X

Search Results

Searching....

13.5 Nuclear–Electronic Orbital Method

13.5.2 Overview of Available NEO Models

(September 1, 2024)

13.5.2.1 NEO-Hartree-Fock

The simplest method within the NEO framework is the Hartree-Fock (NEO-HF) method, where the total nuclear-electronic wavefunction is approximated as a product of electronic (𝚽0e(𝐱e)) and nuclear (𝚽0p(𝐱p)) Slater determinants composed of electronic and protonic spin orbitals, respectively:

𝚿NEO-HF(𝐱e,𝐱p)=𝚽0e(𝐱e)𝚽0p(𝐱p)=|0e0p. (13.33)

Here, 𝐱e and 𝐱p are collective spatial and spin coordinates of the quantum electrons and protons. The NEO-HF energy for a restricted Hartree-Fock (RHF) treatment of the electrons and a high-spin open-shell treatment of the quantum protons is

ENEO-HF=2iNe/2hiie+iNe/2jNe/2(2(ii|jj)-(ij|ij))+INphIIp+12INpJNp((II|JJ)-(IJ|IJ))-2iNe/2INp(ii|II). (13.34)

The i,j,, indices denote occupied spatial electronic orbitals, and the I,J,, indices correspond to occupied spatial protonic orbitals. In Eq. (13.34), hije and (ij|kl) are conventional electronic core Hamiltonian and two-electron integrals, respectively, and the corresponding terms for quantum protons are defined analogously. The last term in Eq. (13.34) is the Coulomb interaction between the electrons and the quantum protons. The spatial electronic and protonic orbitals [ψie(𝐫e) and ψIp(𝐫p)] are expanded as linear combinations of electronic or protonic Gaussian basis functions [ϕμe(𝐫e) or ϕμp(𝐫p)]:

ψie(𝐫e)= μNebfCμieϕμe(𝐫e) (13.35a)
ψIp(𝐫p)= μNpbfCμIpϕμp(𝐫p). (13.35b)

The lower-case Greek letters without and with primes denote basis functions for electrons and protons, respectively, and Cμie and CμIp are electronic and protonic MO expansion coefficients, respectively.

Analogous to the conventional electronic Hartree-Fock method, the electronic and protonic coefficients are determined by variationally minimizing the energy in Eq. (13.34) via the self-consistent field (SCF) procedure. This procedure leads to a set of coupled electronic and protonic Roothaan equations:

𝐅e𝐂e =𝐒e𝐂e𝐄e (13.36a)
𝐅p𝐂p =𝐒p𝐂p𝐄p, (13.36b)

where 𝐒e and 𝐒p are electronic and protonic overlap matrices, respectively. The electronic and protonic Fock elements in Eqs. (13.36a) and (13.36b) are given by

Fμνe =hμνe+ρλPλρe((μν|ρλ)-12(μλ|ρν))-μνPνμp(μν|μν) (13.37a)
Fμνp =hμνp+ρλPλρp((μν|ρλ)-(μλ|ρν))-μνPνμe(μν|μν). (13.37b)

The electronic and protonic density matrix elements in Eqs. (13.37a) and (13.37b) are defined as

Pνμe =2iNe/2CνieCμie* (13.38a)
Pνμp =INpCνIpCμIp*. (13.38b)

The generalization to the unrestricted Hartree-Fock (NEO-UHF) treatment of electrons is accomplished by introducing separate spatial orbitals for α and β electron spins.

The analytical gradients of the NEO-HF energy 1335 Webb S. P., Iordanov T., Hammes-Schiffer S.
J. Chem. Phys.
(2002), 117, pp. 4106.
Link
with respect to the classical nuclear coordinates (or coordinates of the centers of the quantum proton basis functions) are available. These gradients allow geometry optimizations within the NEO framework. The analytical Hessians of the NEO-HF energy with respect to the classical nuclear coordinates are also available. 1134 Schneider P. E. et al.
J. Chem. Phys.
(2021), 154, pp. 054108.
Link
The Hessians can identify whether the optimized geometries are minima or transition states on the ground state vibronic potential energy surface.

13.5.2.2 NEO-DFT

NEO density functional theory (NEO-DFT) 966 Pak M. V., Chakraborty A., Hammes-Schiffer S.
J. Phys. Chem. A
(2007), 111, pp. 4522.
Link
, 213 Chakraborty A., Pak M. V., Hammes-Schiffer S.
Phys. Rev. Lett.
(2008), 101, pp. 153001.
Link
, 214 Chakraborty A., Pak M. V., Hammes-Schiffer S.
J. Chem. Phys.
(2009), 131, pp. 124115.
Link
is an extension of DFT to multicomponent systems within the NEO framework. The Hohenberg-Kohn theorems have been extended to multicomponent systems, where the reference is expressed as the product of electronic and nuclear Slater determinants composed of Kohn-Sham orbitals. The NEO-DFT total energy is

E[ρe,ρp]=Eext[ρe,ρp]+Eref[ρe,ρp]+Eexc[ρe]+Epxc[ρp]+Eepc[ρe,ρp]. (13.39)

In this equation, Eext[ρe,ρp] is the interaction of the electronic and protonic densities with the external potential created by the classical nuclei, and Eref[ρe,ρp] contains the electron–electron, proton–proton, and electron-proton classical Coulomb energies, as well as the noninteracting kinetic energies of the quantum particles. The terms Eexc[ρe], Epxc[ρp], and Eepc[ρe,ρp] are the electron-electron exchange-correlation functional, the proton-proton exchange-correlation functional, and the electron-proton correlation functional, respectively. The quantities

ρe(𝐫1e) =2i=1Ne/2|ψie(𝐫1e)|2 (13.40a)
ρp(𝐫1p) =I=1Np|ψIp(𝐫1p)|2 (13.40b)

are the electron and proton densities, respectively, and ψie(𝐫1e) and ψIp(𝐫1p) are the electronic and protonic Kohn-Sham spatial orbitals, respectively. These orbitals are obtained by solving two sets of coupled Kohn-Sham equations for the electrons and quantum protons:

(-122+veffe(𝐫1e))ψie =ϵieψie (13.41a)
(-12mp2+veffp(𝐫1p))ψIp =ϵIpψIp. (13.41b)

The effective potentials veff and veff are obtained by taking the derivative of the total energy expression in Eq. (13.39) with respect to electron density and proton density, respectively. Analogous to NEO-HF, these electronic and protonic Kohn-Sham orbitals are expanded as linear combinations of electronic or protonic Gaussian basis functions (ϕμe(𝐫e) and ϕμp(𝐫p)). The extension to open-shell electron systems is analogous to the NEO-UHF method.

