Skip to content

Fixing / expanding some Trotter options #86

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
BrendanKKrueger merged 3 commits into from
Jul 29, 2026
Merged

Fixing / expanding some Trotter options #86

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
b56b07a
Starting to fix some Trotter issues
bkkrueger Jul 28, 2026
57baaac
Allow the use of any available Trotter method if step count is provid…
bkkrueger Jul 28, 2026
f9dd771
Additional Trotter methods, additional names for Trotter methods, add…
bkkrueger Jul 29, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 7 additions & 0 deletions analysis/config_types.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -131,7 +131,13 @@ def encode_ramped_trotter(self, **kwargs):
self.trotter_combine_terms = kwargs.get("trotter_combine_terms", True)
self.ordering_method = kwargs.get("ordering_method", None)
self.trotter_order = kwargs.get("trotter_order", None)
self.trotter_steps = kwargs.get("trotter_steps", None)
self.phase_scale_factor = kwargs.get("phase_scale_factor", 1.01)
# Validate trotter_steps
if self.trotter_steps is not None:
if not isinstance(self.trotter_steps, int) or self.trotter_steps < 1:
raise ValueError(
f"trotter_steps must be a positive integer, got {self.trotter_steps}")
# Validate phase_scale_factor
if self.phase_scale_factor <= 0:
raise ValueError(
Expand Down Expand Up @@ -159,6 +165,7 @@ def _generate_TOML_table(self):
self.save_if_present(table, "trotter_combine_terms")
self.save_if_present(table, "ordering_method")
self.save_if_present(table, "trotter_order")
self.save_if_present(table, "trotter_steps")
self.save_if_present(table, "phase_scale_factor")
return table

Expand Down
192 changes: 108 additions & 84 deletions analysis/unitary.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -205,95 +205,119 @@ def encode_ramped_trotter(
assert tevol_hbar >= 0.0
logger.info(f"-- using timestep = {tevol_hbar}")

# Compute the number of Trotter steps based on the Trotter error from arXiv:1912.08854v3
# Using fast implementation for better performance (100-150x more samples per second)
from qhat.analysis.trotter_coefficients_fast import trotter_error_estimator_fast

# Read error coefficient computation mode from config (with backward-compatible defaults)
error_coeff_mode = getattr(config_unitary, 'error_coeff_mode', 'monte_carlo')
error_coeff_auto_exact = getattr(config_unitary, 'error_coeff_auto_exact', False)
error_coeff_time_limit = getattr(config_unitary, 'error_coeff_time_limit', 60)

# Validate user configuration
if error_coeff_mode not in ['monte_carlo', 'exact']:
logger.warning(f"Invalid error_coeff_mode='{error_coeff_mode}', using 'monte_carlo'")
error_coeff_mode = 'monte_carlo'

logger.info(f"-- Error coefficient computation mode: {error_coeff_mode}")
if error_coeff_mode == 'monte_carlo' and error_coeff_auto_exact:
logger.info(f"-- Auto-switch to exact coefficient computation: enabled")
logger.info(f"-- Error coefficient time limit: {error_coeff_time_limit}s")

# Compute error coefficients with user-specified mode
c1, c2 = trotter_error_estimator_fast(
hamiltonian.get_grouped_terms(),
error_coeff_time_limit,
mode=error_coeff_mode,
auto_exact=error_coeff_auto_exact
)
logger.info(f"-- Trotter error coefficients: C1 = {c1}, C2 = {c2}")
assert config_unitary.energy_error is not None
logger.info(f"-- allowable energy error = {config_unitary.energy_error} Hartrees")
logger.info(f"-- effective energy range = {2 * math.pi / tevol_hbar} Hartrees")
eps_trotter = config_unitary.energy_error * tevol_hbar / (2 * math.pi)
logger.info(f"-- allowable fractional Trotter error = {eps_trotter}")
# use config_unitary.energy_error here instead of eps_trotter, because these equations are
# derived in terms of the absolute energy error rather than the fractional energy error
s1 = tevol_hbar * config_unitary.error_scale * c1 / config_unitary.energy_error
s2 = tevol_hbar * math.sqrt(config_unitary.error_scale * c2 / config_unitary.energy_error)
logger.info(f"-- Trotter step count: s1 = {s1}, s2 = {s2}")

