Two-Qubit Non-Hermitian XY Model

• The Two-Qubit Non-Hermitian XY Model is a set of exactly solvable spin-½ dimer Hamiltonians featuring complex exchange couplings, imaginary fields, and measurement-induced dynamics.
• It encompasses diverse formulations including RT/PT-symmetric constructions and heralded no-jump protocols that yield invariant block structures and exceptional spectral degeneracies.
• The model underpins practical applications in entanglement generation and quantum thermodynamic engines by enabling controlled Bell-state selection and thermal mode switching.

Searching arXiv for the specified papers and closely related two-qubit non-Hermitian XY models. First search: measurement-induced non-Hermitian entanglement generation and two-qubit XY realizations. Second search: RT/PT-symmetric non-Hermitian XY dimers, geometric phases, and thermodynamic applications. The two-qubit non-Hermitian XY model denotes a class of spin-$\tfrac12$ dimer Hamiltonians in which the XY exchange sector is rendered non-Hermitian by complex anisotropy, imaginary local $z$-fields, conditional decay terms, or measurement-induced postselection. In the recent literature, this label covers measurement-induced effective dimers with complex exchange couplings and imaginary local fields, RT-symmetric anisotropic XY dimers obtained as the $N=2$ reduction of non-Hermitian chains, PT-symmetric dimers with a staggered imaginary field, and heralded no-jump XY dimers generated by spontaneous decay. Across these realizations, the common technical structures are $2\times2$ invariant subspaces, biorthogonal spectral decompositions, exceptional points in specific parameter regimes, and Bell-state selection under normalized non-unitary dynamics (Grimaudo et al., 2020, Zhang et al., 2012, Li et al., 2022, Lee et al., 2014).

1. Canonical Hamiltonian forms

In the two-qubit setting, the non-Hermitian XY label does not refer to a single universal Hamiltonian. Instead, it refers to several analytically tractable dimer models that share an XY-type spin-exchange core and a non-Hermitian deformation. A general measurement-induced form is

$$H_{\mathrm{XY}} = J_x \sigma_1^x \sigma_2^x + J_y \sigma_1^y \sigma_2^y + h_1^z \sigma_1^z + h_2^z \sigma_2^z + J_z \sigma_1^z \sigma_2^z,$$

with complex $J_x,J_y$ and, in general, complex local fields $h_1^z,h_2^z$. In the symmetric ancilla-induced construction, the identifications are

$$J_x=J_y=\gamma_{xy}-i\tau g_{xy}^2,\qquad J_z=\gamma_z,\qquad h_1^z=h_2^z=g_z+i\tau g_{xy}^2.$$

A distinct RT-symmetric dimer arises from the $N=2$ limit of a non-Hermitian anisotropic XY chain with periodic boundary conditions: $$H_2=J\Big[(1+i\gamma)\,\sigma_1^x\sigma_2^x+(1-i\gamma)\,\sigma_1^y\sigma_2^y\Big] +J\lambda\left(\sigma_1^z+\sigma_2^z\right).$$ Here the non-Hermiticity is encoded directly in the complex anisotropy, with $z$0 and $z$1.

A PT-symmetric formulation uses a real transverse field and a staggered imaginary field,

$z$2

or, equivalently, in reduced parameters $z$3, $z$4,

$z$5

A further realization is the heralded no-jump XY dimer. There the effective Hamiltonian has the generic form

$z$6

so the non-Hermiticity arises from conditioning on no fluorescence events rather than from a static gain-loss term in a closed model. In all of these formulations, the computational basis $z$7 or the equivalent spin basis $z$8 yields a block structure that makes the two-qubit problem exactly solvable (Grimaudo et al., 2020, Zhang et al., 2012, Li et al., 2022, Tusun et al., 21 Jun 2026, Lee et al., 2014).

