Pular para o conteúdo principal

Retropropagação de operador (OBP) para estimativa de valores esperados

Estimativa de uso: 16 minutos em um processador Eagle r3 (NOTA: Esta é apenas uma estimativa. Seu tempo de execução pode variar.)

# Added by doQumentation — required packages for this notebook
!pip install -q matplotlib numpy qiskit qiskit-addon-obp qiskit-addon-utils qiskit-ibm-runtime rustworkx
# This cell is hidden from users;
# it disables linting rules.
# ruff: noqa

Contexto​

A retropropagação de operador é uma técnica que envolve absorver operações do final de um circuito quântico no observável medido, geralmente reduzindo a profundidade do circuito ao custo de termos adicionais no observável. O objetivo é retropropagar o máximo possível do circuito sem permitir que o observável cresça demais. Uma implementação baseada em Qiskit está disponível no addon OBP do Qiskit, mais detalhes podem ser encontrados na documentação correspondente com um exemplo simples para começar.

Considere um circuito de exemplo para o qual um observável O=∑PcPPO = \sum_P c_P P deve ser medido, onde PP são Paulis e cPc_P são coeficientes. Vamos denotar o circuito como um único unitário UU que pode ser logicamente particionado em U=UCUQU = U_C U_Q como mostrado na figura abaixo.

Circuit diagram showing Uq followed by Uc

A retropropagação de operador absorve o unitário UCU_C no observável evoluindo-o como O′=UC†OUC=∑PcPUC†PUCO' = U_C^{\dagger}OU_C = \sum_P c_P U_C^{\dagger}PU_C. Em outras palavras, parte da computação é realizada classicamente através da evolução do observável de OO para O′O'. O problema original agora pode ser reformulado como medir o observável O′O' para o novo circuito de menor profundidade cujo unitário é UQU_Q.

O unitário UCU_C é representado como um número de fatias UC=USUS−1...U2U1U_C = U_S U_{S-1}...U_2U_1. Existem múltiplas maneiras de definir uma fatia. Por exemplo, no circuito de exemplo acima, cada camada de RzzR_{zz} e cada camada de portas RxR_x pode ser considerada como uma fatia individual. A retropropagação envolve o cálculo de O′=Πs=1S∑PcPUs†PUsO' = \Pi_{s=1}^S \sum_P c_P U_s^{\dagger} P U_s classicamente. Cada fatia UsU_s pode ser representada como Us=exp(−iθsPs2)U_s = exp(\frac{-i\theta_s P_s}{2}), onde PsP_s é um Pauli de nn-qubits e θs\theta_s é um escalar. É fácil verificar que

Us†PUs=Pif [P,Ps]=0,U_s^{\dagger} P U_s = P \qquad \text{if} ~[P,P_s] = 0, Us†PUs=cos(θs)P+isin(θs)PsPif {P,Ps}=0U_s^{\dagger} P U_s = \qquad cos(\theta_s)P + i sin(\theta_s)P_sP \qquad \text{if} ~\{P,P_s\} = 0

No exemplo acima, se {P,Ps}=0\{P,P_s\} = 0, então precisamos executar dois circuitos quânticos, em vez de um, para calcular o valor esperado. Portanto, a retropropagação pode aumentar o número de termos no observável, levando a um maior número de execuções de circuito. Uma maneira de permitir uma retropropagação mais profunda no circuito, enquanto evita que o operador cresça demais, é truncar termos com coeficientes pequenos, em vez de adicioná-los ao operador. Por exemplo, no exemplo acima, pode-se escolher truncar o termo envolvendo PsPP_sP desde que θs\theta_s seja suficientemente pequeno. Truncar termos pode resultar em menos circuitos quânticos para executar, mas fazer isso resulta em algum erro no cálculo final do valor esperado proporcional à magnitude dos coeficientes dos termos truncados.

Este tutorial implementa um padrão Qiskit para simular a dinâmica quântica de uma cadeia de spins de Heisenberg usando qiskit-addon-obp.

