Time-dependent density functional theory with ultrasoft pseudopotentials: Real-time electron propagation across a molecular junction

Xiaofeng Qian, Ju Li, Xi Lin, Sidney YipPhysical 2006 · DOI · arXiv

Abstract

A practical computational scheme based on time-dependent density functional theory (TDDFT) and ultrasoft pseudopotentials (USPP) is developed to study electron dynamics in real time. A modified Crank-Nicolson time-stepping algorithm is adopted under a plane-wave basis. The scheme is validated by calculating the optical absorption spectra for a sodium dimer and a benzene molecule. As an application of this USPP-TDDFT formalism, we compute the time evolution of a test electron packet at the Fermi energy of the left metallic lead crossing a benzene-(1,4)-dithiolate junction. A transmission probability of 5–7%, corresponding to a conductance of 4.0–5.6 $\mu S$, is obtained. These results are consistent with complex band structure estimates and Green's function calculation results at small bias voltages.

Introduction

As electronic devices approach the molecular scale, the ability to accurately model electron transport through single-molecule junctions becomes paramount. Traditionally, the Green’s Function (GF) method has served as the standard theoretical framework for these investigations. However, GF methods often struggle to incorporate complex dynamical effects, such as electron-phonon coupling or time-dependent electron-electron correlations.

Time-Dependent Density Functional Theory (TDDFT) offers a promising alternative by allowing researchers to observe electron dynamics in real-time. Despite its potential, the computational cost of TDDFT—particularly when using "norm-conserving" pseudopotentials for transition metals like gold—has limited its applicability. This paper presents a novel framework that integrates Ultrasoft Pseudopotentials (USPP) into the TDDFT formalism, significantly reducing the computational burden and enabling the simulation of large-scale molecular junctions.

Theoretical Framework and Methodology

The primary challenge in implementing TDDFT with USPP lies in the non-identity overlap operator ($S$) inherent to the USPP framework. Unlike standard Schrödinger-like equations, the equations of motion must account for this operator to ensure physical consistency.

Hamiltonian Reformulation

The researchers derived a specific set of equations of motion for the USPP framework using a plane-wave basis set. This approach allows for the efficient treatment of the valence electrons of transition metals, which typically require a very high energy cutoff in norm-conserving schemes.

Time-Propagation Algorithm

To propagate the electron wavefunctions, a modified Crank-Nicolson algorithm was employed. This implicit scheme is preferred over explicit methods due to its numerical stability and its ability to conserve the total number of electrons (norm conservation) over extended simulation periods. To further enhance accuracy, a second-order predictor-corrector step was implemented, which estimates the potential at future time steps before finalizing the wavefunction evolution.

Validation and Benchmarking

The TD-USPP scheme was validated using two well-characterized molecular systems:

  1. Sodium Dimer ($Na_2$): The optical absorption spectra were calculated and compared against previous Time-Dependent Local Density Approximation (TDLDA) results, showing excellent agreement.
  2. Benzene ($C_6H_6$): The team focused on the $\pi \to \pi^*$ electronic transition. The calculated absorption peak at 6.95 eV matched both experimental data and high-level theoretical benchmarks, confirming the accuracy of the USPP implementation for organic molecules.

Application: The Au-Benzene-Au Junction

The framework was applied to a model molecular junction consisting of a benzene-(1,4)-dithiolate (BDT) molecule sandwiched between two one-dimensional gold atomic chains.

The "Computer Experiment" Approach

Rather than solving static frequency-domain equations, the researchers performed a real-time "computer experiment." An electron wave packet was initialized at the Fermi energy of the left metallic lead. The simulation then tracked the packet as it moved toward the junction, scattered through the BDT molecule, and partially transmitted to the right lead.

Results and Conductance

The simulation yielded a transmission probability of 5–7%. This translates to a conductance range of 4.0–5.6 $\mu S$. These findings are highly significant as they align closely with results obtained from traditional Green’s function and complex band structure calculations at low bias voltages. This consistency validates the real-time TD-USPP approach as a reliable tool for transport studies.

Discussion and Future Outlook

A key theoretical advancement presented in this work is the derivation of a "moving spatial gauge" ($P$ operator). This operator accounts for the motion of ions during electron propagation. While the current study kept ions fixed, this derivation provides the necessary foundation for future studies on electron-phonon coupling and Joule heating at the nanoscale.

One identified limitation is the approximation used for the initial electronic "kick" to start the simulation, which slightly violates the f-sum rule. Future refinements will focus on a more rigorous implementation of this initial state.

The ability to simulate electron dynamics in real-time with the efficiency of USPP opens new avenues for studying how molecular wires respond to high currents and how atomic vibrations influence electronic transport. This framework represents a significant step toward a comprehensive, dynamic understanding of molecular electronics.

Summary updated

Abstract

A practical computational scheme based on time-dependent density functional theory (TDDFT) and ultrasoft pseudopotentials (USPP) is developed to study electron dynamics in real time. A modified Crank-Nicolson time-stepping algorithm is adopted, under plane-wave basis. The scheme is validated by calculating the optical absorption spectra for a sodium dimer and a benzene molecule. As an application of this USPP-TDDFT formalism, we compute the time evolution of a test electron packet at the Fermi energy of the left metallic lead crossing a benzene-(1,4)-dithiolate junction. A transmission probability of 5--7%, corresponding to a conductance of $4.0--5.6\phantom{\rule{0.3em}{0ex}}\mathrm{\ensuremath{\mu}}\mathrm{S}$, is obtained. These results are consistent with complex band structure estimates and Green's function calculation results at small bias voltages.
BibTeX
@inproceedings{qian2006time,
  author = {Xiaofeng Qian and Ju Li and Xi Lin and Sidney Yip},
  title = {{Time-dependent density functional theory with ultrasoft pseudopotentials: Real-time electron propagation across a molecular junction}},
  booktitle = {Physical},
  year = {2006},
  doi = {10.1103/physrevb.73.035408},
}
Figure