Skip to main content
NCBI home page
ACS AuthorChoice logoLink to ACS AuthorChoice
. 2022 Feb 2;18(3):1489–1500. doi: 10.1021/acs.jctc.1c01000

Efficient Implementation of Equation-of-Motion Coupled-Cluster Singles and Doubles Method with the Density-Fitting Approximation: An Enhanced Algorithm for the Particle–Particle Ladder Term

Aslı Ünal †,‡, Uğur Bozkaya ‡,*
PMCID: PMC8908769  PMID: 35107297

Abstract

graphic file with name ct1c01000_0006.jpg

An efficient implementation of the density-fitted equation-of-motion coupled-cluster singles and doubles (DF-EOM-CCSD) method is presented with an enhanced algorithm for the particle–particle ladder (PPL) term, which is the most expensive part of EOM-CCSD computations. To further improve the evaluation of the PPL term, a hybrid density-fitting/Cholesky decomposition (DF/CD) algorithm is also introduced. In the hybrid DF/CD approach, four virtual index integrals are constructed on-the-fly from the DF factors; then, their partial Cholesky decomposition is simultaneously performed. The computational cost of the DF-EOM-CCSD method for excitation energies is compared with that of the resolution of the identity EOM-CCSD (RI-EOM-CCSD) (from the Q-chem 5.3 package). Our results demonstrate that DF-EOM-CCSD excitation energies are significantly accelerated compared to RI-EOM-CCSD. There is more than a 2-fold reduction for the C8H18 molecule in the cc-pVTZ basis set with the restricted Hartree-Fock (RHF) reference. This cost savings results from the efficient evaluation of the PPL term. In the RHF based DF-EOM-CCSD method, the number of flops (NOF) is 1/4O2V4, while that of RI-EOM-CCSD was reported (Epifanovsky et al. J. Chem. Phys. 2013, 139, 134105.) to be 5/8O2V4 for the PPL contraction term. Further, the NOF of VVVV-type integral transformation is 1/2V4Naux in our case, while it appears to be V4Naux for RI-EOM-CCSD. Hence, our implementation is 2.5 and 2.0 times more efficient compared to RI-EOM-CCSD for these expensive terms. For the unrestricted Hartree-Fock (UHF) reference, our implementation maintains its enhanced performance and provides a 1.8-fold reduction in the computational time compared to RI-EOM-CCSD for the C7H16 molecule. Our results indicate that our DF-EOM-CCSD implementation is 1.7 and 1.4 times more efficient compared with RI-EOM-CCSD for average computational cost per EOM-CCSD iteration. Moreover, our results show that the new hybrid DF/CD approach improves upon the DF algorithm, especially for large molecular systems. Overall, we conclude that the new hybrid DF/CD PPL algorithm is very promising for large-sized chemical systems.

1. Introduction

It is well-known that coupled-cluster (CC) methods provide accurate results for molecular properties for most chemical systems near equilibrium geometries.1−13 For example, the coupled-cluster singles and doubles (CCSD) method14 provides quite accurate results for most molecular systems at equilibrium geometries. The addition of a perturbative triples excitations correction [CCSD(T)]10,11,15 further enhances CCSD and yields very accurate results for a broad range of molecular systems.12,16−25 However, high computational costs of common CC methods, such as O(N6) and O(N7) for CCSD and CCSD(T) (where N is the number of basis functions), limits their applications to relatively small-sized chemical systems.

Accurate computations of excitation energies (EEs) is one of the most challenging problems in modern quantum chemistry. Equation-of-motion CC (EOM-CC) methods provide accurate results for excited-state properties for a broad range of chemical systems.26−44 The accuracy of the EOM approach based on the CCSD model (EOM-CCSD) has been reported to be 0.1–0.2 eV.28,31 However, as in the case of the ground-state CC methods, the computational cost and disk/memory requirements for the EOM-CC methods scale steeply with the system size.

Tensor decomposition techniques for electron repulsion integrals (ERIs) have been of significant interest in modern computational chemistry.45−70 Density fitting (DF) is one of the most popular ERI decomposition techniques.45−52,59−70 In the DF approach, the ERI tensor of rank-4 is expanded in terms of rank-3 tensors. Another common ERI factorization approach is the partial Cholesky decomposition (CD).55−60,63,64 The DF and CD techniques are very useful to reduce the cost of integral transformations and the storage requirements for the ERI tensor.

In this research, a new implementation of the density-fitted EOM-CCSD method is presented with an enhanced algorithm for the particle–particle ladder (PPL) term, which is the most expensive term. The equations presented have been implemented in a new computer code by the present authors and added to the MacroQC package.71 The computational time of our DF-EOM-CCSD implementation is compared with that of the Q-chem 5.3 software.72 The DF-EOM-CCSD method is applied to a test set for excitation energies.

2. CCSD Energy and Amplitude Equations

At first, we would like to note that all equations reported in this study are in the spin–orbital formalism. The spin-free version of our equations for the restricted closed-shell systems are provided in the Supporting Information. The unrestricted version of the formulas can be readily obtained from the spin–orbital equations.

The correlation energy for the CCSD method can be expressed as follows

2. 1

where ĤN is the normal-ordered Hamiltonian operator,4,73 |0⟩ is the Hartree–Fock (HF) determinant, and T̂ is the sum of single- and double-excitation operators T̂ = T̂1 + T̂2:

2. 2
2. 3

where ↠and î are the creation and annihilation operators and tia and tij are the single and double excitation amplitudes, respectively.

tia and tij can be obtained from the following equations:

2. 4
2. 5

where Φia and Φij are singly and doubly excited Slater determinants, respectively. For explicit equations of our CCSD implementation, one may refer to our previous studies.65−67

3. The EOM-CCSD Model

In the EOM-CCSD framework, the target excited-state wave functions are written as follows:

3. 6
3. 7

where R̂ and L̂ are linear excitation and de-excitation operators, respectively. For EOM-CCSD, R̂ = R̂1 + R̂2:

3. 8
3. 9

where ria and rij are the single and double excitation amplitudes, respectively, and the notation {â†...î} denotes a string of normal-ordered operators with respect to the Fermi vacuum.

For the ground state, we have the following Schrödinger equation:

3. 10

Hence, by multiplying eq 10 by e–T̂, we obtain

3. 11

where H̅ = e–T̂ĤeT̂.

Further, the normal ordered H̅ can be written as follows:

3. 12

Hence, we define:

3. 13

where subscript c means that only connected diagrams should be included. Therefore, we may rewrite eq 11 as follows

3. 14

where ΔE is the ground-state CC correlation energy.

The excited-state eigenvalue equation can be written as follows:

3. 15

where ΔER is the excited-state CC correlation energy. The excitation energy can be written as

3. 16

After performing some algebra, we obtain the EOM-CCSD equation as follows:

3. 17

Equation 17 is equivalent to the following matrix eigenvalue equation for CCSD:

3. 18

However, we solve eq 18 iteratively with the Davidson algorithm.74−77 Hence, we need to introduce the so-called σ vector as follows:

3. 19

where I, J = 0, S, D.

3.1. DF-EOM-CCSD Intermediates

The DF-CCSD intermediates that appear in the DF-EOM-CCSD equations are given in Appendix A.

