Jastrow–AGP Wavefunctions

• Jastrow–AGP wavefunctions are variational ansatzes that combine a Jastrow factor and an antisymmetrized geminal power to capture both dynamic and static electron correlations.
• They leverage a multiplicative Jastrow factor to enforce short-range cusp conditions and an AGP structure that generalizes pairing mechanisms similar to BCS superconductivity.
• Utilized in advanced QMC and tensor network methods, these wavefunctions enable systematic improvements in modeling strongly correlated quantum systems.

A Jastrow–AGP (Antisymmetrized Geminal Power) wavefunction is a highly structured variational ansatz central to modern quantum chemistry and strongly correlated quantum many-body physics. It combines a multiplicative Jastrow correlation factor—encoding explicit many-body correlations—with an AGP, which implements a pairing mechanism via antisymmetrized products of two-body functions (geminals). This construction unifies several crucial ideas: capturing static and dynamic correlation, enforcing proper nodal structure, and allowing systematic improvements by either modifying the Jastrow or extending the geminal expansion. The Jastrow–AGP form serves as a foundation for advanced variational Monte Carlo (VMC) and diffusion Monte Carlo (DMC) calculations and is deeply intertwined with modern developments in quantum simulations, tensor networks, and the stochastic optimization of correlated states.

1. Structure and Mathematical Definition

The Jastrow–AGP wavefunction $\Psi_{\mathrm{JAGP}}$ for an $N$-fermion system (usually electrons) is given by: $$\Psi_{\mathrm{JAGP}}(x_1,\ldots,x_N) = J(x_1,\ldots,x_N)\;\Psi_{\mathrm{AGP}}(x_1,\ldots,x_N)$$ where:

• $J(x_1,\ldots,x_N)$ is a positive, symmetric Jastrow factor encoding explicit correlation among subsets of particles, often written as $\exp[\sum_{i<j} u(x_i,x_j)+\cdots]$ with $u$ a two-body function but potentially with higher-body extensions.
• $\Psi_{\mathrm{AGP}}$ is the antisymmetrized product of $N/2$ geminals: $$\Psi_{\mathrm{AGP}}(x_1,\ldots,x_N) = \mathcal{A}[\,g(x_1,x_2)\,g(x_3,x_4)\cdots\,g(x_{N-1},x_N)\,]$$ with $\mathcal{A}$ the full antisymmetrizer and $N$0 a symmetric two-body (geminal) function.

The AGP generalizes the Bardeen-Cooper-Schrieffer (BCS) superconducting wavefunction, allowing arbitrary (not necessarily factorizable) symmetric pair functions $N$1. The Jastrow pre-factor modifies both the amplitude and the nodal structure (indirectly via amplitude) and captures short- and mid-range correlation effects, essential for an accurate description of electron correlation in both weakly and strongly correlated systems.

2. AGP Formalism and Matrix Representation

For systems with an even number of particles, the AGP can be written in first-quantized language as a geminal determinant: $N$2 or as a determinant, when $N$3 is restricted to a sum over a single-particle orbital basis: $N$4 This gives, for the $N$5-particle case,

$N$6

with the appropriate pairing orbitals (see e.g. references on Pfaffian and BCS forms). The AGP reduces to single-determinant Hartree–Fock for single-pairing or to multi-determinant configurations when $N$7 has higher rank.

3. Jastrow Factor and Correlation Hierarchies

The Jastrow operator $N$8 is constructed to encode explicit $N$9-body correlations. The simplest and still standard form is the two-body electron-electron Jastrow: $$\Psi_{\mathrm{JAGP}}(x_1,\ldots,x_N) = J(x_1,\ldots,x_N)\;\Psi_{\mathrm{AGP}}(x_1,\ldots,x_N)$$0 where $$\Psi_{\mathrm{JAGP}}(x_1,\ldots,x_N) = J(x_1,\ldots,x_N)\;\Psi_{\mathrm{AGP}}(x_1,\ldots,x_N)$$1 is chosen such that it imposes the proper short-range correlation cusp (e.g., for electrons, the Kato cusp condition). Higher-body terms (three-body, electron-nucleus-electron, etc.) are introduced as needed to account for more intricate correlations (important in solids and atoms with heavy elements).

