Dicke State Ansatz in Quantum Many-Body Systems

Updated 9 July 2026

Dicke State Ansatz is a family of symmetry-adapted constructions using permutation-symmetric Dicke states to compress a 2^N state space into an analytically tractable collective subspace.
It is employed in diverse applications ranging from exact integrable models via Bethe and Richardson–Gaudin formulations to variational quantum algorithms that enforce fixed Hamming-weight constraints.
Efficient circuit designs and distributed state-preparation methods have been developed to realize Dicke state ansätze experimentally, enhancing scalability in simulating complex quantum dynamics.

“Dicke state ansatz” denotes a family of symmetry-adapted state constructions built from Dicke states, namely permutation-symmetric fixed-excitation states of $N$ identical two-level systems, together with their direct generalizations to higher local dimension. In the literature, the expression is used in several closely related senses: as a restriction of dynamics to the symmetric $J=N/2$ manifold of a Dicke-type light–matter model; as an exact Bethe- or Richardson–Gaudin-type product state for integrable spin–boson Hamiltonians; and as a variational state family that enforces fixed Hamming-weight constraints in quantum algorithms (Seidov et al., 2023, Tsyplyatyev et al., 2010, Scursulim et al., 19 Aug 2025). The common structure is the replacement of a generic $2^N$ -dimensional many-body description by a collective, permutation-symmetric, or fixed-occupation subspace in which the relevant amplitudes can be represented analytically, semiclassically, or by compact circuits.

1. Dicke states as the underlying state space

For $N$ identical two-level systems, the standard Dicke basis is the fully symmetric angular-momentum basis

$|J,M\rangle,\qquad J=\frac{N}{2},\qquad M=-J,-J+1,\dots,J,$

with

$S^2|J,M\rangle = J(J+1)|J,M\rangle,\qquad S_z|J,M\rangle = M|J,M\rangle.$

Equivalently, a state with $k$ excitations is the equal-weight superposition of all computational-basis strings with exactly $k$ spins up, so that $M=-N/2+k$ (Seidov et al., 2023). In $n$ -qubit notation, the symmetric Dicke state with $J=N/2$ 0 excitations can be written as

$J=N/2$ 1

and the $J=N/2$ 2 case is the $J=N/2$ 3 state (Saha et al., 2018). The symmetric qubit subspace has dimension $J=N/2$ 4, and $J=N/2$ 5 furnishes a natural basis for it (Mukherjee et al., 2020).

The same construction extends to qudits. For local dimension $J=N/2$ 6, a qudit Dicke state is labeled by a composition vector $J=N/2$ 7 with $J=N/2$ 8, and is the normalized equal-weight superposition over all computational-basis strings whose occupation numbers are specified by $J=N/2$ 9. The fully symmetric qudit subspace then has dimension

$2^N$ 0

so the Dicke basis becomes the occupation-number basis of $2^N$ 1 (Nepomechie et al., 2023). In this sense, a Dicke state ansatz is a symmetry-preserving restriction to amplitudes supported on one or more fixed-occupation sectors.

2. Collective-spin ansatz in Dicke and extended Dicke models

In Dicke-model many-body physics, the ansatz typically consists of replacing an ensemble of $2^N$ 2 identical, collectively coupled two-level systems by a single large “superspin” of length $2^N$ 3, confined to the symmetric Dicke manifold. In the quantum-battery analysis of the Dicke model, this restriction is made explicit through the quasiclassical parametrization

$2^N$ 4

together with the statement that the many-body state remains in $2^N$ 5 (Seidov et al., 2023). In the dispersive regime $2^N$ 6, the model reduces to an effective Lipkin–Meshkov–Glick Hamiltonian

$2^N$ 7

for the standard Dicke battery case, and the charging dynamics is then analyzed entirely inside the collective-spin sector (Seidov et al., 2023).

Within that reduced manifold, the “bound luminosity” state is a specific collective dynamical ansatz: a periodically beating state in which a superradiant photonic condensate and the TLS superspin exchange energy coherently. The exact semiclassical solution is written in terms of Jacobi elliptic functions, the charging time scales as $2^N$ 8, and the charging power scales as $2^N$ 9 (Seidov et al., 2023). In the extended Dicke model, the same logic survives after inclusion of the direct $N$ 0 interaction term; under the adiabatic condition $N$ 1, the reduced spin equations again admit Jacobi-elliptic solutions, now describing periodic transfer of energy between the cavity field and the ensemble of two-level systems (S. et al., 2022).

