betaFeedback

WorksPhysics and Astronomy

Strong eigenstate thermalization in Singer-cycle Clifford circuits

Abstract

Eigenstate thermalization hypothesis (ETH) provides a central framework for understanding thermalization in isolated quantum systems. Recently, the infinite-temperature ETH has been established for almost all Floquet eigenstates in linear feedback shift register quantum circuits. Here we introduce brickwork Clifford circuits with LL qubits where the induced transformation of Pauli operators is represented by an irreducible symplectic matrix over the finite field F2\mathbb F_2 of the maximal order 2L+12^L+1, known as a Singer cycle. For these Singer-cycle Clifford circuits, we prove a strong version of ETH: all Floquet eigenstates are thermal at infinite temperature without exceptions. We derive uniform bounds on both diagonal deviations from thermal expectation values and off-diagonal matrix elements of arbitrary observables supported on a fixed number of sites. The results indicate that long-time averages of observables approach the infinite-temperature values in the thermodynamic limit for arbitrary initial states. We also show that the temporal fluctuation of expectation values of observables is exponentially suppressed in the thermodynamic limit, implying that each observable remains close to the infinite-temperature value for almost all times.

Introduction

Understanding the mechanism of thermalization in isolated quantum systems lies at the fundamental core of thermodynamics and statistical mechanics in quantum systems [1, 2]. While we empirically know that expectation values of observables approach a thermal value if a quantum system is left as it is for a long time, it is not obvious how it arises from unitary dynamics where the system’s density matrix does not approach a thermal one in the operator level. Among various approaches, eigenstate thermalization hypothesis (ETH) [3, 4, 5, 6, 7] offers a compelling perspective on this issue: thermal properties are already encoded in individual eigenstates of the Hamiltonian. If ETH holds, one can immediately show that the time average of expectation values of observables approach thermal values at the corresponding energy.

While there have been various numerical studies that support ETH [2, 5], it has been challenging to show ETH rigorously, even for a specific quantum many-body model. There is a weaker version of ETH, stating that almost all eigenstates of the Hamiltonian are thermal (i.e., their expectation values of observables are close to the thermal ones) [8, 9, 10]. This requires only that the fraction of thermal eigenstates approach unity in the thermodynamic limit. The weak ETH can be established under broad conditions, including some integrable systems, and therefore does not distinguish thermalizing systems from non-thermalizing ones (Figure 1). This motivates a sharper estimate on the number of thermal eigenstates, or equivalently, of exceptional nonthermal eigenstates.

Figure 1. Examples of quantum many-body systems classified according to the strong and weak ETH. The strong ETH requires all eigenstates in the relevant energy window and symmetry sector to be thermal, while the weak ETH allows a fraction of nonthermal eigenstates that vanishes in the thermodynamic limit. Here LFSR circuit refers to the one studied in Ref. [23]. Integrable systems can satisfy or violate the weak ETH: representative translation-invariant models can satisfy it, whereas spatially localized free-fermion models violate it [10].

A strong version of ETH excludes such exceptions [11, 12]: all eigenstates within the relevant energy window and symmetry sector are thermal, and hence long-time averages of observables approach the thermal values for initial states confined to those regimes. More precisely, the strong ETH states that the expectation values of an observable OO for all energy eigenstates in each symmetry sector within the energy range of interest approach their corresponding thermal values uniformly:

∣⟨ψk∣O∣ψk⟩−Oth(Ek)∣≤ε(L),(1)\left|\langle\psi_k|O|\psi_k\rangle-O_{\mathrm{th}}(E_k)\right|\leq\varepsilon(L), \tag*{(1)}

where EkE_k is the eigenenergy of ∣ψk⟩|\psi_k\rangle and ε(L)\varepsilon(L) is independent of the eigenstate index kk and vanishes as the system size LL tends to infinity. A more detailed ETH ansatz describes both diagonal and off-diagonal matrix elements, including the scaling and statistical properties of the fluctuations [2, 13]. For periodically driven (Floquet) systems [14, 15], thermalization is expected to lead to an infinite-temperature state. The corresponding Floquet ETH states that expectation values of observables in Floquet eigenstates approach the infinite-temperature expectation values [16, 17].