Requisitos​

Antes de iniciar este tutorial, certifique-se de ter o seguinte instalado:

  • Qiskit SDK v1.2 ou posterior (pip install qiskit)
  • Qiskit Runtime v0.28 ou posterior (pip install qiskit-ibm-runtime)
  • OBP Qiskit addon (pip install qiskit-addon-obp)
  • Qiskit addon utils (pip install qiskit-addon-utils)

Configuração​

import numpy as np
import matplotlib.pyplot as plt

from qiskit.primitives import StatevectorEstimator as Estimator
from qiskit.transpiler.preset_passmanagers import generate_preset_pass_manager
from qiskit.quantum_info import SparsePauliOp
from qiskit.transpiler import CouplingMap
from qiskit.synthesis import LieTrotter

from qiskit_addon_utils.problem_generators import generate_xyz_hamiltonian
from qiskit_addon_utils.problem_generators import (
    generate_time_evolution_circuit,
)
from qiskit_addon_utils.slicing import slice_by_gate_types, combine_slices
from qiskit_addon_obp.utils.simplify import OperatorBudget
from qiskit_addon_obp import backpropagate
from qiskit_addon_obp.utils.truncating import setup_budget

from rustworkx.visualization import graphviz_draw

from qiskit_ibm_runtime import QiskitRuntimeService
from qiskit_ibm_runtime import EstimatorV2, EstimatorOptions

Parte I: Cadeia de spins de Heisenberg em pequena escala​

Etapa 1: Mapear entradas clássicas para um problema quântico​

Mapear a evolução temporal de um modelo quântico de Heisenberg para um experimento quântico.​

O pacote qiskit_addon_utils fornece algumas funcionalidades reutilizáveis para vários propósitos.

Seu módulo qiskit_addon_utils.problem_generators fornece funções para gerar Hamiltonianos tipo Heisenberg em um determinado grafo de conectividade. Este grafo pode ser um rustworkx.PyGraph ou um CouplingMap, facilitando o uso em fluxos de trabalho centrados no Qiskit.

A seguir, geramos um CouplingMap de cadeia linear de 10 qubits.

num_qubits = 10
layout = [(i - 1, i) for i in range(1, num_qubits)]

# Instantiate a CouplingMap object
coupling_map = CouplingMap(layout)
graphviz_draw(coupling_map.graph, method="circo")

Output of the previous code cell

Em seguida, geramos um operador de Pauli modelando um Hamiltoniano XYZ de Heisenberg.

H^XYZ=∑(j,k)∈E(Jxσjxσkx+Jyσjyσky+Jzσjzσkz)+∑j∈V(hxσjx+hyσjy+hzσjz){\hat{\mathcal{H}}_{XYZ} = \sum_{(j,k)\in E} (J_{x} \sigma_j^{x} \sigma_{k}^{x} + J_{y} \sigma_j^{y} \sigma_{k}^{y} + J_{z} \sigma_j^{z} \sigma_{k}^{z}) + \sum_{j\in V} (h_{x} \sigma_j^{x} + h_{y} \sigma_j^{y} + h_{z} \sigma_j^{z})}

Onde G(V,E)G(V,E) é o grafo do mapa de acoplamento fornecido.

# Get a qubit operator describing the Heisenberg XYZ model
hamiltonian = generate_xyz_hamiltonian(
    coupling_map,
    coupling_constants=(np.pi / 8, np.pi / 4, np.pi / 2),
    ext_magnetic_field=(np.pi / 3, np.pi / 6, np.pi / 9),
)
print(hamiltonian)
SparsePauliOp(['IIIIIIIXXI', 'IIIIIIIYYI', 'IIIIIIIZZI', 'IIIIIXXIII', 'IIIIIYYIII', 'IIIIIZZIII', 'IIIXXIIIII', 'IIIYYIIIII', 'IIIZZIIIII', 'IXXIIIIIII', 'IYYIIIIIII', 'IZZIIIIIII', 'IIIIIIIIXX', 'IIIIIIIIYY', 'IIIIIIIIZZ', 'IIIIIIXXII', 'IIIIIIYYII', 'IIIIIIZZII', 'IIIIXXIIII', 'IIIIYYIIII', 'IIIIZZIIII', 'IIXXIIIIII', 'IIYYIIIIII', 'IIZZIIIIII', 'XXIIIIIIII', 'YYIIIIIIII', 'ZZIIIIIIII', 'IIIIIIIIIX', 'IIIIIIIIIY', 'IIIIIIIIIZ', 'IIIIIIIIXI', 'IIIIIIIIYI', 'IIIIIIIIZI', 'IIIIIIIXII', 'IIIIIIIYII', 'IIIIIIIZII', 'IIIIIIXIII', 'IIIIIIYIII', 'IIIIIIZIII', 'IIIIIXIIII', 'IIIIIYIIII', 'IIIIIZIIII', 'IIIIXIIIII', 'IIIIYIIIII', 'IIIIZIIIII', 'IIIXIIIIII', 'IIIYIIIIII', 'IIIZIIIIII', 'IIXIIIIIII', 'IIYIIIIIII', 'IIZIIIIIII', 'IXIIIIIIII', 'IYIIIIIIII', 'IZIIIIIIII', 'XIIIIIIIII', 'YIIIIIIIII', 'ZIIIIIIIII'],
              coeffs=[0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j,
 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j,
 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j,
 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j,
 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j,
 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j,
 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j, 1.04719755+0.j,
 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j,
 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j,
 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j,
 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j,
 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j,
 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j,
 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j,
 0.34906585+0.j])

