Coupling between thermochemical contributions of subvalence correlation and of higher-order post-CCSD(T) correlation effects -- a step toward `W5 theory'

This paper investigates the significant coupling between subvalence correlation and higher-order post-CCSD(T) effects on the total atomization energies of first- and second-row molecules, leading to a proposed "W5 theory" protocol that yields revised, highly accurate thermochemical values consistent with Active Thermochemical Tables (ATcT).

The Gist

Imagine you are trying to weigh a very delicate, intricate piece of jewelry (a molecule) to determine its exact value (its energy). For decades, scientists have had a very good, high-tech scale called W4 Theory. It's accurate enough for most everyday jewelry, but when you get to the really complex, heavy, or strange pieces (molecules with second-row atoms like sulfur or phosphorus), the scale starts to show tiny, confusing errors.

This paper is the story of how scientists upgraded that scale to a new, ultra-precise version called "W5 Theory."

Here is the breakdown of what they did, using simple analogies:

1. The Problem: The "Hidden" Weights

In the old W4 method, scientists calculated the energy of a molecule by looking at its "outer" electrons (the ones doing the bonding). They treated the "inner" electrons (the deep core electrons) as if they were just heavy, static weights that didn't move or interact much.

• The Analogy: Imagine a dance floor. The dancers on the floor are the outer electrons. The people sitting in the VIP boxes in the very back are the inner electrons.
• The Old View: The old theory assumed the VIP guests just sat there quietly.
• The New Discovery: The authors found that for certain "row 2" molecules (like those with Sulfur or Phosphorus), the VIP guests aren't just sitting there. They are actually dancing with the people on the floor and even dancing with each other.
  • When you have a molecule with several second-row atoms next to each other (like a chain of Sulfur atoms), these "inner" interactions become huge. Ignoring them is like weighing a suitcase but forgetting to count the heavy books hidden inside the lining.

2. The Fix: Re-measuring the Shape

To get the weight right, you also need to know the exact shape of the object.

• The Analogy: If you try to weigh a crumpled piece of paper, the result is different than if you weigh it flat.
• The Discovery: The authors realized that when you include those "VIP guest" interactions, the molecule's shape actually changes slightly. The bonds get a tiny bit shorter (like a spring compressing).
• The Result: They re-calculated the shape of the molecules using a more complete view of the electrons. This "re-shaping" changed the final energy numbers significantly, especially for complex molecules.

3. The "Spin" Confusion (The Dance Floor Chaos)

Some molecules are "open-shell," meaning they have unpaired electrons. It's like a dance floor where some dancers are spinning wildly.

• The Issue: There were two ways to calculate this: one method (UHF) was like letting the dancers spin freely but sometimes getting them confused (spin contamination), and another (ROHF) was like putting them in strict pairs.
• The Fix: The authors tested both and found that for the messy, spinning molecules, the "strict pair" method (ROCCSD) actually gave a more accurate picture of the bond lengths, even if the free-spinning method seemed easier. They decided to use the stricter, more careful method for the geometry.

4. The "Post-CCSD(T)" Boost

The main calculation (CCSD(T)) is like the main engine of a car. But for the most precise racing, you need to add a turbocharger and a nitrous system.

• The Boost: These are called "higher-order correlations" (like triple and quadruple excitations).
• The Surprise: In the old days, calculating these "turbo" effects for the inner electrons was too expensive for computers. But computers have gotten faster! The authors finally ran these expensive calculations for the inner electrons.
• The Payoff: For simple molecules, the turbo didn't add much speed. But for the complex "Sulfur chains" and "Phosphorus cages," the turbo added a massive amount of power (energy correction), fixing errors that were previously too big to ignore.

5. The Result: W5 Theory

By combining:

1. Re-measuring the shape (geometry),
2. Counting the VIP guests (subvalence correlation),
3. Adding the turbo (higher-order effects),

They created W5 Theory.

Why does this matter?

• Accuracy: They compared their new numbers against the "Active Thermochemical Tables" (ATcT), which is like the gold standard of experimental data. The new W5 numbers match the gold standard almost perfectly, even for tricky elements like Boron, Silicon, and Sulfur.
• Revisions: For some molecules, the old numbers were off by enough to matter in real-world chemistry. The new numbers fix these discrepancies.
• The Future: This paper is a blueprint. It says, "If you want to do chemistry with sub-kJ/mol accuracy (extreme precision) in the future, you must include these inner-electron interactions and re-optimize the shapes."

