Fix AttributeError when Hessian has no negative eigenvalues

[WHOIM1205]

Fix: Prevent AttributeError in Bonded.__init__ when Hessian has no negative eigenvalues

Problem

self.hess_new was only initialized inside the if np.any(evals < tolerance) branch.
For well-optimized molecules (the normal case), this condition is false and self.hess_new was never created.
However, it was later used unconditionally for unit conversion, causing:

This prevented the Bonded class from being instantiated for typical equilibrium geometries.

Solution

Add an else branch to initialize self.hess_new with a copy of the original Hessian when no eigenvalue correction is required.

Change

if np.any(evals < tolerance):
    ...
    self.hess_new = evecs @ evals_m @ evecs.T
else:
    self.hess_new = self.hess.copy()

self.hess_new = (self.hess_new * 627.509391) / (0.529 ** 2)
self.hess = (self.hess * 627.509391) / (0.529 ** 2)

[WHOIM1205]

hey @ccastilloo
This PR fixes a crash in Bonded.__init__ where self.hess_new was only initialized inside the negative-eigenvalue correction branch. For well-optimized molecules (the common case), this branch is skipped and self.hess_new remained undefined, causing an AttributeError during unit conversion.

The fix adds a fallback initialization (self.hess_new = self.hess.copy()) when no correction is needed, ensuring the attribute is always defined while preserving the existing behavior for negative eigenvalues.

[Copilot]

Pull request overview

This PR updates Bonded.__init__ to ensure self.hess_new is always initialized, avoiding an AttributeError when the Hessian has no sufficiently negative eigenvalues (the common equilibrium case).

Changes:

• Add an else branch to set self.hess_new = self.hess.copy() when no eigenvalue correction is needed.
• Keep the unit-conversion logic operating on both self.hess and self.hess_new after initialization.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[ccastilloo]

@WHOIM1205 Thank you for your contribution! The bond.py module is still a work in progress, so there are several known issues and improvements that are planned.
I don't see any conflicts between your proposed changes and the current implementation, so I'm happy to merge this PR. Thanks again for taking the time to contribute and help improve the project