A partir do operador de qubit, podemos gerar um circuito quântico que modela sua evolução temporal. Mais uma vez, o módulo qiskit_addon_utils.problem_generators vem ao resgate com uma função conveniente para fazer exatamente isso:

circuit = generate_time_evolution_circuit(
    hamiltonian,
    time=0.2,
    synthesis=LieTrotter(reps=2),
)
circuit.draw("mpl", style="iqp", scale=0.6)

Output of the previous code cell

Etapa 2: Otimizar problema para execução em hardware quântico​

Criar fatias de circuito para retropropagar​

Lembre-se, a função backpropagate retropropagará fatias inteiras de circuito por vez, então a escolha de como fatiar pode ter um impacto em quão bem a retropropagação funciona para um determinado problema. Aqui, agruparemos portas do mesmo tipo em fatias usando a função slice_by_gate_types.

Para uma discussão mais detalhada sobre fatiamento de circuito, confira este guia de instruções do pacote qiskit-addon-utils.

slices = slice_by_gate_types(circuit)
print(f"Separated the circuit into {len(slices)} slices.")
Separated the circuit into 18 slices.

Restringir o quanto o operador pode crescer durante a retropropagação​

Durante a retropropagação, o número de termos no operador geralmente se aproximará rapidamente de 4N4^N, onde NN é o número de qubits. Quando dois termos no operador não comutam qubit a qubit, precisamos de circuitos separados para obter os valores esperados correspondentes a eles. Por exemplo, se temos um observável de 2 qubits O=0.1XX+0.3IZ−0.5IXO = 0.1 XX + 0.3 IZ - 0.5 IX, então como [XX,IX]=0[XX,IX] = 0, a medição em uma única base é suficiente para calcular os valores esperados para esses dois termos. No entanto, IZIZ anticomuta com os outros dois termos. Então precisamos de uma medição de base separada para calcular o valor esperado de IZIZ. Em outras palavras, precisamos de dois, em vez de um, circuito para calcular ⟨O⟩\langle O \rangle. À medida que o número de termos no operador aumenta, há a possibilidade de que o número necessário de execuções de circuito também aumente.

O tamanho do operador pode ser limitado especificando o argumento operator_budget da função backpropagate, que aceita uma instância OperatorBudget.

Para controlar a quantidade de recursos extras (tempo) alocados, restringimos o número máximo de grupos de Pauli comutantes qubit a qubit que o observável retropropagado pode ter. Aqui especificamos que a retropropagação deve parar quando o número de grupos de Pauli comutantes qubit a qubit no operador crescer além de 8.

op_budget = OperatorBudget(max_qwc_groups=8)

Retropropagar fatias do circuito​

Primeiro especificamos o observável como MZ=1N∑i=1N⟨Zi⟩M_Z = \frac{1}{N} \sum_{i=1}^N \langle Z_i \rangle, sendo NN o número de qubits. Retropropagaremos fatias do circuito de evolução temporal até que os termos no observável não possam mais ser combinados em oito ou menos grupos de Pauli comutantes qubit a qubit.

