FermiWord and FermiSentence to_mat() method optionally return sparse …

…matrices (#6173)

This PR adds an optional flag to the `FermiWord` and `FermiSentence`
methods `to_mat` which when set will return a sparse matrix. Both are
used in the following way: `mat = fw.to_mat(format="csr")`. This PR
addresses [sc-72169].

---------

Co-authored-by: soranjh <40344468+soranjh@users.noreply.github.com>

doc/releases/changelog-dev.md

@@ -38,6 +38,9 @@
   unique representation of the object.
   [(#6167)](https://github.com/PennyLaneAI/pennylane/pull/6167)
 
+* The `to_mat` methods for `FermiWord` and `FermiSentence` now optionally return
+  a sparse matrix.
+  [(#6173)](https://github.com/PennyLaneAI/pennylane/pull/6173)
 
 <h3>Breaking changes 💔</h3>
 

pennylane/fermi/fermionic.py

@@ -273,12 +273,16 @@ def __pow__(self, value):
 
         return operator
 
-    def to_mat(self, n_orbitals=None):
+    def to_mat(self, n_orbitals=None, format="dense", buffer_size=None):
         r"""Return the matrix representation.
 
         Args:
             n_orbitals (int or None): Number of orbitals. If not provided, it will be inferred from
                 the largest orbital index in the Fermi operator.
+            format (str): The format of the matrix. It is "dense" by default. Use "csr" for sparse.
+            buffer_size (int or None)`: The maximum allowed memory in bytes to store intermediate results
+                in the calculation of sparse matrices. It defaults to ``2 ** 30`` bytes that make
+                1GB of memory. In general, larger buffers allow faster computations.
 
         Returns:
             NumpyArray: Matrix representation of the :class:`~.FermiWord`.
@@ -299,9 +303,10 @@ def to_mat(self, n_orbitals=None):
             )
 
         largest_order = n_orbitals or largest_orb_id
-        mat = qml.jordan_wigner(self, ps=True).to_mat(wire_order=list(range(largest_order)))
 
-        return mat
+        return qml.jordan_wigner(self, ps=True).to_mat(
+            wire_order=list(range(largest_order)), format=format, buffer_size=buffer_size
+        )
 
 
 # pylint: disable=useless-super-delegation
@@ -493,12 +498,16 @@ def simplify(self, tol=1e-8):
             if abs(coeff) <= tol:
                 del self[fw]
 
-    def to_mat(self, n_orbitals=None):
+    def to_mat(self, n_orbitals=None, format="dense", buffer_size=None):
         r"""Return the matrix representation.
 
         Args:
             n_orbitals (int or None): Number of orbitals. If not provided, it will be inferred from
                 the largest orbital index in the Fermi operator
+            format (str): The format of the matrix. It is "dense" by default. Use "csr" for sparse.
+            buffer_size (int or None)`: The maximum allowed memory in bytes to store intermediate results
+                in the calculation of sparse matrices. It defaults to ``2 ** 30`` bytes that make
+                1GB of memory. In general, larger buffers allow faster computations.
 
         Returns:
             NumpyArray: Matrix representation of the :class:`~.FermiSentence`.
@@ -519,9 +528,10 @@ def to_mat(self, n_orbitals=None):
             )
 
         largest_order = n_orbitals or largest_orb_id
-        mat = qml.jordan_wigner(self, ps=True).to_mat(wire_order=list(range(largest_order)))
 
-        return mat
+        return qml.jordan_wigner(self, ps=True).to_mat(
+            wire_order=list(range(largest_order)), format=format, buffer_size=buffer_size
+        )
 
 
 def from_string(fermi_string):

tests/fermi/test_fermionic.py

@@ -17,6 +17,7 @@
 
 import numpy as np
 import pytest
+from scipy import sparse
 
 import pennylane as qml
 from pennylane import numpy as pnp
@@ -133,6 +134,11 @@ def test_to_mat(self):
 
         mat = fw1.to_mat()
         assert np.allclose(mat, expected_mat)
+        assert isinstance(mat, np.ndarray)
+
+        mat = fw1.to_mat(format="csr")
+        assert np.allclose(mat.toarray(), expected_mat)
+        assert isinstance(mat, sparse.csr_matrix)
 
     def test_to_mat_error(self):
         """Test that an error is raised if the requested matrix dimension is smaller than the
@@ -636,6 +642,9 @@ def test_to_mat(self):
         mat = fs7.to_mat()
         assert np.allclose(mat, expected_mat)
 
+        mat = fs7.to_mat(format="csr")
+        assert np.allclose(mat.toarray(), expected_mat)
+
     def test_to_mat_error(self):
         """Test that an error is raised if the requested matrix dimension is smaller than the
         dimension inferred from the largest orbital index.