Exceptional Points in Non-Hermitian Systems

• Exceptional Points (EPs) are singularities in non-Hermitian systems where both eigenvalues and eigenvectors coalesce, resulting in defective operators with fractional power scaling.
• They are pivotal in applications such as enhanced sensing, unidirectional light transport, and topological laser design, as seen in photonic, mechanical, and quantum systems.
• Analytical and numerical techniques, including reduced Hamiltonian models and eigenvalue tracking, reveal the square-root and higher-order branch-point behaviors characteristic of EPs.

Exceptional Points (EPs) are singularities in the parameter space of non-Hermitian systems where both eigenvalues and their corresponding eigenvectors coalesce. In contrast to ordinary (Hermitian) degeneracies—where the eigenvalues may coincide but the eigenvectors remain linearly independent—EPs are characterized by the simultaneous merging of both quantities, leading to a defective (non-diagonalizable) operator. EPs are central to the description of a wide class of physical systems where gain, loss, or openness to the environment renders the effective dynamics non-Hermitian. They have far-reaching implications for quantum physics, optics, mechanics, and nanophotonic engineering, offering routes toward highly sensitive device functionalities and novel topological phenomena.

1. Mathematical Structure and Defining Properties

Mathematically, an EP of order $n$ is a parameter set at which a non-Hermitian matrix, such as a Hamiltonian $\mathcal{H}(\lambda)$ or an effective dynamical operator, becomes defective: $n$ eigenvalues and their eigenvectors coalesce, forming a single $n$-dimensional Jordan block. This implies the characteristic polynomial exhibits a repeated root, but the geometric multiplicity is less than the algebraic multiplicity. For a generic $2 \times 2$ system (Eleuch et al., 2013): $$\mathcal{H}^{(2)} = \begin{pmatrix}\varepsilon_1 & \omega \ \omega & \varepsilon_2 \end{pmatrix}, \quad \varepsilon_i = e_i + (i/2)\gamma_i,$$ where $e_i$ are energies, $\gamma_i$ describe decay (or gain), and $\omega$ quantifies coupling via the environment. The EP occurs when

$$Z \equiv \frac{1}{2}\sqrt{(\varepsilon_1-\varepsilon_2)^2 + 4\omega^2} = 0,$$

so that the two eigenvalues and eigenvectors coalesce, and the matrix becomes non-diagonalizable. The haLLMark of an EP is the loss of phase rigidity (the phases of the eigenfunctions are no longer fixed relative to one another) and the emergence of square-root or higher fractional power branch-point behavior in the spectrum under parameter variation.

2. Physical Realizations and Environmental Effects

EPs inherently arise in open quantum systems, where the system is coupled to an external “environment” or “reservoir,” introducing non-Hermitian features such as decay or gain (Eleuch et al., 2013). In photonics, this is realized in optical cavities, waveguides, or nanostructures subject to radiative loss or engineered gain/loss contrast (Park et al., 2019, Kodigala et al., 2016). In mechanics, EPs appear in metamaterials with loss/gain elements or complex stiffness (Wang et al., 2020). The environment induces nonrigid phase relationships and allows the system to “align” with external dissipation channels, leading to phenomena such as level repulsion in energies, width bifurcation (redistribution of decay rates), and reorganized resonance lifetimes.

In PT-symmetric systems, a balance of gain and loss yields purely real eigenvalues in the symmetry-unbroken regime; EPs mark the transition to complex pairs, corresponding to spontaneous symmetry breaking (Eleuch et al., 2013, Longhi, 2018). Open photonic or mechanical systems with only loss (no explicit gain) can also support EPs due to radiation-induced non-Hermiticity (Abdrabou et al., 2018, Abdrabou et al., 2019).

3. Analytical and Numerical Characterization

Analyses of EPs commonly involve reduced models with $2 \times 2$ or $n \times n$ effective non-Hermitian Hamiltonians subject to environmental coupling. Key observables include the eigenvalues (and their real/imaginary parts) and eigenfunctions, monitored as parameters such as coupling strength, loss/gain, system geometry, or background environment are varied (Eleuch et al., 2013, Kodigala et al., 2016). Analytical expressions for eigenvalues trace out avoided crossings, crossings, or square-root splitting indicative of EPs.

Numerical methods range from direct matrix diagonalization with sensitivity to loss/gain (Kodigala et al., 2016, Abdrabou et al., 2018), to contour-integral-based solvers for nonlinear eigenvalue problems in continuum settings (Deguchi et al., 13 Sep 2024), to mode-matching and multipole expansions for nanostructures and resonators (Abdrabou et al., 2019, Matsushima et al., 2023). The square-root ($\sqrt{\epsilon}$), cube-root ($\epsilon^{1/3}$), or higher-order fractional scaling of eigenvalue splitting near EPs is a universal diagnostic, confirmed both numerically and experimentally (Abdrabou et al., 2018, Matsushima et al., 2023).

4. System Dynamics and Topological Implications

System dynamics near EPs is fundamentally altered by the coalescence of states. In open quantum and optical systems, approaching an EP leads to phenomena such as the transition from Rabi-like oscillations to “loss-induced transparency” (Longhi, 2018), dramatic mode-mixing, and enhanced sensitivity to parameter changes (Park et al., 2019). The “phase rigidity” of eigenfunctions decreases near the EP, leading to non-orthogonal mode interactions and pronounced redistribution effects (Eleuch et al., 2013). In mechanical metamaterials, EPs enable perfect transmission (“bi-directional transparency”) or unidirectional reflection, depending on the interplay of symmetry and non-conservativity (Wang et al., 2020).

The topological structure of the eigenvalue spectrum near an EP consists of branch points on multi-sheeted Riemann surfaces. Encircling an EP in parameter space results in eigenvalue permutation, a feature seen directly in both numerics and experiment (Abdrabou et al., 2019, Deguchi et al., 13 Sep 2024). In synthetic photonic systems, EPs manifest as Fermi arcs and band windings, providing new routes to explore nontrivial band topology in non-Hermitian contexts (Yang et al., 2022).

5. Higher-Order EPs and Symmetry Protection

While second-order EPs (coalescence of two eigenstates) are the most generically observed—requiring the tuning of two real parameters—higher-order EPs (involving three or more coalescing states) are possible with additional symmetry or fine-tuning (Mandal et al., 2021). Parity-time (PT), generalized chiral, or pseudo-Hermitian symmetries can reduce the parameter count required to realize higher-order EPs and enforce specific dispersion relations, such as $k^{1/3}$ scaling in PT-symmetric settings and $k^{1/2}$ in chiral-symmetric models (Mandal et al., 2021). Vector EPs, enabled by symmetry-protected coupling between orthogonal polarizations, allow for the generation of superchiral fields in photonic devices, relevant for chiral sensing applications (Wu et al., 2020).

A distinction arises between defective EPs (non-diagonalizable points), non-defective EPs (degeneracy with diagonalizability at the locus), and ordinary nodal points (remaining diagonalizable in the neighborhood), with the interplay of symmetry class dictating their presence and coexistence (Sayyad et al., 2022).

6. Experimental Observation and Applications

EPs have been observed in nanoscale plasmonic structures (Kodigala et al., 2016, Park et al., 2019), periodic dielectric slabs (Abdrabou et al., 2018), parallel dielectric cylinders (Abdrabou et al., 2019), magnon-polariton systems with synthetic dimensions (Zhang et al., 2019), mechanical metamaterials (Wang et al., 2020), and elastic bodies (cylindrical and spherical) with radiation loss (Matsushima et al., 2023, Deguchi et al., 13 Sep 2024). Experimental methods often detect the coalescence of resonant frequencies and linewidths, sensitivity enhancements, and the permutation of resonances under parameter tuning.

Notable applications include:

• Enhanced Sensing: Due to the fractional-root dependence of eigenvalue splitting near an EP ($\epsilon^{1/n}$), nanoscale and photonic sensors can be dramatically more sensitive to environmental changes (Park et al., 2019, Kodigala et al., 2016, Wu et al., 2020).
• Topological Lasers and Light Control: Robust edge modes and unidirectional transparency tied to EPs enable stable single-mode lasing and asymmetric light transport (Simon et al., 1 Aug 2024).
• Mechanical and Acoustic Devices: EP-induced transparency and wave steering offer routes to filtering, cloaking, and unidirectional energy control (Wang et al., 2020, Matsushima et al., 2023).
• Quantum Phase Transitions: EPs mark boundaries between dynamical phases—such as transitions between damping and oscillatory regimes in magnetic bilayers (Deng et al., 2022) or defect scaling in nonequilibrium quantum criticality (Kibble–Zurek mechanism) (Dóra et al., 2018).

7. Theoretical Advances and Future Research

Recent advances extend the EP paradigm to non-Markovian quantum systems, showing that memory effects in structured reservoirs generate “pure non-Markovian EPs” inaccessible in the Markovian limit, and these can be engineered via environmental spectral properties (Lin et al., 26 Jun 2024). The mathematical description of EPs has been linked to tropical geometry, which enables a systematic classification of EP order and the skin effect via Newton polygons and amoebas, providing new analytical tools for non-Hermitian band theory (Banerjee et al., 2023).

In generalized models, including longer-range hopping (e.g., generalized Hatano–Nelson), any order of EP can be realized independently of system size, with robust features such as the tunable non-Hermitian skin effect and resilience to disorder (Gohsrich et al., 18 Mar 2024). The ability to control the number, location, and order of EPs—using non-Hermitian extensions like hopping amplitude gradients (Simon et al., 1 Aug 2024) or symmetry engineering—suggests a path to customizable spectral and topological responses in a growing range of platforms.

In summary, exceptional points represent structurally robust singularities in non-Hermitian systems with profound consequences for spectral, dynamical, and topological properties. Their paper has unified descriptions across quantum, optical, mechanical, and synthetic systems, offering unprecedented possibilities for device engineering, topological photonics, quantum control, and sensor technologies.