observable = SparsePauliOp.from_sparse_list(
    [("Z", [i], 1 / num_qubits) for i in range(num_qubits)],
    num_qubits=num_qubits,
)
observable
SparsePauliOp(['IIIIIIIIIZ', 'IIIIIIIIZI', 'IIIIIIIZII', 'IIIIIIZIII', 'IIIIIZIIII', 'IIIIZIIIII', 'IIIZIIIIII', 'IIZIIIIIII', 'IZIIIIIIII', 'ZIIIIIIIII'],
              coeffs=[0.1+0.j, 0.1+0.j, 0.1+0.j, 0.1+0.j, 0.1+0.j, 0.1+0.j, 0.1+0.j, 0.1+0.j,
 0.1+0.j, 0.1+0.j])

Abaixo você verá que retropropagamos seis fatias, e os termos foram combinados em seis e não oito grupos. Isso implica que retropropagar mais uma fatia faria com que o número de grupos de Pauli excedesse oito. Podemos verificar que este é o caso inspecionando os metadados retornados. Observe também que nesta porção a transformação do circuito é exata. Isto é, nenhum termo do novo observável O′O' foi truncado. O circuito retropropagado e o operador retropropagado fornecem o resultado exato como o circuito e operador originais.

# Backpropagate slices onto the observable
bp_obs, remaining_slices, metadata = backpropagate(
    observable, slices, operator_budget=op_budget
)
# Recombine the slices remaining after backpropagation
bp_circuit = combine_slices(remaining_slices)

print(f"Backpropagated {metadata.num_backpropagated_slices} slices.")
print(
    f"New observable has {len(bp_obs.paulis)} terms, which can be combined into {len(bp_obs.group_commuting(qubit_wise=True))} groups."
)
print(
    f"Note that backpropagating one more slice would result in {metadata.backpropagation_history[-1].num_paulis[0]} terms "
    f"across {metadata.backpropagation_history[-1].num_qwc_groups} groups."
)
print("The remaining circuit after backpropagation looks as follows:")
bp_circuit.draw("mpl", fold=-1, scale=0.6)
Backpropagated 6 slices.
New observable has 60 terms, which can be combined into 6 groups.
Note that backpropagating one more slice would result in 114 terms across 12 groups.
The remaining circuit after backpropagation looks as follows:

Output of the previous code cell

Em seguida, especificaremos o mesmo problema com as mesmas restrições no tamanho do observável de saída. No entanto, desta vez, alocamos um orçamento de erro para cada fatia usando a função setup_budget. Termos de Pauli com coeficientes pequenos serão truncados de cada fatia até que o orçamento de erro seja preenchido, e o orçamento restante será adicionado ao orçamento da fatia seguinte. Observe que neste caso, a transformação devido à retropropagação é aproximada, pois alguns dos termos no operador são truncados.

Para habilitar este truncamento, precisamos configurar nosso orçamento de erro assim:

truncation_error_budget = setup_budget(max_error_per_slice=0.005)

Observe que ao alocar 5e-3 erro por fatia para truncamento, somos capazes de remover mais 1 fatia do circuito, enquanto permanecemos dentro do orçamento original de oito grupos de Pauli comutantes no observável. Por padrão, backpropagate usa a norma L1 dos coeficientes truncados para limitar o erro total incorrido do truncamento. Para outras opções, consulte o guia de instruções sobre especificação da p_norm.

Neste exemplo particular onde retropropagamos sete fatias, o erro total de truncamento não deve exceder (5e-3 erro/fatia) * (7 fatias) = 3.5e-2. Para uma discussão mais aprofundada sobre distribuição de um orçamento de erro entre suas fatias, confira este guia de instruções.

# Run the same experiment but truncate observable terms with small coefficients
bp_obs_trunc, remaining_slices_trunc, metadata = backpropagate(
    observable,
    slices,
    operator_budget=op_budget,
    truncation_error_budget=truncation_error_budget,
)

