Calling Qobj.eigenstates() of an operator created with energy restricted functionality raises an error. I believe this is due to the dimension refactoring of v5. The dimension of the solved eigenvectors are set by this list representation from Qobj.dims instead of the underlying Dimension/Space instances. Specifically here, the list is used instead of the EnrSpace instance, and thus information about the excitation limit is lost, raising a dimension/shape mismatch error in the Qobj constructor.
The simplest solution, inspired by another ENR issue #2387, would be to use _dims instead of dims in lines 1509-1512 (master at the time of writing) of qobj.py:
if self.type == 'super':
new_dims = [self._dims[0], [1]]
else:
new_dims = [self._dims[0], [1]*len(self.dims[0])]
I quickly checked that the superoperators work as well, but I am not sure how general the approach is.
Code to Reproduce the Bug# Hamiltonian of two uncoupled qubits (harmonic oscillators limited to 2 levels) with energies e1 and e2, and only one excitation allowed in the system. import qutip as qt dims = [2, 2] excitations = 1 e1, e2 = 1, 2 a1, a2 = qt.enr_destroy(dims, excitations) H = e1*a1.dag()*a1 + e2*a2.dag()*a2 H.eigenstates()Code Output
ValueError Traceback (most recent call last)
Cell In[3], line 6
4 a1, a2 = qt.enr_destroy(dims, excitations)
5 H = e1*a1.dag()*a1 + e2*a2.dag()*a2
----> 6 H.eigenstates()
File ~\anaconda3\envs\bare_qt5\Lib\site-packages\qutip\core\qobj.py:1514, in Qobj.eigenstates(self, sparse, sort, eigvals, tol, maxiter, phase_fix)
1512 new_dims = [self.dims[0], [1]*len(self.dims[0])]
1513 ekets = np.empty((evecs.shape[1],), dtype=object)
-> 1514 ekets[:] = [Qobj(vec, dims=new_dims, copy=False)
1515 for vec in _data.split_columns(evecs, False)]
1516 norms = np.array([ket.norm() for ket in ekets])
1517 if phase_fix is None:
File ~\anaconda3\envs\bare_qt5\Lib\site-packages\qutip\core\qobj.py:279, in Qobj.__init__(self, arg, dims, copy, superrep, isherm, isunitary)
277 self._isherm = isherm
278 self._isunitary = isunitary
--> 279 self._initialize_data(arg, dims, copy)
281 if superrep is not None:
282 self.superrep = superrep
File ~\anaconda3\envs\bare_qt5\Lib\site-packages\qutip\core\qobj.py:265, in Qobj._initialize_data(self, arg, dims, copy)
261 self._dims = Dimensions(
262 dims or [[self._data.shape[0]], [self._data.shape[1]]]
263 )
264 if self._dims.shape != self._data.shape:
--> 265 raise ValueError('Provided dimensions do not match the data: ' +
266 f"{self._dims.shape} vs {self._data.shape}")
ValueError: Provided dimensions do not match the data: (np.int64(4), np.int64(1)) vs (3, 1) Expected Behaviour
I expect v5 to work like v4.7.5, where the same code gives the following output:
(array([0., 1., 2.]),
array([Quantum object: dims = [[2, 2], [1, 1]], shape = (3, 1), type = ket
Qobj data =
[[1.]
[0.]
[0.]] ,
Quantum object: dims = [[2, 2], [1, 1]], shape = (3, 1), type = ket
Qobj data =
[[0.]
[0.]
[1.]] ,
Quantum object: dims = [[2, 2], [1, 1]], shape = (3, 1), type = ket
Qobj data =
[[0.]
[1.]
[0.]] ],
dtype=object))
QuTiP Version: 5.1.0 Numpy Version: 2.2.1 Scipy Version: 1.15.0 Cython Version: None Matplotlib Version: 3.10.0 Python Version: 3.12.8 Number of CPUs: 8 BLAS Info: Generic INTEL MKL Ext: None Platform Info: Windows (AMD64)Additional Context
Contributing to the Qutip community is new territory for me, so please bare with me as learn the ropes. I'm happy to create a PR with additional testing for this issue, if the suggested solution seems reasonable.
RetroSearch is an open source project built by @garambo | Open a GitHub Issue
Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo
HTML:
3.2
| Encoding:
UTF-8
| Version:
0.7.4