The practical implementation of the NEO-DFT method requires an electron-electron exchange-correlation functional, a proton-proton exchange-correlation functional, and an electron-proton correlation functional. Any conventional electron-electron exchange-correlation functional can be used within the NEO-DFT framework. 150 Brorsen K. R., Schneider P. E., Hammes-Schiffer S.
J. Chem. Phys.
(2018), 149, pp. 044110.
Link
Because the proton-proton exchange and correlation are negligible in molecular systems, only the exchange at the NEO-Hartree-Fock level is included to eliminate self-interaction error in the NEO-DFT method. A suitable electron-proton correlation functional is essential for obtaining accurate proton densities and energies, and the epc17-2 1412 Yang Y. et al.
J. Chem. Phys.
(2017), 147, pp. 114113.
Link
, 151 Brorsen K. R., Yang Y., Hammes-Schiffer S.
J. Phys. Chem. Lett.
(2017), 8, pp. 3488.
Link
and epc19 1258 Tao Z., Yang Y., Hammes-Schiffer S.
J. Chem. Phys.
(2019), 151, pp. 124102.
Link
functionals are designed to achieve this goal. These two functionals are based on the multicomponent extension of the Colle-Salvetti formalism. The epc17-2 functional is of the local density approximation (LDA) type with the functional form:

Eepc[ρe,ρp]=-𝑑𝐫ρe(𝐫)ρp(𝐫)a-b[ρe(𝐫)ρp(𝐫)]1/2+cρe(𝐫)ρp(𝐫). (13.42)

The epc19 functional is its multicomponent generalized gradient approximation (GGA) extension that depends on the electron and proton density gradients and is of the form:

Eepc[ρe,ρp,^ρe,^ρp]=-d𝐫ρe(𝐫)ρp(𝐫)a-b[ρe(𝐫)ρp(𝐫)]1/2+cρe(𝐫)ρp(𝐫)×{1-d([ρe(𝐫)ρp(𝐫)]-1/3(1+mp)2[mp2^2ρe(𝐫)ρe(𝐫)-2mp^ρe(𝐫)^ρp(𝐫)ρe(𝐫)ρp(𝐫)+^2ρp(𝐫)ρp(𝐫)])exp[-k[ρe(𝐫)ρp(𝐫)]1/6]}. (13.43)

In addition to the parameters a, b, and c in the epc17-2 functional, 151 Brorsen K. R., Yang Y., Hammes-Schiffer S.
J. Phys. Chem. Lett.
(2017), 8, pp. 3488.
Link
the epc19 functional 1258 Tao Z., Yang Y., Hammes-Schiffer S.
J. Chem. Phys.
(2019), 151, pp. 124102.
Link
has the d and k parameters and also depends on the proton mass mp. Analogous to the NEO-HF analytical energy gradients, the NEO-DFT analytical gradients 1257 Tao Z. et al.
J. Chem. Theory Comput.
(2021), 17, pp. 5110.
Link
are available for these two functionals, allowing geometry optimizations on the ground state vibronic potential energy surface. The NEO-DFT analytical Hessians 1257 Tao Z. et al.
J. Chem. Theory Comput.
(2021), 17, pp. 5110.
Link
are available for the epc17-2 functional or when no electron-proton correlation functional is used and allow characterization of the stationary points.

13.5.2.3 NEO-MSDFT

The NEO multistate DFT (NEO-MSDFT) method was developed to describe hydrogen transfer and hydrogen tunneling systems within the NEO framework. 1424 Yu Q., Hammes-Schiffer S.
J. Phys. Chem. Lett.
(2020), 11, pp. 10106.
Link
, 307 Dickinson J. A., Yu Q., Hammes-Schiffer S.
J. Phys. Chem. Lett.
(2023), 14, pp. 6170.
Link
Analogous to the conventional electronic MSDFT method, 894 Mo Y., Bao P., Gao J.
Phys. Chem. Chem. Phys.
(2011), 13, pp. 6760.
Link
, 396 Gao J. et al.
J. Phys. Chem. Lett.
(2016), 7, pp. 5143.
Link
, 461 Grofe A. et al.
J. Chem. Theory Comput.
(2017), 13, pp. 1176.
Link
, 815 Lu Y., Gao J.
J. Phys. Chem. Lett.
(2022), 13, pp. 7762.
Link
the NEO-MSDFT method linearly combines localized NEO-DFT states in a nonorthogonal configuration interaction (NOCI) scheme. 1177 Skone J. H., Pak M. V., Hammes-Schiffer S.
J. Chem. Phys.
(2005), 123, pp. 134108.
Link
This approach captures the delocalized, bilobal proton densities needed to describe hydrogen tunneling and avoids complications of local minima in orbital space.

Consider a system with N transferring protons, where each proton moves in a double-well potential in the Born-Oppenheimer picture. In NEO-DFT calculations, the protonic density tends to localize in one well of a symmetric or nearly symmetric double-well potential instead of delocalizing over both wells. Quantizing each transferring proton with the NEO approach leads to 2N diabatic NEO-DFT states. Each diabatic state has the protonic density of each transferring proton localized in one of the two wells of its corresponding double-well potential. In practice, higher-energy diabatic states can be excluded from the NOCI expansion. The set of all diabatic NEO-DFT states is {|Ψ~0,|Ψ~1,|Ψ~n}, where n=2N-1. Each diabatic state is the product of a Kohn-Sham electronic and protonic determinant, as discussed in Section 13.5.2.2.

The adiabatic NEO-MSDFT states {|Ψ0,|Ψ1,|Ψn} are linear combinations of all diabatic NEO-DFT states:

|Ψ0 =D00|Ψ~0+D10|Ψ~1++Dn0|Ψ~n (13.44)
|Ψ1 =D01|Ψ~0+D11|Ψ~1++Dn1|Ψ~n
              
|Ψn =D0n|Ψ~0+D1n|Ψ~1++Dnn|Ψ~n

The coefficients in Eq. (13.44) are determined by solving the 2N×2N matrix eigenvalue problem

𝐇𝐃=𝐒𝐃𝐄. (13.45)

The overlap matrix 𝐒 and effective Hamiltonian matrix 𝐇 contain the overlap and couplings, respectively, between pairs of localized diabatic states. Note that the diagonal elements of 𝐒 are unity, and the diagonal elements of 𝐇 are the NEO-DFT energies of the diabatic states. For the sake of brevity, the analytical forms of the off-diagonal terms of these matrices are excluded here but are given in previous work. 307 Dickinson J. A., Yu Q., Hammes-Schiffer S.
J. Phys. Chem. Lett.
(2023), 14, pp. 6170.
Link

The limitations of the epc functionals and the resulting inaccuracies of the overlap between two localized NEO-DFT states, as well as the approximate form of the off-diagonal Hamiltonian matrix elements, can be accounted for by applying a simple correction function to the off-diagonal elements of the 𝐒 matrix:

Sij=α(Sij)β. (13.46)

Users can control the values of α and β used in NEO-MSDFT calculations, and their default values were determined though a fitting process discussed in our previous work. 1424 Yu Q., Hammes-Schiffer S.
J. Phys. Chem. Lett.
(2020), 11, pp. 10106.
Link
Setting both these parameters to unity produces the original overlap matrix. However, in general the alpha and beta parameters should be kept at their default values unless another parameterization procedure is performed to determine alternative values.

A NEO-MSDFT calculation can be enabled by setting the $rem variable NEO_MSDFT = 1. Analytical gradients have also been implemented, 1426 Yu Q., Schneider P. E., Hammes-Schiffer S.
J. Chem. Phys.
(2022), 156, pp. 114115.
Link
allowing users to optimize geometries on specified NEO-MSDFT surfaces by setting JOBTYPE = OPT. By default, analytical derivative couplings between NEO-MSDFT states are computed directly following an analytical gradient calculation, 1425 Yu Q., Roy S., Hammes-Schiffer S.
J. Chem. Theory Comput.
(2022), 18, pp. 7132.
Link
but this calculation can be disabled. Note that the only gradients and derivative couplings printed out are those pertaining to the ground and first-excited NEO-MSDFT states, as these are the only states of physical relevance (as explained elsewhere 307 Dickinson J. A., Yu Q., Hammes-Schiffer S.
J. Phys. Chem. Lett.
(2023), 14, pp. 6170.
Link
). Users can also request a semi-numerical Hessian calculation by setting IDERIV = 2. Further control of a NEO-MSDFT calculation is given by the $neo_msdft section, whose job-control options are given below:

OPT_STATE
       Controls which NEO-MSDFT state is geometry-optimized when JOBTYPE = OPT or which NEO-MSDFT surface the semi-numerical Hessian will be calculated on when IDERIV = 2.
INPUT SECTION: $neo_msdft
TYPE:
       INTEGER
DEFAULT:
       0 Ground-State
OPTIONS:
       n Indicates optimization/semi-numerical Hessian calculation will occur on the nth NEO-MSDFT surface.
RECOMMENDATION:
       Ensure that n is strictly less than the number of diabatic states included in the adiabatic state expansion of Eq. (13.44).

ALPHA
       Sets the α parameter used for the overlap correction of Eq. (13.46).
INPUT SECTION: $neo_msdft
TYPE:
       FLOAT
DEFAULT:
       0.0604
OPTIONS:
       User-defined.
RECOMMENDATION:
       Keep default value unless another parameterization procedure is performed.

BETA
       Sets the β parameter used for the overlap correction of Eq. (13.46).
INPUT SECTION: $neo_msdft
TYPE:
       FLOAT
DEFAULT:
       0.492
OPTIONS:
       User-defined.
RECOMMENDATION:
       Keep default value unless another parameterization procedure is performed.

NACV
       Controls if analytical derivative couplings are calculated following an analytical gradient calculation.
INPUT SECTION: $neo_msdft
TYPE:
       INTEGER
DEFAULT:
       1 Enables derivative coupling calculation.
OPTIONS:
       1 Enables derivative coupling calculation. 0 Disables derivative coupling calculation.
RECOMMENDATION:
       None.

DENPLT
       Controls the generation of proton density cube files for NEO-MSDFT states.
INPUT SECTION: $neo_msdft
TYPE:
       INTEGER
DEFAULT:
       -1 Disables generation of proton density cube files.
OPTIONS:
       n Enables generation of proton density cube files for NEO-MSDFT states {0,,n}.
RECOMMENDATION:
       Users can also generate electron density cube files for each of the specified NEO-MSDFT states via the $plots section (refer to Section 10.5.4.1 for details). Also note that if n>0, a directory neo_msdft_denplt will be created in the current working directory where all the generated cube files will be written.

CPSCF_THRESH
       Controls the convergence criteria for the NEO-CPSCF routine. Solving the NEO-CPSCF equations is required for calculating the analytical gradients of the NEO-MSDFT energies. 1426 Yu Q., Schneider P. E., Hammes-Schiffer S.
J. Chem. Phys.
(2022), 156, pp. 114115.
Link
The NEO-CPSCF routine is considered converged when the error is less than 10-CPSCF_THRESH.

INPUT SECTION: $neo_msdft
TYPE:
       INTEGER
DEFAULT:
       8
OPTIONS:
       User-defined.
RECOMMENDATION:
       Tightening the NEO-CPSCF convergence will improve the reliability of the analytical gradients.

CPSCF_NSTEPS
       Controls the maximum number of NEO-CPSCF iterations permitted. Solving the NEO-CPSCF equations is required for calculating the analytical gradients of the NEO-MSDFT energies. 1426 Yu Q., Schneider P. E., Hammes-Schiffer S.
J. Chem. Phys.
(2022), 156, pp. 114115.
Link
The NEO-CPSCF routine will fail if it does not converge within a CPSCF_NSTEPS number of steps.

INPUT SECTION: $neo_msdft
TYPE:
       INTEGER
DEFAULT:
       300
OPTIONS:
       User-defined.
RECOMMENDATION:
       If CPSCF_THRESH is higher than the default, increasing the number of iterations permitted in the NEO-CPSCF routine may be needed.

FDIFF_STEP
       Sets the distance each center is perturbed if a semi-numerical Hessian calculation is requested.
INPUT SECTION: $neo_msdft
TYPE:
       FLOAT
DEFAULT:
       10-4 Å
OPTIONS:
       User-defined.
RECOMMENDATION:
       If INPUT_BOHR = FALSE, then the step size input by the user will be measured in Å. If INPUT_BOHR = TRUE, then the step size input by the user will be measured in bohr.

As discussed in previous work, 1424 Yu Q., Hammes-Schiffer S.
J. Phys. Chem. Lett.
(2020), 11, pp. 10106.
Link
, 307 Dickinson J. A., Yu Q., Hammes-Schiffer S.
J. Phys. Chem. Lett.
(2023), 14, pp. 6170.
Link
each quantum transferring proton in NEO-MSDFT is given two basis function centers: one center localized near the donor and another center localized near the acceptor. When requesting a NEO-MSDFT calculation, in the $molecule  section, each quantum transferring proton must have one center input as a standard H atom center and its other center input as a ghost H atom center. These centers must be listed consecutively in the $molecule section. Additionally, it is assumed that all ghost H atom centers in the $molecule section of a NEO-MSDFT calculation belong to a quantum transferring proton.

By default, all possible 2N NEO-DFT diabatic states are included in the adiabatic state expansion of Eq. (13.44). If this is not desired, users can provide an input to the $neo_msdft_diabat_control section to specify which diabatic states to include. This input section takes in an array of Boolean values (input as ones or zeros), where each row is a unique string corresponding to a particular diabatic state. These Booleans refer to whether or not a quantum transferring proton is localized on each proton basis function center in the order in which they appear in the $molecule section. Thus, if the number of diabatic states the user wants to include is m and the number of quantum transferring protons is N, the $neo_msdft_diabat_control section should contain an array with m rows and 2N columns, where the i-th row specifies which proton basis function centers have a localized quantum proton in the i-th diabatic state. Note that other protons can be quantized with NEO that do not correspond to a transferring proton and therefore are represented by a single proton basis function center; these protons should not be included in the $neo_msdft_diabat_control section. Examples showing how to set up NEO-MSDFT calculations/input files with both one and multiple transferring protons can be found in Section 13.5.4.