# Recombine the slices remaining after backpropagation
bp_circuit_trunc = combine_slices(
    remaining_slices_trunc, include_barriers=False
)

print(f"Backpropagated {metadata.num_backpropagated_slices} slices.")
print(
    f"New observable has {len(bp_obs_trunc.paulis)} terms, which can be combined into {len(bp_obs_trunc.group_commuting(qubit_wise=True))} groups.\n"
    f"After truncation, the error in our observable is bounded by {metadata.accumulated_error(0):.3e}"
)
print(
    f"Note that backpropagating one more slice would result in {metadata.backpropagation_history[-1].num_paulis[0]} terms "
    f"across {metadata.backpropagation_history[-1].num_qwc_groups} groups."
)
print("The remaining circuit after backpropagation looks as follows:")
bp_circuit_trunc.draw("mpl", scale=0.6)
Backpropagated 7 slices.
New observable has 82 terms, which can be combined into 8 groups.
After truncation, the error in our observable is bounded by 3.266e-02
Note that backpropagating one more slice would result in 114 terms across 12 groups.
The remaining circuit after backpropagation looks as follows:

Output of the previous code cell

Observamos que o truncamento nos permite retropropagar mais sem aumentar o número de grupos comutantes no observável.

Agora que temos nosso ansatz reduzido e observáveis expandidos, podemos transpilar nossos experimentos para o backend.​

Aqui usaremos um computador quântico IBM® de 127 qubits para demonstrar como transpilar para um backend QPU.

service = QiskitRuntimeService()
backend = service.least_busy(
    operational=True, simulator=False, min_num_qubits=127
)
pm = generate_preset_pass_manager(backend=backend, optimization_level=1)

# Transpile original experiment
circuit_isa = pm.run(circuit)
observable_isa = observable.apply_layout(circuit_isa.layout)

# Transpile backpropagated experiment
bp_circuit_isa = pm.run(bp_circuit)
bp_obs_isa = bp_obs.apply_layout(bp_circuit_isa.layout)

# Transpile the backpropagated experiment with truncated observable terms
bp_circuit_trunc_isa = pm.run(bp_circuit_trunc)
bp_obs_trunc_isa = bp_obs_trunc.apply_layout(bp_circuit_trunc_isa.layout)

Criamos o Primitive Unified Bloc (PUB) para cada um dos três casos.

pub = (circuit_isa, observable_isa)
bp_pub = (bp_circuit_isa, bp_obs_isa)
bp_trunc_pub = (bp_circuit_trunc_isa, bp_obs_trunc_isa)

Etapa 3: Executar usando primitivas Qiskit​

Calcular valor esperado​

Finalmente, podemos executar os experimentos retropropagados e compará-los com o experimento completo usando o StatevectorEstimator sem ruído.

ideal_estimator = Estimator()

# Run the experiments using Estimator primitive to obtain the exact outcome
result_exact = (
    ideal_estimator.run([(circuit, observable)]).result()[0].data.evs.item()
)
print(f"Exact expectation value: {result_exact}")
Exact expectation value: 0.8871244838989416

Usaremos resilience_level = 2 para este exemplo.

options = EstimatorOptions()
options.default_precision = 0.011
options.resilience_level = 2

estimator = EstimatorV2(mode=backend, options=options)
job = estimator.run([pub, bp_pub, bp_trunc_pub])

Passo 4: Pós-processar e retornar resultado no formato clássico desejado​

result_no_bp = job.result()[0].data.evs.item()
result_bp = job.result()[1].data.evs.item()
result_bp_trunc = job.result()[2].data.evs.item()

std_no_bp = job.result()[0].data.stds.item()
std_bp = job.result()[1].data.stds.item()
std_bp_trunc = job.result()[2].data.stds.item()
print(
    f"Expectation value without backpropagation: {result_no_bp} ± {std_no_bp}"
)
print(f"Backpropagated expectation value: {result_bp} ± {std_bp}")
print(
    f"Backpropagated expectation value with truncation: {result_bp_trunc} ± {std_bp_trunc}"
)
Expectation value without backpropagation: 0.8033194665993642
Backpropagated expectation value: 0.8599808781259016
Backpropagated expectation value with truncation: 0.8868736004169483
methods = [
    "No backpropagation",
    "Backpropagation",
    "Backpropagation w/ truncation",
]
values = [result_no_bp, result_bp, result_bp_trunc]
stds = [std_no_bp, std_bp, std_bp_trunc]

ax = plt.gca()
plt.bar(methods, values, color="#a56eff", width=0.4, edgecolor="#8a3ffc")
plt.axhline(result_exact)
ax.set_ylim([0.6, 0.92])
plt.text(0.2, 0.895, "Exact result")
ax.set_ylabel(r"$M_Z$", fontsize=12)
Text(0, 0.5, '$M_Z$')

Output of the previous code cell

Parte B: Ampliando a escala!​

Vamos agora usar a Retropropagação de Operador para estudar a dinâmica do Hamiltoniano de uma Cadeia de Spin de Heisenberg de 50 qubits.