In a Nutshell

Think of the old method as a high-quality map of a city. It's great for driving from point A to point B. But if you are trying to park a very specific, tiny car in a crowded garage (predicting the exact energy of a complex molecule), the old map has a few inches of error that matter.

This paper draws a new map that accounts for the tiny potholes, the hidden underground tunnels, and the exact curvature of the roads. It's a step toward the ultimate "W5" map, ensuring that when we calculate the energy of molecules, we aren't just guessing—we are knowing.

Technical Summary

Here is a detailed technical summary of the paper "Coupling between thermochemical contributions of subvalence correlation and of higher-order post-CCSD(T) correlation effects — a step toward 'W5 theory'" by Barman et al.

1. Problem Statement

Accurate thermochemistry, particularly for second-row elements (Al–Ar), has reached a point where standard high-accuracy protocols (like W4 theory and HEAT) face limitations. The primary issues addressed in this work are:

• Subvalence Correlation: The neglect of post-CCSD(T) contributions (specifically connected quadruples $(Q)$ and higher-order triples) to the inner-shell (subvalence) correlation energy. While often negligible for first-row molecules, these contributions become significant for second-row species, especially those with adjacent second-row atoms (e.g., $S_4$, $P_4$).
• Reference Geometry Sensitivity: The impact of using geometries optimized with frozen-core CCSD(T) versus those including core-valence (CV) correlation. Previous protocols often froze subvalence electrons during geometry optimization to save computational resources, potentially introducing errors in Total Atomization Energies (TAE).
• Spin Contamination in Radicals: The choice between Unrestricted Hartree-Fock (UHF) and Restricted Open-Shell Hartree-Fock (ROHF) references for open-shell species. UHF is prone to spin contamination, which can affect geometry optimization and energy calculations.
• Benchmark Discrepancies: Discrepancies between previous W4-17 benchmark values and the highly accurate Active Thermochemical Tables (ATcT), particularly for second-row species, necessitating a re-evaluation of theoretical protocols.

2. Methodology

The authors performed high-level ab initio calculations on the W4-08 subset of the W4-17 benchmark (200 species), focusing on first- and second-row molecules.

• Software & Hardware: Calculations utilized development versions of CFOUR, MOLPRO 2024, and MRCC 2024 on the Weizmann Institute's CHEMFARM HPC facility. A significant optimization was implemented in CFOUR to parallelize the weighting of cluster amplitudes by Fock denominators, reducing wall-time by factors of 10–30 for large $(Q)$ steps.
• Basis Sets:
  • Core-Valence: cc-pCVnZ and cc-pwCVnZ ($n=T, Q, 5, 6$) were used for subvalence and core-valence correlation.
  • Valence: Standard cc-pVnZ and augmented variants (aug-cc-pV(n+d)Z) for valence correlation.
  • Extrapolation: Two-point extrapolations (Schwartz and Helgaker formulas) were averaged (the "SuperHEAT" approach) to estimate the Complete Basis Set (CBS) limit.
• Geometry Optimization:
  • Geometries were reoptimized at the CCSD(T)/cc-pwCVQZ level (including core-valence correlation).
  • For open-shell species, a comparison was made between UCCSD(T) and ROCCSD(T). The study concluded that ROCCSD(T) is preferred for radicals with significant spin contamination (e.g., CN, $C_2H$) as it yields geometries closer to the iterative CCSDT limit.
• Energy Components Calculated:
  • Valence Correlation: CCSD(T), higher-order triples $T_3-(T)$, connected quadruples $(Q)$, and the $\Lambda$-coupled cluster corrections (CCSDT(Q)$\Lambda$).
  • Subvalence Correlation: Post-CCSD(T) contributions to the inner-shell energy, calculated at the CCSDT(Q) level using cc-pwCVTZ basis sets.
  • Relativistic Effects: Scalar relativistic corrections using the X2C (Exact Two-Component) method.
  • DBOC: Diagonal Born-Oppenheimer Corrections (DBOC) calculated at the CCSD/cc-pwCVTZ level.
  • Rotational ZPE: Corrections for the "Rotational Zero-Point Energy" in diatomics with spin-orbit splitting (e.g., OH, CH).

