SearcharxivSearch

arXiv subjects

Kejun Liu

Publications and source records attributed to Kejun Liu.

17 recordsLinked to original sources

Metric completion of the Bender--Brody--M\"uller Hamiltonian: dilation spectrum and missing eigenstates

The Bender--Brody--M\"uller (BBM) Hamiltonian was proposed as a non-Hermitian Hilbert--P\'olya operator. We analyze the Hilbert completion induced, on the standard half-line $L^2$ core, by BBM's candidate metric $\hat\eta=\sin^2(\hat p/2)=\Delta^\dagger\Delta/4$. The form $\eta_0=\Delta^\dagger\Delta$ is positive with trivial kernel but is not coercive. Completing $C_c^\infty(0,\infty)$ in the norm $\|\psi\|_{\eta_0}=\|\Delta\psi\|$ gives a Hilbert space canonically unitarily equivalent to $L^2(\mathbb R_+)$. Its free self-adjoint realization is the dilation generator, with simple, purely absolutely continuous spectrum $\mathbb R$. The analysis yields two spectral statements of interest beyond the BBM problem. First, no bounded sandwich $\Delta^\dagger h(D)\Delta$ is boundedly invertible. Second, the transported symmetric operator has deficiency indices $(\infty,\infty)$ and an adjoint with every real point as an eigenvalue of infinite multiplicity, while its free extension is purely continuous. The realization-independent BBM conclusion concerns the candidate eigenfunctions: $\Delta\psi_z=x^{-z}$, so for $\operatorname{Re}z=1/2$ they do not belong to the completed space. Thus the original BBM boundary-condition/eigenfunction mechanism cannot produce point-spectrum Riemann-zero states in this $L^2$-based metric completion.

math-ph

Paraparticles intrinsically exhibit Hardy-space breakdown

The memory kernel of an open quantum system obeys Kramers--Kronig (KK) relations if and only if its Laplace transform is analytic in the upper half-plane -- a property known as Hardy-space analyticity. Here we show that non-unitary exchange statistics, the defining property of paraparticles, intrinsically breaks Hardy-space analyticity. The metric $\eta$ that guarantees a real closed-system spectrum for these particles necessarily differs from the physical Born inner product ($\|\eta - I\|_F / \|I\|_F = 0.51$) -- a mathematical consequence of the R-matrix's non-unitarity, not a parameter choice. This metric is a "shadow metric": Schur's lemma forces it to commute with every bilinear observable, making the distortion physically invisible in the closed system. But when the paraparticle is coupled to a bath, any coupling operator that lies outside the symmetry algebra -- that is, any interaction that sees the internal flavour structure -- exposes the distortion. The memory kernel then develops upper-half-plane poles at coupling $g_c \approx 0.1$, breaking standard dispersion relations before the closed-system spectrum complexifies. Fermions and bosons, whose exchange is unitary ($\eta = I$ as an analytic fact of the canonical anticommutation algebra), are immune at any coupling, because there is no distortion to expose. The violation is intrinsic: it distinguishes non-unitary exchange statistics from ordinary particle statistics at the level of the memory kernel's analytic structure.

quant-ph

A real-space exceptional ring mediates an eigenframe-charge transition in a non-Hermitian skyrmion

The integer topological charge of a skyrmion is the standard example of topological protection. We ask what happens to that protection when the local generator is made non-Hermitian by polarization-selective gain or loss. The texture charge of a smoothly evolving state remains homotopy-protected under the usual fixed-boundary condition: after projective normalization, the gain/loss contribution has the Gilbert relaxation form on the target sphere. The spectral charges of the local generator behave differently. Its biorthogonal eigenframe charge is quantized while the texture avoids exceptional points. At linewidth matching, the skyrmion equator becomes a real-space exceptional ring on which the biorthogonal Bloch field diverges, and the eigenframe class jumps from +1 to 0 in a parameter-driven spectral transition. The closed contour relies on constant splitting along the equator, or an equivalent design constraint; a partial or amplitude-anisotropic realization instead leaves isolated real-space exceptional points. We use an operational Jones parametrization of an isolated polarization doublet and give a Stokes-tomography reconstruction of the left-right frame. The result identifies a spatially resolved exceptional contour that mediates an integer transition of a real-space eigenframe charge.