Passo 1: Mapear entradas clássicas para um problema quântico​

Consideramos um Hamiltoniano de 50 qubits H^XYZ\hat{\mathcal{H}}_{XYZ} para o problema ampliado com os mesmos valores para os coeficientes JJ e hh do exemplo em pequena escala. O observável MZ=1N∑i=1N⟨Zi⟩M_Z = \frac{1}{N} \sum_{i=1}^N \langle Z_i \rangle também é o mesmo de antes. Este problema está além da simulação clássica por força bruta.

num_qubits = 50
layout = [(i - 1, i) for i in range(1, num_qubits)]

# Instantiate a CouplingMap object
coupling_map = CouplingMap(layout)
graphviz_draw(coupling_map.graph, method="circo")

Output of the previous code cell

hamiltonian = generate_xyz_hamiltonian(
    coupling_map,
    coupling_constants=(np.pi / 8, np.pi / 4, np.pi / 2),
    ext_magnetic_field=(np.pi / 3, np.pi / 6, np.pi / 9),
)
print(hamiltonian)
SparsePauliOp(['IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXXI', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYYI', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZZI', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXXIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYYIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZZIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXXIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYYIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZZIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXXIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYYIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZZIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXXIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYYIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZZIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXXIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYYIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZZIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXXIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYYIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZZIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXXIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYYIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZZIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXXIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYYIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZZIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIXXIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIYYIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIZZIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIXXIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIYYIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIZZIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIXXIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIYYIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIZZIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIXXIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIYYIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIZZIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIXXIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIYYIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIZZIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIXXIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIYYIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIZZIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIXXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIYYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIZZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIXXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIYYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIZZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIXXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIYYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIZZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIXXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIYYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIZZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIXXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIYYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIZZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIXXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIYYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIZZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIXXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIYYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIZZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIXXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIYYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIZZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IXXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IYYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IZZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXX', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYY', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZZ', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXXII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYYII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZZII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXXIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYYIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZZIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXXIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYYIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZZIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXXIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYYIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZZIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXXIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYYIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZZIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXXIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYYIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZZIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXXIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYYIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZZIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXXIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYYIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZZIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXXIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYYIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZZIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIXXIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIYYIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIZZIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIXXIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIYYIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIZZIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIXXIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIYYIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIZZIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIXXIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIYYIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIZZIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIXXIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIYYIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIZZIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIXXIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIYYIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIZZIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIXXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIYYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIZZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIXXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIYYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIZZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIXXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIYYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIZZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIXXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIYYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIZZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIXXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIYYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIZZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIXXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIYYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIZZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIXXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIYYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIZZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIXXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIYYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIZZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'XXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'YYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'ZZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIX', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIY', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZ', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXI', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYI', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZI', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIXIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIYIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIXIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIYIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIXIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIYIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIXIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIYIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIXIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIYIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIXIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIYIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIXIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIYIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIXIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIYIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIXIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIYIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIXIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIYIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIXIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIYIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIXIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIYIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IXIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IYIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'XIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'YIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'ZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII'],
              coeffs=[0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j,
 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j,
 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j,
 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j,
 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j,
 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j,
 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j,
 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j,
 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j,
 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j,
 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j,
 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j,
 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j,
 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j,
 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j,
 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j,
 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j,
 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j,
 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j,
 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j,
 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j,
 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j,
 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j,
 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j,
 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j,
 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j,
 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j,
 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j,
 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j,
 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j,
 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j,
 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j,
 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j,
 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j,
 0.78539816+0.j, 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j,
 1.57079633+0.j, 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j,
 0.39269908+0.j, 0.78539816+0.j, 1.57079633+0.j, 1.04719755+0.j,
 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j,
 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j,
 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j,
 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j,
 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j,
 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j,
 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j,
 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j,
 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j,
 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j,
 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j,
 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j,
 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j,
 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j,
 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j,
 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j,
 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j,
 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j,
 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j,
 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j,
 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j,
 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j,
 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j,
 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j,
 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j,
 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j,
 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j,
 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j,
 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j,
 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j,
 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j,
 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j,
 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j,
 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j,
 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j,
 1.04719755+0.j, 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j,
 0.52359878+0.j, 0.34906585+0.j, 1.04719755+0.j, 0.52359878+0.j,
 0.34906585+0.j])
observable = SparsePauliOp.from_sparse_list(
    [("Z", [i], 1 / num_qubits) for i in range(num_qubits)],
    num_qubits,
)
observable
SparsePauliOp(['IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZ', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZI', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IIZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'IZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII', 'ZIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII'],
              coeffs=[0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j,
 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j,
 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j,
 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j,
 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j,
 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j,
 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j, 0.02+0.j,
 0.02+0.j])

