---
title: Kicked Field Ising Model Overview
url: https://www.emergentmind.com/topics/kicked-field-ising-model
type: topic
---

# Kicked Field Ising Model Overview

The kicked field Ising model denotes a family of periodically driven Ising spin systems in which Ising interactions and magnetic-field rotations are applied in separate stroboscopic layers. In its standard spin-\(\tfrac12\) one-dimensional form, the dynamics is generated by a time-dependent Hamiltonian \(H(t)=H_I+H_K\sum_{\tau\in\mathbb Z}\delta(t-\tau)\), so that the one-period Floquet operator factorizes into an interaction layer and a kick layer [2606.13857]. Closely related formulations appear as the kicked Ising chain, the transverse-field kicked Ising chain, the kicked transverse-field Ising model, and dual-unitary kicked chains; the common structure is a Floquet unitary built from noncommuting Ising and field terms [1602.07130]. Across these formulations, the model functions as a paradigmatic setting for quantum chaos, duality, light-cone correlation dynamics, entanglement production, localization diagnostics, and hardware-level quantum simulation [2506.15816].

## 1. Canonical formulations and model families

In the standard nearest-neighbor chain, the interaction and kick terms are
\[
H_I = J \sum_{n=1}^N \sigma_n^z \sigma_{n+1}^z + h_z \sum_{n=1}^N \sigma_n^z,\qquad
H_K = h_x \sum_{n=1}^N \sigma_n^x,
\]
with periodic boundary conditions in the canonical ring geometry [2606.13857]. Other papers use the equivalent Floquet decomposition
\[
U_N = U_I(J)U_K(b,\phi),
\]
where
\[
U_I(J)=\exp\!\left(-iJ\sum_{i=1}^N \sigma_i^z\sigma_{i+1}^z\right),\qquad
U_K(b,\phi)=\exp\!\left(-i\sum_{i=1}^N[b^x\sigma_i^x+b^z\sigma_i^z]\right),
\]
with \(b^x=b\sin\phi\) and \(b^z=b\cos\phi\) [1602.07130]. The order convention varies across the literature: some works write \(U=U_KU_I\), others \(U=U_IU_K\), reflecting different stroboscopic conventions rather than a different physical content [2606.13857][2004.08386].

The same Floquet architecture extends beyond the nearest-neighbor ring. Two-dimensional square-lattice and heavy-hex versions use commuting \(ZZ\) gates on graph edges followed by uniform or site-dependent \(X\)-rotations [1409.2915][2308.03082]. Long-range dual-unitary chains replace the nearest-neighbor Ising term by a distance-\(r\) coupling together with local functions on the \(r-1\) intermediate sites [2606.13857]. Fully connected collective models provide an all-to-all generalization: the kicked \(p\)-spin family has
\[
U_p(\tau)=\exp\!\left[- i \frac{k}{pJ^{p-1}} J_z^p\right]\exp(-i\alpha J_y),
\]
and reduces to the quantum kicked top at \(p=2\) [2103.00748].

| Setting | Representative Floquet structure | Salient feature |
|---|---|---|
| 1D nearest-neighbor chain | \(U_F=e^{-iH_I}e^{-iH_K}\) | canonical KIC/KFIM [1602.07130] |
| Long-range dual-unitary chain | \(U=U_K[g]U_I[h,\vec f]\) | exact dual-unitarity for range \(r>1\) [2606.13857] |
| Heavy-hex kicked TFIM | \(\prod_{\langle i,j\rangle} e^{\,i\pi Z_iZ_j/4}\prod_i e^{-i\theta_h X_i/2}\) | IBM-oriented 2D Floquet circuit [2308.03082] |
| Collective kicked \(p\)-spin | \(\exp[-i(k/(pJ^{p-1}))J_z^p]\exp(-i\alpha J_y)\) | all-to-all transverse Ising-like extension [2103.00748] |

This variety of realizations explains why the term “kicked field Ising model” is used both narrowly for the one-dimensional ring and more broadly for a Floquet Ising class defined by alternating Ising evolution and field kicks.

## 2. Self-duality, particle–time duality, and dual-unitarity

A central structural feature of the kicked Ising chain is particle–time duality. For the homogeneous ring, the trace of the \(T\)-step Floquet evolution on \(N\) spins can be rewritten as
\[
\mathrm{Tr}\,U_N^T=\mathrm{Tr}\,\tilde U_T^N,
\]
where \(\tilde U_T\) acts on \(T\) spins and is generically nonunitary [1602.07130]. This identity converts a \(2^N\times 2^N\) trace problem into a \(2^T\times 2^T\) transfer problem, which is especially effective at fixed small \(T\) and large \(N\) [1602.07479].

