betaFeedback

WorksPhysics and Astronomy

Multi-symmetry projected coupled cluster theory with triples corrections

Abstract

Symmetry projection and coupled cluster theory provide complementary descriptions of static and dynamic correlation. We introduce two approximate perturbative triples corrections to spin- and point-group-projected coupled cluster with single and double excitations (PCCSD). The coherent (T)(T) correction combines triples intermediates over the projection grid, whereas +(T)+(T) evaluates the conventional triples expression using PCCSD amplitudes. Both recover conventional (T)(T) in the absence of projection. We assess these corrections for N2_2 and C2_2 dissociation, 3d3d transition-metal monoxides, and Cr2_2. Projection improves the singles-and-doubles description and distinguishes selected low-lying symmetry sectors. For the six-monoxide benchmark subset, coherent (T)(T) reduces the PCCSD mean absolute error from 13.413.4 to 3.0~\mEh, achieving accuracy comparable to conventional CCSD(T). Both triples corrections restore the short-bond minimum missing from the PCCSD curves of Cr2_2, with coherent (T)(T) giving a well depth closer to extrapolated DMRG. In contrast, +(T)+(T) provides more balanced N2_2 and C2_2 dissociation curves. These results demonstrate the benefits of perturbative triples in PCC, while their system-dependent performance highlights the need for a more systematic treatment.

Introduction

Coupled cluster (CC) theory provides a systematic and accurate description of electron correlation for weakly-correlated molecules.[1, 2, 3] The singles-and-doubles approximation, CCSD,[4] supplemented by a perturbative triples correction, CCSD(T),[5, 6, 7] is widely regarded as the “gold standard” for such systems. Applications of coupled-cluster methods to large systems have been enabled by local correlation and other cost-reduction strategies.[8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24] However, CCSD(T) becomes less impressive when several configurations become important, as in stretched multiple bonds and many transition-metal compounds. In such cases, strong static correlation can require higher-order excitations or alternative treatments that account explicitly for the multireference character of the wave function.

A common approach to strong correlation is to describe the important configurations explicitly within an active space. Complete active space self-consistent field (CASSCF) theory[25, 26] provides such a reference, with dynamic correlation subsequently recovered through methods such as CASPT2[27, 28] or NEVPT2.[29, 30, 31] Multireference CC offers another route to combining a multiconfigurational reference with an exponential treatment of correlation.[32, 33, 34, 35, 36, 37, 38, 39] These developments motivate compact wave functions that describe both static and dynamic correlation across different bonding regimes. However, multireference CC is far from straightforward.

Allowing the reference determinant to break symmetry can improve the description of bond dissociation,[40] but at the cost of losing the correct symmetry quantum numbers—the familiar symmetry dilemma.[41, 42] A subsequent truncated coupled-cluster treatment does not, in general, restore the desired spin or spatial symmetry. Symmetry projection recovers the correct symmetries and provides a different construction of a multiconfigurational state. Following the ideas of Löwdin,[41] projected Hartree–Fock (PHF) restores the symmetries of a broken-symmetry determinant and optimizes that determinant in the presence of projection.[43, 44, 45] The projected state is a linear combination of nonorthogonal determinants whose projection weights are determined by symmetry, while any required component-mixing coefficients are optimized variationally. The broader projected quasiparticle framework also permits particle-number, spin, spatial, and complex-conjugation symmetry restoration.[46, 47] These approaches capture important static correlations without specifying an active space, but additional dynamic correlation is needed for quantitative energies.

Several methods have been developed to recover this missing correlation by incorporating particle-hole excitations on the broken-symmetry reference, including spin-projected configuration interaction,[48] spin-extended coupled electron pair approximations,[49] and projected perturbation theories.[50, 51] Other approaches introduce correlators acting directly on the projected reference, including killing operators,[52, 53] Jastrow operators[54] and non-linear correlators.[55] Polynomial representations of spin-projected unrestricted and generalized Hartree–Fock states have also established connections with cluster expansions and enabled combinations with CC.[56, 57]

Direct symmetry projection of a CC wave function has been developed in several related formulations. Duguet introduced symmetry-broken and restored CC for angular momentum,[58] subsequently extended to particle number.[59] Qiu and co-workers formulated projected CC (PCC) for molecular and model Hamiltonians using disentangled cluster operators[60, 61] and demonstrated the importance of optimizing amplitudes in the presence of projection.[62] An alternative spin-extended CC approach approximates the exponential through a truncated Taylor expansion.[63] The power series formulation combines this strategy with disentangled cluster operators,[64] and subsequent work incorporates complex-conjugation and spin-flip projection.[65]

An important remaining question is how to include triple excitations in PCCSD. Projection generates correlations beyond the excitation rank of the underlying cluster operator, but does not eliminate the need for additional dynamic correlation. Related work on approximate coupled-pair methods likewise highlights the importance of incorporating three-body clusters when treating strong correlation.[66] Here, we compare two approximate perturbative corrections that recover conventional (T) when projection is absent.

We examine spin projection of unrestricted and generalized references, combined with point-group projection where needed to distinguish spatial symmetry sectors. N2_2 and C2_2 dissociation test bond recoupling and competing states. Transition-metal monoxides and Cr2_2, for which accurate reference calculations are available,[67, 68] provide more demanding tests of the balance between static and dynamic correlation. Our aim is to assess the performance of symmetry projection and the two triples prescriptions across these regimes.

Theory

In this section, we briefly review symmetry projection and the basic framework of projected coupled cluster (PCC). More detailed discussions of the disentangled cluster formalism and its power series approximation can be found in Refs. 60,62,64. Throughout, ii, jj, kk, ll, …\ldots, aa, bb, cc, dd, …\ldots, and pp, qq, rr, ss, …\ldots denote occupied, virtual, and general spin orbitals, respectively.

Projection operators

For continuous symmetries, projection operators can be expressed as integrals over symmetry-generated transformations. For discrete symmetries, the integrals are replaced by sums. Here, we consider spin and point-group symmetry projection. The spin projection operators are

Pmks=∣s,m⟩⟨s,k∣=2s+18π2∫dΩ Dmks∗(Ω)R(Ω),(1)P^{s}_{mk}=|s,m\rangle\langle s,k| =\frac{2s+1}{8\pi^{2}}\int d\Omega\,D^{s*}_{mk}(\Omega)R(\Omega), \tag*{(1)}
R(Ω)=eiαSzeiβSyeiγSz.(2)R(\Omega)=e^{i\alpha S_z}e^{i\beta S_y}e^{i\gamma S_z}. \tag*{(2)}

Here, ∣s,m⟩|s,m\rangle is a spin eigenstate, Dmks(Ω)=⟨s,m∣R(Ω)∣s,k⟩D^{s}_{mk}(\Omega)=\langle s,m|R(\Omega)|s,k\rangle is a Wigner rotation matrix element, and Ω=(α,β,γ)\Omega=(\alpha,\beta,\gamma) denotes Euler angles. We use dΩ=dαsin⁡β dβ dγd\Omega=d\alpha\sin\beta\,d\beta\,d\gamma, with 0≤α,γ<2π0\leq\alpha,\gamma<2\pi and 0≤β≤π0\leq\beta\leq\pi.