3. Key Contributions

• Proposal of "W5 Theory": The authors propose two preliminary protocols, W5prelim1 and W5prelim2, which explicitly include:
  • Core-valence optimized geometries.
  • Post-CCSD(T) subvalence correlation (specifically $(Q)$ and $T_3-(T)$).
  • UHF references for energy calculations but ROCCSD(T) for open-shell geometry optimizations.
  • Scalar relativistic (X2C) and DBOC corrections.
• Quantification of Subvalence Post-CCSD(T): The study demonstrates that while subvalence post-CCSD(T) effects are small for first-row molecules, they are non-trivial for second-row molecules, particularly those with adjacent second-row atoms.
• Resolution of Spin Contamination Issues: The work provides a definitive recommendation to use ROCCSD(T) over UCCSD(T) for geometry optimizations of open-shell species to avoid spin contamination artifacts, showing that ROCCSD(T) geometries are closer to the CCSDT limit.
• Basis Set Efficiency: The authors show that subvalence post-CCSD(T) contributions can be accurately estimated using the cc-pwCVTZ basis set with a scaling factor (1.30), avoiding the prohibitive cost of cc-pwCVQZ for large systems.

4. Key Results

• Magnitude of Corrections:
  • Geometry Shift: Reoptimizing geometries with core-valence correlation generally increases TAEs. The effect is most pronounced for second-row compounds with adjacent atoms (e.g., $P_4$: +0.12 kcal/mol, $SO_3$: +0.11 kcal/mol).
  • Subvalence Post-CCSD(T): The largest contributions come from connected quadruples $(Q)$ in systems with multiple adjacent second-row atoms. For $S_4$, the subvalence $(Q)$ contribution is 0.38 kcal/mol, and for $S_3$, it is 0.23 kcal/mol. In contrast, for species like $SiF_4$ (no adjacent second-row atoms), the effect is negligible.
  • Coupling: There is a significant coupling between geometry shifts and post-CCSD(T) subvalence contributions. For $S_4$, the combined effect reaches 0.55 kcal/mol.
• Comparison with ATcT:
  • The new W5prelim values show excellent agreement with the latest ATcT (v1.220) data.
  • Significant gaps between the old W4-17 values and ATcT (e.g., for $Cl_2$, $S_2$, $P_2$, and CN radical) are closed by the new protocol.
  • For Boron, the new calculations support the recent ATcT revision of the atomic heat of formation ($\Delta H^\circ_{f,0K}[B(g)]$) from ~133.8 to 135.13 kcal/mol.
  • For Silicon, the results support a revised $\Delta H^\circ_{f,0K}[Si(g)]$ of 107.63 kcal/mol, higher than previous CODATA values.
• Specific Species Revisions:
  • $S_4$ (Tetrasulfur): The new TAE is significantly revised due to the large subvalence $(Q)$ contribution and strong static correlation.
  • $P_4$ and $P_2$: Show non-negligible corrections requiring the new protocol for kJ/mol accuracy.
  • NO$_2$: The DBOC was suppressed due to a "pole" near the equilibrium geometry caused by vibronic coupling, a finding consistent with the need for diabatic treatments in such systems.

5. Significance

This paper represents a critical step toward the next generation of computational thermochemistry (W5 theory). Its significance lies in:

1. Breaking the "kJ/mol Barrier" for Second-Row Elements: It demonstrates that achieving sub-kJ/mol accuracy for second-row molecules requires treating subvalence correlation at the post-CCSD(T) level, a step previously omitted due to computational cost.
2. Methodological Rigor: It resolves long-standing ambiguities regarding reference geometries (frozen-core vs. all-electron) and reference wavefunctions (UHF vs. ROHF) for open-shell systems.
3. Validation of ATcT: The strong agreement between the new theoretical predictions and the ATcT network validates both the experimental thermochemical network and the new theoretical protocol.
4. Practical Guidelines: It provides a roadmap for handling computationally expensive subvalence post-CCSD(T) effects via basis set scaling, making high-accuracy thermochemistry feasible for larger second-row systems.

In conclusion, the authors establish that for second-row chemistry, subvalence correlation is not just a "core-valence" effect but includes significant higher-order correlation terms that must be included to achieve the highest standards of accuracy.