def test_bravyi_kitaev_transform(self):
# Check that the QubitOperators are two-term.
lowering = bravyi_kitaev(FermionOperator(((3, 0), )))
raising = bravyi_kitaev(FermionOperator(((3, 1), )))
self.assertEqual(len(raising.terms), 2)
self.assertEqual(len(lowering.terms), 2)
# Test the locality invariant for N=2^d qubits
# (c_j majorana is always log2N+1 local on qubits)
n_qubits = 16
invariant = numpy.log2(n_qubits) + 1
for index in range(n_qubits):
operator = bravyi_kitaev(FermionOperator(((index, 0), )), n_qubits)
qubit_terms = operator.terms.items() # Get the majorana terms.
for item in qubit_terms:
coeff = item[1]
# Identify the c majorana terms by real
# coefficients and check their length.
if not isinstance(coeff, complex):
self.assertEqual(len(item[0]), invariant)
# Hardcoded coefficient test on 16 qubits
lowering = bravyi_kitaev(FermionOperator(((9, 0), )), n_qubits)
raising = bravyi_kitaev(FermionOperator(((9, 1), )), n_qubits)
correct_operators_c = ((7, 'Z'), (8, 'Z'), (9, 'X'), (11, 'X'), (15,
'X'))
correct_operators_d = ((7, 'Z'), (9, 'Y'), (11, 'X'), (15, 'X'))
self.assertEqual(lowering.terms[correct_operators_c], 0.5)
self.assertEqual(lowering.terms[correct_operators_d], 0.5j)
self.assertEqual(raising.terms[correct_operators_d], -0.5j)
self.assertEqual(raising.terms[correct_operators_c], 0.5)
def test_bk_jw_majoranas(self):
# Check if the Majorana operators have the same spectrum
# irrespectively of the transform.
n_qubits = 7
a = FermionOperator(((1, 0), ))
a_dag = FermionOperator(((1, 1), ))
c = a + a_dag
d = 1j * (a_dag - a)
c_spins = [jordan_wigner(c), bravyi_kitaev(c)]
d_spins = [jordan_wigner(d), bravyi_kitaev(d)]
c_sparse = [
get_sparse_operator(c_spins[0]),
get_sparse_operator(c_spins[1])
]
d_sparse = [
get_sparse_operator(d_spins[0]),
get_sparse_operator(d_spins[1])
]
c_spectrum = [eigenspectrum(c_spins[0]), eigenspectrum(c_spins[1])]
d_spectrum = [eigenspectrum(d_spins[0]), eigenspectrum(d_spins[1])]
self.assertAlmostEqual(
0., numpy.amax(numpy.absolute(d_spectrum[0] - d_spectrum[1])))
def test_bk_jw_number_operator(self):
# Check if number operator has the same spectrum in both
# BK and JW representations
n = number_operator(1, 0)
jw_n = jordan_wigner(n)
bk_n = bravyi_kitaev(n)
# Diagonalize and make sure the spectra are the same.
jw_spectrum = eigenspectrum(jw_n)
bk_spectrum = eigenspectrum(bk_n)
self.assertAlmostEqual(
0., numpy.amax(numpy.absolute(jw_spectrum - bk_spectrum)))
def test_bk_jw_hopping_operator(self):
# Check if the spectrum fits for a single hoppping operator
n_qubits = 5
ho = FermionOperator(((1, 1), (4, 0))) + FermionOperator(
((4, 1), (1, 0)))
jw_ho = jordan_wigner(ho)
bk_ho = bravyi_kitaev(ho)
# Diagonalize and make sure the spectra are the same.
jw_spectrum = eigenspectrum(jw_ho)
bk_spectrum = eigenspectrum(bk_ho)
self.assertAlmostEqual(
0., numpy.amax(numpy.absolute(jw_spectrum - bk_spectrum)))
def test_bk_jw_number_operator_scaled(self):
# Check if number operator has the same spectrum in both
# JW and BK representations
n_qubits = 1
n = number_operator(n_qubits, 0, coefficient=2) # eigenspectrum (0,2)
jw_n = jordan_wigner(n)
bk_n = bravyi_kitaev(n)
# Diagonalize and make sure the spectra are the same.
jw_spectrum = eigenspectrum(jw_n)
bk_spectrum = eigenspectrum(bk_n)
self.assertAlmostEqual(
0., numpy.amax(numpy.absolute(jw_spectrum - bk_spectrum)))
def test_bk_jw_integration(self):
# This is a legacy test, which was a minimal failing example when
# optimization for hermitian operators was used.
n_qubits = 4
# Minimal failing example:
fo = FermionOperator(((3, 1), ))
jw = jordan_wigner(fo)
bk = bravyi_kitaev(fo)
jw_spectrum = eigenspectrum(jw)
bk_spectrum = eigenspectrum(bk)
self.assertAlmostEqual(
0., numpy.amax(numpy.absolute(jw_spectrum - bk_spectrum)))
def test_bk_jw_number_operators(self):
# Check if a number operator has the same spectrum in both
# JW and BK representations
n_qubits = 2
n1 = number_operator(n_qubits, 0)
n2 = number_operator(n_qubits, 1)
n = n1 + n2
jw_n = jordan_wigner(n)
bk_n = bravyi_kitaev(n)
# Diagonalize and make sure the spectra are the same.
jw_spectrum = eigenspectrum(jw_n)
bk_spectrum = eigenspectrum(bk_n)
self.assertAlmostEqual(
0., numpy.amax(numpy.absolute(jw_spectrum - bk_spectrum)))
def test_bk_jw_integration_original(self):
# This is a legacy test, which was an example proposed by Ryan,
# failing when optimization for hermitian operators was used.
n_qubits = 5
fermion_operator = FermionOperator(((3, 1), (2, 1), (1, 0), (0, 0)),
-4.3)
fermion_operator += FermionOperator(((3, 1), (1, 0)), 8.17)
fermion_operator += 3.2 * FermionOperator()
# Map to qubits and compare matrix versions.
jw_qubit_operator = jordan_wigner(fermion_operator)
bk_qubit_operator = bravyi_kitaev(fermion_operator)
# Diagonalize and make sure the spectra are the same.
jw_spectrum = eigenspectrum(jw_qubit_operator)
bk_spectrum = eigenspectrum(bk_qubit_operator)
self.assertAlmostEqual(0., numpy.amax(numpy.absolute(jw_spectrum -
bk_spectrum)))
def test_bk_n_qubits_too_small(self):
with self.assertRaises(ValueError):
bravyi_kitaev(FermionOperator('2^ 3^ 5 0'), n_qubits=4)
def test_bk_identity(self):
self.assertTrue(
bravyi_kitaev(FermionOperator(())).isclose(QubitOperator(())))
