Abstract
The description and detection of unconventional magnetic states, such as spin liquids, is a recurring topic in condensed matter physics. While much of the efforts have traditionally been directed at geometrically frustrated antiferromagnets, recent studies reveal that systems featuring competing antiferromagnetic and ferromagnetic interactions are also promising candidate materials. We find that this competition leads to the notion of special temperatures, analogous to those of gases, at which the competing interactions balance, and the system is quasiideal. Although induced by weak perturbing interactions, these special temperatures are surprisingly high and constitute an accessible experimental diagnostic of eventual order or spinliquid properties. The well characterised Hamiltonian and extended lowtemperature susceptibility measurement of the canonical frustrated ferromagnet Dy_{2}Ti_{2}O_{7} enables us to formulate both a phenomenological and microscopic theory of special temperatures for magnets. Other members of this class of magnets include kapellasite Cu_{3}Zn(OH)_{6}Cl_{2} and the spinel GeCo_{2}O_{4}.
Introduction
Models of magnetic frustration on regular lattices have naturally tended to focus on the case where there is a single interaction of one sign that is frustrated by the lattice geometry. Examples include the triangular or kagome lattice antiferromagnets^{1,2,3}, the pyrochlore Heisenberg antiferromagnet^{4, 5} and spin ice in the nearneighbour approximation, a frustrated ferromagnet^{6}. While there are many real materials that roughly approximate these ideal models^{7,8,9}, the nature of real magnetic interactions is such that a competition between antiferromagnetic (AF) and ferromagnetic (FM) interactions is commonly encountered. This arises because the superexchange interaction is fundamentally the difference between two large numbers—an AF and a FM part—and small differences in orbital overlap can tip it in one direction or the other^{10}. Also, the dipole–dipole interaction, which is important in rare earth systems, has a sign that depends sensitively on direction. Hence, while nearneighbour interactions are of one sign, further neighbour interactions may be of the opposite sign. In the context of a geometrically frustrated lattice, it has recently been recognised that this competition can produce some interesting effects, including spinliquid behaviour^{11, 12}, magnetic fragmentation^{13}, competing ground states^{14, 15} and spin glass physics^{16}. Many of these materials show a conspicuous broad peak in χT/C (where χ is the magnetic susceptibility and C the Curie parameter), which is the analogue of the product pV/nRT in gas thermodynamics, and the focus of this work.
Classical gases exhibit a number of temperature values that signal transitions between contrasting physical properties^{17, 18}. We label these ‘special temperatures’ to emphasise that they do not simply reflect characteristic or typical energy scales. They include the Boyle and Joule temperatures, with the most notable one being perhaps the Joule–Thomson (or inversion) temperature, T_{JT}, below which a gas may be liquefied by the Linde–Hampson process, which underpins a vast lowtemperature technology. A particularly remarkable aspect of T_{JT} is how large it is. For example, for nitrogen, T_{JT} = 621 K, even though the thermally averaged potential energy that gives rise to the finite T_{JT} accounts for only about one thousandth of the internal energy of the system. In terms of the van der Waals equation of state, p = \(\frac{{RT}}{{V{\mathrm{/}}n  b}}  \frac{{an^2}}{{V^2}}\), T_{JT} = 2a/(bR) ≈ 27T_{c}/4. Hence, T_{JT} presents a surprising signature of the eventual liquid state in a temperature regime where at first sight, the intermolecular interactions are negligible. Until now, the magnetic analogies of these special temperatures foreshadowing phenomena at much lower temperature appear not to have been noticed.
In this work we put forward the concept of a class of 'inverting' frustrated ferromagnets, which exhibit a maximum in χT/C as a function of temperature. In strong analogy with the theory of classical gases, we identify the peak in χT/C with a magnetic Joule temperature, T_{J}, where the system is quasiideal, and the internal energy U is independent of the magnetisation M, (∂U/∂M)_{ T } = 0. The Joule temperature marks the onset of the lowtemperature antiferromagnetic correlations. In addition, we identify and define a magnetic Boyle temperature, T_{B}, at which point χT/C = 1, and the incipient ferromagnetic correlations cross over to antiferromagnetic at low temperature. So while the magnitude of χT/C can be used to classify magnets as ferromagnets (χT/C > 1) or antiferromagnets (χT/C < 1)^{19, 20}, we here focus on the special temperature values (points) of χT/C.
Results
Spin ice as a model inverting magnet
Theoretically, the physics of special temperatures is hard to expose computationally in quantum spin systems due to the sign problem^{21}. For frustrated systems with strong FM interactions, a major challenge is to control demagnetising effects^{22}. This makes the canonical frustrated ferromagnet spin ice^{6, 7, 23,24,25,26,27} Dy_{2}Ti_{2}O_{7} a natural starting point to explore the physics of competing FM–AF interactions. In this material, nearneighbour dipolar and exchange interactions average to a ferromagnetic coupling and a mapping to Pauling’s model of water ice. However, further neighbour exchange and directiondependent dipolar interactions provide competing couplings of opposite sign. By measuring the DC bulk susceptibility to lower temperatures than previously reported and using carefully crafted defectfree spherical singlecrystal samples, which enables full control of demagnetising issues^{22}, we are able to identify the special temperatures in this wellstudied material. Dy_{2}Ti_{2}O_{7} lends itself naturally to this study as it stays close to the ideal paramagnetic limit (χT/C = 1) over a broad temperature range and remains a paramagnet well below its Curie–Weiss temperature on account of its high degree of frustration.
Thanks to the availability of a wellcharacterised Hamiltonian for Dy_{2}Ti_{2}O_{7} (refs.^{15, 28}), we are able to formulate a phenomenological model of the susceptibility which exposes the mechanism that induces the special temperatures and elevates the effects of minutefrustrated exchange interactions to surprisingly high temperature in this dipolarcoupled material. Furthermore, through an explicit numerical decomposition of the microscopic Hamiltonian, we demonstrate that these special temperatures, and eventual antiferromagnetic ordering, are caused by the weak quadrupolar corrections to the primary monopolar (dumbbell) Hamiltonian. Our study therefore establishes χT/C as a measure of weak interaction parameters, which are otherwise difficult to access experimentally. From another broad context, the lowtemperature susceptibility of spin ice is of particular interest in relation to ‘topological sector fluctuations’ of the harmonic component of the magnetisation^{29, 30}. The analogy with the nonideal gas allows an interpretation of the new experimental features in the magnetic susceptibility reported in this study, and have an appreciable impact on the interesting properties of spin ice—its residual entropy^{7}, magnetic monopoles^{26} and Coulomb phase^{27}—as discussed below.
Experimental determination of the magnetic susceptibility
A sphere of diameter 4 mm was commercially handcut from a larger crystal of Dy_{2}Ti_{2}O_{7} (see refs.^{30, 31}). Experimental conditions were carefully controlled to minimise measurement errors; see the Methods section. The experimental susceptibility of the sphere, χ_{exp}, was determined from measurements of the magnetic moment, with a subsequent demagnetising correction to obtain the shapeindependent intrinsic susceptibility, χ_{int},
using the exact result N = 1/3 for a sphere^{22, 32}.
From now on, we shall focus our discussion on the intrinsic susceptibility and suppress the 'int' subscript. The experimental measurement results are shown in Fig. 1. The Curie parameter is given by C = \(\frac{{N\mu _0\mu ^2}}{{3Vk_{\mathrm{B}}}}\) = 3.92 K for Dy_{2}Ti_{2}O_{7}, where N/V is the ion density, and μ is the magnetic moment^{30}. Our current measurements extend the earlier ones^{30} (where the lowest temperature was 2 K), down to 0.5 K. The extended temperature range reveals the important physical phenomena that are the focus of the present study namely, a peak in χT/C at T_{J} ≈ 2.2 K, and a 'transition' from χT/C > 1 to χT/C < 1 at T_{B} ≈ 0.57 K. These define the magnetic Joule and Boyle temperatures, respectively, as explained below. Alternatively, susceptibility measurements are often displayed as 1/χ versus T. In the inset of Fig. 1, we show C/χ versus T, and note that the gradient at T = T_{J} intersects the origin, hence demonstrating that the temperaturedependent Curie–Weiss temperature T_{CW}(T) equals zero at T = T_{J}. In the next section, we analyse the physical interpretation and consequences of these experimental results.
Analogy to classical gases
In this section, we explore the thermodynamic implications of a maximum in χT/C and propose a strong analogy to the theory of classical gases. That there is a peak in χT/C is not entirely surprising given that the sign of the effective nearestneighbour interaction in Dy_{2}Ti_{2}O_{7} is ferromagnetic, while the eventual expected ordering wave vector is most likely nonzero^{15, 24}. What is more surprising is the 'peak temperature' where χT/C reaches a maximum. It occurs at T ≈ 2 K, or about 20 times the expected ordering temperature in Dy_{2}Ti_{2}O_{7} (refs.^{15, 24}), and a factor 2 or so above the wellstudied peak temperature for the specific heat, which signals the rapid crossover from the paramagnetic regime to the spin ice state^{7, 33,34,35}. The main aim of this study is to understand the temperature scale and physical origin of the peak in χT/C shown in Fig. 1.
To start with, we consider the consequences of a peak in χT/C by introducing the thermodynamic potential
with a total differential
Cross differentiating with respect to M and 1/T, we obtain the relation
which implies
We therefore find that an extremum in χT/C implies (∂U/∂M)_{T} = 0. Similarly, it follows that the temperaturedependent Curie–Weiss temperature, T_{CW}(T) vanishes at the peak temperature, as shown in the inset of Fig. 1. That the internal energy, U, is independent of the magnetisation is a strong and intuitive definition of an effectively ideal noninteracting system. This is reminiscent of certain special conditions in gas thermodynamics, the best known defining the Boyle temperature, where the second virial coefficient vanishes and the ideal equation of state is obeyed. In fact, we see in Fig. 1 that there is a special temperature that corresponds to the Boyle temperature, namely the temperature at which χT/C equals unity, at T_{B} = 0.57 K.
In the Methods section, we show in detail how p/T for a gas or H/T for a magnet may be expressed as the sum of the familiar ideal equation of state (ideal gas law or Curie law, respectively) plus a nonideal term, that we label q. In both cases, the sign of the function q reflects the sign of the net interaction in the system. Thus, for a gas, q_{gas} is essentially the virial expansion: q_{gas} = \(\mathop {\sum}\nolimits_{i = 2}^\infty {\kern 1pt} B_i(T)(n{\mathrm{/}}V)^{i  1}\), which for many purposes may be truncated at the second term (i = 2). In that case, B_{2} is an integral over the pair potential u_{ ij }, where the integrand depends on the Mayer function \(\left( {{\mathrm e}^{  u_{ij}/kT}  1} \right)\). The sign of q ∝ B_{2} thus indicates the net interaction: positive for repulsive and negative for attractive. For the Van der Waals gas, B_{2} = b − a/RT and the net interaction switches sign precisely at the Boyle temperature T_{B} = a/(bR), reflecting the crossover from net repulsion at high temperature to net attraction at low temperature. The critical and Joule–Thomson temperatures are determined by the same energy scale with numerical prefactors 8/27 and 2, respectively. Similarly, for a magnet, q_{mag} = \(\mathop {\sum}\nolimits_{{\bf{r}} \ne 0} {\kern 1pt} {\mathrm{\Gamma }}_{\bf{r}}(T)\), where Γ(r) is the pair correlation function. It is therefore positive for net ferromagnetic correlations (analogous to repulsive interactions in the gas as they tend to make make M or V larger) and negative for net antiferromagnetic ones (analogous to attractive interactions in the gas as they tend to make M or V smaller). Therefore, in both a gas and a magnet, the Boyle temperature, T_{B}, marks the temperature at which competing interactions cancel each other to give an apparently ideal equation of state.
We explore further thermodynamic analogies in the Methods section while here we simply summarise the main results in Table 1. In order to work out and understand the microscopic and phenomenological origin of these results, we begin by discussing the microscopic models used to describe spin ice in the next section.
Spin ice models
The interactions in spin ice materials stem from the ions with magnetic moments μ_{ i } which reside on the corners of the pyrochlore lattice of cornersharing tetrahedra^{36}. As a result of the nature of the crystalfield doublet ground state^{37,38,39,40}, the magnetic moments are Isinglike^{40} and confined to point towards the centres of the adjacent tetrahedra. The primary magnetic interactions are the dipolar and shortrange exchange interaction, and the materials are modelled by the dipolar spin ice model (DSM)
where r_{ ij } is the distance between spin i and j, D the dipolar interaction and J_{1} the nearestneighbour exchange interaction. With no dipolar interaction (D = 0) and only nearestneighbour exchange, this model reduces to the nearestneighbour spinice model (NNSI), which describes spin ice quantitatively well down to about 0.6 K^{41}. The NNSI has a completely degenerate ground state and does not order. Together, dipolar and nearestneighbour exchange interaction lead to the standard dipolar spinice model (s–DSM)^{35}. The dipolar interaction weakly breaks the degeneracy of the NNSI and induces a transition to an ordered state at very low temperature. In addition to the nearestneighbour exchange interactions J_{1}, the generalised spin ice model (g–DSM) contains second and third nearestneighbour interactions J_{2}, J_{3a} and J_{3b}. A set of parameter values were previously determined (J_{1} = 3.41 K, J_{2} = −0.14 K, J_{3a} = J_{3b} = 0.025 K), which models a number of experiments at a quantitative level^{28}.
Another model of high conceptual and physical importance that elegantly captures salient features of spin ice systems is the dumbbell model^{26}, obtained by replacing the pointlike dipoles of the spin ice materials by dipoles of finite length. In this manner, the dipolar () Hamiltonian can be written as the sum of the monopolar (⃝) dumbbell model and quadrupolar () corrections^{26, 42},
Having introduced the models commonly used to describe spin ice, we are now in a position to model the experimental susceptibility of Dy_{2}Ti_{2}O_{7} reported in Fig. 1.
Phenomenological susceptibility model
In order to begin describing phenomenologically the experimental downturn in χT/C, we use the Husimi tree solution,
for the susceptibility of the NNSI^{29} as our starting point. The nearestneighbour interaction is here denoted by J_{eff}. We assume that there is an additive correction to the Helmholtzfree energy of the NNSI, \({\cal F}_0\), and that the correction is quadratic in the magnetisation M,
We take \({\cal F}_0\) to be the NNSIfree energy obtained on the pyrochlore cactus and θ is a coupling parameter. Differentiating twice with respect to M yields the sought correction to χ_{0}:
The resulting susceptibility, χ_{ θ }, is thus a product of χ_{0} and a Curie–Weiss like susceptibility (1 − θχ_{0}/C)^{−1}. This model contains two parameters (θ and J_{eff}). In Fig. 2, we show the best fit to experimental data (θ = −0.277 K, J_{eff} = 1.531 K), along with the two separate factors of the product. It is clear that χ_{ θ } models the experimental data well. Furthermore, the peak in χT/C arises from the product of the monotonically decreasing function χ_{0} and the monotonically increasing function, (1 − θχ_{0}/C)^{−1}. Note that χ_{0}T/C has a lowtemperature plateau close to 2 extending out to a temperature of about 1 K. It follows that an infinitesimally small, but finite negative θ induces a peak in χT/C at a temperature O(1) K. This is the 'mechanism' behind the elevation of T_{J} to a surprisingly high temperature by very weak interactions. This phenomenon is a main result of our study.
The physical origin of θ is perhaps most easily viewed as a meanfieldlike correction arising from beyond nearestneighbour interactions and the weak ordering tendencies of the dipolar interaction. It constitutes a meanfield correction to the (Husimi tree) meanfield construct, which apparently, works rather well. In the next section, we discuss the microscopic interpretation of the θcorrection further.
Finally, we would like to point out that this framework bears a close resemblance to a demagnetising correction, where χ_{0} is the external and χ_{ θ } the internal susceptibility. By differentiating Eq. (9) only once, we obtain a relation for the magnetic fields in the two models,
which has exactly the same form as the definition of the demagnetising field with the demagnetising factor equal to θ/C. The demagnetising transformation is a sensitive function of the demagnetising factor^{22}, and evidently a similar sensitivity arises in our phenomenological model.
Microscopic susceptibility model
In the present section, we wish to determine how well our measured susceptibility can be modelled at a microscopic level and to establish a connection between our phenomenological model of the previous section and the microscopic theory. We begin by modelling our experimental data using the g–DSM model^{28}.
As shown in Fig. 1, the g–DSM parameter set results in a susceptibility that overshoots our experimental result. We thus found it necessary to slightly adjust the third nearestneighbour parameter to J_{3a} = 0.030 K and J_{3b} = 0.031 K. As can be seen in Fig. 1, we obtain a close match to the experimental data with this new parameter set labelled g^{+}–DSM. We checked that such an adjustment of J_{3a} and J_{3b} leads to almost imperceptible changes in the neutron structure factor and the specific heat.
Having established a good microscopic model describing the experimental data, we would like to understand the connection between our phenomenological χ_{ θ } model and the microscopic g^{+}–DSM. In order to do this, we consider the monopolar–quadrupolar decomposition of the Hamiltonian, Eq. (7). The dumbbell model, \({\cal H}^\bigcirc\), like the NNSI, features perfectly degenerate spin ice states and a Curie crossover from χT/C = 1 at high temperature to χT/C = 2 at low temperature^{29}. The quadrupolar model, , is short ranged, with interactions decaying as 1/r^{5} (ref.^{26}). From Eq. (7), it follows that the energy of the ice states in the dipolar and quadrupolar model are equivalent, up to a constant shift and, therefore, we expect the shortrange quadrupolar model to capture the lowtemperature behaviour of spin ice. We have numerically decomposed the dipolar Hamiltonian and simulated these three models for finite systems. In Fig. 3, we show that the specific heat of the g^{+}–DSM model is indeed well described as the sum of a lowtemperature part from the quadrupolar model restricted to the ice states and a hightemperature part calculated from the dumbbell model.
The connection to our phenomenological model χ_{ θ } follows from the following observations: Since lim_{T→0} χ^{⃝} = lim_{T→∞} = 2, and the low (high) temperature behaviour of is well described by (χ^{⃝}), it follows that, to a good approximation,
We therefore note that the meanfieldlike correction (1 − θχ_{0}/C)^{−1} to the Husimi susceptibility is closely related to the susceptibility of the quadrupolar corrections to the dumbbell model, χ^{⃝}/2, restricted to the ice states. We show in Fig. 4 that this is indeed a good approximation.
Finally, we note that since the dumbbell model does not order at any finite temperature, it should correspond to our phenomenological model with θ = 0. However, a second nearestneighbour exchange interaction induces a finitetemperature transition, which is ferromagnetic for J_{2} < 0 and antiferromagnetic for J_{2} > 0. Numerical access to the dumbbell model therefore allows us to check how well our phenomenological model captures the transition from an antiferromagnetic to a ferromagnetic ground state as we tune an additional second nearestneighbour J_{2}. As can be seen in Fig. 5, the phenomenological model follows the tuned dumbbell model very closely right through the point of complete frustration (J_{2} = θ = 0). On the ferromagnetic side, the phenomenological parameter θ equals the critical temperature, T_{c} = θ. On the antiferromagentic side, one finds that \({\mathrm{lim}}_{T \to 0}{\kern 1pt} \chi _\theta\)\(\left. {T/C} \right_{T =  \theta } = 2{\mathrm{/}}3\), and in Fig. 5, we see that the ordering transition in the Monte Carlo simulations occurs when χ_{ θ }T/C ≈ 2/3. This relation can therefore be a useful experimental criterion as to when to expect an ordering transition. It also provides further evidence that the phenomenological model χ_{ θ } captures relevant physical aspects of frustrated ferromagnets.
To summarise, in this section we have thus shown that the experimental susceptibility is well matched by the g–DSM model with slightly adjusted third nearestneighbour parameters (g^{+}–DSM). This shows that χT/C provides access to interaction parameters that are otherwise hard to access. In addition, we have demonstrated that our phenomenological model is a good description of the microscopic dipolar model for a wide range of parameters, and that the phenomenological correction term describing the phase transition arises from the quadrupolar corrections to the dumbbell model.
Discussion
If spin ice were the only inverting ferromagnet, the notion of special temperatures could be just a curiosity of limited interest. However, we have identified a number of compounds which feature a peak in χT/C. Kapellasite, a proposed quantum spinliquid^{12}, is formed of kagome planes and features competing FM and AF interactions. There is a clear peak in χT/C which, as in the case of Dy_{2}Ti_{2}O_{7}, hints at an eventual AF ordering. χT/C for the quantum pyrochlore material Nd_{2}Zr_{2}O_{7} increases as T is lowered, but the peak is apparently preempted by a phase transition to an ordered allinallout state^{43,44,45}. Through Monte Carlo simulations, we have also verified that the wellstudied Ising and Heisenberg models of classical spins coupled through dipolar interactions on the cubic lattice features a peak in χT/C, as shown in Fig. 6. The spinel GeCo_{2}O_{4} also belongs to this class of magnets^{46}. Finally, χT/C peaks for a number of spinglass materials such as the organic κ(BEDTTTF)_{2}Hg(SCN)_{2}Br compound^{16} and Eu_{ x }Sr_{1−x}S_{ y }Se_{1−y} (ref.^{47}).
We have therefore demonstrated that there exists a class of inverting frustrated ferromagnets, which feature special temperatures at which the intrinsic competing FM and AF interactions balance and the magnets are quasiideal. At T_{B}, the magnetic Boyle temperature, the ideal equation of state is obeyed, and at T_{J}, the magnetic Joule temperature, the internal energy is independent of the magnetisation. Below T_{J}, the AF interactions start to dominate, and the corresponding peak in χT/C is an indication of eventual AF order, barring further disruptive lowtemperature terms in the Hamiltonian. Since the peak can occur at a high temperature relative to the eventual ordering temperature, it is a useful diagnostic feature in the quest for quantum and classical spin liquids. In a true spinliquid, the competing FM and AF interactions should be delicately balanced so that there is no finite Joule temperature, see Fig. 5. In our case study of Dy_{2}Ti_{2}O_{7}, the peak in χT/C is caused by weak (quadrupolar) perturbations to the primary (monopolar) Hamiltonian, and provides a way to experimentally probe these corrections.
In this context, we also note that a common way to characterise the level of frustration in magnetic systems with strongly competing interactions is the frustration index f ≡ T_{CW}/T_{c}, see ref.^{48}, where T_{CW} is the Curie–Weiss temperature and T_{c} the critical temperature. However, there are many systems for which this measure is not suitable, such as lowdimensional systems for which T_{c} = 0, or systems with either strongly anisotropic components of the g tensor or highly anisotropic exchange^{49}. The field of highly frustrated magnetism would thus benefit from other indicators of operating high frustration, which relies on ratios of temperature scales that are readily experimentally available, such as the Joule temperature for inverting magnets.
By introducing the concept of special temperatures in frustrated ferromagnets, we have filled a notable gap in the wellestablished thermodynamic analogy between magnets and classical gases. While our present investigation has focused on systems featuring a maximum in χT/C, we note that the converse phenomenon occurs in, for example, ferrimagnetic spin chains^{50}. In these systems, antiferromagnetic correlations at high temperature cross over to an eventual lowtemperature ferromagnet.The prevalence of such behaviour is a question we leave for future investigations.
Methods
Susceptibility measurement
The magnetic susceptibility was measured using a Quantum Design SQUID magnetometer and the crystals were positioned in a cylindrical plastic tube to ensure a uniform magnetic environment. Measurements were performed in the reciprocating sample option operating mode to achieve better sensitivity by eliminating lowfrequency noise. The position of the sample was carefully optimised to minimise misalignment with respect to the applied magnetic field. In particular, the sphere was measured at different positions and orientations in order to confirm the isotropic response and to fully reproduce the results of ref.^{30}.
Lowtemperature magnetic susceptibility was measured using a Quantum Design MPMS SQUID magnetometer equipped with an iQuantum ^{3}He insert^{51}. In analogy with ref.^{30}, different measurements were made: lowfield susceptibility (at μ_{0}H_{0} = 0.005, 0.01 and 0.02 T) and fieldcooled (FC) versus zerofieldcooled (ZFC) susceptibility. Also, magnetic field sweeps at fixed temperature were performed in order to evaluate the susceptibility accurately and confirm the linear approximation. The FC versus ZFC susceptibility measurements involved cooling the sample to base temperature 0.5 K in zero field, applying the weak magnetic field, measuring the susceptibility while warming up to 2 K, cooling to base temperature again and finally remeasuring the susceptibility while warming. Before switching the magnetic field off, field scans with small steps were performed in order to estimate the absolute susceptibilities.
To increase the statistics and control for dynamical effects, three measurements were taken at each temperature before warming to the next step point. Furthermore, to test the accuracy of the measurement, some data were acquired with an increased number of raw data points—typically 64 points rather than the usual 24. In fact, at low temperature, the magnetic moment of the sample is close to the saturation value of the instrument, especially at μ_{0}H_{0} = 0.02 T. When measuring under such conditions, it is necessary to increase the number of raw data points to 64 to maintain consistency between measurements.
Data have been compared with the hightemperature measurements described in ref.^{30}, in particular in the overlapping region 1.8 ≤ T ≤ 2 K. Without further manipulation, the two sets of data are in very good agreement with variations of the order of <0.5%. This can be attributed to the uncertainty in the actual field value in each of the two instruments, mainly due to the presence of small frozen fields in the superconducting coils. In Fig. 1, the two sets of measurements were accurately superimposed, by compensating (<0.5%) the actual applied field value of the lowtemperature measurement.
Magnetic thermodynamics
To make an analogy between magnetic and fluid thermodynamics, we define X and x as the extensive and intensive mechanical variable, respectively. For a fluid, X = V (volume) and x = p (pressure) and we assume the mole number n is fixed. For a magnet, we assume an ellipsoidal sample and deal with intrinsic properties (postdemagnetising correction). We have x = H, the internal Hfield and X = −μ_{0}VM, where M is the magnetisation. Here, the minus sign is included to complete the analogy, but it makes no difference to the following results.
A strong and intuitive definition of an ideal noninteracting system is that the internal energy depends only on temperature: (∂U/∂X)_{ T } = 0. This implies x − T(∂x/∂T)_{ X } = 0 as an equivalent definition of ideality. Integration of the latter then shows that the noninteracting equation of state is of the form:
where ϕ is some function. Indeed, this is true for both the ideal gas and the ideal paramagnet, where the functions in question are ϕ_{mag} = M/C, where C is the Curie constant and ϕ_{gas} = nR/V, where R is the gas constant. For a real gas or paramagnet, we write the equation of state as:
where the function q is the nonideal correction. We now define the following special temperatures: these may not be unique, but we will refer to them in the singular for clarity.
The Boyle temperature, T_{B}, is defined as the temperature where q = 0, so the ideal equation of state happens to be obeyed.
The Joule temperature, T_{J}, is the temperature where (∂U/∂X)_{ T } = 0 ⇒ x − T(∂x/∂T)_{ X } = 0, which indicates that the intensive variable p or H is tangentially proportional to absolute temperature T. This temperature is infinite for a Van der Waals gas, but may be finite for some real gases (e.g., helium) and some magnets.
The Joule–Thomson temperature, T_{JT}, is defined as the temperature where: (∂x/∂T)_{ E } = 0 ⇒ X − T(∂X/∂T)_{ x } = 0. where E = U + xX is the enthalpy. For a typical gas, T_{JT} is finite at any density, and in this sense, a real gas never reaches the ideal gas limit. For the magnetic models considered here, T_{JT} is infinite.
We can see that T_{B} and T_{J} indicate quasiideality, where some criteria of ideality are satisfied. The third special temperature, T_{JT}, indicates that X ∝ T tangentially. This corresponds to quasiideality only in the particular case of a gas (V ∝ T) and not in the case of a magnet (M ∝ T). Nevertheless, there is a symmetry between T_{J} and T_{JT}: both are defined by setting to zero a Legendre transform \({\cal L}[z] = z  T(\partial z{\mathrm{/}}\partial T)\) of the intensive variable z → x and the extensive variable z → X, respectively. Hence, both imply tangential linearity of the corresponding variable with absolute temperature.
Starting with these Legendre transforms, we can translate the three special temperatures into conditions on the function Tϕ(X)/x. We are interested in the magnet in the linear regime at low field and magnetisation, where we can define the susceptibility χ = M/H, which is a function of T only: χ = χ(T). Hence, for a magnet, we obtain conditions on χT/C which, in this context, is analogous to pV/nRT. The Boyle temperatures, T_{B}, are located by χT/C = 1 and pV/nRT = 1. The Joule temperature, T_{J}, for a magnet corresponds to an extremum in χT/C as a function of temperature. The Joule–Thomson temperature, T_{JT}, for a gas at fixed pressure corresponds to an extremum in pV/nRT. These relations are summarised in Table 1.
Applicability of the phenomenological model
In order to establish the applicability of the phenomenological χ_{ θ } model, Eq. (10), we compare it here to the standard dipolar spin ice model (s–DSM), Eq. (6), which includes the dipolar interaction D in addition to a nearestneighbour exchange interaction, J_{1}. In this model, spin ice behaviour persists up to J_{1}/D < 6.01, and the dipolar interaction induces a lowtemperature phase transition to a 'singlechain' state^{15, 24}. The model features a corresponding Joule temperature, and as can be seen in Fig. 7, our phenomenological model describes the susceptibility of the s–DSM remarkably well down to, and including, the critical temperature T_{c}, at which \(\chi _\theta \left. {T{\mathrm{/}}C} \right_{T =  \theta } \approx 2{\mathrm{/}}3\).
Determination of model parameters
The parameters for the g^{+}–DSM were chosen in the following manner: J_{1} = 3.41 K and J_{2} = −0.14 K were set to the previoulsy determined values of the g–DSM^{28}. Then, a χT/C RMS chart was calculated for the deviation between our experimental data and Monte Carlo calculations as a function of J_{3a} and J_{3b} (see Fig. 8). From the chart, we determined the point closest to the g–DSM values (J_{3a} = J_{3b} = 0.025 K) located in the minimum RMS valley. This point is indicated by a yellow ring in Fig. 8, corresponding to the g^{+}–DSM values (J_{3a} = 0.030 K, J_{3b} = 0.031 K).
Code availability
The custom computer codes used in this study are available from the corresponding author on reasonable request.
Data availability
The data sets generated and analysed in this study are available from the corresponding author on reasonable request.
References
 1.
Wannier, G. H. Antiferromagnetism. The triangular Ising net. Phys. Rev. 79, 357–364 (1950).
 2.
Anderson, P. W. Resonating valence bonds: a new kind of insulator? Mater. Res. Bull. 8, 153–160 (1973).
 3.
Marston, J. B. & Zeng, C. SpinPeierls and spinliquid phases of kagome quantum antiferromagnets. J. Appl. Phys. 69, 5962–5964 (1991).
 4.
Anderson, P. W. Ordering and antiferromagnetism in ferrites. Phys. Rev. 102, 1008–1013 (1956).
 5.
Villain, J. Insulating spin glasses. Z. für Phys. B Condens. Matter 33, 31–42 (1979).
 6.
Bramwell, S. T. & Gingras, M. J. P. Spin ice state in frustrated magnetic pyrochlore materials. Science 294, 1495–1501 (2001).
 7.
Ramirez, A. P., Hayashi, A., Cava, R. J., Siddharthan, R. B. & Shastry, S. Zeropoint entropy in spin ice. Nature 399, 333–335 (1999).
 8.
Han, T. et al. Fractionalized excitations in the spinliquid state of a kagomelattice antiferromagnet. Nature 492, 406–410 (2012).
 9.
Fu, M., Imai, T., Han, T. & Lee, Y. S. Evidence for a gapped spinliquid ground state in a kagome Heisenberg antiferromagnet. Science 350, 655–658 (2015).
 10.
White, R. M. Quantum Theory of Magnetism: Magnetic Properties of Materials, Springer Series in SolidState Sciences (Springer, Berlin Heidelberg, 2007).
 11.
Balz, C. et al. Physical realization of a quantum spin liquid based on a complex frustration mechanism. Nat. Phys. 12, 942–949 (2016).
 12.
Fåk, B. et al. Kapellasite: a kagome quantum spin liquid with competing interactions. Phys. Rev. Lett. 109, 037208 (2012).
 13.
Petit, S. et al. Observation of magnetic fragmentation in spin ice. Nat. Phys. 12, 746–750 (2016).
 14.
McClarty, P. A. et al. Chainbased order and quantum spin liquids in dipolar spin ice. Phys. Rev. B 92, 094418 (2015).
 15.
Henelius, P. et al. Refrustration and competing orders in the prototypical Dy_{2}Ti_{2}O_{7} spin ice material. Phys. Rev. B 93, 024402 (2016).
 16.
Hemmida, M. et al. Weak ferromagnetism and spin glass in kappa(BEDTTTF)_{2}Hg(SCN)_{2}Br. Preprint at https://arxiv.org/abs/1710.04028v2 (2017).
 17.
Callen, H. B. Thermodynamics and an Introduction to Thermostatics (Wiley India Pvt. Limited, India, 2006).
 18.
Wannier, G. H. Statistical Physics (Dover Books on Physics) (Dover Publications, USA, 2012).
 19.
Fisher, M. E. Relation between the specific heat and susceptibility of an antiferromagnet. Philos. Mag. 7, 1731–1743 (1962).
 20.
De Jongh, L. J. & Miedema, A. R. Experiments on simple magnetic model systems. Adv. Phys. 79, 2554–2557 (1999).
 21.
Henelius, P. & Sandvik, A. W. Sign problem in Monte Carlo simulations of frustrated quantum spin systems. Phys. Rev. B 62, 1102–1113 (2000).
 22.
Twengström, M., Bovo, L., Gingras, M. J. P., Bramwell, S. T. & Henelius, P. Microscopic aspects of magnetic lattice demagnetizing factors. Phys. Rev. Mater. 1, 044406 (2017).
 23.
Harris, M. J., Bramwell, S. T., McMorrow, D. F., Zeiske, T. & Godfrey, K. W. Geometrical frustration in the ferromagnetic pyrochlore Ho_{2}Ti_{2}O_{7}. Phys. Rev. Lett. 64, 43–57 (2010).
 24.
Melko, R. G., den Hertog, B. C. & Gingras, M. J. P. Longrange order at low temperatures in dipolar spin ice. Phys. Rev. Lett. 87, 067203 (2001).
 25.
Ryzhkin, I. A. Magnetic relaxation in rareearth pyrochlores. J. Exp. Theor. Phys. 101, 481–486 (2005).
 26.
Castelnovo, C., Moessner, R. & Sondhi, S. L. Magnetic monopoles in spin ice. Nature 451, 42–45 (2008).
 27.
Fennell, T. et al. Magnetic Coulomb phase in the spin ice Ho_{2}Ti_{2}O_{7}. Science 326, 415–417 (2009).
 28.
Yavors’kii, T., Fennell, T., Gingras, M. J. P. & Bramwell, S. T. Dy_{2}Ti_{2}O_{7} spin ice: a test case for emergent clusters in a frustrated magnet. Phys. Rev. Lett. 101, 037204 (2008).
 29.
Jaubert, L. D. C. et al. Topologicalsector fluctuations and Curielaw crossover in spin ice. Phys. Rev. X 3, 011014 (2013).
 30.
Bovo, L., Jaubert, L. D. C., Holdsworth, P. C. W. & Bramwell, S. T. Crystal shapedependent magnetic susceptibility and Curie law crossover in the spin ices Dy_{2}Ti_{2}O_{7} and Ho_{2}Ti_{2}O_{7}. J. Phys. Condens. Matter 25, 386002 (2013).
 31.
Prabhakaran, D. & Boothroyd, A. T. Crystal growth of spinice pyrochlores by the floatingzone method. Crys. Growth 318, 1053–1056 (2011).
 32.
Osborn, J. A. Demagnetizing factors of the general ellipsoid. Phys. Rev. 67, 351–357 (1945).
 33.
Bramwell, S. T. et al. Spin correlations in Ho_{2}Ti_{2}O_{7}: a dipolar spin ice system. Phys. Rev. Lett. 87, 047205 (2001).
 34.
Higashinaka, R., Fukazawa, H. & Maeno, Y. Specific heat of single crystal of spin ice compound Dy_{2}Ti_{2}O_{7}. Phys. B Condens. Matter 329–333, 1040–1041 (2003).
 35.
den Hertog, B. C. & Gingras, M. J. P. Dipolar interactions and origin of spin ice in Ising pyrochlore magnets. Phys. Rev. Lett. 84, 3430–3433 (2000).
 36.
Subramanian, M. A., Aravamudan, G. & Subba Rao, G. V. Oxide pyrochlores—a review. Progress Solid State Chem. 15, 55–143 (1983).
 37.
Rosenkranz, S. et al. Crystalfield interaction in the pyrochlore magnet Ho_{2}Ti_{2}O_{7}. J. Appl. Phys. 87, 5914–5916 (2000).
 38.
Bertin, A., Chapuis, Y., Dalmas de Réotier, P. & Yaouanc, A. Crystal electric field in the R_{2}Ti_{2}O_{7} pyrochlore compounds. J. Phys. Condens. Matter 24, 256003 (2012).
 39.
Jana, Y. M., Sengupta, A. & Ghosh, D. Estimation of single ion anisotropy in pyrochlore Dy_{2}Ti_{2}O_{7}, a geometrically frustrated system, using crystal field theory. J. Magn. Magn. Mater. 248, 7–18 (2002).
 40.
Rau, J. G. & Gingras, M. J. P. Magnitude of quantum effects in classical spin ices. Phys. Rev. B 92, 144417 (2015).
 41.
Melko, R. G. & Gingras, M. J. P. Monte Carlo studies of the dipolar spin ice model. J. Phys. Condens. Matter 16, R1277–R1319 (2004).
 42.
Byström, D. Ground State Order In Spin Ice Without Longrange Interactions. MSc thesis, Royal Institute of Technology (2013).
 43.
Lhotel, E. et al. Fluctuations and allinallout ordering in dipoleoctupole Nd_{2}Zr_{2}O_{7}. Phys. Rev. Lett. 115, 197202 (2015).
 44.
Benton, O. Quantum origins of moment fragmentation in Nd_{2}Zr_{2}O_{7}. Phys. Rev. B 94, 104430 (2016).
 45.
Xu, J. et al. Magnetic structure and crystalfield states of the pyrochlore antiferromagnet Nd2_{Z}r_{2}O_{7}. Phys. Rev. B 92, 224430 (2015).
 46.
Hubsch, J. & Gavoille, G. First order magnetic phase transition in GeCo_{2}O_{4}. J. Magn. Magn. Mater. 66, 17–22 (1987).
 47.
Westerholt, K. & Bach, H. EuxSr_{1−x}SySe_{1−y}: a model system for studying competing magnetic interactions. Phys. Rev. Lett. 47, 1925–1927 (1981).
 48.
Ramirez, A. P. Strongly geometrically frustrated magnets. Annu. Rev. Mater. Sci. 24, 453–480 (1994).
 49.
Hayre, N. R. et al. Thermodynamic properties of Yb_{2}Ti_{2}O_{7} pyrochlore as a function of temperature and magnetic field: Validation of a quantum spin ice exchange Hamiltonian. Phys. Rev. B 87, 184423 (2013).
 50.
Shiomi, D., Sato, K. & Takui, T. Spin−spin correlation function and magnetic susceptibility of quantum ferrimagnetic spin chains as models for organic moleculebased ferrimagnetics. J. Phys. Chem. B 104, 19611965 (2000).
 51.
Shirakawa, N., Horinouchi, H. & Yoshida, Y. Measuring Sr_{2}RuO_{4} down to 0.5 K with a commercial SQUID magnetometer combined with 3He refrigeration. J. Magn. Magn. Mater. 272276, e149–e150 (2004).
Acknowledgements
We thank D. Prabhakaran for providing crystals from which the samples were cut. The idea to numerically determine the quadrupole correction to the dumbbell model is due to S. Powell^{42}. We are grateful to Addison Richards for providing Monte Carlo simulation data for the dipolar Heisenberg model on a cubic lattice. S.T.B. thanks J. Xu and B. Lake for a useful correspondence concerning Nd_{2}Zr_{2}O_{7}. We thank Wen Jin for pointing out ref.^{50} to us. The simulations were performed on resources provided by the Swedish National Infrastructure for Computing (SNIC) at the Centre for High Performance Computing (PDC) at the Royal Institute of Technology (KTH). M.T. and P.H. are supported by the Swedish Research Council (201303968), M.T. is grateful for funding from Stiftelsen Olle Engkvist Byggmästare (2014/807), and L.B. is supported by The Leverhulme Trust through the Early Career Fellowship programme (ECF2014284). The work at the University of Waterloo was supported by the Canada Research Chair programme (M.J.P.G., Tier 1). This research was supported in part by the Perimeter Institute for Theoretical Physics. Research at the Perimeter Institute is supported by the Government of Canada through Innovation, Science, and Economic Development Canada and by the Province of Ontario through the Ministry of Research, Innovation, and Science.
Author information
Affiliations
Contributions
L.B. performed all hightemperature measurements and data collection. L.B. and O.A.P. performed the lowtemperature SQUID measurements. M.T. and P.H. performed all simulations and the dumbbell–quadrupole decomposition. S.T.B. conceived the phenomenological model and the concept of special temperatures. M.J.P.G. realised the broad applicability of the concept of inverting magnets. T.F. contributed to data analysis. L.B., M.T., M.J.P.G., S.T.B. and P.H. wrote the manuscript with input from all authors.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Electronic supplementary material
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Bovo, L., Twengström, M., Petrenko, O.A. et al. Special temperatures in frustrated ferromagnets. Nat Commun 9, 1999 (2018). https://doi.org/10.1038/s41467018042973
Received:
Accepted:
Published:
Further reading

Characteristic temperatures of a triplon system of dimerized quantum magnets
International Journal of Modern Physics B (2021)

LiHoF4 : Cuboidal demagnetizing factor in an Ising ferromagnet
Physical Review B (2020)

Effects of uniaxial pressure on the spin ice Ho2Ti2O7
Physical Review B (2020)

JouleThomson temperature of a triplon system of dimerized quantum magnets
Physics Letters A (2020)

Investigation of thermal behavior of mixedvalent iron borates vonsenite and hulsite containing [OM 4] n + and [OM 5] n + oxocentred polyhedra by in situ hightemperature Mössbauer spectroscopy, Xray diffraction and thermal analysis
Acta Crystallographica Section B Structural Science, Crystal Engineering and Materials (2020)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.