1/χ(T)1/\chi(T) behaviour. For the (Fe,Ga)2TiO5(\mathrm{Fe},\mathrm{Ga})_{2}\mathrm{TiO}_{5} series, the Fe2TiO5\mathrm{Fe}_{2}\mathrm{TiO}_{5} end member has the highest possible Fe content for the pseudobrookite structure. Randomness arises from mixing of the Fe3+Fe^{3+} and Ti4+Ti^{4+} ions among their respective crystallographic sites. Thus, due to the different oxidation states of FeFe and TiTi, increasing Fe3+Fe^{3+} content beyond 2/32/3 is not possible within the same structure without destroying the insulating phase.

Glass transition in a weakly disordered system

In this section, we estimate the glass transition temperature in a system of spins with exchange interactions with randomly located generic defects and demonstrate its vanishing in the limit of the vanishing density nimpn_{\mathrm{imp}} of the defects. Such defects may be represented, for example, by vacancies in the lattice of the system, spinless foreign atoms or perturbations in the exchange interaction between the spins. In order to analyse the glass transition, we employ the replica trick. The replicated Hamiltonian of the system is given by

H=H0+Hdis,(2)\mathcal{H} = \mathcal{H}_0 + \mathcal{H}_{\mathrm{dis}}, \tag*{(2)}
H0=12∑r,r′,αJr,r′srαsr′α,(3)\mathcal{H}_0 = \frac{1}{2}\sum_{r,r',\alpha} J_{r,r'}s_r^\alpha s_{r'}^\alpha, \tag*{(3)}
Hdis=12∑i,αsρiαKr−ρi,r′−ρisr′α,(4)\mathcal{H}_{\mathrm{dis}} = \frac{1}{2}\sum_{i,\alpha} s_{\rho_i}^\alpha K_{r-\rho_i,r'-\rho_i}s_{r'}^\alpha, \tag*{(4)}

where we consider, for simplicity, classical Ising spins sr=±1s_r=\pm1 labelled by their locations rr and r′r' in the lattice and the replica index α\alpha; the Hamiltonian H0\mathcal{H}_0 describes the pure systems; Hdis\mathcal{H}_{\mathrm{dis}} accounts for the effect of the defects; Jrr′J_{r r'} is the coupling between spins at locations rr and r′r' and is a function of only the relative position r−r′r-r'. Quenched disorder is represented by the randomness of the locations ρi\rho_i of the defects in Eq. (1); the matrix Kr1,r2K_{r_1,r_2} is non-random and describes the modifications of the couplings Jr,r′J_{r,r'} caused by one defect located at ρ=0\rho=0.

The defects thus create identical perturbations near different locations ρi\rho_i. For example, vacancies in a lattice with nearest-neighbour exchange coupling Jr,r′=JJ_{r,r'}=J are described by the elements Kr−ρ,0=K0,r−ρ=−JK_{r-\rho,0}=K_{0,r-\rho}=-J at all sites rr adjacent to the location ρ\rho of the vacancy, effectively cancelling the interactions between the spin at the location of the vacancy and the neighbouring spins, and to vanishing elements of Kr1,r2K_{r_1,r_2} for all other positions. If the defect is represented by a foreign atom with the same spin as the atoms in the compound, the elements of the matrix Kr1,r2K_{r_1,r_2} may represent small perturbations of the exchange couplings around a defect. If quenched disorder is represented by random-bond disorder on all sites coming from, e.g., random strain fields, it can be described by the same model with the defect density set to the density of the sites, nimp∼nn_{\mathrm{imp}}\sim n.

We perform averaging of the replicated partition function of the system over the locations ρi\boldsymbol{\rho}_i of all defects, ⟨Z⟩≡1NNimp∑ρ1∑ρ2⋯∑ρNimpZ\langle Z\rangle\equiv\frac{1}{N^{N_{\mathrm{imp}}}}\sum_{\boldsymbol{\rho}_1}\sum_{\boldsymbol{\rho}_2}\cdots\sum_{\boldsymbol{\rho}_{N_{\mathrm{imp}}}} Z, where NN is the total number of sites in the system and NimpN_{\mathrm{imp}} is the number of defects. Keeping the first two cumulants of the impurity-induced perturbation in the action of the disorder-averaged partition function ⟨Z⟩\langle Z\rangle gives the action