The overlap correction of Eq. (13.46) is parameterized to produce accurate proton densities and tunneling splittings when the proton basis function centers are optimized for separate NEO-DFT diabatic states (see the procedures explained in our previous work). 1424 Yu Q., Hammes-Schiffer S.
J. Phys. Chem. Lett.
(2020), 11, pp. 10106.
Link
, 307 Dickinson J. A., Yu Q., Hammes-Schiffer S.
J. Phys. Chem. Lett.
(2023), 14, pp. 6170.
Link
However, users may wish to perform geometry optimizations on NEO-MSDFT surfaces instead. One can either optimize the positions of just the proton basis function centers for fixed geometries of classical nuclei or optimize the positions of both the classical nuclei and proton basis function centers simultaneously. Both cases require the user to set JOBTYPE = OPT, but in the former case, the $opt section must be used to freeze the classical nuclei during the optimization (see Section 9.4.3 for details). In either case, the input of the $molecule section serves as an initial guess for the optimization, so users must make an informed guess as to where to initially place each of the two proton basis function centers for each quantum transferring proton.

13.5.2.4 NEO-PCM

Bulk solvent effects can be directly incorporated into NEO calculations through the application of various implicit solvation models (Section 11.2) within the NEO framework. 1367 Wildman A. et al.
J. Chem. Theory Comput.
(2022), 18, pp. 1340.
Link
The polarizable continuum model (PCM) constitutes one family of implicit solvation models and itself encompasses several different formulations: 716 Lange A. W., Herbert J. M.
Chem. Phys. Lett.
(2011), 509, pp. 77.
Link
, 529 Herbert J. M.
Wiley Interdiscip. Rev.: Comput. Mol. Sci.
(2021), 11, pp. e1519.
Link
C-PCM, 1282 Truong T. N., Stefanovich E. V.
Chem. Phys. Lett.
(1995), 240, pp. 253.
Link
, 72 Barone V., Cossi M.
J. Phys. Chem. A
(1998), 102, pp. 1995.
Link
IEF-PCM, 228 Chipman D. M.
J. Chem. Phys.
(2000), 112, pp. 5558.
Link
, 171 Cancès E., Mennucci B.
J. Chem. Phys.
(2001), 114, pp. 4744.
Link
SS(V)PE, 228 Chipman D. M.
J. Chem. Phys.
(2000), 112, pp. 5558.
Link
etc. 529 Herbert J. M.
Wiley Interdiscip. Rev.: Comput. Mol. Sci.
(2021), 11, pp. e1519.
Link
In the PCM approach, the solute molecule is placed in a cavity that is embedded in dielectric continuum solvent, and the cavity surface is discretized into i tesserae grid points. The solvent response is represented by a partial charge qi centered at each tesserae grid point 𝐬i. 1270 Tomasi J., Mennucci B., Cammi R.
Chem. Rev.
(2005), 106, pp. 2999.
Link
, 529 Herbert J. M.
Wiley Interdiscip. Rev.: Comput. Mol. Sci.
(2021), 11, pp. e1519.
Link

For each SCF iteration, the current electronic and protonic densities, together with the fixed classical nuclei, define the solute’s charge distribution. This charge distribution gives rise to the solute’s electrostatic potential Vi at each tesserae grid point:

Vi=AZA|𝐑A-𝐬i|-μνPμνe𝑑𝐫eϕμe(𝐫e)ϕνe(𝐫e)|𝐫e-𝐬i|+μνPμνp𝑑𝐫pϕμp(𝐫p)ϕνp(𝐫p)|𝐫p-𝐬i|. (13.47)

The solute electrostatic potential is then used to compute qi using standard PCM methods. Once obtained, the set of tesserae charges is included as an additional one-electron (one-proton) contribution to the electronic (protonic) Fock or analogous Kohn-Sham matrix:

Fμνe,solv =Fμνe,0-iqi𝑑𝐫eϕμe(𝐫e)ϕνe(𝐫e)|𝐫e-𝐬i| (13.48a)
Fμνp,solv =Fμνp,0+iqi𝑑𝐫pϕμp(𝐫p)ϕνp(𝐫p)|𝐫p-𝐬i|, (13.48b)

where Fμνe,0 and Fμνp,0 refer to the gas-phase, electronic [Eq. (13.37a)] and protonic [Eq. (13.37b)] Fock or analogous Kohn-Sham matrix elements, respectively.

NEO-PCM calculations involve iterative, self-consistent convergence of the nuclear-electronic wavefunction in the presence of the dielectric continuum solvent. Both NEO-HF and NEO-DFT PCM energies and analytic gradients are implemented. The calculation can be invoked by setting SOLVENT_METHOD = PCM in the $rem input section, alongside variables for NEO-SCF as described in Section 13.5.3.

In the simplest approach, the cavity surface is discretized into point charges. However, a more sophisticated approach utilizing Gaussian-smeared charges is also supported. 713 Lange A. W., Herbert J. M.
J. Phys. Chem. Lett.
(2010), 1, pp. 556.
Link
, 714 Lange A. W., Herbert J. M.
J. Chem. Phys.
(2010), 133, pp. 244111.
Link
, 711 Lange A. W. et al.
Mol. Phys.
(2020), 118, pp. e1644384.
Link
, 529 Herbert J. M.
Wiley Interdiscip. Rev.: Comput. Mol. Sci.
(2021), 11, pp. e1519.
Link
Selection of these various PCM schemes and related variables can be set in the $pcm and/or $solvent input sections. An example on how to set up a solvated NEO calculation can be found in Section 13.5.4.

13.5.2.5 CNEO

In NEO-SCF calculations, the nuclear solution corresponds to the variational nuclear wavefunction (density) for the specific method and basis sets. The nuclear orbitals in molecular systems are localized, and therefore the quantized nuclei can be treated as distinguishable when analyzing their position expectation values. These expectation values are generally not equal to the input coordinates of the NEO nuclear basis function centers. Control over the position expectation values of the nuclei can be achieved with the constrained (C)NEO approach, 1402 Xu X., Yang Y.
J. Chem. Phys.
(2020), 152, pp. 084107.
Link
based on the original NEO method 1335 Webb S. P., Iordanov T., Hammes-Schiffer S.
J. Chem. Phys.
(2002), 117, pp. 4106.
Link
and in analogy to the conventional constrained DFT (Section 5.11). This is accomplished by imposing the following constraint:

ψIp|𝐫1p|ψIp=𝐑I (13.49)

where 𝐑I is the chosen value for the position expectation value for nucleus I. For geometry optimizations and vibrational frequency calculations, typically 𝐑I is the position of the NEO nuclear basis function center, as defined by the input coordinates. The constraint is imposed by using Lagrange multipliers. In this case, the protonic part of the coupled Kohn-Sham equations of NEO-DFT (Eq. (13.41b)) becomes:

(-12mp2+veffp(𝐫1p)+𝐟I𝐫1p)ψIp=ϵIpψIp (13.50)

where 𝐟I is the Lagrange multiplier. In practice, the value of 𝐟I is optimized numerically to satisfy the constraint in Eq. (13.49).