# Cost factors: r1=1 (single ramp), r2=2 (forward+backward)
r1 = 1
r2 = 2

# Check if user explicitly requested a specific Trotter order
# Check if user provided manual step count
user_provided_steps = (hasattr(config_unitary, 'trotter_steps') and
config_unitary.trotter_steps is not None)

# Get the requested Trotter order/method
requested_order = config_unitary.trotter_order

if requested_order is not None:
# User explicitly requested a specific order
if requested_order.lower() == "first order":
method = "first order"
Nsteps0 = s1
ramps_per_step = 1
elif requested_order.lower() == "second order":
if user_provided_steps:
# User provided step count manually - allow any method from the implementation
Nsteps = config_unitary.trotter_steps
logger.info(f"-- User provided number of Trotter steps: {Nsteps}")

if requested_order is None:
raise ValueError(
"When providing trotter_steps manually, you must also specify trotter_order."
)

# When steps are provided, we use the requested method directly
method = requested_order.lower()
logger.info(f"-- Using user-specified Trotter method: {method}")

else:
# Auto-calculate optimal step count - only certain methods support this
# Compute the number of Trotter steps based on the Trotter error from arXiv:1912.08854v3
# Using fast implementation for better performance (100-150x more samples per second)
from qhat.analysis.trotter_coefficients_fast import trotter_error_estimator_fast

# Read error coefficient computation mode from config (with backward-compatible defaults)
error_coeff_mode = getattr(config_unitary, 'error_coeff_mode', 'monte_carlo')
error_coeff_auto_exact = getattr(config_unitary, 'error_coeff_auto_exact', False)
error_coeff_time_limit = getattr(config_unitary, 'error_coeff_time_limit', 60)

# Validate user configuration
if error_coeff_mode not in ['monte_carlo', 'exact']:
logger.warning(f"Invalid error_coeff_mode='{error_coeff_mode}', using 'monte_carlo'")
error_coeff_mode = 'monte_carlo'

logger.info(f"-- Error coefficient computation mode: {error_coeff_mode}")
if error_coeff_mode == 'monte_carlo' and error_coeff_auto_exact:
logger.info(f"-- Auto-switch to exact coefficient computation: enabled")
logger.info(f"-- Error coefficient time limit: {error_coeff_time_limit}s")

# Compute error coefficients with user-specified mode
c1, c2 = trotter_error_estimator_fast(
hamiltonian.get_grouped_terms(),
error_coeff_time_limit,
mode=error_coeff_mode,
auto_exact=error_coeff_auto_exact
)
logger.info(f"-- Trotter error coefficients: C1 = {c1}, C2 = {c2}")
assert config_unitary.energy_error is not None
logger.info(f"-- allowable energy error = {config_unitary.energy_error} Hartrees")
logger.info(f"-- effective energy range = {2 * math.pi / tevol_hbar} Hartrees")
eps_trotter = config_unitary.energy_error * tevol_hbar / (2 * math.pi)
logger.info(f"-- allowable fractional Trotter error = {eps_trotter}")
# use config_unitary.energy_error here instead of eps_trotter, because these equations are
# derived in terms of the absolute energy error rather than the fractional energy error
s1 = tevol_hbar * config_unitary.error_scale * c1 / config_unitary.energy_error
s2 = tevol_hbar * math.sqrt(config_unitary.error_scale * c2 / config_unitary.energy_error)
logger.info(f"-- Trotter step count: s1 = {s1}, s2 = {s2}")

# Cost factors: r1=1 (single ramp), r2=2 (forward+backward)
r1 = 1
r2 = 2

if requested_order is not None:
# User explicitly requested a specific order for auto-calculation
if requested_order.lower() == "first order":
method = "first order"
Nsteps0 = s1
ramps_per_step = 1
elif requested_order.lower() == "second order":
method = "second order"
Nsteps0 = s2
ramps_per_step = 2
elif requested_order.lower() == "fourth order":
# Fourth-order step count using C2-based heuristic
# NOTE: This is an approximation provided by Claude (not human verified) since we don't
# compute the exact C4 coefficient
x = config_unitary.error_scale * c2 / config_unitary.energy_error
s4 = tevol_hbar * x ** (1.0/3.0)
logger.warning(
"Fourth-order step count (s4) uses an APPROXIMATION based on C2, not an exact C4 "
"coefficient. Unlike first-order (C1) and second-order (C2), this is a heuristic "
"estimate and may be less accurate."
)
logger.info(f"-- Fourth-order step count: s4 = {s4}")
method = "fourth order"
Nsteps0 = s4
ramps_per_step = 10
else:
raise ValueError(
f"Invalid trotter_order '{requested_order}' for automatic step calculation. "
f"Only 'first order', 'second order', and 'fourth order' have step auto-calculation. "
f"To use other methods (third, eighth order, etc.), provide trotter_steps manually."
)
logger.info(f"-- User requested {method}")
else:
# Auto-select between first and second order based on cost
# -- Fourth order is excluded because the step estimate is not as reliable. It should be
# added here once we have a better calculation of C4 and therefore s4.
method = "second order"
Nsteps0 = s2
ramps_per_step = 2
elif requested_order.lower() == "fourth order":
# Fourth-order step count using C2-based heuristic
# NOTE: This is an approximation provided by Claude (not human verified) since we don't
# compute the exact C4 coefficient
x = config_unitary.error_scale * c2 / config_unitary.energy_error
s4 = tevol_hbar * x ** (1.0/3.0)
logger.warning(
"Fourth-order step count (s4) uses an APPROXIMATION based on C2, not an exact C4 "
"coefficient. Unlike first-order (C1) and second-order (C2), this is a heuristic "
"estimate and may be less accurate."
)
logger.info(f"-- Fourth-order step count: s4 = {s4}")
method = "fourth order"
Nsteps0 = s4
ramps_per_step = 10
else:
raise ValueError(f"Invalid trotter_order '{requested_order}'. Must be 'first order', 'second order', or 'fourth order'.")
logger.info(f"-- User requested {method}")
else:
# Auto-select between first and second order based on cost
# -- Fourth order is excluded because the step estimate is not as reliable. It should be
# added here once we have a better calculation of C4 and therefore s4.
method = "second order"
Nsteps0 = s2
ramps_per_step = 2
if r1 * s1 < r2 * s2:
method = "first order"
Nsteps0 = s1
ramps_per_step = 1
logger.verbose("-- auto selecting order based on unoptimized ramp count")

Nsteps = max(1, math.ceil(Nsteps0))
logger.info(f"-- using {method} Trotter formula with {Nsteps} steps ({Nsteps0})")
logger.verbose(f"-- unoptimized ramp count = {Nsteps * ramps_per_step}")
if r1 * s1 < r2 * s2:
method = "first order"
Nsteps0 = s1
ramps_per_step = 1
logger.verbose("-- auto selecting order based on unoptimized ramp count")

Nsteps = max(1, math.ceil(Nsteps0))
logger.info(f"-- Trotter recommendation: use {method} formula with {Nsteps} steps ({Nsteps0})")
logger.verbose(f"-- unoptimized ramp count = {Nsteps * ramps_per_step}")

pauli_strings = hamiltonian.get_all_pauli_strings(return_as='strings')
pauli_strings = reorder_paulis(pauli_strings, config_unitary.ordering_method)
Expand Down
Loading