f(s)T=12T∑r,r′srαsr′α(Jr,r′+nimpkrr′)−nimp2T2∑r1,r2,r1′,r2′,α,βsr1αsr2β(Fr1r2r1′r2′−nimpkr1r1′kr2r2′)sr1′αsr2′β,(5)\frac{f^{(s)}}{T} = \frac{1}{2T}\sum_{\mathbf{r},\mathbf{r}'}s_{\mathbf{r}}^\alpha s_{\mathbf{r}'}^\alpha\left(J_{\mathbf{r},\mathbf{r}'}+n_{\mathrm{imp}}k_{\mathbf{r}\mathbf{r}'}\right)-\frac{n_{\mathrm{imp}}}{2T^2}\sum_{\mathbf{r}_1,\mathbf{r}_2,\mathbf{r}_1',\mathbf{r}_2',\alpha,\beta}s_{\mathbf{r}_1}^\alpha s_{\mathbf{r}_2}^\beta\left(F_{\mathbf{r}_1\mathbf{r}_2\mathbf{r}_1'\mathbf{r}_2'}-n_{\mathrm{imp}}k_{\mathbf{r}_1\mathbf{r}_1'}k_{\mathbf{r}_2\mathbf{r}_2'}\right)s_{\mathbf{r}_1'}^\alpha s_{\mathbf{r}_2'}^\beta, \tag*{(5)}

where we have introduced the ratio nimp=Nimp/Nn_{\mathrm{imp}}=N_{\mathrm{imp}}/N of the number of defects to the total number of sites in the lattice and the quantities

krr′=∑ρKr−ρ,r′−ρ,(6)k_{\mathbf{r}\mathbf{r}'}=\sum_{\boldsymbol{\rho}}K_{\mathbf{r}-\boldsymbol{\rho},\mathbf{r}'-\boldsymbol{\rho}}, \tag*{(6)}
Fr1r2r1′r2′=14∑ρKr1−ρ,r1′−ρKr2−ρ,r2′−ρ.(7)F_{\mathbf{r}_1\mathbf{r}_2\mathbf{r}_1'\mathbf{r}_2'}=\frac{1}{4}\sum_{\boldsymbol{\rho}}K_{\mathbf{r}_1-\boldsymbol{\rho},\mathbf{r}_1'-\boldsymbol{\rho}}K_{\mathbf{r}_2-\boldsymbol{\rho},\mathbf{r}_2'-\boldsymbol{\rho}}. \tag*{(7)}

Taking into account higher-order cumulants of the disorder-induced perturbation may change coefficients of order unity in the mean-field phase boundary of the glass phase but will not lead to qualitatively new effects.

Decoupling the variables srαsr′βs_{\mathbf{r}}^\alpha s_{\mathbf{r}'}^\beta by the Hubbard-Stratonovich field Qrr′αβQ_{\mathbf{r}\mathbf{r}'}^{\alpha\beta} gives

f(s,Q)T=12TsrαJ~r,r′sr′α−nimp2T2Qr1r2αβF~r1r2r1′r2′sr1αsr2β−nimp2T2sr1αsr2βF~r1r2r1′r2′Qr1′r2′αβ+nimp2T2Qr1r2αβF~r1r2r1′r2′Qr1′r2′αβ,(8)\begin{aligned} \frac{f(s,Q)}{T}={}&\frac{1}{2T}s_{\mathbf{r}}^\alpha\widetilde{J}_{\mathbf{r},\mathbf{r}'}s_{\mathbf{r}'}^\alpha-\frac{n_{\mathrm{imp}}}{2T^2}Q_{\mathbf{r}_1\mathbf{r}_2}^{\alpha\beta}\widetilde{F}_{\mathbf{r}_1\mathbf{r}_2\mathbf{r}_1'\mathbf{r}_2'}s_{\mathbf{r}_1}^\alpha s_{\mathbf{r}_2}^\beta\\ &-\frac{n_{\mathrm{imp}}}{2T^2}s_{\mathbf{r}_1}^\alpha s_{\mathbf{r}_2}^\beta\widetilde{F}_{\mathbf{r}_1\mathbf{r}_2\mathbf{r}_1'\mathbf{r}_2'}Q_{\mathbf{r}_1'\mathbf{r}_2'}^{\alpha\beta}+\frac{n_{\mathrm{imp}}}{2T^2}Q_{\mathbf{r}_1\mathbf{r}_2}^{\alpha\beta}\widetilde{F}_{\mathbf{r}_1\mathbf{r}_2\mathbf{r}_1'\mathbf{r}_2'}Q_{\mathbf{r}_1'\mathbf{r}_2'}^{\alpha\beta}, \tag*{(8)} \end{aligned}