CNEO calculations can be invoked by setting CNEO = TRUE, and NEO = TRUE must also be set. Analytical CNEO gradients 1403 Xu X., Yang Y.
J. Chem. Phys.
(2020), 153, pp. 074106.
Link
and Hessians 1404 Xu X., Yang Y.
J. Chem. Phys.
(2021), 154, pp. 244110.
Link
are available, allowing geometry optimizations with JOBTYPE = OPT and vibrational frequency calculations with JOBTYPE = FREQ. CNEO can be used with Hartree-Fock or with DFT in conjunction with the epc17-2 or epc19 electron-proton correlation functional and the electron exchange-correlation functionals available in Q-Chem. CNEO is currently implemented with only one quantum nucleus.

13.5.2.6 LR-NEO-TDDFT

The LR-NEO-TDDFT method 1413 Yang Y., Culpitt T., Hammes-Schiffer S.
J. Phys. Chem. Lett.
(2018), 9, pp. 1765.
Link
is a multicomponent extension of the linear response TDDFT method within the NEO framework. It allows the simultaneous calculation of the electronic and proton vibrational excitation energies. In the (LR)-NEO-TDDFT method, the linear response of the NEO Kohn-Sham system to perturbative external fields is computed. The NEO-TDDFT working equation is

[𝐀e𝐁e𝐂𝐂𝐁e𝐀e𝐂𝐂𝐂T𝐂T𝐀p𝐁p𝐂T𝐂T𝐁p𝐀p][𝐗e𝐘e𝐗p𝐘p]=ω[𝐈0000-𝐈0000𝐈0000-𝐈][𝐗e𝐘e𝐗p𝐘p] (13.51)

where

Aia,jbe =(ϵa-ϵi)δabδij+aj|ib+δ2EexcδPjbeδPaie+δ2EepcδPjbeδPaie (13.52)
Bia,jbe =ab|ij+δ2EexcδPjbeδPiae+δ2EepcδPjbeδPiae (13.53)
AIA,JBp =(ϵA-ϵI)δABδIJ+AJ|IB+δ2EpxcδPJBpδPAIp+δ2EepcδPJBpδPAIp (13.54)
BIA,JBp =AB|IJ+δ2EpxcδPJBpδPIAp+δ2EepcδPJBpδPIAp (13.55)
Cia,JB =-aB|iJ+δ2EepcδPJBpδPaie (13.56)

Here, the occupied electronic orbitals are denoted with indices i and j, whereas the unoccupied electronic orbitals are denoted with indices a and b. The analogous upper case indices denote protonic orbitals. The solution of Eq. (13.51) provides the electronic and proton vibrational excitation energies ω, as well as the transition excitation and de-excitation amplitudes, 𝐗 and 𝐘, respectively. Analogous to the TDDFT method, the Tamm-Dancoff approximation (TDA) can be imposed within the NEO framework, defining the NEO-TDDFT-TDA method that is represented by

[𝐀e𝐂𝐂T𝐀p][𝐗e𝐗p]=ω[𝐗e𝐗p]. (13.57)

The extension of the NEO-TDDFT and NEO-TDDFT-TDA approaches to open-shell electron systems is straightforward. 271 Culpitt T. et al.
J. Chem. Phys.
(2019), 150, pp. 201101.
Link
NEO-TDHF and NEO-CIS have similar forms as NEO-TDDFT and NEO-TDA without electron-proton, electron-electron, or proton-proton correlation. The analytical gradients for NEO-CIS/NEO-TDA/NEO-TDHF/NEO-TDDFT are available, 1257 Tao Z. et al.
J. Chem. Theory Comput.
(2021), 17, pp. 5110.
Link
enabling geometry optimizations on the excited state vibronic potential energy surfaces. For NEO-TDA and NEO-TDDFT, analytical gradients are available for the epc17-2 functional or when no electron-proton correlation functional is used. The transition densities can be analyzed to determine the percentages of electronic and protonic character for each vibronic excited state.

13.5.2.7 RT-NEO-TDDFT

Real-time NEO time-dependent density functional theory (RT-NEO-TDDFT, or RT-NEO for brevity) is a multicomponent extension of conventional electronic RT-TDDFT within the NEO framework 1440 Zhao L. et al.
J. Phys. Chem. Lett.
(2020), 11, pp. 4052.
Link
and is an approach that enables the modeling of nonequilibrium, non-Born-Oppenheimer nuclear-electronic quantum dynamics. This approach assumes the form of the single Slater determinant product ansatz of the NEO-DFT reference system, where 𝐱e and 𝐱p are collective spatial and spin coordinates of the quantum electrons and protons, 𝐏e (or 𝐏p) is the single particle density matrix and 𝐅e (or 𝐅p) is the Kohn-Sham matrix for the electrons (or protons), and t is the time. The time-dependent Schrödinger equation

itΨNEO(𝐱e , 𝐱p ; t)=H(𝐱e , 𝐱p ; t)ΨNEO(𝐱e , 𝐱p ; t) (13.58)

can be propagated according to the set of multicomponent von Neumann equations:

it𝐏e(t)=[𝐅e(t,𝐏e(t),𝐏p(t)),𝐏e(t)] (13.59a)
it𝐏p(t)=[𝐅p(t,𝐏p(t),𝐏e(t)),𝐏p(t)] (13.59b)

Note that these equations are not propagated independently, but are actually coupled through Fock terms that depend on both particle densities. The electron and proton densities are evolved together in time. The resulting electronic and protonic dipole moments can be obtained at every time step and analyzed via post-processing to provide relevant spectral features. Fourier transformation of these signals from the corresponding time-domain to the frequency-domain can yield both electronic and vibrational spectra that have been shown to agree well with LR-NEO-TDDFT. In addition, this approach provides the nonequilbrium, real-time dynamics of the electronic and protonic densities.

Up to this point, the equations of motion presented above [Eq. (13.59a) and Eq. (13.59b) assume that the classical nuclei are held at a fixed geometry throughout the entire simulation. To combine the real-time dynamics of the electrons and protons together with the mean-field motion of the classical nuclei, the RT-NEO-TDDFT-Ehrenfest approach has been formulated. 1441 Zhao L. et al.
J. Chem. Phys.
(2020), 153, pp. 224111.
Link
In this approach, the classical nuclei are evolved according to Newtonian dynamics, while the nonequilibrium electrons and protons are still being propagated according to Eq. (13.59a) and Eq. (13.59b), respectively:

MA𝐑¨A=-A𝐸 (13.60)

Here we denote MA and 𝐑A to be the mass and the position, respectively, of the Ath classical nucleus, and A𝐸 is the abbreviated form of the total nuclear gradient expression. We support both of these approaches, i.e., RT-NEO with fixed classical nuclei and RT-NEO with moving classical nuclei, and we have designated those methods as METHOD = REALTIME and METHOD = EHRENFEST, respectively.

