Read A Fragment Interaction Report

The water-dimer calculation gives five interaction energies. Reading them together shows how the method and active space affect the result. This tutorial explains the energy differences and the associated resource estimates.

You can read the tables below without running Python. To reproduce them, follow Fragment Interaction Energy with cc-pVDZ and mode="full". The generated report uses Chinese headings. The explanations below follow its table order.

Start With The Interaction Energy

The first table compares five methods for the same water-dimer geometry. Each row gives the counterpoise-corrected interaction energy, \(\Delta E = E(AB) - E(A) - E(B)\). AB denotes the dimer, and A and B denote the two water molecules. All three energies use the dimer basis.

Method	Calculation	ΔE (kcal/mol)
HF	Hartree–Fock	−3.682
MP2	Second-order Møller–Plesset perturbation theory with a frozen core	−3.935
CASCI	Exact diagonalization within the selected active space	−3.574
SQD	Sample-based quantum diagonalization within the same active space	−3.557
CCSD(T)	Coupled cluster with perturbative triples and a frozen core	−3.699

All five values are negative. Each method therefore places the dimer below the separate fragments at this geometry. A more negative value represents a stronger calculated interaction. The spread from −3.935 to −3.557 kcal/mol shows how the result changes with the electronic structure method.

The report shows each energy in Ha and kcal/mol, with six and three decimal places, respectively. A dash means that the calculation has no value for that entry. The adjacent column explains why. For example, classical_only mode supplies HF and MP2 energies. In full mode, CCSD(T) runs when its estimated cost fits the calculation budget.

Compare SQD With CASCI

CASCI provides an exact solution within the selected active space. SQD approximates that solution in the same space. Their difference therefore measures the SQD algorithm error for this calculation:

\[\epsilon_{\mathrm{algorithm}} = \Delta E_{\mathrm{SQD}} - \Delta E_{\mathrm{CASCI}}.\]

The example report gives +0.016 kcal/mol. SQD predicts a slightly less negative interaction energy than CASCI. This comparison isolates the algorithm’s contribution because both methods use the same active space.

The report calculates differences before rounding the energies. Subtracting the rounded table entries can therefore give a slightly different last digit.

Compare The Active Space With A Larger Calculation

The next comparison asks how much the selected active space changes the interaction energy. In this example, the reference is frozen-core CCSD(T) in the full orbital space with the same cc-pVDZ basis:

\[\epsilon_{\mathrm{active\ space}} = \Delta E_{\mathrm{CASCI}} - \Delta E_{\mathrm{CCSD(T)}}.\]

This difference is +0.125 kcal/mol. The active-space CASCI calculation predicts a less negative interaction energy than the CCSD(T) reference. Here, this difference exceeds the SQD–CASCI difference. Choosing the active space therefore deserves attention alongside the SQD settings when you refine this calculation.

The report calls this quantity the active-space error. It measures the difference from the chosen reference, which here also changes the correlation method. CCSD(T) remains an approximate reference. The source label pyscf_ccsd_t identifies the CCSD(T) calculation from the same run. If that calculation has no value, q2m3 can use an imported reference with the same basis.

Add An External Reference

A reference with another basis extends the comparison beyond the current calculation. The report combines it with the same-basis reference to give two further differences:

Quantity	Definition
Basis error	Same-basis reference − external reference
Model error	CASCI − external reference

Each term uses interaction energies. When all references are available, the model error equals the active-space error plus the basis error. The report shows each difference with its sign and its absolute value, in Ha and kcal/mol.

The water-dimer example supplies no external reference, so these two entries show a dash and missing external reference. To fill them, supply a reference file as described in Adapt The Calculation. Choose a reference with the same geometry and energy definition. If its method also differs, the reported basis error includes that method difference.

Relate The Active Space To Computational Cost

The final two tables help plan calculations for the dimer and its monomers. They describe QPE resources and SQD workload separately.

The QPE table estimates the cost of quantum phase estimation with a double-factorized active-space Hamiltonian. It lists the Hamiltonian norm parameter λ in Ha, logical qubits, and Toffoli gates for AB, A, and B. These estimates describe a possible QPE calculation. The energies above come from the electronic structure methods in the first table.

The SQD table connects the active-space size to the simulated workload:

Subsystem	SQD qubits	Full CI dimension	Prepare RSS (MB)	Sample RSS (MB)
AB	16	784	1031.7	1031.7
A	8	16	1024.5	1024.5
B	8	16	1024.5	1024.5

SQD uses two qubits per active orbital. The full configuration-interaction (CI) dimension counts the determinants for the active space and its spin populations. RSS means resident set size. The two RSS columns predict peak memory for the prepare and sample stages in decimal MB.

Read these estimates when you choose a larger active space or plan a series of calculations. For an energy comparison, start with the interaction energies and their differences from the selected references.

Include Experimental Ks Values

If you also have experimental Ks values, pass a mapping of labels to values as ks_table in render_report. The report appends them in the supplied order, so collaborators can read the measurements alongside the calculations. Statistical analysis of their relationship to ΔE remains a separate step. The water-dimer example has no Ks table.