Quantum Phase Estimation-Based CASCI

• QPE-CASCI is a hybrid quantum chemistry approach that integrates quantum phase estimation with complete active space configuration interaction to directly prepare ground and excited electronic states.
• The methodology employs Trotterized quantum circuits and qubit mapping via Jordan–Wigner or Bravyi–Kitaev transformations to simulate the active-space Hamiltonian with optimized resource scaling.
• This approach offers scalable simulation, robust state extraction, and advanced error mitigation, making it effective for strongly correlated, multireference, and supramolecular systems.

Quantum Phase Estimation-Based Complete Active Space Configuration Interaction (QPE-CASCI) is an approach that combines the complete active space configuration interaction methodology of quantum chemistry with quantum phase estimation algorithms implemented on quantum computers. QPE-CASCI enables direct preparation of ground and excited electronic states in an active space, facilitates robust extraction of molecular energies, and offers efficient scaling for systems intractable by classical means. This methodology is extensible to fragmented, supramolecular, and strongly correlated systems, leveraging algorithmic innovations, resource scaling heuristics, and advanced error mitigation.

1. Theoretical Framework and Circuit Architecture

QPE-CASCI operates on the second-quantized active-space electronic Hamiltonian: $$H_{\rm CAS} = \sum_{p,q=1}^M h_{pq} a_p^\dagger a_q + \frac{1}{2}\sum_{p,q,r,s=1}^M g_{pqrs} a_p^\dagger a_q^\dagger a_s a_r$$ where $M$ is the number of active orbitals, $h_{pq}$ and $g_{pqrs}$ are one- and two-electron integrals, and $a_p^\dagger, a_q$ denote fermionic operators. Qubit encoding via Jordan-Wigner or Bravyi-Kitaev transformations yields a Hamiltonian of the form $H_{\rm qub} = \sum_\ell \omega_\ell P_\ell$ with Pauli strings $P_\ell$ (Tachi et al., 4 Dec 2025).

Quantum phase estimation (QPE) circuits comprise:

• An ancilla (phase) register of $n_a$ qubits (precision parameter).
• A system register of $N_f$ qubits holding the active-space wavefunction.
• Hadamard gates on the ancilla register, controlled-$U^{2^j}$ gates (with $U = e^{-i H_{\rm qub} t}$ Trotterized in $m$ slices), an inverse Quantum Fourier Transform (QFT), and measurement of the ancillas (D'Cunha et al., 2023).

After the IQFT and measurement, the system register collapses onto an eigenstate $|E_k\rangle$ with probability $|\langle\psi_0|E_k\rangle|^2$, extracting the molecular energy $E_k = 2\pi\,\phi_k / t$ where $\phi_k$ is the measured phase (Ino et al., 2023).

2. Active Space Selection and Initial State Preparation

Choosing the optimal active space is achieved via several protocols:

• MP2/Pseudo-Natural Orbitals: Diagonalizing the MP2 one-body density matrix yields PNOs, enabling compact orbital basis selection (Ino et al., 2023).
• Boys Localization: Minimizing the spread $\langle r_1 - r_2 \rangle^2$ separates occupied and virtual orbitals into locally centered sets, facilitating supramolecular partitioning (Tachi et al., 4 Dec 2025).

For initial state design, classical CISD calculations within the active space provide dominant configurations for the excited state. Construction of $$|\psi_\text{init}\rangle = \sum_{\mu=1}^k c_\mu |\Phi_\mu\rangle$$ (with $\sum |c_\mu|^2 \approx 0.8-0.85$) balances overlap versus circuit complexity (Ino et al., 2023). For fragment-based approaches, each fragment $f$ is initialized in a guess $|\psi_0\rangle_f$ (e.g., LASSCF or RHF) (D'Cunha et al., 2023).

3. Quantum Circuit Implementation and Resource Scalability

Hamiltonian simulation is effected through second- or higher-order Trotter-Suzuki decompositions. Each Trotter slice applies exponentials of Pauli strings, incurring $O(M^4)$ gates per slice (for $M$ orbitals) (Kang et al., 2022, Ino et al., 2023).

Resource scaling is summarized below:

Algorithm	Qubit Count	CNOT Complexity	Scaling Behavior
Direct Init.	$N_f$	$G_{\rm DI}(N_f) = 4^{N_f} - \frac{3}{2}\,2^{N_f}$	Exponential in $N_f$ (D'Cunha et al., 2023)
Fragmented QPE	$N_f + n_a$	$G_{\rm QPE} = n_U(N_f) \times m \times 2^{n_a-1}$	Poly in $N_f$, exponential in $n_a$ (D'Cunha et al., 2023)
Supramolecular QPE	$N_s+N_a$	$>10^7$ (uncompressed), $<10^5$ (compressed)	Compressible via gate fusion and commutation (Tachi et al., 4 Dec 2025)

For large fragments ($N_f>20$), fragmented QPE is asymptotically favorable compared to DI. Ancilla reuse further reduces total qubit requirements in the fragmentation protocol (D'Cunha et al., 2023).