In addition to electrons, it is important that the proton dynamics is adequately described throughout the course of a given trajectory. This is especially important in cases where proton delocalization may start to occur, such as during a proton transfer reaction or when the geometry of the classical nuclei is no longer held fixed. To provide the flexibility for describing proton delocalization, one may use a larger protonic basis set and/or utilize multiple fixed proton basis (FPB) function centers placed at pre-designated positions. Alternatively, a traveling proton basis (TPB) approach has been formulated and will be made available in a future release. 1441 Zhao L. et al.
J. Chem. Phys.
(2020), 153, pp. 224111.
Link
Examples of using either one or multiple FPB function centers can be found in the examples section of this manual.

In certain applications, the requirement of using a small time step due to the fast electronic dynamics can prohibit long simulation times. For systems that are electronically adiabatic, the electronic Born-Oppenheimer (BO) approximation can be invoked by quenching the electrons to an instantaneous electronic ground state via the standard self-consistent-field (SCF) procedure instead of propagating Eq. (13.59a) at every time step:

𝐏e(t)=SCF[𝐏p(t),{𝐑A(t)}] (13.61)

This approach enables the use of a time step that is several orders of magnitude larger than that required for integrating Eq. (13.59a), therefore rendering longer simulation times more tractable. The electronic BO approximation can be invoked for RT-NEO with fixed classical nuclei or for RT-NEO with moving classical nuclei by toggling the methods METHOD = BO-REALTIME or METHOD = BO-EHRENFEST, respectively.

A RT-NEO calculation can be invoked by setting NEO_TDKS = TRUE in the $rem section, and additional job control is provided in its own input section. The input file for a RT-NEO propagation is illustrated in several examples in Section 13.5.4, however, the available controls are also discussed below. Note that the keywords and field types defined for $tdks (Section 7.4) are not the same parameters as those defined for the $neo_tdks, and vice versa. Only parameters/keywords documented below are supported. The parameters of the jobs are set using options specified in the $neo_tdks input section. The format of the $neo_tdks section is analogous to the $rem section:

$neo_tdks
  <Keyword>  <parameter/option>
$end

Note:  The following job control variables belong only in the $neo_tdks  section. Do not place them in the $rem section.

METHOD
       Specifies which RT-NEO dynamics approach will be used to propagate the system.
INPUT SECTION: $neo_tdks
TYPE:
       STRING
DEFAULT:
       NONE
OPTIONS:
       Realtime Fixed classical nuclei 1440 Zhao L. et al.
J. Phys. Chem. Lett.
(2020), 11, pp. 4052.
Link
Ehrenfest Moving classical nuclei 1441 Zhao L. et al.
J. Chem. Phys.
(2020), 153, pp. 224111.
Link
BO-Realtime Fixed classical nuclei with electronic BO approximation 770 Li T. E., Hammes-Schiffer S.
J. Chem. Phys.
(2023), 158, pp. 114118.
Link
BO-Ehrenfest Moving classical nuclei with electronic BO approximation 770 Li T. E., Hammes-Schiffer S.
J. Chem. Phys.
(2023), 158, pp. 114118.
Link

RECOMMENDATION:
       None.

DT
       Specifies the time step Δt, in atomic units.
INPUT SECTION: $neo_tdks
TYPE:
       DOUBLE
DEFAULT:
       0.04
OPTIONS:
       Δt Δt>0 User-selected.
RECOMMENDATION:
       None.

MAXITER
       Specifies the maximum number of time steps to simulate.
INPUT SECTION: $neo_tdks
TYPE:
       INTEGER
DEFAULT:
       NONE
OPTIONS:
       n n>0 User-selected.
RECOMMENDATION:
       The total propagation length is Δt×𝐌𝐀𝐗𝐈𝐓𝐄𝐑.

ELECTRONIC_HOMO_TO_LUMO
       Performs a HOMO LUMO excitation for a single electron prior to time propagation. An unrestricted calculation is performed when this keyword is invoked.
INPUT SECTION: $neo_tdks
TYPE:
       LOGICAL
DEFAULT:
       FALSE
OPTIONS:
       TRUE FALSE
RECOMMENDATION:
       Should not use in conjunction with the electronically adiabatic approximation.

To perturb the system, a time-dependent electric field pulse 𝓔(t) can be applied via a dipole coupling term added to the electronic and/or protonic Fock matrices,

𝐅e =𝐅e,0-𝝁e𝓔(t) (13.62a)
𝐅p =𝐅p,0-𝝁p𝓔(t), (13.62b)

where 𝐅e,0 (or 𝐅p,0) is the unperturbed Kohn-Sham matrix and 𝝁e (or 𝝁p) is the dipole moment matrix for the electrons (or protons). Note that the dipole moment (i.e., 𝝁e and 𝝁p) have the opposite sign for electrons and protons. Tunable parameters for FIELD_TYPE are provided in the additional keywords found below. For the supported field types below, note that the electric field vector 𝓔=(x,y,z) is a quantity whose magnitude is controlled by FIELD_AMP and whose direction is controlled by FIELD_DIRECTION.

  1. 1.

    Delta simulates a Dirac δ-function kick with a field that is turned on only at time t0=0:

    𝓔(t)=𝐀0δ(t-t0) (13.63)
    • The amplitude of 𝐀0 and its direction are set using FIELD_AMP and FIELD_DIRECTION, respectively.

  2. 2.

    Gaussian simulates the following impulse field that is turned on at all times t0:

    𝓔(t)=𝐀0exp((t-tpeak)2τ2)sin(ωt) (13.64)
    • The amplitude of 𝐀0 and its direction are set using FIELD_AMP and FIELD_DIRECTION, respectively.

    • The center of the pulse tpeak is set using the FIELD_PEAK keyword.

    • The pulse half-width τ is set using the FIELD_TAU keyword.

    • The pulse frequency ω is set using the FIELD_FREQUENCY keyword.

FIELD_TYPE
       The external applied field
INPUT SECTION: $tdks
TYPE:
       STRING
DEFAULT:
       NONE
OPTIONS:
       DELTA δ-function kick GAUSSIAN Impulse field (Gaussian envelope) NONE No field
RECOMMENDATION:
       No recommendation.

FIELD_AMP
       The amplitude of the external field, in atomic units.
INPUT SECTION: $tdks
TYPE:
       DOUBLE
DEFAULT:
       0.01
OPTIONS:
       NONE
RECOMMENDATION:
       No recommendation.

FIELD_PEAK
       The peak position tpeak (in atomic units of time) for the Gaussian impulse field.
INPUT SECTION: $tdks
TYPE:
       DOUBLE
DEFAULT:
       0.0
OPTIONS:
       NONE
RECOMMENDATION:
       No recommendation.