cond-mat.mes-hall

Exceptional Points as Manifestations of Analyticity Breakdown in the 't Hooft Model

We use the exactly-solvable t Hooft model of 1+1D large-N_c QCD as a rigorous laboratory for the breakdown of analyticity of a causal response function, the meson two-point function. A PT-symmetric deformation i gamma(x-1/2) of the light-cone meson operator, the analogue of an imaginary chemical potential, drives the lowest two mesons to an exceptional point (EP) at gamma_c. Recasting the resolvent as a Jacobi continued fraction yields gamma_c in closed form: 2 pi g^2 N_c at the two-pole level, converging to 7.966 g^2 N_c by depth five -- an analytic, not numerical, threshold. The square-root exponent nu=1/2 is fixed by the 2x2 Jordan form and confirmed by finite-size scaling to N=1999. The breakdown has an unambiguous time-domain signature: the propagator norm is bounded for gamma < gamma_c, grows linearly at gamma_c (the Jordan secular law), and exponentially beyond -- observable, since the deformed operator is a non-Hermitian Wannier-Stark ladder, in photonic and topolectrical analogues. The threshold is locked to confinement, gamma_c propto g^2 N_c, and recurs as a uniform EP cascade; a second, non-reciprocal deformation yields an exactly-exponential non-Hermitian skin effect. This is the first analytically-controlled instance of exceptional-point analyticity breakdown in a confining gauge theory.

math-ph

Algebraic Spectral Curves of Two-Channel Operator Pencils with Power-Law Response

A finite matrix can acquire a nontrivial algebraic spectral curve when its entries contain a multivalued response. We study this mechanism for a two-channel pencil with scalar response $z^\beta$ and $J$-self-adjoint matrices. If $\beta=r/n$ is in lowest terms and the coefficients are generic, the normalized determinant curve in $(z,\lambda)$ has genus $n-1$ and is a double cover. It is rational for $n=1$, elliptic for $n=2$, and hyperelliptic for $n\geq3$. The generic genus depends only on the denominator $n$. The proof reduces birationally to a classical hyperelliptic normal form. The numerator appears as the cyclic character of the distinguished coupling and controls its divisor and boundary singularity. Here $\lambda$ is a complexified coupling coordinate: the curve is the joint determinant locus, not a fixed-coupling spectrum. For irrational exponents the local monodromy has infinite order, while rational approximants converge uniformly on compact sets and preserve isolated zeros on a fixed branch. Negative exponents in $(-1,0)$ arise from causal Volterra kernels and cutoff continuum-bath Schur complements, separating finite channels from nonlocal memory. As an application, the exact leading infrared monomial model of a zero-temperature conformal response of dimension $h_{\rm probe}$ has, for rational $2h_{\rm probe}$, genus $\operatorname{den}(2h_{\rm probe})-1$. For the fundamental fermion in the even-$q>2$ Sachdev--Ye--Kitaev saddle, this becomes $q/2-1$. The core statements, including determinant algebra, branch counts, and causal-response identities, are machine-checked from raw matrix definitions.

math.NT

The Predictive-Causal Gap: An Impossibility Theorem and Large-Scale Neural Evidence

We report a systematic failure mode in predictive representation learning. Across 2695 neural network configurations trained to predict linear-Gaussian dynamics, the optimal encoder tracks the environment rather than the system it is meant to model. The mean causal fidelity -- the fraction of encoder sensitivity allocated to system degrees of freedom -- is 0.49, and only 2.5% of configurations exceed 0.70. The failure intensifies with dimension: at N=100, the optimal encoder becomes causally blind (fidelity ~10^{-8}) while achieving 92% lower prediction error than the causal representation. We prove this is not an optimization artifact but a structural property of the predictive objective: when environment modes are slower or less noisy than system modes, every minimizer of the population risk encodes the former. The set of dynamics exhibiting this predictive-causal gap is open and of positive measure in parameter space. In a nonlinear Duffing-GRU sweep, unconstrained predictors learn environment-dominant representations in 55% of tasks (95% CI 41--68%) versus 24% under operational grounding (p=2.3e-3); the median out-of-distribution MSE inflation under environment shift is 1.82x versus 1.00x. Operational grounding -- restricting the loss to system observables -- partially suppresses the gap, but causal fidelity is never recovered without an explicit system-environment boundary. The results identify the predictive-causal gap as a structural limit of learning, with implications for self-supervised representation learning, world models, and the scaling paradigm.

cs.LG

Dispersion Relations Across the Unitarity Boundary

Kramers-Kronig (KK) relations rest on a binary premise: a response function is either analytic in the upper half-plane or it is not. We show that a single reduced-state transform organizes both outcomes into a sharp dichotomy controlled by microscopic unitarity. One closed-form function carries, simultaneously, a zero and a pole in the upper half-plane; the spectral abscissa alpha of the reduced propagator decides which is realized. For alpha < 0 (unitary reduction) the upper-half-plane object is a protected zero: KK holds, yet the zero is directly measurable from a finite-time coherence record by a damped Fourier transform (no analytic continuation), obeying a closed law Im(zeta) = 0.3092 g. For alpha > 0 (gain-driven non-unitary reduction) the zero is replaced by a genuine pole, the Blaschke winding number jumps from 0 to 1, and KK acquires a Lorentzian residue correction scaling as a power law with negative exponent nu ~ -1.08, peaking at threshold. The protected zero is not inert: any scalar single-channel kernel extraction is forced to reproduce a phantom resonance -- a refractive feature with no absorptive origin, at a protocol-independent frequency -- without any initial system-bath correlation. We give the closed-form criteria, a measurable terahertz signature (31-1391 GHz), and the solvable dimer and Jaynes-Cummings models that realize both sides of the boundary.

quant-ph

A symmetry-protected pseudo-Hermitian phase of quantum memory-kernel generators

The Jaynes-Cummings (JC) model, introduced in 1963 and central to cavity quantum electrodynamics, describes a two-level system coupled to a single bosonic mode under the rotating-wave approximation. When the mode is projected out via the Nakajima-Zwanzig (NZ) formalism, the memory-kernel generator QLQ is manifestly non-Hermitian -- yet we prove analytically that its spectrum is strictly real at every coupling and every finite truncation, for both vacuum and thermal baths. For the vacuum bath the characteristic polynomial factorises completely; the nonzero eigenvalues reproduce the JC dressed-state ladder for n >= 2, while the lowest mode is suppressed by exactly sqrt(2) relative to the bare-Hamiltonian prediction. For any thermal Gibbs state, the squared generator reduces to an asymmetric rank-one perturbation symmetrised by a closed-form diagonal metric, with eigenvalues guaranteed real by Cauchy interlacing. We construct a positive-definite metric eta_osc intertwining QLQ with its adjoint on the nonzero spectral subspace, proving hidden pseudo-Hermiticity. A classification theorem extends these results to the full U(1)-conserving single-photon-exchange class with arbitrary complex couplings. The phase boundary under counter-rotating deformation is mapped analytically at resonance and numerically across the coupling-truncation plane, revealing a weak-coupling protected wedge, re-entrant real-spectrum bubbles, and N-universal plateaus organised by a three-family band catalog with closed-form level spacings. The full phase structure is proved well-defined in the N -> infinity thermodynamic limit. The sqrt(2) suppression provides a platform-independent experimental falsification target spanning seven orders of magnitude in coupling strength.

quant-ph

Operator-Valued Hardy Spaces and Kramers--Kronig Relations for Non-Markovian Quantum Memory Kernels

Retarded support, upper-half-plane holomorphy, and Hardy boundary control are distinct properties of a memory kernel. We give sufficient conditions linking them for the Nakajima--Zwanzig kernel of an open system with finite-dimensional system Liouville space. If the projected kernel has an absolutely continuous real-axis representation with density w in L1 intersect Lp0, no singular part, and p0 > 1, its transform lies in the operator-valued Hardy class Hp(B) for 1 < p <= p0. A finite-dimensional componentwise argument yields the corresponding principal-value and once-subtracted Kramers--Kronig (KK) boundary formulas; the H1 endpoint is separate. Microscopic unitary evolution gives holomorphy and a trace-norm bound for reduced-state transforms for arbitrary trace-class initial states, without that spectral hypothesis. On a common analytic domain, a perturbative force-fit equation has a first-order quotient pole at a simple baseline state-transform zero zeta only when the first-order inhomogeneous numerator I_tilde^(1)(zeta) is nonzero; at finite perturbation the actual zero and numerator must be checked, while matrix reconstructions additionally require adjugate non-cancellation. For rational reconstructions, the printed local or global clearing and no-cancellation hypotheses convert a rational kernel pole into a genuine upper-half-plane propagator pole. Such a pole is incompatible with a uniformly bounded CPTP family when the transforms agree on an open set, with Vieta's formula giving the imaginary-part budget. We formalize these paper-specific implications in Lean 4 from named classical inputs. A finite Jaynes--Cummings calculation provides a shifted, distributional KK check with a 0.024 percent fixed-truncation residual.

quant-ph

Chemical Medium-Range Order Enables Stoichiometric Rigidity

Maxwell counting predicts an isostatic threshold at $\langle r\rangle = 2.4$ for covalent network glasses, but which structural correlations actually produce rigidity near this point is still unclear. In this work, we test four candidates: enthalpic stress, chemical defects, geometric interlocking, and medium-range order (MRO). We use a locally tree-like configuration model as a zero-MRO baseline and apply perturbations to test each candidate. We find that (i) enthalpic stress delays rigidity rather than enabling it; (ii) chemical defects require fractions ($\sim$40%) far above experimental values ($\sim$16% in GeSe$_2$); (iii) geometric linking density does not govern the threshold location, which is instead set by loop-induced redundancy; and (iv) only phenomenological MRO proxies recover rigidity at experimentally accessible strengths. Consequently, chalcogenide intermediate-phase data and amorphous SiO$_2$ ring statistics positively implicate chemical MRO, while DNA spatial networks independently rule out pure geometric entanglement. We conclude that rigidity near the Maxwell threshold requires chemistry-specific correlations beyond pure connectivity.

cond-mat.mtrl-sci

Smile on the Face, Sadness in the Eyes: Bridging the Emotion Gap with a Multimodal Dataset of Eye and Facial Behaviors

Emotion Recognition (ER) is the process of analyzing and identifying human emotions from sensing data. Currently, the field heavily relies on facial expression recognition (FER) because visual channel conveys rich emotional cues. However, facial expressions are often used as social tools rather than manifestations of genuine inner emotions. To understand and bridge this gap between FER and ER, we introduce eye behaviors as an important emotional cue and construct an Eye-behavior-aided Multimodal Emotion Recognition (EMER) dataset. To collect data with genuine emotions, spontaneous emotion induction paradigm is exploited with stimulus material, during which non-invasive eye behavior data, like eye movement sequences and eye fixation maps, is captured together with facial expression videos. To better illustrate the gap between ER and FER, multi-view emotion labels for mutimodal ER and FER are separately annotated. Furthermore, based on the new dataset, we design a simple yet effective Eye-behavior-aided MER Transformer (EMERT) that enhances ER by bridging the emotion gap. EMERT leverages modality-adversarial feature decoupling and a multitask Transformer to model eye behaviors as a strong complement to facial expressions. In the experiment, we introduce seven multimodal benchmark protocols for a variety of comprehensive evaluations of the EMER dataset. The results show that the EMERT outperforms other state-of-the-art multimodal methods by a great margin, revealing the importance of modeling eye behaviors for robust ER. To sum up, we provide a comprehensive analysis of the importance of eye behaviors in ER, advancing the study on addressing the gap between FER and ER for more robust ER performance. Our EMER dataset and the trained EMERT models will be publicly available at https://github.com/kejun1/EMER.

cs.CV

A Compact Dual-Beam Zeeman Slower for High-Flux Cold Atoms

We present a compact design of dual-beam Zeeman slower optimized for efficient production of cold atom applications. Traditional single-beam configurations face challenges from substantial residual atomic flux impacting downstream optical windows, resulting in increased system size, atomic deposition contamination, and a reduced operational lifetime. Our approach employs two oblique laser beams and a capillary-array collimation system to address these challenges while maintaining efficient deceleration. For rubidium ($^{87}$Rb), simulations demonstrate a significant increase in the fraction of atoms captured by a two-dimensional magneto-optical trap (2D-MOT) and nearly eliminate atom-induced contamination probability at optical windows, all within a compact Zeeman slower length of 44 cm. Experimental validation with Rb and Yb demonstrates highly efficient atomic loading within the same compact design. This advancement represents a substantial improvement for high-flux cold atom applications, providing reliable performance for high-precision metrology, quantum computation and simulation.

physics.atom-ph

Smile upon the Face but Sadness in the Eyes: Emotion Recognition based on Facial Expressions and Eye Behaviors

Emotion Recognition (ER) is the process of identifying human emotions from given data. Currently, the field heavily relies on facial expression recognition (FER) because facial expressions contain rich emotional cues. However, it is important to note that facial expressions may not always precisely reflect genuine emotions and FER-based results may yield misleading ER. To understand and bridge this gap between FER and ER, we introduce eye behaviors as an important emotional cues for the creation of a new Eye-behavior-aided Multimodal Emotion Recognition (EMER) dataset. Different from existing multimodal ER datasets, the EMER dataset employs a stimulus material-induced spontaneous emotion generation method to integrate non-invasive eye behavior data, like eye movements and eye fixation maps, with facial videos, aiming to obtain natural and accurate human emotions. Notably, for the first time, we provide annotations for both ER and FER in the EMER, enabling a comprehensive analysis to better illustrate the gap between both tasks. Furthermore, we specifically design a new EMERT architecture to concurrently enhance performance in both ER and FER by efficiently identifying and bridging the emotion gap between the two.Specifically, our EMERT employs modality-adversarial feature decoupling and multi-task Transformer to augment the modeling of eye behaviors, thus providing an effective complement to facial expressions. In the experiment, we introduce seven multimodal benchmark protocols for a variety of comprehensive evaluations of the EMER dataset. The results show that the EMERT outperforms other state-of-the-art multimodal methods by a great margin, revealing the importance of modeling eye behaviors for robust ER. To sum up, we provide a comprehensive analysis of the importance of eye behaviors in ER, advancing the study on addressing the gap between FER and ER for more robust ER performance.

cs.CV

Interlayer charge transfer in graphene 2D polyimide heterostructures

The vertical integration of multiple two-dimensional (2D) materials in heterostructures, held together by van der Waals forces, has opened unprecedented possibilities for modifying the (opto-)electronic properties of nanodevices. Graphene, with its remarkable opto-electronic properties, is an ideal candidate for such applications. Further candidates are 2D polymers, crystalline polymeric materials with customizable structure and electronic properties that can be synthesized in all mathematically possible Bravais lattices. In this study, we investigated the optoelectronic properties of a heterostructure created by pristine graphene and a rectangular 2D polyimide (2DPI) film. This imprints a new superlattice on graphene in conjunction with a direct influence on its electronic properties. Theoretical and experimental analyses reveal that interlayer charge exchange between the 2D polymer and graphene induces hole doping in the graphene layer. We have also observed that the properties of the heterostructure are dependent on the substrate used in experiments, likely due to the porous character of the 2DPI allowing direct interaction of graphene with the support. These findings highlight the unique ability to tailor functionalities in 2D polymers-based heterostructures, allowing the development of optoelectronic devices with precisely engineered properties and stimulating further exploration of the diverse phenomena accessible through tailored designs of the 2D polymers.

cond-mat.mes-hall

Air-Water Interface-Assisted Synthesis and Charge Transport Characterization of Quasi-2D Polyacetylene Films with Enhanced Electron Mobility via Ring-Opening Polymerization of Pyrrole

Water surfaces catalyze some organic reactions more effectively, making them unique for 2D organic material synthesis. This report introduces a new synthesis method via surfactant-monolayer-assisted interfacial synthesis on water surfaces for ring-opening polymerization of pyrrole, producing distinct polypyrrole derivatives with polyacetylene backbones and ionic substitutions. The synthesis result in quasi 2D polyacetylene (q2DPA) film with enhanced charge transport behavior. We employed time-of-flight photoconductivity (TOFP) measurements using pulsed laser light of tunable wavelength for photoexcitation of the charge carriers within the q2DPA film. The charge transport was measured in the lateral direction as a function of external bias voltage ranging from 0 V to 200 V. We observed high electron mobility ({\mu}) of q2DPA reaching values of 375 cm2 V-1 s-1 at bias voltage Vb = -20V and photon energy of 3.8 eV.

cond-mat.mtrl-sci

Learning Language-guided Adaptive Hyper-modality Representation for Multimodal Sentiment Analysis

Though Multimodal Sentiment Analysis (MSA) proves effective by utilizing rich information from multiple sources (e.g., language, video, and audio), the potential sentiment-irrelevant and conflicting information across modalities may hinder the performance from being further improved. To alleviate this, we present Adaptive Language-guided Multimodal Transformer (ALMT), which incorporates an Adaptive Hyper-modality Learning (AHL) module to learn an irrelevance/conflict-suppressing representation from visual and audio features under the guidance of language features at different scales. With the obtained hyper-modality representation, the model can obtain a complementary and joint representation through multimodal fusion for effective MSA. In practice, ALMT achieves state-of-the-art performance on several popular datasets (e.g., MOSI, MOSEI and CH-SIMS) and an abundance of ablation demonstrates the validity and necessity of our irrelevance/conflict suppression mechanism.

cs.AI

Pose-disentangled Contrastive Learning for Self-supervised Facial Representation

Self-supervised facial representation has recently attracted increasing attention due to its ability to perform face understanding without relying on large-scale annotated datasets heavily. However, analytically, current contrastive-based self-supervised learning (SSL) still performs unsatisfactorily for learning facial representation. More specifically, existing contrastive learning (CL) tends to learn pose-invariant features that cannot depict the pose details of faces, compromising the learning performance. To conquer the above limitation of CL, we propose a novel Pose-disentangled Contrastive Learning (PCL) method for general self-supervised facial representation. Our PCL first devises a pose-disentangled decoder (PDD) with a delicately designed orthogonalizing regulation, which disentangles the pose-related features from the face-aware features; therefore, pose-related and other pose-unrelated facial information could be performed in individual subnetworks and do not affect each other's training. Furthermore, we introduce a pose-related contrastive learning scheme that learns pose-related information based on data augmentation of the same image, which would deliver more effective face-aware representation for various downstream tasks. We conducted linear evaluation on four challenging downstream facial understanding tasks, ie, facial expression recognition, face recognition, AU detection and head pose estimation. Experimental results demonstrate that our method significantly outperforms state-of-the-art SSL methods. Code is available at https://github.com/DreamMr/PCL}{https://github.com/DreamMr/PCL

cs.CV