where J~r,r′=Jr,r′+nimpkrr′\widetilde{J}_{\mathbf{r},\mathbf{r}'}=J_{\mathbf{r},\mathbf{r}'}+n_{\mathrm{imp}}k_{\mathbf{r}\mathbf{r}'} and F~r1r2r1′r2′=Fr1r2r1′r2′−nimpkr1r1′kr2r2′\widetilde{F}_{\mathbf{r}_1\mathbf{r}_2\mathbf{r}_1'\mathbf{r}_2'}=F_{\mathbf{r}_1\mathbf{r}_2\mathbf{r}_1'\mathbf{r}_2'}-n_{\mathrm{imp}}k_{\mathbf{r}_1\mathbf{r}_1'}k_{\mathbf{r}_2\mathbf{r}_2'} and summation over repeated indices is implied. We note that in the limit of low concentrations nimpn_{\mathrm{imp}} of the defects, their effect on the quantities J~r,r′≈Jr,r′\widetilde{J}_{\mathbf{r},\mathbf{r}'}\approx J_{\mathbf{r},\mathbf{r}'} and F~r1r2r1′r2′≈Fr1r2r1′r2′\widetilde{F}_{\mathbf{r}_1\mathbf{r}_2\mathbf{r}_1'\mathbf{r}_2'}\approx F_{\mathbf{r}_1\mathbf{r}_2\mathbf{r}_1'\mathbf{r}_2'} in Eq. (8) may be neglected.

Summing over the spin degrees of freedom leads to the free energy as a function of the fields QQ given by

F(Q)=nimp2TQr1r2αβF~r1r2r1′r2′Qr1′r2′αβ−nimp28T3Qr1r2αβF~r1r2r1′r2′Gr1r1′Gr2r2′F~r1′mathbfr2′r1r2Qr1r2αβ−nimp28T3Qr1r2αβF~r1r2r1′r2′Gr1′r1Gr2′r2F~r2′r1′r1r2Qr1r2βα+O(Q3),(9)\begin{aligned} F(Q)={}&\frac{n_{\mathrm{imp}}}{2T}Q_{\mathbf{r}_1\mathbf{r}_2}^{\alpha\beta}\widetilde{F}_{\mathbf{r}_1\mathbf{r}_2\mathbf{r}_1'\mathbf{r}_2'}Q_{\mathbf{r}_1'\mathbf{r}_2'}^{\alpha\beta}\\ &-\frac{n_{\mathrm{imp}}^2}{8T^3}Q_{\mathbf{r}_1\mathbf{r}_2}^{\alpha\beta}\widetilde{F}_{\mathbf{r}_1\mathbf{r}_2\mathbf{r}_1'\mathbf{r}_2'}G_{\mathbf{r}_1\mathbf{r}_1'}G_{\mathbf{r}_2\mathbf{r}_2'}\widetilde{F}_{\mathbf{r}_1'\mathbf{mathbf{r}_2'}\mathbf{r}_1\mathbf{r}_2}Q_{\mathbf{r}_1\mathbf{r}_2}^{\alpha\beta}\\ &-\frac{n_{\mathrm{imp}}^2}{8T^3}Q_{\mathbf{r}_1\mathbf{r}_2}^{\alpha\beta}\widetilde{F}_{\mathbf{r}_1\mathbf{r}_2\mathbf{r}_1'\mathbf{r}_2'}G_{\mathbf{r}_1'\mathbf{r}_1}G_{\mathbf{r}_2'\mathbf{r}_2}\widetilde{F}_{\mathbf{r}_2'\mathbf{r}_1'\mathbf{r}_1\mathbf{r}_2}Q_{\mathbf{r}_1\mathbf{r}_2}^{\beta\alpha}+O(Q^3), \tag*{(9)} \end{aligned}

where Grr′=⟨srαsr′α⟩G_{\mathbf{r}\mathbf{r}'}=\langle s_{\mathbf{r}}^\alpha s_{\mathbf{r}'}^\alpha\rangle is the correlator of spins in the same replica in the pure system. To obtain the phase boundary of the glass phase, we look for the mean-field solution for the fields Qr1r2αβQ_{\mathbf{r}_1\mathbf{r}_2}^{\alpha\beta}. This solution has the same symmetry Qr1r2αβ=Qr2r1αβ=Qr1r2βαQ_{r_1r_2}^{\alpha\beta}=Q_{r_2r_1}^{\alpha\beta}=Q_{r_1r_2}^{\beta\alpha} as the averaged correlators ⟨sr1αsr2β⟩\langle s_{r_1}^{\alpha}s_{r_2}^{\beta}\rangle. Furthermore, although the glass phase breaks replica symmetry[10, 11, 12], it is sufficient to look for the replica-symmetric solution to obtain the phase boundary.