The minimal \(L=2\) dual-unitary realization is the kicked Ising chain at the self-dual line
\[
J=b=\frac{\pi}{4},
\]
with arbitrary longitudinal field \(h\) [2001.01298]. At this point the model becomes dual-unitary: evolution is unitary both in time and in the dual spatial direction [2001.01298]. In standard nearest-neighbor KIC this property is tied to exact space–time symmetry, but that symmetry is highly restrictive and permits only nearest-neighbor interactions [2606.13857].

Long-range dual-unitary chains circumvent that restriction by building the kick and interaction layers from a pair of complex Hadamard matrices. In the formulation of [2606.13857], one defines
\[
U=U_K[g]\,U_I[h,\vec f],
\]
with \(U_K[g]\) a tensor product of on-site Hadamard matrices and \(U_I[h,\vec f]\) diagonal in the computational basis. For spin-\(\tfrac12\), the choice
\[
g(s,s')=h(s,s')=\frac{\pi}{4}\big[(2s-3)(2s'-3)-1\big]
\]
reproduces \(h_x=J=\pi/4\), and for arbitrary local functions \(f_n\) and arbitrary \(h_z\) the resulting Floquet unitary remains dual-unitary even for interaction range \(r>1\) [2606.13857]. The corresponding trace duality becomes
\[
\mathrm{Tr}\,U^T=\mathrm{Tr}\prod_{n=1}^{N/r}\tilde U_n,
\]
or \(\mathrm{Tr}\,U^T=\mathrm{Tr}\,\tilde U^N\) in the homogeneous case [2606.13857].

A recurring misconception is that exact space–time symmetry and dual-unitarity are synonymous. The long-range construction shows otherwise: for \(r>1\), exact space–time symmetry is lost, but dual-unitarity is retained through the Hadamard constraints [2606.13857].

## 3. Local correlations, transfer matrices, and light-cone edges

The exact treatment of local correlations is one of the most distinctive analytical features of the kicked field Ising literature. In dual-unitary nearest-neighbor chains, strictly local traceless operators have vanishing two-point correlators for sufficiently long chains:
\[
\langle \sigma_i^\alpha(t)\sigma_j^\beta(0)\rangle=0
\]
under \(N>|t|+|i-j|\) [2001.01298]. Nontrivial signals survive for operators supported on adjacent pairs, and then only on the light-cone edge \(n=t\), where the correlator reduces to a finite-dimensional transfer-matrix expression,
\[
C_T=\langle \overline{\Phi}_{q_1q_2}|\mathcal T^{T-2}|\Phi_{q_3q_4}\rangle
\]
[2001.01298].

For the self-dual kicked Ising chain, the relevant transfer matrix has spectrum \(\{1,\cos(2h),0,0\}\), leading to explicit edge correlators of the form
\[
C(n,T)=\delta_{n,T}\,\mathcal C_{\alpha\beta}^{\delta\gamma}\cos^T(2h)
\]
for specific spin-component choices [2001.01298]. This light-cone confinement is not limited to the exactly self-dual point. For the kicked Ising chain at \(b=\pi/4\) but arbitrary \(J\) and \(h\), exact edge correlators remain available beyond dual-unitarity:
\[
C_t=\mathcal C_{\alpha\beta}^{\gamma\delta}\,[\cos(2h)\sin^2(2J)]^t
\]
on the edge \(n=t\), while near-edge strictly local correlators at \(n=t-1\) acquire an additional \(\cot^2(2J)\) factor [2004.08386]. The analytical control survives because the light-cone edge still reduces to a small transfer matrix, even though correlations inside the cone need no longer vanish [2004.08386].

The long-range dual-unitary extension replaces the nearest-neighbor light cone by
\[
|n|=r|t|.
\]
For interaction range \(r\), causality constrains correlations to \(|k|\le r|t|\), and dual-unitarity collapses traceless local two-point correlators to the cone edges \(|n|=r|t|\) [2606.13857]. In the \(r=2\) construction, edge correlators take the form
\[
C(t)=\langle \overline{\Phi}_{\overline\Sigma}|\hat T^{\,t-2}|\Phi_\Sigma\rangle,
\]
with \(\hat T\) acting on a local \(L^4\times L^4\) space [2606.13857]. For spin-\(\tfrac12\), the subleading eigenvalues of \(\hat T\) set the exponential decay: in the non-homogeneous case \(\lambda_{1,2}=\cos(2J)\cos(2h)\), \(\lambda_3=\cos^2(2h)\); in the homogeneous case \(\lambda_{1,2}=\cos(2h)\cos^2(2J)\), \(\lambda_3=\cos^2(2h)\cos^2(2J)\) [2606.13857].

| Interaction range | Correlation front | Exact edge structure |
|---|---|---|
| \(r=1\) | \(|n|=|t|\) | adjacent-pair correlators on light-cone edge [2001.01298] |
| generic \(r=1\), \(b=\pi/4\) | edge \(n=t\) | exact beyond dual-unitarity [2004.08386] |
| \(r>1\) dual-unitary | \(|n|=r|t|\) | edge-only propagation for traceless local operators [2606.13857] |

These results make the KFIM unusual among interacting Floquet systems: certain real-time correlation functions are not merely numerically accessible but analytically closed.

## 4. Spectral statistics, chaos, boundaries, and localization

The kicked Ising model is also a benchmark system for many-body spectral statistics. In chaotic regimes, the dual operator \(\tilde U_T\) is generically nonunitary and its eigenvalues fill a region within the unit disk, whereas integrable limits display highly structured spectra, including collapse toward zero or concentric rings in the free-fermion transverse-field case [1602.07130]. The dual formulation explains both the oscillatory density of states and anomalous short-time behavior of the spectral form factor through the largest-modulus eigenvalues of \(\tilde U_T\) [1602.07479].

At the self-dual point, the spectral form factor admits exact thermodynamic-limit control. For periodic boundary conditions, the disorder-averaged second moment reproduces the linear ramp \(2t\), but higher moments do not belong to COE. Instead, the appropriate random-matrix description is a symmetric-space ensemble: \(Sp(N)/U(N)\) for odd \(L\) and \(O(2N)/(O(N)\times O(N))\) for even \(L\) [2009.03199]. Equivalently, the trace behaves as a real Gaussian random variable at the self-dual point with periodic boundaries, not as a complex Gaussian [2506.15816]. With open boundaries, this changes: at the same self-dual point, the trace behaves as a complex Gaussian random variable, as expected from COE universality [2506.15816]. Boundary conditions therefore affect not only finite-size details but the universality class of global spectral observables.

Another important qualification comes from the two-dimensional kicked quantum Ising model on the square lattice. There, the level spacing distribution agrees with the Wigner surmise for almost all values of parameters except where the model is essentially noninteracting, even in regions where local observables are not ergodic or where the spectral density is non-flat [1409.2915]. This directly undermines the common identification of Wigner-Dyson level statistics with local ergodicity.

Localization diagnostics provide a complementary perspective. In the disordered one-dimensional kicked Ising chain with strong longitudinal disorder and \(b_x=\pi/4\), the largest eigenvalue \(\lambda_0(t)\) of the double dual operator distinguishes localized and ergodic behavior: \(\lambda_0(t)>1\) with a gap indicates localization, while \(\lambda_0(t)\to1\) at large \(t\) indicates ergodicity [2101.10057]. The same analysis identifies a critical \(\Delta J_c\approx0.58\) and a Thouless time
\[
t_{\mathrm{Th}}\sim \frac{1}{\eta}\ln N
\]
for the spectral-form-factor diagnostic [2101.10057]. On heavy-hex hardware, a disordered Floquet kicked Ising model probed by late-time OTOCs yields an experimentally extracted many-body localization crossover at
\[
W_c\approx0.18
\]
on a 60-qubit device-compatible graph [2510.01983].

## 5. Entanglement production, coherence control, and energy-storage protocols

At special kicking angles, the KFIM supports exact statements about multipartite entanglement. For the open-chain protocol with
\[
U=\exp\!\left[-i\frac{\pi}{4}\sum_{j=1}^{L-1}\sigma_j^x\sigma_{j+1}^x\right]
\exp\!\left[-i\frac{\pi}{4}\sum_{j=1}^L \sigma_j^z\right],
\]
the bipartite entropy between equal halves obeys
\[
S_M^{\mathrm{Open}}(n)=n+(M-n)\Theta(n-M)\Theta(2M-n),
\]
so one ebit is added per kick until saturation, followed by unraveling at the same rate [1408.1677]. For periodic chains,
\[
S_M^{\mathrm{Closed}}(n)=2n+(M-2n)\Theta(n-M/2)\Theta(M-n),
\]
corresponding to two ebits per kick until saturation [1408.1677]. These states are stabilizer or cluster-like states with extensive multipartite entanglement but vanishing pairwise concurrence in the periodic case [1408.1677].

Time-dependent quench protocols reveal a different entanglement mechanism. In the kicked-field Ising chain with continuous longitudinal field and pulsed transverse field,
\[
h_x(t)=h_{x0}\cos(\alpha t),\qquad h_z(t)=h_{z0}\sin(\alpha t),
\]
a quench from large transverse field to strong longitudinal field produces multipartite entanglement close to the random-state Page value away from resonance [1305.5129]. At the commensurate condition
\[
J_z\tau=\frac{\pi}{2}m,
\]
the interaction gate becomes a phase times the identity for periodic chains, and entanglement generation freezes [1305.5129].