FIELD_TAU
       The value of τ (in atomic units of time) for the Gaussian impulse field.
INPUT SECTION: $tdks
TYPE:
       DOUBLE
DEFAULT:
       800.0
OPTIONS:
       NONE
RECOMMENDATION:
       No recommendation.

FIELD_FREQUENCY
       The frequency ω of the external field, in eV units.
INPUT SECTION: $tdks
TYPE:
       DOUBLE
DEFAULT:
       6.0
OPTIONS:
       NONE
RECOMMENDATION:
       No recommendation.

FIELD_DIRECTION
       The direction of the external applied field vector.
INPUT SECTION: $tdks
TYPE:
       STRING
DEFAULT:
       NONE
OPTIONS:
       XYZ Pulse along the direction {1, 1, 1} X Pulse along the direction {1, 0, 0} Y Pulse along the direction {0, 1, 0} Z Pulse along the direction {0, 0, 1} NONE
RECOMMENDATION:
       No recommendation.

FIELD_PARTICLE
       The subsystem on which to apply the external field perturbation.
INPUT SECTION: $tdks
TYPE:
       STRING
DEFAULT:
       NONE
OPTIONS:
       BOTH Electrons and quantized protons ELECTRONS PROTONS
RECOMMENDATION:
       No recommendation.

13.5.2.8 NEO-DFT(V)

Within the NEO framework, select nuclei are treated quantum mechanically at the same level as the electrons. This removes the Born-Oppenheimer separation between the quantum nuclei and the electrons and naturally includes nonadiabatic effects between the quantum nuclei and the electrons. At the same time, quantizing the select nuclei gives rise to a potential energy surface with fewer nuclear degrees of freedom, which prevents a direct calculation of the vibrational frequencies of the entire molecule. Consequently, diagonalization of a coordinate Hessian in the NEO framework yields vibrational frequencies and accompanying normal modes of only the classical nuclei, with the quantum nuclei responding instantaneously to the motion of the classical nuclei. 1134 Schneider P. E. et al.
J. Chem. Phys.
(2021), 154, pp. 054108.
Link
Although the fundamental anharmonic vibrational frequencies of the quantum nuclei can be accurately obtained through LR-NEO-TDDFT, 271 Culpitt T. et al.
J. Chem. Phys.
(2019), 150, pp. 201101.
Link
the couplings between the vibrations of the classical and quantum nuclei are missing. To obtain the fully coupled molecular vibrations, an effective strategy denoted NEO-DFT(V) was developed. 1414 Yang Y. et al.
J. Phys. Chem. Lett.
(2019), 10, pp. 1167.
Link
, 272 Culpitt T. et al.
J. Chem. Theory Comput.
(2019), 15, pp. 6840.
Link
The NEO-DFT(V) method has been shown to incorporate key anharmonic effects in full molecular vibrational analyses and to produce accurate molecular vibrational frequencies compared to experiments. 1414 Yang Y. et al.
J. Phys. Chem. Lett.
(2019), 10, pp. 1167.
Link
, 272 Culpitt T. et al.
J. Chem. Theory Comput.
(2019), 15, pp. 6840.
Link

The NEO-DFT(V) method involves diagonalization of an extended NEO Hessian composed of partial second derivatives of the coordinates of the classical nuclei (𝐫c) and the expectation values of the quantum nuclei (𝐫q). This extended Hessian matrix is composed of three sub-matrices: 𝐇0=(2E/𝐫c2)|𝐫c=𝐫q, 𝐇1=2E/𝐫q𝐫c, and 𝐇2=2E/𝐫q2, where in each case, all other coordinates of the classical nuclei and expectation values of the quantum nuclei are fixed. The extended Hessian has the following structure:

𝐇DFT(V)=[𝐇0𝐇1𝐇1𝐇2] (13.65)

where

𝐇2 =𝐔𝛀𝐌𝐔 (13.66a)
𝐇1 =-𝐇2𝐑 (13.66b)
𝐇0 =𝐇NEO+𝐑𝐇2𝐑. (13.66c)

The quantity 𝐑=d𝐫q/d𝐫c and the NEO Hessian matrix is 𝐇NEO=d2E/d𝐫c2 (without the constraint that the expectation values of the quantum nuclei are fixed). In the expression for the 𝐇2 matrix, 𝐌 is the diagonal mass matrix, and 𝛀 is the diagonal matrix with elements corresponding to the squares of the LR-NEO-TDDFT fundamental vibrational frequencies. 1414 Yang Y. et al.
J. Phys. Chem. Lett.
(2019), 10, pp. 1167.
Link
𝐔 is a unitary matrix that transforms 𝛀 to the target coordinate system and is approximated with the transition dipole moment vectors afforded by a LR-NEO-TDDFT calculation. 272 Culpitt T. et al.
J. Chem. Theory Comput.
(2019), 15, pp. 6840.
Link

Diagonalization of 𝐇DFT(V) produces the fully coupled molecular vibrational frequencies including anharmonic effects associated with the quantum protons. The NEO-DFT(V) method is available for use with the epc17-2 functional or when no electron-proton correlation functional is used. The NEO-HF(V) method, which involves building the extended NEO-Hessian based on the NEO-HF Hessian and inputs from NEO-TDHF, is also available.

13.5.2.9 NEO-CC

An alternative route for inclusion of correlation effects between quantum particles (i.e., electrons and protons) is with wave functions methods that are systematically improvable and parameter-free. 1335 Webb S. P., Iordanov T., Hammes-Schiffer S.
J. Chem. Phys.
(2002), 117, pp. 4106.
Link
, 978 Pavošević F., Culpitt T., Hammes-Schiffer S.
Chem. Rev.
(2020), 120, pp. 4222.
Link
Among the various developed multicomponent wave function methods, the NEO coupled cluster (NEO-CC) methods have been particularly successful. 977 Pavošević F., Culpitt T., Hammes-Schiffer S.
J. Chem. Theory Comput.
(2018), 15, pp. 338.
Link
, 979 Pavošević F., Hammes-Schiffer S.
J. Chem. Phys.
(2019), 151, pp. 074104.
Link
, 980 Pavošević F., Rousseau B. J. G., Hammes-Schiffer S.
J. Phys. Chem. Lett.
(2020), 11, pp. 1578.
Link
, 981 Pavošević F., Tao Z., Hammes-Schiffer S.
J. Phys. Chem. Lett.
(2021), 12, pp. 1631.
Link
The NEO-CC wave function is given by

|ΨNEO-CC=eT^|0e0p, (13.67)

where T^ is the cluster operator that incorporates the correlation effects between quantum particles, and |0e0p is the NEO-HF reference wave function. In the NEO-CCSD method, 977 Pavošević F., Culpitt T., Hammes-Schiffer S.
J. Chem. Theory Comput.
(2018), 15, pp. 338.
Link
the cluster operator is given by

