betaFeedback

WorksPhysics and Astronomy

Two-Qutrit Quantum Battery Charging through Ising Model and Anisotropic XZ Heisenberg Interactions

Abstract

In this paper, we propose a model of a two-qutrit quantum battery composed of two spin-1 systems coupled to an external field via anisotropic Heisenberg interactions. From the Ising model, we derive an exact analytical expression of ergotropy in the parallel-charging case. However, the collective charging process offers an advantage over the parallel one. Moreover, increasing the collective exchange coupling accelerates the evolution of ergotropy and enhances the maximum average power, as measured by global coherence and mutual information. Conversely, the Dzyaloshinskii-Moriya (DM) interaction primarily acts as a detrimental mechanism, essentially reducing all of these quantities. In the XZ model, the obtained results show that the DM interaction is constructive when it cooperates with the longitudinal anisotropy and when the exchange coupling is sufficiently strong. Therefore, by optimising the maximum reachable ergotropy and maximum average power, we conclude that the Ising model without DM interaction is sufficient to maximise the ergotropy. The anisotropic XZ model allows the DM interaction to be exploited constructively to improve the maximum average power through the joint control of the exchange coupling, longitudinal anisotropy, and chiral interaction.

Introduction

At small scales, the study of energy exchanges does not always rely solely on the assumptions of classical thermodynamics [1, 2]. It must also incorporate the dynamics of quantum systems, where fluctuations, coherence effects and nonclassical correlations can modify energy evolution [2, 3, 4, 5]. Hence, this theoretical framework gives rise to the development of new energy-storage devices, particularly quantum batteries [6, 7, 8, 9]. In this regard, R. Alicki et al. showed that using global unitary operations generating entanglement among several independent copies of a quantum battery allows more work to be extracted than using local unitary operations applied separately to each battery [10]. Then, Binder et al. introduced the concept of the Quantacell and showed that collective charging protocols increase significantly the charging power compared with local operations [11]. Subsequently, Campaioli et al. formalized the notion of collective quantum advantage in power and established fundamental bounds showing that this advantage is limited when charging protocols rely on finite-order (kk-body) interactions, as is generally the case in realistic physical systems [12]. After that, numerous quantum-battery architectures have been proposed in the literature; including Dicke models coupled to cavities, two-level systems charged by an optical mode [13], spin chains [14], stable adiabatic protocols [15], many-body localized systems [16], as well as strongly correlated Sachdev–Ye–Kitaev-type models [17].

One of the most widely used energetic quantities for characterizing the performance of a quantum battery is ergotropy, which quantifies the maximum extractable work from a quantum system using unitary transformations [18, 19]. In many studies, the amount of ergotropy is studied together with the average charging power, which quantifies the rate at which extractable work is accumulated [7, 20, 21]. However, early studies on quantum batteries mainly focused on two-level systems (qubits), commonly used for studying quantum-battery charging processes [11, 13, 16, 20, 21, 22, 23, 24]. Recently, recent advances in quantum batteries deal with higher-dimensional systems, like three-level Dicke quantum batteries [25] and resonator–qutrit architectures [26]. In fact, qutrit quantum batteries offer a richer energy structure, additional transition channels and a larger Hilbert space [23, 25, 26, 27].

Among the models used to describe collective quantum batteries, Heisenberg spin chains constitute a relevant system used to investigate the effect of spin–spin interactions on the charging process and energetic performance. The Ising model provides a simple framework for analyzing the collective advantage associated with transverse coupling. Whereas anisotropic models, like the XZ model, offer additional control parameters for modulating the charging process and extractable work [14, 20, 21, 28, 29]. Moreover, antisymmetric interactions by means of DM coupling were introduced to explain weak ferromagnetism in antiferromagnets and were subsequently derived microscopically by Moriya from anisotropic superexchange induced by spin–orbit coupling [30, 31]. In the context of spin quantum batteries, several recent studies have shown that the DM interaction can significantly modify ergotropy, energy capacity, and charging power. However, its effect depends on the Heisenberg model considered, the coupling regime and the Hamiltonian parameters [20, 28]. However, the influence of DM interaction in quantum batteries remains largely unexplored, particularly when different exchange anisotropies are compared within the same system, which will be investigated in the present work.

More specifically, Zhang et al. [20] studied a two-qubit quantum battery with anisotropic Heisenberg and Dzyaloshinskii–Moriya interactions, whereas Bhattacharya et al. [21] investigated qubit networks and emphasized the role of geometry and connectivity. But, in our work, we address a complementary question by examining how the local spin-1 structure modifies the charging dynamics in a minimal two-cell quantum battery. The intermediate qutrit level changes the transitions induced by both the transverse charger and the intercell interaction. In particular, it allows the longitudinal anisotropy to tune the energetic mismatch between configurations directly coupled by the DM interaction, a local mechanism that is absent in the corresponding two-qubit bond, where these configurations have the same longitudinal interaction energy. Moreover, in the case of parallel-charging process, the two-qutrit and two-qubit batteries reach the same absolute ergotropy bound, whereas the two-qutrit battery exhibits a first full-charging time twice as long and a maximum average power reduced by a factor of two. Further, the interaction analysis shows that the parameters favoring the maximum reachable ergotropy and those maximizing the average power do not generally coincide.

In fact, this work investigates the charging process using a two-qutrit quantum battery (TQQB) composed of two spin-1 systems initially prepared in the ground state, which are interacted with an external field responsible for their charging dynamics. First, we consider the parallel-charging regime, in which the two qutrits evolve independently in the absence of mutual interaction, before examining the effect of an Ising-type exchange interaction on the charging performance. Fundamentally, we have derived an exact expression of time-dependent ergotropy associated with parallel charging, which is in excellent agreement with numerical simulations. This provides a reliable basis for comparison in the analysis of collective regimes. Then, our results show that the collective charging process exceeds the parallel one by accelerating the charging dynamics, which increases the maximum average power. Importantly, we show that increasing the collective exchange coupling improves the charging performance significantly. Conversely, the DM interaction mainly reduces the ergotropy and average power, particularly when the exchange coupling is weak.

To explain the opposing effects of the collective exchange coupling and the DM interaction in the Ising model, we analyze global coherence and mutual information associated with the charging dynamics [33, 34, 35]. We find that the performance improvement induced by the collective coupling is accompanied by an increase in these resources, whereas the DM interaction tends to reduce them. We then extend our study to the anisotropic model to examine the joint effect of the DM interaction and longitudinal anisotropy on the charging performance. Contrary to the Ising model, we show that DM coupling becomes constructive in the XZ model when the exchange coupling is sufficiently strong. In this regime, the longitudinal anisotropy and the chiral interaction cooperate to increase the maximum average power. However, this enhancement in power does not necessarily correspond to full charging. Thus, we show that the type of Heisenberg interaction, the XZ-type Heisenberg interaction considered, determines the role of the DM interaction in the charging performance, making this interaction regime an essential step for analyzing the charging of TQQB.

The paper is organized as follows. Sec. II presents the theoretical model and the physical quantities used to characterize the charging performance. Sec. III includes quantum resources by means of global coherence and mutual information. Sec. IV is devoted to the collective Ising model as well as the comparison between the parallel and collective charging processes. Sec. V investigates the anisotropic XZ model, the joint optimization of the maximum reachable ergotropy and the maximum average power. Sec. VI includes the microscopic mechanism of the DM interaction. Finally, Sec. VII summarizes the main results and discusses the future perspectives of this work.

TQQB Charging Model

TQQB Charging Dynamics

Free Storage Hamiltonian

Let us consider a TQQB, where each qutrit is modeled as a spin-1 system comprising three energy levels. Its Hilbert space has dimension 9, with the ordered product basis

B={∣+1,+1⟩,∣+1,0⟩,∣+1,−1⟩,∣0,+1⟩,∣0,0⟩,∣0,−1⟩,∣−1,+1⟩,∣−1,0⟩,∣−1,−1⟩}.(1)\mathcal{B}=\{ |+1,+1\rangle, |+1,0\rangle, |+1,-1\rangle, |0,+1\rangle, |0,0\rangle, |0,-1\rangle, |-1,+1\rangle, |-1,0\rangle, |-1,-1\rangle\}. \tag*{(1)}

The corresponding free Hamiltonian defines the energy stored in the battery as [20, 21]

H0=ℏω0(Sz(1)+Sz(2))=ℏω0(Sz⊗I3+I3⊗Sz),(2)H_0=\hbar\omega_0\left(S_z^{(1)}+S_z^{(2)}\right)=\hbar\omega_0\left(S_z\otimes I_3+I_3\otimes S_z\right), \tag*{(2)}

where ω0\omega_0 defines the Larmor frequency common to the two qutrits. Moreover, the spin-1 operators Sz(1)S_z^{(1)} and Sz(2)S_z^{(2)} act on the first and second spins along the zz direction, respectively. While, I3I_3 denotes the 3×33\times3 identity matrix.

It is worth mentioning that the charging dynamics of the battery can be interpreted as a transition between the ground state and the excited state, corresponding in our case to the absolute energy-capacity bound Wabs=4ℏω0\mathcal{W}_{\mathrm{abs}}=4\hbar\omega_0. In fact, the initial state, ρ(0)=∣−1,−1⟩⟨−1,−1∣\rho(0)=|-1,-1\rangle\langle-1,-1| represents a completely discharged battery with an energy associated with the free Hamiltonian H0H_0: −2ℏω0-2\hbar\omega_0. However, the excited state ∣+1,+1⟩⟨+1,+1∣|+1,+1\rangle\langle+1,+1| corresponds to a battery that is fully charged with the energy +2ℏω0+2\hbar\omega_0.

Charging Hamiltonian

The battery charging process is provided via a homogeneous transverse magnetic field, acting locally on each qutrit, as shown in the following Figure 1. The charging Hamiltonian is given as [20, 21]

A comparative scheme between parallel and collective charging protocols in a TQQB

Figure 1. A comparative scheme between parallel and collective charging protocols in a TQQB.

Hch=ℏΩ(Sx(1)+Sx(2))=ℏΩ(Sx⊗I3+I3⊗Sx),(3)H_{\mathrm{ch}}=\hbar\Omega\left(S_x^{(1)}+S_x^{(2)}\right)=\hbar\Omega\left(S_x\otimes I_3+I_3\otimes S_x\right), \tag*{(3)}

where Ω\Omega denotes the strength of the external field. While, Sx(1)S_x^{(1)}, Sx(2)S_x^{(2)} are spin-1 operators acting on the first and second spin along the xx direction, respectively. This term generates coherent rotations around the xx axis, which gives rise to a parallel charging of the battery in the absence of internal interactions.

Anisotropic Heisenberg Interaction

Let us shed light on the internal interactions between the two qutrits of the quantum battery. These interactions are described by an anisotropic Heisenberg Hamiltonian of the form [20, 21, 28, 29, 36, 37]

HJ=ℏJ[(1+γ)Sx(1)Sx(2)+(1−γ)Sy(1)Sy(2)+ΔSz(1)Sz(2)],(4)H_J=\hbar J\left[(1+\gamma)S_x^{(1)}S_x^{(2)}+(1-\gamma)S_y^{(1)}S_y^{(2)}+\Delta S_z^{(1)}S_z^{(2)}\right], \tag*{(4)}

where JJ denotes the exchange coupling constant, γ\gamma and Δ\Delta characterize the transverse anisotropy in xyxy space and the longitudinal anisotropy along zz axis, respectively. Depending on the values of γ\gamma and Δ\Delta, different effective models are recovered. In this work, we are particularly interested in studying the Ising model, such that γ=1\gamma=1 and Δ=0\Delta=0, which selects a dominant term of the Sx(1)Sx(2)S_x^{(1)}S_x^{(2)} type. Moreover, we aim to investigate the anisotropic XZ-type model, which corresponds to γ=1\gamma=1 and Δ≠0\Delta\ne0, describing a coupling along the xx direction and an anisotropic coupling along the zz direction controlled by the strength of the parameter Δ\Delta.