KFIM-based tripartite environments have been used to study decoherence control. In the nested-environment construction
\[
H(t)=H_0(t)+\lambda V_{CE}+\gamma V_{EF},
\]
increasing the near–far coupling \(\gamma\) slows decoherence of the central system, increases purity at fixed time, and delays entanglement sudden death [1512.07683]. For multiple \(E\)–\(F\) links, the empirical rescaling
\[
\gamma'=\frac{\gamma}{\sqrt{\nu}}
\]
collapses purity and concurrence trajectories across different connectivities [1512.07683].

A more recent application is the kicked-Ising quantum battery. In the self-dual operator regime
\[
J=\frac{\pi}{4},\qquad b=-\frac{\pi}{4},
\]
the injected energy can be computed exactly, and for the \(H_1^{xx}\) charger it simplifies to
\[
E_N(m)=\omega_0\sum_k \sin^2(mk)
\]
[2511.17835]. The charging pattern depends sharply on boundary conditions and parity. Under periodic boundaries, even-\(N\) \(H_1^{xx}\) chains reach \(E_N/N=1\) at half-integer multiples of \(N\), whereas odd-\(N\) chains remain at \(0.5\) except at integer multiples \(qN\), where they return to the ground state [2511.17835]. Under \(H_1^{zz}\) with open boundaries, GHZ-like states
\[
\ket{\mathrm{GHZ}^{\pm i}}=\frac{\ket{0}^{\otimes N}\pm i\,\ket{1}^{\otimes N}}{\sqrt2}
\]
appear at specified kick numbers [2511.17835]. This suggests a direct link between charging, delocalization, and kick-induced operator spreading.

## 6. Quantum hardware realizations and classical simulation

The heavy-hex kicked transverse-field Ising model has become a standard hardware-oriented realization. On the 127-qubit heavy-hex layout, one Floquet step is
\[
U_F(\theta_h)
=
\Bigg(\prod_{\langle i,j\rangle\in E} e^{\,i\frac{\pi}{4}Z_iZ_j}\Bigg)
\Bigg(\prod_{i=1}^{127} e^{-i\frac{\theta_h}{2}X_i}\Bigg),
\]
with the \(ZZ\) angle fixed at \(\pi/4\) and the field angle \(\theta_h\) varied [2308.03082]. The same logical structure underlies tensor-network simulations on the IBM Eagle heavy-hex experiment [2306.14887]. On this geometry, sparse connectivity and relatively long loops generate “tree-like correlations,” which sharply improve the accuracy of graph-aligned tensor-network methods [2306.14887].

Two classical approaches have been especially prominent. A belief-propagation-assisted tensor network on the heavy-hex graph simulated the 127-qubit experiment with high precision; in the \(n=5\) regime, average magnetization was computed to approximately \(10^{-14}\) absolute error in less than \(10\) seconds on a laptop, and a weight-17 observable at \(n=6\) was within \(10^{-4}\) of exact values at bond dimension \(\chi=500\) [2306.14887]. A Heisenberg-picture PEPO method exploited the low-rank and near-Clifford structure of the same circuit; for the “5+1” circuit, PEPO with \(\chi=184\) produced exact results in \(3\) seconds on a single CPU, and \(\chi=2\) already matched or exceeded the accuracy of CPT with \(K=10\) and Heisenberg MPO with \(\chi=1024\) [2308.03082].

The same heavy-hex Floquet model has also been emulated on D-Wave Pegasus hardware by matching the ratio of transverse and Ising energy scales. The key mapping condition is
\[
\frac{A(s_p)}{B(s_p)J_a}=-\frac{2\theta_h}{\pi},
\]
with pause durations chosen from the corresponding \(X\)- and \(ZZ\)-angle matching conditions [2311.01657]. This annealer-based emulation reproduced magnetization dynamics for \(N=20,50,100,126,200\), and up to \(10{,}000\) time steps, and yielded reasonable agreement with heavy-hex single-site magnetization at \(20\) steps [2311.01657].

Hardware studies have also pushed disordered kicked Ising dynamics beyond verification by exact diagonalization. On 60 qubits of ibm\_fez, a heavy-hex disordered Floquet circuit with
\[
U_F=e^{-iH_ZT/2}e^{-iH_XT/2}
\]
was used to measure OTOCs under operator renormalization and zero-noise extrapolation, locating the MBL crossover at \(W_c\approx0.18\) [2510.01983]. This establishes the KFIM not only as a theoretical benchmark but as a practical testbed for error mitigation, causal-cone circuit optimization, and large-scale nonequilibrium quantum simulation [2510.01983].

Taken together, these developments place the kicked field Ising model at the intersection of exactly solvable Floquet many-body theory and hardware-constrained quantum dynamics. Its importance derives not from a single canonical Hamiltonian, but from a stable algebraic architecture—Ising entanglers plus field kicks—that remains analytically tractable in special regimes, physically rich away from them, and directly implementable across digital and analog quantum platforms.

Source: https://www.emergentmind.com/topics/kicked-field-ising-model