The point-group projection operators are

PλκJ=lJh∑g∈GΓλκJ∗(g)Rg,(3)P^{J}_{\lambda\kappa}=\frac{l^{J}}{h}\sum_{g\in G}\Gamma^{J*}_{\lambda\kappa}(g)R_g, \tag*{(3)}

where hh is the order of the group GG, lJl^{J} is the dimension of the irreducible representation ΓJ\Gamma^{J}, and ΓλκJ(g)\Gamma^{J}_{\lambda\kappa}(g) is the corresponding representation matrix element for the group operation gg, where RgR_g is the action of gg on the electronic wave function.

For either symmetry, projected components can be combined as

∣Ψ⟩=∑ycyPxyZ∣Φ⟩.(4)|\Psi\rangle=\sum_y c_yP^Z_{xy}|\Phi\rangle. \tag*{(4)}

Here, ZZ denotes the target irreducible representation, and x,yx,y label its components. This notation includes spin projection, with (Z,x,y)=(s,m,k)(Z,x,y)=(s,m,k), and point-group projection, with (Z,x,y)=(J,λ,κ)(Z,x,y)=(J,\lambda,\kappa). For simultaneous projection, the labels are composite, and the operators are products of the corresponding spin and point-group operators. The coefficients cyc_y are determined from a generalized eigenvalue problem. This mixing removes the dependence of the projected energy on the orientation of the underlying determinant within the symmetry group.[45]

Projected Hartree–Fock

We incorporate the mixing coefficients from both the bra and ket into the contracted operator

P=∑yy′cy′∗cyPyy′Z.(5)P=\sum_{yy'}c^{*}_{y'}c_yP^{Z}_{yy'}. \tag*{(5)}

Using the multiplication and conjugation properties of the projection operators, the projected Hartree–Fock (PHF) energy is then

EPHF=min⁡∣0⟩⟨0∣HP∣0⟩⟨0∣P∣0⟩.(6)E_{\mathrm{PHF}}=\min_{\lvert0\rangle}\frac{\langle0|HP|0\rangle}{\langle0|P|0\rangle}. \tag*{(6)}

where ∣0⟩|0\rangle is a broken-symmetry determinant. The coefficients cyc_y are optimized through the generalized eigenvalue problem described above and held fixed at their converged PHF values during the subsequent PCC calculation.

We minimize this functional using an effective-Fock formulation.[45] The effective Fock matrix is

F=F0+ρˉF+ρˉ+ρF−ρ.(7)F=F_0+\bar{\rho}F_+\bar{\rho}+\rho F_-\rho. \tag*{(7)}

where ρpq=⟨0∣cq†cp∣0⟩\rho_{pq} = \langle0|c_q^\dagger c_p|0\rangle is the one-particle density matrix of the reference and ρˉ=1−ρ\bar{\rho}=1-\rho. The first term is the orbital gradient,

(F0)pq=⟨0∣cq†cp(H−E)P∣0⟩⟨0∣P∣0⟩+h.c.,(8)(F_0)_{pq}=\frac{\langle0|c_q^\dagger c_p(H-E)P|0\rangle}{\langle0|P|0\rangle}+\text{h.c.}, \tag*{(8)}

and has only occupied–virtual and virtual–occupied blocks that vanish at convergence. Here, EE is the PHF energy at the current determinant. The remaining terms are associated with electron attachment and removal via a generalized Koopmans construction.[69, 70] We have defined

(F+)pq=12⟨0∣cp(H−E)cq†P∣0⟩⟨0∣P∣0⟩+h.c.,(9)(F_+)_{pq}=\frac{1}{2}\frac{\langle0|c_p(H-E)c_q^\dagger P|0\rangle}{\langle0|P|0\rangle}+\text{h.c.}, \tag*{(9)}
(F−)pq=−12⟨0∣cq†(H−E)cpP∣0⟩⟨0∣P∣0⟩+h.c.,(10)(F_-)_{pq}=-\frac{1}{2}\frac{\langle0|c_q^\dagger(H-E)c_pP|0\rangle}{\langle0|P|0\rangle}+\text{h.c.}, \tag*{(10)}

and these terms, due to the presence of the density matrices ρ\rho and ρˉ\bar{\rho}, define the occupied-occupied and virtual-virtual blocks of FF at convergence. Diagonalization of FF yields the PHF orbitals and the orbital energies used in the coherent perturbative triples correction below.

Projected coupled cluster

The projected coupled cluster energy and amplitude equations are[60]

E=⟨0∣HPeT∣0⟩⟨0∣PeT∣0⟩,(11)E=\frac{\langle0|HP e^T|0\rangle}{\langle0|P e^T|0\rangle}, \tag*{(11)}
Rμ=⟨μ∣(H−E)PeT∣0⟩=0.(12)R_\mu=\langle\mu|(H-E)P e^T|0\rangle=0. \tag*{(12)}

Here, ⟨μ∣\langle\mu| denotes an excited determinant relative to ∣0⟩|0\rangle, and the cluster operator TT contains single and double excitations. The cluster amplitudes are optimized in the presence of the projection operator, with the PHF reference held fixed.

The central computational task is to evaluate the Hamiltonian and overlap kernels involving a symmetry rotation and the coupled cluster wave function. We use the disentangled cluster formalism[60] schematically

ReT∣0⟩=eW∣0⟩,(13)R e^T|0\rangle=e^W|0\rangle, \tag*{(13)}

where WW contains a scalar term and pure excitation operators. A power series expansion gives[64]

ReT∣0⟩=(1+TˉR+12TˉR2+⋯ )R∣0⟩=(1+TˉR+12TˉR2+⋯ )eU0+U1∣0⟩=eU0+U1(1+TˉRU+12TˉRU2+⋯ )∣0⟩.(14)\begin{aligned} R e^T|0\rangle&=\left(1+\bar{T}_R+\frac{1}{2}\bar{T}_R^2+\cdots\right)R|0\rangle\\ &=\left(1+\bar{T}_R+\frac{1}{2}\bar{T}_R^2+\cdots\right)e^{U_0+U_1}|0\rangle\\ &=e^{U_0+U_1}\left(1+\bar{T}_{RU}+\frac{1}{2}\bar{T}_{RU}^2+\cdots\right)|0\rangle. \tag*{(14)} \end{aligned}

The barred operators denote the successive similarity transformations of TT, and the second line follows from Thouless’ theorem.[71] After acting on the reference, the resulting state can be written as C0(1+C1+C2+⋯ )∣0⟩C_0(1+C_1+C_2+\cdots)|0\rangle, where CnC_n creates nn-fold excitations. The disentangled cluster operators are then written as

eW0=C0,(15)e^{W_0}=C_0, \tag*{(15)}
W1=C1,(16)W_1=C_1, \tag*{(16)}
W2=C2−12C12,(17)W_2=C_2-\frac{1}{2}C_1^2, \tag*{(17)}

with analogous expressions at higher excitation rank. In this work, the singles contribution is treated through Thouless’ theorem, and the remaining transformed doubles exponential is expanded through second order. We retain only W0W_0, W1W_1, and W2W_2.

