Reaching precise proton affinities in non-Born-Oppenheimer calculations

This study demonstrates that non-Born-Oppenheimer proton affinity calculations achieve high accuracy (within 0.1 kcal/mol) primarily by using uncontracted electronic basis sets on quantum protons, which significantly improves convergence while rendering the choice of protonic basis sets less critical.

The Gist

The Big Picture: Treating Protons Like Ghosts

Imagine you are trying to understand how a molecule holds together. In standard chemistry (the "old way"), we treat electrons as fuzzy clouds of probability, but we treat the nucleus (the proton) as a tiny, solid, stationary billiard ball sitting right in the center. This is called the Born-Oppenheimer approximation.

However, protons are actually very light and wiggly. They don't just sit still; they "jiggle" and spread out like a cloud of mist. This is called a quantum effect.

To get a super-precise answer, the authors of this paper decided to stop treating the proton like a billiard ball and start treating it like a ghost—a fuzzy cloud just like the electrons. This is called a Non-Born-Oppenheimer (Non-BO) calculation.

The Problem: The "Fitting" Mismatch

To calculate these fuzzy clouds, scientists use mathematical building blocks called basis sets. Think of these like a set of LEGO bricks.

• Electronic Basis Set: Bricks used to build the electron cloud.
• Protonic Basis Set: Bricks used to build the proton cloud.

The authors wanted to know: How many bricks do we need to get a perfect model?

They discovered a major problem with how people were using the electronic bricks.

The "Contracted" vs. "Uncontracted" Analogy

Imagine you have a box of LEGO bricks.

• Contracted (The Old Way): To save time, manufacturers glue certain bricks together into pre-made clusters. For example, they glue three small bricks together to make one "medium" brick. This is great for building a house where the foundation is solid and predictable. In standard chemistry, the proton is a solid point, so these glued bricks work fine.
• Uncontracted (The New Way): The authors realized that when the proton is a "ghost" (a fuzzy cloud), those glued bricks are too rigid. The electron cloud needs to wiggle and change shape right next to the proton. If the bricks are glued together, the electron can't move freely enough to fit the shape of the proton cloud.

The Discovery: The authors found that if you un-glue the bricks (uncontract the basis set) specifically around the quantum proton, the model becomes incredibly accurate almost instantly. It's like taking apart a pre-made LEGO wall so you can mold the bricks perfectly around a weirdly shaped rock.

The Results: Faster, Cheaper, Better

The paper tested this on 13 different molecules to see how well they could predict Proton Affinity (basically, how much a molecule "wants" to grab a proton).

1. The Magic of Un-gluing: By simply un-gluing the electronic bricks around the proton, they got results that were one level of precision higher than before, without spending much more computer time.

   • Analogy: It's like realizing you don't need a Ferrari to win a race; you just need to take the tires off your sedan and put on racing tires. You get the speed of the Ferrari for the price of a sedan.

2. The Proton Bricks: They also tested different sets of "proton bricks." They found that you don't need a massive, complicated set of bricks for the proton. A smaller, simpler set works just fine because the proton doesn't need as much detail as the electron does.

3. The "Mixed" Trap: Some scientists try to use a huge set of bricks for the proton and a small set for the rest of the molecule. The authors showed this is a trap. It often gives a "lucky" correct answer by accident (canceling out two big errors), but it's not reliable. If you use the "un-glued" method consistently, you get the real correct answer every time.

The Bottom Line

This paper is a "user manual" for the future of high-precision chemistry.

• The Rule: If you are modeling a proton as a fuzzy quantum cloud, do not use glued-together (contracted) electronic bricks around it. Un-glue them.
• The Benefit: You get results accurate to within 0.1 kcal/mol (which is basically perfect for chemistry) using standard, off-the-shelf tools, rather than needing custom-made, expensive super-computers.

In short: Stop forcing the electron to fit a rigid mold around a wiggly proton. Let the electron flow freely, and the math will solve itself.

Technical Summary

1. Problem Statement

Non-Born–Oppenheimer (non-BO) methods, such as the Nuclear-Electronic Orbital (NEO) approach, treat select nuclei (typically protons) quantum mechanically alongside electrons. While promising for capturing nuclear quantum effects (e.g., zero-point energy, tunneling), these calculations require two distinct basis sets: an electronic basis set and a protonic basis set.

The paper addresses two critical gaps in current non-BO methodology:

1. Lack of Systematic Convergence Studies: There is insufficient understanding of how proton affinities (PAs) converge with respect to both electronic and protonic basis sets.
2. Inappropriate Basis Set Usage: Standard electronic basis sets are typically contracted based on the assumption of point-like nuclear charges. However, in non-BO calculations, the quantum proton is delocalized, creating a finite charge distribution rather than a point charge. The authors hypothesize that standard contracted basis sets introduce significant errors when applied to these delocalized protons, leading to slow convergence and inaccurate results.

2. Methodology

The authors investigated the basis set convergence of Proton Affinities (PAs) for a set of 13 molecules (including anions like $CN^-$, $NO_2^-$, and neutral molecules like $H_2O$, $NH_3$) using non-BO Density Functional Theory (DFT).

