mu reassignment breaks shape handling in HamHeisenberg when using connectivity

[Manav939]

🐛 Bug: mu reassignment breaks shape handling in HamHeisenberg when using connectivity

Description

When using the HamHeisenberg class with a connectivity matrix and scalar parameters (J_eq, J_ax, mu), the initialization logic appears to incorrectly overwrite self.mu, leading to shape inconsistencies and runtime errors during one-body integral generation. It seems to me that it was added for the second method for defining the Hamiltonian by providing exchange couplings, $J^{ax}$ $J^{eq}$ , and $\mu$ as numpy arrays. as mentioned in the tutorials.

Relevant Code

The issue seems to originate from:

if connectivity is not None:
    self.connectivity = connectivity
    self.n_sites = connectivity.shape[0]
    if isinstance(J_eq, (int, float)):
        self.J_eq = J_eq * connectivity
        self.J_ax = J_ax * connectivity
        self.mu = mu * np.ones(self.n_sites)
    else:
        raise TypeError("Connectivity matrix is provided, "
                        "J_eq, J_ax, and mu should be floats")

However, later in the code, self.mu is reassigned:

self.mu = np.array(mu)

This overrides the earlier initialization and can collapse self.mu into a scalar array when mu = 0.

---

Steps to Reproduce

1. Provide a valid connectivity matrix.
2. Use scalar values for J_eq, J_ax, and mu (e.g., mu = 0).
3. Call:

h1 = ham.generate_one_body_integral(dense=True, basis='spinorbital basis')

---

Observed Behavior

A TypeError is raised:

TypeError: mu array has wrong basis

This occurs due to:

if self.mu.shape != (2 * self.n_sites,) and \
   self.mu.shape == (self.n_sites,):
    mu = np.hstack([self.mu, self.mu])
else:
    raise TypeError("mu array has wrong basis")

Because self.mu becomes a scalar array after reassignment, it no longer satisfies the expected shape conditions.

---

Expected Behavior

• When connectivity is provided and mu is scalar, self.mu should remain an array of shape (n_sites,).
• It should not be overwritten later in initialization.
• generate_one_body_integral should run without errors.

---

Proposed Fix

Avoid overwriting self.mu after it has been initialized based on connectivity.

if connectivity is not None:
    self.connectivity = connectivity
    self.n_sites = connectivity.shape[0]
    if isinstance(J_eq, (int, float)):
        self.J_eq = J_eq * connectivity
        self.J_ax = J_ax * connectivity
        self.mu = mu * np.ones(self.n_sites)
    else:
        raise TypeError("Connectivity matrix is provided, "
                        "J_eq, J_ax, and mu should be floats")
else:
    if isinstance(J_eq, np.ndarray) and \
       isinstance(J_ax, np.ndarray) and \
       isinstance(mu, np.ndarray) and \
       J_eq.shape == J_ax.shape and \
       mu.shape[0] == J_eq.shape[0]:
        self.n_sites = J_eq.shape[0]
        self.J_eq = J_eq
        self.J_ax = J_ax
        self.mu = mu
    else:
        raise TypeError("J_eq and J_ax should be numpy arrays of the same shape")

---

Can anyone please help me with this issue?

[PaulWAyers]

Your thoughts @RichRick1 ?