Signatures of spinon dynamics and phase structure of
dipolar-octupolar quantum spin ices in two-dimensional coherent spectroscopy

Abstract

We study how sharp signatures of fractionalization emerge in nonlinear spectroscopy experiments on spin liquids with separated energy scales. Our model is that of dipolar-octupolar rare earth pyrochlore materials, prime candidates for realising quantum spin ice. This family of three dimensional quantum spin liquids exhibits fractionalization of spin degrees of freedom into spinons charged under an emergent $U(1)$ gauge field. We show that the technique of two dimensional coherent spectroscopy (2DCS) can identify clear signatures of fractionalised spinon dynamics in dipolar-octupolar quantum spin ices. However, at intermediate temperatures, spinon dynamics are heavily constrained in the presence of an incoherent spin background, leading to a broad 2DCS response. At lower temperatures, a sharp signal emerges as the system enters a coherent spin liquid state. This lower temperature signal can in turn distinguish between zero-flux and $\pi$-flux forms of quantum spin ice.

The emergence of fractional excitations Rajaraman (2001); Moessner and Moore (2021) is a feature of many exotic quantum states, including quantum spin liquids Anderson (1973); Moessner and Sondhi (2001); Balents (2010); Knolle and Moessner (2019) and topologically ordered phases of matter Laughlin (1983); Wen (2002). Establishing clear signatures of such excitations is then key to the unambiguous identification of these phases in real materials. Fractional excitations can only be created in multiplets, and never singly, meaning that, from kinematic considerations, linear response functions are broad. These broad responses are difficult to unambiguously distinguish from the response of conventional excitations broadened by (e.g.) disorder or finite lifetime effects.

The technique of two-dimensional coherent spectroscopy (2DCS) Wan and Armitage (2019); Kuehn et al. (2011); Woerner et al. (2013) has been proposed as a method that avoids this issue. 2DCS is a non-linear optical technique which produces responses that are functions of two frequencies $\omega_{t}$ and $\omega_{\tau}$. The continuum response from the excitation of zero-momentum pairs of fractionalised excitations becomes a sharp extended signal along the $\omega_{t}=-\omega_{\tau}$ line in the two-dimensional frequency plane, termed the ‘spinon-echo’ or ‘rephasing’ signal. The effects of disorder, or the finite lifetime of quasiparticles, are responsible for any broadening of the signal transverse to this line, allowing the existence of a continuum from fractionalized excitations to be distinguished from these other effects.

The technique has been investigated theoretically in a number of settings Wan and Armitage (2019); Choi et al. (2020); Hart and Nandkishore (2023); Qiang et al. , though primarily in exactly solvable models such as the Kitaev honeycomb spin liquid, or certain one dimensional models. Here, we discuss the application of 2DCS to quantum spin ice (QSI) - a class of three dimensional spin liquids on the pyrochlore lattice Hermele et al. (2004); Shannon et al. (2012); Benton et al. (2012); Savary and Balents (2012); Gingras and McClarty (2014). In recent years several materials have been suggested as promising candidates for systems with QSI ground states Gingras and McClarty (2014); Kimura et al. (2013); Sibille et al. (2018); Gaudet et al. (2019); Sibille et al. (2020); Porée et al. (2022, 2024), although definitively establishing the presence of such a spin liquid in any specific material remains challenging.

The most promising QSI candidate materials have been the rare earth pyrochlores, in particular the Ce based materials Ce₂Zr₂O₇, Ce₂Sn₂O₇ and Ce₂Hf₂O₇ Gaudet et al. (2019); Gao et al. (2019); Smith et al. (2023); Sibille et al. (2020); Porée et al. (2022, 2024); Yahne et al. (2024). These are all examples of so-called dipolar-octupolar pyrochlores due to the particular symmetry properties of their pseudospin-1/2 magnetic degrees of freedom.

In this Letter we discuss the capacity of 2DCS to identify fractionalised excitations within QSI phases. We present a theory for the 2DCS of dipolar-octupolar pyrochlores in two settings: (i) an intermediate temperature regime where spinon excitations are quantum coherent, but the initial state of the system can be taken to be an incoherent ensemble of classical spin ice ground states, which is treated using degenerate perturbation theory and exact diagonalization; (ii) the low energy regime where quantum coherence between spin ice states is important, which is treated using a gauge mean field theory (GMFT) Savary and Balents (2012); Lee et al. (2012); Desrochers et al. (2023); Desrochers and Kim (2024a).

We demonstrate that the 2DCS response produces a (perhaps surprising) broad response in the intermediate temperature regime [Fig. 1], as the incoherent spin background constrains the motion of spinon excitations such that they no longer posses a well defined real-space momentum. Instead, their dynamics can be approximated by motion on a Husimi cactus graph [Fig. 2], with respect to which spinons have a well defined pseudo-momentum. Due to geometrical coupling factors between the field and local magnetic moments, the probe field does not conserve this pseudo-momentum, resulting in a broad signal in the 2DCS response.

By contrast, in the low temperature regime, the ground state is expected to be a coherent superposition of classical ice states, removing the constraints on spinon motion, and allowing spinons to acquire a definite real-space momentum. Sharp features are thus obtained in the 2DCS response, distinguishing this regime from the higher temperature regime [Fig. 3]. Further, we show that the 2DCS response can clearly discriminate between the two possible quantum spin ice phases predicted for dipolar-ocutpolar pyrochlores in GMFT, the $0$-flux and $\pi$-flux $U(1)$ QSLs. In the $\pi$-flux phase, an increased unit cell size results in multiple spinon bands Desrochers and Kim (2024a), and transitions between these bands produce streaks away from the main rephasing diagonal in the 2DCS response.

Following a brief overview of dipolar-octupolar pyrochlores, we will discuss these items in turn. The most general nearest neighbour Hamiltonian for dipolar-octupolar pyrochlores is Rau and Gingras (2019); Desrochers and Kim (2024a):

$$H=\sum_{\langle i,j\rangle}J_{yy}S^{y}_{i}S^{y}_{j}-J_{\pm}(S^{+}_{i}S^{-}_{j}+S^{-}_{i}S^{+}_{j})+J_{\pm\pm}(S^{+}_{i}S^{+}_{j}+S^{-}_{i}S^{-}_{j}),\$$ (1)

with $\langle i,j\rangle$ denoting nearest neighbour sites on the pyrochlore lattice, and the ladder operators being defined relative to the $y$ direction of pseudospin space. Motivated by the estimated parameter ranges for Ce pyrochlores Smith et al. (2022); Sibille et al. (2020); Porée et al. (2024), we assume that the dominant coupling is $J_{yy}>0$. The $S^{\mu}_{j}$ operators are pseudo-spin operators acting in the lowest energy Kramers doublet of the magnetic ions. Under the action of the point group of the lattice and time reversal, the $S^{z}$ and $S^{x}$ pseudo-spin operators transform as magnetic dipoles, whilst $S^{y}$ transforms as a magnetic octupole Rau and Gingras (2019); Huang et al. (2014). For the basis chosen for Eq. (1), the dipolar coupling between the magnetic degrees of freedom and an external probe field $\mathbf{H}$ is given by: $\mu_{B}g_{z}\ \mathbf{H}\cdot\mathbf{z}_{i}\ (\cos\theta\ S^{z}_{i}+\sin\theta\ S^{x}_{i})$. Here $\theta$ is a material dependent parameter, and the $\mathbf{z}_{i}$ are unit vectors directed along the lines connecting the centres of neighbouring tetrahedra on the pyrochlore lattice. These tetrahedra sit on two fcc sublattices, which we define as the ‘A’ and ‘B’ sublattices. This geometrical coupling factor between the external field and local degrees of freedom endows the probe field operator with some site dependence, which we find has a significant effect on the linear and non-linear responses of the system.

The rephasing signal that we calculate is only part of the full 2DCS response Wan and Armitage (2019), and is obtained from the imaginary part of the Fourier transform of the non-linear susceptibility $\chi^{(3)}(t,t,t+\tau)$:

$$\frac{i^{3}\theta(t)\theta(\tau)}{N}\langle[[[M(t+\tau),M(\tau)],M(\tau)],M(0)]\rangle,$$ (2)

with $M(t)$ the time-dependent operator coupling the probe field and the pseudo-spin degrees of freedom.

Figure 1: 2DCS response of dipolar-octupolar QSI in the intermediate temperature regime with ((a)) and without ((b)) the geometrical factors $\mathbf{H}\cdot\mathbf{z}_{i}$ included in the field probe operator, as calculated using exact diagonalization on a finite cluster of 32 sites. A significant broad signal is observed if the correct geometrical coupling factors are included, due to the failure of the probe field operator to conserve the spinons’ pseudo-momentum. By contrast, a much sharper rephasing streak is observed if the probe is modified to conserve pseudo-momentum by removing site dependent factors. The spectral weight of the response is shifted to the upper end of the spinon band, which can be understood in terms of an effective pseudo-momentum vertex factor (supplementary material sup ) In both cases $J_{\pm}=-0.1,J_{yy}=2$, and the Lorentzian linewidth $\Gamma=0.001$

Figure 2: (a) Section of the pyrochlore lattice with a single spinon indicated with a full orange circle. Each tetrahedron has 12 nearest neighbour tetrahedra on the same fcc sublattice, however the constrained hop** introduced in first order degenerate perturbation theory only allows the spinon to hop to the six sites indicated with empty blue circles. (b): The Husimi cactus approximation for the many-body state space of a single spinon. Each state is connected to six other states, and the system can return to the same state via cycles of 3.

Intermediate temperature regime- In the limit of large $J_{yy}$, the Hamiltonian of Eq. (1) can be treated in degenerate perturbation theory. The low energy states then obey a ‘2-in-2-out’ ice rule on $S^{y}_{i}$ – making the system an octupolar spin ice. Tunnelling between such configurations happens perturbatively and is of order $\sim|J_{\pm}|^{3}/J_{yy}^{2}$, while local violations of the ice rule (spinons) hop with matrix element $\sim J_{\pm}$.

At intermediate temperatures such that $|J_{\pm}|^{3}/J_{yy}^{2},|J_{\pm\pm}|^{3}/J_{yy}^{2}\ll T\ll|J_{\pm}|$, only the first order spinon hop** will be relevant, and the equilibrium state will be an incoherent ensemble of classical spin ice states. A schematic of the spinon hop** generated by the $J_{\pm}$ effective Hamiltonian is presented in Fig. 2. Note that this term in the Hamiltonian is symmetric under rotations in the $S^{x}$, $S^{z}$ plane, and so we can freely set $\theta$ to zero.

To obtain the 2DCS response of the system in this regime, the $J_{\pm}$ perturbation (the second term in Eq. (1)) is exactly diagonalized in the two and four spinon sectors on a 32-site cluster of the pyrochlore lattice. The calculated 2DCS response in this regime is presented in the first panel of Fig. 1. Whilst a streak resembling the expected response for a system with fractionalised excitations is observed, it is accompanied by a broad region of non-zero intensity, which is unexpected for a system possessing coherent fractionalised excitations.

We can understand the broad response obtained by making use of an approximate analytic treatment of the spinon dynamics, wherein spinon dynamics are modelled using a Bethe lattice-like graph.

To motivate this, we first note that the hop** generated by the perturbation Hamiltonian does not allow spinons to freely move across the lattice. A spinon on the ‘A’ fcc sublattice of tetrahedra will always remain on the ‘A’ sublattice. Moreover, a spinon must hop by flip** two neighbouring spins of opposite sign or of opposite colour in Fig. 2 (into or out of an ‘A’ tetrahedron), and must not generate any further violations of the ice rule as they move. This allows spinons to access only 6 of the 12 nearest neighbour fcc sites in a single hop, with these 6 sites dependent on the local spin configuration.

There are interaction effects between spinons: spinons flip spins as they move, which then influences the paths other spinons can take across the pyrochlore lattice; neighbouring spinons that share a majority spin on their respective tetrahedra impose further restrictions on the number of states the other can hop to Udagawa and Moessner (2019).

One of the most important aspects of the dynamics is that if a spinon traverses a closed loop on the fcc lattice, it may leave a loop of flipped spins in its wake, meaning that the system has not returned to the same many-body state, and thus this process does not correspond to a closed loop on a graph of the many-body states of the system.

We consider such a graph in more detail for the hop** of one isolated spinon. If each state is a vertex on this graph, it is then connected to 6 other vertices. Spinons can hop in 3-cycles on the fcc lattice in such a way that the background spin configuration is unchanged, meaning that these are also loops in the many-body graph. The graph is then approximately a tree of triangles, with three triangles attached to each vertex. The many-body state graph will have other larger loops (of six hops or larger), corresponding to the double traversal of loops on the fcc lattice with spins connected ‘head to tail’. As the presence of these larger loops depends on the local spin configuration, they do not appear uniformly across the graph, and so the tree of triangles, or cactus, is a reasonable approximation to the state graph for single spinon hop**.

The resulting Husimi cactus graph is presented in the second panel of Fig. 2, as described in Ref. Udagawa and Moessner (2019). Within this approximation, spinon dynamics undergo an effective dimension reduction, and obtain a one-dimensional dispersion indexed by a pseudo-momentum $p$. This model for the spinon dynamics has been shown to successfully reproduce the linear responses of QSI within this temperature regime Udagawa and Moessner (2019) when compared with exact diagonalization, a posteriori justifying its use.

From the Husimi cactus viewpoint, we can explain the broad features observed in the 2DCS response of Fig. 1. The probe field operator $M(t)$ is proportional to $\sum_{i}\mathbf{H}\cdot\mathbf{z}_{i}S^{z}_{i}$ for dipolar-octupolar pyrochlores. The spatial dependence of this coupling prevents $M(t)$ from conserving spinon pseudo-momentum on the Husimi cactus. Without the conservation of pseudo-momentum, the ‘rephasing’ that occurs in Eq. (2) to ensure a sharp streak response along $\omega_{t}=-\omega_{\tau}$ fails, and a broad response results. Further details concerning the representation of the probe field operator $M(t)$ in the Husimi cactus picture, and a demonstration that it does not conserve pseudo-momentum is presented in supplementary material sup .

That the broad response is a direct consequence of the pseudo-momentum non-conservation is verified by modifying the probe field operator such that it has a definite pseudo-momentum. This is achieved by replacing the $\mathbf{H}\cdot\mathbf{z}_{i}$ coupling factors with a uniform constant H to remove any site dependence from $M(t)$. Using this operator, we find that the computed 2DCS response has a sharp rephasing streak, as shown in the second panel of Fig. 1, supporting the conclusion that the broad features in the 2DCS response in the intermediate temperature regime as calculated for the realistic probe field operator are indicative of the constrained spinon dynamics.

Figure 3: 2DCS rephasing signals in the low temperature regime from gauge mean field theory. (a) response in the $0$-flux phase (with $J_{\pm}/J_{yy}=0.04$); (b) $\pi$-flux response (for $J_{\pm}/J_{yy}=-0.04)$. In the $0$-flux phase, a single extended streak is observed, whilst in the $\pi$-flux phase, the much narrower bandwidth of the two spinon bands, and transitions between them, result in multiple rephasing peaks, and additional peaks away from the main diagonal. Note that, if $\epsilon_{\pm}$ are the mean energies of each spinon band, the peaks in the $\pi$-flux response occur at frequencies $\omega_{i}=\epsilon_{-},(\epsilon{-}+\epsilon_{+})/2$. Matrix elements for processes where $\omega_{i}=\epsilon_{+}$ are heavily suppressed. In both cases the Lorenztian linewidth $\Gamma=0.01$

Figure 4: Linear optical response functions for the intermediate temperature regime ((a)), $0$-flux ground state ((b)), and $\pi$-flux ground state ((c)). For the intermediate temperature regime and $0$-flux response functions, $J_{\pm}/J_{yy}=0.04$, and for the $\pi$-flux $J_{\pm}/J_{yy}=-0.04$. The two ground state phases remain distinguishable in the linear optical response, with the $\pi$-flux response possessing multiple clear maxima. However, the qualitatively different behaviour of the 2DCS signal in the intermediate temperature regime to the low energy regime is not evident at linear response, and there is little qualitative difference between the $0$-flux response and the response from spinons moving on an incoherent spin background, despite the stark differences in the underlying dynamics. For panel (a), the Lorentzian linewidth $\Gamma=0.002$, and in panels (b) and (c) $\Gamma=0.01$.

Low temperature regime- The 2DCS response has significant broad features when quantum coherence between classical spin-ice states is suppressed by thermal fluctuations. If instead we work in a $T=0$ limit, this anomalous broadening is removed, and characteristic sharp rephasing signals are recovered. In this limit, both the gauge field and spinon degrees of freedom are quantum coherent. We analyse this regime of the problem by appealing to gauge mean field theory, making use of the formulation described in Ref. Desrochers and Kim (2024a). By the nature of mean field theories, this calculation is less controlled that the exact diagonalization analysis used in the intermediate temperature regime, however we expect it to still provide a useful qualitative description of the physics close to $T=0$.

In this GMFT formulation, a gauge field $A_{i,j}$ and spinon field $\Phi_{i}$ are introduced through the assignment: $S^{+}_{i,j}=\Phi^{\dagger}e^{iA_{i,j}}\Phi_{j}/2$, where $i$ and $j$ label neighbouring tetrahedra on the ‘A’ and ‘B’ sublattices respectively. The spin component $S^{y}_{i,j}$ is correspondingly mapped to an emergent electric field $E_{i,j}$, which obeys an emergent Gauss’ law. Technical details of the construction used can be found in Ref. Desrochers and Kim (2024a), where it is also demonstrated that there exist two possible mean field solutions for gapped spinon effective actions consistent with the gauge, time-reversal and lattice symmetries. These are the $0$-flux and $\pi$-flux $U(1)$ dipolar-ocutpolar spin ice phases. In both phases, spinons on the ‘A’ and ‘B’ sublattices decouple, and the terms generated by the $J_{\pm\pm}$ term in Eq. (1) vanish in the mean field, again allowing us to freely set the mixing angle $\theta$ to zero. In the 0-flux phase, the ‘A’ and ‘B’ sublattice spinons both have a single dispersive band, whilst in the $\pi$-flux phase, the threading of $\pi$-flux through each hexagonal plaquette of the pyrochlore lattice leads to an enlargement of the unit cell, resulting in two doubly degenerate bands per sublattice.

The two-band dispersion of spinons in the $\pi$-flux phase has already been identified as a key signature distinguishing it from the $0$-flux regime in linear response Lee et al. (2012); Desrochers and Kim (2024b), and we find that spinon transitions between these two bands produce features in the 2DCS response that clearly distinguish the two phases. Figure 3 presents the 2DCS responses for both phases calculated analytically using GMFT. For the 0-flux phase, a single rephasing streak is obtained, whilst the $\pi$-flux response resembles a cluster of isolated peaks. The reason for this stark contrast is two-fold. First, each of the two distinct spinon bands in the $\pi$-flux phase are relatively flat, leading to a density of states with two well defined maxima, and hence relatively narrow bandwidth optical responses. Second, the probe operator has non-zero matrix elements to excite a spinon from one band to another, and these inter-band transitions result in the additional peaks off of the main rephasing diagonal. Details of this calculation are given in the supplementary materialsup .

Discussion- The 2DCS responses of spinons in dipolar-octupolar QSI are found to be sensitive to both temperature regime and phase of the system. Sharp responses reflecting the presence of fractionalized spinon excitations are predicted for the lowest $T$, and these responses are demonstrated to clearly distinguish between the spinon dynamics found in the 0-flux and $\pi$-flux phases of dipolar-octupolar spin ice. Conversely, in the intermediate temperature regime, a broad non-linear response is observed indicative of heavily constrained spinon motion in the presence of an incoherent spin background.

Whilst this letter has focused on the example of dipolar-octupolar pyrochlores, we hypothesise that the appearance of broad signatures in 2DCS where fractionalised excitations move across an incoherent spin background may occur more generally in similar frustrated models, presenting an obstacle to observing sharp signatures of fractionalisation at all but the lowest temperatures. On the other hand, by measuring the 2DCS signal as a function of temperature one could use the crossover between these regimes to demonstrate the onset of a quantum coherent low-temperature state. The GMFT calculation however does not include the effects of quantum fluctuations in the gauge field, which will be necessary to understand the cross-over between these two regimes.

A natural question is then whether the clear distinctions between the different spinon dynamical regimes observable in 2DCS can be observed more directly in the linear optical response to a uniform magnetic field, which is more simply accessed experimentally. In Fig. 4 we present the linear responses for all phases and regimes considered above. The linear response varies between each regime, and in particular the existence of multiple maxima in the $\pi$-flux response distinguishes it from the $T=0$ $0$-flux response, as already noted in Ref. Desrochers and Kim (2024a). However, the stark difference between the behaviour of the response in the intermediate temperature regime and the low temperature regime is entirely absent; the intermediate temperature response and the low temperature $0$-flux response are not qualitatively different, despite very different underlying spinon dynamics. By contrast, the existence of broad features in the intermediate temperature 2DCS response clearly distinguishes this regime from the low-temperature regime.

To summarise, we find that two-dimensional coherent spectroscopy is a suitable experimental setting to probe the dynamics of fractionalised excitations in candidate quantum spin ice materials, providing clear signatures distinguishing the regime where fractionalised quasiparticles move ballistically on the pyrochlore lattice at the lowest temperatures from that where their motion is heavily constrained by an incoherent spin background. Further, the appearance of response streaks away from the main rephasing diagonal in 2DCS offers a direct signature of the transition of quasiparticles between bands, which for dipolar-octupolar quantum spin ices allows one to distinguish experimentally between a $0$-flux or $\pi$-flux QSI ground states. We believe that this further demonstrates the potential utility of 2DCS as a probe of the dynamics of fractionalised excitations in quantum spin ice candidate materials, and other systems potentially hosting exotic quantum phases of matter.