Unlike previous formulations,[60, 64] we apply the rotation operator RR to the ket rather than the bra. This avoids the additional similarity transformation of the Hamiltonian generated by the rotation-dependent Thouless de-excitation operator associated with the bra. The Hamiltonian kernels are therefore evaluated directly using the original integrals and the disentangled operators. The resulting formulation is equivalent to the previous one but is more compact and better suited to the perturbative triples treatment below.

Symmetry projection introduces linear dependencies among the projected excitations. Following the treatment of redundant residual directions in earlier work,[63, 72] we solve reduced residual equations,

rμ=∑νSμνRν=0,(18)r_\mu=\sum_\nu S_{\mu\nu}R_\nu=0, \tag*{(18)}

where Sμν=⟨μ∣P∣ν⟩S_{\mu\nu}=\langle\mu|P|\nu\rangle is the overlap matrix between projected excited determinants. This procedure addresses redundancies in the residual equations but does not entirely eliminate the corresponding redundant components of the cluster operator.

Perturbative triples with projection

Once the single and double amplitudes have been obtained, we estimate the contribution of triple excitations perturbatively. The projection operator complicates a direct extension of conventional CCSD(T). We therefore introduce an approximate correction whose unprojected limit recovers the usual (T)(T) expression.

Using the disentangled representation, the PCCSD energy in Eq. 9a can be written as

EPCCSD=∑gwgeW0(g)⟨0∣HˉW(g)∣0⟩∑gwgeW0(g).(19)E_{\mathrm{PCCSD}}=\frac{\sum_g w_g e^{W_0^{(g)}}\langle0|\bar{H}_{W(g)}|0\rangle}{\sum_g w_g e^{W_0^{(g)}}}. \tag*{(19)}

Here, wgw_g contains the projection and quadrature weights, and HˉW(g)\bar{H}_{W(g)} denotes the Hamiltonian similarity-transformed by the disentangled excitation operator.

Motivated by this kernel-weighted expression, we define an approximate triples intermediate by combining the connected triples kernels constructed from W2(g)W_2(g) with the same normalized weights:

Wijkabc=∑gwgeW0(g)⟨0∣a^ijkabc(VNW2(g))c∣0⟩∑gwgeW0(g).(20)\mathcal{W}_{ijk}^{abc}=\frac{\sum_g w_g e^{W_0(g)}\left\langle0\left| \hat{a}_{ijk}^{abc}\left(V_N W_2(g)\right)_c\right|0\right\rangle}{\sum_g w_g e^{W_0(g)}}. \tag*{(20)}

Here, VNV_N is the normal-ordered two-body part of the Hamiltonian, and the subscript c selects connected contributions. This expression is a proposed approximation motivated by the structure of conventional (T)(T), rather than an exact projected perturbative triples. It retains the conventional connected contraction evaluated with W2(g)W_2(g) and neglects contributions involving higher-rank disentangled operators.

Following the same prescription, we define the disconnected intermediate as

Vijkabc=∑gwgeW0(g)⟨0∣a^ijkabc(VNW1(g)+FNW2(g))disc∣0⟩∑gwgeW0(g).(21)\mathcal{V}_{ijk}^{abc}=\frac{\sum_g w_g e^{W_0(g)}\left\langle0\left| \hat{a}_{ijk}^{abc}\left(V_N W_1(g)+F_N W_2(g)\right)_{\mathrm{disc}}\right|0\right\rangle}{\sum_g w_g e^{W_0(g)}}. \tag*{(21)}

Here, FNF_N is the normal-ordered one-body part of the Hamiltonian, and the subscript disc selects the disconnected triples contributions. The correction is evaluated as

E(T)=136∑ijkabc(Wijkabc+Vijkabc)∗WijkabcDijkabc,(22)E_{(T)}=\frac{1}{36}\sum_{ijkabc}\frac{\left(\mathcal{W}_{ijk}^{abc}+\mathcal{V}_{ijk}^{abc}\right)^*\mathcal{W}_{ijk}^{abc}}{D_{ijk}^{abc}}, \tag*{(22)}

where the orbital-energy denominator is

Dijkabc=ϵi+ϵj+ϵk−ϵa−ϵb−ϵc.(23)D_{ijk}^{abc}=\epsilon_i+\epsilon_j+\epsilon_k-\epsilon_a-\epsilon_b-\epsilon_c. \tag*{(23)}

Motivated by earlier work on Brueckner coupled cluster theory[73] and the multireference character of the projected state, we construct this denominator using orbital energies obtained from the PHF effective Fock operator. We call this the coherent (T)(T) correction because the projection-grid contributions are combined at the intermediate level before the final energy contraction.

A simpler alternative is to evaluate the conventional, unprojected (T)(T) expression using the converged PCCSD amplitudes and add the resulting correction to the PCCSD energy. We denote this procedure by +(T)+(T). Its denominators are obtained from the semicanonicalized Fock matrix of the broken-symmetry reference.

In either case, the T1T_1 and T2T_2 amplitudes are those of PCCSD. It is easy to see that when there is no projection, PCCSD and PCCSD(T) reduce to the common UCCSD and UCCSD(T). When the deformed determinant is RHF, they reduce to RCCSD and RCCSD(T).

Results

Computational details

The PHF and PCC calculations were performed with an in-house Python/Fortran implementation[74] using PySCF[75] and Libcint[76] for the electronic integrals. Spin projection used 8-point Gauss–Legendre quadrature for unrestricted references and a 14×614 \times6 Lebedev–Gaussian grid for generalized-reference PCC.[77] The PHF references were optimized separately before solving the PCC equations. For generalized references, we applied the gauge-fixing procedure described in Appendix A to select the underlying determinant before the PCC calculation.

SUCCSD and SGCCSD denote spin-projected CCSD based on unrestricted and generalized determinants, respectively. The labels +coh(T)+\mathrm{coh}(T) and +(T)+(T) distinguish the triples corrections, and a C4C_4 prefix indicates point-group projection. We use C4C_4 to distinguish the rotational symmetry sectors considered here; the molecular term symbols identify the target states.

PCCSD convergence is assessed using the maximum absolute reduced residual, max⁡∣rμ∣\max|r_\mu|, with a threshold of 5×10−55 \times10^{-5}. We adopt this relatively loose threshold as a compromise between energy convergence and iteration count: PCCSD energy changes near this threshold are typically of order 10−4 Eh10^{-4}\ E_\mathrm{h}, while tighter thresholds require substantially more iterations.