Para este problema ampliado, consideramos o tempo de evolução como 0.20.2 com 44 passos de Trotter. O problema é selecionado de forma que esteja além da simulação clássica por força bruta, mas possa ser simulado pelo método de rede de tensores. Isso nos permite verificar o resultado obtido via retropropagação em um computador quântico com o resultado ideal.

O valor de expectativa ideal para este problema, obtido via simulação de rede de tensores, é ≃0.89\simeq 0.89.

circuit = generate_time_evolution_circuit(
    hamiltonian,
    time=0.2,
    synthesis=LieTrotter(reps=4),
)
circuit.draw("mpl", style="iqp", fold=-1, scale=0.6)

Output of the previous code cell

Passo 2: Otimizar problema para execução em hardware quântico​

slices = slice_by_gate_types(circuit)
print(f"Separated the circuit into {len(slices)} slices.")
Separated the circuit into 36 slices.

Especificamos max_error_per_slice como 0.005, assim como antes. No entanto, como o número de fatias para este problema em grande escala é muito maior do que o problema em pequena escala, permitir um erro de 0.005 por fatia pode acabar criando um grande erro geral de retropropagação. Podemos limitar isso especificando max_error_total, que limita o erro total de retropropagação, e definimos seu valor como 0.03 (que é aproximadamente o mesmo do exemplo em pequena escala).

Para este exemplo em grande escala, permitimos um valor maior para o número de grupos comutativos, e o definimos como 15.

op_budget = OperatorBudget(max_qwc_groups=15)
truncation_error_budget = setup_budget(
    max_error_total=0.03, max_error_per_slice=0.005
)

Vamos primeiro obter o circuito retropropagado e o observável sem qualquer truncamento.

bp_obs, remaining_slices, metadata = backpropagate(
    observable, slices, operator_budget=op_budget
)
bp_circuit = combine_slices(remaining_slices)

print(f"Backpropagated {metadata.num_backpropagated_slices} slices.")
print(
    f"New observable has {len(bp_obs.paulis)} terms, which can be combined into {len(bp_obs.group_commuting(qubit_wise=True))} groups."
)
print(
    f"Note that backpropagating one more slice would result in {metadata.backpropagation_history[-1].num_paulis[0]} terms "
    f"across {metadata.backpropagation_history[-1].num_qwc_groups} groups."
)
print("The remaining circuit after backpropagation looks as follows:")
bp_circuit.draw("mpl", fold=-1, scale=0.6)
Backpropagated 7 slices.
New observable has 634 terms, which can be combined into 12 groups.
Note that backpropagating one more slice would result in 1246 terms across 27 groups.
The remaining circuit after backpropagation looks as follows:

Output of the previous code cell

Agora, permitindo truncamento, obtemos:

bp_obs_trunc, remaining_slices_trunc, metadata = backpropagate(
    observable,
    slices,
    operator_budget=op_budget,
    truncation_error_budget=truncation_error_budget,
)

