Skip to content

Bonding Character as a Descriptor for Huang–Rhys Factors in Optically Active Defects

FieldValue
JournalThe Journal of Physical Chemistry Letters
Published2026-02-19
AuthorsFatimah Habis, Yuanxi Wang
DOI10.1021/acs.jpclett.5c03935

A ground-state COHP descriptor estimated Huang-Rhys factors across a reference range of 0 to 20 without explicit excited-state relaxation or full phonon calculations. The method reproduced an NV-center transition with a partial-factor root mean square error of 0.001 and required one ground-state relaxation.

Why bonding character serves as a descriptor

Section titled “Why bonding character serves as a descriptor”

The dimensionless Huang-Rhys factor represents the average number of phonons emitted during an electronic transition. For defect-based single-photon emitters, a factor below 2 is desirable for greater photon indistinguishability.

For qubits, the factor affects excited-state dynamics, non-radiative lifetimes, optical initialization, optical readout, and phonon-induced decoherence. Existing protocols can suffer from excited-state convergence problems, level degeneracies, and costly phonon calculations.

They also offer little guidance for engineering a local defect structure. The proposed descriptor uses Crystal orbital Hamilton populations, or COHPs, to compare bonding characters of occupied and unoccupied states.

Negative COHP values indicate bonding interactions, positive values indicate antibonding interactions, and zero values indicate nonbonding interactions. The hypothesis is that a smaller bonding-character change corresponds to a smaller Huang-Rhys factor.

In the negatively charged VBON defect in monolayer hBN, the lowest-energy excitation had a factor of 14.3. The next-lowest transition had a factor of 2.5 and joined two antibonding states with a smaller bonding-character change.

Converting COHP changes into excited-state forces

Section titled “Converting COHP changes into excited-state forces”

The excitation-energy derivative is approximated through the occupied and unoccupied Kohn-Sham eigenvalue difference. This simplification neglects excitonic effects and assumes quasiparticle eigenvalues follow Kohn-Sham eigenvalues under distortion.

Localized defect orbitals allow a discrete sum over bands rather than an energy integral. Pairwise COHP changes are assigned along bonds and combined into an estimate of the excited-state force on each atom.

The linear force estimate uses a proportionality factor A. Bond-strength and bond-length relations suggested a value on the order of 2 Å⁻¹. Calibration against ΔSCF forces for the hBN and diamond defects gave A = 3.33 Å⁻¹.

Calibration with a finite displacement of 0.01 Å gave a similar value for A with 1% deviation. COHP forces reproduced the magnitudes and directions of ΔSCF forces well for the lowest-energy VBON transition.

They were not intended as direct replacements for excited-state forces, and some systems showed large errors on selected atoms. For the VNCB HOMO-1 to LUMO transition, the partial-factor root mean square error remained about 0.02.

Force errors projected mainly onto modes above 100 meV, beyond the modes within 100 meV that dominate the spectral lineshape. The descriptor therefore acts as a proxy for calculating factors and spectral functions.

Finding the excited-state geometry without relaxation

Section titled “Finding the excited-state geometry without relaxation”

Ground-Excited Reflective Deformation finds the excited-state equilibrium without an excited-state energy minimization. It applies the displaced harmonic oscillator approximation already used when ground-state phonon frequencies enter the factor calculation.

At ground-state equilibrium, the excited-state force is assumed equal and opposite to the ground-state force at excited-state equilibrium. The method searches the ground-state surface for a configuration whose restoring force balances the imposed excited-state force.

Deformation continues until the forces balance, and the atomic displacement supplies the one-dimensional configuration coordinate. That coordinate determines an effective frequency and total Huang-Rhys factor through the standard one-dimensional model.

Deformation displacements agreed with direct ΔSCF excited-state relaxations. Factors from this deformation agreed with full-phonon references when the initial forces came from ΔSCF. The agreement remained when the initial forces came from the COHP descriptor.

All structures were relaxed with GGA-PBE density functional theory, projector-augmented wave pseudopotentials, and VASP. The plane-wave cutoff was 500 eV, and ionic forces converged to 0.002 eV Å⁻¹. Ground-state COHP values were calculated with LOBSTER.

Reference excited-state forces came from single Γ-point ΔSCF calculations. The hBN defects used a 7 × 7 × 1 supercell, while the NV center used a 4 × 4 × 4 supercell.