All electrons were correlated for N2N_2 and C2C_2 in cc-pVDZ. The metal oxides were treated in def2-SVP at the geometries of Ref. 67, with the metal 1s1s, 2s2s, and 2p2p shells and oxygen 1s1s shell frozen. C2C_2 calculations used cc-pVDZ-DK, a spin-free one-electron exact two-component (X2C) Hamiltonian,[78, 79] and a frozen Ne core on each atom.

Main-group molecules

We begin with the dissociation of N2N_2 and C2C_2, two familiar tests of strong correlation in stretched bonds. Because the total-energy curves are closely spaced, the upper panels of Figure 1 and Figure 2 show PHF, PCCSD and the coherent triples correction, while the lower panels compare both triples prescriptions with their corresponding reference energies.

Figure 1. N2_2 dissociation in cc-pVDZ. (a) Total energies from CDFCI, PHF, PCCSD, and PCCSD+coh(TT). (b) Errors relative to CDFCI for PCCSD and both triples corrections.

Figure 2. C2_2 dissociation in cc-pVDZ. (a) Total energies from iCI, PHF, PCCSD, and PCCSD+coh(TT) for the 1Σg+^1\Sigma_g^+ and 1Δg^1\Delta_g states, targeted by C4(A)C_4(A) and C4(B)C_4(B) projection, respectively. (b) Errors relative to the corresponding iCI state for PCCSD and both triples corrections. The C4(A)C_4(A) PCCSD curve follows the lower of the two solution branches; triples corrections are evaluated on the selected branch.

The N2N_2 ground-state curve has 1Σg+{}^{1}\Sigma_g^+ symmetry throughout the range considered. At the HF level, the reference becomes unrestricted beyond the Coulson–Fischer point and remains so at dissociation. In contrast, variation-after-projection (VAP) optimization yields a deformed UHF reference throughout the entire bond-length range. We further allow SzS_z symmetry breaking to obtain a GHF reference and recover additional correlation energy upon projection. Using these optimized references, we compare spin-projected unrestricted and generalized CC with coordinate descent full configuration interaction (CDFCI) reference energies.[80] Figure 1 shows the total energies and errors for the 1Σg+{}^{1}\Sigma_g^+ ground state.

As at the PHF level, SGCCSD recovers more correlation than SUCCSD, although the deliberately introduced SzS_z symmetry breaking yields an improvement of only a few mEhE_\mathrm{h}. The largest PCCSD errors occur in the bondrecoupling region, reaching approximately 11.5 mEhE_{\mathrm{h}} for SUCCSD and 8.9 mEhE_{\mathrm{h}} for SGCCSD. The relatively large errors at equilibrium and in the recoupling region indicate the need for higher-order correlation. Both triples corrections recover the missing correlation near equilibrium. Coherent (TT), however, overcorrelates in the recoupling region, by about 4.7 mEhE_{\mathrm{h}} for SUCCSD and 8.3 mEhE_{\mathrm{h}} for SGCCSD. Thus, the better PCCSD reference does not necessarily lead to a more accurate triples-corrected energy. The simpler +(TT) prescription provides a more balanced curve: despite overcorrelation at short bond lengths, its errors remain within approximately 3 mEhE_{\mathrm{h}} throughout the plotted range.

C2_2 presents an additional challenge because low-lying states with the same spin quantum number but of different spatial symmetry cross along the dissociation coordinate. We target 1Σg+^1\Sigma_g^+ and 1Δg^1\Delta_g states with C4(A)C_4(A) and C4(B)C_4(B) projection and compare with iCI results [81] in Figure 2. Here, we focus on spatial symmetry and only break S2S^2 symmetry to have a UHF reference.

At the PHF level, two branches of C4(A)C_4(A) solutions cross near 3.3 Bohr, while the C4(A)C_4(A) and C4(B)C_4(B) curves cross near 2.9 bohr. At short bond lengths, the second C4(A)C_4(A) SUHF solution becomes unstable and collapses to the first. Mulliken analysis indicates that electrons preferentially occupy the 2px2p_x and 2py2p_y orbitals in the C4(A)C_4(A) SUHF1 solution, but the 2px2p_x and 2pz2p_z orbitals in SUHF2. Complex orbitals and complex-conjugation projection may help avoid the crossing of the competing C4(A)C_4(A) solutions, as suggested by earlier applications,[65] but are not considered here.

For C4(A)C_4(A) PCCSD, we select the lower-energy solution from the two branches at each bond length. PCCSD reproduces the overall shape of both states and places their crossing near 3.0 bohr, compared with approximately 3.1 bohr in iCI. Both triples corrections shift the crossing to between 3.1 and 3.2 bohr. Near this crossing, the C4(A)C_4(A) curve switches between two PCCSD branches, producing a cusp that is amplified into a discontinuity by the triples corrections. Coherent (T)(T) strongly overcorrelates in the recoupling region of the C4(A)C_4(A) state and at large bond lengths in the C4(B)C_4(B) state. Once again, +(T)+(T) gives more uniform errors, within approximately 3 mEhE_{\mathrm{h}} for both states. Both molecular examples therefore favor +(T)+(T) for a balanced description along the dissociation coordinate.

Transition-metal monoxides

The transition-metal monoxides provide a complementary test involving different spin multiplicities and spatial symmetries. We combine spin and C4C_4 projection to target the states listed in Table I. Although projection onto the C4C_4 subgroup cannot distinguish all irreducible representations of the full molecular point group, it still helps reduce spatial-symmetry contamination. We compare with the extrapolated ASCI+PT2 reference energies of Hait and co-workers.[67] The conventional benchmarks labeled UCCSD and UCCSD(T) in the table use restricted open-shell Hartree–Fock (ROHF) reference orbitals, as specified in that work.

PCCSD convergence is more difficult for FeO, CoO, and CuO: the energies change by less than 0.3 mEhE_{\mathrm{h}} over the final 20 iterations, while the largest reduced residual magnitudes remain above 10−410^{-4}. Projection improves every singles-and-doubles entry in Table I, reducing the full-set MAE from 29.9 to 15.9 mEhE_{\mathrm{h}}. PCCSD convergence remains incomplete for FeO, CoO, and CuO, and the state assignment of the NiO benchmark remains uncertain. Excluding these four species, coherent (T)(T) reduces the PCCSD MAE from 13.4 to 3.0 mEhE_{\mathrm{h}}, comparable to 2.6 mEhE_{\mathrm{h}} for UCCSD(T). Thus, the principal benefit of the triples correction is the recovery of correlation missing from PCCSD, bringing its accuracy close to that of conventional CCSD(T) for these equilibrium geometries.

In contrast to N2_2 and C2_2, +(T)+(T) tends to overcorrelate the six-species subset defined in footnote c of Table I, with CrO being the exception. Its MAE is 9.5 mEhE_{\mathrm{h}}, compared with 3.0 mEhE_{\mathrm{h}} for coherent (T)(T). The larger errors may partly arise from contributions associated with unwanted spatial-symmetry components, which are not explicitly removed in the unprojected +(T)+(T) expression. By combining intermediates with the projection weights, coherent (T)(T) may partially suppress these contributions.