There have been various theoretical studies towards establishing thermalization and ETH in isolated quantum systems. Among them, in Ref. [18] certain nonintegrable spin chains were shown to possess exact eigenstates that have infinite-temperature expectation values for spatially local observables. This approach was extended to a broader class of stabilizer states [19]. Rigorous results on macroscopic equilibration have been obtained for coarse-grained particle densities in free-fermion systems [20, 21, 22].

Another recent work, which is closely related to the present paper, has shown that for a family of quantum circuits implementing linear feedback shift registers (LFSRs) all Floquet eigenstates except for those in a two-dimensional eigenspace obey the infinite-temperature ETH [23]. As we will see, these circuits have no local conserved quantity, so the exceptional nonthermal eigenstates can be viewed as quantum many-body scar states [24, 25, 26, 27, 28]. In scarred systems, although the number of nonthermal eigenstates is negligible, an initial state may have substantial overlap with nonthermal eigenstates, thereby preventing thermalization (weak ergodicity breaking, Figure 1) [26, 27]. Motivated by these developments, we address the following question: Can one construct an explicit quantum many-body model for which the strong ETH can be rigorously established, without any exceptional nonthermal eigenstates, and hence observables reach thermal values irrespective of the choice of initial states?

In this paper, we consider brickwork Clifford circuits composed of CNOT and SWAP gates in the bulk and Hadamard and phase gates at the right endpoint (see Figure 2(a), with each gate represented in (c)). While the LFSR circuits (Figure 2(b)) permute computational-basis states, the endpoint gates in our circuit generate superpositions and relative phases among them, allowing for richer state dynamics. On the operator side, the endpoint gates mix XX and ZZ Pauli operators (distinction between shift-register and brickwork arrangements is not important [29]). We choose the gate coefficients so that the induced transformation of Pauli operators is represented by an irreducible symplectic matrix over the finite field F2\mathbb{F}_2 with the maximal order 2L+12^L+1 (LL is the number of qubits), known as a Singer cycle [30, 31, 32]. Consequently, every (non-identity) Pauli operator has an orbit of length 2L+12^L+1, up to phase factors.

Figure 2. Circuit diagrams of (a) a Singer-cycle Clifford circuit and (b) an LFSR circuit illustrated for L=8L = 8. (c) Representation of the bulk gates G(ai)G(a_i) and G(bi)G(b_i): each gate consists of a SWAP gate when the binary coefficient is 0, and a combination of SWAP and CNOT gates when it is 1. The endpoint gate G(aL)G(a_L) marked by pink (where we always choose aL=1a_L = 1) in (a) consists of phase (S) and Hadamard (H) gates. The coefficient sets {ai}\{a_i\} and {bi}\{b_i\} are chosen according to different criteria.

For these Singer-cycle Clifford circuits, we rigorously establish the strong ETH without any exceptional nonthermal eigenstates (Figure 1). Despite the resemblance to the LFSR circuits, the proof of the strong ETH for the present circuits is not a straightforward extension, since the circuit dynamics can no longer be described as a permutation of computational-basis states. Instead of explicitly writing down the eigenstates on the computational basis, we keep track of the evolution of Pauli operators and describe squared matrix elements of observables in terms of finite fields.