Related uses of the same collective-state restriction appear outside the battery setting. In optical lattices, a probe photon can be absorbed collectively to create a timed Dicke state

$N$ 2

whose subsequent deformation under Bose– or Fermi–Hubbard dynamics is read out through superradiant emission, thereby turning the Dicke-state construction into a nondestructive probe of many-body coherence (Brinke et al., 2015). In ESQPT studies of Dicke superradiance models, the partition sum is explicitly restricted to Dicke states with maximal $N$ 3, and the resulting semiclassical density of states and its singularities are derived from that restriction (Brandes, 2013).

3. Exact integrable meanings of the ansatz

A distinct use of “Dicke state ansatz” appears in integrable formulations of the Dicke model, where the ansatz is not variational but exact. In the inhomogeneous Dicke model studied by Tsyplyatyev, von Delft, and Loss, the eigenstate is assumed to have the Bethe-product form

$N$ 4

with coefficients chosen as

$N$ 5

Commutation relations then yield the exact Bethe equations

$N$ 6

and the many-body eigenenergy is $N$ 7 (Tsyplyatyev et al., 2010). Here the ansatz is a product of collective spin–photon excitations, generalizing symmetric Dicke states to the inhomogeneous case.

The Richardson–Gaudin construction yields the same theme in a more algebraic form. By pseudo-deforming one $N$ 8 quasispin copy into a bosonic mode and taking the contraction limit, the Dicke Hamiltonian emerges together with a Bethe ansatz state

$N$ 9

where the rapidities $|J,M\rangle,\qquad J=\frac{N}{2},\qquad M=-J,-J+1,\dots,J,$ 0 satisfy the associated Richardson–Gaudin equations

$|J,M\rangle,\qquad J=\frac{N}{2},\qquad M=-J,-J+1,\dots,J,$ 1

(Claeys et al., 2014). In this integrable usage, “Dicke state ansatz” refers to an exact algebraic wavefunction built from collective spin–boson creation operators, not to a low-dimensional approximation.

This distinction is conceptually important. In the collective-spin semiclassical literature, the ansatz is a restriction to the symmetric sector followed by quasiclassical dynamics; in the Bethe and Richardson–Gaudin literature, the ansatz is an exact eigenstate construction inside the full integrable model (Seidov et al., 2023, Claeys et al., 2014).

4. Variational and constraint-preserving ansätze in quantum algorithms

In variational quantum computing, Dicke states serve as problem-inspired ansätze because fixed Hamming weight translates directly into hard combinatorial constraints. For Maximum $|J,M\rangle,\qquad J=\frac{N}{2},\qquad M=-J,-J+1,\dots,J,$ 2-Vertex Cover under the Quantum Alternating Operator Ansatz, the initial state is the Dicke state

$|J,M\rangle,\qquad J=\frac{N}{2},\qquad M=-J,-J+1,\dots,J,$ 3

which lives entirely in the feasible subspace $|J,M\rangle,\qquad J=\frac{N}{2},\qquad M=-J,-J+1,\dots,J,$ 4. Combined with Hamming-weight-preserving $|J,M\rangle,\qquad J=\frac{N}{2},\qquad M=-J,-J+1,\dots,J,$ 5 mixers, this keeps the evolution inside the legal solution space. Numerically, Dicke states improve performance compared to easy-to-prepare classical starting states, and the complete graph mixer improves performance relative to the ring mixer (1910.13483).

A more explicitly variational usage appears in multiclass portfolio optimization via VQE. There the single-class parameterized Dicke ansatz is

$|J,M\rangle,\qquad J=\frac{N}{2},\qquad M=-J,-J+1,\dots,J,$ 6

and the multiclass construction is the tensor product

$|J,M\rangle,\qquad J=\frac{N}{2},\qquad M=-J,-J+1,\dots,J,$ 7

Because every computational basis state in the support has exactly $|J,M\rangle,\qquad J=\frac{N}{2},\qquad M=-J,-J+1,\dots,J,$ 8 ones in class $|J,M\rangle,\qquad J=\frac{N}{2},\qquad M=-J,-J+1,\dots,J,$ 9, diversification constraints are satisfied by construction, allowing the penalty coefficient to be set to $S^2|J,M\rangle = J(J+1)|J,M\rangle,\qquad S_z|J,M\rangle = M|J,M\rangle.$ 0. The parameter count is

$S^2|J,M\rangle = J(J+1)|J,M\rangle,\qquad S_z|J,M\rangle = M|J,M\rangle.$ 1

and among the tested optimizers CMA-ES performs best overall (Scursulim et al., 19 Aug 2025).

The mixed-state extension generalizes this principle from equality to inequality constraints. A coherent ancilla-assisted superposition over Dicke sectors is traced down to the density matrix

$S^2|J,M\rangle = J(J+1)|J,M\rangle,\qquad S_z|J,M\rangle = M|J,M\rangle.$ 2

so the pure Dicke ansatz is recovered as the special case with a single nonzero $S^2|J,M\rangle = J(J+1)|J,M\rangle,\qquad S_z|J,M\rangle = M|J,M\rangle.$ 3. Tensor products of such pure or mixed Dicke blocks encode multiple equality and inequality constraints simultaneously, eliminating penalty terms in the objective function and preserving feasibility structurally rather than energetically (Scursulim, 7 Jun 2026). This suggests a broader design principle: Dicke-state ansätze are especially effective when the feasible set is itself a fixed-occupation manifold.

5. Preparation circuits, scalability, and architectural realizations

The practical utility of a Dicke-state ansatz depends on whether the corresponding state family can be prepared efficiently. For qubits, deterministic preparation circuits with improved gate counts were derived by exploiting partially defined unitary transformations. For $S^2|J,M\rangle = J(J+1)|J,M\rangle,\qquad S_z|J,M\rangle = M|J,M\rangle.$ 4, the improved circuit has

$S^2|J,M\rangle = J(J+1)|J,M\rangle,\qquad S_z|J,M\rangle = M|J,M\rangle.$ 5

A divide-and-conquer strategy refines this further by first distributing Hamming weight between blocks and then applying improved Dicke-state unitaries on those blocks. On IBM Quantum Sydney and Montreal devices, this approach yielded significantly higher state fidelity than earlier results for systems up to $k$ 3 qubits, while remaining compatible with linear nearest-neighbor topologies favored by heavy-hex connectivity (Aktar et al., 2021). For higher local dimension, deterministic qudit Dicke-state preparation is also available through a recursive universal unitary $k$ 4 satisfying

$k$ 5

together with explicit elementary-gate decompositions for the qubit and qutrit cases (Nepomechie et al., 2023).

Distributed preparation extends the ansatz beyond a single processor. For $k$ 6 QPUs, the state $k$ 7 can be prepared with communication complexity $k$ 8, circuit size $k$ 9, and depth

$k$ 0

and for $k$ 1 the lower bound $k$ 2 on communication complexity is matched exactly (Chen et al., 28 Jan 2026). These results turn Dicke states from formal symmetry objects into scalable circuit modules that can function as state-preparation layers inside larger ansatz architectures.

6. Experimental realization, certification, and conceptual boundaries

Experimentally, symmetric Dicke states have been realized and characterized directly. A six-photon implementation of the symmetric state

$k$ 3

reached a fidelity of $k$ 4, while a two-setting witness yielded $k$ 5, proving genuine six-photon entanglement. The same resource state was shown to generate lower-qubit Dicke, $k$ 6, and GHZ-like states through local projective measurements, emphasizing its role as a structured multipartite resource (0903.2213). Device-independent certification has also been developed through permutation-invariant Bell operators tailored to Dicke states; for example, a three-qubit $k$ 7 state is an eigenstate of the corresponding Bell operator with expectation value $k$ 8 (Saha et al., 2018).

Hamiltonian engineering provides a different preparation route. In counterdiabatic driving, an initial spin coherent state—the ground state of a linear Hamiltonian—is driven toward the ground state of a quadratic Hamiltonian $k$ 9, whose ground state is the target Dicke state $M=-N/2+k$ 0. With a small set of compensating operators $M=-N/2+k$ 1, the protocol can suppress diabatic leakage strongly; for $M=-N/2+k$ 2 and $M=-N/2+k$ 3, the four-operator protocol reached $M=-N/2+k$ 4 and squeezing of about $M=-N/2+k$ 5 dB for the central Dicke target (Opatrný et al., 2015).

The scope of the ansatz also has clear boundaries. Superradiant phenomena do not require an initial Dicke state with zero dipole moment: non-Dicke initial states with nonzero dipole moment can also display a superradiant burst, with the relevant organizing principle being reduction of quantum-phase dispersion and constructive interference of envelope phases (Nefedkin et al., 2016). A common misconception is therefore to identify “Dicke state ansatz” with a single universal object. In practice, the term names a family of constructions: exact integrable product states, symmetric-sector semiclassical reductions, and constraint-preserving variational circuits all qualify, but they solve different problems and should not be conflated (Tsyplyatyev et al., 2010, Seidov et al., 2023, Scursulim, 7 Jun 2026).