T^=T^1+T^2=taiaia+tAIaIA+14tabijaijab+14tABIJaIJAB+taAiIaiIaA=αtαaα, (13.68)

where aα=aα={aia,aIA,aijab,aIJAB,aiIaA} are the excitation operators expressed in terms of creation/annihilation (ap/ap) fermionic operators, and α is the excitation rank. Here, the i,j, indices denote occupied electronic orbitals, the a,b, indices denote unoccupied electronic orbitals, and the p,q, indices denote general electronic orbitals. The protonic orbitals are denoted analogously using the capitalized indices. The unknown tα wave function parameters (amplitudes) are determined by solving the set of nonlinear equations for each α: 977 Pavošević F., Culpitt T., Hammes-Schiffer S.
J. Chem. Theory Comput.
(2018), 15, pp. 338.
Link

0e0p|aαe-T^1-T^2H^NEOeT^1+T^2|0e0p=0. (13.69)

In this equation, HNEO=hqpapq+12grspqapqrs+hQPapq+12gRSPQaPQRS-gqQpPapPqQ is the second-quantized NEO Hamiltonian, where hqp=q|h^e|p and grspq=rs|pq are conventional electronic core Hamiltonian and two-electron integrals, respectively. The remaining protonic (hQP and gRSPQ) and electron-proton (gqQpP) integrals are defined analogously. Lastly, the NEO-CCSD energy is calculated from

ENEO-CCSD=0e0p|e-T^1-T^2H^NEOeT^1+T^2|0e0p. (13.70)

To increase the computational efficiency and reduce the memory requirements for the NEO-CCSD method, the two-particle integrals can be approximated with the density fitting (DF) approximation, 981 Pavošević F., Tao Z., Hammes-Schiffer S.
J. Phys. Chem. Lett.
(2021), 12, pp. 1631.
Link
in which the two-particle four-center integrals are factorized into a sum of products of three-center and two-center two-particle integrals. In particular, the four-center two-electron integrals are approximated by

(μν|ρσ)=μρ|νσXY(μν|X)(X|Y)-1(Y|ρσ), (13.71)

where (μν|X) and (X|Y) are three-center and two-center two-electron integrals, respectively. In this equation, μ,ν, and X,Y, indices denote electronic and auxiliary electronic basis functions, respectively. The four-center two-proton integrals are approximated analogously by

(μν|ρσ)=μρ|νσXY(μν|X)(X|Y)-1(Y|ρσ), (13.72)

where primed indices denote protonic basis functions and (μν|X) and (X|Y) are three-center and two-center two-proton integrals, respectively. Finally, the four-center electron-proton integrals are approximated as

(μν|μν)=μμ|ννXY(μν|X)(X|Y)-1(Y|μν). (13.73)

By employing the DF approximation, the memory requirements for storing four-center two-particle integrals are reduced from Nbf4 to Nbf2×Naux, where Nbf and Naux are the number of electronic or protonic basis functions and auxiliary basis functions, respectively. 981 Pavošević F., Tao Z., Hammes-Schiffer S.
J. Phys. Chem. Lett.
(2021), 12, pp. 1631.
Link

13.5.2.10 NEO-MP2

Second-order Møller-Plesset perturbation theory provides a useful framework for wave function based correlation effects with lower cost than NEO-CC 1244 Swalina C., Pak M. V., Hammes-Schiffer S.
Chem. Phys. Lett.
(2005), 404, pp. 394.
Link
. The NEO-MP2 correlation energy is composed of the electron-electron, proton-proton and electron-proton contributions:

ENEO(2)=Eee(2)+Epp(2)+Eep(2), (13.74)

where

Eee(2) =14ijab|ij|ab-ij|ba|2εi+εj-εa-εb (13.75a)
Epp(2) =14IJAB|IJ|AB-IJ|BA|2εI+εJ-εA-εB (13.75b)
Eep(2) =iIaA|iI|aA|2εi+εI-εa-εA. (13.75c)

Here, i and j are occupied electronic spin orbitals with energies εi and εj, a and b are virtual (unoccupied) electronic spin orbitals with energies εa and εb, and uppercase indices indicate the analogous protonic spin orbitals and energies.

To provide accuracy that is competitive with NEO-CCSD, NEO-MP2 has been extended to include orbital optimization and empirical spin-component and electron-proton correlation scaling. 980 Pavošević F., Rousseau B. J. G., Hammes-Schiffer S.
J. Phys. Chem. Lett.
(2020), 11, pp. 1578.
Link
, 981 Pavošević F., Tao Z., Hammes-Schiffer S.
J. Phys. Chem. Lett.
(2021), 12, pp. 1631.
Link
This leads to the NEO equivalents of SOS-OOMP2 (O2) 805 Lochan R. C., Head-Gordon M.
J. Chem. Phys.
(2007), 126, pp. 164101.
Link
and SOS-MP2 455 Grimme S.
J. Chem. Phys.
(2003), 118, pp. 9095.
Link
, 619 Jung Y. et al.
J. Chem. Phys.
(2004), 121, pp. 9793.
Link
(Sections 6.6.5 and 6.6.6). The NEO equivalents are referred to as NEO-SOS-OOMP2 and NEO-SOS-MP2 to indicate the inclusion of an electron-proton correlation scaling factor:

ENEO(2)=cssEeess(2)+cosEeeos(2)+cepEep(2)+Epp(2), (13.76)

where Eeess(2) and Eeeos(2) are the same- and opposite-spin parts of the electron-electron correlation energy, which are scaled by css and cos, respectively, and cep is the electron-proton scaling factor.

Orbital optimization is performed by minimizing the NEO-MP2 energy with respect to the unitary operator eX^-X^ formed from the rotation operator X^=X^e+X^p=xaiaia+xAIaIA, with 𝐱={xai,xAI} being the set of unknown electronic and protonic orbital rotation parameters. Optimal orbital rotation parameters are solved for by finding the stationary points of electronic and protonic orbital gradients 𝐰e and 𝐰p, which have elements

(𝐰e)ai=ENEO-MP2(𝐱)xia|𝐱e=0 (13.77a)
and
(𝐰p)AI=ENEO-MP2(𝐱)xIA|𝐱p=0, (13.77b)

and are solved in an alternating fashion until self-consistency is achieved. For computational efficiency, all NEO-MP2 methods are implemented with the density fitting approximation for two-particle integrals, 359 Fetherolf J. H. et al.
J. Phys. Chem. Lett.
(2022), 13, pp. 5563.
Link
as described in more detail in Section 13.5.2.9.

NEO-MP2 methods are invoked with NEO_RIMP2 = 1 to run without orbital optimization and NEO_RIMP2 = 2 to run with orbital optimization. NEO = TRUE must also be set. Spin-component scaling settings are controlled with the SCS variable. Custom css and cos can be set with SSS_FACTOR and SOS_FACTOR, respectively, when SCS = 3 is set (arbitrary scaling). By default, cep is set to the same value as cos, but a custom electron-proton scaling factor can be input with the EP_FACTOR variable, which will always override the default set by SCS.