Explain the adverse [T] correction for stretched BH

Determine whether the deterioration of the q-sc-EOM-[T] excitation-energy prediction for the lowest triplet 93a3cb state of BH at bond length 2.0 R_e is caused by the baseline q-sc-EOM-UCCSD result underpredicting the CASCI excitation energy and thereby negatively biasing the triples correction.

Background

The paper benchmarks the leading-order classical triples correction [T] to q-sc-EOM-UCCSD for excited states of BH at equilibrium and stretched geometries. For the lowest 93a3cb state, which is dominated by single excitations, baseline q-sc-EOM-UCCSD already gives errors of approximately 1013 meV. At 2.0 R_e, however, adding [T] slightly worsens the result for both the full and trotterized anstze variants.

The authors explicitly conjecture that this behavior is related to the baseline method slightly underpredicting the CASCI excitation energy, which may induce an unfavorable bias in the perturbative triples correction. Establishing whether this mechanism explains the deterioration would clarify the reliability of [T] even for predominantly singly excited states.

References

We conjecture this is influenced by the fact that the baseline q-sc-t/EOM-UCCSD result at 2.0 $R_e$ slightly underpredicts the CASCI solution, which in turn negatively biases the corresponding [T] correction.

— Resolving molecular biexcited states using classical leading-order triples corrections to quantum equation-of-motion UCCSD  (2610.09260 - Windom et al., 7 Oct 2026) in Section 4.1, Results, subsection BH