Dzyaloshinskii–Moriya Interaction

In addition, we include the DM interaction, which corresponds to an antisymmetric exchange term between localized spins, arising in the presence of spin–orbit coupling in the absence of inversion symmetry between the magnetic sites [20, 21, 29, 31, 36, 38]. In fact, the corresponding Hamiltonian is described as follows:

HD=∑⟨i,j⟩Dij⋅(S(i)×S(j)),(5)H_D=\sum_{\langle i,j\rangle}\mathbf{D}^{ij}\cdot\left(\mathbf{S}^{(i)}\times\mathbf{S}^{(j)}\right), \tag*{(5)}

where S(i)\mathbf{S}^{(i)} and S(j)\mathbf{S}^{(j)} are the spin operators at sites ii and jj. While, Dij\mathbf{D}^{ij} defines the DM vector determined by the local symmetry of the lattice. Here, we consider the standard case of a uniform uniaxial DM interaction oriented along the zz axis, namely Dij=Dz=D\mathbf{D}^{ij}=D_z=D, which leads to

HD(z)=D∑⟨i,j⟩(Sx(i)Sy(j)−Sy(i)Sx(j)).(6)H_D^{(z)}=D\sum_{\langle i,j\rangle}\left(S_x^{(i)}S_y^{(j)}-S_y^{(i)}S_x^{(j)}\right). \tag*{(6)}

In the particular case of two spins, one can obtain the following Hamiltonian:

HD(z)=D(Sx(1)Sy(2)−Sy(1)Sx(2)).(7)H_D^{(z)}=D\left(S_x^{(1)}S_y^{(2)}-S_y^{(1)}S_x^{(2)}\right). \tag*{(7)}
Effective Battery Hamiltonian

According to the previous Hamiltonians, we are now able to introduce the charging Hamiltonian of the TQQB, which is described by the dynamical Hamiltonian: [14, 20]

Hdyn=Hch+Hint,(8)H_{\mathrm{dyn}}=H_{\mathrm{ch}}+H_{\mathrm{int}}, \tag*{(8)}

where HintH_{\mathrm{int}} reflects the interaction Hamiltonian between the two qutrits, such that

Hint=HJ+HD,(9)H_{\mathrm{int}}=H_J+H_D, \tag*{(9)}

where HJH_J denoting the interaction Hamiltonian in the Heisenberg spin-chain model, and HDH_D DM one already defined in Eqs. (4) and (7), respectively. Hence, the evolution of the quantum battery is described by the following unitary dynamics [39]:

ρ(t)=e−iHt/ℏρ(0)e+iHt/ℏ.(10)\rho(t)=e^{-iHt/\hbar}\rho(0)e^{+iHt/\hbar}. \tag*{(10)}

Thermodynamic Quantities of the TQQB

To evaluate the performance of the proposed quantum battery, two quantities are introduced in our case, namely, ergotropy and average charging power.

Ergotropy for the TQQB

Quantum ergotropy is defined as the maximum work that can be extracted from a given quantum state ρ(t)\rho(t). In fact, the amount of ergotropy can be derived using a unitary operation relative to the free Hamiltonian H0H_0 as [11, 14, 19]

W(t)=Tr⁡[ρ(t)H0]−Tr⁡[ρp(t)H0],(11)\mathcal{W}(t)=\operatorname{Tr}[\rho(t)H_0]-\operatorname{Tr}[\rho_{\mathrm{p}}(t)H_0], \tag*{(11)}

where ρp(t)\rho_{\mathrm{p}}(t) denotes the passive state associated with ρ(t)\rho(t).

More precisely, if the eigenvalues of the density matrix are ordered as r1≥⋯≥r9r_1 \ge\cdots\ge r_9 and the eigenenergies of H0H_0 as ϵ1≤⋯≤ϵ9\epsilon_1 \le\cdots\le\epsilon_9, the passive state and its energy are given as follows:

ρp(t)=∑j=19rj∣ϵj⟩⟨ϵj∣,Ep=∑j=19rjϵj.(12)\rho_{\mathrm{p}}(t)=\sum_{j=1}^{9}r_j\lvert\epsilon_j\rangle\langle\epsilon_j\rvert,\qquad E_{\mathrm{p}}=\sum_{j=1}^{9}r_j\epsilon_j. \tag*{(12)}

Thus, the passive-state construction depends on the eigenvalue spectrum of the density matrix, rather than merely on its diagonal populations in the energy basis [19]. This expression can be written in the equivalent form [6, 40, 41, 42]

W(t)=Tr⁡[ρ(t)H0]−min⁡U∈UTr⁡[Uρ(t)U†H0],(13)\mathcal{W}(t)=\operatorname{Tr}[\rho(t)H_0]-\min_{U\in\mathcal{U}}\operatorname{Tr}[U\rho(t)U^\dagger H_0], \tag*{(13)}

where U\mathcal{U} denotes the set of all unitary operations.

For the pure initial state introduced above and the closed, unitary charging dynamics considered here, the spectrum of ρ(t)\rho(t) remains (1,0,…,0)(1,0,\ldots,0). Consequently, the passive state remains the nondegenerate ground state ∣−1,−1⟩\lvert-1,-1\rangle, with energy −2ℏω0-2\hbar\omega_0, where

W(t)=Tr⁡[ρ(t)H0]+2ℏω0≤Wabs,Wabs=4ℏω0.(14)\mathcal{W}(t)=\operatorname{Tr}[\rho(t)H_0]+2\hbar\omega_0\leq W_{\mathrm{abs}},\qquad W_{\mathrm{abs}}=4\hbar\omega_0. \tag*{(14)}

The absolute bound is reached only when the state is completely transferred to the nondegenerate upper level ∣+1,+1⟩\lvert+1,+1\rangle. This condition is specific to the pure-state unitary dynamics considered here. For a mixed initial state, unitary evolution preserves the eigenvalue spectrum of the density matrix and the maximum extractable work is generally constrained by this spectrum and by passivity [19].

In the absence of any interaction between the two spins, J=D=0J=D=0. In this case, each qutrit evolves independently under the effect of the external field, where the battery operates in parallel mode. Hence, the corresponding ergotropy follows a sinusoidal law as (see Appendix A)

W∥qtr(t)=4ℏω0sin⁡2(Ωt2),(15)\mathcal{W}_{\parallel}^{\mathrm{qtr}}(t)=4\hbar\omega_0\sin^2\left(\frac{\Omega t}{2}\right), \tag*{(15)}

where the maximum reachable ergotropy W∥qtr(max⁡)=Wabs=4ℏω0\mathcal{W}_{\parallel}^{\mathrm{qtr}}(\max)=W_{\mathrm{abs}}=4\hbar\omega_0 is reached for a minimum charging time tmin⁡=π/Ωt_{\min}=\pi/\Omega. This latter time will serve as a comparative reference in the following studies concerning collective charging, in the absence and presence of DM interactions.

Average Charging Power

The average charging power Pavg(t)P_{\mathrm{avg}}(t) is defined as the time-averaged value of the instantaneous power P(t)P(t) over a time interval [0,t][0,t] of the charging process. It is expressed as [41]

Pavg(t)=1t∫0tP(t′) dt′=W(t)−W(0)t,(16)P_{\mathrm{avg}}(t)=\frac{1}{t}\int_0^tP(t')\,\mathrm{d}t'=\frac{\mathcal{W}(t)-\mathcal{W}(0)}{t}, \tag*{(16)}

W(t)\mathcal{W}(t) denotes the maximum extractable work after the charging time tt. In the case where the initial extractable energy is zero, this expression reduces to [14, 20, 37]

Pavg(t)=W(t)t.(17)P_{\mathrm{avg}}(t)=\frac{\mathcal{W}(t)}{t}. \tag*{(17)}

In the parallel-charging case, the average charging power is calculated in our case straightforwardly as follows:

Pavg,∥qtr(t)=4ℏω0sin⁡2(Ωt2)t.(18)P_{\mathrm{avg},\parallel}^{\mathrm{qtr}}(t)=\frac{4\hbar\omega_0\sin^2\left(\frac{\Omega t}{2}\right)}{t}. \tag*{(18)}

The average power constitutes a relevant quantity for evaluating the overall efficiency of the charging process of a quantum battery, taking into account both the amount of stored energy and the evolution time of the process.

Quantum Correlation

To highlight the influence of the Heisenberg coupling JJ and the DM interaction on the system dynamics, we examine different forms of correlations by means of global coherence and mutual information.

Global Coherence

Within the formalism introduced by Baumgratz [32], we use the l1l_1 norm to quantify the coherence defined relative to the energy basis of the storage Hamiltonian H0H_0. In this framework, the l1l_1 coherence is defined as the distance between the quantum state ρ\rho and its closest incoherent state σ\sigma, belonging to the set I\mathcal{I} of incoherent states that are diagonal in the eigenbasis of H0H_0. Hence, the l1l_1 norm of coherence is explicitly written as [32, 43]

Cl1(H0)(ρ)=min⁡σ∈I∥ρ−σ∥l1=∑i≠j∣ρij(H0)∣,(19)C_{l_1}^{(H_0)}(\rho)=\min_{\sigma\in\mathcal{I}}\lVert\rho-\sigma\rVert_{l_1}=\sum_{i\ne j}\left\lvert\rho_{ij}^{(H_0)}\right\rvert, \tag*{(19)}

where ρij(H0)=⟨Ei∣ρ∣Ej⟩\rho_{ij}^{(H_0)}=\langle E_i\lvert\rho\lvert E_j\rangle denotes the off-diagonal elements of the density matrix in the energy basis of H0H_0, namely {∣Ei⟩}\{\lvert E_i\rangle\}. This quantity vanishes if and only if the state ρ\rho is incoherent in the energy basis of H0H_0, and increases with the creation of quantum superpositions between distinct energy levels generated by the charging dynamics. For a quantum battery composed of two qutrits, the maximum coherence associated with the l1l_1 norm is reached for a maximally coherent pure state of the form ∣ψmax⁡⟩=(1/d)∑i=1d∣Ei⟩|\psi_{\max}\rangle= (1/\sqrt{d})\sum_{i=1}^{d}|E_i\rangle and is equal to Cl1,max⁡(H0)=d−1=8C_{l_1,\max}^{(H_0)} = d-1=8 [32].

To obtain a bounded measure independent of the system’s dimension, we introduce the normalized global coherence as follows: [40]

Cl1,norm(H0)(t)=1Cl1,max⁡(H0)Cl1(H0)(ρ)=18∑i≠j∣ρij(H0)∣.(20)C_{l_1,\mathrm{norm}}^{(H_0)}(t)=\frac{1}{C_{l_1,\max}^{(H_0)}}C_{l_1}^{(H_0)}(\rho)=\frac{1}{8}\sum_{i\ne j}\left|\rho_{ij}^{(H_0)}\right|. \tag*{(20)}

The above normalized global coherence satisfies 0≤Cl1,norm(H0)(t)≤10\le C_{l_1,\mathrm{norm}}^{(H_0)}(t)\le1. This quantity provides a dimensionless quantification of the coherence resource defined relative to the storage Hamiltonian H0H_0 and will be used to analyze the evolution of global coherence during the TQQB’s charging process.

MutualInformation

Let us examine the total correlations between the two qutrits using the so-called quantum mutual information, [6, 33, 34, 35, 39]

I(ρAB)=S(ρA)+S(ρB)−S(ρAB),(21)I(\rho_{AB})=S(\rho_A)+S(\rho_B)-S(\rho_{AB}), \tag*{(21)}

where [44] S(ρ)=−Tr⁡(ρln⁡ρ)S(\rho)=-\operatorname{Tr}(\rho\ln\rho) defines the von-Neumann entropy. For a qutrit-qutrit system, the theoretical maximum value of the mutual information is achieved for a maximally entangled pure global state and equals Imax⁡=2ln⁡3I_{\max}=2\ln3. However, to facilitate the comparison between different dynamical regimes, we introduce the normalized mutual information as follows:

Inorm(ρAB)=S(ρA)+S(ρB)−S(ρAB)2ln⁡3,(22)I_{\mathrm{norm}}(\rho_{AB})=\frac{S(\rho_A)+S(\rho_B)-S(\rho_{AB})}{2\ln3}, \tag*{(22)}

where Inorm=0I_{\mathrm{norm}}=0 corresponds to two independent qutrits, and Inorm=1I_{\mathrm{norm}}=1 characterizes a maximally entangled pure global state.