The Jastrow operator strongly influences the efficiency and accuracy of QMC calculations, suppressing the weight of high-energy (short-range) configurations and dramatically accelerating statistical convergence.

4. Connections to Quantum Monte Carlo and Variational Optimization

Jastrow–AGP wavefunctions constitute the primary ansatz for variational Monte Carlo (VMC) and fixed-node diffusion Monte Carlo (DMC) methods. The variational parameters in both the geminal and the Jastrow factor are optimized—traditionally by energy minimization or stochastic reconfiguration—to minimize the variational energy or variance [see e.g. discussions in (Qin, 2020)].

The AGP structure allows for capturing static correlation (degeneracy between configurations), while the Jastrow factor is essential for dynamic correlation (short-range electron-electron interactions). In fixed-node DMC, the AGP determinant enforces the nodal structure, while the Jastrow factor modifies the amplitude and does not affect nodes directly but impacts the quality of the trial function and the fixed-node error.

Contemporary QMC methods employ increasingly flexible Jastrow forms (using neural-network-inspired or tensor-network-enhanced parameterizations) and extend the AGP to Pfaffian forms for pairing with broken time-reversal or spin symmetry.

5. Theoretical Properties and Relationship to Tensor Network States

The Jastrow–AGP ansatz can be viewed as a prototype for correlated paired-wavefunction constructions underlying diverse approaches, including tensor networks such as PEPS and MPS [see e.g. (Orus, 2013, Mortier et al., 2020)]. The AGP is itself a special case of a Fermionic Gaussian state, and applying a symmetric, local Jastrow operator corresponds to an exponential map of Hamiltonian-like operators (e.g., for two-body Jastrow exponential of a sum of projectors). The connection between the Jastrow–AGP ansatz and projected entangled pair states (PEPS) becomes explicit by interpreting the Jastrow as a diagonal multiplier and the AGP as a PEPS with bond dimension corresponding to the number of pairing channels. Advanced tensor network algorithms can variationally optimize generalized Jastrow–AGP-like states, parameterizing the Gutzwiller projector or longer-range multi-body Jastrow terms via matrix-product operators or higher-dimensional generalizations.

6. Physical Applications and Extensions

Jastrow–AGP wavefunctions are routinely applied to:

• Strongly correlated atoms, molecules, and solids: capturing both dynamical and static (near-degenerate) correlation, improving over Hartree–Fock and traditional single-determinant approximations.
• Superconducting states: AGP/BCS serves as the variational ground state for s-wave and other superconductors; Jastrow factors suppress double occupation and encode interactions beyond mean-field approximations.
• Resonating Valence Bond (RVB) and quantum spin liquids: PEPS or MPS representations of RVB states correspond to Jastrow–AGP forms, where the Jastrow captures projective Gutzwiller projection constraints (Schuch et al., 2012).
• Quantum chemical and condensed-matter QMC: state-of-the-art variational energies for realistic systems are achieved using systematically improved Jastrow–AGP constructions (Qin, 2020).

Extensions include:

• Pfaffian wavefunctions (for broken spin/number symmetry);
• Correlation-consistent geminal expansions (multi-AGP or full configuration interaction limits);
• Neural-network parameterizations of both Jastrow and pairing parts.

7. Limitations and Perspectives

While the Jastrow–AGP ansatz is powerful, its exact optimization remains challenging due to the high-dimensional, non-convex nature of the parameter landscape. The quality of the node structure—determined by the AGP—limits DMC accuracy, though systematic improvements are possible by enlarging the pairing function space. The Jastrow factor, while essential for efficiency, is variationally flexible and does not affect the nodal structure directly. For systems with significant multi-reference character, extensions to multi-AGP or Pfaffian forms are required. Interpretation in tensor network language (as Gutzwiller–projected BCS states or as PEPS formalisms) continues to provide insight into the expressiveness and limitations of the ansatz, especially as tensor network and QMC approaches converge.

---

References:

• (Orus, 2013) for foundational tensor network definitions and PEPS/MPS context.
• (Qin, 2020) for QMC implementations and the role of variationally optimized PEPS/Jastrow-Geminal states.
• (Schuch et al., 2012) for connections to RVB and symmetry-projected paired states in PEPS language.
• (Mortier et al., 2020) for fermionic PEPS as a generalization of pairing/AGP structures in tensor networks.