Chromium dimer

Finally, we consider Cr2_2, a longstanding challenge in quantum chemistry that requires an accurate treatment of both static and dynamic correlation. Its 12 valence electrons participate in a formal sextuple bond, giving rise to strong entanglement upon dissociation, while

higher-order excitations are needed to describe correlation involving nonvalence electrons. We compare our potential energy curves with the largest-bond-dimension and extrapolated DMRG results of Larsson and co-workers.[68] Figure 3 shows total energies and binding energies. For PCC, binding energies are referenced to the energy of each method at R=50R = 50 Å, since the singlet projection does not apply separately to ground-state Cr atoms and the projection methods are not generally size-consistent. For DMRG, we use the reported atomic reference.

Figure 3. Cr2_2 dissociation in cc-pVDZ-DK with the spin-free X2C Hamiltonian. (a) Total energies shifted by 2099 EhE_{\mathrm{h}}. (b) Binding energies. DMRG(Dmax⁡D_{\max}) and DMRG(extrap.) denote the largest bond dimension in the optimization and extrapolated results, respectively.

Both PCCSD methods miss the short-bond well and instead predict a broad minimum near 2.5 Å, indicating that substantial dynamic correlation remains unrecovered at the singles-and-doubles level. Adding either triples correction restores the short-bond minimum and substantially improves agreement with DMRG. Coherent (T)(T) gives the more accurate well depth: SGCCSD+coh(T)(T) yields approximately 1.55 eV, close to

SpeciesUCCSDC4_4-SUCCSDUCCSD(T)C4_4-SUCCSD +coh(T)C4_4-SUCCSD +(T)Extrapolated ASCI+PT2
ScO (2Σ+^2\Sigma^+)-834.8648-834.8793-834.8869-834.8906-834.9058-834.8886
TiO (3Δ^3\Delta)-923.5252-923.5429-923.5530-923.5566-923.5668-923.5547
VO (4Σ−^4\Sigma^-)-1017.9835-1018.0054-1018.0169-1018.0224-1018.0291-1018.0195
CrO (5Π^5\Pi)-1118.3567-1118.3733-1118.3875-1118.3944-1118.3879-1118.3909
MnO (6Σ+^6\Sigma^+)-1224.8511-1224.8566-1224.8708-1224.8698-1224.8776-1224.8753
FeO (5Δ^5\Delta)2-1337.4430-1337.4460-1337.4663-1337.4648-1337.4710-1337.4710
CoO (4Δ^4\Delta)2-1456.3987-1456.4236-1456.4314-1456.4433-1456.4453-1456.4401
NiO (3Π^3\Pi)3-1581.8317-1581.8454-1581.8533-1581.8670-1581.8628-1581.8677
CuO (2Π^2\Pi)2-1713.9215-1713.9280-1713.9402-1713.9496-1713.9635-1713.9427
ZnO (1Σ+^1\Sigma^+)-1852.7477-1852.7631-1852.7704-1852.7742-1852.7845-1852.7718
MAE (all ten)29.915.94.63.58.8–
MAE (six species)428.613.42.63.09.5–

Table 1. Total energies (EhE_\mathrm{h}) and mean absolute errors (MAEs, mEhE_\mathrm{h}) for the 3d transition-metal monoxides in def2-SVP, relative to extrapolated ASCI+PT2. Geometries and frozen-core shells follow Ref. 67. Conventional CC totals were reconstructed from published bond dissociation energies and corresponding atomic CC energies.

extrapolated DMRG, whereas +(T)+(T) gives shallower wells of about 1.2–1.3 eV. The four corrected curves approach one another beyond 2.25 Å and reproduce the shelf region. The remaining discrepancies occur mainly at intermediate bond lengths, where the triples-corrected methods still underestimate the correlation energy and more closely follow the shape of the unextrapolated DMRG curve.

Breaking SzS_z symmetry recovers additional correlation at short bond lengths in both PCCSD and its triples-corrected variants, with the differences diminishing upon stretching. However, the N2_2 results suggest that the close agreement between SGCCSD+coh(TT) and extrapolated DMRG in the well depth may partly be due to error cancellation involving overcorrelation from the triples correction.

Discussion and Conclusion

The results show that symmetry projection and perturbative triples address complementary parts of the correlation problem. Projection improves the singles-and-doubles description and distinguishes selected low-lying symmetry sectors, while triples recover additional correlation near equilibrium and restore the short-bond well of Cr2_2. The choice of triples correction is system-dependent: +(T)+(T) is more balanced for N2_2 and C2_2, whereas coherent (T)(T) performs better for the monoxide set and the Cr2_2 well depth. A universal non-iterative triples correction would be desirable.

Both corrections recover conventional (T)(T) in the absence of the projection operator, but the coherent construction involves additional approximations. Only the scalar, single, and double disentangled operators are retained here. The influence of projection-generated W3W_3, W4W_4, and higher terms on a triples correction remains to be determined. A more systematic treatment of the projected left state and the coupling to higher excitation manifolds would help assess the approximations used in the present construction.

Gauge invariances of the projected reference and linear dependencies among projected excitations are important practical limitations.[62, 63, 64, 72] Distinct broken-symmetry parameterizations can represent the same projected mean-field state,[82] yet behave differently after truncating the disentangled cluster operator. Although the gauge-fixing procedure in Appendix A helps select the underlying reference, a satisfactory PHF state does not guarantee a stable PCCSD solution or a reliable triples correction. The reduced residual equations mitigate one aspect of the redundancy but do not remove the redundant components from the cluster amplitudes. Large or poorly determined amplitudes can therefore persist, with potentially amplified effects in +(T)+(T).

The remaining overcorrelation, branch discontinuities, and convergence sensitivity show that our triples corrections are not yet uniformly reliable. Nevertheless, they support further development of PCC as a framework for combining symmetry restoration with coupled cluster correlation.

Data Availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Acknowledgments

This work was supported by the U.S. Department of Energy, Office of Basic Energy Sciences, under Award DE-SC0001474. G.E.S. is a Welch Foundation Chair (C-0036). We thank Guo Chen and Yanbo Fang for helpful discussions of projected Hartree–Fock theory, and Ning Zhang for providing the iCI data for C2_2.

Gauge fixing in projected Hartree–Fock

Gauge freedom allows different broken-symmetry determinants to represent the same projected state, but these determinants need not be equally suitable for subsequent correlation treatments. Here, we propose a procedure for selecting a reference within this manifold. Taking SGHF as an example, we transform the optimized PHF reference according to

∣α,β,γ⟩=eαSzeiβSyeγSz∣0⟩,(24)|\alpha,\beta,\gamma\rangle= e^{\alpha S_z}e^{i\beta S_y}e^{\gamma S_z}|0\rangle, \tag*{(24)}

where α\alpha, β\beta, and γ\gamma are real parameters. The nonunitary factors explore gauge directions associated with the dual representation of the Goldstone manifold.[82] Because the SzS_z transformations are nonunitary, they can change the norm and expectation values of the unprojected determinant.