Phonon frequencies came from finite displacements with PHONOPY. The hBN phonon calculation used a 7 × 7 × 1 supercell and a Γ-centered 2 × 2 × 1 k-point grid. The NV phonon calculation used a 3 × 3 × 3 supercell with one Γ point.

Reference factors used force-constant embedding to approach the dilute-defect limit. The embedded hBN supercell was 15 × 15 × 1, and the embedded diamond supercell was 10 × 10 × 10.

For hBN, converged embedding used R1 = 6.28 Å and R2 = 8.79 Å after comparison with a 9 × 9 × 1 reference. For diamond, the embedding used R1 = R2 = 5 Å, consistent with earlier tests for the NV center.

Validation covered three hBN defects and the bulk-diamond NV center across the listed optical transitions. The table compares total factors from ΔSCF forces with values from COHP-estimated forces, using full-phonon frequencies in both cases.

Defect and transitionSpin channelΔSCF totalCOHP totalPartial-factor RMSE
NV, HOMO to LUMO24.514.670.001
NV, HOMO to LUMO plus 123.984.670.001
VNCB, HOMO to LUMO12.192.970.01
VNCB, HOMO-1 to LUMO19.117.570.018
VBON, HOMO-2 to LUMO23.122.870.008
VBON, HOMO-1 to LUMO25.65.260.014
VBON, HOMO to LUMO216.8216.430.014
VB, HOMO to LUMO23.032.540.008
VB, HOMO-3 to LUMO210.2610.310.014

The VNCB HOMO-1 to LUMO case shows the largest listed disagreement, with totals of 9.11 and 7.57. It also has the largest partial-factor root mean square error, 0.018.

The NV HOMO to LUMO totals are 4.51 from ΔSCF and 4.67 from COHP. The partial-factor root mean square error is 0.001. The comparison shows that partial-factor agreement does not rely on cancellation between modes.

An earlier high-throughput study found that Huang-Rhys factors were largely independent of other defect properties except the configuration coordinate. A related study linked higher host stiffness to lower average factors across hosts.

The paper argues that host averages conceal larger differences among defects within the same host. Local bonding-character differences address that within-host variation and connect the factor to a defect orbital property.

The validated range of 0 to 20 encompasses most Huang-Rhys factors reported for hBN defects. The authors propose moving factor estimation earlier than detailed phonon calculations during high-throughput screening.

This could retain defects with low factors and high formation energies that might be engineered to be more stable. The descriptor does not replace formation-energy calculations, so defect stability still requires separate assessment.

The descriptor uses ground-state Kohn-Sham orbitals and a linear approximation for distortions in covalent B, C, N, and O systems. The cited excited-state displacement scale is about 0.1 Å, which supports the linear treatment.

The method assumes the displaced harmonic oscillator approximation for the relevant phonon modes. It does not include excitonic effects in the excited-state force approximation and targets ΔSCF rather than many-body excited-state forces.

For correlated excitations, the authors propose transition weights from time-dependent DFT, the Bethe-Salpeter equation, or a one-particle transition density matrix. Validation was limited to prototypical defects in monolayer hBN and bulk diamond.

Large errors on selected atomic forces limit direct force interpretation, even when total and partial factors remain accurate. The paper reports agreement with computational references, but no experimental device validation.

An NV-center spin qubit or single-photon emitter requires a bulk diamond host. 6CCVD’s diamond materials catalog lists MPCVD-grown single-crystal diamond plates and wafers up to 125 mm, with Au, Pt, Pd, Ti, W, or Cu metallization.

Engineering support can help define the host and contact specifications for defect-device integration.

View Original Abstract

The electron-phonon coupling of a defect, characterized by its Huang-Rhys (HR) factor, is a crucial metric determining its excited-state dynamics, relevant to defect applications as qubits and quantum emitters. However, HR factors remain challenging to calculate from first principles, complicated by convergence issues in excited-state relaxation and time-consuming phonon calculations. Even when calculated, HR factors lack a rational design principle. Here we show that an orbital-based descriptor can be used to rationalize and efficiently estimate HR factors. Combining this descriptor with a ground-state deformation technique allows circumvention of both excited-state relaxation and full phonon calculations. Specifically, our descriptor for HR factors is constructed using bonding character differences obtained from ground-state density functional theory, quantified by crystal orbital Hamilton populations. We demonstrate this descriptor for prototypical hBN defects and the diamond NV - center. This orbital-based descriptor can be potentially used in high-throughput computational screening to identify ideal candidates for spin qubits and SPEs.