Gate-depth for practical examples (benzene, M=6): depth $\sim$ 21,600 CNOTs (Ino et al., 2023). Compression techniques give $>$99% reduction in two-qubit gate count (Tachi et al., 4 Dec 2025).

4. Error Analysis and Mitigation Strategies

Two principal error sources are

• Phase-estimation resolution: $\Delta E_{\rm phase} = O(2^{-n_a}/t)$
• Trotter error: $\Delta E_{\rm Trot} = O(t^2/m)$ (second-order decomposition).

Optimal ancilla and Trotter parameters follow:

• $$n_a \approx \lceil\log_2(2\pi /(\Delta E_{\rm target} t))\rceil$$
• $m$ chosen so that $C t^2/m \lesssim \Delta E_{\rm target}$

Empirically, $n_a = 6$–$8$ and $m = 4$–$9$ suffice for chemical accuracy ($<1.6$ m$E_h$) for weakly correlated fragments. Strong correlation regimes may require $n_a \sim 15$–21 (D'Cunha et al., 2023).

Algorithmic Error Mitigation (AEM) utilizes extrapolation $E'(1/M) = a(1/M)^2 + b$ to estimate $M \to \infty$ limit, minimizing second-order Trotter artifacts. Readout refinement uses “WgtAve” bit-string averaging to reduce phase bias (Tachi et al., 4 Dec 2025).

Numerical benchmarks on H$_2$O (15 orbitals): QPE with nMP2 orbitals yields $E_0 = -76.2225$ Hartree, within 0.02 Hartree of CCSDTQ(24), and excitation energies accurate to within 2–3 eV (Kang et al., 2022). Water dimer interaction energies: QPE–CASCI gives $E_\text{int}(QPE) = -5.1333$ kcal/mol, with $+$0.0197 kcal/mol error relative to CASCI (Tachi et al., 4 Dec 2025).

5. Extensions to Fragmentation, Downfolding, and Advanced Protocols

Fragmentation divides the system into smaller subsystems $f$, each with its own QPE subcircuit, facilitating multireference state preparation and enabling polynomial scaling of gate requirements. Ancilla qubits may be reused sequentially between fragments, enhancing qubit efficiency (D'Cunha et al., 2023).

DUCC (double unitary coupled-cluster) downfolding constructs state-specific similarity-transformed effective Hamiltonians $H_\text{eff}(K)$: $$H_\text{eff}(K) = P e^{-\sigma_\text{ext}(K)} H e^{\sigma_\text{ext}(K)} P$$ where $P$ projects onto the active space, and amplitudes $\sigma_\text{ext}$ are determined from classical EOM-CCSD. This reduces required qubits and captures dynamical correlation outside the CAS. QPE on $H_\text{eff}$ yields ground/excited state energies within single-digit millihartree error, outperforming bare CASCI (Bauman et al., 2019).

Stochastic spectrum extraction via repeated QPE collapses on the phase register enables multi-state resolution without separate state-specific circuits (Bauman et al., 2019).

6. Practical Implementation Workflow and Applications

A standard implementation pipeline comprises:

1. Classical preprocessing: HF $\to$ MP2, diagonalize density matrix, obtain PNOs/Boys orbitals, select active space.
2. Calculate $h_{pq}$, $h_{pqrs}$ in chosen basis; perform CAS-CISD to guide initial-state configurations.
3. Map second-quantized Hamiltonian to Pauli strings via Jordan–Wigner; apply symmetry tapering if possible.
4. State-preparation: single determinant or compact multi-Slater CI expansion balanced for overlap and circuit cost.
5. Quantum circuit construction: controlled-Trotterized $U$, ancilla register appropriate for desired energy precision.
6. QPE run, extract phase $\phi$, compute energy $E = 2\pi \phi / t$, repeat for multiple electronic states.

Tested applications include ground and excited $\pi$–$\pi^*$ state calculations in benzene and derivatives, water dimer interaction energies, and H$_2$/H$_4$ strongly correlated models. Chemical accuracy and sub-kcal/mol precision have been substantiated in numerical experiments up to 18 qubits (Ino et al., 2023, Tachi et al., 4 Dec 2025).

7. Limitations and Outlook

Key limitations include:

• Exponential scaling in ancilla count for high precision
• State-specific classical downfolding required for DUCC-based QPE-CASCI
• Need for accurate classical amplitudes; EOM-CCSD truncation can be limiting for strongly multireference scenarios (Bauman et al., 2019)

Open research avenues involve:

• Development of state-averaged DUCC for simultaneous multi-state downfolding
• Advanced circuit compression and error-mitigation strategies allowing feasible execution on future fault-tolerant quantum hardware (Tachi et al., 4 Dec 2025)
• Systematic exploration of larger active spaces via fragmentation and ancilla reuse.

QPE-CASCI thus provides a rigorous, scalable protocol for directly accessing strongly correlated, multireference molecular states and energy landscapes, with robust resource scaling and extensibility to advanced quantum simulation paradigms (D'Cunha et al., 2023, Ino et al., 2023, Tachi et al., 4 Dec 2025, Kang et al., 2022, Bauman et al., 2019).