---

name: "specialized-open-systems" description: "QuTiP PIQS, HEOM, environment, Bloch-Redfield, and non-Markovian workflows." metadata: disco-role: operating disable-model-invocation: true license: BSD 3-Clause

Specialized open systems

Use this subskill when the task is about QuTiP's higher-level open-system model families, especially bath/environment construction, PIQS, HEOM, Bloch-Redfield, or non-Markovian transfer-tensor workflows.

Read this subskill when the prompt mentions

• PIQS, Dicke, permutational invariance, collective emission, collective pumping, or Dicke basis states
• HEOM, hierarchical equations of motion, Drude-Lorentz baths, Lorentzian baths, bath exponents, or hierarchy depth
• BosonicEnvironment, FermionicEnvironment, DrudeLorentzEnvironment, OhmicEnvironment, spectral density, correlation function, or power spectrum
• Bloch-Redfield, brmesolve, bath coupling operators, or non-Markovian transfer-tensor methods

What to decide first

1. Is the user asking for a bath/environment object, a solver, or a symmetry-reduced state representation?
2. Does the workflow stay in the Dicke basis or require the full Hilbert space?
3. Is the requested bath model bosonic, fermionic, Drude-Lorentz, Lorentzian, Ohmic, under-damped, or custom exponent based?
4. Is a small CPU example enough, or is the task asking for large-scale physics that needs a performance warning?

Core workflow

• Use qutip.core.environment for spectral-density, correlation-function, and power-spectrum objects.
• Use qutip.solver.heom for HEOM baths and HEOMSolver when the hierarchy is part of the model.
• Use qutip.piqs.piqs for Dicke-basis states, collective spin operators, and the Dicke system wrapper.
• Return to dynamics-and-solvers when the task is only a generic mesolve, steadystate, or correlation calculation.

Typical success signals

• Environment functions accept both scalars and arrays.
• Dicke-basis dimensions match the expected num_dicke_states(N) rather than the full 2**N space.
• HEOM bath exponent counts and hierarchy depth are explicit before launching a solve.
• Optional acceleration is treated as a performance feature, not a correctness requirement.

Boundaries

Use this subskill for advanced open-system model construction. Do not use it as the main route for:

• Generic Qobj algebra before the bath or symmetry model is defined; start in core-objects.
• Plain mesolve, sesolve, or steadystate questions with no specialized bath or PIQS/HEOM term; use dynamics-and-solvers.
• Plotting or serializing final states; use analysis-and-io after the model result exists.

Answer shape

When responding from this subskill, give:

1. The model family: environment, HEOM, PIQS, Bloch-Redfield, or transfer tensor.
2. The required parameters and their physical meaning.
3. A tiny construction or validation snippet before any expensive solve.
4. The dimensionality or exponent-count check that bounds the cost.
5. Any optional acceleration or unavailable-backend note.

Validation hints

• For PIQS, compare dimensions against num_dicke_states(N).
• For environments, evaluate spectral density or correlation functions on both scalars and small arrays.
• For HEOM, print bath exponent counts and use small Nk / hierarchy depth in examples.

Reference files

• references/api-reference.md for environment, PIQS, HEOM, and non-Markovian entry points.
• references/workflows.md for compact environment and PIQS examples.
• references/troubleshooting.md for basis, bath, and hierarchy pitfalls.

Helper script

• scripts/open_systems_smoke.py constructs a Drude-Lorentz environment, a PIQS Dicke-basis object, and a small HEOM bath without launching an expensive solve.