Using the Weil bound on mixed character sums over finite fields [33, 34, 35], we derive uniform bounds of order D−1/2D^{-1/2} [36] on both diagonal deviations from infinite-temperature expectation values and off-diagonal matrix elements of observables OO supported on a fixed number of sites,

∣⟨ψk∣O∣ψk′⟩−Tr⁡ODδk,k′∣≤cD−1/2,(2)\left|\langle\psi_k|O|\psi_{k'}\rangle-\frac{\operatorname{Tr}O}{D}\delta_{k,k'}\right|\le cD^{-1/2}, \tag*{(2)}

where cc is a constant that does not depend on the system size and the eigenstate labels kk, k′k', and D=2LD=2^L is the Hilbert-space dimension. We further show that, for any initial state, the temporal fluctuations of these expectation values from the thermal values vanish in the thermodynamic limit. Thus, each observable remains close to its infinite-temperature value for almost all times, regardless of the choice of initial states. As a byproduct of the temporal fluctuation analysis, we obtain an alternative (and simpler) proof of the strong ETH but with a weaker bound of order D−1/4D^{-1/4}.

The paper is organized as follows. In Sec. II, we review the LFSR circuits studied in Ref. [23]. In Sec. III, we introduce Singer-cycle Clifford circuits and describe their properties, including the characteristic polynomial, gate coefficients, eigenvalues and eigenstates of the Floquet operator, and phase convention in terms of finite fields. In Sec. IV, we establish the strong ETH by bounding matrix elements of Pauli operators using the Weil bound on finite-field character sums. In Sec. V, we derive bounds on temporal fluctuations of expectation values of observables. Finally, Sec. VI summarizes the paper.

LFSR circuits

In this section, we briefly review the LFSR circuit developed by Ippoliti and Long [23], for which ETH holds for almost all eigenstates (except for two degenerate eigenstates) under suitable conditions on the circuit coefficients.

In the LFSR circuit (Figure 2(b)), the unitary operator generating one period of time evolution (Floquet operator) takes the form of

U=GL−1(bL−1)GL−2(bL−2)⋯G1(b1),(3)U=G_{L-1}(b_{L-1})G_{L-2}(b_{L-2})\cdots G_1(b_1), \tag*{(3)}

where bi=0,1∈F2b_i=0,1\in\mathbb{F}_2 are binary coefficients (F2={0,1}\mathbb{F}_2=\{0,1\} denotes the finite field of characteristic 22) and

Gi(bi)=(CNOTi,i+1)biSWAPi,i+1=SWAPi,i+1(CNOTi+1,i)bi(i=1,2,…,L−1)(4)G_i(b_i)=({\rm CNOT}_{i,i+1})^{b_i}{\rm SWAP}_{i,i+1} ={\rm SWAP}_{i,i+1}({\rm CNOT}_{i+1,i})^{b_i}\qquad(i=1,2,\ldots,L-1) \tag*{(4)}

are gate operators. CNOTi,j{\rm CNOT}_{i,j} denotes a controlled-NOT gate with control qubit ii and target qubit jj, and SWAPi,j{\rm SWAP}_{i,j} represents a swap operator that exchanges states at sites ii and jj. The action of each gate on the two-site computational basis is given by Gi(bi)∣zi,zi+1⟩=∣zi+1,zi+bizi+1⟩G_i(b_i)|z_i,z_{i+1}\rangle= |z_{i+1},z_i+b_i z_{i+1}\rangle, where addition and multiplication is performed in F2\mathbb{F}_2.

We can write the total computational basis in terms of a bit string as ∣z⟩|z\rangle with z=(z1,…,zL)T∈F2Lz=(z_1,\ldots,z_L)^T \in\mathbb{F}_2^L. In the LFSR circuit, a state ∣z⟩|z\rangle evolves after one period as U∣z⟩=∣Az⟩U|z\rangle= |Az\rangle, where AA is an F2\mathbb{F}_2-valued matrix defined by

A=(010⋯0001⋯000⋯01⋮⋮⋱⋱⋮00⋯011b1b2⋯bL−1)∈GL⁡(L,F2).(5)A=\begin{pmatrix} 0 & 1 & 0 & \cdots& 0\\ 0 & 0 & 1 & \cdots& 0\\ 0 & 0 & \cdots& 0 & 1\\ \vdots& \vdots& \ddots& \ddots& \vdots\\ 0 & 0 & \cdots& 0 & 1\\ 1 & b_1 & b_2 & \cdots& b_{L-1} \end{pmatrix}\in\operatorname{GL}(L,\mathbb{F}_2). \tag*{(5)}