Taking into account the symmetry Qr1r2αβ=Qr2r1αβ=Qr1r2βαQ_{r_1r_2}^{\alpha\beta}=Q_{r_2r_1}^{\alpha\beta}=Q_{r_1r_2}^{\beta\alpha}, the free energy (6) can be rewritten as

f(Q^)=nimp2TQ^F~Q^−nimp24T3Q^F~(G^⊗G^)F~Q^+O(Q3),(10)f(\widehat{Q})=\frac{n_{\mathrm{imp}}}{2T}\widehat{Q}\widetilde{F}\widehat{Q}-\frac{n_{\mathrm{imp}}^2}{4T^3}\widehat{Q}\widetilde{F}(\widehat{G}\otimes\widehat{G})\widetilde{F}\widehat{Q}+O(Q^3), \tag*{(10)}

where Q^\widehat{Q} is our convention for the matrix Qr1r2αβQ_{r_1r_2}^{\alpha\beta} and F~\widetilde{F} and G^⊗G^\widehat{G}\otimes\widehat{G} are superoperators acting on matrices of the respective dimension; Q^F~Q^≡Qr1r2αβF~r1r2r1′r2′Qr1′r2′αβ\widehat{Q}\widetilde{F}\widehat{Q}\equiv Q_{r_1r_2}^{\alpha\beta}\widetilde{F}_{r_1r_2r'_1r'_2}Q_{r'_1r'_2}^{\alpha\beta} and Q^F~(G^⊗G^)F~Q^≡Qr1r2αβF~r1r2r1′r2′Gr1′R1′Gr2′R2′F~R1′R2′R1R2QR1R2αβ\widehat{Q}\widetilde{F}(\widehat{G}\otimes\widehat{G})\widetilde{F}\widehat{Q}\equiv Q_{r_1r_2}^{\alpha\beta}\widetilde{F}_{r_1r_2r'_1r'_2}G_{r'_1R'_1}G_{r'_2R'_2}\widetilde{F}_{R'_1R'_2R_1R_2}Q_{R_1R_2}^{\alpha\beta}. The mean-field transition temperature is given by

T=(12nimpλ)12,(11)T=\left(\frac{1}{2}n_{\mathrm{imp}}\lambda\right)^{\frac{1}{2}}, \tag*{(11)}

where λ\lambda is the minimum eigenvalue of the (positive) operator F~(G^⊗G^)\widetilde{F}(\widehat{G}\otimes\widehat{G}). Because this operator is independent of the the defect concentration in the limit of low defect density nimpn_{\mathrm{imp}}, the glass transition temperature vanishes in the limit of vanishing defect concentration.

We emphasise that the result (8) can be applied to models of disorder of various nature, including vacancies in GF systems as well as random-bond disorder, or combinations of different sources of disorder. For weak random-bond disorder on all bonds (n∼nimpn\sim n_{\mathrm{imp}}), Eq. (8) predicts a transition temperature proportional to the amplitude of the fluctuation of the exchange coupling, in agreement with the result of Refs. 6 and 7.

On the possibility of glassiness in 2D

Some of the GF compounds discussed in our work (SrCr8Ga4O19SrCr_{8}Ga_{4}O_{19}, BaCr9−9xGa3+9xO19BaCr_{9-9x}Ga_{3+9x}O_{19} and NiGa2S4NiGa_{2}S_{4}) are layered with weak interlayer coupling and, therefore, may be considered as effectively two-dimensional (2D) systems at short distances. In this section, we argue that the 3D nature of those systems, which accounts for the interlayer coupling, is essential to observe the glass transition in them.

There is abundant numerical evidence (see, for example, Refs. 13–18) of the absence of a glass state at a finite temperature T>0T>0 in 2D spin systems with bond disorder. This changes in the presence of field disorder, for a system described by the Hamiltonian

H=H0−∑ibisi,(12)\mathcal{H}=\mathcal{H}_0-\sum_i b_i s_i, \tag*{(12)}

where bib_i is a random field. In this case, the system has a finite glass order parameter

qαβ=⟨siαsiβ⟩=Tr⁡[siαsiβexp⁡(−1T∑γH0γ−ϰ2T2∑j,γ,γ′sjγsjγ′)]>0(13)q^{\alpha\beta}=\left\langle s_i^\alpha s_i^\beta\right\rangle=\operatorname{Tr}\left[s_i^\alpha s_i^\beta\exp\left(-\frac{1}{T}\sum_\gamma H_0^\gamma-\frac{\varkappa}{2T^2}\sum_{j,\gamma,\gamma'}s_j^\gamma s_j^{\gamma'}\right)\right]>0 \tag*{(13)}

due to the identity

∂qαβ∂ϰ=−1T2(14)\frac{\partial q^{\alpha\beta}}{\partial\varkappa}=-\frac{1}{T^2} \tag*{(14)}

in the limit of weak disorder strength ϰ=⟨bj2⟩dis\varkappa=\langle b_j^2\rangle_{\mathrm{dis}}. The presence of a finite glass order parameter in a system with random-field disorder is similar, in some sense, to the existence of a finite magnetisation in a magnet in a magnetic field. Because the order parameter qαβq^{\alpha\beta} is nonzero at all temperatures, there is no phase transition between the glass (qαβ≠0q^{\alpha\beta}\ne0) and non-glass (qαβ=0q^{\alpha\beta}=0) phases.

