Hidden Hermiticity of the Jaynes–Cummings Projected Liouvillian
Kejun Liu · 2026
Read Hidden Hermiticity of the Jaynes–Cummings Projected Liouvillian by Kejun Liu (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-25801.txt.
Introduction
Liu proves a surprising reality result for the Jaynes-Cummings memory-kernel generator. The model couples a two-level system to one bosonic mode in cavity quantum electrodynamics. Projecting out the mode leaves a seemingly non-Hermitian generator. Yet its spectrum stays strictly real at every coupling and finite truncation. For a vacuum bath the polynomial factors fully, recovering the dressed-state ladder, lowest mode cut by root two. For thermal states, Cauchy interlacing forces real eigenvalues and a metric proves hidden pseudo-Hermiticity.
Catalog record
| CREATOR | Kejun Liu |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 21097 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.25801
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Hidden Hermiticity of the Jaynes–Cummings Projected Liouvillian Kejun Liu Abstract The Jaynes–Cummings model has been studied for six decades, yet its Nakajima–Zwanzig projected Liouvillian Q L Q QLQ hides an unexpected structure: although manifestly non-Hermitian, its spectrum is purely real across all numerically accessible parameters. We resolve this anomaly by constructing a positive-definite metric η > 0 \eta>0 such that ( Q L Q ) † η = η ( Q L Q ) (QLQ)^{\dagger}\eta=\eta(QLQ) , proving that the generator of quantum memory is pseudo-Hermitian. The effect is genuine, not a block-diagonal artefact: the Δ N = 0 \Delta N=0 and Δ N = ± 2 \Delta N=\pm 2 sectors remain strongly non-Hermitian, yet the global spectrum is protected by η \eta . Building on Ng and Rabani’s identification of exceptional-point physics in Q L Q QLQ , we show that the Jaynes–Cummings point sits inside a symmetry-protected pseudo-Hermitian phase. Under continuous deformation toward the full Rabi model this protection breaks and re-enters, producing a U ( 1 ) → U(1)\to complex → Z 2 \to Z_{2} phase structure bracketed by exceptional-point boundaries. The result supplies a structural reason for Hardy-space analyticity of the memory kernel in the canonical light–matter model and identifies a phase transition between oscillatory and genuinely dissipative quantum memory. The Nakajima–Zwanzig (NZ) projection-operator formalism nakajima1958 ; zwanzig1960 is foundational to non-Markovian open quantum systems. The projected Liouvillian QLQ=Q\,\mathcal{L}\,Q,\qquad Q=\mathbb{I}-P, (1) where \mathcal{L}=[H,\cdot] and P is the stationary NZ projector breuer2002 , generates the memory kernel \mathcal{K}(t)\sim P\mathcal{L}Q\,e^{iQ\mathcal{L}Qt}\,Q\mathcal{L}P . While \mathcal{K}(t) and the reduced state \sigma(t) have been studied extensively, QLQ itself has attracted little direct attention. This is a consequential gap: the recent Hardy-space programme liu2026hardy has shown that the analytic structure of the memory kernel is controlled by the spectrum of QLQ : a purely real spectrum guarantees the Cauchy spectral representation and hence Hardy-class membership, which in turn validates Kramers–Kronig (KK) dispersion relations. The community default is that QLQ , being manifestly non-Hermitian whenever the bath reference state \rho_{B}\neq\mathbb{I}/d_{B} , has a generically complex spectrum. A numerical survey liu2026hardy confirmed this for the spin-boson model ( |\Imag\lambda|\sim 10^{-2} – 10^{-1} at moderate coupling). The anomaly is the Jaynes–Cummings (JC) model jaynes1963 : QLQ has a purely real spectrum across all numerically accessible parameters ( N_{\rm max}=3 – 20 , g=0.1 – 2.0 , vacuum and thermal baths), despite \|QLQ-(QLQ)^{\dagger}\|_{F}\sim 10^{1} – 10^{2} . Below we resolve this anomaly. Ng and Rabani identified the first exceptional-point physics of QLQ in a many-body-localized central-spin problem, where non-Hermiticity drives exponential growth of the memory kernel rather than spectral protection ng2021 . Building on that observation, we show that the JC model realizes the opposite regime: a previously unknown positive-definite metric \eta renders QLQ pseudo-Hermitian and protects its real spectrum. Under counter-rotating deformation toward the full Rabi model this protection breaks and re-enters, producing a U(1)\to complex \to Z_{2} phase structure bracketed by exceptional-point boundaries. The point is not that QLQ can be pathological; it is that quantum memory generators themselves admit symmetry-protected phases. Model. —The single-mode JC Hamiltonian ( \hbar=1 ) is H_{\rm JC}=\frac{\omega_{0}}{2}\sigma_{z}+\omega_{c}\,a^{\dagger}a+g(\sigma_{+}a+\sigma_{-}a^{\dagger}) . The total Hilbert space has dimension d=2(N_{\rm max}+1) . The Liouvillian L=\mathbb{I}\otimes H-H^{\mathsf{T}}\otimes\mathbb{I} (column-stacking convention) is Hermitian. The NZ projector P\rho=(\Tr_{B}\rho)\otimes\rho_{B} with vacuum bath \rho_{B}=|0\rangle\langle 0| selects system operators tensored with the vacuum; Q=\mathbb{I}-P . The nontrivial spectrum of QLQ lives on \operatorname{range}(Q) , of dimension d^{2}-4 . Pseudo-Hermiticity. —An operator A is \eta -pseudo-Hermitian mostafazadeh2002a if there exists \eta>0 such that A^{\dagger}\eta=\eta A , implying \eta^{1/2}A\eta^{-1/2} is Hermitian. For a diagonalizable A with real eigenvalues \lambda_{n} and biorthonormal left/right eigenvectors \langle l_{m}|r_{n}\rangle=\delta_{mn} , any metric \eta=\sum_{\lambda_{n}\neq 0}c_{n}|l_{n}\rangle\langle l_{n}|,\qquad c_{n}>0, (2) satisfies the intertwining relation. We adopt the canonical choice c_{n}\equiv 1 . We construct QLQ numerically, perform the biorthonormal diagonalisation, and verify: (i) \max|\Imag\lambda_{n}|<10^{-14} for all tested N_{\rm max} ; (ii) the intertwining residual \|(QLQ)^{\dagger}\eta-\eta QLQ\|_{F}<5\times 10^{-12} ; (iii) \eta is positive-definite on the nontrivial-eigenvalue subspace of QLQ ; and (iv) the similarity-transformed operator is Hermitian to <10^{-12} . Table 1 reports the basic convergence. The important structural point is not the raw growth of the global condition number but its resolution: because QLQ is block-diagonal in the excitation-number difference \Delta N , the metric can be constructed independently within each sector, and the apparent divergence of \kappa(\eta) is revealed to be a global normalisation artefact. Figure 1 (a) shows that the \Delta N=\pm 2 sectors remain perfectly conditioned while \kappa_{\Delta N=0} stays moderate, so the pseudo-Hermitian protection does not deteriorate with the accessible truncation. The full convergence analysis is given in the Supplemental Material. Table 1: Convergence of the metric construction (vacuum, g=0.3 , \omega_{0}=\omega_{c}=1.0 ). Intertwining: \|(QLQ)^{\dagger}\eta-\eta QLQ\|_{F} . N_{\rm max} d^{2} |\Imag\lambda|_{\rm max} Intertwining \kappa(\eta) 3 64 0 5.2\times 10^{-14} 43 7 256 1.6\times 10^{-33} 4.2\times 10^{-13} 165 10 484 2.3\times 10^{-31} 5.3\times 10^{-12} 2.2\times 10^{3} 16 1156 2.3\times 10^{-15} 1.8\times 10^{-12} 495 20 1764 6.8\times 10^{-15} 2.8\times 10^{-12} 766 Genuineness: beyond the U(1) sector. —The JC model has a U(1) symmetry generated by the excitation number N_{\rm exc}=(\sigma_{z}+\mathbb{I})/2+a^{\dagger}a (so |g,n\rangle\!\to\!n , |e,n\rangle\!\to\!n+1 ). L=[H,\cdot] commutes with L_{N}=[N_{\rm exc},\cdot\,] , and the NZ projector P also commutes with L_{N} for any bath state diagonal in the bath number basis; hence QLQ is block-diagonal along the eigensectors \Delta N=N_{\rm exc}({\rm left})-N_{\rm exc}({\rm right}) . The natural suspicion is that the global real spectrum is a trivial consequence of this U(1) block structure—each block real-spectrum because individually Hermitian. We test this by computing the per-sector Hermiticity residual \|QLQ_{\Delta N}-QLQ_{\Delta N}^{\dagger}\|_{F} (Table 2 ). Table 2: Sector-by-sector Hermiticity of QLQ at (N_{\rm max},g)=(3,0.3) . Sector dimensions sum to d^{2}=64 ; off-sector leakage is zero to machine precision; \sum_{\Delta N}\|QLQ_{\Delta N}\|_{F}^{\,2}=\|QLQ\|_{F}^{\,2}=226.91 . \Delta N Dimension \|QLQ_{\Delta N}\|_{F} Hermiticity residual 0 14 2.25 1.70 \pm 1 12 3.95 <10^{-13} \pm 2 8 6.53 5.06 \pm 3 4 6.06 <10^{-13} \pm 4 1 4.00 <10^{-13} The hypothesis fails. The sectors \Delta N=\pm 1,\pm 3,\pm 4 are individually Hermitian, but the \Delta N=0 and \Delta N=\pm 2 sectors are substantially non-Hermitian: at \Delta N=\pm 2 the Hermiticity residual is 5.06 , 77\% of the sector’s total Frobenius norm 6.53 ; at \Delta N=0 the residual is 1.70 , 76\% of the sector norm. Summed in quadrature, 1.70^{2}+2(5.06^{2})=54.1=\|QLQ-(QLQ)^{\dagger}\|_{F}^{\,2} —these three sectors carry the entire global non-Hermitian content. Yet the full QLQ has a purely real spectrum. This is the defining signature of genuine pseudo-Hermiticity. The U(1) block structure is necessary but not sufficient for spectral reality; the metric \eta is what renders the non-Hermitian sectors into a globally Hermitian operator under similarity, without the individual sectors themselves being Hermitian. A further structural observation: QLQ is a purely real matrix (all entries real), so (QLQ)^{\dagger}=(QLQ)^{\mathsf{T}} and Eq. ( 2 ) simplifies to (QLQ)^{\mathsf{T}}\eta=\eta(QLQ) , a generalised transpose-symmetry. Figure 1: Sector-resolved metric condition number and spectral stability. (a) Global \kappa(\eta) (squares) grows as d^{1.8} (dashed), while the \Delta N=\pm 2 sectors remain perfectly conditioned ( \kappa=1 , diamonds) and \Delta N=0 fluctuates between 6 and 24 (circles). The maximum per-sector \kappa (stars) stays below 24 at all N_{\rm max} . The growth is a normalisation