Thus, the bits are shifted as zi′=zi+1z_i^\prime= z_{i+1} for i<Li < L and the end bit changes as zL′=z1+∑i=1L−1bizi+1z_L^\prime= z_1+\sum_{i=1}^{L-1}b_i z_{i+1}. One can see that UU permutes the computational-basis states without introducing phases or superpositions, which allows one to construct the eigenstates of the Floquet operator directly from the orbits of bit strings.

The characteristic polynomial of AA is

gL(t)=det⁡(tI−A)=tL+∑j=1L−1bjtj+1.(6)g_L(t)=\det(tI-A)=t^L+\sum_{j=1}^{L-1}b_jt^j+1. \tag*{(6)}

To derive ETH, the gate coefficients bjb_j are chosen in such a way that gL(t)g_L(t) becomes a primitive polynomial over F2\mathbb{F}_2. That is, gL(t)g_L(t) is irreducible, and every root of gL(t)g_L(t) has multiplicative order 2L−12^L-1. These conditions ensure that an orbit starting from a nonzero bit string, z0→Az0→A2z0→⋯z_0\to Az_0\to A^2z_0\to\cdots, forms a periodic cycle with a period,

N=D−1,(7)N=D-1, \tag*{(7)}

where D=2LD=2^L is the dimension of the Hilbert space. In particular, AN=IA^N=I and hence UN=IU^N=I, implying that the eigenvalues of UU are NNth roots of unity,

ωk=e2πik/N(k=0,1,…,N−1).(8)\omega^k=e^{2\pi i k/N}\qquad(k=0,1,\ldots,N-1). \tag*{(8)}

By choosing an arbitrary nonzero bit string z0z_0, one can construct NN eigenstates of UU as

∣ψk⟩=1N∑j=0N−1ωjk∣Ajz0⟩(k=0,1,…,N−1)(9)|\psi_k\rangle=\frac{1}{\sqrt{N}}\sum_{j=0}^{N-1}\omega^{jk}|A^jz_0\rangle\qquad(k=0,1,\ldots,N-1) \tag*{(9)}

with eigenvalues ω−k\omega^{-k}. They are orthogonal to each other (⟨ψk∣ψk′⟩=δk,k′\langle\psi_k|\psi_{k^\prime}\rangle=\delta_{k,k^\prime}).

The remaining eigenstate is ∣0⟩=∣0⟩⊗L|0\rangle=|0\rangle^{\otimes L}, which is invariant under UU (U∣0⟩=∣0⟩U|0\rangle=|0\rangle). Thus, the eigenvalue 1 is twofold degenerate, with the eigenspace spanned by ∣ψ0⟩|\psi_0\rangle and ∣0⟩|0\rangle, while all other eigenvalues are nondegenerate. The degenerate eigenspace is also spanned by ∣0⟩⊗L|0\rangle^{\otimes L} and ∣+⟩⊗L|+\rangle^{\otimes L}, since ∣+⟩⊗L=1D∣0⟩+D−1D∣ψ0⟩|+\rangle^{\otimes L}=\frac{1}{\sqrt{D}}|0\rangle+\sqrt{\frac{D-1}{D}}|\psi_0\rangle. The states ∣0⟩⊗L|0\rangle^{\otimes L} and ∣+⟩⊗L|+\rangle^{\otimes L} are nonthermal eigenstates, since they have nonzero expectation values for some Pauli operators (e.g., ⟨0∣Zj∣0⟩=1\langle0|Z_j|0\rangle=1, ⟨+∣Xj∣+⟩=1\langle+|X_j|+\rangle=1). Recall that all (non-identity) Pauli operators have vanishing expectation values at the infinite-temperature thermal state.