Both in the case of bond- and random-field disorder, the glass transition is absent; the system either has the glass order parameter at all temperatures or never has it. Because the glass transition is observed in all geometrically frustrated compounds discussed in the main text of this paper, the three-dimensional nature of the discussed layered compounds (SrCr8Ga4O19SrCr_{8}Ga_{4}O_{19}, BaCr9−9xGa3+9xO19BaCr_{9-9x}Ga_{3+9x}O_{19} and NiGa2S4NiGa_{2}S_{4}) is, therefore, important for the existence of the transition.

Crossover in the magnetic permeability in the Coulomb phase

In this section, we describe the mechanism of a crossover in the behaviour of a GF medium that drives the glass transition in a spin ice or a system of classical vector spins. Such a crossover is a property of the clean system and, albeit insufficient to cause a transition on its own, determines the conditions favourable for the onset of the glass phase.

We assume that the medium is described by the Coulomb phase[19, 20, 21, 22, 23]. The role of the variables in this description is played by a coarse-grained classical field B\mathbf{B} associated with the polarisations of the spins. In the absence of impurities and interactions other than the exchange interactions between the spins, the classical ground states of the system (states that minimise the total exchange energy) are hugely degenerate, corresponding to all configurations of the field B\mathbf{B} that satisfy the condition div⁡B=0\operatorname{div}\mathbf{B}=0.

Flipping a spin creates a pair of monopoles with opposite charges that can be further separated by local rotations of the spins. Creating an isolated monopole requires a large energy of the order of the AF coupling and will not be central to our arguments here. Defects with spins that mismatch the spins of the parent compound also correspond to monopole charges QiQ_i at the locations of the defects.

Interactions other than the antiferromagnetic exchange interactions between the spins may lift the degeneracy of the abovementioned classical ground states. Here, we assume that such interactions are present and, in the limit of small B\mathbf{B}, lead to the temperature-independent contribution 18π∫B2dr\frac{1}{8\pi}\int\mathbf{B}^{2}\mathrm{d}r (measured in appropriate units) to the free energy of the system. It may come, for example, from the magnetic dipole-dipole interactions between the spins or from interactions beyond nearest neighbours[24]. Another possible mechanism for generating this energy is quantum tunnelling[25, 26, 27] between classical ground states of the regions with div⁡B=0\operatorname{div}\mathbf{B}=0. While it does not require the existence of an energy scale other than the exchange energy, the amplitude of the tunnelling is in general exponentially suppressed by the number of spins that need to be rotated in order to arrive from one classical ground state to another [17, 18, 19]. The exact nature of the energy of the field B\mathbf{B} is not important for our consideration and cannot be uniquely inferred from the available data, which is why we leave the investigation of the origin of this energy for future studies.

The free energy of the systems also includes the “entropic” contribution 18πTT~∫B2dr\frac{1}{8\pi}\frac{T}{\widetilde{T}}\int\mathbf{B}^{2}\mathrm{d}r linear in the temperature TT, scaled by a constant prefactor 1/T~1/\widetilde{T}. This contribution accounts for the decrease of the entropy of the systems with increasing the parameter B\mathbf{B}.

Combining all the discussed contributions gives the free energy of low-energy excitations in a GF material in the form

F(ϕ,B)=18πT+T~T~∫B2dr−14π∫ϕdiv⁡B dr+∑i[ϕ(ri)+EQ]Qi−∫h⋅B dr−18π∫h0B2 dr(15)F(\phi,\mathbf{B})=\frac{1}{8\pi}\frac{T+\widetilde{T}}{\widetilde{T}}\int\mathbf{B}^{2}\mathrm{d}r-\frac{1}{4\pi}\int\phi\operatorname{div}\mathbf{B}\,\mathrm{d}r+\sum_i[\phi(\mathbf{r}_i)+E_Q]Q_i-\int\mathbf{h}\cdot\mathbf{B}\,\mathrm{d}r-\frac{1}{8\pi}\int h_0\mathbf{B}^{2}\,\mathrm{d}r \tag*{(15)}

where QiQ_i and ri\mathbf{r}_i are the charges and the locations of the monopoles introduced by the defects; EQE_Q is the energy of the defect, which may depend on the state of the defect spin; h(r)\mathbf{h}(r) and h0(r)h_0(r) are weak quenched random fields that account for disorder other than monopole charges created by magnetic vacancies, e.g. random strain. In Eq. (S1), we have introduced the field ϕ\phi, minimisation with respect to which enforces the Gauss theorem in the form div⁡B=4π∑iQiδ(r−ri)\operatorname{div}\mathbf{B}=4\pi\sum_i Q_i\delta(\mathbf{r}-\mathbf{r}_i). For states that minimise the free energy (12), the scalar field ϕ\phi plays the role of the potential of the Coulomb field, ∇ϕ=−T+T~T~B+4πh+h0B\nabla\phi=-\frac{T+\widetilde{T}}{\widetilde{T}}\mathbf{B}+4\pi\mathbf{h}+h_0\mathbf{B}, where the random fields 4πh4\pi\mathbf{h} and h0Bh_0\mathbf{B} come from the fluctuations of the effective magnetisation and magnetic permeability. The free energy (S1) describes a system of randomly located magnetic charges (defects) in a disordered medium with the average magnetic permeability determined by the properties of the disorder-free system and given by Eq. (1) in the main text.

We emphasise that the permeability μ(T)\mu(T) given by Eq. (1) characterises the interaction between excitations in the system and is in general unrelated to the magnetic susceptibility χ(T)\chi(T), which characterises response to the external magnetic field. The energy E≈E0+1μ(T)∑i,j<iQiQj∣ri−rj∣E \approx E_0 + \frac{1}{\mu(T)}\sum_{i,j<i}\frac{Q_iQ_j}{\lvert r_i-r_j\rvert} of a system of monopoles in the material consists of a temperature-independent contribution E0E_0 determined by the exchange coupling between spins and the interaction contribution ∝[μ(T)]−1\propto[\mu(T)]^{-1}. The temperature dependence of the permeability μ(T)\mu(T) can thus be independently verified via spin resonance spectroscopy of multi-charge, e.g, dipole, excitations in the system.

The glass state in the system described by the free energy (S1) requires sufficiently strong disorder represented by randomly located impurities, which create monopole charges QiQ_i, or other sources of randomness represented by the quenched parameters h\mathbf{h} and h0h_0 in Eq. (1). We emphasise that the hidden energy scale T∗T^*, the glass transition temperature at zero impurity concentration, cannot be explained solely by the residual disorder distinct from impurities because the glass transition temperature grows with decreasing disorder. This trend, the exact microscopic mechanism of which we leave for future studies, suggests also a significant role of the disorder-free medium for the “hidden” energy scale T∗T^* discussed in this paper.

References

References

  1. [1]P. W. Anderson, Ordering and antiferromagnetism in ferrites, Physical Review 102, 1008 (1956).
  2. [2]A. P. Ramirez, Thermodynamic measurements on geometrically frustrated magnets, Journal of Applied Physics 70, 5952 (1991).DOI
  3. [3]A. Kitaev, Anyons in an exactly solved model and beyond, Annals of Physics 321, 2 (2006).DOI
  4. [4]L. Savary, and L. Balents, Quantum spin liquids: a review, Reports on Progress in Physics 80, 016502, (2017).DOI
  5. [5]A. P. Ramirez, G. P. Espinosa, and A. S. Cooper, Strong Frustration and Dilution-Enhanced Order in a Quasi-2d Spin-Glass, Physical Review Letters 64, 2070 (1990).pubmed.ncbi.nlm.nih.gov/10041569
  6. [6]S. Nakatsuji, Y. Nambu, H. Tonomura, O. Sakai, S. Jonas, C. Broholm, H. Tsunetsugu, Y. M. Qiu, and Y. Maeno, Spin disorder on a triangular lattice, Science 309, 1697 (2005).pubmed.ncbi.nlm.nih.gov/16151004
  7. [7]P. W. Anderson, B. I. Halperin, and C. M. Varma, Anomalous low-temperature thermal properties of glasses and spin glasses, Philosophical Magazine 25, 1 (1972).
  8. [8]A. Andreanov, J. T. Chalker, T. E. Saunders, and D. Sherrington, Spin-glass transition in geometrically frustrated antiferromagnets with weak disorder, Physical Review B 81, 014406, (2010).DOI
  9. [9]J. E. Greedan, M. Sato, Y. Xu, and F. S. Razavi, Spin-glass-like behavior in Y₂Mo₂O₇, a concentrated, crystalline system with negligible apparent disorder, Solid State Communications 59, 895 (1986).DOI
  10. [10]M. J. P. Gingras, C. V. Stager, B. D. Gaulin, N. P. Raju, and J. E. Greedan, Nonlinear susceptibility measurements at the spin-glass transition of the pyrochlore antiferromagnet Y₂Mo₂O₇, Journal of Applied Physics 79, 6170 (1996).DOI
  11. [11]M. Mezard, G. A. Parisi, and M. Virasoro, Spin Glass Theory and Beyond (World Scientific, Singapore, 1986), Vol. 9, World Scientific Lecture Notes in Physics.
  12. [12]V. S. Dotsenko, Physics of a spin-glass state, Uspekhi Fizicheskikh Nauk 163, 1 (1993).
  13. [13]Supplemental Information.
  14. [14]A. P. Ramirez, Strongly Geometrically Frustrated Magnets, Annual Review of Materials Science 24, 453 (1994).
  15. [15]A. Sen, K. Damle, and R. Moessner, Fractional Spin Textures in the Frustrated Magnet SrCr₉pGa₁₂-₉pO₁₉, Physical Review Letters 106, 127203, (2011).DOI
  16. [16]A. D. LaForge, S. H. Pulido, R. J. Cava, B. C. Chan, and A. P. Ramirez, Quasispin Glass in a Geometrically Frustrated Magnet, Physical Review Letters 110, 017203, (2013).pubmed.ncbi.nlm.nih.gov/23383834
  17. [17]P. M. M. Thygesen, J. A. M. Paddison, R. H. Zhang, K. A. Beyer, K. W. Chapman, H. Y. Playford, M. G. Tucker, D. A. Keen, M. A. Hayward, and A. L. Goodwin, Orbital Dimer Model for the Spin-Glass State in Y₂Mo₂O₇, Physical Review Letters 118, 067201, (2017).DOI
  18. [18]Y. Ying, W. Zhang, J. Yu, L. Qiao, J. W. Zheng, W. C. Li, J. Li, W. Cai, and S. L. Che, Structure and Magnetic Properties of the Ti-Doped Pyrochlore Molybdate Y₂Mo₂(1-x)Ti₂xO₇, Journal of Superconductivity and Novel Magnetism 32, 3563 (2019).DOI
  19. [19]Y. Okamoto, M. Nohara, H. Aruga-Katori, and H. Takagi, Spin-liquid state in the S=1/2 hyperkagome antiferromagnet Na₄Ir₃O₈, Physical Review Letters 99, 137207, (2007).DOI
  20. [20]H. Maletta, and W. Felsch, Insulating spin-glass system EuₓSr₁-xS, Physical Review B 20, 1245 (1979).DOI
  21. [21]Y. Nambu, S. Nakatsuji, and Y. Maeno, Coherent behavior and nonmagnetic impurity effects of spin disordered state in NiGa₂S₄, Journal of the Physical Society of Japan 75, 043711, (2006).DOI
  22. [22]V. Cannella, and J. A. Mydosh, Magnetic ordering in gold-iron alloys, Physical Review B-Solid State 6, 4220 (1972).DOI
  23. [23]S. Nagata, P. H. Keesom, and H. R. Harrison, Low dc-field susceptibility of CuMn spin glass, Physical Review B 19, 1633 (1979).DOI
  24. [24]D. Fiorani, J. L. Dormann, J. L. Tholence, and J. L. Soubeyroux, From the antiferromagnetic regime to the spin-glass state in the frustrated spinel system ZnCr₂xGa₂-2xO₄ Journal of Physics C-Solid State Physics 18, 3053 (1985).DOI
  25. [25]B. Martinez, F. Sandiumenge, A. Rouco, A. Labarta, J. Rodriguez-Carvajal, M. T. T. Causa, S. Gali, and X. Obradors, Magnetic dilution in the strongly frustrated kagome antiferromagnet SrGa₁₂-xCrxO₁₉, Physical Review B 46, 10786 (1992).
  26. [26]S. B. Oseroff, Magnetic susceptibility and EPR measurements in concentrated spin-glasses: Cd₁₋ₓMnₓTe and Cd₁₋ₓMnₓSe, Physical Review B 25, 6584 (1982).DOI
  27. [27]W. Ratcliff, S. H. Lee, C. Broholm, S. W. Cheong, and Q. Huang, Freezing of spin correlated nanoclusters in a geometrically frustrated magnet, Physical Review B 65, 220406, (2002).DOI
  28. [28]D. Phelan, P. G. LaBarre, F. Ye, T. Besara, T. Siegrist, S. Rosenkranz, S. V. Syzranov, and A. P. Ramirez, Behavior of the spin glass phase in the pseudobrookite Fe₍₃₋₃ₓ₎Ga₍₃ₓ₋₁₎TiO₅, to appear in arXiv (2022).
  29. [29]A. Balodhi, A. Thamizhave l, and Y. Singh, Evolution of magnetic, transport, and thermal properties in Na₄₋ₓIr₃O₈, Physical Review B 91, 224409, (2015).DOI
  30. [30]P. G. LaBarre, D. Phelan, Y. Xin, F. Ye, T. Besera, T. Siegrist, S. V. Syzranov, S. Rosenkranz, and A. P. Ramirez, Fluctuation-induced interactions and the spin-glass transition in Fe₂TiO₅, Physical Review B 103, L220404, (2021).DOI
  31. [31]C. Broholm, G. Aeppli, G. P. Espinosa, and A. S. Cooper, Antiferromagnetic fluctuations and short-range order in a kagome lattice, Physical Review Letters 65, 3173 (1990).
  32. [32]J. S. Gardner, B. D. Gaulin, S. H. Lee, C. Broholm, N. P. Raju, and J. E. Greedan, Glassy statics and dynamics in the chemically ordered pyrochlore anti-ferromagnet Y₂Mo₂O₇, Physical Review Letters 83, 211 (1999).
  33. [33]C. Stock, S. Jonas, C. Broholm, S. Nakatsuji, Y. Nambu, K. Onuma, Y. Maeno, and J. H. Chung, Neutron-Scattering Measurement of Incommensurate Short-Range Order in Single Crystals of the S=1 Triangular Antiferromagnet NiGa₂S₄, Physical Review Letters 105, 037402, (2010).pubmed.ncbi.nlm.nih.gov/20867806
  34. [34]H. Martinho, N. O. Moreno, J. A. Sanjurjo, C. Rettori, A. J. Garcia-Adeva, D. L. Huber, S. B. Oseroff, W. Ratcliff, S. W. Cheong, P. G. Pagliuso, J. L. Sarrao, and G. B. Martins, Magnetic properties of the frustrated antiferromagnetic spinel ZnCr₂O₄ and the spin-glass Zn₁₋ₓCdₓCr₂O₄ (x=0.05,0.10), Physical Review B 64, 024408, (2001).
  35. [35]J. J. Yang, A. M. Samarakoon, K. W. Hong, J. R. D. Copley, Q. Z. Huang, A. Tennant, T. J. Sato, and S. H. Lee, Glassy Behavior and Isolated Spin Dimers in a New Frustrated Magnet BaCr₉ₚGa₁₂₋₉ₚO₁₉, Journal of the Physical Society of Japan 85, 094712, (2016).DOI
  36. [36]J. Villain, Spin glass with nonrandom interactions, Journal of Physics C-Solid State Physics 10, 1717 (1977).
  37. [37]C. Broholm, R. J. Cava, S. A. Kivelson, D. G. Nocera, M. R. Norman, and T. Senthil, Quantum spin liquids, Science 367, 263 aay0668, (2020).
  38. [38]J. R. Chamorro, T. M. McQueen, and T. T. Tran, Chemistry of Quantum Spin Liquids, Chemical Reviews 121, 2898 (2021).DOI
  39. [39]J. S. Helton, K. Matan, M. P. Shores, E. A. Nytko, B. M. Bartlett, Y. Yoshida, Y. Takano, A. Suslov, Y. Qiu, J. H. Chung, D. G. Nocera, and Y. S. Lee, Spin dynamics of the spin-1/2 Kagome lattice antiferromagnet ZnCu₃(OH)₆Cl₂, Physical Review Letters 98, 107204, (2007).pubmed.ncbi.nlm.nih.gov/17358563
  40. [40]I. Kimchi, J. P. Sheckelton, T. M. McQueen, and P. A. Lee, Scaling and data collapse from local moments in frustrated disordered quantum spin systems, Nature Communications 9, 4367 (2018).DOI
  41. [41]K. Kitagawa, T. Takayama, Y. Matsumoto, A. Kato, R. Takano, Y. Kishimoto, S. Bette, R. Dinnebier, G. Jackeli, and H. Takagi, A spin–orbital-entangled quantum liquid on a honeycomb lattice, Nature 554, 341 (2018).DOI
  42. [42]Z. L. Feng, Z. Li, X. Meng, W. Yi, Y. Wei, J. Zhang, Y. C. Wang, W. Jiang, Z. Liu, S. Y. Li, F. Liu, J. L. Luo, S. L. Li, G. Q. Zheng, Z. Y. Meng, J. W. Mei, and Y. G. Shi, Gapped Spin-1/2 Spinon Excitations in a New Kagome Quantum Spin Liquid Compound Cu₃Zn(OH)₆FBr, Chinese Physics Letters 34, 077502, (2017).DOI
  43. [43]L. A. Fernandez, E. Marinari, V. Martin-Mayor, G. Parisi, and J. J. Ruiz-Lorenzo, Universal critical behavior of the two-dimensional Ising spin glass, Physical Review B 94, (2016).DOI
  44. [44]B. S. Shastry, and B. Sutherland, Exact Ground-State of a Quantum-Mechanical Antiferromagnet, Physica B & C 108, 1069 (1981).DOI
  45. [45]L. Savary, and L. Balents, Disorder-Induced Quantum Spin Liquid in Spin Ice Pyrochlores, Phys Rev Lett 118, 087203 (2017).
  46. [46]S. V. Isakov, K. Gregor, R. Moessner, and S. L. Sondhi, Dipolar spin correlations in classical pyrochlore magnets, Physical Review Letters 93, 167204, (2004).
  47. [47]M. Hermele, M. P. A. Fisher, and L. Balents, Pyrochlore photons: The U(1) spin liquid in a S=1/2 three-dimensional frustrated magnet, Physical Review B 69, 064404, (2004).DOI
  48. [48]C. L. Henley, Power-law spin correlations in pyrochlore antiferromagnets, Physical Review B 71, 014424, (2005).DOI

Paper details

vlnc:W_xjd88kks
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents