Foundations of quantum mechanics · Theory
Open quantum time crystals without external thermodynamic resources: go and no-go theorems
arXiv:2611.09540v1 ↗PDF ↗Submitted 6 November 2026
- Question
- Every known time crystal in an open quantum system is fed by something outside equilibrium: a periodic drive, a hot-cold temperature difference, or a reservoir prepared with coherence. That leaves open whether coupling a system to an ordinary equilibrium environment, with nothing else supplied, can make it oscillate forever when the isolated system does not.
- Result
- Yes, but only through the Hamiltonian. The authors build a chain of three-level sites, coupled to a zero-temperature environment through a time-independent, energy-conserving interaction, whose two-time correlation oscillates as \(\cos(\Omega t)/12\) in the thermodynamic limit while the isolated chain's is constant. They then prove that for any Lindblad equation with time-independent generators, time-translation covariance and detailed balance, the dissipator never creates a time crystal: in every stationary state, the open system oscillates only if the isolated system with the same, environment-modified Hamiltonian already does.
The paper in detail
A time crystal, in the definition of Watanabe and Oshikawa, is a stationary state in which a macroscopic order parameter \(A_N\), a sum of local terms, has a two-time correlation \(N^{-2}\langle A_N(t)A_N\rangle\) that tends to a nonconstant periodic function as the system grows. Isolated systems with short-range interactions have none in their ground or Gibbs states. Open-system proposals evade this, but each needs an outside supply of free energy: a time-dependent drive needs a clock, broken detailed balance needs reservoirs at different temperatures, and covariance-breaking jumps need a coherent reservoir.
The authors turn "no outside supply" into three conditions on the Lindblad equation: time-independent Hamiltonian and jump operators, each jump operator shifting energy by a definite amount (time-translation covariance), and detailed balance at some temperature. Standard derivations from an equilibrium bath with a time-independent coupling produce exactly this class, and a collision-model argument shows that breaking any condition costs nonequilibrium free energy.
The go theorem uses a ring of three-level sites (empty, up, down). The isolated Hamiltonian only counts occupied sites and occupied neighbours, so every arrangement of up and down spins on a full ring has the same energy, and the collective spin \(S^x\) does not move. The environment adds a field \(-\Omega S^{z}\), and dephasing jumps \(2n_i-1\) commute with everything relevant. In the uniform mixture of Dicke states, stationary for both dynamics, the correlation of \(S^x\) is exactly \(\frac{N+2}{12N}\cos\Omega t\) under the open dynamics and constant without the field. A microscopic version couples each site to a ground-state two-level system (which supplies the field) and a maximally mixed \(d\)-level system (which supplies the dephasing as \(d\to\infty\)).
The no-go theorem says the dissipator can only damp. Detailed balance makes the dissipator self-adjoint in the Gibbs inner product, and covariance makes it preserve each frequency sector, so every eigenmode of the open dynamics oscillates at a Hamiltonian frequency with an added decay. A correlation that stays periodic must come from undamped modes, and those give the same function as the purely unitary evolution. Whether the result holds beyond Markovian dynamics, and whether a physical system can realise the go model, are open.
Transferable methods
Dynamical symmetry fixes an undamped frequency
To show an oscillation that never decays, the authors need an operator that the full Lindbladian rotates without damping. A collective raising operator that the dissipator leaves alone and the Hamiltonian shifts by a fixed energy does exactly this, and its expectation in a stationary state is then a pure cosine.
With \(S^{+}=\sum_i|{\uparrow}\rangle_i\langle{\downarrow}|\), the model satisfies \([H_N,S^{\pm}]=\mp\Omega S^{\pm}\) and \([W_i,S^{\pm}]=0\). The Heisenberg dissipator \(\gamma\sum_i(W_iAW_i-A)\) vanishes on \(S^{\pm}\) because \(W_i^{2}=I\), so \(S^{\pm}(t)=e^{\mp i\Omega t}S^{\pm}\) and \[G(t)=\frac{1}{4N^{2}}\Bigl(e^{-i\Omega t}\langle S^{+}S^{-}\rangle+e^{i\Omega t}\langle S^{-}S^{+}\rangle\Bigr).\]
On a Dicke state with \(q\) up spins, \(\langle S^{+}S^{-}\rangle=q(N-q+1)\) and \(\langle S^{-}S^{+}\rangle=(q+1)(N-q)\). Averaging uniformly over \(q=0,\dots,N\) makes the two equal, and the sum gives \(\frac{N+2}{12N}\cos\Omega t\). Editorial check: exact evolution of the three-site Lindbladian (27 levels, generic \(u,J,\gamma,\Omega\)) reproduces this value at \(t=0\), \(0.7\) and \(2\) to \(10^{-15}\).
How to reuse it
To certify a persistent oscillation in any Lindblad model, look for an operator \(X\) with \([H,X]=\omega X\) that every dissipative term annihilates; such dynamical symmetries give undamped modes regardless of the rates.
Mode decomposition under detailed balance
The no-go theorem has to compare open and closed evolutions in the same state for every observable. Splitting operators into frequency sectors (modes of asymmetry) and using detailed balance to make the dissipator self-adjoint within each sector turns the open correlation into the closed one with extra decay factors.
For \(X=\sum_\omega X_\omega\) with \(X_\omega=\sum_EP_EXP_{E+\omega}\), covariance means \(\mathcal L^{\dagger}\) maps each sector to itself, and detailed balance makes \(\mathcal D^{\dagger}\) self-adjoint and negative in \(\langle X,Y\rangle_\pi=\mathrm{Tr}[\pi X^{\dagger}Y]\). Choosing eigenoperators \(R_{\omega,a}\) with \(\mathcal L^{\dagger}R_{\omega,a}=(-i\omega-\lambda_{\omega,a})R_{\omega,a}\), the Gibbs-state correlations are \[G^{\mathcal L}(t)=\sum_{\omega,a}v_{\omega,a}e^{(i\omega-\lambda_{\omega,a})t},\qquad G^{\mathcal H}(t)=\sum_{\omega,a}v_{\omega,a}e^{i\omega t},\] with the same nonnegative weights \(v_{\omega,a}\).
Editorial remark: because the weights are nonnegative and shared, any part of \(G^{\mathcal L}\) that survives as \(t\to\infty\) comes from modes with \(\lambda_{\omega,a}\to0\), which contribute identically to \(G^{\mathcal H}\). The paper turns this into a statement about thermodynamic limits by contradiction, and extends it to arbitrary stationary states by showing that, under these conditions, any open-system stationary state is also stationary for the Hamiltonian alone.
How to reuse it
When comparing a Markovian open system with its coherent part, decompose by frequency and look for an inner product in which the dissipator is self-adjoint; the open spectrum is then the closed spectrum shifted into the left half-plane, which bounds what dissipation can add.
Context & limitations
Why it matters. The two results split the question cleanly. Coupling to equilibrium can switch a time crystal on, so external resources are not logically required; but the switch is always the shift the environment induces in the system's Hamiltonian, never damping or noise. The no-go theorem also extends the Watanabe–Oshikawa absence of time crystals in ground and Gibbs states to open systems whose coupling leaves the Hamiltonian unchanged.
Limitations. The activating model is constructed for the proof rather than proposed for an experiment, as the authors state. Its stationary state is a long-range-correlated mixture of Dicke states, not a Gibbs state, and the oscillation comes from an exact dynamical symmetry, so it is fine-tuned in the sense that perturbations must commute with the collective raising operator. The microscopic environment reproduces the dissipator only in the limit of an infinite-dimensional ancilla per site. The no-go theorem covers Markovian (Lindblad) dynamics; non-Markovian environments are left open.