3.1.1. DF-EOM-CCSD 3-Index Intermediates

1- and 3-index intermediates that were used for EOM-CCSD are defined as follows:

3.1.1. 20
3.1.1. 21
3.1.1. 22
3.1.1. 23
3.1.1. 24
3.1.1. 25
3.1.1. 26
3.1.1. 27
3.1.1. 28

where Q runs over auxiliary basis functions and the bpqQ terms are the molecular orbital (MO) basis DF factors, which are defined in our previous studies.65−67

3.1.2. 4-Index Intermediates

4-index intermediates are defined as follows:

3.1.2. 29
3.1.2. 30
3.1.2. 31
3.1.2. 32
3.1.2. 33

3.1.3. 2-Index Intermediates

2-index intermediates are defined as follows:

3.1.3. 34
3.1.3. 35
3.1.3. 36

3.2. DF-EOM-CCSD σ Equations

The DF-EOM-CCSD σ0 equation can be written as

3.2. 37

With the DF approximation, the EOM-CCSD σia equation can be written as

3.2. 38

With the DF approximation, the EOM-CCSD σijab equation can be written as

3.2. 39

3.3. PPL Algorithm with the DF Approach

The most expensive terms of the T2 and σ2 amplitude equations are the PPL terms. Our PPL algorithm for the σ2 tensor originated from the PPL algorithm used for the T2 amplitude equation in our 2016 study.65 Here, we employ the same algorithm to σ2 amplitudes. For example, for the closed-shell case, the PPL term can be written as

3.3. 40

Following the previous studies of Saebø and Pulay78 and Scuseria at al.79 and our previous DF-CCSD studies,65,66 we employ the following algorithm for the evaluation of σ-PPL:

3.3. 41

where S is the symmetric component, while A is the antisymmetric component. Now let us define

3.3. 42
3.3. 43
3.3. 44
3.3. 45
3.3. 46

where S and A have the following symmetry properties.

3.3. 47
3.3. 48

Hence, we can always keep i ≥ j and a ≥ b.

3.3. 49
3.3. 50

The pseudo code for the σ-PPL algorithm is

3.3. 51

With this algorithm, the cost of PPL is 1/4O2V4 + 1/2V4Naux + OV3Naux + OV4 + 1/4V4. The most expensive term is 1/2V4Naux. Finally, we note that our DF-EOM-CCSD code has a shared-memory parallelism feature through threaded BLAS calls as well as OpenMP parallelization of all tensor manipulations.

3.3.1. DF/CD Hybrid PPL Algorithm

In the common CD approach for ERIs, the CD factors are generated from the AO basis ERI tensor (μν|λσ), and the number of CD factors is generally higher than that of DF factors. Hence, it does not seem to speed up our DF algorithm. However, we have observed that, for the large molecules, the CD technique can be beneficial to take advantage of the sparsity of the ERI tensor if it is applied to the MO basis ERIs generated from the DF integrals. More specifically, if we perform the Cholesky decomposition of the (ab|cd)-type integrals, we may get a reduced number of auxiliary basis functions, which is especially true for large molecular systems. Hence, in our DF/CD hybrid approach, we build the (ab|cd)-type integrals from the DF factors, on-the-fly, and perform Cholesky decomposition simultaneously.

4. Results and Discussion

Results from the DF-EOM-CCSD and RI-EOM-CCSD80 methods were obtained for a set of alkanes to compare the computational cost for the excitation energy computations. For the alkanes set, Dunning’s correlation-consistent polarized valence triple-ζ basis set (aug-cc-pVTZ) was used with the frozen core approximation.81,82 For the aug-cc-pVTZ basis sets, aug-cc-pVTZ-JKFIT50 and aug-cc-pVTZ-RI83 auxiliary basis set pairs were employed for reference and correlation energies, respectively. Additionally, the DF-EOM-CCSD, resolution of the identity EOM-CCSD (RI-EOM-CCSD),80 and EOM-CCSD(fT)84 methods were applied to a set of molecules to compare the excitation energies.

4.1. Efficiency of DF-EOM-CCSD

A set of alkanes is considered to assess the efficiency of the RI-EOM-CCSD and DF-EOM-CCSD methods. The RI-EOM-CCSD computations were performed with the Q-chem 5.3 package.72 The computational time for the RI-EOM-CCSD and DF-EOM-CCSD methods are presented graphically in Figures 1 and 2 for restricted and unrestricted Hartree-Fock (RHF and UHF) references, respectively. Timing computations were carried out for a single root with a 10–7 energy and 10–7 EOM eigenvalue convergence tolerances on a single node (1 core) Intel(R) Xeon(R) CPU E5-2620 v4 @ 2.10 GHz computer (memory ∼ 64 GB). For the RI-CCSD code of Q-chem 5.3, MEM_TOTAL 64000, MEM_STATIC 2000, and CC_MEMORY 51200 options are used. We start our assessment with the RHF versions of the RI-EOM-CCSD and DF-EOM-CCSD implementations. The DF-EOM-CCSD method significantly reduces the computational cost compared to RI-EOM-CCSD, and there is more than a 2-fold reduction in the computational time for DF-EOM-CCSD for the largest member (C8H18) of the alkanes set. For the C8H18 molecule, the CCSD times are 750 and 471 min for RI-EOM-CCSD and DF-EOM-CCSD, respectively; there is a 1.6-fold reduction in the computational time for DF-EOM-CCSD. Further, for the C8H18 molecule, the EOM times are 1351 min (RI-EOM-CCSD) and 537 min (DF-EOM-CCSD); hence, there is a 2.5-fold reduction in the computational time for DF-EOM-CCSD.

Figure 1.

Figure 1

Open in a new tab

Total, CCSD, and EOM wall times (in min) for computations of excitation energies for the CnH2n+2 (n = 1–8) set from the RI-EOM-CCSD (from Q-Chem(72)) and DF-EOM-CCSD methods with the cc-pVTZ basis set. The RHF reference is used for these computations. All computations were performed for a single root with 10–7 energy and EOM eigenvalue convergence tolerances on a single node (1 core) Intel(R) Xeon(R) CPU E5-2620 v4 @ 2.10 GHz computer (memory ∼ 64 GB).

Figure 2.

Figure 2

Open in a new tab

Total, CCSD, and EOM wall time (in min) for computations of excitation energies for the CnH2n+2 (n = 1–7) set from the RI-EOM-CCSD (from Q-Chem(72)) and DF-EOM-CCSD methods with the cc-pVTZ basis set. The UHF reference is used for these computations. All computations were performed for a single root with 10–7 energy and EOM eigenvalue convergence tolerances on a single node (1 core) Intel(R) Xeon(R) CPU E5-2620 v4 @ 2.10 GHz computer (memory ∼ 64 GB).

The number of iterations for the CCSD part are 11 (DF-EOM-CCSD) and 12 (RI-EOM-CCSD). The average computational time per CCSD iteration (tccsd/niter) for C8H18 is 42.8 min (DF-EOM-CCSD) and 62.5 min (RI-EOM-CCSD). Hence, our new DF-EOM-CCSD implementation is 1.5 times faster than the RI-EOM-CCSD code for average computational cost per CCSD iteration. Similarly, the number of Davidson iterations for the EOM part is 12 (DF-EOM-CCSD) and 18 (RI-EOM-CCSD). The average computational time per a Davidson iteration (teom/niter) for C8H18 is 44.8 min (DF-EOM-CCSD) and 75.1 min (RI-EOM-CCSD). Hence, our new DF-EOM-CCSD implementation is 1.7 times faster than the RI-EOM-CCSD code for average computational cost per EOM-CCSD iteration.

The efficiency of our DF-EOM-CCSD method compared to that of RI-EOM-CCSD is attributed to the our more efficient PPL algorithm. For the closed-shell case, the number of flops (NOF) for our DF-CCSD method65,66 is 1/4O2V4 + 2O3V3 + 1/4O4V2, while that of RI-CCSD80 was reported to be 5/8O2V4 + 4O3V3 + 27/8O4V2. When one compares the cost of implementation, our DF-CCSD implementation65,66 is 2.5 times more efficient than that of RI-CCSD80 for the PPL contraction term. Further, our implementation is 2 times more efficient compared to that of RI-CCSD for the particle-hole ladder (PHL)terms. Moreover, the cost of VVVV-type integral transformation, on-the-fly of course, is 1/2V4Naux in our case, while it appears to be V4Naux for RI-CCSD.80 In fact, the most expensive term is this integral transformation step for large-scale computations with optimized auxiliary basis sets. Hence, our algorithm appears to be 2 times more efficient for this term. Basically, we follow the same algorithm for the PPL term of EOM; the same is also true for RI-EOM-CCSD. Hence, the efficiency of our DF-EOM-CCSD implementation was maintained.

As the second step of our timing assessment, we consider the UHF versions of the RI-EOM-CCSD and DF-EOM-CCSD implementations. The DF-EOM-CCSD method noticeably reduces the computational cost compared with RI-EOM-CCSD (Figure 2); there is a 1.8-fold reduction in the computational time for DF-EOM-CCSD for the C7H16 molecule. For the C7H16 molecule, the CCSD time is 1168 and 920 min for RI-EOM-CCSD and DF-EOM-CCSD, respectively; there is a 1.3-fold reduction in the computational time for DF-EOM-CCSD. Further, for the C7H16 molecule, the EOM time is 3221 min (RI-EOM-CCSD) and 1568 min (DF-EOM-CCSD); hence, there is a 2.1-fold reduction in the computational time for DF-EOM-CCSD. The number of iterations for the UHF-CCSD part are 11 (DF-EOM-CCSD) and 12 (RI-EOM-CCSD) for the C7H16 molecule. The average computational time per UHF-CCSD iteration (tccsd/niter) for C7H16 is 83.7 min (DF-EOM-CCSD) and 97.3 min (RI-EOM-CCSD). Hence, our new DF-EOM-CCSD implementation is 1.2 times faster than the RI-EOM-CCSD code for the average computational cost per UHF-CCSD iteration. Similarly, the number of Davidson iterations for the EOM part are 12 (DF-EOM-CCSD) and 18 (RI-EOM-CCSD). The average computational time per Davidson iteration (teom/niter) for C7H16 is 130.7 min (DF-EOM-CCSD) and 178.9 min (RI-EOM-CCSD). Hence, our new DF-EOM-CCSD implementation is 1.4 times faster than the RI-EOM-CCSD code for the average computational cost per EOM-CCSD iteration. Hence, our new DF-EOM-CCSD implementation maintains its efficiency for the UHF reference.

4.1.1. Assessment of the DF/CD Hybrid PPL Algorithm

As the final step of our assessment for the efficiency of our new implementations, we present benchmark timing results for comparisons of DF and DF/CD hybrid approaches for the evaluation of the PPL terms of the CCSD and EOM parts. The ratios of the number of auxiliary basis functions employed in the PPL term of EOM-CCSD for the DF and DF/CD approaches are presented graphically in Figure 3. Since the most expensive term of the PPL algorithm scales linearly with the number of auxiliary basis functions, lets call it M, the reduction of M may yield significant improvements in the evaluation of the PPL term. For example, for the C9H20 molecule with the cc-pVTZ primary basis set, the M values are 1329 and 1208 for our canonical DF and hybrid DF/CD algorithms, respectively. Hence, the ratio of MDF/MDF/CD is 1.10, which indicates a more than 10% reduction in the number of auxiliary basis functions. For the alkanes set considered, CnH2n+2 (n = 1–9), we plot the MDF/MDF/CD values with respect to the n values and obtain a linear relation for this fit. The equation and the R2 value for the linear fit are MDF/MDF/CD = 0.0122n + 0.9883 and R2 = 0.9939. At first, one should note that as the n value increases the MDF/MDF/CD ratio increases as well. The reason for this correlation is that as molecular size increases the hybrid DF/CD algorithm makes better use of the sparsity of the ERI tensor. Hence, the obtained linear equation indicates that, if we proceed to larger molecules, the hybrid DF/CD algorithm will have a larger impact on the computational time. For example, the MDF/MDF/CD ratio will be approximately 1.23 and 1.60 for the C20H42 and C50H102 molecules, which indicates up to 23% and 60% acceleration in the PPL terms can be achieved.

Figure 3.

Figure 3

Open in a new tab

Ratio of the number of auxiliary basis functions, M, employed in the PPL term of DF-EOM-CCSD from the DF and hybrid DF/CD approaches (with the CD tolerances of 10–4, 10–3, and 10–2) for computations of excitation energies for the CnH2n+2 (n = 1–9) set. The RHF reference is used for these computations along with the cc-pVTZ basis set.

For the alkanes set, the computational time for the DF-EOM-CCSD and hybrid DF/CD-EOM-CCSD approaches with the CD tolerances of 10–4, 10–3, and 10–2 are presented graphically in Figure 4. For the largest member of the test set, C9H20, the computational times are 2205.9 min (DF-EOM-CCSD), 2186.6 min (DF/CD-EOM-CCSD with tolCD = 10–4), 1326.6 min (DF/CD-EOM-CCSD with tolCD = 10–3), and 1232.0 min (DF/CD-EOM-CCSD with tolCD = 10–2). With tolCD = 10–4, the cost of DF/CD-EOM-CCSD is slightly reduced compared with that of DF-EOM-CCSD, while with tolCD = 10–3 and tolCD = 10–2, the cost of DF/CD-EOM-CCSD is reduced 39.9% and 44.1% compared with that of the canonical DF-EOM-CCSD. Hence, our new hybrid approach provides significant improvements in efficiency compared to the that of the canonical DF algorithm. Further, our above discussion suggests that for the larger molecules further improvements may be observed. Hence, our new hybrid DF/CD PPL algorithm appears to be very promising for large-sized chemical systems.

Figure 4.

Figure 4

Open in a new tab

Total wall time (in min) for computations of excitation energies for the CnH2n+2 (n = 1–9) set from the DF-EOM-CCSD and hybrid DF/CD-EOM-CCSD (with the CD tolerances of 10–4, 10–3, and 10–2) methods with the cc-pVTZ basis set. The RHF reference is used for these computations. All computations were performed for a single root with 10–7 energy and EOM eigenvalue convergence tolerances on a single node (1 core) Intel(R) Xeon(R) CPU E5-2620 v4 @ 2.10 GHz computer (memory ∼ 64 GB).

We would like to note why we did not prefer the CD decomposition of the original 4-index ERIs. The number of CD factors generated from the 4-index ERIs are generally much higher than that of the DF factors achieving the same accuracy. For example, for the C8H18 molecule with the cc-pVTZ basis set, the number of auxiliary basis functions are 1188 (DF), 2018 (tolCD = 10–4), 1621 (tolCD = 10–3), and 1005 (tolCD = 10–2). Hence, the number auxiliary basis functions obtained from the partial CD decomposition of the conventional 4-index ERIs may yield a similar number with DF only if it is used with a loose CD tolerance of 10–2. However, with our hybrid approach, the number of auxiliary basis functions are 1095 (tolCD = 10–4), 593 (tolCD = 10–3), and 63 (tolCD = 10–2). Hence, our hybrid DF/CD approach significantly reduces the number of auxiliary basis functions.

4.2. Accuracy of DF-EOM-CCSD

In this section, we consider a test set to assess the accuracy of the DF-EOM-CCSD method. Chemical names of the molecules considered are given in the Supporting Information. Excitation energies (in eV) for the test set considered from the DF-EOM-CCSD, DF/CD-EOM-CCSD, RI-EOM-CCSD, and EOM-CCSD(fT) methods with the aug-cc-pVTZ basis set are reported in Table 1. The mean absolute errors (MAEs) with respect to EOM-CCSD(fT) are depicted in Figure 5. The MAE values with respect to EOM-CCSD(fT) are 0.71 eV (CIS), 0.26 eV (DF-EOM-CCSD), 0.26 eV (DF/CD-EOM-CCSD with tolCD = 10–4), 0.27 eV (DF/CD-EOM-CCSD with tolCD = 5 × 10–4), 0.27 eV (DF/CD-EOM-CCSD with tolCD = 10–3), 0.30 eV (DF/CD-EOM-CCSD with tolCD = 5 × 10–3), 0.33 eV (DF/CD-EOM-CCSD with tolCD = 10–2), and 0.27 (RI-EOM-CCSD). Hence, the results of DF-EOM-CCSD, RI-EOM-CCSD, and DF/CD-EOM-CCSD (with tolCD = 10–4–10–3) are almost identical. Further, the errors of DF/CD-EOM-CCSD with tolCD = 5 × 10–3 and 10–2 are only 0.03 and 0.06 eV, deviating from the DF-EOM-CCSD approach. When these results and the computational efficiency are considered, the DF/CD-EOM-CCSD approach may be used with such loose CD tolerances. The differences between DF-EOM-CCSD and RI-EOM-CCSD results are in between 0.00 and 0.06 eV for most cases.

