This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Magnetization oscillations in a periodically driven transverse field Ising chain

Xiao Wang gbsn(王骁) 0000-0003-2898-3355 Tsung-Dao Lee Institute, Shanghai Jiao Tong University, Shanghai, 201210, China Shanghai Branch, Hefei National Laboratory, Shanghai 201315, China    Masaki Oshikawa gbsn(押川正毅) 0000-0002-7637-7432 Institute for Solid State Physics, The University of Tokyo. Kashiwa, Chiba 277-8581, Japan Kavli Institute for the Physics and Mathematics of the Universe (WPI), The University of Tokyo, Kashiwa, Chiba 277-8583, Japan    Márton Kormos 0000-0002-1884-2818 [email protected] Department of Theoretical Physics, Institute of Physics, Budapest University of Technology and Economics, 1111 Budapest, Műegyetem rkp. 3, Hungary HUN-REN–BME Quantum Dynamics and Correlations Research Group, Budapest University of Technology and Economics, 1111 Budapest, Műegyetem rkp. 3, Hungary    Jianda Wu gbsn(吴建达) 0000-0002-3571-3348 [email protected] Tsung-Dao Lee Institute, Shanghai Jiao Tong University, Shanghai, 201210, China Shanghai Branch, Hefei National Laboratory, Shanghai 201315, China School of Physics & Astronomy, Shanghai Jiao Tong University, Shanghai, 200240, China
Abstract

We investigate the nonequilibrium dynamics of the magnetization in an Ising chain subjected to a slowly rotating transverse field. The magnetization oscillations are found to be explained by the contributions from different particle excitations in the quantum E8E_{8} model. We study the magnetization in the frequency domain in detail, uncovering a series of singular peaks for the zz (Ising) component. These singular peaks are split into two sets for the magnetization along xx and yy directions with frequency shifts set by the rotational-field frequency. The peaks include both δ\delta-function type and edge-singularity type peaks. The δ\delta-function peaks can be attributed to particle excitations involving an E8E_{8} particle with either the vacuum or a different particle. The edge-singularity peaks are contributed by particle excitations of two E8E_{8} particles with either the vacuum or another particle, or by particle excitations that contain two sets of two particles with each set including at least a particle of the same type. We propose a Rydberg qubit array for possible experimental investigation.

Introduction.—Time-dependent quantum systems have attracted continuous attention in the past few decades, leading to rich experimental and theoretical progress. These include time crystals [1, 2, 3, 4], nonlinear electron spin resonance and photon-induced phase transitions [5, 6, 7, 8, 9, 10], measurement induced exotic phenomena [11, 12, 13, 14, 15], as well as Kibble-Zurek mechanism [16, 17, 18, 19, 20]. Along the course the transverse-field quantum Ising chain (TFIC) serves as a prototypical system for investigating out of equilibrium dynamics [21, 22, 23]. For a static transverse field, the system is one of the simplest solvable models that accommodates quantum phase transition [24, 25]. At its quantum critical point (QCP), the system exhibits conformal invariance [26]. Introducing a perturbation field along the Ising-spin direction at the QCP gives rise, in the scaling limit, to an integrable field theory known as quantum E8E_{8} integrable model [27, 28, 29, 30, 31, 32, 33, 34]. When the transverse field is time dependent, a complete analytical understanding poses a significant challenge.

There are two different widely-studied forms of the time-dependent magnetic field. One is a sudden change of the magnetic field at the initial time [35, 21, 36, 37, 38, 39], and the other is a periodically driven magnetic field [40, 10, 41]. For a sudden quench involving a nonzero longitudinal field, the dynamics of both the one-point functions and the entanglement entropy reveal fingerprints of the confinement phenomenon [36, 37, 38, 42, 43, 44, 45, 46]. For the periodic driving, it has been proved that for a rotating field in the transverse plane, the time evolution can be related to the TFIC with the presence of an effective longitudinal field proportional to the rotation frequency [41, 40].

In this letter, we focus on the critical Ising chain with a periodically driven magnetic field that slowly rotates in the transverse plane. Under a rotational wave transformation [47], the time evolution process can be effectively related to a time-independent critical TFIC perturbed by an effective longitudinal field proportional to the rotation frequency [40]. In the scaling limit, the effective model becomes the quantum E8E_{8} integrable model [31]. Using the exact form factors of the model and the post-quench perturbative framework [28, 48, 42, 17, 31, 49, 50, 38], we analytically determine the time evolution of the magnetization components MαM^{\alpha} (α=x,y,z)(\alpha=x,y,z) in both the time and frequency domains, where the dominant low-energy spectrum can be understood through contributions from different particle states of the quantum E8E_{8} integrable model.

For the magnetization spectra Mz(ω)M^{z}(\omega), we find that the low-energy collective excitations in frequency domain are dominated by cascades of singular peaks of two different types: the δ\delta-function type and the edge-singularity type. The former singular peaks are contributed by the particle excitations of an E8E_{8} particle with either the vacuum or a different particle. The latter are due to the particle excitations of two particles with either the vacuum or an E8E_{8} particle, or due to the particle excitations of two sets of two particles, where the two sets contain particles of the same type  [51]. For the magnetization spectra Mx(ω)M^{x}(\omega) and My(ω)M^{y}(\omega), cascades of singular peaks also appear with similar structure and the same origin, but split into two sets with respect to the peaks of Mz(ω)M^{z}(\omega), with their corresponding frequency shifts matching the rotating frequency of the transverse field. Numerical verification using the infinite time-evolving block decimation (iTEBD) algorithm is also presented, supporting the analytical results. Finally, a Rydberg-atom quantum simulator is proposed for future experimental study, providing a promising way to simulate and investigate the interplay of non-equilibrium physics and quantum integrability.

Model.—The Hamiltonian of a critical Ising chain subjected to a periodically driven transverse field B(t)=(cos(ω0t)\vec{B}(t)=({\cos(\omega_{0}t)}, sin(ω0t),0)-\sin(\omega_{0}t),0) is given by

=Ji[σizσi+1z+cos(ω0t)σixsin(ω0t)σiy],\mathcal{H}=-J\sum_{i}\left[\sigma^{z}_{i}\sigma^{z}_{i+1}+\cos(\omega_{0}t)\sigma^{x}_{i}-\sin(\omega_{0}t)\sigma^{y}_{i}\right], (1)

where σiα(α=x,y,z)\sigma_{i}^{\alpha}(\alpha=x,~{}y,~{}z) represent the Pauli matrices at site ii, and we assume that the rotation frequency ()ω0J(\hbar)\omega_{0}\ll J. Hereafter, \hbar and JJ are set as 1. At t=0t=0, the system is a critical TFIC, and we set the initial state to be its ground state. By using a unitary transformation [41, 40, 16, 51],

𝒰D(t)=exp{i2ω0tiσiz},\mathcal{U}_{D}(t)=\exp\left\{-\frac{i}{2}\omega_{0}t\sum_{i}\sigma_{i}^{z}\right\}, (2)

the time-evolved wave function |ψ(t)|\psi(t)\rangle is given by

|ψ(t)=𝒰D1(t)eiefft|ψ(0)|\psi(t)\rangle=\mathcal{U}_{D}^{-1}(t)e^{-i\mathcal{H}_{\text{eff}}t}|\psi(0)\rangle (3)

with the effective static Hamiltonian

eff=i(σizσi+1z+σix12ω0σiz),\mathcal{H}_{\text{eff}}=-\sum_{i}\left(\sigma_{i}^{z}\sigma_{i+1}^{z}+\sigma_{i}^{x}-\frac{1}{2}\omega_{0}\sigma_{i}^{z}\right), (4)

which is a quantum critical TFIC perturbed by an effective longitudinal field ω0/2\omega_{0}/2. We consider the time-dependent magnetization along the xx, yy, and zz directions,

Mα(t)=σα(t)=limN1Ni=1Nσiα(t).M^{\alpha}(t)=\langle\sigma^{\alpha}(t)\rangle=\lim_{N\rightarrow\infty}\frac{1}{N}\sum_{i=1}^{N}\langle\sigma_{i}^{\alpha}(t)\rangle. (5)

Introducing the time-dependent magnetic components Dα(t)D^{\alpha}(t) as

Dα(t)=ψ(0)|eiHefftσαeiHefft|ψ(0),D^{\alpha}(t)=\langle\psi(0)|e^{iH_{\text{eff}}t}\sigma^{\alpha}e^{-iH_{\text{eff}}t}|\psi(0)\rangle, (6)

the time-dependent magnetization Mα(t)M^{\alpha}(t) components are

Mx(t)\displaystyle M^{x}(t) =cos(ω0t)Dx(t)+sin(ω0t)Dy(t),\displaystyle=\cos(\omega_{0}t)D^{x}(t)+\sin(\omega_{0}t)D^{y}(t), (7)
My(t)\displaystyle M^{y}(t) =cos(ω0t)Dy(t)sin(ω0t)Dx(t),\displaystyle=\cos(\omega_{0}t)D^{y}(t)-\sin(\omega_{0}t)D^{x}(t),
Mz(t)\displaystyle M^{z}(t) =Dz(t).\displaystyle=D^{z}(t).