• Theoretical Framework:
  • Method: Non-BO-DFT using the B3LYP functional (and others for cross-assessment) with the epc17-2 electron-proton correlation functional.
  • Hamiltonian: The Schrödinger equation is solved for coupled electrons and quantum protons.
  • Basis Sets:
    • Electronic: Jensen's polarization-consistent (aug-pc-n), Dunning's correlation-consistent (aug-cc-pVXZ), and Karlsruhe (def2-XZPD) families. Diffuse functions were included for all.
    • Protonic: Yu et al.'s PB sets (PB4–PB6) and even-tempered sets (Yang et al., Khan and Tonner-Zech).
• Key Innovation (Uncontraction):
  • The authors introduced a scheme denoted as uncHq, where the electronic basis functions centered on the quantum protons are uncontracted, while the rest of the molecule uses standard contracted basis sets.
  • This removes the "contraction error" inherent in basis sets designed for point charges, allowing the electronic wavefunction greater flexibility near the delocalized proton center.
• Computational Details:
  • Calculations performed using Q-Chem 6.1.
  • Geometries optimized using standard DFT (Born-Oppenheimer).
  • Target precision: Convergence within 0.1 kcal/mol of the Complete Basis Set (CBS) limit.

3. Key Contributions

1. Demonstration of Uncontraction Necessity: The paper proves that uncontracting electronic basis sets on quantum protons is essential for accurate non-BO calculations. Standard contracted sets fail to describe the electron density correctly near the delocalized proton, leading to large errors.
2. Efficiency Breakthrough: Uncontracting the basis set on the quantum proton yields results of at least one $\zeta$-level higher quality (e.g., a triple-$\zeta$ uncontracted set performs like a quadruple-$\zeta$ contracted set) with negligible additional computational cost.
3. Debunking Mixed Basis Set Strategies: The study shows that using "mixed basis sets" (a larger basis for the quantum proton and smaller for others) often leads to fortuitous error cancellation (Pauling points) rather than true convergence.
4. Evaluation of Protonic Basis Sets: The authors critically evaluate existing protonic basis sets, finding that many (specifically the PB5 family) do not form a systematic hierarchy and may be suboptimal due to inconsistent optimization methodologies.

4. Key Results

• Electronic Basis Set Convergence:
  • Contracted Sets: Required quintuple-$\zeta$ (aug-pc-4) levels to reach the 0.1 kcal/mol convergence criterion.
  • Uncontracted (uncHq) Sets: Triple-$\zeta$ (aug-pc-3) uncontracted sets achieved convergence within 0.1 kcal/mol. Quadruple-$\zeta$ uncontracted sets converged to within 0.03 kcal/mol.
  • Physical Insight: Uncontraction corrects the electron and proton densities near the quantum proton, which were qualitatively incorrect in contracted calculations (as shown in density plots for $FHF^-$).
• Protonic Basis Set Convergence:
  • Most tested protonic basis sets (including PB4-F1 and PB6-H) were sufficient to reach the 0.1 kcal/mol limit when paired with a sufficiently large electronic basis.
  • The PB5 family showed non-systematic behavior and poorer precision compared to PB4 and PB6, likely due to inconsistent optimization targets (mixing non-BO and BO methodologies).
  • Even-tempered sets (e.g., 8s8p8d) performed well, suggesting dedicated, smaller protonic basis sets could be developed.
• Mixed Basis Sets:
  • Using a larger basis on the quantum proton while keeping others small resulted in systematic positive bias. The apparent improvement in some cases was due to the cancellation of this bias against the negative bias of the contracted basis, a classic "Pauling point" error.
  • Once the quantum proton basis was uncontracted, mixed basis sets offered no advantage over consistent, uncontracted pairings.
• Transferability:
  • The convergence behavior and the benefits of uncontraction were consistent across different electronic functionals (PW92, PBE0, B3LYP, r2SCAN, HF) and electron-proton correlation treatments.

5. Significance

• Standardization of Non-BO Calculations: The paper establishes a clear protocol for achieving chemical accuracy (0.1 kcal/mol) in non-BO calculations: use uncontracted electronic basis sets on quantum nuclei.
• Cost-Effectiveness: By allowing researchers to use smaller basis sets (e.g., triple-$\zeta$ instead of quintuple-$\zeta$) while maintaining high precision, the method significantly reduces computational costs for studying nuclear quantum effects.
• Guidance for Future Development: The findings highlight the need for new, systematically optimized protonic basis sets that are balanced with electronic basis sets. The authors suggest that current PB sets are often too large for low-$\zeta$ electronic sets and that dedicated, smaller sets could be developed.
• Methodological Rigor: The study warns against relying on error cancellation in mixed-basis approaches and emphasizes the importance of reaching the true CBS limit to validate density functionals and train machine-learned potentials for non-BO systems.

In summary, Nikkanen and Lehtola demonstrate that the primary bottleneck in non-BO precision is not the size of the basis set, but the contraction scheme applied to the quantum protons. Uncontracting these basis sets is a simple, low-cost modification that drastically improves accuracy and convergence.