Ising Model

To study collective effects in the TQQB charging process, we consider a collective Ising model with interaction along the xx direction as [20, 21]

HintIsing=2ℏJSx(1)Sx(2).(23)H_{\mathrm{int}}^{\mathrm{Ising}}=2\hbar J S_x^{(1)}S_x^{(2)}. \tag*{(23)}

This model constitutes a minimal reference for evaluating the collective advantage in the dynamics of W(t)W(t) and Pavg(t)P_{\mathrm{avg}}(t), which allows us to investigate the influence of the coupling JJ on the charging kinetics to be analyzed. It also provides the basis for studying additional effects induced by the antisymmetric DM interaction and the role of quantum resources, such as global coherence and mutual information, in the efficiency of energy transfer.

Parallel Charging versus Collective Charging

First, let us compare the charging dynamics of the TQQB in two distinct regimes, a noninteracting parallel regime and a collective regime governed by an Ising-type interaction. The corresponding results are presented in Figure 2, where W(t)W(t) and Pavg(t)P_{\mathrm{avg}}(t) are plotted versus the dimensionless time Ωt\Omega t. In the following, we set ℏ=ω0=Ω=1\hbar=\omega_0=\Omega=1

Figure 2. Comparison between parallel and collective charging in the Ising model for a two qutrits quantum battery. (a) Ergotropy W(t)W(t) and (b) average charging power Pavg(t)P_{\mathrm{avg}}(t) versus the dimensionless time Ωt\Omega t, where J=1J=1.

In the parallel regime, each qutrit is charged independently by the local transverse field. Figure. Figure 2(a) shows that W∥qtr(t)W_{\parallel}^{\mathrm{qtr}}(t) increases monotonically from the initial uncharged state and reaches its maximum value W∥,max⁡qtr=4ℏω0W_{\parallel,\max}^{\mathrm{qtr}}=4\hbar\omega_0 at time Ωt=π\Omega t=\pi. This behavior is in perfect agreement with our analytical expression given in Eq. (15), which describes the independent charging of the two qutrits. The associated average power, shown in Figure 2(b), exhibits a single maximum Pavg,∥qtr,max⁡≃1.45ℏω0ΩP_{\mathrm{avg},\parallel}^{\mathrm{qtr},\max}\simeq1.45\hbar\omega_0\Omega reached around Ωt≃2.33\Omega t\simeq2.33. However, when the collective Ising-type interaction is activated, the charging dynamics is strongly accelerated. Although the maximum reachable ergotropy remains unchanged and saturates the absolute bound, Wabs=4ℏω0W_{\mathrm{abs}}=4\hbar\omega_0, because it is imposed by the spectrum of the free Hamiltonian H0H_0, the ergotropy increases much more rapidly at short times compared to the parallel process, as illustrated in Figure 2(a).

The collective effect is particularly visible in the average power. The maximum power increases from Pavg,∥qtr,max⁡≃1.45ℏω0ΩP_{\mathrm{avg},\parallel}^{\mathrm{qtr},\max}\simeq1.45\hbar\omega_0\Omega in the parallel case to Pavg,collqtr,max⁡≃1.69ℏω0ΩP_{\mathrm{avg},\mathrm{coll}}^{\mathrm{qtr},\max}\simeq1.69\hbar\omega_0\Omega in the collective regime, where the optimal time is reduced from Ωt≃2.33\Omega t\simeq2.33 to Ωt≃0.94\Omega t\simeq0.94. Thus, the Ising interaction does not increase the maximum reachable ergotropy, which already saturates WabsW_{\mathrm{abs}}, but modifies the charging rate profoundly by reducing the optimal time and increasing Pavgmax⁡P_{\mathrm{avg}}^{\max}. This behavior constitutes a clear signature of a collective charging advantage in the TQQB.

In Sec. IVB, these results will motivate us to analyze the impact of the exchange coupling strength JJ on the dynamics of the TQQB.

Effect of the exchange coupling JJ

As discussed before, it is essential to investigate the impact of the Ising coupling strength JJ on the collective dynamics in the absence of the DM interaction. Indeed, we evaluate in Figure 3 W(t)W(t) and Pavg(t)P_{\mathrm{avg}}(t) for weak, intermediate and strong coupling, namely J=0.2ΩJ=0.2\Omega, J=1ΩJ=1\Omega and J=2ΩJ=2\Omega, respectively.

Figure 3. Impact of Ising coupling strength JJ on the charging dynamics of a two-qutrit quantum battery in the absence of DM interaction (D=0D=0). (a) Time evolution of W(t)W(t). (b) Time evolution of Pavg(t)P_{\mathrm{avg}}(t).

Obviously, increasing JJ accelerates the initial growth of the ergotropy in Figure 3(a). For J=0.2ΩJ=0.2\Omega, the dynamics remains close to a weakly collective regime dominated by the local excitation imposed by the external field (charger), where the ergotropy increases then slowly. When JJ becomes comparable to Ω\Omega, the collective effects become stronger and the TQQB reaches the absolute bound Wabs=4ℏω0W_{\mathrm{abs}}=4\hbar\omega_0 at time Ωt=π\Omega t=\pi. For J=2ΩJ=2\Omega, the ergotropy increases more rapidly at short times and exhibits oscillations of larger amplitude, which reaches the maximum bound 4ℏω04\hbar\omega_0 at almost the same time Ωt=π\Omega t=\pi.

In Figure 3(b), the influence of the coupling JJ appears even more clearly in the average power. The maximum power increases from Pavgmax⁡≃1.35ℏω0ΩP_{\mathrm{avg}}^{\max}\simeq1.35\hbar\omega_0\Omega for J=0.2ΩJ=0.2\Omega to Pavgmax⁡≃1.69ℏω0ΩP_{\mathrm{avg}}^{\max}\simeq1.69\hbar\omega_0\Omega for J=1.0ΩJ=1.0\Omega, and then to Pavgmax⁡≃3.00ℏω0ΩP_{\mathrm{avg}}^{\max}\simeq3.00\hbar\omega_0\Omega for J=2.0ΩJ=2.0\Omega. Simultaneously, the optimal time is reduced from Ωtopt≃2.05\Omega t_{\mathrm{opt}}\simeq2.05 to 0.940.94, and then to 0.550.55. This shows that the enhancement in the quantum battery’s performance depends strongly on the ratio J/ΩJ/\Omega. In the strong-coupling regime, the average power is higher than in weak and intermediate couplings without altering the maximum ergotropy bound but allowing it to be reached more rapidly. This amplification is accompanied by more oscillatory dynamics, reflecting faster energy exchanges between the charger and TQQB, which reduces the duration during which the maximum power can be exploited.

From these results, the exchange coupling in the Ising model acts as a collective dynamical resource that can accelerate energy transfer and increase maximum power. Moreover, the present results constitute an important reference for comparison in analyzing the effect of more general interactions. This will make it possible to evaluate in Sec. IV C whether the antisymmetric DM interaction can enhance or not the collective advantage observed in the Ising model.

Influence of the DM Interaction on Charging

Let us analyze the effect of DM interaction on the charging dynamics of TQQB in the Ising model. Figure. Figure 4 provides the time evolution of W(t)W(t) as well as Pavg(t)P_{\mathrm{avg}}(t), for different values of the DM interaction strength DD and for three characteristic regimes of the exchange coupling JJ. The results highlight a competition between the parameters DD and JJ. For the model considered, increasing DD leads to a progressive degradation of the charging process in ergotropy as well as a weakening of the maximum average power. This degradation is particularly apparent for low values of JJ. Whereas it becomes less pronounced when the exchange coupling is strengthened and the collective Ising interaction increasingly dominates the dynamics. In contrast, increasing values of JJ significantly improve the charging efficiency. In particular, when JJ increases from 0.2Ω0.2\Omega to 2Ω2\Omega, W(t)W(t) increases significantly. Meanwhile, the average power maxima increase and are reached at shorter times.

Figure 4. Dynamics of W(t)W(t) (a, c, e) and Pavg(t)P_{\mathrm{avg}}(t) (b, d, f) using Ising-type Heisenberg model in the presence of DM interaction. Panels (a,b), (c,d), and (e,f) correspond to J=0.2ΩJ=0.2\Omega, J=1ΩJ=1\Omega, and J=2ΩJ=2\Omega, respectively.

In summary, one can conclude that the DM interaction modifies the charging dynamics by reducing the efficiency of energy transfer via decreasing W(t)W(t) and Pavg(t)P_{\mathrm{avg}}(t). Conversely, strengthening JJ enhances efficient energy transfer of the TQQB. Nevertheless, this behavior depends on the nature of the coupling and could be modified in other types of the Heisenberg model, as will be discussed later in § 2. However, the competition between JJ and DD will be interpreted Sec. IV D subsection from the dynamics of coherence and quantum mutual information.

Quantum Resources in the Ising Model

To explain the performance observed in the previous subsection, we examine in Figure 5 the quantum resources by means of global quantum coherence and quantum correlations in order to identify their contribution to the charging dynamics of the TQQB. Figures. Figure 5(a,Figure 5c,Figure 5e) represent the normalized global coherence, namely Cl1,norm(H0)(t)C_{l_{1},\mathrm{norm}}^{(H_0)}(t). While Figure 5(b,Figure 5d,Figure 5f) illustrate the normalized mutual information Inorm(t)I_{\mathrm{norm}}(t) for different values of JJ and DD. We observe that increasing the DM interaction tends to reduce Cl1,norm(H0)(t)C_{l_{1},\mathrm{norm}}^{(H_0)}(t) and Inorm(t)I_{\mathrm{norm}}(t). This effect is particularly shown in the weak-coupling regime, namely J=0.2ΩJ=0.2\Omega, where the curves corresponding to large values of DD remain below those obtained for D=0D=0 in Figure 5(a,Figure 5b). This lowering becomes less pronounced when JJ is strong, although the dynamics becomes more oscillatory and non-monotonic in Figure 5(c–Figure 5f).

Figure 5. Dynamics of Cl1,norm(H0)(t)C^{(H_0)}_{l_1,\mathrm{norm}}(t) (left panels) and Inorm(t)I_{\mathrm{norm}}(t) (right panels) for the TQQB in the Ising-type model in the presence of DM interaction. Panels (a,b), (c,d), and (e,f) correspond to J=0.2ΩJ=0.2\Omega, J=1ΩJ=1\Omega, and J=2ΩJ=2\Omega, respectively.

However, increasing JJ enhances global coherence and correlations between the two qutrits. When JJ increases from 0.2Ω0.2\Omega to 2Ω2\Omega, Cl1,norm(H0)(t)C_{l_{1},\mathrm{norm}}^{(H_0)}(t) reaches higher values and exhibits more pronounced oscillations. Similarly, Inorm(t)I_{\mathrm{norm}}(t) increases basically when JJ increases, which supports the presence of collective correlations in the system. This trend is consistent with the enhancement in W(t)W(t) and Pavg(t)P_{\mathrm{avg}}(t) previously highlighted in Figure 3 and Figure 4, respectively. Thus, we conclude that the DM interaction limits the generation of global coherence and correlations between the qutrits. This basically also shows the amounts of ergotropy and average power highlighted in Sec. IVC in Figure 4. Conversely, the coupling JJ supports basically the presence of quantum correlations and coherence, allowing the enhancement of the performance of the charging process observed in Figure 3 and Figure 4.

XZ Model

After analyzing the effect of DM interaction strength DD and the exchange coupling JJ in the Ising model, we aim in this part to extend our study to an anisotropic XZ-type Heisenberg model. This extension is motivated by the fact that DM interaction can become constructive when the exchange coupling JJ is sufficiently strong. In fact, unlike the Ising model, which contains only a coupling along the xx direction, the XZ model additionally introduces a longitudinal coupling along the zz direction, controlled by the longitudinal anisotropy parameter Δ\Delta. This makes it possible to examine how Δ\Delta and DD jointly

act on W(t)W(t) and Pavg(t)P_{\mathrm{avg}}(t). In this case, the interaction Hamiltonian turns out to be [20]

HintXZ=ℏJ(2Sx(1)Sx(2)+ΔSz(1)Sz(2)).(24)H_{\mathrm{int}}^{\mathrm{XZ}}=\hbar J\left(2S_x^{(1)}S_x^{(2)}+\Delta S_z^{(1)}S_z^{(2)}\right). \tag*{(24)}

Note that in the above Hamiltonian, the case of Δ=0\Delta=0 recovers the Ising case studied previously.

Simultaneous impact of DD and Δ\Delta

Figure. Figure 6 exhibits dynamics of the maximum reachable ergotropy Wmax⁡W_{\max} and maximum average power Pavgmax⁡P_{\mathrm{avg}}^{\max} based on the simultaneous impact of DD and Δ\Delta within the XZ model. Panels (a, b, e, f) correspond to J=1ΩJ=1\Omega regime. While, (c, d, g, h) reflect the stronger-coupling regime, namely J=2ΩJ=2\Omega. For J=1ΩJ=1\Omega, the configuration (Δ,D)=(0,0)(\Delta,D)=(0,0) yields a maximum reachable ergotropy that saturates the absolute ergotropy bound, namely Wabs=4ℏω0W_{\mathrm{abs}}=4\hbar\omega_0 as shown in Figure 6(a). This limit corresponds to an Ising-type regime in which the exchange coupling allows for a complete transfer to the fully charged state. Figure. Figure 6(e) confirms that the maximum reachable ergotropy, which saturates the absolute bound 4ℏω04\hbar\omega_0, is reached at time ΩtWmax⁡≃3.142\Omega t_{W_{\max}}\simeq3.142. For the same configuration, the maximum average power Pavgmax⁡≃1.691ℏω0ΩP_{\mathrm{avg}}^{\max}\simeq1.691\hbar\omega_0\Omega is reached earlier at time ΩtPavgmax⁡≃0.939ℏω0Ω\Omega t_{P_{\mathrm{avg}}^{\max}}\simeq0.939\hbar\omega_0\Omega as shown in Figure 6(f). Thus, (Δ,D)=(0,0)(\Delta,D)=(0,0) maximizes the extractable work but does not maximize the whole average power.

Figure 6. Plots of maximum reachable ergotropy (left panels) and maximum average power (right panels) using XZ model. We set J=1ΩJ = 1\Omega and J=2ΩJ = 2\Omega for (a,b) and (c,d), respectively. In (e,f), we compare these results when (Δ,D)=(0,0)(\Delta, D) = (0, 0) and (Δ,D)=(2.36,0.00)(\Delta, D) = (2.36, 0.00), respectively. However, (g,h) compare the results when (Δ,D)=(0,0)(\Delta, D) = (0, 0) and (Δ,D)=(1.98,2.36)(\Delta, D) = (1.98, 2.36), respectively.

In addition, the average power is maximized for J=1ΩJ = 1\Omega and (Δ,D)=(2.36,0.00)(\Delta, D) = (2.36, 0.00) as shown in Figure 6(b). At this point, the maximum average power increases to Pavgmax⁡≃2.228ℏω0ΩP_{\mathrm{avg}}^{\max} \simeq2.228\hbar\omega_{0}\Omega at time ΩtPavgmax⁡≃1.356\Omega t_{P_{\mathrm{avg}}^{\max}} \simeq1.356 [see Figure 6(f)]. However, this enhancement in power occurs at the cost of a reduced maximum reachable ergotropy, namely Wmax⁡≃3.699ℏω0W_{\max} \simeq3.699\hbar\omega_{0} at ΩtWmax⁡≃2.042\Omega t_{W_{\max}} \simeq2.042 as clearly observed in panel (e). Thus, in this intermediate-coupling regime, the enhancement in Pavgmax⁡P_{\mathrm{avg}}^{\max} is mainly governed by the longitudinal anisotropy Δ\Delta. Whereas the DM coupling does not yet play a constructive role, since the optimum is obtained for D=0D = 0.

For J=2ΩJ = 2\Omega, the configuration (Δ,D)=(0,0)(\Delta, D) = (0, 0) gives rise to the absolute ergotropy bound, namely Wabs=4ℏω0W_{\mathrm{abs}} = 4\hbar\omega_{0}. For the same point, the maximum average power is strongly increased compared to J=1ΩJ = 1\Omega. This allows for Pavgmax⁡≃2.999ℏω0ΩP_{\mathrm{avg}}^{\max} \simeq2.999\hbar\omega_{0}\Omega at time ΩtPavgmax⁡≃0.548\Omega t_{P_{\mathrm{avg}}^{\max}} \simeq0.548. Interestingly enough, increasing the coupling JJ already accelerates the charging process in the limit (Δ,D)=(0,0)(\Delta, D) = (0, 0), while reaching the absolute ergotropy bound. However, for J=2ΩJ = 2\Omega, the global maximum of the average power is no longer obtained at (Δ,D)=(0,0)(\Delta, D) = (0, 0), but at (Δ,D)=(1.98,2.36)(\Delta, D) = (1.98, 2.36). At this point, the maximum average power reaches Pavgmax⁡≃3.360ℏω0ΩP_{\mathrm{avg}}^{\max} \simeq3.360\hbar\omega_{0}\Omega at time ΩtPavgmax⁡≃0.773\Omega t_{P_{\mathrm{avg}}^{\max}} \simeq0.773. As in the case of J=1ΩJ = 1\Omega, this enhancement in power is accompanied by a lower maximum reachable ergotropy, Wmax⁡≃3.029ℏω0W_{\max} \simeq3.029\hbar\omega_{0} at ΩtWmax⁡≃2.994\Omega t_{W_{\max}} \simeq2.994. Thus, in the strong-coupling regime, the DM interaction contributes to maximizing the average power. Combined with Δ\Delta, it accelerates the charging process without necessarily reaching the upper ergotropy bound.

A central result emerges from Figure 6 reflects that the optimal conditions associated with Wmax⁡W_{\max} and Pavgmax⁡P_{\mathrm{avg}}^{\max} do not coincide. In both proposed coupling regimes, the absolute ergotropy bound, Wabs=4ℏω0W_{\mathrm{abs}} = 4\hbar\omega_{0}, is reached in the limit (Δ,D)=(0,0)(\Delta, D) = (0, 0), where the XZ model approaches collective Ising dynamics. The optimization of the average power requires moving away from this limit. For J=1ΩJ = 1\Omega, it is obtained mainly through the introduction of a finite longitudinal anisotropy in the absence of DM interaction. For J=2ΩJ = 2\Omega, it results from the joint action of Δ\Delta and a finite value of DD. Ultimately, the XZ model highlights an important qualitative difference from the Ising model. In both cases, increasing the coupling JJ increases the charging process of TQQB. However, the role of DM interaction depends strongly on the nature of the exchange coupling. More precisely, it becomes constructive in the XZ model when the collective coupling is sufficiently strong to efficiently exploit the chiral component that it introduces.

Population Analysis of Wmax⁡W_{\max} and Pavgmax⁡P_{\mathrm{avg}}^{\max} Optima

This subsection provides more detailed physical explanations of the results obtained in § 2 A based on population dynamics. Indeed, the populations are projected onto the eigensectors of the free Hamiltonian H0H_{0} given by E/ℏω0=−2,−1,0,+1,+2E/\hbar\omega_{0} = -2, -1, 0, +1, +2. Figure 7 shows that the optimum of Wmax⁡W_{\max} and Pavgmax⁡P_{\mathrm{avg}}^{\max} are based on two distinct physical mechanisms. The first one shows that the optimum of Wmax⁡W_{\max} is obtained when the population is almost entirely transferred to the fully charged state ∣+1,+1⟩\lvert+1,+1\rangle, which belongs to E/ℏω0=+2E/\hbar\omega_{0} = +2. The second mechanism reflects that the optimum of Pavgmax⁡P_{\mathrm{avg}}^{\max} does not require full charging, but it is favored by a faster partial transfer to the high-energy values, particularly E/ℏω0=+1E/\hbar\omega_{0} = +1 and E/ℏω0=+2E/\hbar\omega_{0} = +2.

Figure 7. Population dynamics at the optimal charging times. (a,b) Energy-level populations at tWmax⁡t_{W_{\max}} for J=1ΩJ = 1\Omega and J=2ΩJ = 2\Omega, respectively. (c,d) display populations at tPavgmax⁡t_{P_{\mathrm{avg}}^{\max}}. While, (e,f) represent time evolution of the fully charged-state population p∣+1,+1⟩(t)p_{\lvert+1,+1\rangle}(t) for J=1ΩJ = 1\Omega and J=2ΩJ = 2\Omega.

Figure 7(a,Figure 7b) show the energy-resolved populations, pE(tWmax⁡)p_{E}(t_{W_{\max}}), which correspond to the population at the time when the maximum reachable ergotropy is attained, for J=1ΩJ = 1\Omega and J=2ΩJ = 2\Omega, respectively. Figure 7(c,Figure 7d) show the same populations at time tavgPmax⁡t_{\mathrm{avg}}^{P_{\max}} when the average power now reaches its maximum. Finally, Figure 7(e,Figure 7f) display the population of the fully charged state, p+2(t)≡p∣+1,+1⟩(t)p_{+2}(t) \equiv p_{\lvert+1,+1\rangle}(t), for J=1ΩJ = 1\Omega and J=2ΩJ = 2\Omega, respectively. For J=1ΩJ = 1\Omega, the optimal value of Wmax⁡W_{\max} when (Δ,D)=(0,0)(\Delta,D) = (0,0) leads to a full charging process, as illustrated in § 2 A. At time ΩtWmax⁡≃3.142\Omega t_{W_{\max}} \simeq3.142, the entire population is transferred to E/ℏω0=+2E/\hbar\omega_0 = +2, with p+2(tWmax⁡)=1p_{+2}(t_{W_{\max}}) = 1 [see red color in Figure 7(a,Figure 7e)]. This complete occupation of the state ∣+1,+1⟩\lvert+1,+1\rangle explains straightforwardly why the ergotropy reaches its absolute ergotropy bound, Wabs=4ℏω0W_{\mathrm{abs}} = 4\hbar\omega_0. However, this same point does not maximize Pavgmax⁡P_{\mathrm{avg}}^{\max}, which is reached much earlier at ΩtPavgmax⁡≃0.939\Omega t_{P_{\mathrm{avg}}^{\max}} \simeq0.939. In this case, the population of ∣+1,+1⟩\lvert+1,+1\rangle is still very low around p+2(tPavgmax⁡)≃0.029p_{+2}(t_{P_{\mathrm{avg}}^{\max}}) \simeq0.029 [red color in Figure 7(c,Figure 7e)]. We conclude that full charging is achieved for ergotropy, but it occurs too late around ΩtWmax⁡≃3.142\Omega t_{W_{\max}} \simeq3.142 to maximize Pavgmax⁡P_{\mathrm{avg}}^{\max}.

The behavior is different at the point (Δ,D)=(2.36,0.00)(\Delta,D) = (2.36,0.00), which maximizes Pavgmax⁡P_{\mathrm{avg}}^{\max} for J=1ΩJ = 1\Omega, which does not allow the full charging process. Indeed, at time tWmax⁡t_{W_{\max}}, the population of ∣+1,+1⟩\lvert+1,+1\rangle is p+2(tWmax⁡)≃0.875p_{+2}(t_{W_{\max}}) \simeq0.875, which limits the maximum reachable ergotropy to Wmax⁡≃3.699ℏω0W_{\max} \simeq3.699\hbar\omega_0. However, at tPavgmax⁡t_{P_{\mathrm{avg}}^{\max}}, the population is already strongly redistributed toward the high-energy values, with p+2(tPavgmax⁡)≃0.407p_{+2}(t_{P_{\mathrm{avg}}^{\max}}) \simeq0.407 and a total population in E/ℏω0=+1E/\hbar\omega_0 = +1 and E/ℏω0=+2E/\hbar\omega_0 = +2 of approximately 0.817. This faster redistribution explains why the average power increases to Pavgmax⁡≃2.228ℏω0ΩP_{\mathrm{avg}}^{\max} \simeq2.228\hbar\omega_0\Omega, even though the battery does not reach full charging.