It is in general difficult to determine Eq. (6) analytically. However, for small ω0\omega_{0} the effective Hamiltonian Eq. (4) turns out to be the quantum E8E_{8} Hamiltonian in the scaling limit [27, 28, 31],

E8=c=1/2+h𝑑xσ(x).\mathcal{H}_{E_{8}}=\mathcal{H}_{c=1/2}+h\int dx\sigma(x). (8)

Here, c=1/2\mathcal{H}_{c=1/2} represents the Hamiltonian of conformal field theory (CFT) with central charge c=1/2c=1/2. The coupling intensity hh corresponds to ω0/2\omega_{0}/2 in the scaling limit. The Pauli matrices σz\sigma^{z} and σx\sigma^{x} correspond to the primary fields of the Ising and energy density fields σ\sigma and ε\varepsilon, respectively. In the quantum E8E_{8} integrable model, the matrix elements of these two primary fields can be determined using the form factor bootstrap method [28, 42, 31, 51]. Taking the scaling limit of Dα(t)D^{\alpha}(t) [Eq. (6)]

𝒟𝒪(t)=n,m=0Anm𝒪(t)\displaystyle\mathscr{D}^{\mathcal{O}}(t)=\sum_{n,m=0}^{\infty}A^{\mathcal{O}}_{nm}(t) (9)

for 𝒪=ε,σ\mathcal{O}=\varepsilon,\sigma, hereafter the math script font of corresponding physical quantity refers to the scaling limit. We have inserted two complete E8E_{8} bases {|n;j}\{|n;j\rangle\} and {|m;k}\{|m;k\rangle\} where we organize the states according to their particle numbers nn and mm and j,kj,k label states within these sectors. Then Anm𝒪(t)=j,kei(En;jEm;k)tψs(0)|n;jn;j|𝒪|m;km;k|ψs(0)A^{\mathcal{O}}_{nm}(t)=\sum_{j,k}e^{i(E_{n;j}-E_{m;k})t}\langle\psi_{s}(0)|n;j\rangle\langle n;j|\mathcal{O}|m;k\rangle\langle m;k|\psi_{s}(0)\rangle, with |ψs(0)|\psi_{s}(0)\rangle being |ψ(0)|\psi(0)\rangle in the scaling limit. In addition, En;jE_{n;j} and Em;kE_{m;k} are eigenenergies of the two E8E_{8} eigenstates |n;j|n;j\rangle and |m;k|m;k\rangle, respectively. The overlaps m;k|ψs(0)\langle m;k|\psi_{s}(0)\rangle can be determined using a post-quench perturbative framework [38, 42, 17, 51], and the explicit form of the Anm𝒪(t)A^{\mathcal{O}}_{nm}(t) can be found in [51]. In the following, we call the terms in Eq. (9) with fixed particle numbers particle channels. The remaining challenge is to determine Dy(t)D^{y}(t) in the scaling limit, as σy\sigma^{y} does not correspond to a primary field in the Ising CFT. However, starting from the lattice formalism, we notice [52, 31, 51]

Dy(t)=12dDz(t)dtscaling limit𝒟y(t)=12d𝒟σ(t)dt.D^{y}(t)=\frac{1}{2}\frac{dD^{z}(t)}{dt}\stackrel{{\scriptstyle\text{scaling limit}}}{{\longrightarrow}}\mathscr{D}^{y}(t)=\frac{1}{2}\frac{d\mathscr{D}^{\sigma}(t)}{dt}. (10)

Therefore, by taking the system back to the scaling limit we can simultaneously determine 𝒟σ\mathscr{D}^{\sigma} and 𝒟y\mathscr{D}^{y}.

Refer to caption
Figure 1: (a), (d) Time-dependent magnetization calculated in the quantum E8E_{8} field theory [α(t),α=ε,y,σ\mathscr{M}^{\alpha}(t),\alpha=\varepsilon,y,\sigma] and iTEBD simulation for the lattice model [Eq. (1), (Mα(t),α=x,y,zM^{\alpha}(t),\alpha=x,y,z)], respectively. (b), (e) |σ(ω)||\mathscr{M}^{\sigma}(\omega)| and |Mz(ω)||M^{z}(\omega)|, respectively. For better illustration, σ(t)σ\mathscr{M}^{\sigma}(t)-\langle\sigma\rangle is enlarged with a factor 10. Correspondingly the spectral weights for ω>0\omega>0 in |σ(ω)||\mathscr{M}^{\sigma}(\omega)| are also enlarged with this factor. (c), (f) |ε,y(ω)||\mathscr{M}^{\varepsilon,y}(\omega)| and |Mx,y(ω)||M^{x,y}(\omega)|, respectively. In (e), the dashed magenta and green lines label singular peaks coming from channels of an E8E_{8} particle with either the E8E_{8} vacuum (i\mathcal{R}_{i}) or a different E8E_{8} particle (ij\mathcal{R}_{ij}), with detailed frequencies listed in Table. 1. In (f) the two groups of singular peaks +\mathcal{R}^{+} (black dashed line) and \mathcal{R}^{-} (grey dashed line) for single E8E_{8} particles with corresponding frequency m±=m±ω0m_{\mathcal{R}^{\pm}}=m_{\mathcal{R}}\pm\omega_{0}. NOTE: For the numerical results the peak positions match the analytical results accurately. Due to the finite truncated dimension for long-time evolution in iTEBD algorithm, the spectral weights at low-frequency (ω<1\omega<1) should not be compared with analytical ones. For the analytical results due to heavy cost of multi-particle form factor calculations, some multi-particle’s contribution are not shown in the figure.

Singular peaks.—Using Eq. (7) and Eq. (10), the results for 𝒪(t)\mathscr{M}^{\mathcal{O}}(t), with 𝒪=ε,y\mathcal{O}=\varepsilon,y, and σ\sigma, are displayed in Fig. 1 (a), after normalization with m1=Cωω08/15m_{1}=C_{\omega}{\omega_{0}}^{8/15}, σ=Cσω01/15\langle\sigma\rangle=-C_{\sigma}{\omega_{0}}^{1/15} and ε=εh=0Cεω08/15\langle\varepsilon\rangle=\langle\varepsilon\rangle_{h=0}-C_{\varepsilon}{\omega_{0}}^{8/15}. Here we set ω0=0.03\omega_{0}=0.03, and the explicit values of CωC_{\omega}, CσC_{\sigma} and CεC_{\varepsilon} can be found in [51]. To better understand the collective excitations in the time-dependent magnetization, we consider magnetization in the frequency domain following the one-sided Fourier transformation,

𝒪(ω)=0𝑑teiωt𝒪(t).\mathscr{M}^{\mathcal{O}}(\omega)=\int_{0}^{\infty}dte^{i\omega t}\mathscr{M}^{\mathcal{O}}(t). (11)

We present our analytical results alongside those obtained from numerical iTEBD simulations in Figure 1. To facilitate a clearer comparison between the analytical and numerical outcomes, we applied Gaussian broadening with a broadening factor of α=2%\alpha=2\%. The detailed analytical expressions are provided in [51]. As can be seen in the figure, the dominant singular peaks predicted by the analytical field-theoretical approach are verified by the numerical simulations, which, overall, are given by contributions from single- and two-E8E_{8}-particle channels in the frequency domain [42, 51]. For σ(ω)\mathscr{M}^{\sigma}(\omega) [Mz(ω)M^{z}(\omega) for the lattice simulation], cascades of singular peaks {}\{\mathcal{R}\} are found in the low-energy excitations. Such singular peaks are corresponding to two different types of singular behaviors: the δ\delta-function type and the edge-singularity type. For the δ\delta-function peaks, besides those peaks from an E8E_{8} particle and the E8E_{8} vacuum, other peaks emerge from the contributions of two different E8E_{8} particles, with the corresponding frequency being their mass difference [36, 38, 49, 51]. For the edge-singularity peaks, they have two distinct origins. The first one is due to the divergence in the density of states of E8E_{8} particles, also known as Van Hove singularity [31, 53]. Such edge-singularity peaks are contributed by two different E8E_{8} particles mi,mjm_{i},~{}m_{j}, and the other one being either the E8E_{8} vacuum or a different E8E_{8} particle mkm_{k}, showing a singular behavior as 1/ωωth1/\sqrt{\omega-\omega_{\text{th}}}, where ωth=mi+mj\omega_{\text{th}}=m_{i}+m_{j} (or =mi+mjmk=m_{i}+m_{j}-m_{k}) being the threshold [51]. The second one originates from the kinematic singularities of the multi-particle form factor of the quantum E8E_{8} field theory [51, 28, 42, 31, 53], associated with the particle channels of two distinct sets of E8E_{8} particles. One set comprises two E8E_{8} particles mim_{i} and mkm_{k}, while the other set includes two E8E_{8} particles with at least one of the same type as mjm_{j} and mkm_{k}. These contribute ln(ω+mimj)\ln(\omega+m_{i}-m_{j}) edge-singularities in addition to the previous δ\delta-function peaks at the thresholds of ωth=mjmi\omega_{\text{th}}=m_{j}-m_{i}. It is worth noticing that in the particle channels of a single to two E8E_{8} particles, the kinematic poles also appear and seem to contribute additional singularities, which will eventually be cancelled due to the exchange property of the multi-particle form factors  [51]. In the numerical simulation of the Hamiltonian Eq. (1), by comparing the corresponding frequencies of the peaks with the analytical predictions, the dominant contributions are identified as δ\delta-function peaks [dashed vertical lines in Fig. 1 (e)], with their detailed frequencies listed in Table. 1. Additionally, other singularity peaks are identified as edge-singularity peaks contributed by the vacuum-to-two-particle channels of m1m2m_{1}m_{2}, the one-to-two-particle channels of m5m1m4m_{5}-m_{1}m_{4} and m1m2m5m_{1}m_{2}-m_{5} [51]. However, for the later two classes, the heavy E8E_{8}-particle channels are not shown in the analytical results due to the high computational cost of form factor calculations.

Refer to caption
Figure 2: (a) iTEBD simulation on the magnetization of the Rydberg Ising Hamiltonian [Eq. (13), (Mα(t),α=x,y,zM^{\alpha}(t),\alpha=x,y,z)]. (b, c) |Mz(ω)||M^{z}(\omega)| and |Mx,y(ω)||M^{x,y}(\omega)|. All the peaks have the same structure as Fig. 1(e,f).
Table 1: Ratios of the magnetization peaks in frequency domain correspond to single E8E_{8} particles (1st{}^{\text{st}} row) or the combinations of two E8E_{8} particles (2nd{}^{\text{nd}} row) are given in the 3rd{}^{\text{rd}} row along with their expected values for a specific rotation frequency ω0=0.03\omega_{0}=0.03 (4th{}^{\text{th}} row). The peaks simulated by iTEBD algorithm for ω0=0.03\omega_{0}=0.03 according to the original lattice model [Eq. (1)] and the proposed Rydberg Ising Hamiltonian [Eq. (13)] are also listed (5th{}^{\text{th}} and 6th{}^{\text{th}} row), shown in Fig. (1) (d) and Fig. (2) (b), as the vertical dashed lines in the spectra, respectively.
One particle m1m_{1} m2m_{2} m3m_{3} m4m_{4} m5m_{5} m6m_{6}
Two particles m3m_{3}-m2m_{2} m4m_{4}-m3m_{3} m2m_{2}-m1m_{1} m4m_{4}-m2m_{2} m4m_{4}-m1m_{1} m6m_{6}-m1m_{1} m1m_{1}+m2m_{2}
Theoretical ratio mim_{i}/m1m_{1} 0.3710 0.4158 0.6180 0.7868 1 1.4049 1.6180 1.9890 2.2183 2.4049 2.6180 2.9563 3.2183
Theoretical frequency 0.2139 0.2398 0.3563 0.4537 0.5766 0.8100 0.9329 1.1468 1.2790 1.3866 1.5095 1.7045 1.8556
lattice simulation 0.2238 0.2238 0.3597 0.4563 0.5763 0.8024 0.9293 1.1485 1.2700 1.3809 1.5166 1.6989 1.8628
Rydberg simulation 0.2166 0.2427 0.3608 0.4593 0.5838 0.8201 0.9446 1.1611 1.2950 1.4039 1.5283 1.7258 1.8788

In the spectra of ε(ω)\mathscr{M}^{\varepsilon}(\omega) and y(ω)\mathscr{M}^{y}(\omega), the stable singular peaks {}\{\mathcal{R}\} are further separated into two distinct groups of new singular peaks {±}\{\mathcal{R}^{\pm}\}, where the corresponding excitation frequencies have been shifted as m±ω0m_{\mathcal{R}}\pm\omega_{0}, shown in Fig. 1 (c). This special feature is also captured by the numerical simulation for Mx(ω)M^{x}(\omega) and My(ω)M^{y}(\omega) [Fig. 1 (f)], where the separated single E8E_{8} particles are also highlighted by dashed vertical lines. The numerical verification agrees well with the analytical results.

Implementation in a Rydberg qubit array.— For possible experimental study of the E8E_{8} physics in a periodically driven system, we propose a many-body time-dependent Rydberg Hamiltonian [54, 55],

Ry(t)\displaystyle\mathcal{H}_{Ry}(t) =Ω(t)2ieiϕ(t)|giri|+eiϕ(t)|rigi|\displaystyle=\frac{\Omega(t)}{2}\sum_{i}e^{i\phi(t)}|g_{i}\rangle\langle r_{i}|+e^{-i\phi(t)}|r_{i}\rangle\langle g_{i}| (12)
Δ(t)ini+i<jC6|xixj|6ninj,\displaystyle-\Delta(t)\sum_{i}n_{i}+\sum_{i<j}\frac{C_{6}}{|x_{i}-x_{j}|^{6}}n_{i}n_{j},

where |ri=|1i|r_{i}\rangle=|1_{i}\rangle and |gi=|0i|g_{i}\rangle=|0_{i}\rangle are the representation of the Rydberg states and the ground states on the ithi^{th} qubit, respectively. The amplitude and phase of the driving laser field is Ω(t)\Omega(t) and ϕ(t)\phi(t), respectively. The operator ni=|riri|n_{i}=|r_{i}\rangle\langle r_{i}| counts the Rydberg excitations, and xi=iax_{i}=ia are the coordinate for the ii-th qubit with aa being the lattice spacing. Ω(t)\Omega(t), ϕ(t)\phi(t), Δ(t)\Delta(t) and C6C_{6} are the corresponding (time-dependent) coefficients, whose values can be found in Ref. [54]. By applying a transformation between the Rydberg states and an effective Ising basis σi+=|giri|,σi=|rigi|,σiz=𝟏2ni\sigma_{i}^{+}=|g_{i}\rangle\langle r_{i}|,~{}\sigma_{i}^{-}=|r_{i}\rangle\langle g_{i}|,~{}\sigma_{i}^{z}=\mathbf{1}-2n_{i}, this Rydberg Hamiltonian can be transformed to a periodically driven transverse field Ising chain. Since the interaction strength decays as 1/|ij|61/|i-j|^{6}, we neglect spin interaction term with distance larger than 2 lattice sites, so the effective Hamiltonian follows [55],

Ising/|C6|=\displaystyle\mathcal{H}_{\text{Ising}}/{|C_{6}|}= (13)
iσizσi+1z+λσizσi+2zgc[cos(ω0t)σixsin(ω0t)σiy],\displaystyle-\sum_{i}\sigma^{z}_{i}\sigma^{z}_{i+1}+\lambda\sigma^{z}_{i}\sigma^{z}_{i+2}-g_{c}[\cos(\omega_{0}t)\sigma^{x}_{i}-\sin(\omega_{0}t)\sigma^{y}_{i}],

where λ=1/26\lambda=1/2^{6}. Using iTEBD algorithm, we find that the model hosts a quantum phase transition at the quantum critical point gc=1.0275g_{c}=1.0275, which also falls into the TFIC universality class. The rotating frequency is set as ω0=0.03\omega_{0}=0.03. All the corresponding numerical results are shown in Fig. 2 (a), (b) and (c), with a Gaussian broadening α=2%\alpha=2\%. The dominated low-energy physics proposed by the quantum E8E_{8} integrable field theory are well captured by the numerical lattice simulation results, and the corresponding peaks are also shown in Table. 1. It can be found that the additional next nearest neighboring Ising interaction only slightly modifies the frequency of the singular peaks. Our numerical simulations on Eq. (13) implies that the E8E_{8} physics revealed in the time-dependent model [Eq. (1)] is fully accessible through a practical Rydberg qubit array experimental setup.

Conclusions.—In this letter, we showed that the time evolution of the magnetization in the Ising chain under a slowly rotating transverse field can be described via the integrable E8E_{8} quantum field theory and its particle content. Based on a post-quench perturbation framework, along with the E8E_{8} form factor theory, we analytically determined the magnetizations in both the time and frequency domains. Along the zz direction, we identified two types of singular peaks in the frequency domain: the δ\delta-function and the edge-singularity types. The frequencies of the δ\delta-function peaks are given by the masses and mass differences of E8E_{8} particles. The edge-singularity peaks are essentially the Van Hove singularities given by the divergence of the density of states, or the kinematic singularities of the multi-particle E8E_{8} form factors. A significant contribution given by the particle combination m1m2m_{1}m_{2} is found in the numerical data. Moreover, for the magnetizations along the xx and yy directions, in the frequency domain all the singular peaks are split into two sets of new peaks with their corresponding frequency shifts matching the rotation frequency.

For a possible experimental implementation, we proposed a neutral atom Rydberg quantum simulator that can realize a modified periodically driven transverse field Ising chain. We confirmed numerically that despite the possible presence of next-nearest-neighbor couplings, it can realize the expected magnetization spectra, thus providing a novel and practical route to access the quantum E8E_{8} physics in a time-dependent system.

Switching on a static transverse field provides an exciting opportunity to study the dynamics of confinement, which we leave for future work. Our work may also inspire further studies of non-equilibrium physics via quantum integrability.

Acknowledgements.—We thank Ning Xi, Zongping Gong and Zilong Li for discussions, Congjun Wu for helpful discussion during an early stage of the work. The work at Shanghai Jiao Tong University is sponsored by National Natural Science Foundation of China No. 12274288 and the Innovation Program for Quantum Science and Technology Grant No. 2021ZD0301900. MK was supported by the National Research, Development and Innovation Office of Hungary (NKFIH) through the OTKA Grant K138606. The work of M.O. was partially supported by the JSPS KAKENHI Grants No. JP23K25791 and No. JP24H00946.

References

Appendix A Supplemental Material—Magnetization oscillations in a periodically driven transverse field Ising chain

Appendix B A. The emergent quantum E8E_{8} integrable model

In this appendix, we review the derivation of the effective static Hamiltonian and its scaling limit originally given in Ref. [40]. We start from the original periodically driven transverse field Ising chain to derive the emergent quantum E8E_{8} model. The Schro¨\ddot{\rm{o}}dinger equation is

it|ψ(t)=(t)|ψ(t).i\partial_{t}|\psi(t)\rangle=\mathcal{H}(t)|\psi(t)\rangle. (14)

To solve this equation, considering a time-dependent unitary transformation,

it𝒰D(t)|ψ(t)=[𝒰D(t)(t)𝒰D1(t)+i𝒰D˙(t)𝒰D1(t)]𝒰D(t)|ψ(t),i\partial_{t}\mathcal{U}_{D}(t)|\psi(t)\rangle=\left[\mathcal{U}_{D}(t)\mathcal{H}(t)\mathcal{U}_{D}^{-1}(t)+i\dot{\mathcal{U}_{D}}(t)\mathcal{U}_{D}^{-1}(t)\right]\mathcal{U}_{D}(t)|\psi(t)\rangle, (15)

with

𝒰D(t)=exp{i2ω0tiσiz}.\mathcal{U}_{D}(t)=\exp\left\{-\frac{i}{2}\omega_{0}t\sum_{i}\sigma_{i}^{z}\right\}. (16)

Substituting Eq. (16) into Eq. (15), obtaining

𝒰D(t)(t)𝒰D1(t)+i𝒰D˙(t)𝒰D1(t)=i(σizσi+1z+σix)+12ω0iσiz.\mathcal{U}_{D}(t)\mathcal{H}(t)\mathcal{U}_{D}^{-1}(t)+i\dot{\mathcal{U}_{D}}(t)\mathcal{U}_{D}^{-1}(t)=-\sum_{i}(\sigma_{i}^{z}\sigma_{i+1}^{z}+\sigma_{i}^{x})+\frac{1}{2}\omega_{0}\sum_{i}\sigma_{i}^{z}. (17)

In the scaling limit with ω00\omega_{0}\sim 0, this effective Hamiltonian becomes a quantum E8E_{8} Hamiltonian,

E8=c=1/2+h𝑑xσ(x),\mathcal{H}_{E_{8}}=\mathcal{H}_{c=1/2}+h\int dx\sigma(x), (18)

where hh corresponds to ω0/2\omega_{0}/2 in the scaling limit. The Schro¨\ddot{\rm{o}}dinger equation in the scaling limit becomes

it𝒰D,s(t)|ψs(t)=E8𝒰D,s(t)|ψs(t)i\partial_{t}\mathcal{U}_{D,s}(t)|\psi_{s}(t)\rangle=\mathcal{H}_{E_{8}}\mathcal{U}_{D,s}(t)|\psi_{s}(t)\rangle (19)

with 𝒰D,s(t)=exp{ihtσ(x)𝑑x}\mathcal{U}_{D,s}(t)=\exp\left\{-iht\int\sigma(x)dx\right\} being the correspondence of 𝒰D(t)\mathcal{U}_{D}(t) in the scaling limit. Solving this, we can obtain

|ψs(t)=𝒰D,s1(t)eiE8t|ψs(0),|\psi_{s}(t)\rangle=\mathcal{U}_{D,s}^{-1}(t)e^{-i\mathcal{H}_{E_{8}}t}|\psi_{s}(0)\rangle, (20)

where |ψs(0)|\psi_{s}(0)\rangle is the ground state of the driven Ising chain at time t=0t=0 in the scaling limit.

Appendix C B. post-quench perturbation framework and the singularity regularization

In this part, we provide the necessary formalism for solving the overlap between the initial state and complete quantum E8E_{8} basis via perturbation theory, known as the post-quench framework [42]. The detailed derivations and results are well organized and performed in [42], and here we just quote the necessary results for our calculations. The post-quench overlap between the single-E8E_{8}-particle basis and the perturbed wave function is given by,

i={0}|ψ(0)ai=hFiσmai2+h2(j=18FjσFijσ(iπ,0)mai2mbj2+higher form factor contributions)+O(h3),\mathcal{F}_{i}={}_{a_{i}}\langle\{0\}|\psi(0)\rangle=h\frac{F^{\sigma}_{i}}{m_{a_{i}}^{2}}+h^{2}\left(\sum_{j=1}^{8}\frac{F_{j}^{\sigma}F_{ij}^{\sigma}(i\pi,0)}{m_{a_{i}}^{2}m_{b_{j}}^{2}}+\text{higher form factor contributions}\right)+O(h^{3}), (21)

where |{0}|\{0\}\rangle implies that this single particle state carries 0 momentum trivially due to the momentum constraint in the perturbed wave function. The form factor Fa1,,an𝒪F^{\mathcal{O}}_{a_{1},...,a_{n}} is given by the definition as

Fa1,,an𝒪=0|𝒪|θ1,,θna1,,an.F^{\mathcal{O}}_{a_{1},...,a_{n}}=\langle 0|\mathcal{O}|\theta_{1},...,\theta_{n}\rangle_{a_{1},...,a_{n}}. (22)

The basic E8E_{8} form factor theory and a complete form factor bootstrapping process can be found in many previous works, for instance, Ref. [28, 42, 31], etc. The post-quench overlap between the two-E8E_{8}-particle basis and the perturbed wave function is given by,

ij=dθ2π{θ,θij}|ψ(0)ai,aj=dθ2πKij(θ,θij),\mathcal{F}_{ij}=\int\frac{d\theta}{2\pi}{}_{a_{i},a_{j}}\langle\{\theta,\theta_{ij}\}|\psi(0)\rangle=\int\frac{d\theta}{2\pi}K_{ij}(\theta,\theta_{ij}), (23)

where θij\theta_{ij} is given by the momentum conservation misinhθ+mjsinhθij=0m_{i}\sinh\theta+m_{j}\sinh\theta_{ij}=0. The explicit form of Kij(θ,θij)K_{ij}(\theta,\theta_{ij}) is

Kij(θ,θij)=hFijσ(θ,θij)Cij(θ,θij)+O(h2),K_{ij}(\theta,\theta_{ij})=h\frac{F^{\sigma*}_{ij}(\theta,\theta_{ij})}{C_{ij}(\theta,\theta_{ij})}+O(h^{2}), (24)

with

Cij(θ,θij)=(micoshθ+mjcoshθij)mjcoshθij.C_{ij}(\theta,\theta_{ij})=(m_{i}\cosh\theta+m_{j}\cosh\theta_{ij})m_{j}\cosh\theta_{ij}. (25)

Appendix D C. Calculation of the time-dependent magnetization

In the scaling limit, the time-dependent magnetization α(t)\mathscr{M}^{\alpha}(t)s turn to be

ε(t)\displaystyle\mathscr{M}^{\varepsilon}(t) =cos(ω0t)𝒟ε(t)+sin(ω0t)𝒟y(t),\displaystyle=\cos(\omega_{0}t)\mathscr{D}^{\varepsilon}(t)+\sin(\omega_{0}t)\mathscr{D}^{y}(t), (26)
y(t)\displaystyle\mathscr{M}^{y}(t) =cos(ω0t)𝒟y(t)sin(ω0t)𝒟ε(t),\displaystyle=\cos(\omega_{0}t)\mathscr{D}^{y}(t)-\sin(\omega_{0}t)\mathscr{D}^{\varepsilon}(t),
σ(t)\displaystyle\mathscr{M}^{\sigma}(t) =𝒟σ(t),\displaystyle=\mathscr{D}^{\sigma}(t),

where 𝒟𝒪(t)\mathscr{D}^{\mathcal{O}}(t)s are given by

𝒟𝒪(t)=r,s=0ψs(0)|rr|eiE8t𝒪eiE8t|ss|ψs(0).\mathscr{D}^{\mathcal{O}}(t)=\sum_{r,s=0}^{\infty}\langle\psi_{s}(0)|r\rangle\langle r|e^{i\mathcal{H}_{E_{8}}t}\mathcal{O}e^{-i\mathcal{H}_{E_{8}}t}|s\rangle\langle s|\psi_{s}(0)\rangle. (27)

The r=0|rr|\sum_{r=0}^{\infty}|r\rangle\langle r| and s=0|ss|\sum_{s=0}^{\infty}|s\rangle\langle s| are two E8E_{8} complete bases labeled by the total account of the particles included in the basis. It is proved that in the post-quench overlap, only single- and two-particle basis will make significant contributions [42], thus in the following calculation we only include the E8E_{8} complete bases with total particle number up to 2. Since ε\varepsilon and σ\sigma are the primary fields of Ising conformal field theory, the exact form factors can be determined using the form factor bootstrap [28, 42, 31]. Plugging the above post-quench overlap formulas into Eq. (26) and Eq. (27), the results are given by

𝒟ε(t)=r,s=02Arsε(t),𝒟σ(t)=r,s=02Arsσ(t),\mathscr{D}^{\varepsilon}(t)=\sum_{r,s=0}^{2}A^{\varepsilon}_{rs}(t),\mathscr{D}^{\sigma}(t)=\sum_{r,s=0}^{2}A^{\sigma}_{rs}(t), (28)

with the detailed components,

A00ε(t)\displaystyle A_{00}^{\varepsilon}(t) =ε,A00σ(t)=σ,\displaystyle=\langle\varepsilon\rangle,A_{00}^{\sigma}(t)=\langle\sigma\rangle, (29)
A01ε(t)\displaystyle A_{01}^{\varepsilon}(t) =ieimitFiεi,A01σ(t)=ieimitFiσi,\displaystyle=\sum_{i}e^{-im_{i}t}F^{\varepsilon}_{i}\mathcal{F}_{i},A_{01}^{\sigma}(t)=\sum_{i}e^{-im_{i}t}F^{\sigma}_{i}\mathcal{F}_{i},
A10ε(t)\displaystyle A_{10}^{\varepsilon}(t) =A01ε(t),A10σ(t)=A01σ(t),\displaystyle=A_{01}^{\varepsilon*}(t),A_{10}^{\sigma}(t)=A_{01}^{\sigma*}(t),
A11ε(t)\displaystyle A_{11}^{\varepsilon}(t) =i,jei(mimj)tFijε(iπ,0)ij,A11σ(t)=i,jei(mimj)tFijσ(iπ,0)ij,\displaystyle=\sum_{i,j}e^{i(m_{i}-m_{j})t}F^{\varepsilon}_{ij}(i\pi,0)\mathcal{F}^{*}_{i}\mathcal{F}_{j},A_{11}^{\sigma}(t)=\sum_{i,j}e^{i(m_{i}-m_{j})t}F^{\sigma}_{ij}(i\pi,0)\mathcal{F}^{*}_{i}\mathcal{F}_{j},
A02ε(t)\displaystyle A_{02}^{\varepsilon}(t) =a,bdθ2πeiEabtFabε(θa,θab)Kabσ(θa,θab),A02σ(t)=a,bdθ2πeiEabtFabσ(θa,θab)Kabσ(θa,θab),\displaystyle=\sum_{a,b}\int\frac{d\theta}{2\pi}e^{-iE_{ab}t}F^{\varepsilon}_{ab}(\theta_{a},\theta_{ab})K^{\sigma}_{ab}(\theta_{a},\theta_{ab}),A_{02}^{\sigma}(t)=\sum_{a,b}\int\frac{d\theta}{2\pi}e^{-iE_{ab}t}F^{\sigma}_{ab}(\theta_{a},\theta_{ab})K^{\sigma}_{ab}(\theta_{a},\theta_{ab}),
A20ε(t)\displaystyle A_{20}^{\varepsilon}(t) =A02ε(t),A20σ(t)=A02σ(t),\displaystyle=A_{02}^{\varepsilon*}(t),A_{20}^{\sigma}(t)=A_{02}^{\sigma*}(t),
A12ε(t)\displaystyle A_{12}^{\varepsilon}(t) =r,a,bdθa2πei(mrEab)tFrabε(iπ,θa,θab)rKabσ(θa,θab),\displaystyle=\sum_{r,a,b}\int\frac{d\theta_{a}}{2\pi}e^{i(m_{r}-E_{ab})t}F^{\varepsilon}_{rab}(i\pi,\theta_{a},\theta_{ab})\mathcal{F}^{*}_{r}K^{\sigma}_{ab}(\theta_{a},\theta_{ab}),
A12σ(t)\displaystyle A_{12}^{\sigma}(t) =r,a,bdθa2πei(mrEab)tFrabσ(iπ,θa,θab)rKabσ(θa,θab),\displaystyle=\sum_{r,a,b}\int\frac{d\theta_{a}}{2\pi}e^{i(m_{r}-E_{ab})t}F^{\sigma}_{rab}(i\pi,\theta_{a},\theta_{ab})\mathcal{F}^{*}_{r}K^{\sigma}_{ab}(\theta_{a},\theta_{ab}),
A21ε(t)\displaystyle A_{21}^{\varepsilon}(t) =A12ε(t),A21σ(t)=A12σ(t),\displaystyle=A_{12}^{\varepsilon*}(t),A_{21}^{\sigma}(t)=A_{12}^{\sigma*}(t),
A22ε(t)\displaystyle A_{22}^{\varepsilon}(t) =r,s,a,bdθr2πdθa2πei(ErsEab)tFrsabε(θr+iπ,θrs+iπ,θa,θab)Krsσ(θr,θrs)Kabσ(θa,θab),\displaystyle=\sum_{r,s,a,b}\int\frac{d\theta_{r}}{2\pi}\int\frac{d\theta_{a}}{2\pi}e^{i(E_{rs}-E_{ab})t}F^{\varepsilon}_{rsab}(\theta_{r}+i\pi,\theta_{rs}+i\pi,\theta_{a},\theta_{ab})K^{\sigma*}_{rs}(\theta_{r},\theta_{rs})K^{\sigma}_{ab}(\theta_{a},\theta_{ab}),
A22σ(t)\displaystyle A_{22}^{\sigma}(t) =r,s,a,bdθr2πdθa2πei(ErsEab)tFrsabσ(θr+iπ,θrs+iπ,θa,θab)Krsσ(θr,θrs)Kabσ(θa,θab).\displaystyle=\sum_{r,s,a,b}\int\frac{d\theta_{r}}{2\pi}\int\frac{d\theta_{a}}{2\pi}e^{i(E_{rs}-E_{ab})t}F^{\sigma}_{rsab}(\theta_{r}+i\pi,\theta_{rs}+i\pi,\theta_{a},\theta_{ab})K^{\sigma*}_{rs}(\theta_{r},\theta_{rs})K^{\sigma}_{ab}(\theta_{a},\theta_{ab}).

The vacuum expectation values ε\langle\varepsilon\rangle and σ\langle\sigma\rangle are related with the rotation frequency ω0\omega_{0}, and can be obtained from the iTEBD simulation. mim_{i}s are the mass of the quantum E8E_{8} particles in unit of the mass of the lightest particle m1m_{1}, where m1m_{1} can be determined through the scaling relation m1=Cω08/15m_{1}=C\omega_{0}^{8/15}. Eab(rs)E_{ab(rs)} are the total energy of a two particle complete basis, given by Eab(rs)=ma(r)coshθa(r)+mb(s)coshθab(rs)E_{ab(rs)}=m_{a(r)}\cosh\theta_{a(r)}+m_{b(s)}\cosh\theta_{ab(rs)}, together with the momentum conservation 0=ma(r)sinhθa(r)+mb(s)sinhθab(rs)0=m_{a(r)}\sinh\theta_{a(r)}+m_{b(s)}\sinh\theta_{ab(rs)}. The only difficulty would then be the calculation of 𝒟y(t)\mathscr{D}^{y}(t). By considering the lattice formalism, we have,

dDz(t)dt=ddtψ(0)|eiHefftσzeiHefft|ψ(0)=iψ(0)|eiHefft[Heff,σz]eiHefft|ψ(0)=2Dy(t).\frac{dD^{z}(t)}{dt}=\frac{d}{dt}\langle\psi(0)|e^{iH_{\text{eff}}t}\sigma^{z}e^{-iH_{\text{eff}}t}|\psi(0)\rangle=i\langle\psi(0)|e^{iH_{\text{eff}}t}\left[H_{\text{eff}},\sigma^{z}\right]e^{-iH_{\text{eff}}t}|\psi(0)\rangle=2D^{y}(t). (30)

Sending the system back to the scaling limit, we have

𝒟y(t)=12d𝒟σ(t)dt.\mathscr{D}^{y}(t)=\frac{1}{2}\frac{d\mathscr{D}^{\sigma}(t)}{dt}. (31)

Appendix E D. Correspondence between lattice formalism and field theory

In the quantum E8E_{8} field theory as Eq. (18), an exact scaling relation between the longitudinal field deformation hh and the mass of the lightest E8E_{8} particle m1m_{1} is given by m1=Csh8/15m_{1}=C_{s}h^{8/15}, obtained by V. A. Fateev in 1994 [56] for the first time, where the exact value of hh is given as h=4.40490858h=4.40490858.... Such scaling relation also holds for the lattice Hamiltonian as the small longitudinal field perturbed critical transverse field quantum Ising chain [Eq. (17)], given by m1=Cωhz8/15m_{1}=C_{\omega}h^{8/15}_{z}. Here hz=ω0/2h_{z}=\omega_{0}/2 for Eq. (17), and the coefficient CωC_{\omega} differs with CsC_{s} due to the essential lattice formalism. For solving the new coefficient CωC_{\omega}, let’s start from a more general lattice Hamiltonian,

lat=Ji(σizσi+1z+gσix+hzσiz),\mathcal{H}_{\text{lat}}=-J\sum_{i}\left(\sigma^{z}_{i}\sigma^{z}_{i+1}+g\sigma^{x}_{i}+h_{z}\sigma^{z}_{i}\right), (32)

with g1g\lesssim 1 and hz0h_{z}\rightarrow 0. Taking the scaling limit, the corresponding field theory Hamiltonian reads

s=Hc=1/2+(1g)ϵ(x)𝑑x+hσ(x)𝑑x.\mathcal{H}_{s}=H_{c=1/2}+(1-g)\int\epsilon(x)dx+h\int\sigma(x)dx. (33)

For connecting the longitudinal field deformation between the lattice formalism and the field theory, we consider the correspondence between the two expectation values as

hσ(x)𝑑xJhziσiz.\langle h\int\sigma(x)dx\rangle\Leftrightarrow\langle Jh_{z}\sum_{i}\sigma_{i}^{z}\rangle. (34)

For the field theory expectation value,

hσ(x)𝑑x=hLσ¯=hLs¯M1/8,\langle h\int\sigma(x)dx\rangle=hL\bar{\sigma}=hL\bar{s}M^{1/8}, (35)

where L=NaL=Na\rightarrow\infty is the length of the system, with aa being the lattice spacing. σ¯\bar{\sigma} is the expectation value of the spin density operator σ\sigma, M=2J(1g)M=2J(1-g), and s¯=21/12e1/8A3/2\bar{s}=2^{1/12}e^{-1/8}A^{3/2}, A=1.2824271291A=1.2824271291... known as Glaisher’s constant [zam2003]. During the process for taking the scaling limit, we have chosen the convention as 2Ja=12Ja=1. For the lattice expectation value, since σz=(1g2)1/8\langle\sigma^{z}\rangle=(1-g^{2})^{1/8} [25], we have

Jhziσiz=JhzNσz=JhzN(1g2)1/8.\langle Jh_{z}\sum_{i}\sigma_{i}^{z}\rangle=Jh_{z}N\langle\sigma^{z}\rangle=Jh_{z}N(1-g^{2})^{1/8}. (36)

Combining Eq. (35) and Eq. (36), we obtain the correspondence between hh and hzh_{z},

h=2s¯J15/8hz.h=\frac{2}{\bar{s}}J^{15/8}h_{z}. (37)

Considering the correspondence of the scaling relation, we have m1=Clhz8/15=Csh8/15m_{1}=C_{l}h^{8/15}_{z}=C_{s}h^{8/15}, thus

Cl=CsJ(2s¯)8/15=5.4154.C_{l}=C_{s}J(\frac{2}{\bar{s}})^{8/15}=5.4154...~{}. (38)

Since hz=ω0/2h_{z}=\omega_{0}/2, for the lattice Hamiltonian Eq. (17), the scaling relation is give by

m1=Cωω08/15m_{1}=C_{\omega}\omega^{8/15}_{0} (39)

with Cω=Cl/28/15=3.7418C_{\omega}=C_{l}/2^{8/15}=3.7418... . Similarly, in the quantum E8E_{8} field theory, it was determined that [28]

σ=1.27758h1/15.\langle\sigma\rangle=1.27758...h^{1/15}. (40)

On the lattice, we can determine

σz=s¯1J1/8σ=s¯1J1/8(2s¯J15/8ω02)1/15=Cσω01/15\langle\sigma^{z}\rangle=\bar{s}^{-1}J^{-1/8}\langle\sigma\rangle=\bar{s}^{-1}J^{-1/8}\left(\frac{2}{\bar{s}}J^{15/8}\frac{\omega_{0}}{2}\right)^{1/15}=C_{\sigma}{\omega_{0}}^{1/15} (41)

with Cσ=0.9219C_{\sigma}=0.9219... . The scaling relation of σx\langle\sigma^{x}\rangle with ω0\omega_{0} can only be determined with iTEBD algorithm, as

σx=σxω0=0Cεω08/15,\langle\sigma^{x}\rangle=\langle\sigma^{x}\rangle_{\omega_{0}=0}-C_{\varepsilon}\omega_{0}^{8/15}, (42)

where σxω0=0=2/π\langle\sigma^{x}\rangle_{\omega_{0}=0}=2/\pi can be determined analytically [24, 57], and Cε=0.2342C_{\varepsilon}=0.2342... is fixed by iTEBD algorithm. The results of Eq. (39), Eq. (41) and Eq. (42) are used for normalization of the field theory results and comparison with lattice simulations.

Appendix F E. Performing iTEBD simulation

For verification of the analytical results obtained via field theory and post-quench perturbation approach, we use infinite time-evolving block decimation (iTEBD) algorithm to realize the lattice simulation for both the original lattice model and the Rydberg lattice model. The iTEBD algorithm can simulate the ground state wave function of an infinite large one dimensional quantum system and calculate time evolution process or correlation function of certain operators. Following the lattice formalism,

Mα(t)=ψ(0)|eiHefft𝒪latα(t)eiHefft|ψ(0),α=x,y,z.M^{\alpha}(t)=\langle\psi(0)|e^{iH_{\text{eff}}t}\mathcal{O}_{\text{lat}}^{\alpha}(t)e^{-iH_{\text{eff}}t}|\psi(0)\rangle,\alpha=x,y,z. (43)

The effective Hamiltonian is given by Eq. (17), and the corresponding operators 𝒪latα\mathcal{O}_{\text{lat}}^{\alpha} is given as following,

𝒪latx(t)\displaystyle\mathcal{O}_{\text{lat}}^{x}(t) =cosω0tσx+sinω0tσy,\displaystyle=\cos{\omega_{0}t}\sigma^{x}+\sin{\omega_{0}t}\sigma^{y}, (44)
𝒪laty(t)\displaystyle\mathcal{O}_{\text{lat}}^{y}(t) =cosω0tσysinω0tσx,\displaystyle=\cos{\omega_{0}t}\sigma^{y}-\sin{\omega_{0}t}\sigma^{x},
𝒪latz(t)\displaystyle\mathcal{O}_{\text{lat}}^{z}(t) =σz.\displaystyle=\sigma^{z}.

For performing the lattice simulation of the time-evolution process, we first generate |ψ(0)|\psi(0)\rangle, the ground state of a critical transverse field quantum Ising chain, with bond dimension χ=64\chi=64, and the convergence of the singular values being lower than 101110^{-11}. Then we use the effective lattice Hamiltonian with a small longitudinal field h=0.015h=0.015 (corresponding to the rotating frequency being ω0=0.03\omega_{0}=0.03) to do time evolution process on this wave function. and calculate the expectation values of 𝒪latα\mathcal{O}_{\text{lat}}^{\alpha} at every time point, with a fifth order Trotter-Suzuki decomposition for total time t=1000000t=1000000 and dt=0.1dt=0.1. Furthermore, the vacuum expectation values of σz\langle\sigma^{z}\rangle and σx\langle\sigma^{x}\rangle of the ground state of the effective Hamiltonian Eq. (17) are also used for the normalization of σ\langle\sigma\rangle and ε\langle\varepsilon\rangle in the quantum E8E_{8} field theory.

Appendix G F. Theoretical proof of the singular peaks

Since the periodically driven time evolution process starts at t=0t=0, for better revealing the low energy collective excitations of this non-equilibrium system, and comparing the analytical results with the numerical simulation, we apply a one-sided Fourier transformation to both the analytical and numerical data. By defining 𝒟𝒪(ω)=0𝑑teiωt𝒟𝒪(t)\mathscr{D}^{\mathcal{O}}(\omega)=\int_{0}^{\infty}dte^{i\omega t}\mathscr{D}^{\mathcal{O}}(t) and recalling the relation between 𝒟y(t)\mathscr{D}^{y}(t) and 𝒟σ(t)\mathscr{D}^{\sigma}(t) as Eq. (31), it is straightforward to obtain that 𝒟y(ω)=iω𝒟σ(ω)/2\mathscr{D}^{y}(\omega)=-i\omega\mathscr{D}^{\sigma}(\omega)/2. Then, for the analytical results Eq. (26), after the one-sided Fourier transformation, we have

ε(ω)\displaystyle\mathscr{M}^{\varepsilon}(\omega) =12[𝒟ε(ω+ω0)+𝒟ε(ωω0)]14[(ω+ω0)𝒟σ(ω+ω0)(ωω0)𝒟σ(ωω0)],\displaystyle=\frac{1}{2}\left[\mathscr{D}^{\varepsilon}(\omega+\omega_{0})+\mathscr{D}^{\varepsilon}(\omega-\omega_{0})\right]-\frac{1}{4}\left[(\omega+\omega_{0})\mathscr{D}^{\sigma}(\omega+\omega_{0})-(\omega-\omega_{0})\mathscr{D}^{\sigma}(\omega-\omega_{0})\right], (45)
y(ω)\displaystyle\mathscr{M}^{y}(\omega) =i2[𝒟ε(ω+ω0)𝒟ε(ωω0)]i4[(ω+ω0)𝒟σ(ω+ω0)+(ωω0)𝒟σ(ωω0)],\displaystyle=\frac{i}{2}\left[\mathscr{D}^{\varepsilon}(\omega+\omega_{0})-\mathscr{D}^{\varepsilon}(\omega-\omega_{0})\right]-\frac{i}{4}\left[(\omega+\omega_{0})\mathscr{D}^{\sigma}(\omega+\omega_{0})+(\omega-\omega_{0})\mathscr{D}^{\sigma}(\omega-\omega_{0})\right],
σ(ω)\displaystyle\mathscr{M}^{\sigma}(\omega) =𝒟σ(ω),\displaystyle=\mathscr{D}^{\sigma}(\omega),

By making use of the following expression for one-sided Fourier transformation, we prove that there are two types of singular peaks inside the whole spectra, as the delta-function type and the edge-singularity type.

0𝑑teiωtS=Slimϵ00𝑑te(iωϵ)t=limϵ0iSω+iϵ=𝒫(iSω)+Sπδ(ω).\int_{0}^{\infty}dte^{i\omega t}S=S\lim_{\epsilon\rightarrow 0}\int_{0}^{\infty}dte^{(i\omega-\epsilon)t}=\lim_{\epsilon\rightarrow 0}\frac{iS}{\omega+i\epsilon}=\mathcal{P}\left(i\frac{S}{\omega}\right)+S\pi\delta(\omega). (46)

where SS could be some constants or functions that do not have tt dependence. In the following we will use σ(ω)=σ(ω)\mathscr{M}^{\sigma}(\omega)=\mathscr{M}^{\sigma}(\omega) as an example to show these two types of singular peaks. ε(ω)\mathscr{M}^{\varepsilon}(\omega) and y(ω)\mathscr{M}^{y}(\omega) are given by shifting of σ(ω)\mathscr{M}^{\sigma}(\omega) with the rotating frequency ω0\omega_{0} according to Eq. (45). For the detailed analysis of the singular peaks, we take 𝒪=σ\mathcal{O}=\sigma as an example, and the form for 𝒪=ε\mathcal{O}=\varepsilon only differs with the operator.

G.1 a. The delta-function type singular peak

The delta-function type singular peaks are contributed by the particle channels of (A00σ+A01σ+A10σ+A11σ)(ω)(A^{\sigma}_{00}+A^{\sigma}_{01}+A^{\sigma}_{10}+A^{\sigma}_{11})(\omega), where A(ω)=0𝑑teiωtA(t)A(\omega)=\int_{0}^{\infty}dte^{i\omega t}A(t). More explicitly, for A00σ(ω)A^{\sigma}_{00}(\omega),

A00σ(ω)=𝒫(iσω)+σπδ(ω),A^{\sigma}_{00}(\omega)=\mathcal{P}\left(i\frac{\langle\sigma\rangle}{\omega}\right)+\langle\sigma\rangle\pi\delta(\omega), (47)

which trivially contributes to the delta-function peak at ω=0\omega=0. For A01σ(ω)A^{\sigma}_{01}(\omega) and A10σ(ω)A^{\sigma}_{10}(\omega),

A01σ(ω)=i𝒫(iFiσiωmi)+Fiσiπδ(ωmi),A10σ(ω)=i𝒫(iFiσiω+mi)+Fiσiπδ(ω+mi).A^{\sigma}_{01}(\omega)=\sum_{i}\mathcal{P}\left(i\frac{F^{\sigma}_{i}\mathcal{F}_{i}}{\omega-m_{i}}\right)+F^{\sigma}_{i}\mathcal{F}_{i}\pi\delta(\omega-m_{i}),A^{\sigma}_{10}(\omega)=\sum_{i}\mathcal{P}\left(i\frac{F^{\sigma}_{i}\mathcal{F}_{i}}{\omega+m_{i}}\right)+F^{\sigma}_{i}\mathcal{F}_{i}\pi\delta(\omega+m_{i}). (48)

They give rise to the eight delta-function peaks with the absolute frequencies corresponding to the mass of the eight single E8E_{8} particles, both in the frequency region of ω>0\omega>0 and ω<0\omega<0. For A11σ(ω)A^{\sigma}_{11}(\omega),

A11σ(ω)=i,j𝒫(iFijσ(iπ,0)ijω+mimj)+Fijσ(iπ,0)ijπδ(ω+mimj).A^{\sigma}_{11}(\omega)=\sum_{i,j}\mathcal{P}\left(i\frac{F^{\sigma}_{ij}(i\pi,0)\mathcal{F}^{*}_{i}\mathcal{F}_{j}}{\omega+m_{i}-m_{j}}\right)+F^{\sigma}_{ij}(i\pi,0)\mathcal{F}^{*}_{i}\mathcal{F}_{j}\pi\delta(\omega+m_{i}-m_{j}). (49)

For the case i=ji=j, i.e., the included single E8E_{8} particles being the same in the two complete bases, it will contribute to the spectral weight for the delta-function singular peak at ω=0\omega=0. For the case iji\neq j, i.e., different single E8E_{8} particles being included in the two complete bases, it will give rise to exotic delta-function singular peaks at ω=mjmi\omega=m_{j}-m_{i}. All the delta-function singular peaks are the dominant structures that can be found in both the analytical and numerical results, due to the fast decaying behavior of the form factors.

G.2 b. The edge-singularity type peak of the Van Hove singularity

The edge-singularity type singular peaks of Van Hove’s singularity are contributed by the other particle channels as (A02σ+A12σ)(ω)(A^{\sigma}_{02}+A^{\sigma}_{12})(\omega) (Here only the case of ω>0\omega>0 are considered for simplicity). It is essentially due to the divergence of the density of states near the threshold of the spectra in the frequency domain, known as Van Hove singularity [31, 53]. Taking A02σ(ω)A^{\sigma}_{02}(\omega) as an example, we have

A02σ(ω)=a,b𝒫(idθa2πFabσ(θa,θab)Kabσ(θa,θab)ωEab)+dθa2Fabσ(θa,θab)Kabσ(θa,θab)δ(ωEab),A^{\sigma}_{02}(\omega)=\sum_{a,b}\mathcal{P}\left(i\int\frac{d\theta_{a}}{2\pi}\frac{F^{\sigma}_{ab}(\theta_{a},\theta_{ab})K^{\sigma}_{ab}(\theta_{a},\theta_{ab})}{\omega-E_{ab}}\right)+\int\frac{d\theta_{a}}{2}F^{\sigma}_{ab}(\theta_{a},\theta_{ab})K^{\sigma}_{ab}(\theta_{a},\theta_{ab})\delta(\omega-E_{ab}), (50)

where Eab=macoshθa+mbcoshθabE_{ab}=m_{a}\cosh\theta_{a}+m_{b}\cosh\theta_{ab}. Integrating out the delta function, we finally obtain that

A02σ(ω)=a,b𝒫(idθa2πFabσ(θa,θab)Kabσ(θa,θab)ωEab)+12Fabσ(θa,θab)mamb|sinh(θaθab)|Kabσ(θa,θab).A^{\sigma}_{02}(\omega)=\sum_{a,b}\mathcal{P}\left(i\int\frac{d\theta_{a}}{2\pi}\frac{F^{\sigma}_{ab}(\theta_{a},\theta_{ab})K^{\sigma}_{ab}(\theta_{a},\theta_{ab})}{\omega-E_{ab}}\right)+\frac{1}{2}\frac{F^{\sigma}_{ab}(\theta_{a},\theta_{ab})}{m_{a}m_{b}|\sinh(\theta_{a}-\theta_{ab})|}K^{\sigma}_{ab}(\theta_{a},\theta_{ab}). (51)

When aba\neq b, i.e., the two E8E_{8} particles included in the complete bases are different, it is proved that 1/|sinh(θaθab)|1/ω(ma+mb)1/|\sinh(\theta_{a}-\theta_{ab})|\approx 1/\sqrt{\omega-(m_{a}+m_{b})} under the additional energy conservation of ω=macoshθa+mbcoshθab\omega=m_{a}\cosh\theta_{a}+m_{b}\cosh\theta_{ab} [31]. Recalling the explicit form of Eq. (24), in the denominator of Kabσ(θa,θab)K^{\sigma}_{ab}(\theta_{a},\theta_{ab}) we have Cij(θ,θij)=(micoshθ+mjcoshθij)mjcoshθij=ωmjcoshθijC_{ij}(\theta,\theta_{ij})=(m_{i}\cosh\theta+m_{j}\cosh\theta_{ij})m_{j}\cosh\theta_{ij}=\omega m_{j}\cosh\theta_{ij}, which gives the final divergence behavior of the density of states as 1/ωω(ma+mb)1/\omega\sqrt{\omega-(m_{a}+m_{b})} near the threshold of ω=ma+mb\omega=m_{a}+m_{b}. Thus for A02σ(ω)A^{\sigma}_{02}(\omega), when the two particles inside the complete bases are different, they give rise to an edge-singularity type of singular peaks with the singular behavior 1/ωω(ma+mb)1/\omega\sqrt{\omega-(m_{a}+m_{b})} at the threshold ω=ma+mb\omega=m_{a}+m_{b}.

For A12σ(ω)A^{\sigma}_{12}(\omega), the explicit result is given by

A12σ(ω)=r,a,b𝒫(idθa2πFrabσ(iπ,θa,θab)rKabσ(θa,θab)ω+mrEab)+12Frabσ(iπ,θa,θab)rmamb|sinh(θaθab)|Kabσ(θa,θab),A^{\sigma}_{12}(\omega)=\sum_{r,a,b}\mathcal{P}\left(i\int\frac{d\theta_{a}}{2\pi}\frac{F^{\sigma}_{rab}(i\pi,\theta_{a},\theta_{ab})\mathcal{F}^{*}_{r}K^{\sigma}_{ab}(\theta_{a},\theta_{ab})}{\omega+m_{r}-E_{ab}}\right)+\frac{1}{2}\frac{F^{\sigma}_{rab}(i\pi,\theta_{a},\theta_{ab})\mathcal{F}^{*}_{r}}{m_{a}m_{b}|\sinh(\theta_{a}-\theta_{ab})|}K^{\sigma}_{ab}(\theta_{a},\theta_{ab}), (52)

where the conservation of energy is given by ω+mr=macoshθa+mbcoshθab\omega+m_{r}=m_{a}\cosh\theta_{a}+m_{b}\cosh\theta_{ab}. For the case when rabr\neq a\neq b, by changing the variable as ωω+mr\omega\rightarrow\omega+m_{r}, it is straightforward to show that if ab,ma+mb>mra\neq b,~{}m_{a}+m_{b}>m_{r}, the edge-singularity behavior as 1/(ω+mr)ω+mr(ma+mb)1/(\omega+m_{r})\sqrt{\omega+m_{r}-(m_{a}+m_{b})} at the threshold ω=ma+mbmr\omega=m_{a}+m_{b}-m_{r}.

G.3 c. The edge-singularity type singular peak as kinematic poles in form factor

For the case of aba\neq b and r=ar=a (or bb) in A12σ(ω)A^{\sigma}_{12}(\omega), additional edge-singular behavior comes from the form factor Frabσ(iπ,θa,θab)=Frrbσ(iπ,θr,θrb)F^{\sigma}_{rab}(i\pi,\theta_{a},\theta_{ab})=F^{\sigma}_{rrb}(i\pi,\theta_{r},\theta_{rb}). Recalling the kinematic singularities of the form factor [28, 42, 31, 53],

ilimθaθb(θaθb)Fa,b,a1,,an𝒪(θa+iπ,θb,θa1,,θan)=(1nSb,ai(θbθai))Fa1,,an𝒪(θa1,,θan),-i\lim_{\theta_{a}\rightarrow\theta_{b}}(\theta_{a}-\theta_{b})F^{\mathcal{O}}_{a,b,a_{1},...,a_{n}}(\theta_{a}+i\pi,\theta_{b},\theta_{a_{1}},...,\theta_{a_{n}})=\left(1-\prod_{n}S_{b,a_{i}}(\theta_{b}-\theta_{a_{i}})\right)F^{\mathcal{O}}_{a_{1},...,a_{n}}(\theta_{a_{1}},...,\theta_{a_{n}}), (53)

Frrbσ(iπ,θr,θrb)F^{\sigma}_{rrb}(i\pi,\theta_{r},\theta_{rb}) will show an edge-singularity behavior as 1/θr1/ωmb1/\theta_{r}\approx 1/\sqrt{\omega-m_{b}} near the threshold at ω=mr+mbmr=mb\omega=m_{r}+m_{b}-m_{r}=m_{b}. Together with the edge-singularity given by the Van Hove’s singularity, it seems that there will be a singular behavior of 1/ω1/\omega appears at the threshold. However, such unphysical edge-singularity peaks would not appear, since due to the exchange behavior of the form factor,

Fa1,,aj,aj+1,,an𝒪(θa1,,θj,θj+1,,θan)=Sj,j+1(θjθj+1)Fa1,,aj+1,aj,,an𝒪(θa1,,θj+1,θj,,θan),F^{\mathcal{O}}_{a_{1},...,a_{j},a_{j+1},...,a_{n}}(\theta_{a_{1}},...,\theta_{j},\theta_{j+1},...,\theta_{a_{n}})=S_{j,j+1}(\theta_{j}-\theta_{j+1})F^{\mathcal{O}}_{a_{1},...,a_{j+1},a_{j},...,a_{n}}(\theta_{a_{1}},...,\theta_{j+1},\theta_{j},...,\theta_{a_{n}}), (54)

we will have Frrbσ(iπ,θr,θrb)=Srb(θrθrb)Frbrσ(iπ,θrb,θr)F^{\sigma}_{rrb}(i\pi,\theta_{r},\theta_{rb})=S_{rb}(\theta_{r}-\theta_{rb})F^{\sigma}_{rbr}(i\pi,\theta_{rb},\theta_{r}), where Sj,j+1(θjθj+1)S_{j,j+1}(\theta_{j}-\theta_{j+1}) and Srb(θrθrb)S_{rb}(\theta_{r}-\theta_{rb}) are the S-matrices for the quantum E8E_{8} field theory. Due to the momentum conservation that mrsinhθr+mbsinhθrb=0m_{r}\sinh\theta_{r}+m_{b}\sinh\theta_{rb}=0, in the limit of θr0\theta_{r}\rightarrow 0 we also have θrb0\theta_{rb}\rightarrow 0. Since Sij(0)=1S_{ij}(0)=-1 for the S-matrices, at the threshold the pair of two edge-singularities Frrbσ(iπ,θr,θrb)+Frbrσ(iπ,θrb,θr)=0F^{\sigma}_{rrb}(i\pi,\theta_{r},\theta_{rb})+F^{\sigma}_{rbr}(i\pi,\theta_{rb},\theta_{r})=0 cancel with each other, and thus will not show up in the final results.

For A22σ(ω)A^{\sigma}_{22}(\omega), the explicit result is given by

A22σ(ω)\displaystyle A^{\sigma}_{22}(\omega) =r,s,a,b𝒫(idθr2πdθa2πFrsabσ(θr+iπ,θrs+iπ,θa,θab)Krsσ(θr,θrs)Kabσ(θa,θab)ω+ErsEab)\displaystyle=\sum_{r,s,a,b}\mathcal{P}\left(i\int\frac{d\theta_{r}}{2\pi}\frac{d\theta_{a}}{2\pi}\frac{F^{\sigma}_{rsab}(\theta_{r}+i\pi,\theta_{rs}+i\pi,\theta_{a},\theta_{ab})K^{\sigma*}_{rs}(\theta_{r},\theta_{rs})K^{\sigma}_{ab}(\theta_{a},\theta_{ab})}{\omega+E_{rs}-E_{ab}}\right) (55)
+dθr2Frsabσ(θr+iπ,θrs+iπ,θa,θab)mambms|coshθssinh(θaθab)|Krsσ(θr,θrs)Kabσ(θa,θab),\displaystyle+\int\frac{d\theta_{r}}{2}\frac{F^{\sigma}_{rsab}(\theta_{r}+i\pi,\theta_{rs}+i\pi,\theta_{a},\theta_{ab})}{m_{a}m_{b}m_{s}|\cosh\theta_{s}\sinh(\theta_{a}-\theta_{ab})|}K^{\sigma*}_{rs}(\theta_{r},\theta_{rs})K^{\sigma}_{ab}(\theta_{a},\theta_{ab}),

with the energy conservation given by ω=macoshθa+mbcoshθab(mrcoshθr+mscoshθrs)\omega=m_{a}\cosh\theta_{a}+m_{b}\cosh\theta_{ab}-(m_{r}\cosh\theta_{r}+m_{s}\cosh\theta_{rs}). It seems that the integration will cancel the Van Hove’s singularity. However, according to the kinematic singularities, in the case of r=ar=a, (or bb, or s=as=a, or bb), additional edge-singularity arise at the threshold of θr=θaω0=mbms\theta_{r}=\theta_{a}\rightarrow\omega_{0}=m_{b}-m_{s} as 1/(θrθa)1/ω+msmb1/(\theta_{r}-\theta_{a})\approx 1/\sqrt{\omega+m_{s}-m_{b}}. Such edge-singularities cannot be canceled with the exchange property, due to the varying range of θr\theta_{r} and θa\theta_{a}. With considering the Van Hove’s singularity, after integration, in the case of r=ar=a and mbmsm_{b}\geq m_{s} the final edge-singularity behavior is given by a ln(ωω0)\ln(\omega-\omega_{0}) at the threshold ω0=mbms\omega_{0}=m_{b}-m_{s}, which enhances the delta-function singular peaks at the same frequency.