PT-Symmetric Open Dimer Dynamics

• PT-symmetric open dimer is a minimal two-mode non-Hermitian system with balanced gain and loss that transitions from real to complex eigenvalue regimes at its exceptional point.
• The model underpins studies on symmetry structures beyond PT, including non-Hermitian particle-hole and chiral symmetries, with applications in optics, scattering, and lattice systems.
• Extensions to nonlinear dynamics, Floquet gain–loss modulations, and coupled dimer lattices reveal integrable behavior, topological transitions, and precise power transfer control.

A PT-symmetric open dimer is the minimal non-Hermitian two-mode system with balanced gain and loss, typically realized as two coupled oscillators, waveguides, wells, resonators, or circuits related by parity exchange and time reversal. In its canonical linear form, the dimer is described by a two-site Hamiltonian

$$H=\begin{pmatrix} i\gamma & t \ t & -i\gamma \end{pmatrix},$$

with coupling $t>0$ and balanced gain/loss strength $\gamma>0$; equivalently, in normalized optical notation,

$$-i\partial_\zeta \begin{pmatrix} \mathcal E_1\ \mathcal E_2 \end{pmatrix} = \begin{pmatrix} i\gamma & 1\ 1 & -i\gamma \end{pmatrix} \begin{pmatrix} \mathcal E_1\ \mathcal E_2 \end{pmatrix}.$$

Its eigenvalues

$E_\pm=\pm\sqrt{t^2-\gamma^2}$

or, after normalization, $\varepsilon_\pm=\pm\Omega$ with $\Omega=\sqrt{1-\gamma^2}$, are real in the PT-symmetric phase, coalesce at an exceptional point, and become purely imaginary in the broken phase (Ge, 2024). The dimer has consequently become the canonical building block for a wide range of non-Hermitian problems, including open-boundary quantization, nonlinear integrability, Floquet gain–loss management, scattering and causality, and lattice/topological extensions (Morales et al., 2016).

1. Canonical formulation and meanings of openness

The basic PT-symmetric dimer consists of two coupled sites or modes with opposite imaginary onsite terms. In the tight-binding or coupled-mode representation, parity exchanges the two sites and time reversal complex-conjugates amplitudes, so PT symmetry is expressed as

$$[PT,H]=0 \qquad\text{or equivalently}\qquad PH^*P=H,$$

with $P=\sigma_x$ and $T=K$ in the two-site model (Ge, 2024). In optics, the same structure arises after removing the average phase of a two-waveguide coupler with equal real refractive-index parts and opposite imaginary parts, leading to a mode-coupling matrix

$t>0$0

in normalized units (Morales et al., 2016).

The expression open dimer is used in more than one precise sense. In the nonlinear Schrödinger literature, it denotes a two-site system that exchanges energy with the environment through balanced gain and loss,

$t>0$1

so that the dynamics are nonconservative even though the gain and loss are exactly balanced (Barashenkov, 2014). In scattering and input–output settings, openness refers instead to explicit coupling of the dimer to external transmission lines or waveguides, turning the two-mode core into a non-Hermitian scattering target (Schindler et al., 2012). In the finite-chain quantization problem, openness has yet another technical meaning: an open chain of concatenated PT dimers with no wraparound coupling between the endpoints (Ge, 2024). This usage suggests that “open” is not a single physical category but a structural qualifier whose meaning depends on whether the emphasis is internal gain/loss, boundary conditions, or external coupling.

The dimer’s status as a minimal model is reinforced by its physical realizations. The literature summarized here places it in coupled optical waveguides, double-well condensates, inductively coupled LRC circuits, classical oscillator dimers, and generalized collective-spin dimers (Morales et al., 2016). In each case, the balance of amplification and attenuation is the defining PT constraint, while the coupling term determines whether the system remains in the exact PT phase or enters the broken regime.

2. Linear spectrum, exceptional points, and exact quantization

For the isolated two-site linear dimer, the spectrum is

$t>0$2

real for $t>0$3, purely imaginary for $t>0$4, and degenerate at the exceptional point $t>0$5 (Ge, 2024). In normalized optical notation, the corresponding propagator is

$t>0$6

with the exceptional-point form

$t>0$7

and the broken-phase form

$t>0$8

showing oscillatory amplified exchange below threshold, power-law growth at the exceptional point, and exponential amplification above threshold (Morales et al., 2016).

A distinctive exact result emerges when the dimer is embedded in a finite open chain of concatenated PT dimers. For sites $t>0$9, the Hamiltonian is

$\gamma>0$0

and the finite-size spectrum follows from a recurrence relation

$\gamma>0$1

Writing $\gamma>0$2 and imposing boundary conditions yields the exact quantization condition

$\gamma>0$3

equivalently

$\gamma>0$4

The resulting finite-chain energies are

$\gamma>0$5

The central consequence is that the chain behaves, for energy quantization, as if its size were

$\gamma>0$6

not $\gamma>0$7 (Ge, 2024).

For a single dimer, $\gamma>0$8, so $\gamma>0$9 and

$$-i\partial_\zeta \begin{pmatrix} \mathcal E_1\ \mathcal E_2 \end{pmatrix} = \begin{pmatrix} i\gamma & 1\ 1 & -i\gamma \end{pmatrix} \begin{pmatrix} \mathcal E_1\ \mathcal E_2 \end{pmatrix}.$$0

which reproduces the two-site spectrum while showing that the quantized “box” size is effectively three rather than two (Ge, 2024). The explanation is that the endpoint equations are equivalent to implicit Dirichlet conditions at ghost sites,

$$-i\partial_\zeta \begin{pmatrix} \mathcal E_1\ \mathcal E_2 \end{pmatrix} = \begin{pmatrix} i\gamma & 1\ 1 & -i\gamma \end{pmatrix} \begin{pmatrix} \mathcal E_1\ \mathcal E_2 \end{pmatrix}.$$1

so the confining interval extends one lattice spacing beyond each physical end. The same paper stresses that the reality of the lattice momentum $$-i\partial_\zeta \begin{pmatrix} \mathcal E_1\ \mathcal E_2 \end{pmatrix} = \begin{pmatrix} i\gamma & 1\ 1 & -i\gamma \end{pmatrix} \begin{pmatrix} \mathcal E_1\ \mathcal E_2 \end{pmatrix}.$$2 means the spatial eigenstate profile remains sinusoidal across the chain; the complex energy controls temporal or propagation amplification/attenuation rather than spatial exponential growth (Ge, 2024).

3. Symmetry structure beyond PT

Besides PT symmetry, the dimer admits additional non-Hermitian spectral symmetries. One is non-Hermitian particle-hole symmetry, written as

$$-i\partial_\zeta \begin{pmatrix} \mathcal E_1\ \mathcal E_2 \end{pmatrix} = \begin{pmatrix} i\gamma & 1\ 1 & -i\gamma \end{pmatrix} \begin{pmatrix} \mathcal E_1\ \mathcal E_2 \end{pmatrix}.$$3

with $$-i\partial_\zeta \begin{pmatrix} \mathcal E_1\ \mathcal E_2 \end{pmatrix} = \begin{pmatrix} i\gamma & 1\ 1 & -i\gamma \end{pmatrix} \begin{pmatrix} \mathcal E_1\ \mathcal E_2 \end{pmatrix}.$$4 complex conjugation and, for the PT dimer, $$-i\partial_\zeta \begin{pmatrix} \mathcal E_1\ \mathcal E_2 \end{pmatrix} = \begin{pmatrix} i\gamma & 1\ 1 & -i\gamma \end{pmatrix} \begin{pmatrix} \mathcal E_1\ \mathcal E_2 \end{pmatrix}.$$5. This implies the spectral pairing

$$-i\partial_\zeta \begin{pmatrix} \mathcal E_1\ \mathcal E_2 \end{pmatrix} = \begin{pmatrix} i\gamma & 1\ 1 & -i\gamma \end{pmatrix} \begin{pmatrix} \mathcal E_1\ \mathcal E_2 \end{pmatrix}.$$6

Real eigenvalues therefore occur in $$-i\partial_\zeta \begin{pmatrix} \mathcal E_1\ \mathcal E_2 \end{pmatrix} = \begin{pmatrix} i\gamma & 1\ 1 & -i\gamma \end{pmatrix} \begin{pmatrix} \mathcal E_1\ \mathcal E_2 \end{pmatrix}.$$7 pairs, while purely imaginary eigenvalues may be self-paired on the imaginary axis (Ge, 2024).

A second is non-Hermitian chiral symmetry, obtained by combining $$-i\partial_\zeta \begin{pmatrix} \mathcal E_1\ \mathcal E_2 \end{pmatrix} = \begin{pmatrix} i\gamma & 1\ 1 & -i\gamma \end{pmatrix} \begin{pmatrix} \mathcal E_1\ \mathcal E_2 \end{pmatrix}.$$8 with parity $$-i\partial_\zeta \begin{pmatrix} \mathcal E_1\ \mathcal E_2 \end{pmatrix} = \begin{pmatrix} i\gamma & 1\ 1 & -i\gamma \end{pmatrix} \begin{pmatrix} \mathcal E_1\ \mathcal E_2 \end{pmatrix}.$$9: $E_\pm=\pm\sqrt{t^2-\gamma^2}$0 This yields the symmetry

$E_\pm=\pm\sqrt{t^2-\gamma^2}$1

so the spectrum is symmetric about the origin in the complex plane (Ge, 2024). These structures extend, in more complicated form, to the concatenated PT-dimer chain, where they coexist with the open-boundary quantization law.

The optical review places these symmetries in a broader group-theoretic context. The dimer is described there as the smallest member of a planar $E_\pm=\pm\sqrt{t^2-\gamma^2}$2-waveguide class governed by

$E_\pm=\pm\sqrt{t^2-\gamma^2}$3

with the complexified $E_\pm=\pm\sqrt{t^2-\gamma^2}$4 generators furnishing a finite-dimensional non-unitary representation of $E_\pm=\pm\sqrt{t^2-\gamma^2}$5 in $E_\pm=\pm\sqrt{t^2-\gamma^2}$6 dimensions (Morales et al., 2016). In that formulation, the dimer is not an isolated curiosity but the minimal representative of a Lorentz-group-related family with the same spectral collapse at the exceptional point.

The same paper also formulates a non-Hermitian generalization of Ehrenfest’s theorem. For an operator $E_\pm=\pm\sqrt{t^2-\gamma^2}$7, the non-Hermitian commutator is

$E_\pm=\pm\sqrt{t^2-\gamma^2}$8

and the expectation value obeys

$E_\pm=\pm\sqrt{t^2-\gamma^2}$9

For the dimer Stokes components

$\varepsilon_\pm=\pm\Omega$0

this framework gives

$\varepsilon_\pm=\pm\Omega$1

making explicit that the total intensity is not conserved (Morales et al., 2016).

4. Nonlinear dimers, integrability, and reduced pendulum dynamics

The standard nonlinear PT-symmetric dimer is

$\varepsilon_\pm=\pm\Omega$2

or, in a common normalized form,

$\varepsilon_\pm=\pm\Omega$3

In terms of amplitudes and phase difference, with

$\varepsilon_\pm=\pm\Omega$4

the reduced equations are

$\varepsilon_\pm=\pm\Omega$5

$\varepsilon_\pm=\pm\Omega$6

$\varepsilon_\pm=\pm\Omega$7

and the equilibrium conditions for $\varepsilon_\pm=\pm\Omega$8 are

$\varepsilon_\pm=\pm\Omega$9

Accordingly, there are no fixed points with $\Omega=\sqrt{1-\gamma^2}$0, and for $\Omega=\sqrt{1-\gamma^2}$1 no equilibrium points exist at all (Pickton et al., 2013).

A key structural fact is that these gain–loss dimers remain integrable. One conserved quantity is

$\Omega=\sqrt{1-\gamma^2}$2

and a geometrically useful reduction introduces an arc-angle variable $\Omega=\sqrt{1-\gamma^2}$3 such that

$\Omega=\sqrt{1-\gamma^2}$4

The dynamics then collapse to a single equation,

$\Omega=\sqrt{1-\gamma^2}$5

with first integral

$\Omega=\sqrt{1-\gamma^2}$6

This is a pendulum equation with a linear potential and a constant drive (Pickton et al., 2013). The reduced phase portrait contains periodic orbits around centers, separatrices through saddles, and unbounded trajectories when the linear drive dominates. The same paper identifies the thresholds

$\Omega=\sqrt{1-\gamma^2}$7

with bounded, mixed, and fully unbounded regimes occurring in the intervals stated in the source (Pickton et al., 2013).

The dimer also admits Hamiltonian formulations despite gain and loss. For the standard nonlinear Schrödinger dimer, introducing Stokes variables

$\Omega=\sqrt{1-\gamma^2}$8

and suitable canonical variables yields the Hamiltonian

$\Omega=\sqrt{1-\gamma^2}$9

This establishes that the system is Hamiltonian even though the obvious canonical structure in $$[PT,H]=0 \qquad\text{or equivalently}\qquad PH^*P=H,$$0 is lost when $$[PT,H]=0 \qquad\text{or equivalently}\qquad PH^*P=H,$$1 (Barashenkov, 2014). A related result constructs two four-parameter cubic PT-symmetric dimer families—cross-gradient and straight-gradient—whose balanced gain–loss extensions remain completely integrable Hamiltonian systems and, in certain parameter regimes, keep all trajectories confined for arbitrary $$[PT,H]=0 \qquad\text{or equivalently}\qquad PH^*P=H,$$2 (Barashenkov et al., 2015).

A distinct exactly solvable Hamiltonian dimer arises from a pair of nonlinear oscillators with balanced gain and loss. Its envelope equations are

$$[PT,H]=0 \qquad\text{or equivalently}\qquad PH^*P=H,$$3

$$[PT,H]=0 \qquad\text{or equivalently}\qquad PH^*P=H,$$4

with conserved $$[PT,H]=0 \qquad\text{or equivalently}\qquad PH^*P=H,$$5 and $$[PT,H]=0 \qquad\text{or equivalently}\qquad PH^*P=H,$$6, linearizable Stokes dynamics, and a power-dependent threshold

$$[PT,H]=0 \qquad\text{or equivalently}\qquad PH^*P=H,$$7

In that model, stable periodic and quasiperiodic states with sufficiently large amplitudes persist for arbitrarily large gain–loss coefficient (Barashenkov et al., 2014). This suggests that nonlinear coupling can soften, or in some families effectively restore, PT-symmetric bounded dynamics far beyond the linear exceptional point.

5. Time-periodic gain/loss, Floquet structure, and fractional-time variants

When the gain–loss coefficient is time dependent, the PT dimer develops a Floquet structure not present in the autonomous case. For the generalized nonlinear dimer

$$[PT,H]=0 \qquad\text{or equivalently}\qquad PH^*P=H,$$8

writing $$[PT,H]=0 \qquad\text{or equivalently}\qquad PH^*P=H,$$9, defining

$P=\sigma_x$0

and using the conserved quantity

$P=\sigma_x$1

gives the reduced system

$P=\sigma_x$2

This is the perturbed-pendulum form on which Melnikov-type and topological-degree analyses are built (Battelli et al., 2014).

A central qualitative result is that time-periodic gain/loss can preserve rotation modes that do not exist in the constant-gain–loss dimer. If the unperturbed pendulum orbit satisfies the resonance condition $P=\sigma_x$3, a simple zero of the Melnikov function

$P=\sigma_x$4

or its shift-periodic analogue implies persistence of nearby periodic or shift-periodic solutions (Battelli et al., 2014). The numerical results summarized in the source show periodic islands, chaotic layers near heteroclinic separatrices, and bounded shift-periodic rotation modes for $P=\sigma_x$5, whereas bounded rotations are destroyed by constant $P=\sigma_x$6 (Battelli et al., 2014).

A related linear-Floquet study uses a piecewise-constant periodic protocol,

$P=\sigma_x$7

usually specialized to $P=\sigma_x$8. In the linear Schrödinger dimer

$P=\sigma_x$9

the PT-transition boundaries satisfy

$T=K$0

in the exact phase, while for $T=K$1 only a single small stable region remains (Psiachos et al., 2014). In contrast, the classical oscillator dimer displays decreased stability at first, then re-entrant transitions between exact and broken PT phases as the coupling is increased, showing that identical modulation protocols do not produce identical stability diagrams across different dimer realizations (Psiachos et al., 2014).

Rapid modulation of the PT-symmetric part can also be treated by averaging. For

$T=K$2

one obtains an averaged autonomous dimer with renormalized couplings and nonlinearities. For cosine modulation $T=K$3,

$T=K$4

so the exact PT region expands from $T=K$5 to

$T=K$6

In the averaged model the nonlinear PT transition is a saddle-center bifurcation at $T=K$7 (Horne et al., 2013).

A different generalization replaces the first-order time derivative by a Caputo derivative of order $T=K$8. For the linear PT dimer

$T=K$9

the exact solution is expressed in Mittag-Leffler functions. The bounded/unbounded distinction remains controlled by the sign of $t>0$00, but for $t>0$01 oscillations acquire a monotonically decreasing envelope, so the standard sustained exchange of the ordinary dimer is replaced by damped exchange (Molina, 2021).

6. Fluxes, scattering, causality, and experimentally open dimers

In photonic dimers, flux-based observables distinguish internal gain/loss from inter-site transport. For coupled amplitudes

$t>0$02

the longitudinal flux is

$t>0$03

while the transverse flux is

$t>0$04

The PT transition follows from the intensity pattern of the eigenmode, because for an eigenstate

$t>0$05

Thus real versus complex eigenvalues are determined by whether intensities on the gain and loss waveguides are equal, and the paper describing this identifies the transition as “classical” in that specific sense (Ge et al., 2017).

The same flux analysis shows that the commonly cited “giant amplification” near the exceptional point is sub-exponential and does not outperform a single gain waveguide with the same gain coefficient. It also shows that apparent oscillatory exchange can conceal unidirectional transverse transport: if

$t>0$06

then $t>0$07 never changes sign and flows only from the gain waveguide to the loss waveguide (Ge et al., 2017). Another correction to common intuition is that power transfer does not become arbitrarily fast near the exceptional point; the shortest peak-to-peak transfer distance remains bounded by $t>0$08 (Ge et al., 2017).

When the dimer is coupled to transmission lines or external ports, it becomes a genuinely open scatterer. In PT-symmetric electronics, two inductively coupled active LRC circuits—one with negative resistance and one with matching positive resistance—realize the dimer experimentally. The scattering amplitudes from left and right obey the generalized PT relation

$t>0$09

and the same device can function as an amplifier or an absorber depending on the direction and phase relation of the interrogating waves (Schindler et al., 2012). In the two-port configuration, a transfer-matrix condition

$t>0$10

identifies a simultaneous laser/CPA point (Schindler et al., 2012).

A recent one-port formulation sharpens the relation between PT symmetry and analyticity. There the dimer

$t>0$11

is opened by coupling site 1 only to an external waveguide, producing

$t>0$12

The exact one-port reflection coefficient is

$t>0$13

with poles given by the roots of the denominator $t>0$14 stated in the source (Liu, 30 Apr 2026). When

$t>0$15

one pole enters the upper half-plane, the Blaschke winding number jumps from $t>0$16 to $t>0$17, standard Kramers–Kronig reconstruction fails, and the missing contribution is a residue-controlled Lorentzian term,

$t>0$18

The violation norm obeys

$t>0$19

in the single-port geometry (Liu, 30 Apr 2026). This places the PT-symmetric open dimer at the intersection of exceptional-point physics, meromorphic response theory, and experimentally measurable causality diagnostics.

7. Extended dimers, lattices, and topological or many-body generalizations

The PT-symmetric dimer is also a building block for larger structures. In the PT-symmetric dimer lattice, each unit cell contains a gain site and a loss site with intra-dimer coupling $t>0$20, inter-dimer coupling $t>0$21, and balanced onsite terms $t>0$22. The Bloch Hamiltonian is

$t>0$23

with eigenvalues

$t>0$24

The PT-breaking threshold is

$t>0$25

in the infinite-lattice limit, while the topological transition occurs at

$t>0$26

The literature summarized here emphasizes that these are independent transitions, and that the topological transition of the PT-symmetric dimer lattice occurs inside the PT-broken phase because $t>0$27 at $t>0$28 (Harter et al., 2016). Sublattice-resolved intensity profiles then reveal bimodal and trimodal structures at the topological point, depending on whether the initial state is on the gain or loss site.

Cyclic arrays of replicated dimers furnish another extension. In necklaces of PT-symmetric dimers, a $t>0$29-waveguide cyclic array decomposes exactly into $t>0$30 independent effective dimers after Fourier transformation of the cyclic sector. Each block has Hamiltonian

$t>0$31

with effective coupling

$t>0$32

The corresponding eigenvalues are

$t>0$33

For even $t>0$34 and homogeneous couplings $t>0$35, at least one effective dimer is always in the broken-symmetry phase, whereas changing $t>0$36 or the coupling phases can restore PT symmetry (Stevens et al., 2017). Because a properly phased input selects a single Fourier sector, the necklace acts as an output-port replicator of the chosen effective dimer.

The dimer paradigm also extends beyond single-particle or waveguide settings. A coupled pair of open Lipkin–Meshkov–Glick models forms a PT-symmetric open dimer of collective spins with one lossy and one gain channel,

$t>0$37

Its mean-field dynamics combine an LMG transition driven by $t>0$38 with a PT transition driven by $t>0$39, producing a phase diagram with a normal phase, two LMG symmetry-broken phases, and a PT-dominated regime with limit cycles and chaos at mean-field level (Kothe et al., 25 Apr 2025). This suggests that the dimer concept remains structurally useful even in many-body dissipative settings where the natural observables are steady-state order parameters rather than modal intensities.

Across these extensions, one misconception is consistently corrected: the dimer is not merely a two-site toy model. In the open-chain quantization problem it behaves as an effective size-3 system; in planar photonics it is the smallest representative of a Lorentz-group-related $t>0$40-waveguide family; in lattices it coexists with independent topological structure; and in open scattering setups it supports analyticity transitions detectable through reflection poles (Ge, 2024). Its continued use across such settings reflects not simplicity alone, but the fact that balanced gain/loss, exceptional points, symmetry-enforced spectral pairings, and nonlinear or topological deformations can all be studied in exact or nearly exact form within the dimer architecture.