Table 1. Excitation Energies for the First Five Excited States (in eV) of the Test Set Considered from the DF-EOM-CCSD, DF/CD-EOM-CCSD, RI-EOM-CCSD, and EOM-CCSD(fT) Methods with the aug-cc-pVTZ Basis Set.

  DF-EOM-CCSD DF/CD-EOM-CCSDa DF/CD-EOM-CCSDb DF/CD-EOM-CCSDc DF/CD-EOM-CCSDd DF/CD-EOM-CCSDe RI-EOM-CCSDf EOM-CCSD(fT)f
1 5.78 5.78 5.78 5.78 5.78 5.79 5.83 5.51
  6.67 6.67 6.67 6.68 6.71 6.73 6.70 6.41
  6.72 6.72 6.72 6.73 6.75 6.78 6.72 6.47
  7.33 7.33 7.33 7.33 7.34 7.36 7.39 7.07
  7.69 7.69 7.70 7.70 7.73 7.76 7.73 7.48
2 4.53 4.53 4.53 4.53 4.53 4.54 4.57 4.25
  6.57 6.57 6.57 6.58 6.61 6.63 6.60 6.36
  7.55 7.55 7.55 7.56 7.59 7.62 7.58 7.36
  7.60 7.60 7.61 7.61 7.64 7.67 7.63 7.40
  7.68 7.68 7.68 7.69 7.72 7.75 7.71 7.50
3 5.69 5.69 5.70 5.70 5.75 5.78 5.70 5.30
  5.87 5.87 5.88 5.89 5.96 6.01 5.86 5.58
  6.49 6.49 6.50 6.50 6.58 6.64 6.47 6.21
  6.54 6.54 6.55 6.56 6.63 6.69 6.52 6.25
  6.65 6.65 6.65 6.66 6.74 6.79 6.63 6.36
4 6.79 6.79 6.79 6.79 6.81 6.84 6.80 6.50
  6.92 6.92 6.92 6.92 6.95 6.98 6.91 6.62
  7.03 7.03 7.04 7.04 7.09 7.13 7.02 6.80
  7.42 7.42 7.42 7.43 7.47 7.51 7.41 7.18
  7.45 7.45 7.45 7.46 7.50 7.54 7.43 7.21
5 6.36 6.36 6.36 6.37 6.42 6.47 6.39 6.02
  6.45 6.45 6.45 6.46 6.53 6.60 6.43 6.16
  6.76 6.76 6.76 6.77 6.84 6.91 6.74 6.47
  6.92 6.92 6.93 6.93 7.00 7.08 6.90 6.63
  7.21 7.21 7.22 7.22 7.28 7.34 7.18 6.69
6 7.48 7.48 7.49 7.49 7.51 7.54 7.46 7.23
  8.09 8.09 8.09 8.09 8.10 8.12 8.08 7.81
  8.13 8.13 8.13 8.13 8.16 8.19 8.11 7.89
  8.20 8.20 8.20 8.20 8.23 8.26 8.18 7.95
  8.55 8.55 8.55 8.56 8.57 8.59 8.56 8.36
7 4.07 4.07 4.07 4.07 4.07 4.07 4.07 3.82
  7.20 7.20 7.20 7.20 7.21 7.22 7.22 7.07
  8.09 8.09 8.09 8.09 8.09 8.10 8.11 7.97
  8.18 8.18 8.18 8.18 8.19 8.21 8.20 8.07
  8.61 8.61 8.61 8.61 8.62 8.63 8.64 8.52
8 5.70 5.70 5.70 5.70 5.70 5.70 5.74 5.45
  6.92 6.92 6.93 6.93 6.94 6.96 6.92 6.68
  6.99 6.99 7.00 7.00 7.01 7.03 7.01 6.71
  7.51 7.51 7.51 7.51 7.52 7.54 7.57 7.31
  7.71 7.71 7.71 7.72 7.73 7.74 7.76 7.53
9 6.19 6.19 6.20 6.20 6.27 6.32 6.18 5.88
  6.57 6.57 6.57 6.58 6.62 6.65 6.53 6.09
  6.73 6.73 6.74 6.74 6.81 6.86 6.72 6.43
  6.89 6.89 6.89 6.89 6.91 6.93 6.90 6.60
  6.92 6.92 6.92 6.93 7.00 7.05 7.34 7.03
10 5.82 5.82 5.83 5.84 5.89 5.93 5.81 5.50
  6.59 6.59 6.60 6.60 6.65 6.70 6.58 6.29
  6.63 6.63 6.63 6.63 6.67 6.70 6.59 6.16
  6.85 6.85 6.85 6.86 6.89 6.90 6.83 6.54
  6.87 6.87 6.87 6.88 6.91 6.96 6.91 6.49
Open in a new tab
a

These computations were performed with the hybrid DF/CD algorithm employing a CD tolerance of 1 × 10–4.

b

These computations were performed with the hybrid DF/CD algorithm employing a CD tolerance of 5 × 10–4.

c

These computations were performed with the hybrid DF/CD algorithm employing a CD tolerance of 1 × 10–3.

d

These computations were performed with the hybrid DF/CD algorithm employing a CD tolerance of 5 × 10–3.

e

These computations were performed with the hybrid DF/CD algorithm employing a CD tolerance of 1 × 10–2.

f

These computations were performed with the Q-chem 5.3 program.

Figure 5.

Figure 5

Open in a new tab

Mean absolute errors (in eV) in excitation energies for the test set from the DF-EOM-CCSD, DF/CD-EOM-CCSD, and RI-EOM-CCSD methods with respect to EOM-CCSD(fT) (the aug-cc-pVTZ basis set was employed).

5. Conclusions

In this study, a new implementation of the density-fitted EOM-CCSD (DF-EOM-CCSD) method has been presented with an enhanced algorithm for the particle–particle ladder (PPL) term, which is the most expensive term of the EOM-CCSD computations. To further improve the evaluation of the PPL term, a hybrid DF/CD algorithm has also been introduced. The computational time of the DF-EOM-CCSD excitation energies has been compared with that of the RI-EOM-CCSD method (from Q-chem 5.3 package72).

The DF-EOM-CCSD method significantly reduces the computational cost compared to that of the RI-EOM-CCSD method; there is more than a 2-fold reduction for the C8H18 molecule in a cc-pVTZ basis set with the RHF reference. This cost savings results from the accelerated evaluation of the PPL term. In our RHF based DF-EOM-CCSD method, the number of flops (NOFs) for the PPL contraction term is 2.5 times lower than that of the RI-EOM-CCSD method. Further, the prefactor of the VVVV-type transformation step used in the PPL term has a reduced NOF value by a factor of 2. For the UHF reference, our implementation maintains its better performance and provides a 1.8-fold reduction in the computational cost compared to that of the RI-EOM-CCSD method for the C7H16 molecule. Furthermore, our results show that the suggested hybrid DF/CD algorithm further improves the canonical DF algorithm, and the degree of improvement increases as the molecular size increases. The preliminary results indicate that the new hybrid DF/CD PPL algorithm is very promising for large-sized chemical systems.

Finally, the DF-EOM-CCSD and DF/CD-EOM-CCSD methods are applied to a test set to compare the excitation energies with those from the CIS, RI-EOM-CCSD, and EOM-CCSD(fT) methods. Our results demonstrate that the DF-EOM-CCSD and DF/CD-EOM-CCSD methods (with CD tolerances of 10–4–10–3) provide identical results with to those of the RI-EOM-CCSD method. Further, the DF/CD-EOM-CCSD approach yields tolerable errors (0.03 and 0.06 eV) compared with those of the DF-EOM-CCSD method for the test set considered with loose CD tolerances, such as 5 × 10–3 and 10–2.