2. Measurement-induced construction from system–ancilla dynamics

A physically explicit route to a two-qubit non-Hermitian XY model starts from a Hermitian three-spin Hamiltonian and repeated measurements on an ancilla spin. The system–ancilla unitary is

$z$9

and a projective measurement on the ancilla with selected outcome $N=2$0 defines the Kraus operator

$N=2$1

After $N=2$2 successful repetitions, the non-normalized conditional state is

$N=2$3

In the stroboscopic limit,

$N=2$4

with

$N=2$5

so that repeated measurements generate the effective non-Hermitian Hamiltonian

$N=2$6

For the symmetric three-spin model with pairwise XXZ-like couplings,

$N=2$7

preparing and repeatedly measuring the ancilla in the $N=2$8-basis with outcome $N=2$9 yields, up to an identity shift,

$2\times2$0

The imaginary terms arise from the positive semidefinite decay operator $2\times2$1 generated by ancilla flips due to the XY couplings $2\times2$2. Dropping the identity term affects the postselection probability but not the normalized conditional dynamics.

The anisotropic generalization starts from

$2\times2$3

which induces

$2\times2$4

This yields an anisotropic XY-like dimer with complex couplings and complex local $2\times2$5-fields. A key structural point is that XX/YY-type system–ancilla couplings together with ancilla measurements in $2\times2$6 generate precisely the imaginary XY exchange terms and imaginary $2\times2$7-fields characteristic of the non-Hermitian two-qubit XY model (Grimaudo et al., 2020).

3. Symmetry classes, invariant blocks, and exceptional points

Two-qubit non-Hermitian XY dimers split naturally into symmetry sectors. In the RT-symmetric anisotropic dimer, the operator

$2\times2$8

satisfies $2\times2$9, although the Hamiltonian is not PT-symmetric with the usual lattice parity. In the computational basis, $$H_{\mathrm{XY}} = J_x \sigma_1^x \sigma_2^x + J_y \sigma_1^y \sigma_2^y + h_1^z \sigma_1^z + h_2^z \sigma_2^z + J_z \sigma_1^z \sigma_2^z,$$0 is block diagonal: $$H_{\mathrm{XY}} = J_x \sigma_1^x \sigma_2^x + J_y \sigma_1^y \sigma_2^y + h_1^z \sigma_1^z + h_2^z \sigma_2^z + J_z \sigma_1^z \sigma_2^z,$$1 The single-excitation block $$H_{\mathrm{XY}} = J_x \sigma_1^x \sigma_2^x + J_y \sigma_1^y \sigma_2^y + h_1^z \sigma_1^z + h_2^z \sigma_2^z + J_z \sigma_1^z \sigma_2^z,$$2 has real eigenvalues $$H_{\mathrm{XY}} = J_x \sigma_1^x \sigma_2^x + J_y \sigma_1^y \sigma_2^y + h_1^z \sigma_1^z + h_2^z \sigma_2^z + J_z \sigma_1^z \sigma_2^z,$$3 with eigenvectors $$H_{\mathrm{XY}} = J_x \sigma_1^x \sigma_2^x + J_y \sigma_1^y \sigma_2^y + h_1^z \sigma_1^z + h_2^z \sigma_2^z + J_z \sigma_1^z \sigma_2^z,$$4. The zero/double-excitation block $$H_{\mathrm{XY}} = J_x \sigma_1^x \sigma_2^x + J_y \sigma_1^y \sigma_2^y + h_1^z \sigma_1^z + h_2^z \sigma_2^z + J_z \sigma_1^z \sigma_2^z,$$5 has

$$H_{\mathrm{XY}} = J_x \sigma_1^x \sigma_2^x + J_y \sigma_1^y \sigma_2^y + h_1^z \sigma_1^z + h_2^z \sigma_2^z + J_z \sigma_1^z \sigma_2^z,$$6

so the unbroken RT-symmetric regime is $$H_{\mathrm{XY}} = J_x \sigma_1^x \sigma_2^x + J_y \sigma_1^y \sigma_2^y + h_1^z \sigma_1^z + h_2^z \sigma_2^z + J_z \sigma_1^z \sigma_2^z,$$7, the broken regime is $$H_{\mathrm{XY}} = J_x \sigma_1^x \sigma_2^x + J_y \sigma_1^y \sigma_2^y + h_1^z \sigma_1^z + h_2^z \sigma_2^z + J_z \sigma_1^z \sigma_2^z,$$8, and the exceptional point occurs at $$H_{\mathrm{XY}} = J_x \sigma_1^x \sigma_2^x + J_y \sigma_1^y \sigma_2^y + h_1^z \sigma_1^z + h_2^z \sigma_2^z + J_z \sigma_1^z \sigma_2^z,$$9, where the block becomes a Jordan block. At $J_x,J_y$0, the coalesced eigenvector is

$J_x,J_y$1

The generalized RT-symmetric chain with deformation parameters $J_x,J_y$2 and $J_x,J_y$3 reduces at $J_x,J_y$4 to a pair sector with

$J_x,J_y$5

under periodic boundary conditions. The exceptional line is $J_x,J_y$6, and the corresponding unbroken region is $J_x,J_y$7. In this formulation, RT symmetry is generated by

$J_x,J_y$8

with $J_x,J_y$9.