For J=2ΩJ = 2\Omega, the same distinction between ergotropy and power remains valid. Again, (Δ,D)=(0,0)(\Delta,D) = (0,0) gives rise to a full charging process (Wabs=4ℏω0W_{\mathrm{abs}} = 4\hbar\omega_0), where at time ΩtWmax⁡≃3.142\Omega t_{W_{\max}} \simeq3.142, the population is entirely localized with respect to E/ℏω0=+2E/\hbar\omega_0 = +2, with p+2(tWmax⁡)=1p_{+2}(t_{W_{\max}}) = 1. However, for J=1ΩJ = 1\Omega, the maximum power associated with (Δ,D)=(0,0)(\Delta,D) = (0,0) appears before full charging. At ΩtPavgmax⁡≃0.548\Omega t_{P_{\mathrm{avg}}^{\max}} \simeq0.548, the population of the fully charged state remains low, even though the power already reaches Pavgmax⁡≃2.999ℏω0ΩP_{\mathrm{avg}}^{\max} \simeq2.999\hbar\omega_0\Omega. This confirms that the average power is controlled by the rapidity of the transfer to excited levels and not by the final full charging.

Hence, one can conclude that the absolute ergotropy bound is reached at (Δ,D)=(0,0)(\Delta,D) = (0,0), where the population is entirely transferred to E/ℏω0=+2E/\hbar\omega_0 = +2, which corresponds to a complete occupation of the state ∣+1,+1⟩\lvert+1,+1\rangle. However, the parameters maximizing PavgP_{\mathrm{avg}} favor a faster redistribution of the population toward the high-energy values, namely E/ℏω0=+1E/\hbar\omega_0 = +1 and E/ℏω0=+2E/\hbar\omega_0 = +2 without complete occupation of the state ∣+1,+1⟩\lvert+1,+1\rangle. For J=1ΩJ = 1\Omega, this acceleration is mainly induced by the longitudinal anisotropy Δ\Delta in the absence of DM interaction. For J=2ΩJ = 2\Omega, it results from the joint action of Δ\Delta and a finite value of DD. Thus, in the strong-coupling regime, the DM interaction contributes constructively to accelerate the redistribution of populations toward high-energy levels, which increases average power without leading to fully charging the battery.

Microscopic mechanism of the DM interaction: comparison between the Ising and XZ models

To clarify the microscopic origin of the different DM interaction roles in the Ising and XZ models, we analyze the relevant selection rules and symmetry properties, together with the restructuring of the spectrum and the dressed charger couplings. The parity matrices, detailed analytical derivations and resolved levels are reported in Appendix B. Using the spin-1 ladder operators, the DM interaction can be written as

HD=iℏD2(S+(1)S−(2)−S−(1)S+(2)),(25)H_D = \frac{i\hbar D}{2}\left(S_{+}^{(1)}S_{-}^{(2)} - S_{-}^{(1)}S_{+}^{(2)}\right), \tag*{(25)}

which leads to the selection rule

Δm1=−Δm2=±1,ΔMz=0.(26)\Delta m_1 = -\Delta m_2 = \pm1,\qquad\Delta M_z = 0. \tag*{(26)}

It follows that [HD,Mz]=[HD,H0]=0[H_D,M_z] = [H_D,H_0] = 0, with Mz=Sz(1)+Sz(2)M_z = S_z^{(1)} + S_z^{(2)} and H0=ℏω0MzH_0 = \hbar\omega_0 M_z. Therefore, the DM interaction transfers the spin projection between the two qutrits while conserving the total magnetization. These bare selection rules are identical in the Ising and XZ models. Moreover, two symmetry properties are relevant. Under exchange of the two cells,

P12HDP12†=−HD,(27)\mathcal{P}_{12}H_D\mathcal{P}_{12}^{\dagger}=-H_D, \tag*{(27)}

whereas the transverse and longitudinal terms are even under exchange. Thus, the DM interaction can couple states belonging to opposite exchange-symmetry sectors. Moreover, the interaction Hamiltonian conserves the magnetization parity,

Πz=eiπMz,[Hint,Πz]=0.(28)\Pi_z=e^{i\pi M_z},\qquad[H_{\mathrm{int}},\Pi_z]=0. \tag*{(28)}

However, the transverse charger reverses this parity, ΠzHchΠz†=−Hch\Pi_zH_{\mathrm{ch}}\Pi_z^\dagger=-H_{\mathrm{ch}}. These selection rules and symmetry properties determine the allowed pathways, but, being common to both regimes, they are not sufficient by themselves to explain the opposite effect of DM on the power. The discriminating element is the longitudinal anisotropy HΔ=ℏJΔSz(1)Sz(2)H_\Delta=\hbar J\Delta S_z^{(1)}S_z^{(2)}. It does not create new bare DM channels, but modifies the relative energies of the configurations already coupled by DM. For example, HΔ∣0,0⟩=0H_\Delta|0,0\rangle=0, whereas HΔ∣±1,∓1⟩=−ℏJΔ∣±1,∓1⟩H_\Delta|\pm1,\mp1\rangle=-\hbar J\Delta|\pm1,\mp1\rangle. At fixed JJ and DD, the corresponding off-diagonal elements remain ℏ(J±iD)\hbar(J\pm iD), with modulus ℏJ2+D2\hbar\sqrt{J^2+D^2}.

This difference appears directly in the exact spectrum of the interaction. The four eigenvalues of the odd sector are

Es,η(o)=ℏ(sJ+ηJ2+D2),s,η=±1,(29)E_{s,\eta}^{(o)}=\hbar\left(sJ+\eta\sqrt{J^2+D^2}\right),\qquad s,\eta=\pm1, \tag*{(29)}

which are independent of Δ\Delta. In the Ising limit (Δ=0\Delta=0), the spectrum of the even sector is

spec⁡(Hint(e)/ℏ)={0,0,0,±4J2+2D2}.(30)\operatorname{spec}\left(H_{\mathrm{int}}^{(e)}/\hbar\right)=\left\{0,0,0,\pm\sqrt{4J^2+2D^2}\right\}. \tag*{(30)}

In the XZ regime, two even eigenvalues are E=±ℏJΔE=\pm\hbar J\Delta, whereas the other three eigenvalues satisfy

λ3−[(JΔ)2+4J2+2D2]λ+2JΔD2=0.(31)\lambda^3-\left[(J\Delta)^2+4J^2+2D^2\right]\lambda+2J\Delta D^2=0. \tag*{(31)}

Therefore, the mixed term 2JΔD22J\Delta D^2 highlights the joint spectral action of DM and the longitudinal anisotropy. The continuous restructuring of the nine interaction-energy levels as a function of DD is shown in Figure 8(a,Figure 8b), whereas the representative resolved levels and their multiplicities are reported in Figure 9 of the Appendix § 1.3.3.

Figure 8. (a,b) Interaction energy spectra for the Ising (Δ=0\Delta=0) and XZ (Δ=1.98\Delta=1.98) models as a function of DD. The nine eigenvalues are ordered in ascending energy and expressed in units of ℏJ\hbar J. (c–f) Charger coupling weights between opposite-parity sectors at D=0D=0 (c,d) and D=2.36D=2.36 (e,f). Each bar sums KabK_{ab} over equal energy gaps; the total weight is six per panel. Energies and gaps are in units of ℏJ\hbar J, with J=2ΩJ=2\Omega.

Figure 9. Nine interaction energy levels for the Ising model and the XZ model with Δ=1.98\Delta= 1.98, at D=0D = 0 and D=2.36D = 2.36. Each horizontal segment represents one eigenvalue, including degenerate levels. Energies are expressed in units of ℏJ\hbar J, with J=2ΩJ = 2\Omega.

In the Ising limit, the interaction Hamiltonian also satisfies the spectral anticommutation relation QHintQ†=−HintQH_{\mathrm{int}}Q^\dagger=-H_{\mathrm{int}}, with Q=eiπSz(2)Q=e^{i\pi S_z^{(2)}}, which imposes the pairing E↔−EE\leftrightarrow-E. For Δ≠0\Delta\ne0, QHintQ†=−Hint+2HΔQH_{\mathrm{int}}Q^\dagger=-H_{\mathrm{int}}+2H_\Delta and this exact spectral symmetry of the Ising regime is lost.

The same restructuring modifies the charger transition matrix elements of the charger in the dressed basis. If Hint∣ϕa(e)⟩=ϵa(e)∣ϕa(e)⟩H_{\mathrm{int}}|\phi_a^{(e)}\rangle=\epsilon_a^{(e)}|\phi_a^{(e)}\rangle and Hint∣ϕb(e)⟩=ϵb(e)∣ϕb(e)⟩H_{\mathrm{int}}|\phi_b^{(e)}\rangle=\epsilon_b^{(e)}|\phi_b^{(e)}\rangle, the relevant amplitudes are

Mab(Δ,D)=⟨ϕa(e)(Δ,D)∣Hch∣ϕb(o)(Δ,D)⟩.(32)M_{ab}(\Delta,D)=\langle\phi_a^{(e)}(\Delta,D)|H_{\mathrm{ch}}|\phi_b^{(o)}(\Delta,D)\rangle. \tag*{(32)}

Note that although HchH_{\mathrm{ch}} itself is independent of DD and Δ\Delta, the eigenvectors of the interaction are not (more details are given in Appendix B.

Conclusion

In this work, we have investigated the charging dynamics of a two-qubit quantum battery composed of two spin-1 systems with anisotropic exchange interactions as well as a DM interaction. Using ergotropy and average power as indicators of charging performance, we have analyzed the parallel-charging regime, the Ising model, and the anisotropic XZ model to highlight the roles of the exchange coupling, chiral interaction, and longitudinal anisotropy in the charging performance of the proposed quantum battery. First, we have derived an exact analytical expression for ergotropy in the parallel-charging regime, whose validity is confirmed by our numerical results. Then, our analysis highlighted the advantage of collective charging over parallel charging, which accelerates the charging dynamics and increases the maximum average power. In the Ising model, we showed that strengthening the exchange coupling improves battery performance, whereas the DM interaction degrades it. This enhancement is accompanied by increases in global coherence and mutual information, whereas the DM interaction degrades these two quantum resources. Because the DM interaction degrades quantum thermodynamic quantities, coherence, and quantum correlations, we have extended our study to the anisotropic XZ model to examine the joint effect of the longitudinal anisotropy and the DM interaction. In this model, we showed that the joint action of the longitudinal anisotropy and the chiral interaction generally increases the maximum average power.

Beyond the dimensional extension from qubits to qutrits, the present results connected the local spin-1 transition structure to the charging performance of the battery. For the parallel chargers considered here, normalized to the same operator norm, the two-qutrit and two-qubit batteries reached the same absolute ergotropy bound. However, the two-qutrit battery exhibited a first full-charging time twice as long and a maximum average power reduced by a factor of two. This difference reflected the transition structure of the chargers: the spin-1 transverse charger couples adjacent levels through the intermediate state, whereas the Pauli charger connected directly the two qubit levels. Besides, the conditions favoring the maximum reachable ergotropy do not necessarily coincide with those maximizing the average power. The population analysis showed that the former is associated with complete transfer to the fully charged state. The latter can result from a faster partial redistribution toward high-energy sectors.

Importantly, the XZ model is considered a separate limit between the conditions that optimize the maximum reachable ergotropy associated with the Ising model and those optimizing the maximum average power. In fact, according to the population analysis, we showed that the maximum ergotropy is obtained when the population is completely transferred to the fully charged state. However, the maximum average power results from a partial faster transfer of the population toward the high-energy levels. Thus, the optimization of extractable work and that of charging speed rely on two distinct physical mechanisms that do not necessarily coincide. Therefore, we have concluded that the type Heisenberg interaction is essential for the enhancement of the charging process of two-qutrit quantum batteries. Indeed, we showed that when the objective is to maximize the ergotropy, the Ising model without DM interaction is the most appropriate choice. However, when the priority is to maximize the average power, the anisotropic XZ model offered greater advantage by allowing the DM interaction to be exploited constructively through the joint control of the exchange coupling, longitudinal anisotropy and chiral interaction.

Analytical Derivation of the Exact Expression of TQQB Ergotropy in the Parallel-Charging Regime

In this appendix, we derive the expression of TQQB ergotropy analytically in the parallel-charging case. Indeed, the charging Hamiltonian is given as

Hch=ℏΩ(Sx(1)+Sx(2)).(33)H_{\mathrm{ch}}=\hbar\Omega\left(S_x^{(1)}+S_x^{(2)}\right). \tag*{(33)}

While the free Hamiltonian of the battery is written as below:

H0=ℏω(Sz(1)+Sz(2)).(34)H_0=\hbar\omega\left(S_z^{(1)}+S_z^{(2)}\right). \tag*{(34)}

Initially, we consider that the state of the TQQB is prepared in the ground state as

ρ(0)=∣−1,−1⟩⟨−1,−1∣.(35)\rho(0)=\lvert-1,-1\rangle\langle-1,-1\rvert. \tag*{(35)}

Now, to calculate the average energy of the quantum battery, we adopt the Heisenberg representation where the time evolution of the operator Sz(1)S_z^{(1)} is given as follows:

Sz(1)(t)=eiℏHchtSz(1)e−iℏHcht.(36)S_z^{(1)}(t)=e^{\frac{i}{\hbar}H_{\mathrm{ch}}t}S_z^{(1)}e^{-\frac{i}{\hbar}H_{\mathrm{ch}}t}. \tag*{(36)}

Substituting HchH_{\mathrm{ch}}, we obtain

Sz(1)(t)=eiΩt(Sx(1)+Sx(2))Sz(1)e−iΩt(Sx(1)+Sx(2)).(37)S_z^{(1)}(t)=e^{i\Omega t(S_x^{(1)}+S_x^{(2)})}S_z^{(1)}e^{-i\Omega t(S_x^{(1)}+S_x^{(2)})}. \tag*{(37)}

Since the operators Sx(1)S_x^{(1)} and Sx(2)S_x^{(2)} act on two distinct qutrits of two different Hilbert spaces. Then, they commute, namely [Sx(1),Sx(2)]=0\left[S_x^{(1)},S_x^{(2)}\right]=0. Therefore, the unity operator can be factorized in the following form:

U(t)=e−iΩt(Sx(1)+Sx(2))=e−iΩtSx(1)e−iΩtSx(2).(38)U(t)=e^{-i\Omega t(S_x^{(1)}+S_x^{(2)})}=e^{-i\Omega tS_x^{(1)}}e^{-i\Omega tS_x^{(2)}}. \tag*{(38)}

According to Eq. (37), one can straightforwardly obtain:

Sz(1)(t)=eiΩtSx(1)eiΩtSx(2)Sz(1)e−iΩtSx(2)e−iΩtSx(1).(39)S_z^{(1)}(t)=e^{i\Omega tS_x^{(1)}}e^{i\Omega tS_x^{(2)}}S_z^{(1)}e^{-i\Omega tS_x^{(2)}}e^{-i\Omega tS_x^{(1)}}. \tag*{(39)}

Furthermore, since Sx(2)S_x^{(2)} and Sz(1)S_z^{(1)} live in two different Hilbert spaces, then [Sx(2),Sz(1)]=0\left[S_x^{(2)},S_z^{(1)}\right]=0. Therefore, one can deduce:

eiΩtSx(2)Sz(1)e−iΩtSx(2)=Sz(1).(40)e^{i\Omega tS_x^{(2)}}S_z^{(1)}e^{-i\Omega tS_x^{(2)}}=S_z^{(1)}. \tag*{(40)}

This leads to

Sz(1)(t)=eiΩtSx(1)Sz(1)e−iΩtSx(1).(41)S_z^{(1)}(t)=e^{i\Omega tS_x^{(1)}}S_z^{(1)}e^{-i\Omega tS_x^{(1)}}. \tag*{(41)}

However, since the spin operators satisfy [Si,Sj]=iεijkSk\left[S_i,S_j\right]=i\varepsilon_{ijk}S_k, we obtain [Sx,Sz]=−iSy\left[S_x,S_z\right]=-iS_y and [Sx,[Sx,Sz]]=Sz\left[S_x,\left[S_x,S_z\right]\right]=S_z. Then, by using the Baker–Campbell–Hausdorff expansion, the operator Sz(t)S_z(t) turns out to be

Sz(t)=Sz+iΩt[Sx,Sz]+(iΩt)22![Sx,[Sx,Sz]]+⋯=Szcos⁡(Ωt)+Sysin⁡(Ωt).(42)\begin{aligned} S_z(t)&=S_z+i\Omega t[S_x,S_z]+\frac{(i\Omega t)^2}{2!}[S_x,[S_x,S_z]]+\cdots\\ &=S_z\cos(\Omega t)+S_y\sin(\Omega t). \tag*{(42)} \end{aligned}

Consequently,

⟨Sz(t)⟩=⟨Sz⟩cos⁡(Ωt)+⟨Sy⟩sin⁡(Ωt).(43)\langle S_z(t)\rangle= \langle S_z\rangle\cos(\Omega t) + \langle S_y\rangle\sin(\Omega t). \tag*{(43)}

Let’s suppose that the qutrit is initially prepared in the state ∣−1⟩\lvert-1\rangle, which is an eigenstate of SzS_z. Then, in the eigenstate basis of SzS_z, we have Sz=diag⁡(1,0,−1)S_z=\operatorname{diag}(1,0,-1). Thus, ⟨Sz⟩=⟨−1∣Sz∣−1⟩=−1\langle S_z\rangle=\langle-1\lvert S_z\lvert-1\rangle=-1. Moreover, the operator SyS_y has no diagonal elements in this basis. Hence, ⟨Sy⟩=⟨−1∣Sy∣−1⟩=0\langle S_y\rangle=\langle-1\lvert S_y\lvert-1\rangle=0. Therefore, it follows that

⟨Sz(t)⟩=−cos⁡(Ωt).(44)\langle S_z(t)\rangle=-\cos(\Omega t). \tag*{(44)}

Since the two qutrits evolve independently, the average energy of the battery is

⟨H0(t)⟩=ℏω0[⟨Sz(1)(t)⟩+⟨Sz(2)(t)⟩]=−2ℏω0cos⁡(Ωt).(45)\begin{aligned} \langle H_0(t)\rangle&= \hbar\omega_0\left[\langle S_z^{(1)}(t)\rangle+\langle S_z^{(2)}(t)\rangle\right]\\ &=-2\hbar\omega_0\cos(\Omega t). \tag*{(45)} \end{aligned}

However, at the initial time, we have ⟨H0(0)⟩=−2ℏω0\langle H_0(0)\rangle=-2\hbar\omega_0. As mentioned before, the ergotropy is defined as the maximum work extractable through unitary transformations as:

W(t)=Tr⁡[ρ(t)H0]−min⁡UTr⁡[Uρ(t)U†H0].(46)\mathcal{W}(t)=\operatorname{Tr}[\rho(t)H_0]-\min_U\operatorname{Tr}[U\rho(t)U^\dagger H_0]. \tag*{(46)}

In the present case, the evolution generated by HchH_{\mathrm{ch}} is unitary. Consequently, the state ρ(t)\rho(t) remains pure at all times. Therefore, the associated passive state is obtained by placing the unit population in the lowest energy level of H0H_0, namely the state ∣−1,−1⟩\lvert-1,-1\rangle. Hence, the corresponding passive energy is given as

Ep=−2ℏω0.(47)E_{\mathrm{p}}=-2\hbar\omega_0. \tag*{(47)}

The ergotropy is then written as

W(t)=⟨H0(t)⟩−Ep=−2ℏω0cos⁡(Ωt)+2ℏω0.(48)\begin{aligned} \mathcal{W}(t)&=\langle H_0(t)\rangle-E_{\mathrm{p}}\\ &=-2\hbar\omega_0\cos(\Omega t)+2\hbar\omega_0. \tag*{(48)} \end{aligned}

Thus,

W(t)=2ℏω0[1−cos⁡(Ωt)].(49)\mathcal{W}(t)=2\hbar\omega_0[1-\cos(\Omega t)]. \tag*{(49)}

By using the trigonometric identity, 1−cos⁡(Ωt)=2sin⁡2(Ωt2)1-\cos(\Omega t)=2\sin^2\left(\frac{\Omega t}{2}\right), one can obtain the TQQB ergotropy in the parallel-charging regime as follows:

W∥qtr(t)=4ℏω0sin⁡2(Ωt2).(50)\mathcal{W}_{\parallel}^{\mathrm{qtr}}(t)=4\hbar\omega_0\sin^2\left(\frac{\Omega t}{2}\right). \tag*{(50)}

Symmetry decomposition, exact spectrum, and dressed transition matrix elements

This appendix presents the analytical details supporting the microscopic interpretation developed in Sec. VI. Indeed, we examine the selection rules and symmetry properties of the DM interaction, the exact decomposition of the interaction Hamiltonian according to magnetization parity, its nine-level spectrum and the charger transition elements in the interaction eigenbasis.

First, let decompose the interaction Hamiltonian as

Hint=Hx+HD+HΔ,(51)H_{\mathrm{int}}=H_x+H_D+H_\Delta, \tag*{(51)}

where

Hx=2ℏJSx(1)Sx(2),HD=ℏD(Sx(1)Sy(2)−Sy(1)Sx(2)),HΔ=ℏJΔSz(1)Sz(2).(52)\begin{aligned} H_x&=2\hbar J S_x^{(1)}S_x^{(2)},\\ H_D&=\hbar D\left(S_x^{(1)}S_y^{(2)}-S_y^{(1)}S_x^{(2)}\right),\\ H_\Delta&=\hbar J\Delta S_z^{(1)}S_z^{(2)}. \tag*{(52)} \end{aligned}

To simplify the notation, we introduce

q=JΔ,r=J2+D2.(53)q=J\Delta,\qquad r=\sqrt{J^2+D^2}. \tag*{(53)}

Selection rules and symmetry properties

Using the spin-1 ladder operators, the DM term can be written as

HD=iℏD2(S+(1)S−(2)−S−(1)S+(2)).(54)H_D=\frac{i\hbar D}{2}\left(S_+^{(1)}S_-^{(2)}-S_-^{(1)}S_+^{(2)}\right). \tag*{(54)}

Therefore, any nonzero DM matrix element obeys the following selection rule

Δm1=−Δm2=±1,ΔMz=0,(55)\Delta m_1=-\Delta m_2=\pm1,\qquad\Delta M_z=0, \tag*{(55)}

where Mz=Sz(1)+Sz(2)M_z=S_z^{(1)}+S_z^{(2)}. It follows that

[HD,Mz]=0,[HD,H0]=0,H0=ℏω0Mz.(56)[H_D,M_z]=0,\qquad[H_D,H_0]=0,\qquad H_0=\hbar\omega_0M_z. \tag*{(56)}

Thus, the DM interaction transfers spin projection between the two qutrits, which conserves the total magnetization. In particular,

HD∣−1,−1⟩=HD∣+1,+1⟩=0.(57)H_D\lvert-1,-1\rangle=H_D\lvert+1,+1\rangle=0. \tag*{(57)}

Moreover, the longitudinal term HΔH_\Delta conserves MzM_z. However, the transverse term HxH_x allows processes with ΔMz=0,±2\Delta M_z=0,\pm2. Hence, the interaction Hamiltonian conserves magnetization parity as

Πz=eiπMz,[Hint,Πz]=0.(58)\Pi_z=e^{i\pi M_z},\qquad[H_{\mathrm{int}},\Pi_z]=0. \tag*{(58)}

A distinct symmetry property concerns the exchange of the two cells. If P12\mathcal{P}_{12} denotes the permutation operator, we have

P12∣m1,m2⟩=∣m2,m1⟩.(59)\mathcal{P}_{12}\lvert m_1,m_2\rangle=\lvert m_2,m_1\rangle. \tag*{(59)}

Then, HxH_x and HΔH_\Delta are even under exchange, whereas

P12HDP12†=−HD.(60)\mathcal{P}_{12}H_D\mathcal{P}_{12}^{\dagger}=-H_D. \tag*{(60)}

Hence, the DM term is odd under exchange of the two cells and can couple components belonging to opposite exchange-symmetry sectors. For example, defining

∣a0⟩=∣+1,−1⟩−∣−1,+1⟩2(61)|a_0\rangle= \frac{|+1,-1\rangle-|-1,+1\rangle}{\sqrt{2}} \tag*{(61)}

one can obtain the following result:

HD∣0,0⟩=i2ℏD∣a0⟩.(62)H_D|0,0\rangle=i\sqrt{2}\hbar D|a_0\rangle. \tag*{(62)}

This property is common to Ising and XZ regimes. It identifies the coupling pathways allowed by DM, but cannot by itself explain the different behavior of the charging power in the two regimes.

Decomposition according to magnetization parity

The commutator [Hint,Πz]=0[H_{\mathrm{int}},\Pi_z]=0 allows the Hilbert space of the two qutrits to be decomposed exactly into an even sector of dimension five and an odd sector of dimension four. In the ordered even basis Be=(∣+1,+1⟩,∣+1,−1⟩,∣0,0⟩,∣−1,+1⟩,∣−1,−1⟩)\mathcal{B}_e=(|+1,+1\rangle,|+1,-1\rangle,|0,0\rangle,|-1,+1\rangle,|-1,-1\rangle), the even block of Hint/ℏH_{\mathrm{int}}/\hbar is

Me=(q0J000−qJ+iD00JJ−iD0J+iDJ00J−iD−q000J0q).(63)M_e= \begin{pmatrix} q & 0 & J & 0 & 0\\ 0 & -q & J+iD & 0 & 0\\ J & J-iD & 0 & J+iD & J\\ 0 & 0 & J-iD & -q & 0\\ 0 & 0 & J & 0 & q \end{pmatrix}. \tag*{(63)}

In the ordered odd basis Bo=(∣+1,0⟩,∣0,+1⟩,∣0,−1⟩,∣−1,0⟩)\mathcal{B}_o=(|+1,0\rangle,|0,+1\rangle,|0,-1\rangle,|-1,0\rangle), the odd block is

Mo=(0J+iDJ0J−iD00JJ00J+iD0JJ−iD0).(64)M_o= \begin{pmatrix} 0 & J+iD & J & 0\\ J-iD & 0 & 0 & J\\ J & 0 & 0 & J+iD\\ 0 & J & J-iD & 0 \end{pmatrix}. \tag*{(64)}

The absence of q=JΔq=J\Delta from MoM_o shows immediately that the longitudinal anisotropy does not modify the odd block, whereas it selectively restructures the even sector.

Exact spectrum in the Ising and XZ regimes

The characteristic polynomial of the odd sector factorizes as

det⁡(λI−Mo)=[λ2−(r+J)2][λ2−(r−J)2].(65)\det(\lambda I-M_o)=[\lambda^2-(r+J)^2][\lambda^2-(r-J)^2]. \tag*{(65)}

Therefore, the four odd-sector eigenvalues are

Es,η(o)=ℏ(sJ+ηJ2+D2),s,η=±1.(66)E^{(o)}_{s,\eta}=\hbar\left(sJ+\eta\sqrt{J^2+D^2}\right),\qquad s,\eta=\pm1. \tag*{(66)}

They are independent of Δ\Delta. The difference between the Ising and XZ models appears in the even sector.

In the Ising limit (Δ=0\Delta=0), the even-sector spectrum is exactly

spec⁡(Hint(e)/ℏ)={0,0,0,±4J2+2D2}.(67)\operatorname{spec}\left(H_{\mathrm{int}}^{(e)}/\hbar\right)=\left\{0,0,0,\pm\sqrt{4J^2+2D^2}\right\}. \tag*{(67)}

The DM interaction shifts the two nonzero branches, whereas three even-sector eigenvalues remain exactly zero as DD varies.

In the XZ regime, two eigenvalues of the even sector are

E=±ℏJΔ,(68)E=\pm\hbar J\Delta, \tag*{(68)}

whereas the other three are given by E=ℏλE=\hbar\lambda, where

λ3−[(JΔ)2+4J2+2D2]λ+2JΔD2=0.(69)\lambda^3-\left[(J\Delta)^2+4J^2+2D^2\right]\lambda+2J\Delta D^2=0. \tag*{(69)}

The mixed term 2JΔD22J\Delta D^2 explicitly shows that, in the even sector, the spectral action of DM depends on the longitudinal anisotropy. The same DM operator acts in different spectral environments in the Ising and XZ models. This continuous restructuring is shown in Figure 8(a,Figure 8b), whereas the nine resolved levels and their multiplicities for representative parameter sets are shown in Figure 9.

In the Ising limit, an additional spectral symmetry is satisfied,

QHintQ†=−Hint,Q=eiπSz(1).(70)QH_{\mathrm{int}}Q^\dagger=-H_{\mathrm{int}},\qquad Q=e^{i\pi S_z^{(1)}}. \tag*{(70)}

This imposes the spectral pairing E↔−EE\leftrightarrow-E, but it does not constitute a conservation law. For Δ≠0\Delta\ne0,

QHintQ†=−Hint+2HΔ,(71)QH_{\mathrm{int}}Q^\dagger=-H_{\mathrm{int}}+2H_\Delta, \tag*{(71)}

and the exact spectral symmetry of the Ising regime is lost.

Bare matrix elements and dressed transition elements

The longitudinal anisotropy modifies neither the DM selection rules nor its bare matrix elements. For example,

⟨+1,−1∣HD∣0,0⟩=iℏD,\langle+1,-1|H_D|0,0\rangle=i\hbar D,
⟨−1,+1∣HD∣0,0⟩=−iℏD.(72)\langle-1,+1|H_D|0,0\rangle=-i\hbar D. \tag*{(72)}

When the transverse exchange interaction is included,

⟨+1,−1∣Hint∣0,0⟩=ℏ(J+iD),\langle+1,-1|H_{\mathrm{int}}|0,0\rangle=\hbar(J+iD),
⟨−1,+1∣Hint∣0,0⟩=ℏ(J−iD),(73)\langle-1,+1|H_{\mathrm{int}}|0,0\rangle=\hbar(J-iD), \tag*{(73)}

and their modulus, ℏJ2+D2\hbar\sqrt{J^2+D^2}, remains independent of Δ\Delta. The Ising–XZ difference does not originate from new bare DM channels.

In contrast, the eigenstates of HintH_{\mathrm{int}} depend on the interaction parameters. Thus, even though HDH_D does not explicitly contain Δ\Delta, its matrix elements in the dressed basis,

Dμνdr=⟨ϕμ(Δ,D)∣HD∣ϕν(Δ,D)⟩,(74)D_{\mu\nu}^{\mathrm{dr}}=\langle\phi_\mu(\Delta,D)|H_D|\phi_\nu(\Delta,D)\rangle, \tag*{(74)}

acquire a dependence on Δ\Delta through the eigenvectors. The transverse charger

Hch=ℏΩ(Sx(1)+Sx(2))(75)H_{\mathrm{ch}}=\hbar\Omega\left(S_x^{(1)}+S_x^{(2)}\right) \tag*{(75)}

reverses the magnetization parity,

ΠzHchΠz†=−Hch,(76)\Pi_z H_{\mathrm{ch}}\Pi_z^\dagger=-H_{\mathrm{ch}}, \tag*{(76)}

and therefore couples the even and odd sectors. In the bases BeB_e and BoB_o, its nonzero block is (Hch)oe=(ℏΩ/2)L(H_{\mathrm{ch}})_{oe}=(\hbar\Omega/\sqrt{2})L, where

L=(11100101100110100111).(77)L=\begin{pmatrix} 1 & 1 & 1 & 0 & 0\\ 1 & 0 & 1 & 1 & 0\\ 0 & 1 & 1 & 0 & 1\\ 0 & 0 & 1 & 1 & 1 \end{pmatrix}. \tag*{(77)}

Let UeU_e and UoU_o denote the unitary matrices containing the eigenvectors of the even and odd sectors, respectively. In the eigenbasis of HintH_{\mathrm{int}}, the charger transition matrix becomes

T=ℏΩ2Uo†LUe.(78)T=\frac{\hbar\Omega}{\sqrt{2}}U_o^\dagger L U_e. \tag*{(78)}

The bare matrix LL is independent of the interaction parameters. However, Ue=Ue(J,D,Δ)U_e=U_e(J,D,\Delta), whereas Uo=Uo(J,D)U_o=U_o(J,D). The longitudinal anisotropy therefore modifies the dressed charger transition matrix elements by restructuring the eigenstates of the even sector.

Invariant coupling weights

When energy levels are degenerate, the individual eigenvectors are not unique. To obtain a quantity that is independent of the basis choice within each degenerate eigenspace, we define

Kab=∥Pa(e)HchPb(o)∥F2ℏ2Ω2,(79)K_{ab}=\frac{\left\|P_a^{(e)}H_{\mathrm{ch}}P_b^{(o)}\right\|_F^2}{\hbar^2\Omega^2}, \tag*{(79)}

where Pa(e)P_a^{(e)} and Pb(o)P_b^{(o)} are the spectral projectors onto the complete eigenspaces of opposite parities. For nondegenerate levels,

Kab=∣⟨ϕa(e)|Hch|ϕb(o)⟩∣2ℏ2Ω2.(80)K_{ab}=\frac{\left|\left\langle\phi_a^{(e)}\middle|H_{\mathrm{ch}}\middle|\phi_b^{(o)}\right\rangle\right|^2}{\hbar^2\Omega^2}. \tag*{(80)}

These weights satisfy the exact sum rule

∑a,bKab=6.(81)\sum_{a,b}K_{ab}=6. \tag*{(81)}

In Figure 8 (c–Figure 8f), the weights are grouped according to the corresponding absolute interaction-energy gaps. For J/Ω=2J/\Omega=2 and D/Ω=2.36D/\Omega=2.36, corresponding to panels (e,f), six distinct gap groups carry nonzero weight in the Ising and DM case, whereas eight are resolved in the XZ and DM case at Δ=1.98\Delta=1.98. Since the sum in Eq. (81) remains equal to six, this difference reflects a redistribution of the same total charger coupling weight among different spectral gaps, rather than an increase in its overall strength.

Connection with the coherent charging dynamics

The connection between the spectral restructuring and the dynamics appears directly in the interaction picture generated by HintH_{\mathrm{int}}:

HchI(t)=∑a,bei(ϵa(e)−ϵb(o))t/ℏPa(e)HchPb(o)+H.c.(82)H_{\mathrm{ch}}^{I}(t)=\sum_{a,b}e^{i(\epsilon_a^{(e)}-\epsilon_b^{(o)})t/\hbar}P_a^{(e)}H_{\mathrm{ch}}P_b^{(o)}+\mathrm{H.c.} \tag*{(82)}

This expression shows that the dynamics depends simultaneously on the energy gaps, the complex transition amplitudes, the relative dynamical phases, and the projections of the initial state. Consequently, neither the number of distinct gaps, nor the breaking of a spectral symmetry, nor an isolated matrix element is sufficient by itself to determine the sign of the change in charging power.

Hence, we conclude that microscopic mechanism can be summarized as follows: the symmetries and selection rules allowed by DM interaction. Whereas, the longitudinal anisotropy modifies the spectral environment and the eigenstates. This restructuring redistributes the dressed transition elements of the charger and modifies the dynamical phases, leading to different coherent dynamics in the Ising and XZ models.

Spectral structure of the interaction Hamiltonian

The nine interaction-energy levels corresponding to the representative parameters considered in the manuscript are shown in Figure 9. This representation makes the degeneracies and their modification in the presence of the DM interaction explicit.

Numerical implementation and reproducibility

All numerical calculations are performed with Python, NumPy, and SciPy in the complete two-qutrit Hilbert space H=C3⊗C3\mathcal{H}=\mathbb{C}^{3}\otimes\mathbb{C}^{3}, with dim⁡H=9\dim\mathcal{H}=9, without Hilbert-space truncation. The initial state is ∣ψ0⟩=∣−1,−1⟩\lvert\psi_0\rangle=\lvert-1,-1\rangle. For each fixed parameter set, the time-independent dynamical Hamiltonian Hdyn=Hch+HintH_{\mathrm{dyn}}=H_{\mathrm{ch}}+H_{\mathrm{int}} is diagonalized once using scipy.linalg.eigh,

Hdyn=Vdiag⁡(ε1,…,ε9)V†,c(0)=V†∣ψ0⟩.(83)H_{\mathrm{dyn}}=V\operatorname{diag}(\varepsilon_1,\ldots,\varepsilon_9)V^\dagger,\qquad c(0)=V^\dagger\lvert\psi_0\rangle. \tag*{(83)}

The state is then propagated spectrally as

cn(tk)=e−iεntk/ℏcn(0),∣ψ(tk)⟩=Vc(tk).(84)c_n(t_k) = e^{-i\varepsilon_n t_k/\hbar}c_n(0), \qquad|\psi(t_k)\rangle= Vc(t_k). \tag*{(84)}

Using H0(e)=V†H0VH_0^{(e)} = V^\dagger H_0 V, the stored energy is evaluated from E(tk)=c†(tk)H0(e)c(tk)E(t_k) = c^\dagger(t_k)H_0^{(e)}c(t_k). Thus, the temporal grid samples the observables and their extrema rather than numerically integrating the Schrödinger equation step by step. The storage Hamiltonian H0H_0 is not included in HdynH_{\mathrm{dyn}}.

The numerical implementation uses ℏ=ω0=Ω=1\hbar= \omega_0 = \Omega= 1. Time, ergotropy, and average power are reported through τ=Ωt\tau= \Omega t, W/(ℏω0)W/(\hbar\omega_0), and Pavg/(ℏω0Ω)P_{\mathrm{avg}}/(\hbar\omega_0\Omega), respectively. Accordingly, JJ and DD are expressed in units of Ω\Omega, whereas Δ\Delta is dimensionless. The dynamical curves are evaluated at 2 000 uniformly spaced times over τ∈[10−6,π+0.05]\tau\in[10^{-6}, \pi+ 0.05].

The maximum reachable ergotropy and maximum average power are extracted independently using numpy.argmax,

kW=arg max⁡k W(tk),kP=arg max⁡k Pavg(tk),(85)k_W = \underset{k}{\operatorname{arg\,max}}\ W(t_k), \qquad k_P = \underset{k}{\operatorname{arg\,max}}\ P_{\mathrm{avg}}(t_k), \tag*{(85)}

with the corresponding maximum values obtained by array indexing or numpy.max. In particular, Pavgmax⁡P_{\mathrm{avg}}^{\max} is determined directly from Pavg(tk)=W(tk)/tkP_{\mathrm{avg}}(t_k) = W(t_k)/t_k and is not obtained from Wmax⁡/tWmax⁡W_{\max}/t_{W_{\max}}.

Declaration of Interest

The authors declare that they have no conflict of interest.

Data availability statement

No data statement is available.

References

References

  1. [1]R. Kosloff, Quantum Thermodynamics: A Dynamical Viewpoint, Entropy 15, 2100 (2013).DOI
  2. [2]S. Vinjanampathy and J. Anders, Quantum thermodynamics, Contemporary Physics 57, 545 (2016).DOI
  3. [3]M. H. Ben Abdou Chakour, K. El Anouz, Y. Hassouni, and A. El Allati, Improving different thermodynamic quantities in non-Markovian regime, Quantum Inf. Process. 24, 211 (2025).
  4. [4]J. Goold, M. Huber, A. Riera, L. del Rio, and P. Skrzypczyk, The role of quantum information in thermodynamics—a topical review, J. Phys. A: Math. Theor. 49, 143001 (2016).DOI
  5. [5]Y. Khlifi, K. El Anouz, and A. El Allati, Operational diagnosis of an autonomous quantum thermal machine via non-Markovianity and heat currents, Opt. Laser Technol. 201, 115246 (2026).
  6. [6]A. Khoudiri, A. Oularabi, K. El Anouz, İ. Demir, and A. El Allati, Coherence-driven quantum battery charging via autonomous thermal machines: Energy transfer, memory effects, and ergotropy enhancement, Phys. Lett. A 565, 131178 (2026).
  7. [7]F. Campaioli, S. Gherardini, J. Q. Quach, M. Polini, and G. M. Andolina, Colloquium: Quantum batteries, Rev. Mod. Phys. 96, 031001 (2024).DOI
  8. [8]A. Khoudiri, A. El Allati, Y. Khlifi, K. El Anouz, and Ö. E. Müste­kaplıoğlu, Correlation enhanced autonomous quantum battery charging via structured reservoirs, Phys. Rev. E 114, 034116 (2026), arXiv:2508.18086 [quant-ph].DOI
  9. [9]A. Lemou, A. El Allati, K. El Anouz, A. Ali, S. Al-Kuwari, and A. Belfakir, Quantum speed limit time in thermodynamic contexts: Linking CPTP dynamics and repetitive interaction, AVS Quantum Sci. 8, 022002 (2026).DOI
  10. [10]R. Alicki and M. Fannes, Entanglement boost for extractable work from ensembles of quantum batteries, Phys. Rev. E 87, 042123 (2013).pubmed.ncbi.nlm.nih.gov/23679388
  11. [11]F. C. Binder, S. Vinjanampathy, K. Modi, and J. Goold, Quantacell: Powerful charging of quantum batteries, New J. Phys. 17, 075015 (2015).DOI
  12. [12]F. Campaioli, F. A. Pollock, F. C. Binder, L. Celeri, J. Goold, S. Vinjanampathy, and K. Modi, Enhancing the charging power of quantum batteries, Phys. Rev. Lett. 118, 150601 (2017).DOI
  13. [13]D. Ferraro, M. Campisi, G. M. Andolina, V. Pellegrini, and M. Polini, High-power collective charging of a solid-state quantum battery, Phys. Rev. Lett. 120, 117702 (2018).DOI
  14. [14]T. P. Le, J. Levinsen, K. Modi, M. M. Parish, and F. A. Pollock, Spin-chain model of a many-body quantum battery, Phys. Rev. A 97, 022106 (2018).DOI
  15. [15]A. C. Santos, B. Çakmak, S. Campbell, and N. T. Zinner, Stable adiabatic quantum batteries, Phys. Rev. E 100, 032107 (2019).DOI
  16. [16]D. Rossini, G. M. Andolina, and M. Polini, Many-body localized quantum batteries, Phys. Rev. B 100, 115142 (2019).DOI
  17. [17]D. Rossini, G. M. Andolina, D. Rosa, M. Carrega, and M. Polini, Quantum advantage in the charging process of Sachdev–Ye–Kitaev batteries, Phys. Rev. Lett. 125, 236402 (2020).DOI
  18. [18]A. Oularabi, A. El Allati, and K. El Anouz, Enhancing ergotropy of quantum batteries through coherence and non-Markovianity, Physica A 679, 131003 (2025).DOI
  19. [19]A. E. Allahverdyan, R. Balian, and T. M. Nieuwenhuizen, Maximal work extraction from finite quantum systems, Europhys. Lett. 67, 565 (2004).DOI
  20. [20]X.-L. Zhang, X.-K. Song, and D. Wang, Quantum Battery in the Heisenberg Spin Chain Models with Dzyaloshinskii–Moriya Interaction, Adv. Quantum Technol. 7, 2400114 (2024).DOI
  21. [21]S. Bhattacharya, V. B. Sabale, and A. Kumar, Heisenberg spin networks for realizing quantum battery with the aid of Dzyaloshinskii–Moriya interaction, New J. Phys. 28, 014508 (2026).DOI
  22. [22]G. M. Andolina, M. Keck, A. Mari, M. Campisi, V. Giovannetti, and M. Polini, Extractable work, the role of correlations, and asymptotic freedom in quantum batteries, Phys. Rev. Lett. 122, 047702 (2019).DOI
  23. [23]G. Gemme, M. Grossi, S. Vallecorsa, M. Sassetti, and D. Ferraro, Quantum battery: Comparing different charging protocols, Phys. Rev. Research 6, 023091 (2024).DOI
  24. [24]Y. Kurman, K. Hymas, A. Fedorov, W. J. Munro, and J. Quach, Powering Quantum Computation with Quantum Batteries, Phys. Rev. X 16, 011016 (2026).DOI
  25. [25]D.-L. Yang, F.-M. Yang, and F.-Q. Dou, Three-level Dicke quantum battery, Phys. Rev. B 109, 235432 (2024).DOI
  26. [26]F.-M. Yang and F.-Q. Dou, Resonator-qutrit quantum battery, Phys. Rev. A 109, 062432 (2024).DOI
  27. [27]S. Ghosh and A. Sen(De), Dimensional enhancements in a quantum battery with imperfections, Phys. Rev. A 105, 022628 (2022).DOI
  28. [28]A. Ali, S. Al-Kuwari, M. I. Hussain, T. Byrnes, M. T. Rahim, J. Q. Quach, M. Ghominejad, and S. Haddadi, Ergotropy and capacity optimization in Heisenberg spin-chain quantum batteries, Phys. Rev. A 110, 052404 (2024).DOI
  29. [29]M. Bode, M. Heide, K. von Bergmann, P. Ferriani, S. Heinze, G. Bihlmayer, A. Kubetzka, O. Pietzsch, S. Blügel, and R. Wiesendanger, Chiral magnetic order at surfaces driven by inversion asymmetry, Nature 447, 190 (2007).pubmed.ncbi.nlm.nih.gov/17495922
  30. [30]I. Dzyaloshinsky, A thermodynamic theory of “weak” ferromagnetism of antiferromagnetics, J. Phys. Chem. Solids 4, 241 (1958).DOI
  31. [31]T. Moriya, Anisotropic superexchange interaction and weak ferromagnetism, Phys. Rev. 120, 91 (1960).DOI
  32. [32]T. Baumgratz, M. Cramer, and M. B. Plenio, Quantifying coherence, Phys. Rev. Lett. 113, 140401 (2014).pubmed.ncbi.nlm.nih.gov/25325620
  33. [33]V. Vedral, The role of relative entropy in quantum information theory, Rev. Mod. Phys. 74, 197 (2002).DOI
  34. [34]M. A. Nielsen and I. L. Chuang, Quantum Computation and Quantum Information, 10th ed. (Cambridge University Press, Cambridge, 2010).
  35. [35]K. Modi, A. Brodutch, H. Cable, T. Paterek, and V. Vedral, The classical–quantum boundary for correlations: Discord and related measures, Rev. Mod. Phys. 84, 1655 (2012).DOI
  36. [36]G.-H. Liu, J.-Y. Dou, and P. Lu, Effect of Dzyaloshinskii–Moriya interaction on phase diagrams of spin-1 Heisenberg–Ising alternating chains, J. Magn. Magn. Mater. 401, 796 (2016).DOI
  37. [37]S. Ghosh, T. Chanda, and A. Sen(De), Enhancement in the performance of a quantum battery by ordered and disordered interactions, Phys. Rev. A 101, 032115 (2020).DOI
  38. [38]V. E. Dmitrienko, E. N. Ovchinnikova, S. P. Collins, G. Nisbet, G. Beutier, Y. O. Kvashnin, V. V. Mazurenko, A. I. Liechtenstein, and M. I. Katsnelson, Measuring the Dzyaloshinskii–Moriya interaction in a weak ferromagnet, Nature Phys. 10, 202 (2014).DOI
  39. [39]A. Khoudiri, K. El Anouz, and A. El Allati, Quantum entanglement generation in autonomous thermal machines: Effects of non-Markovianity, Hilbert space structure, and quantum coherence, AVS Quantum Sci. 8, 022001 (2026).DOI
  40. [40]M. B. Arjimandi, H. Mohammadi, A. Saguia, M. S. Sarandy, and A. C. Santos, Localization effects in disordered quantum batteries, Phys. Rev. E 108, 064106 (2023).
  41. [41]G. Bhanja, D. Tiwari, and S. Banerjee, Impact of non-Markovian quantum Brownian motion on quantum batteries, Phys. Rev. A 109, 012224 (2024).
  42. [42]A. Jad, A. El Allati, and M. B. Arjimandi, Suppressing self-discharging of quantum batteries by cavity interactions, arXiv:2606.23999 [quant-ph].arxiv.org/abs/2606.23999
  43. [43]B. Çakmak, Ergotropy from coherences in an open quantum system, Phys. Rev. E 102, 042111 (2020).DOI
  44. [44]A. Khoudiri, A. El Allati, Ö. E. Müsteçaplıoğlu, and K. El Anouz, Non-Markovianity and a generalized Landauer bound for a minimal quantum autonomous thermal machine with a work qubit, Phys. Rev. E 111, 044124 (2025).DOI

Paper details

vlnc:W_3uqd8rz4
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