# Recombine the slices remaining after backpropagation
bp_circuit_trunc = combine_slices(
    remaining_slices_trunc, include_barriers=False
)

print(f"Backpropagated {metadata.num_backpropagated_slices} slices.")
print(
    f"New observable has {len(bp_obs_trunc.paulis)} terms, which can be combined into {len(bp_obs_trunc.group_commuting(qubit_wise=True))} groups.\n"
    f"After truncation, the error in our observable is bounded by {metadata.accumulated_error(0):.3e}"
)
print(
    f"Note that backpropagating one more slice would result in {metadata.backpropagation_history[-1].num_paulis[0]} terms "
    f"across {metadata.backpropagation_history[-1].num_qwc_groups} groups."
)
print("The remaining circuit after backpropagation looks as follows:")
bp_circuit_trunc.draw("mpl", fold=-1, scale=0.6)
Backpropagated 10 slices.
New observable has 646 terms, which can be combined into 14 groups.
After truncation, the error in our observable is bounded by 2.998e-02
Note that backpropagating one more slice would result in 1226 terms across 29 groups.
The remaining circuit after backpropagation looks as follows:

Output of the previous code cell

Observamos que permitir o truncamento leva à retropropagação de mais três fatias. Podemos verificar a profundidade de 2 qubits do circuito original, do circuito retropropagado e do circuito retropropagado com truncamento após a transpilação.

# Transpile original experiment
circuit_isa = pm.run(circuit)
observable_isa = observable.apply_layout(circuit_isa.layout)

# Transpile the backpropagated experiment
bp_circuit_isa = pm.run(bp_circuit)
bp_obs_isa = bp_obs_trunc.apply_layout(bp_circuit_isa.layout)

# Transpile the backpropagated experiment with truncated observable terms
bp_circuit_trunc_isa = pm.run(bp_circuit_trunc)
bp_obs_trunc_isa = bp_obs_trunc.apply_layout(bp_circuit_trunc_isa.layout)
print(
    f"2-qubit depth of original circuit: {circuit_isa.depth(lambda x:x.operation.num_qubits==2)}"
)
print(
    f"2-qubit depth of backpropagated circuit: {bp_circuit_isa.depth(lambda x:x.operation.num_qubits==2)}"
)
print(
    f"2-qubit depth of backpropagated circuit with truncation: {bp_circuit_trunc_isa.depth(lambda x:x.operation.num_qubits==2)}"
)
2-qubit depth of original circuit: 48
2-qubit depth of backpropagated circuit: 40
2-qubit depth of backpropagated circuit with truncation: 36

Passo 3: Executar usando primitivas Qiskit​

pubs = [
    (circuit_isa, observable_isa),
    (bp_circuit_isa, bp_obs_isa),
    (bp_circuit_trunc_isa, bp_obs_trunc_isa),
]
options = EstimatorOptions()
options.default_precision = 0.01
options.resilience_level = 2
options.resilience.zne.noise_factors = [1, 1.2, 1.4]
options.resilience.zne.extrapolator = ["linear"]

estimator = EstimatorV2(mode=backend, options=options)
job = estimator.run(pubs)

Passo 4: Pós-processar e retornar resultado no formato clássico desejado​

result_no_bp = job.result()[0].data.evs.item()
result_bp = job.result()[1].data.evs.item()
result_bp_trunc = job.result()[2].data.evs.item()
print(f"Expectation value without backpropagation: {result_no_bp}")
print(f"Backpropagated expectation value: {result_bp}")
print(f"Backpropagated expectation value with truncation: {result_bp_trunc}")
Expectation value without backpropagation: 0.7887194658035515
Backpropagated expectation value: 0.9532818300978584
Backpropagated expectation value with truncation: 0.8913400398926913
methods = [
    "No backpropagation",
    "Backpropagation",
    "Backpropagation w/ truncation",
]
values = [result_no_bp, result_bp, result_bp_trunc]

ax = plt.gca()
plt.bar(methods, values, color="#a56eff", width=0.4, edgecolor="#8a3ffc")
plt.axhline(0.89)
ax.set_ylim([0.6, 0.98])
plt.text(0.2, 0.895, "Exact result")
ax.set_ylabel(r"$M_Z$", fontsize=12)
Text(0, 0.5, '$M_Z$')

Output of the previous code cell

Pesquisa do tutorial​

Por favor, responda a esta breve pesquisa para fornecer feedback sobre este tutorial. Seus insights nos ajudarão a melhorar nossas ofertas de conteúdo e experiência do usuário.

Link to survey