Acknowledgments

This research was supported by the Scientific and Technological Research Council of Turkey (TÜBİTAK-118Z916). This work is a part of A.Ü.’s Ph.D. thesis study. Authors thank Tuğba Bozkaya for performing RI-EOM-CCSD and EOM-CCSD(fT) excitation energy computations with the cc-pVTZ basis set.

Appendix A

DF-CCSD 3-Index Intermediates

1- and 3-index intermediates used for DF-CCSD are given as follows:65,66

graphic file with name ct1c01000_m052.jpg 52
graphic file with name ct1c01000_m053.jpg 53
graphic file with name ct1c01000_m054.jpg 54
graphic file with name ct1c01000_m055.jpg 55
graphic file with name ct1c01000_m056.jpg 56
graphic file with name ct1c01000_m057.jpg 57
graphic file with name ct1c01000_m058.jpg 58
graphic file with name ct1c01000_m059.jpg 59
graphic file with name ct1c01000_m060.jpg 60
graphic file with name ct1c01000_m061.jpg 61
graphic file with name ct1c01000_m062.jpg 62

where

graphic file with name ct1c01000_m063.jpg 63

and brsQ values are the molecular orbital (MO) DF factors, which are defined in our previous studies.65−67

F and Inline graphic Intermediates

Density-fitted F and Inline graphic intermediates are65,66

graphic file with name ct1c01000_m066.jpg 64
graphic file with name ct1c01000_m067.jpg 65
graphic file with name ct1c01000_m068.jpg 66
graphic file with name ct1c01000_m069.jpg 67
graphic file with name ct1c01000_m070.jpg 68

where fpq is the MO basis Fock matrix. We would like to note that in our previous studies65,66 the diagonal parts of the Fock matrix were removed from the definitions of the Fmi and Fae intermediates since the diagonal Fock terms were moved into the tijabDij definition in the amplitude equation.

W Intermediates

W intermediates, with the DF approximation, are65,66

graphic file with name ct1c01000_m071.jpg 69
graphic file with name ct1c01000_m072.jpg 70
graphic file with name ct1c01000_m073.jpg 71

where P± (pq) is defined by

graphic file with name ct1c01000_m074.jpg 72

and Inline graphic acts to permute the indices p and q.

Inline graphic and Inline graphic Intermediates

The Inline graphic and Inline graphic intermediates are defined as follows:65,66

graphic file with name ct1c01000_m080.jpg 73
graphic file with name ct1c01000_m081.jpg 74
graphic file with name ct1c01000_m082.jpg 75

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jctc.1c01000.

  • Optimized geometries of the species; excitation energies with the cc-pVTZ basis set for the test set; CIS excitation energies with the aug-cc-pVTZ basis set for the test set (PDF)

  • Spin-free RHF-DF-EOM-CCSD equations (PDF)

Author Contributions

A.U.: Investigation, methodology, software, validation, and visualization. U.B: Conceptualization, funding acquisition, methodology, project administration, resources, software, supervision, and writing (original draft, review, and editing).

The authors declare no competing financial interest.

Supplementary Material

ct1c01000_si_001.pdf (210.9KB, pdf)
ct1c01000_si_002.pdf (190.7KB, pdf)

References

  1. Cizek J. On the Correlation Problem in Atomic and Molecular Systems. Calculation of Wavefunction Components in Ursell-Type Expansion Using Quantum-Field Theoretical Methods. J. Chem. Phys. 1966, 45, 4256. 10.1063/1.1727484. [DOI] [Google Scholar]
  2. Bartlett R. J. Many-Body Perturbation Theory and Coupled Cluster Theory for Electron Correlation in Molecules. Annu. Rev. Phys. Chem. 1981, 32, 359–401. 10.1146/annurev.pc.32.100181.002043. [DOI] [Google Scholar]
  3. Bartlett R. J. Coupled-Cluster Approach to Molecular Structure and Spectra: A Step Toward Predictive Quantum Chemistry. J. Phys. Chem. 1989, 93, 1697–1708. 10.1021/j100342a008. [DOI] [Google Scholar]
  4. Crawford T. D.; Schaefer H. F. An Introduction to Coupled Cluster Theory for Computational Chemists. Rev. Comp. Chem. 2007, 14, 33–136. 10.1002/9780470125915.ch2. [DOI] [Google Scholar]
  5. Bartlett R. J.; Musiał M. Coupled-Cluster Theory in Quantum Chemistry. Rev. Mod. Phys. 2007, 79, 291–352. 10.1103/RevModPhys.79.291. [DOI] [Google Scholar]
  6. Bartlett R. J. Coupled-Cluster Theory and Its Equation-of-Motion Extensions. WIREs Comput. Mol. Sci. 2012, 2, 126–138. 10.1002/wcms.76. [DOI] [Google Scholar]
  7. Barlett R. J.; Sekino H.; Purvis G. D. Comparison of MBPT and Coupled-cluster Methods with Full CI. Importance of Triplet Excitation and Infinite Summations. Chem. Phys. Lett. 1983, 98, 66–71. 10.1016/0009-2614(83)80204-8. [DOI] [Google Scholar]
  8. Lee Y. S.; Kucharski S. A.; Bartlett R. J. A Coupled Cluster Approach with Triple Excitations. J. Chem. Phys. 1984, 81, 5906–5912. 10.1063/1.447591. [DOI] [Google Scholar]
  9. Pople J. A.; Head-Gordon M.; Raghavachari K. Quadratic Configuration Interaction. A General Technique for Determining Electron Correlation Energies. J. Chem. Phys. 1987, 87, 5968–5975. 10.1063/1.453520. [DOI] [Google Scholar]
  10. Raghavachari K.; Trucks G. W.; Pople J. A.; Head-Gordon M. A Fifth-Order Perturbation Comparison of Electronic Correlation Theories. Chem. Phys. Lett. 1989, 157, 479–483. 10.1016/S0009-2614(89)87395-6. [DOI] [Google Scholar]
  11. Bartlett R. J.; Watts J. D.; Kucharski S. A.; Noga J. Non-Iterative Fifth-Order Triple and Quadruple Excitation Energy Corrections in Correlated Methods. Chem. Phys. Lett. 1990, 165, 513–522. 10.1016/0009-2614(90)87031-L. [DOI] [Google Scholar]
  12. Scuseria G. E.; Lee T. J. Comparison of Coupled-Cluster Methods which Include the Effects of Connected Triple Excitations. J. Chem. Phys. 1990, 93, 5851–5855. 10.1063/1.459684. [DOI] [Google Scholar]
  13. Scuseria G. E.; Hamilton T. P.; Schaefer H. F. An Assessment for the Full Coupled Cluster Method Including All Single, Double, and Triple Excitations: The Diatomic Molecules LiH, Li2, BH, LiF, C2, BeO, CN+, BF, NO+, and F2. J. Chem. Phys. 1990, 92, 568–573. 10.1063/1.458407. [DOI] [Google Scholar]
  14. Purvis G. D.; Bartlett R. J. A Full Coupled-Cluster Singles and Doubles Model: The Inclusion of Disconnected Triples. J. Chem. Phys. 1982, 76, 1910–1918. 10.1063/1.443164. [DOI] [Google Scholar]
  15. Urban M.; Noga J.; Cole S. J.; Bartlett R. J. Towards a Full CCSDT Model for Electron Correlation. J. Chem. Phys. 1985, 83, 4041–4046. 10.1063/1.449067. [DOI] [Google Scholar]
  16. Lee T. J.; Scuseria G. E. In Quantum Mechanical Electronic Structure Calculations with Chemical Accuracy; Langhoff S. R., Ed.; Kluwer Academic: Dordrecht, 1995; pp 47–108. [Google Scholar]
  17. Watts J. D.; Stanton J. F.; Bartlett R. J. A Benchmark Coupled-Cluster Single, Double, and Triple Excitation (CCSDT) Study of the Structure and Harmonic Vibrational Frequencies of the Ozone Molecule. Chem. Phys. Lett. 1991, 178, 471–474. 10.1016/0009-2614(91)87004-U. [DOI] [Google Scholar]
  18. Scuseria G. E. Ab Initio Theoretical Predictions of the Equilibrium Geometries of C60, C60H60 and C60F60. Chem. Phys. Lett. 1991, 176, 423–427. 10.1016/0009-2614(91)90231-W. [DOI] [Google Scholar]
  19. Gauss J.; Lauderdale W. J.; Stanton J. F.; Watts J. D.; Bartlett R. J. Analytic Energy Gradients for Open-Shell Coupled-Cluster Singles and Doubles (CCSD) Calculations Using Restricted Open-Shell Hartree–Fock (ROHF) Reference Functions. Chem. Phys. Lett. 1991, 182, 207–215. 10.1016/0009-2614(91)80203-A. [DOI] [Google Scholar]
  20. Watts J. D.; Bartlett R. J. Coupled-Cluster Calculations on the C2 molecule and the C2+ and C2– Molecular Ions. J. Chem. Phys. 1992, 96, 6073–6084. 10.1063/1.462649. [DOI] [Google Scholar]
  21. Thomas J. R.; DeLeeuw B. J.; Vacek G.; Crawford T. D.; Yamaguchi Y.; Schaefer H. F. The Balance Between Theoretical Method and Basis Set Quality: A Systematic Study of Equilibrium Geometries, Dipole Moments, Harmonic Vibrational Frequencies, and Infrared Intensities. J. Chem. Phys. 1993, 99, 403–416. 10.1063/1.465764. [DOI] [Google Scholar]
  22. Watts J. D.; Gauss J.; Bartlett R. J. Coupled-Cluster Methods With Noniterative Triple Excitations For Restricted Open-Shell Hartree-Fock And Other General Single Determinant Reference Functions. Energies And Analytical Gradients. J. Chem. Phys. 1993, 98, 8718–8733. 10.1063/1.464480. [DOI] [Google Scholar]
  23. Crawford T. D.; Schaefer H. F. A Comparison of Two Approaches to Perturbational Triples Corrections to the Coupled-Cluster Singles and Doubles Method for High-Spin Open-Shell Systems. J. Chem. Phys. 1996, 104, 6259–6264. 10.1063/1.471287. [DOI] [Google Scholar]
  24. Crawford T. D.; Lee T. J.; Schaefer H. F. A New Spin-Restricted Perturbative Triple Excitation Correction for Coupled Cluster Theory. J. Chem. Phys. 1997, 107, 7943–7950. 10.1063/1.475081. [DOI] [Google Scholar]
  25. Bozkaya U.; Schaefer H. F. Symmetric and Asymmetric Triple Excitation Corrections for The Orbital-Optimized Coupled-Cluster Doubles Method: Improving Upon CCSD(T) and CCSD(T)Λ: Preliminary Application. J. Chem. Phys. 2012, 136, 204114. 10.1063/1.4720382. [DOI] [PubMed] [Google Scholar]
  26. Stanton J. F.; Bartlett R. J. The Equation of Motion Coupled-Cluster Method. A Systematic Biorthogonal Approach to Molecular Excitation Energies, Transition Probabilities, and Excited State Properties. J. Chem. Phys. 1993, 98, 7029–7039. 10.1063/1.464746. [DOI] [Google Scholar]
  27. Watts J. D.; Bartlett R. J. Economical Triple Excitation Equation-of-Motion Coupled-Cluster Methods for Excitation Energies. Chem. Phys. Lett. 1995, 233, 81–87. 10.1016/0009-2614(94)01434-W. [DOI] [Google Scholar]
  28. Gwaltney S. R.; Nooijen M.; Bartlett R. J. Simplified Methods for Equation-of-Motion Coupled-Cluster Excited State Calculations. Chem. Phys. Lett. 1996, 248, 189–198. 10.1016/0009-2614(95)01329-6. [DOI] [Google Scholar]
  29. Hirata S.; Nooijen M.; Grabowski I.; Bartlett R. J. Perturbative Corrections to Coupled-Cluster and Equation-of-Motion Coupled-Cluster Energies: A Determinantal Analysis. J. Chem. Phys. 2001, 114, 3919–3928. 10.1063/1.1346578. [DOI] [Google Scholar]
  30. Sattelmeyer K. W.; Stanton J. F.; Olsen J.; Gauss J. A Comparison of Excited State Properties for Iterative Approximate Triples Linear Response Coupled Cluster Methods. Chem. Phys. Lett. 2001, 347, 499–504. 10.1016/S0009-2614(01)01013-2. [DOI] [Google Scholar]
  31. Larsen H.; Hald K.; Olsen J.; Jørgensen P. Triplet Excitation Energies in Full Configuration Interaction and Coupled-Cluster Theory. J. Chem. Phys. 2001, 115, 3015–3020. 10.1063/1.1386415. [DOI] [Google Scholar]
  32. Hirata S. Higher-Order Equation-of-Motion Coupled-Cluster Methods. J. Chem. Phys. 2004, 121, 51. 10.1063/1.1753556. [DOI] [PubMed] [Google Scholar]
  33. Smith C. E.; King R. A.; Crawford T. D. Coupled Cluster Methods Including Triple Excitations for Excited States of Radicals. J. Chem. Phys. 2005, 122, 054110. 10.1063/1.1835953. [DOI] [PubMed] [Google Scholar]
  34. Gour J. R.; Piecuch P.; Włoch M. Active-Space Equation-of-Motion Coupled-Cluster Methods for Excited States of Radicals and Other Open-Shell Systems: EA-EOMCCSDt and IP-EOMCCSDt. J. Chem. Phys. 2005, 123, 134113. 10.1063/1.2042452. [DOI] [PubMed] [Google Scholar]
  35. Slipchenko L. V.; Krylov A. I. Spin-Conserving and Spin-Flipping Equation-of-Motion Coupled-Cluster Method with Triple Excitations. J. Chem. Phys. 2005, 123, 084107. 10.1063/1.2006091. [DOI] [PubMed] [Google Scholar]
  36. Musiał M.; Bartlett R. J. Addition by Subtraction in Coupled Cluster Theory. II. Equation-of-Motion Coupled Cluster Method for Excited, Ionized, and Electron-Attached States Based on The nCC Ground State Wave Function. J. Chem. Phys. 2007, 127, 024106. 10.1063/1.2747245. [DOI] [PubMed] [Google Scholar]
  37. Krylov A. I. Equation-of-Motion Coupled-Cluster Methods for Open-Shell and Electronically Excited Species: The Hitchhiker’s Guide to Fock Space. Annu. Rev. Phys. Chem. 2008, 59, 433–462. 10.1146/annurev.physchem.59.032607.093602. [DOI] [PubMed] [Google Scholar]
  38. Musial M.; Bartlett R. J. Multireference Fock-Space Coupled-Cluster and Equation-of-Motion Coupled-Cluster Theories: The Detailed Interconnections. J. Chem. Phys. 2008, 129, 134105. 10.1063/1.2982788. [DOI] [PubMed] [Google Scholar]
  39. Kuś T.; Lotrich V. F.; Bartlett R. J. Parallel Implementation of The Equation-of-Motion Coupled-Cluster Singles and Doubles Method and Application for Radical Adducts of Cytosine. J. Chem. Phys. 2009, 130, 124122. 10.1063/1.3091293. [DOI] [PubMed] [Google Scholar]
  40. Musiał M. The Excited, Ionized and Electron Attached States Within The EOM-CC Approach with Full Inclusion of Connected Triple Excitations. Mol. Phys. 2010, 108, 2921–2931. 10.1080/00268976.2010.510854. [DOI] [Google Scholar]
  41. Sneskov K.; Christiansen O. Excited State Coupled Cluster Methods. Wiley Interdisciplinary Reviews: Computational Molecular Science 2012, 2, 566–584. 10.1002/wcms.99. [DOI] [Google Scholar]
  42. Bartlett R. J. Coupled-Cluster Theory and Its Equation-of-Motion Extensions. WIREs Computational Molecular Science 2012, 2, 126–138. 10.1002/wcms.76. [DOI] [Google Scholar]
  43. Musiał M.; Olszówka M.; Lyakh D. I.; Bartlett R. J. The Equation-of-Motion Coupled Cluster Method for Triple Electron Attached States. J. Chem. Phys. 2012, 137, 174102. 10.1063/1.4763354. [DOI] [PubMed] [Google Scholar]
  44. Matthews D. A.; Stanton J. F. A New Approach to Approximate Equation-of-Motion Coupled Cluster with Triple Excitations. J. Chem. Phys. 2016, 145, 124102. 10.1063/1.4962910. [DOI] [PubMed] [Google Scholar]
  45. Whitten J. L. Coulombic Potential Energy Integrals and Approximations. J. Chem. Phys. 1973, 58, 4496–4501. 10.1063/1.1679012. [DOI] [Google Scholar]
  46. Dunlap B. I.; Connolly J. W. D.; Sabin J. R. On Some Approximations in Applications Xα Theory. J. Chem. Phys. 1979, 71, 3396–3402. 10.1063/1.438728. [DOI] [Google Scholar]
  47. Feyereisen M.; Fitzgerald G.; Komornicki A. Use of Approximate Integrals in Ab Initio Theory. An Application in MP2 Energy Calculations. Chem. Phys. Lett. 1993, 208, 359–363. 10.1016/0009-2614(93)87156-W. [DOI] [Google Scholar]
  48. Vahtras O.; Almlöf J.; Feyereisen M. W. Integral Approximations for LCAO-SCF Calculations. Chem. Phys. Lett. 1993, 213, 514–518. 10.1016/0009-2614(93)89151-7. [DOI] [Google Scholar]
  49. Rendell A. P.; Lee T. J. Coupled-Cluster Theory Employing Approximate Integrals: An Approach to Avoid The Input/Output and Storage Bottlenecks. J. Chem. Phys. 1994, 101, 400–408. 10.1063/1.468148. [DOI] [Google Scholar]
  50. Weigend F. A Fully Direct RI-HF Algorithm: Implementation, Optimised Auxiliary Basis Sets, Demonstration of Accuracy and Efficiency. Phys. Chem. Chem. Phys. 2002, 4, 4285–4291. 10.1039/b204199p. [DOI] [Google Scholar]
  51. Sodt A.; Subotnik J. E.; Head-Gordon M. Linear Scaling Density Fitting. J. Chem. Phys. 2006, 125, 194109. 10.1063/1.2370949. [DOI] [PubMed] [Google Scholar]
  52. Werner H.-J.; Manby F. R.; Knowles P. J. Fast Linear Scaling Second-Order Møller-Plesset Perturbation Theory (MP2) Using Local and Density Fitting Approximations. J. Chem. Phys. 2003, 118, 8149–8160. 10.1063/1.1564816. [DOI] [Google Scholar]
  53. Schütz M.; Manby F. R. Linear Scaling Local Coupled Cluster Theory with Density Fitting. Part I: 4-External Integrals. Phys. Chem. Chem. Phys. 2003, 5, 3349. 10.1039/B304550A. [DOI] [Google Scholar]
  54. Werner H.-J.; Schütz M. An Efficient Local Coupled Cluster Method for Accurate Thermochemistry of Large Systems. J. Chem. Phys. 2011, 135, 144116. 10.1063/1.3641642. [DOI] [PubMed] [Google Scholar]
  55. Beebe N. H. F.; Linderberg J. Simplifications in The Generation and Transformation of Two-Electron Integrals in Molecular Calculations. Int. J. Quantum Chem. 1977, 12, 683–705. 10.1002/qua.560120408. [DOI] [Google Scholar]
  56. Røeggen I.; Wisløff-Nilssen E. On the Beebe-Linderberg Two-Electron Integral Approximation. Chem. Phys. Lett. 1986, 132, 154–160. 10.1016/0009-2614(86)80099-9. [DOI] [Google Scholar]
  57. Koch H.; de Meras A. S.; Pedersen T. B. Reduced Scaling in Electronic Structure Calculations Using Cholesky Decompositions. J. Chem. Phys. 2003, 118, 9481–9484. 10.1063/1.1578621. [DOI] [Google Scholar]
  58. Aquilante F.; Pedersen T. B.; Lindh R. Low-Cost Evaluation of The Exchange Fock Matrix from Cholesky and Density Fitting Representations of The Electron Repulsion Integrals. J. Chem. Phys. 2007, 126, 194106. 10.1063/1.2736701. [DOI] [PubMed] [Google Scholar]
  59. DePrince A. E.; Sherrill C. D. Accuracy and Efficiency of Coupled-Cluster Theory Using Density Fitting/Cholesky Decomposition, Frozen Natural Orbitals, and a t1-Transformed Hamiltonian. J. Chem. Theory Comput. 2013, 9, 2687–2696. 10.1021/ct400250u. [DOI] [PubMed] [Google Scholar]
  60. Bozkaya U. Orbital-Optimized Second-Order Perturbation Theory with Density-Fitting and Cholesky Decomposition Approximations: An Efficient Implementation. J. Chem. Theory Comput. 2014, 10, 2371–2378. 10.1021/ct500231c. [DOI] [PubMed] [Google Scholar]
  61. Bozkaya U. Derivation of General Analytic Gradient Expressions for Density-Fitted Post-Hartree-Fock Methods: An Efficient Implementation for the Density-Fitted Second-Order Møller–Plesset Perturbation Theory. J. Chem. Phys. 2014, 141, 124108. 10.1063/1.4896235. [DOI] [PubMed] [Google Scholar]
  62. Bozkaya U. Analytic Energy Gradients and Spin Multiplicities for Orbital-Optimized Second-Order Perturbation Theory with Density-Fitting Approximation: An Efficient Implementation. J. Chem. Theory Comput. 2014, 10, 4389–4399. 10.1021/ct500634s. [DOI] [PubMed] [Google Scholar]
  63. Bozkaya U. Orbital-Optimized MP3 and MP2.5 with Density-Fitting and Cholesky Decomposition Approximations. J. Chem. Theory Comput. 2016, 12, 1179–1188. 10.1021/acs.jctc.5b01128. [DOI] [PubMed] [Google Scholar]
  64. Bozkaya U. Orbital-Optimized Linearized Coupled-Cluster Doubles with Density-Fitting and Cholesky Decomposition Approximations: An Efficient Implementation. Phys. Chem. Chem. Phys. 2016, 18, 11362–11373. 10.1039/C6CP00164E. [DOI] [PubMed] [Google Scholar]
  65. Bozkaya U.; Sherrill C. D. Analytic Energy Gradients for the Coupled-Cluster Singles and Doubles Method with the Density-Fitting Approximation. J. Chem. Phys. 2016, 144, 174103. 10.1063/1.4948318. [DOI] [PubMed] [Google Scholar]
  66. Bozkaya U. A Noniterative Asymmetric Triple Excitation Correction for The Density-Fitted Coupled-Cluster Singles and Doubles Method: Preliminary Applications. J. Chem. Phys. 2016, 144, 144108. 10.1063/1.4945706. [DOI] [PubMed] [Google Scholar]
  67. Bozkaya U.; Sherrill C. D. Analytic Energy Gradients for The Coupled-Cluster Singles and Doubles with Perturbative Triples Method with the Density-Fitting Approximation. J. Chem. Phys. 2017, 147, 044104–044114. 10.1063/1.4994918. [DOI] [PubMed] [Google Scholar]
  68. Bozkaya U. Analytic Energy Gradients for Orbital-Optimized MP3 and MP2.5 with the Density-Fitting Approximation: An Efficient Implementation. J. Comput. Chem. 2018, 39, 351–360. 10.1002/jcc.25122. [DOI] [PubMed] [Google Scholar]
  69. Bozkaya U.; Ünal A.; Alagöz Y. Energy and Analytic Gradients for the Orbital-Optimized Coupled-Cluster Doubles Method with the Density-Fitting Approximation: An Efficient Implementation. J. Chem. Phys. 2020, 153, 244115. 10.1063/5.0035811. [DOI] [PubMed] [Google Scholar]
  70. Alagöz Y.; Ünal A.; Bozkaya U. Efficient Implementations of the Symmetric and Asymmetric Triple Excitation Corrections for the Orbital-Optimized Coupled-Cluster Doubles Method with the Density-Fitting Approximation. J. Chem. Phys. 2021, 155, 114104. 10.1063/5.0061351. [DOI] [PubMed] [Google Scholar]
  71. Bozkaya U.; Ermiş B.; Alagöz Y.; Ünal A.; Uyar A. K. MacroQC 1.0: An Electronic Structure Theory Software for Large-Scale Applications. J. Chem. Phys. 2022, 156, 044801. 10.1063/5.0077823. [DOI] [PubMed] [Google Scholar]
  72. Shao Y.; Gan Z.; Epifanovsky E.; Gilbert A. T.; Wormit M.; Kussmann J.; Lange A. W.; Behn A.; Deng J.; Feng X.; Ghosh D.; Goldey M.; Horn P. R.; Jacobson L. D.; Kaliman I.; Khaliullin R. Z.; Kuś T.; Landau A.; Liu J.; Proynov E. I.; Rhee Y. M.; Richard R. M.; Rohrdanz M. A.; Steele R. P.; Sundstrom E. J.; Woodcock H. L.; Zimmerman P. M.; Zuev D.; Albrecht B.; Alguire E.; Austin B.; Beran G. J. O.; Bernard Y. A.; Berquist E.; Brandhorst K.; Bravaya K. B.; Brown S. T.; Casanova D.; Chang C.-M.; Chen Y.; Chien S. H.; Closser K. D.; Crittenden D. L.; Diedenhofen M.; DiStasio R. A.; Do H.; Dutoi A. D.; Edgar R. G.; Fatehi S.; Fusti-Molnar L.; Ghysels A.; Golubeva-Zadorozhnaya A.; Gomes J.; Hanson-Heine M. W.; Harbach P. H.; Hauser A. W.; Hohenstein E. G.; Holden Z. C.; Jagau T.-C.; Ji H.; Kaduk B.; Khistyaev K.; Kim J.; Kim J.; King R. A.; Klunzinger P.; Kosenkov D.; Kowalczyk T.; Krauter C. M.; Lao K. U.; Laurent A. D.; Lawler K. V.; Levchenko S. V.; Lin C. Y.; Liu F.; Livshits E.; Lochan R. C.; Luenser A.; Manohar P.; Manzer S. F.; Mao S.-P.; Mardirossian N.; Marenich A. V.; Maurer S. A.; Mayhall N. J.; Neuscamman E.; Oana C. M.; Olivares-Amaya R.; O’Neill D. P.; Parkhill J. A.; Perrine T. M.; Peverati R.; Prociuk A.; Rehn D. R.; Rosta E.; Russ N. J.; Sharada S. M.; Sharma S.; Small D. W.; Sodt A.; Stein T.; Stück D.; Su Y.-C.; Thom A. J.; Tsuchimochi T.; Vanovschi V.; Vogt L.; Vydrov O.; Wang T.; Watson M. A.; Wenzel J.; White A.; Williams C. F.; Yang J.; Yeganeh S.; Yost S. R.; You Z.-Q.; Zhang I. Y.; Zhang X.; Zhao Y.; Brooks B. R.; Chan G. K.; Chipman D. M.; Cramer C. J.; Goddard W. A.; Gordon M. S.; Hehre W. J.; Klamt A.; Schaefer H. F.; Schmidt M. W.; Sherrill C. D.; Truhlar D. G.; Warshel A.; Xu X.; Aspuru-Guzik A.; Baer R.; Bell A. T.; Besley N. A.; Chai J.-D.; Dreuw A.; Dunietz B. D.; Furlani T. R.; Gwaltney S. R.; Hsu C.-P.; Jung Y.; Kong J.; Lambrecht D. S.; Liang W.; Ochsenfeld C.; Rassolov V. A.; Slipchenko L. V.; Subotnik J. E.; Voorhis T. V.; Herbert J. M.; Krylov A. I.; Gill P. M.; Head-Gordon M. Advances in Molecular Quantum Chemistry Contained in the Q-Chem 4 Program Package. Mol. Phys. 2015, 113, 184–215. 10.1080/00268976.2014.952696. [DOI] [Google Scholar]
  73. Shavitt I.; Bartlett R. J.. Many-Body Methods in Chemistry and Physics, 1st ed.; Cambridge Press: New York, 2009; pp 443–449. [Google Scholar]
  74. Davidson E. R. The Iterative Calculation of a Few of the Lowest Eigenvalues and Corresponding Eigenvectors of Large Real-Symmetric Matrices. J. Comput. Phys. 1975, 17, 87–94. 10.1016/0021-9991(75)90065-0. [DOI] [Google Scholar]
  75. Liu B. In Numerical Algorithms in Chemistry: Algebraic Methods; Moler C., Shavitt I., Eds.; Technical Report LBL-8158, Lawrence Berkeley Laboratory, University of California: Berkeley, 1978; pp 49–53. [Google Scholar]
  76. Sherrill C. D.; Schaefer H. F. The Configuration Interaction Method: Advances in Highly Correlated Approaches. Adv. Quantum Chem. 1999, 34, 143–269. 10.1016/S0065-3276(08)60532-8. [DOI] [Google Scholar]
  77. Leininger M. L.; Sherrill C. D.; Allen W. D.; Schaefer H. F. Systematic Study of Selected Diagonalization Methods for Configuration Interaction Matrices. J. Comput. Chem. 2001, 22, 1574–1589. 10.1002/jcc.1111. [DOI] [Google Scholar]
  78. Saebø S.; Pulay P. Fourth-Order Møller–Plessett Perturbation Theory in The Local Correlation Treatment. I. Method. J. Chem. Phys. 1987, 86, 914. 10.1063/1.452293. [DOI] [Google Scholar]
  79. Scuseria G. E.; Janssen C. L.; Schaefer H. F. An Efficient Reformulation of the Closed-Shell Coupled Cluster Single and Double Excitation (CCSD) Equations. J. Chem. Phys. 1988, 89, 7382–7387. 10.1063/1.455269. [DOI] [Google Scholar]
  80. Epifanovsky E.; Zuev D.; Feng X.; Khistyaev K.; Shao Y.; Krylov A. I. General Implementation of the Resolution-of-the-Identity and Cholesky Representations of Electron Repulsion Integrals within Coupled-Cluster and Equation-of-Motion Methods: Theory and Benchmarks. J. Chem. Phys. 2013, 139, 134105. 10.1063/1.4820484. [DOI] [PubMed] [Google Scholar]
  81. Dunning T. H. Gaussian Basis Sets for Use in Correlated Molecular Calculations. I. The Atoms Boron Through Neon and Hydrogen. J. Chem. Phys. 1989, 90, 1007–1023. 10.1063/1.456153. [DOI] [Google Scholar]
  82. Woon D. E.; Dunning T. H. Gaussian Basis Sets for Use in Correlated Molecular Calculations. V. Core-Valence Basis Sets for Boron through Neon. J. Chem. Phys. 1995, 103, 4572–4585. 10.1063/1.470645. [DOI] [Google Scholar]
  83. Weigend F.; Köhn A.; Hättig C. Efficient Use of the Correlation Consistent Basis Sets in Resolution of the Identity MP2 Calculations. J. Chem. Phys. 2002, 116, 3175–3183. 10.1063/1.1445115. [DOI] [Google Scholar]
  84. Manohar P. U.; Krylov A. I. A Noniterative Perturbative Triples Correction for the Spin-Flipping and Spin-Conserving Equation-of-Motion Coupled-Cluster Methods with Single and Double Substitutions. J. Chem. Phys. 2008, 129, 194105. 10.1063/1.3013087. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

ct1c01000_si_001.pdf (210.9KB, pdf)
ct1c01000_si_002.pdf (190.7KB, pdf)

Articles from Journal of Chemical Theory and Computation are provided here courtesy of American Chemical Society

RESOURCES