In the PT-symmetric staggered-field dimer, parity is the qubit-swap operator and time reversal is complex conjugation. The Hamiltonian decomposes exactly as

$h_1^z,h_2^z$0

with

$h_1^z,h_2^z$1

Its spectrum is

$h_1^z,h_2^z$2

The working pair is always real. The idle pair is real for $h_1^z,h_2^z$3, degenerate at zero for $h_1^z,h_2^z$4, and forms a complex-conjugate pair for $h_1^z,h_2^z$5. Thus $h_1^z,h_2^z$6 is the PT exceptional point.

The heralded no-jump dimer provides a different non-Hermitian mechanism. For $h_1^z,h_2^z$7 with periodic boundary conditions,

$h_1^z,h_2^z$8

The pairing block $h_1^z,h_2^z$9 has

$$J_x=J_y=\gamma_{xy}-i\tau g_{xy}^2,\qquad J_z=\gamma_z,\qquad h_1^z=h_2^z=g_z+i\tau g_{xy}^2.$$0

and exhibits an exceptional point at

$$J_x=J_y=\gamma_{xy}-i\tau g_{xy}^2,\qquad J_z=\gamma_z,\qquad h_1^z=h_2^z=g_z+i\tau g_{xy}^2.$$1

By contrast, the flip-flop block has

$$J_x=J_y=\gamma_{xy}-i\tau g_{xy}^2,\qquad J_z=\gamma_z,\qquad h_1^z=h_2^z=g_z+i\tau g_{xy}^2.$$2

and its level meeting at $$J_x=J_y=\gamma_{xy}-i\tau g_{xy}^2,\qquad J_z=\gamma_z,\qquad h_1^z=h_2^z=g_z+i\tau g_{xy}^2.$$3 is not an exceptional point because the block remains diagonalizable. One recurring misconception is therefore avoided explicitly in this literature: eigenvalue degeneracy alone is not sufficient for an exceptional point; non-diagonalizability is essential (Zhang et al., 2012, Zhang et al., 2013, Li et al., 2022, Tusun et al., 21 Jun 2026, Lee et al., 2014).

4. Entanglement generation, Bell sectors, and long-time selection

In the measurement-induced symmetric dimer, the invariant $$J_x=J_y=\gamma_{xy}-i\tau g_{xy}^2,\qquad J_z=\gamma_z,\qquad h_1^z=h_2^z=g_z+i\tau g_{xy}^2.$$4 block is

$$J_x=J_y=\gamma_{xy}-i\tau g_{xy}^2,\qquad J_z=\gamma_z,\qquad h_1^z=h_2^z=g_z+i\tau g_{xy}^2.$$5

Its right eigenvalues and right eigenvectors are

$$J_x=J_y=\gamma_{xy}-i\tau g_{xy}^2,\qquad J_z=\gamma_z,\qquad h_1^z=h_2^z=g_z+i\tau g_{xy}^2.$$6

Because

$$J_x=J_y=\gamma_{xy}-i\tau g_{xy}^2,\qquad J_z=\gamma_z,\qquad h_1^z=h_2^z=g_z+i\tau g_{xy}^2.$$7

postselection-normalized dynamics selects the growing eigenmode $$J_x=J_y=\gamma_{xy}-i\tau g_{xy}^2,\qquad J_z=\gamma_z,\qquad h_1^z=h_2^z=g_z+i\tau g_{xy}^2.$$8, so

$$J_x=J_y=\gamma_{xy}-i\tau g_{xy}^2,\qquad J_z=\gamma_z,\qquad h_1^z=h_2^z=g_z+i\tau g_{xy}^2.$$9

For an initial state $N=2$0, the concurrence evolves as

$N=2$1

with $N=2$2 and $N=2$3, so $N=2$4. In the anisotropic construction, suitable choices of $N=2$5 can instead select

$N=2$6

This asymptotic Bell-state purification has a direct postselection cost. The success probability is

$N=2$7

Increasing $N=2$8 while keeping $N=2$9 increases the selection rate $$H_2=J\Big[(1+i\gamma)\,\sigma_1^x\sigma_2^x+(1-i\gamma)\,\sigma_1^y\sigma_2^y\Big] +J\lambda\left(\sigma_1^z+\sigma_2^z\right).$$0 but decreases $$H_2=J\Big[(1+i\gamma)\,\sigma_1^x\sigma_2^x+(1-i\gamma)\,\sigma_1^y\sigma_2^y\Big] +J\lambda\left(\sigma_1^z+\sigma_2^z\right).$$1 more rapidly. In the Hermitian limit $$H_2=J\Big[(1+i\gamma)\,\sigma_1^x\sigma_2^x+(1-i\gamma)\,\sigma_1^y\sigma_2^y\Big] +J\lambda\left(\sigma_1^z+\sigma_2^z\right).$$2 or $$H_2=J\Big[(1+i\gamma)\,\sigma_1^x\sigma_2^x+(1-i\gamma)\,\sigma_1^y\sigma_2^y\Big] +J\lambda\left(\sigma_1^z+\sigma_2^z\right).$$3, the same $$H_2=J\Big[(1+i\gamma)\,\sigma_1^x\sigma_2^x+(1-i\gamma)\,\sigma_1^y\sigma_2^y\Big] +J\lambda\left(\sigma_1^z+\sigma_2^z\right).$$4 sector undergoes only unitary oscillations with

$$H_2=J\Big[(1+i\gamma)\,\sigma_1^x\sigma_2^x+(1-i\gamma)\,\sigma_1^y\sigma_2^y\Big] +J\lambda\left(\sigma_1^z+\sigma_2^z\right).$$5

so no stationary Bell state emerges.

The EP-based Bell mechanism in the open-chain single-excitation model is structurally different. For $$H_2=J\Big[(1+i\gamma)\,\sigma_1^x\sigma_2^x+(1-i\gamma)\,\sigma_1^y\sigma_2^y\Big] +J\lambda\left(\sigma_1^z+\sigma_2^z\right).$$6, the relevant effective Hamiltonian is

$$H_2=J\Big[(1+i\gamma)\,\sigma_1^x\sigma_2^x+(1-i\gamma)\,\sigma_1^y\sigma_2^y\Big] +J\lambda\left(\sigma_1^z+\sigma_2^z\right).$$7

with eigenvalues

$$H_2=J\Big[(1+i\gamma)\,\sigma_1^x\sigma_2^x+(1-i\gamma)\,\sigma_1^y\sigma_2^y\Big] +J\lambda\left(\sigma_1^z+\sigma_2^z\right).$$8

At $$H_2=J\Big[(1+i\gamma)\,\sigma_1^x\sigma_2^x+(1-i\gamma)\,\sigma_1^y\sigma_2^y\Big] +J\lambda\left(\sigma_1^z+\sigma_2^z\right).$$9, the exceptional point produces the coalescing eigenvector

$z$00

or

$z$01

These states are Bell states up to local phase rotations about $z$02. In the PT-broken regime $z$03, normalized non-unitary evolution converges exponentially to the eigenvector with positive imaginary part, with convergence rate $z$04.

The staggered imaginary-field dimer yields a complementary entanglement diagnostic. In the odd sector,

$z$05

so the concurrence is exactly unity in the PT-unbroken region and develops a cusp at the exceptional point. In the even sector,

$z$06

which is independent of $z$07. This indicates that maximal entanglement is not tied uniquely to exceptionality: in some formulations it is selected asymptotically by imaginary spectral ordering, while in others it already characterizes an entire unbroken symmetry sector (Grimaudo et al., 2020, Li et al., 2015, Li et al., 2022).

5. Geometric phase, thermal structure, and quantum-engine applications

In the generalized RT-symmetric non-Hermitian XY model, the geometric phase is formulated biorthogonally through

$z$08

For the three-parameter example in cylindrical coordinates $z$09, the curvature has only an azimuthal component,

$z$10

At $z$11, only one nontrivial mode contributes, and the corresponding $z$12 diverges as $z$13 when approaching the exceptional line $z$14. The phase boundary is therefore identified with a divergence of the Berry curvature rather than merely with an algebraic change in the roots of the characteristic polynomial.

The PT-symmetric staggered-field dimer also supports a fully explicit thermal theory in the unbroken region. With

$z$15

the biorthogonal Gibbs partition function and internal energy are

$z$16

$z$17

This spectrum separates into a working pair $z$18, controlled by the real field $z$19, and an idle pair $z$20, controlled only by the non-Hermitian parameter $z$21.

That separation underlies the two-qubit non-Hermitian XY quantum Otto engine. Writing

$z$22

the exact work per cycle is

$z$23

The terms proportional to $z$24 cancel exactly from the numerator. The dependence on $z$25 enters only through the $z$26 denominators. Since

$z$27

decreases as $z$28 increases within the PT-unbroken phase, the idle-level gap is compressed and both work output and efficiency increase. For the parameter set $z$29, $z$30, $z$31, $z$32, $z$33, $z$34, the critical value is

$z$35

with $z$36 for $z$37 and $z$38 for $z$39. In the same example, the efficiency rises from about $z$40 at $z$41 to $z$42 at $z$43, and the limit $z$44 yields

$z$45

which is the Otto bound and remains below the Carnot bound. This suggests a broader significance of the two-qubit non-Hermitian XY model: it is not only a minimal EP-bearing spin dimer, but also a microscopic platform in which non-Hermiticity controls geometry, thermal occupation, and thermodynamic mode switching (Zhang et al., 2013, Tusun et al., 21 Jun 2026).

6. Inverse engineering, experimental platforms, and limitations

A general inverse-construction recipe makes the two-qubit non-Hermitian XY model physically simulable from a Hermitian enlarged system. Given a target finite-dimensional non-Hermitian Hamiltonian $z$46 on the system $z$47, one decomposes it into Hermitian and anti-Hermitian parts, chooses $z$48 such that the maximum Bohr frequency $z$49 of $z$50 satisfies $z$51, defines

$z$52

and constructs the Hermitian Hamiltonian on $z$53,

$z$54

Repeatedly measuring the ancilla in the $z$55-basis and postselecting outcome $z$56 reproduces $z$57 in the stroboscopic limit. For an anisotropic non-Hermitian XY target, the mapping is

$z$58

together with

$z$59

Several experimental platforms are identified as feasible. Trapped ions, superconducting qubits, and photonic qubits can implement spin-$z$60 interactions together with fast, high-fidelity projective measurements. For the PT-symmetric Otto-engine dimer, trapped-ion realizations use two hyperfine or Zeeman levels per ion, Mølmer–Sørensen interactions to engineer anisotropic XY couplings, AC-Stark shifts for the real field, and quantum-trajectory postselection to generate the effective staggered imaginary field. For the heralded no-jump model, a three-level scheme with $z$61, $z$62, and an auxiliary state $z$63 implements the effective non-Hermitian Hamiltonian by conditioning on the absence of fluorescence; trapped ions, cavity QED arrays, and Rydberg atoms in optical lattices are listed as candidate platforms.

The main limitations are formulation-dependent but technically sharp. In the measurement-induced construction, the stroboscopic approximation requires $z$64; larger $z$65 or explicitly time-dependent $z$66 invalidates the simple $z$67 form. Measurement errors and finite detection efficiency reduce postselection success probability and can perturb $z$68, although moderate imperfections primarily increase postselection cost rather than alter the selected Bell state. Decoherence during the inter-measurement intervals adds Lindbladian terms not captured by the idealized effective Hamiltonian. In the PT-symmetric engine model, one must remain within $z$69 for a real spectrum, and adiabaticity degrades as the idle gap $z$70 near the exceptional point; the explicit recommendation is to truncate experimentally at about $z$71. These caveats do not diminish the central result: two-qubit non-Hermitian XY models constitute a controlled and experimentally grounded setting in which non-Hermitian spectral theory, postselected dynamics, and entanglement generation can be analyzed exactly (Grimaudo et al., 2020, Tusun et al., 21 Jun 2026, Lee et al., 2014).