For singlet projection, this transformation leaves the projected state unchanged. For general target spin JJ, the transformation preserves the space spanned by the projected components, provided that the mixing coefficients are transformed consistently using the inverse Wigner DD matrix. In practice, these coefficients can be reoptimized by solving the generalized eigenvalue problem for the transformed determinant. The optimized PHF energy is then unchanged under exact projection.

We select the transformation parameters by minimizing

min⁡α,β,γ(⟨α,β,γ∣S2∣α,β,γ⟩⟨α,β,γ∣α,β,γ⟩−J(J+1))2,(25)\min_{\alpha,\beta,\gamma}\left(\frac{\langle\alpha,\beta,\gamma|S^2|\alpha,\beta,\gamma\rangle}{\langle\alpha,\beta,\gamma|\alpha,\beta,\gamma\rangle}-J(J+1)\right)^2 , \tag*{(25)}

where JJ is the target spin quantum number. This criterion brings the mean value of S2S^2 closer to its target value without changing the optimized projected energy. We also apply an analogous gauge-fixing step to bring the normalized expectation value of SzS_z closer to its target value. These steps do not generally enforce spin purity, but are intended to select a more suitable broken-symmetry reference for subsequent PCC calculations. An analogous gauge-fixing procedure for SUHF references or discrete-symmetry projection has not been developed here.

References

  1. [1]J. Čížek, On the correlation problem in atomic and molecular systems, calculation of wavefunction components in Ursell-type expansion using quantum-field theoretical methods, The Journal of Chemical Physics 45, 4256 (1966).
  2. [2]R. J. Bartlett and M. Musial, Coupled-cluster theory in quantum chemistry, Reviews of Modern Physics 79, 291 (2007).DOI
  3. [3]J. Paldus and X. Li, A critical assessment of coupled cluster method in quantum chemistry, in Advances in Chemical Physics, Vol. 110, edited by I. Prigogine and S. A. Rice (John Wiley & Sons, 1999) pp. 1–175.DOI
  4. [4]G. D. Purvis III and R. J. Bartlett, A full coupled-cluster singles and doubles model: The inclusion of disconnected triples, The Journal of Chemical Physics 76, 1910 (1982).DOI
  5. [5]K. Raghavachari, G. W. Trucks, J. A. Pople, and M. Head-Gordon, A fifth-order perturbation comparison of electron correlation theories, Chemical Physics Letters 157, 479 (1989).DOI
  6. [6]J. D. Watts, J. Gauss, and R. J. Bartlett, Coupled-cluster methods with noniterative triple excitations for restricted open-shell Hartree–Fock and other general single determinant reference functions. energies and analytical gradients, The Journal of Chemical Physics 98, 8718 (1993).DOI
  7. [7]G. E. Scuseria, Analytic evaluation of energy gradients for the singles and doubles coupled cluster method including perturbative triple excitations: Theory and applications to FOOF and Cr₂, The Journal of Chemical Physics 94, 442 (1991).DOI
  8. [8]C. Riplinger and F. Neese, An efficient and near linear scaling pair natural orbital based local coupled cluster method, The Journal of Chemical Physics 138, 034106 (2013).
  9. [9]C. Riplinger, B. Sandhoefer, A. Hansen, and F. Neese, Natural triple excitations in local coupled cluster calculations with pair natural orbitals, The Journal of Chemical Physics 139, 134101 (2013).DOI
  10. [10]S. Li, J. Ma, and Y. Jiang, Linear scaling local correlation approach for solving the coupled cluster equations of large systems, Journal of Computational Chemistry 23, 237 (2002).pubmed.ncbi.nlm.nih.gov/11924736
  11. [11]S. Li, J. Shen, W. Li, and Y. Jiang, An efficient implementation of the “cluster-in-molecule” approach for local electron correlation calculations, The Journal of Chemical Physics 125, 074109 (2006).pubmed.ncbi.nlm.nih.gov/16942324
  12. [12]Y. Guo, C. Riplinger, U. Becker, D. G. Liakos, Y. Minenkov, L. Cavallo, and F. Neese, Communication: An improved linear scaling perturbative triples correction for the domain based local pair-natural orbital based singles and doubles coupled cluster method [DLPNO-C(CSD(T)], The Journal of Chemical Physics 148, 011101 (2018).DOI
  13. [13]P. R. Nagy and M. Kállay, Optimization of the linear-scaling local natural orbital CCSD(T) method: Redundancy-free triples correction using Laplace transform, The Journal of Chemical Physics 146, 214106 (2017).
  14. [14]P. R. Nagy and M. Kállay, Approaching the basis set limit of CCSD(T) energies for large molecules with lo- cal natural orbital coupled-cluster methods, Journal of Chemical Theory and Computation 15, 5275 (2019).DOI
  15. [15]cal natural orbital coupled-cluster methods, Journal of Chemical Theory and Computation 15, 5275 (2019). [15] L. Gyévi-Nagy, M. Kállay, and P. R. Nagy, Accurate reduced-cost ccsd(t) energies: Parallel implementation, benchmarks, and large-scale applications, Journal of Chemical Theory and Computation 17, 860 (2021).DOI
  16. [16]R. Song, X. Gong, A. Bakry, and H.-Z. Ye, Random phase approximation-based local natural orbital coupled cluster theory, The Journal of Chemical Physics 164, 214114 (2026).arxiv.org/abs/2601.00131
  17. [17]H.-Z. Ye and T. C. Berkelbach, Periodic local coupled-cluster theory for insulators and metals, Journal of Chemical Theory and Computation 20, 8948 (2024).DOI
  18. [18]J. Yang, N. Zhang, S. Yuan, J. Yu, H.-Z. Ye, and G. K.-L. Chan, Ab initio many-body quantum embedding and local correlation in crystalline materials using interpolative separable density fitting, The Journal of Chemical Physics 164, 154104 (2026).
  19. [19]R. Schutski, J. Zhao, T. M. Henderson, and G. E. Scuseria, Tensor-structured coupled cluster theory, The Journal of Chemical Physics 147, 184113 (2017).DOI
  20. [20]R. M. Parrish, Y. Zhao, E. G. Hohenstein, and T. J. Martínez, Rank reduced coupled cluster theory. I. ground state energies and wavefunctions, The Journal of Chemical Physics 150, 164118 (2019).
  21. [21]M. Lesiuk, Quintic-scaling rank-reduced coupled cluster theory with single and double excitations, The Journal of Chemical Physics 156, 064103 (2022).
  22. [22]E. G. Hohenstein, B. S. Fales, R. M. Parrish, and T. J. Martínez, Rank-reduced coupled-cluster. III. tensor hypercontraction of the doubles amplitudes, The Journal of Chemical Physics 156, 054102 (2022).
  23. [23]M. Lesiuk, Implementation of the coupled-cluster method with single, double, and triple excitations using tensor decompositions, Journal of Chemical Theory and Computation 16, 453 (2020).
  24. [24]A. Jiang, J. M. Turney, and H. F. Schaefer III, Tensor hypercontraction form of the perturbative triples energy in coupled-cluster theory, Journal of Chemical Theory and Computation 19, 1476 (2023).
  25. [25]B. O. Roos, P. R. Taylor, and P. E. M. Siegbahn, A complete active space SCF (CASSCF) using a density matrix formulated super-CI approach, Chemical Physics 48, 157 (1980).
  26. [26]P. E. M. Siegbahn, J. Almlöf, A. Heiberg, and B. O. Roos, The complete active space SCF (CASSCF) method in a Newton–Raphson formulation with application to the HNO molecule, The Journal of Chemical Physics 74, 2384 (1981).
  27. [27]K. Andersson, P. A. Malmqvist, B. O. Roos, A. J. Sadlej, and K. Wolinski, Second-order perturbation theory with a CASSCF reference function, The Journal of Physical Chemistry 94, 5483 (1990).
  28. [28]K. Andersson, P.-Å. Malmqvist, and B. O. Roos, Second-order perturbation theory with a complete active space self-consistent field reference function, The Journal of Chemical Physics 96, 1218 (1992).
  29. [29]C. Angeli, R. Cimiraglia, S. Evangelisti, T. Leininger, and J.-P. Malrieu, Introduction of n-electron valence states for multireference perturbation theory, The Journal of Chemical Physics 114, 10252 (2001).
  30. [30]C. Angeli, R. Cimiraglia, and J.-P. Malrieu, N-electron valence state perturbation theory: A fast implementation of the strongly contracted variant, Chemical Physics Letters 350, 297 (2001).
  31. [31]C. Angeli, R. Cimiraglia, and J.-P. Malrieu, N-electron valence state perturbation theory: A spinless formulation and an efficient implementation of the strongly contracted and of the partially contracted variants, The Journal of Chemical Physics 117, 9138 (2002).
  32. [32]B. Jeziorski and H. J. Monkhorst, Coupled-cluster method for multideterminantal reference states, Physical Review A 24, 1668 (1981).
  33. [33]D. I. Lyakh, M. Musial, V. F. Lotrich, and R. J. Bartlett, Multireference nature of chemistry: The coupled-cluster view, Chemical Reviews 112, 182 (2012).DOI
  34. [34]F. A. Evangelista, E. Prochnow, J. Gauss, and H. F. Schaefer III, Perturbative triples corrections in state-specific multireference coupled cluster theory, The Journal of Chemical Physics 132, 074107 (2010).
  35. [35]F. A. Evangelista and J. Gauss, An orbital-invariant internally contracted multireference coupled cluster approach, The Journal of Chemical Physics 134, 114102 (2011).
  36. [36]F. A. Evangelista, M. Hanauer, A. Köhn, and J. Gauss, A sequential transformation approach to the internally contracted multireference coupled cluster method, The Journal of Chemical Physics 136, 204108 (2012).
  37. [37]F. A. Evangelista, Perspective: Multireference coupled cluster theories of dynamical electron correlation, The Journal of Chemical Physics 149, 030901 (2018).
  38. [38]M. Hanauer and A. Köhn, Pilot applications of internally contracted multireference coupled cluster theory, and how to choose the cluster operator properly, The Journal of Chemical Physics 134, 204111 (2011).
  39. [39]M. Hanauer and A. Köhn, Perturbative treatment of triple excitations in internally contracted multireference coupled cluster theory, The Journal of Chemical Physics 136, 204107 (2012).
  40. [40]C. A. Jiménez-Hoyos, T. M. Henderson, and G. E. Scuseria, Generalized Hartree–Fock description of molecular dissociation, Journal of Chemical Theory and Computation 7, 2667 (2011).
  41. [41]P.-O. Löwdin, Quantum theory of many-particle systems. III. extension of the Hartree-Fock scheme to include degenerate systems and correlation effects, Physical Review 97, 1509 (1955).
  42. [42]P. Lykos and G. W. Pratt, Discussion on the Hartree–Fock approximation, Reviews of Modern Physics 35, 496 (1963).DOI
  43. [43]I. Mayer, The spin-projected extended Hartree–Fock method, Advances in Quantum Chemistry 12, 189 (1980).
  44. [44]K. W. Schmid, T. Dahm, J. Margueron, and H. Müther, Symmetry-projected variational approach to the one-dimensional Hubbard model, Physical Review B 72, 085116 (2005).
  45. [45]C. A. Jiménez-Hoyos, T. M. Henderson, T. Tsuchimochi, and G. E. Scuseria, Projected Hartree–Fock theory, The Journal of Chemical Physics 136, 164109 (2012).DOI
  46. [46]K. W. Schmid, On the use of general symmetry-projected Hartree–Fock–Bogoliubov configurations in variational approaches to the nuclear many-body problem, Progress in Particle and Nuclear Physics 52, 565 (2004).
  47. [47]G. E. Scuseria, C. A. Jiménez-Hoyos, T. M. Henderson, K. Samanta, and J. K. Ellis, Projected quasiparticle theory for molecular electronic structure, The Journal of Chemical Physics 135, 124108 (2011).DOI
  48. [48]T. Tsuchimochi and S. Ten-no, Communication: Configuration interaction combined with spin-projection for strongly correlated molecular electronic structures, The Journal of Chemical Physics 144, 011101 (2016).DOI
  49. [49]T. Tsuchimochi and S. Ten-no, Bridging single- and multireference domains for electron correlation: Spin-extended coupled electron pair approximation, Journal of Chemical Theory and Computation 13, 1667 (2017).arxiv.org/abs/1612.02945
  50. [50]T. Tsuchimochi and S. L. Ten-no, Second-order perturbation theory with spin-symmetry-projected Hartree–Fock, Journal of Chemical Theory and Computation 15, 6688 (2019).DOI
  51. [51]T. Tsuchimochi, K. Yoshimura, Y. Shimomoto, and S. L. Ten-no, Improved description and efficient implementation of spin-projected perturbation theory for practical applications, Journal of Chemical Theory and Computation 17, 3471 (2021).DOI
  52. [52]T. M. Henderson and G. E. Scuseria, Geminal-based configuration interaction, The Journal of Chemical Physics 151, 051101 (2019).DOI
  53. [53]T. M. Henderson and G. E. Scuseria, Correlating the antisymmetrized geminal power wave function, The Journal of Chemical Physics 153, 084111 (2020).DOI
  54. [54]R. Dutta, T. M. Henderson, and G. E. Scuseria, Geminal replacement models based on AGP, Journal of Chemical Theory and Computation 16, 6358 (2020).DOI
  55. [55]A. Khamoshi, G. P. Chen, T. M. Henderson, and G. E. Scuseria, Exploring non-linear correlators on AGP, The Journal of Chemical Physics 154, 074113 (2021).DOI
  56. [56]Y. Qiu, T. M. Henderson, and G. E. Scuseria, Projected Hartree-Fock theory as a polynomial of particle-hole excitations and its combination with variational coupled cluster theory, The Journal of Chemical Physics 146, 184105 (2017).DOI
  57. [57]T. M. Henderson and G. E. Scuseria, Spin-projected generalized Hartree-Fock method as a polynomial of particle-hole excitations, Physical Review A 96, 022506 (2017).DOI
  58. [58]T. Duguet, Symmetry broken and restored coupled-cluster theory: I. rotational symmetry and angular momentum, Journal of Physics G: Nuclear and Particle Physics 42, 025107 (2015).DOI
  59. [59]T. Duguet and A. Signoracci, Symmetry broken and restored coupled-cluster theory: II. global gauge symmetry and particle number, Journal of Physics G: Nuclear and Particle Physics 44, 015103 (2017).arxiv.org/abs/1512.02878
  60. [60]Y. Qiu, T. M. Henderson, J. Zhao, and G. E. Scuseria, Projected coupled cluster theory, The Journal of Chemical Physics 147, 064111 (2017).
  61. [61]Y. Qiu, T. M. Henderson, T. Duguet, and G. E. Scuseria, Particle-number projected Bogoliubov-coupled-cluster theory: Application to the pairing Hamiltonian, Physical Review C 99, 044301 (2019).DOI
  62. [62]Y. Qiu, T. M. Henderson, J. Zhao, and G. E. Scuseria, Projected coupled cluster theory: Optimization of cluster amplitudes in the presence of symmetry projection, The Journal of Chemical Physics 149, 164108 (2018).DOI
  63. [63]T. Tsuchimochi and S. L. Ten-no, Orbital-invariant spin-extended approximate coupled-cluster for multi-reference systems, The Journal of Chemical Physics 149, 044109 (2018).DOI
  64. [64]R. Song, T. M. Henderson, and G. E. Scuseria, A power series approximation in symmetry projected coupled cluster theory, The Journal of Chemical Physics 156, 104105 (2022).DOI
  65. [65]R. Song, T. M. Henderson, and G. E. Scuseria, Combining complex conjugation, time-reversal, and spin-flip symmetry projection of coupled cluster wave functions, The Journal of Physical Chemistry A 128, 6593 (2024).
  66. [66]I. Magoulas, J. Shen, and P. Piecuch, Addressing strong correlation by approximate coupled-pair methods with active-space and full treatments of three-body clusters, Molecular Physics 120, e2057365 (2022).DOI
  67. [67]D. Hait, N. M. Tubman, D. S. Levine, K. B. Whaley, and M. Head-Gordon, What levels of coupled cluster theory are appropriate for transition metal systems? a study using near-exact quantum chemical values for 3d transition metal binary compounds, Journal of Chemical Theory and Computation 15, 5370 (2019).DOI
  68. [68]H. R. Larsson, H. Zhai, C. J. Umrigar, and G. K.-L. Chan, The chromium dimer: Closing a chapter of quantum chemistry, Journal of the American Chemical Society 144, 15932 (2022).DOI
  69. [69]O. W. Day, D. W. Smith, and C. Garrod, A generalization of the Hartree-Fock one-particle potential, International Journal of Quantum Chemistry 8, 501 (1974).
  70. [70]D. W. Smith and O. W. Day, Extension of Koopmans’ theorem. I. derivation, The Journal of Chemical Physics 62, 113 (1975).DOI
  71. [71]D. J. Thouless, Stability conditions and nuclear rotations in the Hartree-Fock theory, Nuclear Physics 21, 225 (1960).DOI
  72. [72]T. Tsuchimochi and S. L. Ten-no, Extending spin-symmetry projected coupled-cluster to large model spaces using an iterative null-space projection technique, Journal of Computational Chemistry 40, 265 (2019).DOI
  73. [73]G. E. Scuseria, An alternative expression for the perturbative triples correction to Brueckner coupled-cluster doubles theory, Chemical Physics Letters 226, 251 (1994).DOI
  74. [74]R. Song, PHF_PCC: Projected Hartree–Fock and projected coupled-cluster code, GitHub repository.
  75. [75]Q. Sun, X. Zhang, S. Banerjee, P. Bao, M. Barbry, N. S. Blunt, N. A. Bogdanov, G. H. Booth, J. Chen, Z.-H. Cui, J. J. Eriksen, Y. Gao, S. Guo, J. Hermann, M. R. Hermes, K. Koh, P. Koval, S. Lehtola, Z. Li, J. Liu, N. Mardirossian, J. D. McClain, M. Motta, B. Mussard, H. Q. Pham, A. Pulkin, W. Purwanto, P. J. Robinson, E. Ronca, E. R. Sayfutyarova, M. Scheurer, H. F. Schurkus, J. E. T. Smith, C. Sun, S.-N. Sun, S. Upadhyay, L. K. Wagner, X. Wang, A. White, J. D. Whitfield, M. J. Williamson, S. Wouters, J. Yang, J. M. Yu, T. Zhu, T. C. Berkelbach, S. Sharma, A. Y. Sokolov, and G. K.-L. Chan, Recent developments in the PySCF program package, The Journal of Chemical Physics 153, 024109 (2020).DOI
  76. [76]Q. Sun, Libcint: An efficient general integral library for Gaussian basis functions, Journal of Computational Chemistry 36, 1664 (2015).arxiv.org/abs/1412.0649
  77. [77]P. J. Lestrange, D. B. Williams-Young, A. Petrone, C. A. Jiménez-Hoyos, and X. Li, Efficient implementation of variation after projection generalized Hartree–Fock, Journal of Chemical Theory and Computation 14, 588 (2018).DOI
  78. [78]D. Peng and M. Reiher, Exact decoupling of the relativistic Fock operator, Theoretical Chemistry Accounts 131, 1081 (2012).
  79. [79]W. Kutzelnigg and W. Liu, Quasirelativistic theory equivalent to fully relativistic theory, The Journal of Chemical Physics 123, 241102 (2005).
  80. [80]Z. Wang, Y. Li, and J. Lu, Coordinate descent full configuration interaction, Journal of Chemical Theory and Computation 15, 3558 (2019).DOI
  81. [81]N. Zhang, W. Liu, and M. R. Hoffmann, Iterative configuration interaction with selection, Journal of Chemical
  82. [82]C. A. Jiménez-Hoyos, R. R. Rodríguez-Guzmán, T. M. Henderson, and G. E. Scuseria, On a dual representation of the Goldstone manifold (2020), arXiv:2004.05047.

Paper details

vlnc:W_79f73utz
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents