Chinese Physics Letters, 2022, Vol. 39, No. 12, Article code 120501Express Letter Theory of Critical Phenomena with Memory Shaolong Zeng, Sue Ping Szeto, and Fan Zhong* Affiliations State Key Laboratory of Optoelectronic Materials and Technologies, School of Physics, Sun Yat-Sen University, Guangzhou 510275, China Received 21 October 2022; accepted manuscript online 15 November 2022; published online 20 November 2022 *Corresponding author. Email: stszf@mail.sysu.edu.cn Citation Text: Zeng S L, Szeto S P, and Zhong F 2022 Chin. Phys. Lett. 39 120501    Abstract Memory is a ubiquitous characteristic of complex systems, and critical phenomena are one of the most intriguing phenomena in nature. Here, we propose an Ising model with memory, develop a corresponding theory of critical phenomena with memory for complex systems, and discover a series of surprising novel results. We show that a naive theory of a usual Hamiltonian with a direct inclusion of a power-law decaying long-range temporal interaction violates radically a hyperscaling law for all spatial dimensions even at and below the upper critical dimension. This entails both indispensable consideration of the Hamiltonian for dynamics, rather than the usual practice of just focusing on the corresponding dynamic Lagrangian alone, and transformations that result in a correct theory in which space and time are inextricably interwoven, leading to an effective spatial dimension that repairs the hyperscaling law. The theory gives rise to a set of novel mean-field critical exponents, which are different from the usual Landau ones, as well as new universality classes. These exponents are verified by numerical simulations of the Ising model with memory in two and three spatial dimensions.
cpl-39-12-120501-fig1.png
cpl-39-12-120501-fig2.png
DOI:10.1088/0256-307X/39/12/120501 © 2022 Chinese Physics Society Article Text Equilibrium critical phenomena are intriguing and have been well understood.[1-5] Classical critical dynamics is usually imposed by a Langevin equation with a Gaussian noise on the basis of instantaneous interactions and has also been well established.[4-8] Quantum critical dynamics of closed systems is determined by their Hamiltonians and can be obtained from analytical continuation through a quantum-classical correspondence.[9] However, open quantum systems may contain temporally nonlocal interactions when their coupled environmental degrees of freedom are integrated away.[10-22] In fact, such a long-range temporal correlation or memory is ubiquitous in complex systems such as reaction-diffusion systems,[23,24] disordered systems,[25-37] active matters,[38,39] epidemic spreading,[40-43] and many others,[44-49] not to mention brain science and biology[50-55] in general. It can result in effects such as anomalous diffusions that are qualitatively different from those exhibited by a memoryless system. What then will one expect memory or non-Markovian dynamics to bring to critical behavior? To this end, we study a model with a time-dependent Hamiltonian, \begin{eqnarray} \mathcal{H}_I(t)=-J\sum_{t_1 < t}\sum_{\langle ij \rangle }\frac{s_{i}(t)s_{j}(t_{1})}{(t-t_{1})^{1+\theta}}, \tag {1} \end{eqnarray} which extends the standard Ising model with a nearest-neighbor coupling constant $J$ to account for a long-range temporal interaction parameterizing by a constant $\theta>0$, which ensures convergence at long times[56] but demands $t_1 < t$ to avoid apparent divergence at $t_1=t$. In Eq. (1), the Ising spin $s_i=\pm1$ at site $i$ may model, besides spins,[11] the phase of a Josephson junction in an array,[10] a density of a chemical species in intracellular processes[54] or complex reaction-diffusion systems,[23,34] positions of active particles,[38,39] or even a memorized pattern in a Hopfield neuron network,[55] and can be generalized to more realistic forms. We have chosen a power-law decay as a simplification of various possible memory types.[10-55] The reason is that it is an analogy of a spatial counterpart[57] and thus the single parameter $\theta$ anticipatorily enables us to classify different universality classes, which can then provide clues to the interactions in real systems. Such an interaction may be realized at least in artificial systems noticing that the decay exponent of the spatial counterpart can even be continuously tuned in trapped ions.[58-63] Model (1) is to be simulated using (random-)update Monte Carlo with Metropolis algorithm,[64,65] which is interpreted as dynamics[65,66] with the time unit of the usual Monte Carlo step per spin. Due to the inherent memory in the model, time average is most likely different from ensemble average over samples. Consequently, we sample configurations only at time $t$ but not the earlier configurations at any $t_1 < t$, resulting in $t$-dependent observable quantities. In fact, such ensemble averages have been successfully exploited to tackle Hamiltonian with a time-dependent term in finite-time scaling (FTS),[67-69] in which scaling laws are found to be satisfied even if the fluctuation-dissipation theorem is not.[69] We will come back to FTS when we test the theory. Theoretically, the long-wavelength low-frequency behavior of model (1) is described by a naïve effective Hamiltonian[70] \begin{align} \!\mathcal{H}(t)=\,&\int\!{d^dx}\Big\{\frac{1}{2}\Big[\tau\phi({\boldsymbol x},t)^2+[\nabla\phi({\boldsymbol x},t)]^2\Big]+\frac{1}{4!}u\phi({\boldsymbol x},t)^4 \notag\\ \!&+\frac{1}{2}v\int_{-\infty}^{t}{dt_1}\frac{\phi({\boldsymbol x},t)\phi({\boldsymbol x},t_1)}{(t-t_1)^{1+\theta}}-h({\boldsymbol x},t)\phi({\boldsymbol x},t)\Big\}, \tag {2} \end{align} which is the usual $\phi^4$ theory for an order-parameter field $\phi({\boldsymbol x},t)$ at position ${\boldsymbol x}$ and time $t$, where $\tau$, $v$, $u$, $h$, and $d$ denote a reduced temperature, the strength of the long-range interaction, a coupling constant, an additional ordering field, and the spatial dimensionality, respectively. In accordance with the Metropolis method, dynamics is governed by the Langevin equation \begin{eqnarray} \lambda_1\partial\phi({\boldsymbol x},t)/\partial t=-\delta\mathscr{H}/\delta \phi({\boldsymbol x},t)+\zeta({\boldsymbol x},t) \tag {3} \end{eqnarray} with a Gaussian white noise $\zeta$ of zero mean $\langle\zeta({\boldsymbol x},t)\rangle=0$ and correlator \begin{eqnarray} \langle\zeta({\boldsymbol x},t)\zeta({\boldsymbol x}_1,t_1)\rangle=2\lambda_2\delta({\boldsymbol x}-{\boldsymbol x}_1)\delta(t-t_1) \tag {4} \end{eqnarray} under the ensemble average. Here, $\mathscr{H}\equiv\int{dt}\mathcal{H}(t)$, and we have introduced two kinetic coefficients $\lambda_1$ and $\lambda_2$ in consideration of the memory effect; $\lambda_1=\lambda_2$ in the usual dynamics and an equilibrium Boltzmann distribution is retrieved. The dynamics has been well-known to be equivalent to a dynamic Lagrangian[4-7,71,72] \begin{eqnarray} \mathcal{L}=\!\int\!{dt d^dx}\{\tilde{\phi}\![\lambda_1\partial\phi/\partial t+ \delta\mathscr{H}/\delta\phi]\!-\!\lambda_2\tilde{\phi}^2\}, \tag {5} \end{eqnarray} where $\tilde{\phi}$ is an auxiliary response field.[73] Hamiltonian (2) is similar to its spatial counterpart, which correctly describes critical behavior of long-range spatial interaction systems.[57,74] Here, surprisedly, we find that the naïve theory of its temporal counterpart is not correct anymore. We show below that such a theory violates a hyperscaling law[1-5,75] \begin{eqnarray} \beta/\nu+\beta\delta/\nu=d, \tag {6} \end{eqnarray} for all $d$, once the memory dominates, where $\beta$, $\nu$, and $\delta$ are critical exponents. This is in sharp contrast to the usual Landau mean-field exponents which break Eq. (6) only above the upper critical dimension $d_{\rm c}$. From the singular part of a time-dependent free energy \begin{eqnarray} f(\tau,h,t)=b^{-d}f(\tau b^{1/\nu}, hb^{\beta\delta/\nu},tb^{-z}). \tag {7} \end{eqnarray} Equation (6) ensures a fundamental thermodynamic relation $M=-(\partial f/\partial h)_{\tau}\sim b^{-d+\beta\delta/\nu}\sim b^{-\beta/\nu}$ between the order parameters $M$ and $h$, and violation of the hyperscaling law then implies violation of the thermodynamics, where $z$ is the dynamic critical exponent. Moreover, to correctly describe the critical behavior of model (1), we demonstrate that a proper dimensional correction of the theory is indispensable. This leads to a series of novel results in the corrected theory such as the role of $\mathcal{H}$ in dynamics, the intimate relation between space and time, and a set of novel mean-field critical exponents different from the usual Landau ones. We will utilize the renormalization-group (RG) technique. To this end, we write Eqs. (2) and (5) together as \begin{align} \mathcal{L}=\,&\int\!{dt d^dx}\Big\{\tilde{\phi}\Big[\lambda_1\frac{\partial\phi}{\partial t}+\tau\phi-K\nabla^2\phi+\frac{1}{3!}u\phi^3\notag\\ &+v\int{dt_1}\frac{\phi(t_1)}{(t-t_1)^{1+\theta}}-h-\lambda_2\tilde{\phi}\Big]\Big\}. \tag {8} \end{align} Here we have inserted a new constant $K$ for renormalization. The RG method is first to integrate out the degrees of freedom within the momentum shell $\varLambda/b\le k\le\varLambda$ and then to rescale $k$ to $k_{\rm r}=kb$ so that the cutoff $\varLambda$ of the remaining degrees of freedom recovers its original value, where $b>1$ is a rescaling factor and $r$ denotes renormalized quantity. To keep the form of $\mathcal{L}$ after renormalization, we also rescale \begin{align} &x_{\rm r}=xb^{-1},~t_{\rm r}=tb^{\gamma_t},~\phi_{\rm r}=\phi b^{\gamma_{\phi}},~\tilde{\phi}_{\rm r}=\tilde{\phi}b^{\gamma_{\tilde{\phi}}},\notag\\ &h_{\rm r}=hb^{\gamma_h}, \tag {9} \end{align} where the constant $\gamma_o$ is referred to hereafter as the dimension of $o$ defined as $o_{\rm r}=ob^{\gamma_o}$. Critical exponents can be obtained through \begin{eqnarray} \beta/\nu=\gamma_{\phi},\qquad\beta\delta/\nu=\gamma_h,\qquad z=-\gamma_t, \tag {10} \end{eqnarray} via scaling hypothesis[1-5] similar to Eq. (7). For an infinitesimal transformation, $b=1+dl$ with $dl\ll1$, keeping only terms to one-loop order,[1-5] we obtain \begin{align} &\partial_lv=(d-\gamma_t-\gamma_{\phi}-\gamma_{\tilde{\phi}}+\theta\gamma_t)v, \tag {11a}\\ &\partial_lu=(d-\gamma_t-3\gamma_{\phi}-\gamma_{\tilde{\phi}})u-3u^2I_2/dl, \tag {11b}\\ &\partial_l\ln K=d-2-\gamma_t-\gamma_{\phi}-\gamma_{\tilde{\phi}}, \tag {11c}\\ &\partial_l\ln\lambda_{2}=d-\gamma_t-2\gamma_{\tilde{\phi}}, \tag {11d}\\ &\partial_l\ln\lambda_{1}=d-\gamma_{\phi}-\gamma_{\tilde{\phi}}=\gamma_{\lambda_1}, \tag {11e}\\ &\partial_l\ln h=d-\gamma_t-\gamma_{\tilde{\phi}}-\gamma_h, \tag {11f}\\ &\partial_l\ln\tau=d-\gamma_t-\gamma_{\phi}-\gamma_{\tilde{\phi}}+uI_1/2\tau dl, \tag {11g} \end{align} where $\partial_l o\equiv\partial o/\partial l=(o_{\rm r}-o)/dl$ and the two integrals are defined as \begin{align} &I_1=\frac{1}{(2\pi)^{d+1}}\!\int_{_{\scriptstyle \varLambda/b}}^{\varLambda}{dk}\!\int{d\omega}C_0(k,\omega)\equiv\!\int_{k}^{b}\!\int_{\omega}\!C_0(k,\omega),\notag\\ &I_2=\int_{k}^{b}\!\int_{\omega}\!C_0(k,\omega)G_0(-k,-\omega), \tag {12} \end{align} with the response $\langle\phi'({\boldsymbol k},\omega)\tilde{\phi}'(-{\boldsymbol k},-\omega)\rangle$ and correlation $\langle\phi'({\boldsymbol k},\omega)\phi'(-{\boldsymbol k},-\omega)\rangle$ functions[4-7,71,72,76] given by \begin{eqnarray} G_0(k,\omega)^{-1}=\tau+Kk^2+v\varGamma(-\theta)(-i\omega)^{\theta}-i\lambda_1\omega \tag {13} \end{eqnarray} ($\varGamma$ is the gamma function) and $C({\boldsymbol k},\omega)=2\lambda_2G_0(k,\omega)G_0(k,-\omega)\equiv \lambda_2C_0(k,\omega)$, respectively. Since even the mean-field results of the naïve theory breach the fundamental law, we need essentially a Gaussian analysis without the results of the two integrals. We now analyze the fixed points of the RG equations and their properties. This is obtained by equating all equations in Eq. (11a) with zero.[77] There are two different cases. The first case is $v^*=0$ from Eq. (11), where the star denotes the fixed-point value. The memory is irrelevant and we recover the usual short-range interaction $\phi^4$ theory. Indeed, solving the fixed-point equations from Eqs. (11c)-(11f) yields \begin{eqnarray} \gamma_t^*=-2,~~\gamma_{\phi}^*=(d-2)/2,~~\gamma_{\tilde{\phi}}^*=\gamma_h^*=(d+2)/2, \tag {14} \end{eqnarray} and hence $u^*(4-d-3I_2u^*/dl)=0$ from Eq. (11b). Standard textbook analysis[1-5] shows that the Gaussian fixed point $u^*=0$ is stable above $d_{c0}=4$ with the Gaussian dimensions, Eq. (14), and hence, from Eq. (10), the correct Gaussian critical exponents $\beta/\nu$, $\beta\delta/\nu$, and $z$ satisfy the hyperscaling law (6). For $d < d_{c0}$, since $I_2>0$, there exists a new fixed point $u^*=dl(4-d)/3I_2$ that controls a nonclassical regime.[1-5] For the long-range fixed point, $v^*\neq0$, instead one finds \begin{eqnarray} u^*(6-2/\theta-d-3I_2u^*/dl)=0, \tag {15} \end{eqnarray} using the fixed-point equations from Eqs. (11a)-(11d). The same analysis shows that the Gaussian fixed point $u^*=0$ is now stable above $d_{\rm c}=6-2/\theta$ with the Gaussian dimensions now changing to \begin{eqnarray} \gamma_t^*=-\frac{2}{\theta},~~\gamma_{\phi}^*=\frac{d}{2}+\frac{1}{\theta}-2,~~\gamma_{\tilde{\phi}}^*=\gamma_h^*=\frac{d}{2}+\frac{1}{\theta}. \tag {16} \end{eqnarray} As a result, we have now $z=2/\theta$ and \begin{eqnarray} \beta/\nu+\beta\delta/\nu=d-2+2/\theta. \tag {17} \end{eqnarray} Equation (17) indicates that the hyperscaling law is radically violated for all $d$ except $\theta=1$ and therefore the naïve theory cannot be right. We note that both sets of the exponents, Eqs. (14) and (16), give rise to $\gamma_{\tau}^*=d-\gamma_t^*-\gamma_{\phi}^*-\gamma_{\tilde{\phi}}^*=2$ in Eq. (11g). This indicates that $\tau$ increases with $b$ and thus is relevant. Moreover, the Gaussian exponent $\nu=1/\gamma_{\tau}^*=1/2$ for both the short-rang and the long-range fixed points.[1-5] Similarly, $\gamma_{\lambda_1}^*=2-2/\theta$ for the long-range fixed point from Eq. (11e). Accordingly, it is negative for $\theta < 1$ and therefore is irrelevant. We now focus on the theory with memory and show that the violation of the hyperscaling law originates from the dimension of $\mathcal{H}$. Indeed, using Eq. (9) and the short-range results, Eq. (14), one sees that all terms in $\mathcal{H}$ have a zero dimension except inconsistently the memory term. However, with the long-range results, Eq. (16), all terms then share an identical dimension \begin{eqnarray} \gamma_{_{\scriptstyle \mathcal{H}}}=2-2/\theta=\gamma_a, \tag {18} \end{eqnarray} which is minus the extra term in Eq. (17). This prompts us to eliminate $\gamma_{_{\scriptstyle \mathcal{H}}}$ in order to save the hyperscaling law and the theory. To this end, we multiply $\mathscr{H}$ by a constant $a$ such that $\gamma_{_{\scriptstyle \mathcal{H}'}}=-\gamma_a+\gamma_{_{\scriptstyle \mathcal{H}}}=0$. However, introducing an additional scaling field to the theory breaks the correspondence between Eqs. (1) and (2). This means that $a$ just contributes a factor $b^{-\gamma_a}$ (note the special minus sign to emphasize its speciality) instead of changing to $a_{\rm r}$ upon rescaling. In addition, $\zeta$ ought to be multiplied by $a$ accordingly. Its correlator in Eq. (4) is then proportional to $a^2$. Consequently, $\mathcal{L}$, Eq. (5), becomes \begin{eqnarray} \mathcal{L}'=\int{dtd^dx}[\tilde{\phi}({\boldsymbol x},t)a\delta\mathscr{H}/\delta\phi-a^2\tilde{\phi}({\boldsymbol x},t)^2], \tag {19} \end{eqnarray} where we have omitted $\lambda$ and the irrelevant time derivative term for $\theta < 1$. One sees from Eq. (19) that if one defines $\tilde{\phi}_1=a\tilde{\phi}$, $\mathcal{L}$ recovers its original form. This is reasonable as $\tilde{\phi}$ originates from integration out of $\zeta$.[73] The replacement implies that one ought to replace $\gamma_{\tilde{\phi}}$ with $\gamma_{\tilde{\phi}}+\gamma_a$ in Eq. (11a) and thus the new hyperscaling law (17) does hold. However, this is not enough. To see why, we make a transformation, \begin{eqnarray} \phi=\phi' a^{-n},\qquad\tilde{\phi}=\tilde{\phi}' a^{-\tilde{n}}, \tag {20} \end{eqnarray} with two constants $n$ and $\tilde{n}$ and obtain \begin{align} \mathcal{H}'=\,&a^{1-2n}\int_k\Big[\frac{1}{2}G_0^{-1}|\phi'|^2+\frac{1}{4!}a^{-2n}u\mathcal{H}_u-a^{n}h\phi'\Big],\notag\\ \mathcal{L}'=\,&\int_{\omega,k}\tilde{\phi'}\Big[ a^{1-n-\tilde{n}}G_0^{-1}\phi'+\frac{1}{3!}a^{1-3n-\tilde{n}}u\frac{\delta\mathcal{H}_u}{\delta\phi'}\notag\\ &-a^{1-\tilde{n}}h-a^{2-2\tilde{n}}\tilde{\phi'}\Big], \tag {21} \end{align} after Fourier transform, where $\mathcal{H}_u$ denotes the Fourier transform of $\int\!dtd^dx\phi^4$ with one frequency and wavenumber integration extracted, and we have used identical symbols for both direct spacetime and its reciprocal and suppressed the arguments for clarity. In Eq. (21), we have drawn $a^{1-2n}$ out of the integration for $\mathcal{H}'$. In this way, we find that the perturbation contribution to the renormalized vertex, $\tilde{u}$, becomes \begin{eqnarray} \tilde{u}=u\Big\{1-\frac{3}{4}a^{1-4n}u\!\int_k\!G_0(k,0)^2\Big\}, \tag {22} \end{eqnarray} to one-loop order for all $t$, $t_1$, and $(t-t_1)$ in Eq. (2) approaching infinity. It can be shown that all higher orders contain powers of $a^{1-4n}u$ combinatorially.[70] Therefore, if we choose $n=1/4$, $a$ does not contribute to the renormalization. This is also true for other parameters such as $\tau$ and substantially disentangles the involvement of $a$ in the theory. For $\mathcal{L}'$ in Eq. (21), the transformation Eq. (20) represents only a redistribution of $\gamma_a$ between the fields and must not introduce an overall factor. Accordingly, from Eq. (21), we have $n=1-\tilde{n}$, or $\tilde{n}=3/4$ to ensure identical $h$ transformations ($h'=ha^{1/4}$), propagators, and $u$ for $\mathcal{H}'$ and its $\mathcal{L}'$. Although the last term in $\mathcal{L}'$ leads to an $a$-dependent correlation function $C({\boldsymbol k},\omega)=a^{2n}C_0(k,\omega)$, one can again show that all of $a$ are exactly canceled in the two integrals in Eq. (12) for the renormalized $\tau_{\rm r}$ and $u_{\rm r}$ to one-loop order and to all orders combinatorially.[70] Conversely, without the transformation (20), $a$ could not serve as a simple constant only, since it would be differently involved in $\mathcal{H}$ and $\mathcal{L}$, which are then inconsistent with each other. We note that for $\mathcal{H}'$ and $\mathcal{L}'$ to be really consistent, each integration over wavenumber for $\mathcal{L}'$ and hence in Eq. (12) must also contain the same $a^{1-2n}=a^{1/2}$ factor. However, it must be exactly canceled by $a^{-1/2}$ for the frequency integration. The same happens also to the integrations in $\mathcal{H}_u$. Accordingly, the correct theory in place of Eqs. (2) and (5) is \begin{align} \mathcal{L}'=\,&\int{\big(dta^{-\frac{1}{2}}\big)\big(d^dxa^{\frac{1}{2}}\big)}\tilde{\phi'} \Big\{\tau\phi'-\nabla^2\phi'+\frac{1}{3!}a^{-\frac{1}{2}}u\phi'^3\notag\\ &+v\int_{-\infty}^{t}{dt_1}\frac{\phi'({\boldsymbol x},t_1)}{(t-t_1)^{1+\theta}}-h'-a^{\frac{1}{2}}\tilde{\phi'}\Big\}. \tag {23} \end{align} Equation (23) and its associated $\mathcal{H}'$ constitute our corrected theory of critical phenomena with memory. The RG transformations of the theory are similar to Eq. (11a) except that the transformations for $u$ and $\lambda_2$ need add and subtract $\gamma_a/2$, respectively, because of the extra factors. The fixed-point solution for $u$ is again Eq. (15) while \begin{align} &\gamma_t^*=-2/\theta,~~\gamma_{\phi'}^*=d/2+1/2\theta-3/2,\notag\\ &\gamma_{\tilde{\phi}}^*=d/2+3/2\theta-1/2,~~\gamma_{h'}^*=d/2+1/2\theta+1/2,\tag {24} \end{align} which leads to \begin{eqnarray} \beta/\nu+\beta\delta/\nu=d-\gamma_a/2\equiv d_{\rm eff}, \tag {25} \end{eqnarray} using Eq. (10). Although Eq. (25) appears to violate both Eqs. (6) and (17), $d_{\rm eff}$ is just the dimension of $d^dxa^{1/2}$ and hence an effective spatial dimension. Accordingly, the hyperscaling law (6) holds again in it! This is corroborated by the fact that the coupling term $a^{-1/2}u$ now scales as $b^{d_{\rm c}-\gamma_a/2-d}\equiv b^{d_{\rm c, eff}-d}$, viz., the effective upper critical dimension $d_{\rm c, eff}$ changes consistently to $d_{\rm c}-\gamma_a/2=5-1/\theta$. Similarly, the usual hyperscaling law ought to be modified to $d_{\rm eff}\nu=(d-\gamma_a/2)\nu=2-\alpha$ accordingly, while the others such as $\gamma=(2-\eta)\nu$[1-5] remain unchanged, where $\alpha$ and $\gamma$ are the critical exponents for specific heat and susceptibility $\chi$, respectively. From Eq. (18), $d_{\rm eff}$ is always larger than $d$ in the memory-dominant regime, a fact which is qualitatively reasonable as the temporal correlation inevitably suppresses spatial fluctuations and contributes an effective spatial dimension. Moreover, in the memory-dominated regime, the correct mean-field critical exponents, Eq. (24) at $d_{\rm c}$, become \begin{eqnarray} \beta/\nu=3/2-1/{2\theta},\qquad \beta\delta/\nu=7/2-1/2\theta, \tag {26} \end{eqnarray} though $\nu=1/2$ since $\gamma_{\tau}^*=2$ too and $\gamma=1$ as usual. For $\theta=1/2$ and $2/3$, for example, $d_{\rm c}=2$ and $3$ and $\beta/\nu=1/2$ and $3/4$, respectively, different from the Landau result $\beta/\nu=1$ obtained from both Eq. (14) and Eq. (16). In addition, since the fixed-point Eq. (15) remains intact, $d_{\rm c}$ again separates the classical regime above it from and nonclassical regimes below it. Also, from Eq. (24), $\gamma_{\lambda_1}^*=2-2/\theta$ too, the division of the short-range and long-range regimes remains intact. In fact, to $\epsilon^2=(d_{\rm c}-d)^2$, we find that this division shifts to $\theta=1-\kappa$ with $\kappa=\ln(4/3)\epsilon^2/18$.[78] These divisions of different regimes are similar to the spatial counterpart.[57,74] However, the central difference is that $\gamma_{_{\scriptstyle \mathcal{H}}}=0$ in the latter case and hence no corresponding behavior as presented here emerges. Detailed comparison of the two systems are given elsewhere.[78] Turning to $a^{-1/2}$ in the temporal integration in Eq. (23), we need to transform $t$ to $t'=a^{-1/2}t$, which just accounts for the last extra $a^{1/2}$ in Eq. (23) because $a^{1/2}\delta(t)=\delta(t')$ in Eq. (4). Indeed, an RG analysis with $t'$ instead of $t$ results again in Eq. (24) with a new $z'=-\gamma_{t'}^*=z+\gamma_a/2$,[79] which indicates that part of the temporal dimension has been transferred to the space and which leads correctly to $G_0=b^2/[\tau_{\rm r}+K_Rk^2_{\rm r}+v_{\rm r}a^{-\theta/2}\varGamma(-\theta)(-i\omega'_{\rm r})^{\theta}]$ upon neglecting the irrelevant $\lambda_1$ term in Eq. (13), where $\omega'_{\rm r}=a^{1/2}\omega b^{-\gamma_{t'}^*}$. A similar scaling obviously holds for $G_0$ in terms of $t$ as well. We have seen that $\mathcal{H}$ plays an essential role due to Eq. (18). This draws attention to the eluded role played by $\mathcal{H}$ in dynamics, as nowadays one usually solely focuses on $\mathcal{L}$ for critical dynamics though $\mathcal{H}$ controls the equilibrium properties.[4-7,71,72] Since $a$ arises from the memory, the scaling laws thus connect both static and dynamic properties, demonstrating the intimate relation between space and time similar to quantum phase transitions but adjustable through $\theta$ in the memory-dominated regime. To verify the exponents, we utilize FTS.[67-69] FTS is to change a controlling parameter in Hamiltonian linearly with time at a rate $R$ through a critical point and is perfect for dealing with the time-dependent Hamiltonian (1). The driving imposes on the system a controllable external finite timescale that plays the role of system size in finite-size scaling and leads to FTS.[67] FTS has been successfully applied to many systems to efficiently study their equilibrium and nonequilibrium critical properties.[13,67-69,80-98] Here, we change the temperature $T$ through $\tau=T-T_{\rm c}=Rt'$[70] and choose $R$ as a variable in place of $t'$. Accordingly, Eq. (7) becomes[67-69] \begin{eqnarray} f(\tau,h,R)=b^{-d_{\rm eff}}f(\tau b^{1/\nu},h'b^{\beta\delta/\nu},Rb^{r}), \tag {27} \end{eqnarray} where $r$ is a rate exponent.[99,100] Since $t'$ scales as $t'b^{-z'}$, one finds $r=z'+1/\nu=3+1/\theta$, which enables us to verify $z'$. Consequently, one arrives at the FTS forms \begin{eqnarray} M=R^{\beta/r\nu}f_{1}(\tau R^{-1/r\nu}),~ \chi=R^{-\gamma/r\nu}f_{2}(\tau R^{-1/r\nu}), \tag {28} \end{eqnarray} for $M$ and $\chi$ at $h'=0$, where $f_1$ and $f_{2}$ are scaling functions. At the peak of $\chi$, $T_{\rm p}=T_{\rm c}+cR^{1/r\nu}$ with a constant $c$ satisfying $df_{2}(x)/dx|_{x=c}=0$ from Eq. (28).[67,68] This yields $1/r\nu$ and $T_{\rm c}$, at which $M_{\rm c}\propto R^{\beta/r\nu}$ and $\chi_{\rm c}\propto R^{-\gamma/r\nu}$, and thereby $\beta/r\nu$ and $\gamma/r\nu$. However, $T_{\rm c}$ so estimated has a relatively large error and hence affects others. Thus, we apply the theoretical $1/r\nu$ to fit $T_{\rm c}$ and then use it to determine other exponents and check the consistency by curve collapsing. We simulate model (1) with $J=0.2$ and a cutoff ${\cal T}$ of the long-range interaction. Periodic boundary conditions are applied throughout and initial ordered states are carefully chosen to be far away from $T_{\rm c}$ and equilibrated before heating and thus do not affect results. In Fig. 1, we display the results for the 2D model. We estimate, with the above-mentioned method, $T_{\rm c}=1.402(18)$, which leads to Figs. 1(a)–1(c), where the straight lines have slopes of $1/r\nu=0.405(22)$, $\beta/r\nu=0.0941(4)$, and $-\gamma/r\nu=-0.372(9)$, close to the theoretical values of $2/5$, $1/10$, and $-2/5$, respectively. Note that if $t$ were not transformed to $t'$, the three exponents would be $1/3$, $1/12$, and $-1/3$, respectively, and further, if $\beta$ were $1/2$, $\beta/r\nu=1/6$, all far away from the numerical values. Moreover, these estimated exponents result in the rather good curve collapses in Figs. 1(d) and 1(e) around $T_{\rm c}$, though far away from it, the collapses are not quite good near the peaks where fluctuations are large as expected.[93,95] Similar results are obtained for ${\cal T}=15$.[78] This is reasonable once the driven time is shorter than the cutoff memory time so that the latter becomes sub-leading. Similar condition applies to the lattice sizes as well.[68,69,82] Although these sub-leading factors may still contribute to the small difference of the numerical results, any logarithmic correction needed in the spatial counterpart[101] is found to deteriorate the collapses.
cpl-39-12-120501-fig1.png
Fig. 1. Results of (a) $\tau_{\rm p}=T_{\rm p}-T_{\rm c}$, (b) $|M_{\rm c}|$, (c) $\chi_{\rm c}$, (d) and (e) rescaled $M$ and $\chi$, respectively, of the Ising model with memory with $\theta=0.5$ and ${\cal T}=20$ and subject to heating at a series of $R$ given in the legend for 40000 samples on $100\times100$ square lattices. The insets in (d) and (e) are the respective original curves. Each column shares identical abscissa axes. Double logarithmic scales are used in (a)–(c).
cpl-39-12-120501-fig2.png
Fig. 2. Results of the 3D Ising model with $\theta=2/3$ and ${\cal T}=18$ for 40000 samples on $25\times25\times25$ simple cubic lattices. All style formats are the same as Fig. 1.
The 3D exponents are estimated to be $1/r\nu=0.45(6)$, $\beta/r\nu=0.1580(7)$, and $-\gamma/r\nu=-0.435(14)$ shown in Figs. 2(a)–2(c), again close to the theoretical values of $4/9$, $1/6$, and $-4/9$, respectively. They again yield rather good curve collapses in Figs. 2(d) and 2(e) around $T_{\rm c}$. In summary, we have proposed a model and derived the corresponding theory for critical phenomena with memory arising from a temporal power-law decaying interaction. Due to the dimension of the Hamiltonian, the theory contains a unique dimensional constant that serves (1) to inextricably interweave space and time into an effective dimension, (2) to repair the radically violated hyperscaling law, (3) to transform the time, (4) to change the Gaussian and mean-field critical exponents, (5) to draw attention to the eluded role played by the Hamiltonian in dynamics, (6) to dramatically distinguish the theory from its spatial counterpart, and yet, (7) not to be a scaling field that alters the fixed points. New classical and nonclassical universality classes divided by $d_{\rm c}$ emerge for $\theta < 1-\kappa$. Acknowledgment. This work was supported by the National Natural Science Foundation of China (Grant Nos. 11575297 and 12175316).
References Critical dynamics: a field-theoretical approachTheory of dynamic critical phenomenaOnset of Global Phase Coherence in Josephson-Junction Arrays: A Dissipative Phase TransitionPhase Diagram and Critical Exponents of a Dissipative Ising Spin Chain in a Transverse Magnetic FieldNonequilibrium quantum criticality in open systems: The dissipation rate as an additional indispensable scaling variableQuantum channels and memory effectsColloquium : Non-Markovian dynamics in open quantum systemsQuantum memories at finite temperatureNumerical Renormalization Group for Bosonic Systems and Application to the Sub-Ohmic Spin-Boson ModelQuantum Phase Transition in the Sub-Ohmic Spin-Boson Model: Quantum Monte Carlo Study with a Continuous Imaginary Time Cluster AlgorithmFinite-Size Scaling of Classical Long-Ranged Ising Chains and the Criticality of Dissipative Quantum Impurity ModelsQuantum criticality in spin chains with non-Ohmic dissipationQuantum phase transitions in the spin-boson model: Monte Carlo method versus variational approach à la FeynmanQuantum tricritical point emerging in the spin-boson model with two dissipative spins in staggered biasesSpatiotemporal memory in a diffusion–reaction systemFractional dynamics of systems with long-range space interaction and temporal memoryAnomalous diffusion in disordered media: Statistical mechanisms, models and physical applicationsThe random walk's guide to anomalous diffusion: a fractional dynamics approachColloquium : Fractional calculus view of complexity: A tutorialMemory formation in matterNoise-Induced Dynamics in Bistable Systems with DelayNoise-Induced Resonance in Delayed Feedback SystemsDistribution of Residence Times of Time-Delayed Bistable Systems Driven by NoiseMemory driven Ginzburg-Landau modelElephants can always remember: Exact long-range memory effects in a non-Markovian random walkMemory-controlled diffusionSimple Measure of Memory for Dynamical Processes Described by a Generalized Langevin EquationRejuvenation and Memory Effects in a Structural GlassGiant leaps and long excursions: Fluctuation mechanisms in systems with long-range memoryMemory-Induced Transition from a Persistent Random Walk to Circular Motion for Achiral MicroswimmersActive particles sense micromechanical properties of glassesEpidemic processes in complex networksNon-Markovian Infection Spread Dramatically Alters the Susceptible-Infected-Susceptible Epidemic Threshold in NetworksEffects of Network Structure, Competition and Memory Time on Social Spreading PhenomenaNon-Markovian recovery makes complex networks more resilient against large-scale failuresAngular Memory of Photonic MetasurfacesMultistability in Coupled Nonlinear MetasurfacesMemory-induced acceleration and slowdown of barrier crossingFeedback control of intercellular signalling in developmentMarkovian approaches to modeling intracellular reaction processes with molecular memoryAssociative memory model with arbitrary Hebbian lengthStatistical mechanics and dynamics of solvable models with long-range interactionsCritical Exponents for Long-Range InteractionsQuasiparticle engineering and entanglement propagation in a quantum many-body systemEngineered two-dimensional Ising interactions in a trapped-ion quantum simulator with hundreds of spinsEmergence and Frustration of Magnetism with Variable-Range Interactions in a Quantum SimulatorNon-local propagation of correlations in quantum systems with long-range interactionsQuantum spin dynamics and entanglement generation with hundreds of trapped ionsAchieving continuously tunable critical exponents for long-range spin systems simulated with trapped ionsEquation of State Calculations by Fast Computing MachinesTime‐Dependent Statistics of the Ising ModelFinite-time scaling via linear drivingTheory of driven nonequilibrium critical phenomenaTheory of Critical Phenomena with MemoryStatistical Dynamics of Classical SystemsRecursion Relations and Fixed Points for Ferromagnets with Long-Range InteractionsRenormalization-group theory of first-order phase transition dynamics in field-driven scalar modelFinite-time scaling of dynamic quantum criticalityDynamic scaling at classical phase transitions approached through nonequilibrium quenchingKibble-Zurek mechanism and finite-time scalingUniversal dynamic scaling in three-dimensional Ising spin glassesQuantum versus Classical Annealing: Insights from Scaling Theory and Results for Spin Glasses on 3-Regular GraphsKibble-Zurek mechanism beyond adiabaticity: Finite-time scaling with critical initial slipOff-equilibrium scaling behaviors driven by time-dependent external fields in three-dimensional O ( N ) vector modelsDynamic scaling of topological ordering in classical systemsUniversal driven critical dynamics across a quantum phase transition in ferromagnetic spinor atomic Bose-Einstein condensatesScaling theory of entanglement entropy in confinements near quantum critical pointsDynamical Ginzburg criterion for the quantum-classical crossover of the Kibble-Zurek mechanismDriving driven lattice gases to identify their universality classesActivating critical exponent spectra with a slow driveSelf-Similarity Breaking: Anomalous Nonequilibrium Finite-Size Scaling and Finite-Time ScalingPhases fluctuations and anomalous finite-time scaling in an externally applied field on finite-sized latticesPhases fluctuations, self-similarity breaking and anomalous scalings in driven nonequilibrium critical phenomenaScaling theory of the Kosterlitz-Thouless phase transitionUniversal space-time scaling symmetry in the dynamics of bosons across a quantum phase transitionQuantum Kibble–Zurek mechanism and critical dynamics on a programmable Rydberg simulatorMonte Carlo renormalization group study of the dynamic scaling of hysteresis in the two-dimensional Ising modelProbing criticality with linearly varying external fields: Renormalization group theory of nonequilibrium critical dynamics under drivingClassical critical behavior of spin models with long-range interactions
[1]Ma S K 1976 Modern Theory of Critical Phenomena (Canada: W. A. Benjamin, Inc.)
[2]Cardy J 1996 Scaling and Renormalization in Statistical Physics (Cambridge: Cambridge University Press)
[3]Amit D J and Martin-Mayer V 2005 Field Theory, the Renormalization Group, and Critical Phenomena 3rd edn (Singapore: World Scientific)
[4]Zinn-Justin J 2021 Quantum Field Theory and Critical Phenomena 5th edn (Oxford: Oxford University Press)
[5]Vasil'ev A N 2004 The Field Theoretic Renormalization Group in Critical Behavior Theory and Stochastic Dynamics (London: Chapman and Hall/CRC)
[6]Täuber U C 2014 Critical Dynamics (Cambridge: Cambridge University Press)
[7] Folk R and Moser G 2006 J. Phys. A 39 R207
[8] Hohenberg P C and Halperin B I 1977 Rev. Mod. Phys. 49 435
[9]Sachdev S 1999 Quantum Phase Transitions (Cambridge: Cambridge University Press)
[10] Chakravarty S, Ingold G L, Kivelson S, and Luther A 1986 Phys. Rev. Lett. 56 2303
[11] Werner P, Völker K, Troyer M, and Chakravarty S 2005 Phys. Rev. Lett. 94 047201
[12]Weiss U 2008 Quantum Dissipative Systems 3rd edn (Singapore: World Scientific)
[13] Yin S, Mai P, and Zhong F 2014 Phys. Rev. B 89 094108
[14] Caruso F, Giovannetti V, Lupo C, and Mancini S 2014 Rev. Mod. Phys. 86 1203
[15] Breuer H P, Laine E M, Piilo J, and Vacchini B 2016 Rev. Mod. Phys. 88 021002
[16] Brown B J, Loss D, Pachos J K, Self C N, and Wootton J R 2016 Rev. Mod. Phys. 88 045005
[17] Bulla R, Tong N H, and Vojta M 2003 Phys. Rev. Lett. 91 170601
[18] Winter A, Rieger H, Vojta M, and Bulla R 2009 Phys. Rev. Lett. 102 030601
[19] Kirchner S, Si Q, and Ingersent K 2009 Phys. Rev. Lett. 102 166405
[20] Sperstad I B, Stiansen E B, and Sudbø A 2012 Phys. Rev. B 85 214302
[21] De Filippis G, de Candia A, Cangemi L M, Sassetti M, Fazio R, and Cataudella V 2020 Phys. Rev. B 101 180408(R)
[22] Wang Y Z, He S, Duan L, and Chen Q H 2021 Phys. Rev. B 103 205106
[23] Schulz M, Trimper S, and Zabrocki K 2007 J. Phys. A 40 3369
[24] Tarasov V E and Zaslavsky G M 2007 Phys. A 383 291
[25] Bouchaud J P and Georges A 1990 Phys. Rep. 195 127
[26] Metzler R and Klafter J 2000 Phys. Rep. 339 1
[27] West B J 2014 Rev. Mod. Phys. 86 1169
[28] Keim N C, Paulsen J D, Zeravcic Z, Sastry S, and Nagel S R 2019 Rev. Mod. Phys. 91 035002
[29] Tsimring L S and Pikovsky A 2001 Phys. Rev. Lett. 87 250602
[30] Masoller C 2002 Phys. Rev. Lett. 88 034102
[31] Masoller C 2003 Phys. Rev. Lett. 90 020601
[32] Trimper S, Zabrocki K, and Schulz M 2002 Phys. Rev. E 66 026114
[33] Schütz G M and Trimper S 2004 Phys. Rev. E 70 045101(R)
[34] Trimper S, Zabrocki K, and Schulz M 2004 Phys. Rev. E 70 056133
[35] Mokshin A V, Yulmetyev R M, and Hänggi P 2005 Phys. Rev. Lett. 95 200601
[36] Scalliet C and Berthier L 2019 Phys. Rev. Lett. 122 255502
[37] Jack R L and Harris R J 2020 Phys. Rev. E 102 012154
[38] Narinder N, Bechinger C, and Gomez-Solano J R 2018 Phys. Rev. Lett. 121 078003
[39] Lozano C, Gomez-Solano J R, and Bechinger C 2019 Nat. Mater. 18 1118
[40] Pastor-Satorras R, Castellano C, Van Mieghem P, and Vespignani A 2015 Rev. Mod. Phys. 87 925
[41] Van Mieghem P and van de Bovenkamp R 2013 Phys. Rev. Lett. 110 108701
[42] Gleeson J P, O'Sullivan K P, Baños R A, and Moreno Y 2016 Phys. Rev. X 6 021019
[43] Lin Z H, Feng M, Tang M, Liu Z, Xu C, Hui P M, and Lai Y C 2020 Nat. Commun. 11 2490
[44]Zaslavsky G M 2005 Hamiltonian Chaos and Fractional Dynamics (Oxford: Oxford University Press)
[45]Rangarajan G and Ding M (eds) 2003 Processes with Long-Range Correlations: Theory and Applications, Lecture Notes in Physics vol 621 (Berlin: Springer-Verlag)
[46]Beran J, Feng Y, Ghosh S, and Kulik R 2013 Long-Memory Processes: Probabilistic Properties and Statistical Methods (Berlin: Springer-Verlag)
[47]Wunsch C 2015 Modern Observational Physical Oceanography: Understanding the Global Ocean (Princeton, NJ: Princeton University Press)
[48] Valagiannopoulos C, Sarsen A, and Alù A 2021 IEEE Trans. Anten. Propag. 69 7720
[49] Valagiannopoulos C 2022 IEEE Trans. Anten. Propag. 70 5534
[50]Murray J D 2000 Mathematical Biology, part I (Berlin: Springer)
[51]Murray J D 2003 Mathematical Biology, part II (Berlin: Springer)
[52] Kappler J, Daldrop J O, Brünig F N, Boehle M D, and Netz R R 2018 J. Chem. Phys. 148 014903
[53] Freeman M 2000 Nature 408 313
[54] Zhang J J and Zhou T 2019 Proc. Natl. Acad. Sci. USA 116 23542
[55] Jiang Z J, Zhou J W, Hou T Q, Wong K Y M, and Huang H P 2021 Phys. Rev. E 104 064306
[56] Campa A, Dauxois T, and Ruffo S 2009 Phys. Rep. 480 57
[57] Fisher M E, Ma S K, and Nickel B G 1972 Phys. Rev. Lett. 29 917
[58] Jurcevic P, Lanyon B P, Hauke P, Hempel C, Zoller P, Blatt R, and Roos C F 2014 Nature 511 202
[59] Britton J W, Sawyer B C, Keith A C, Wang C C J, Freericks J K, Uys H, Biercuk M J, and Bollinger J J 2012 Nature 484 489
[60] Islam R, Senko C, Campbell W, Korenblit S, Smith J, Lee A, Edwards E, Wang C C, Freericks J, and Monroe C 2013 Science 340 583
[61] Richerme P, Gong Z X, Lee A, Senko C, Smith J, Foss-Feig M, Michalakis S, Gorshkov A V, and Monroe C 2014 Nature 511 198
[62] Bohnet J G, Sawyer B C, Britton J W, Wall M L, Rey A M, Foss-Feig M, and Bollinger J J 2016 Science 352 1297
[63] Yang F, Jiang S J, and Zhou F 2019 Phys. Rev. A 99 012119
[64] Metropolis N, Rosenbluth A W, Rosenbluth M N, Teller A M, and Teller E 1953 J. Chem. Phys. 21 1087
[65]Landau D P and Binder K 2005 A Guide to Monte Carlo Simulations in Statistical Physics 2nd edn (Cambridge: Cambridge University Press)
[66] Glauber R J 1963 J. Math. Phys. 4 294
[67] Gong S R, Zhong F, Huang X Z, and Fan S L 2010 New J. Phys. 12 043036
[68]Zhong F 2011 Applications of Monte Carlo Method in Science and Engineering, edited by Mordechai S (Intech, Rijeka, Croatia) p 469 available at http://www.dwz.cn/B9Pe2
[69] Feng B Q, Yin S, and Zhong F 2016 Phys. Rev. B 94 144103
[70] Zeng S S, Szeto S P, and Zhong F 2022 arXiv:2203.16243 [cond-mat.stat-mech] for a proof of the equivalence of the two models
[71]Janssen H K 1979 Dynamical Critical Phenomena and Related topics, Enz C P (ed) Lecture Notes in Physics vol 104 (Berlin: Springer)
[72]Janssen H K 1992 From Phase Transition to Chaos, Györgyi G, Kondor I, Sasvári L, and Tél T (eds) (Singapore: World Scientific)
[73] Martin P C, Siggia E D, and Rose H A 1973 Phys. Rev. A 8 423
[74] Sak J 1973 Phys. Rev. B 8 281
[75]It is derived from the standard scaling laws $d\nu=2-\alpha$, $\alpha+2\beta+\gamma=2$, and $\gamma=\beta(\delta-1)$ (see Refs. [1-5])
[76] Zhong F 2017 Front. Phys. 12 126402
[77]In the long-range fixed point $v^{*}\neq0$ and to the present one-loop order, Eq. (11e) is equal to $\gamma_{\lambda_{1}}|=2-2/\theta$ rather than zero because it is irrlevant (see the text). However, to higher orders, it is again equal to zero and leads to crossover, see Ref. [78] for details.
[78]Zeng S, Szeto S P, and Zhong F 2022 in preparation
[79]Note that although $\gamma_{t}$ becomes $\gamma_{t'}=\gamma_{t}-\gamma_{a}/2$, $d$ also changes to $d-\gamma_{a}/2$. Consequently, $\gamma_{h}$ keeps unchanged.
[80] Yin S, Qin X, Lee C, and Zhong F 2012 arXiv:1207.1602 [cond-mat.stat-mech]
[81] Liu C W, Polkovnikov A, and Sandvik A W 2014 Phys. Rev. B 89 054307
[82] Huang Y Y, Yin S, Feng B Q, and Zhong F 2014 Phys. Rev. B 90 134108
[83] Liu C W, Polkovnikov A, Sandvik A W, and Young A P 2015 Phys. Rev. E 92 022128
[84] Liu C W, Polkovnikov A, and Sandvik A W 2015 Phys. Rev. Lett. 114 147203
[85] Huang Y Y, Yin S, Hu Q J, and Zhong F 2016 Phys. Rev. B 93 024103
[86] Pelissetto A and Vicari E 2016 Phys. Rev. E 93 032141
[87] Xu N, Castelnovo C, Melko R G, Chamon C, and Sandvik A W 2018 Phys. Rev. B 97 024432
[88] Xue M, Yin S, and You L 2018 Phys. Rev. A 98 013619
[89] Cao X M, Hu Q J, and Zhong F 2018 Phys. Rev. B 98 245124
[90] Gerster M, Haggenmiller B, Tschirsich F, Silvi P, and Montangero S 2019 Phys. Rev. B 100 024311
[91] Li Y H, Zeng Z D, and Zhong F 2019 Phys. Rev. E 100 020105(R)
[92] Mathey S and Diehl S 2020 Phys. Rev. Res. 2 013150
[93] Yuan W L, Yin S, and Zhong F 2021 Chin. Phys. Lett. 38 026401
[94] Yuan W L and Zhong F 2021 J. Phys.: Condens. Matter 33 375401
[95] Yuan W L and Zhong F 2021 J. Phys.: Condens. Matter 33 385401
[96] Zuo Z Y, Yin S, Cao X M, and Zhong F 2021 Phys. Rev. B 104 214108
[97] Clark L W, Feng L, and Chin C 2016 Science 354 606
[98] Keesling A, Omran A, Levine H, Bernien H, Pichler H, Choi S, Samajdar R, Schwartz S, Silvi P, Sachdev S, Zoller P, Endres M, Greiner M, Vuletić V, and Lukin M D 2019 Nature 568 207
[99] Zhong F 2002 Phys. Rev. B 66 060401(R)
[100] Zhong F 2006 Phys. Rev. E 73 047102
[101] Luijten E and Blöte H W J 1997 Phys. Rev. B 56 8945