References

  1. [1]C. Gogolin and J. Eisert, “Equilibration, thermalisation, and the emergence of statistical mechanics in closed quantum systems,” Rep. Prog. Phys. 79, 056001 (2016).DOI
  2. [2]L. D’Alessio, Y. Kafri, A. Polkovnikov, and M. Rigol, “From quantum chaos and eigenstate thermalization to statistical mechanics and thermodynamics,” Adv. Phys. 65, 239 (2016).DOI
  3. [3]J. M. Deutsch, “Quantum statistical mechanics in a closed system,” Phys. Rev. A 43, 2046 (1991).pubmed.ncbi.nlm.nih.gov/9905246
  4. [4]M. Srednicki, “Chaos and quantum thermalization,” Phys. Rev. E 50, 888 (1994).DOI
  5. [5]M. Rigol, V. Dunjko, and M. Olshanii, “Thermalization and its mechanism for generic isolated quantum systems,” Nature 452, 854 (2008).
  6. [6]J. M. Deutsch, “Eigenstate thermalization hypothesis,” Rep. Prog. Phys. 81, 082001 (2018).DOI
  7. [7]T. Mori, T. N. Ikeda, E. Kaminishi, and M. Ueda, “Thermalization and prethermalization in isolated quantum systems: a theoretical overview,” J. Phys. B: At. Mol. Opt. Phys. 51, 112001 (2018).DOI
  8. [8]G. Biroli, C. Kollath, and A. M. Läuchli, “Effect of Rare Fluctuations on the Thermalization of Isolated Quantum Systems,” Phys. Rev. Lett. 105, 250401 (2010).
  9. [9]E. Iyoda, K. Kaneko, and T. Sagawa, “Fluctuation Theorem for Many-Body Pure Quantum States,” Phys. Rev. Lett. 119, 100601 (2017).DOI
  10. [10]P. Lydzba, R. SwIATEK, M. Mierzejewski, M. Rigol, and L. Vidmar, “Normal weak eigenstate thermalization,” Phys. Rev. B 110, 104202 (2024).DOI
  11. [11]H. Kim, T. N. Ikeda, and D. A. Huse, “Testing whether all eigenstates obey the eigenstate thermalization hypothesis,” Phys. Rev. E 90, 052105 (2014).
  12. [12]T. Yoshizawa, E. Iyoda, and T. Sagawa, “Numerical Large Deviation Analysis of the Eigenstate Thermalization Hypothesis,” Phys. Rev. Lett. 120, 200604 (2018).DOI
  13. [13]M. Srednicki, “The approach to thermal equilibrium in quantized chaotic systems,” J. Phys. A: Math. Gen. 32, 1163 (1999).
  14. [14]T. Oka and S. Kitamura, “Floquet Engineering of Quantum Materials,” Annu. Rev. Condens. Matter Phys. 10, 387 (2019).DOI
  15. [15]N. Tsuji, “Floquet states,” in Encyclopedia of Condensed Matter Physics (Second Edition), edited by T. Chakraborty (Academic Press, Oxford, 2024) second edition ed., pp. 967–980.
  16. [16]A. Lazarides, A. Das, and R. Moessner, “Equilibrium states of generic quantum systems subject to periodic driving,” Phys. Rev. E 90, 012110 (2014).DOI
  17. [17]L. D’Alessio and M. Rigol, “Long-time Behavior of Isolated Periodically Driven Interacting Lattice Systems,” Phys. Rev. X 4, 041048 (2014).
  18. [18]Y. Chiba and Y. Yoneta, “Exact Thermal Eigenstates of Nonintegrable Spin Chains at Infinite Temperature,” Phys. Rev. Lett. 133, 170404 (2024).DOI
  19. [19]A. Hokkyō, “Exact Thermal Stabilizer Eigenstates at Infinite Temperature,” (2026), arXiv:2601.16177 [quant-ph].arxiv.org/abs/2601.16177
  20. [20]N. Shiraishi and H. Tasaki, “Nature Abhors a Vacuum: A Simple Rigorous Example of Thermalization in an Isolated Macroscopic Quantum System,” J. Stat. Phys. 191, 82 (2024).DOI
  21. [21]H. Tasaki, “Macroscopic irreversibility in quantum systems: Free expansion in a fermionic chain,” Phys. Rev. E 111, L062103 (2025).DOI
  22. [22]T. Hara and T. Koike, “Timescale for Macroscopic Equilibration in Isolated Quantum Systems: A Rigorous Derivation for Free Fermions,” Phys. Rev. Lett. 137, 030406 (2026).DOI
  23. [23]M. Ippoliti and D. M. Long, “Infinite Temperature at Zero Energy,” Phys. Rev. X 16, 031030 (2026).DOI
  24. [24]H. Bernien, S. Schwartz, A. Keesling, H. Levine, A. Omran, H. Pichler, S. Choi, A. S. Zibrov, M. Endres, M. Greiner, V. Vuletić, and M. D. Lukin, “Probing many-body dynamics on a 51-atom quantum simulator,” Nature 551, 579 (2017).
  25. [25]N. Shiraishi and T. Mori, “Systematic Construction of Counterexamples to the Eigenstate Thermalization Hypothesis,” Phys. Rev. Lett. 119, 030601 (2017).DOI
  26. [26]C. J. Turner, A. A. Michailidis, D. Abanin, M. Serbyn, and Z. Papić, “Weak ergodicity breaking from quantum many-body scars,” Nat. Phys. 14, 745 (2018).DOI
  27. [27]M. Serbyn, D. A. Abanin, and Z. Papić, “Quantum many-body scars and weak breaking of ergodicity,” Nat. Phys. 17, 675 (2021).DOI
  28. [28]S. Moudgalya, B. A. Bernevig, and N. Regnault, “Quantum many-body scars and Hilbert space fragmentation: a review of exact results,” Rep. Prog. Phys. 85, 086501 (2022).DOI
  29. [29]One can rearrange the shift register configuration into a brickwork circuit by cyclicly permuting Clifford gates, which can be represented by a unitary transformation.
  30. [30]J. Singer, “A theorem in finite projective geometry and some applications to number theory”, Trans. Amer. Math. Soc. 43, 377 (1938).DOI
  31. [31]B. Huppert, “Singer-Zyklen in klassischen Gruppen,” Math. Z. 117, 141 (1970).DOI
  32. [32]Á. Bereczky, “Maximal Overgroups of Singer Elements in Classical Groups,” J. Algebra 234, 187 (2000).DOI
  33. [33]A. Weil, “On Some Exponential Sums”, Proc. Natl. Acad. Sci. U.S.A. 34, 204 (1948).
  34. [34]F. Castro and C. Moreno, “Maxim exponential sums over finite fields”, Proc. Am. Math. Soc. 128, 2529 (2000).
  35. [35]T. Cochrane and C. Pinner, “Using Stepanov’s method for exponential sums involving rational functions,” J. Number Theory 116, 270 (2006).DOI
  36. [36]The scaling of D−1/2 is natural since 1/D² ∑_{k,k′} |⟨ψk|O|ψk′⟩ − Tr O/D δk,k′|² ≤ ||O||²op/D. holds for arbitrary bounded observables. If every term inside the sum has the same order, then the D−1/2 scaling is naturally expected.
  37. [37]H. D. Kloosterman, “On the representation of numbers in the form ax² + by² + cz² + dt²”, Acta Math. 49, 407 (1926).DOI
  38. [38]K. Conrad, “On Weil’s proof of the bound for Kloosterman sums,” J. Number Theory 97, 439 (2002).DOI

Paper details

vlnc:W_873mae4u
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents