Time-Dependent NEGF: Real-Time Quantum Transport

• TD-NEGF is a quantum many-body framework that describes real-time electronic, spin, and excitonic dynamics in open systems.
• It employs contour-ordered Green’s functions and the Kadanoff–Baym equations to capture both transient responses and asymptotic steady-state behavior.
• Advanced numerical schemes, including memory-kernel reduction and ODE formulations, enable efficient simulation of ultrafast transport in nanoscale devices.

Time-dependent nonequilibrium Green's function (TD-NEGF) theory is a quantum many-body framework that enables the real-time description of electronic, spin, and excitonic transport in open systems subjected to arbitrary time-dependent perturbations. TD-NEGF generalizes the conventional steady-state NEGF formalism to problems far from equilibrium, allowing for the treatment of transient dynamics, ultrafast response, quantum pumping, and dynamical functional phenomena in nanostructures, molecular junctions, and mesoscale devices. The approach leverages contour-ordered Green's functions, self-energy techniques, and memory-kernel equations of motion to capture the influence of leads, interactions, and external drives, and incorporates both transient and asymptotic steady-state behavior (Foldi, 2015, Ridley et al., 2022, Cornean et al., 2017).

1. Fundamental Formalism: Contour-Ordered and Real-Time Green’s Functions

TD-NEGF theory is fundamentally built on the single-particle (or, more generally, many-body) Green's function defined on the Keldysh contour: $$G(\tau, \tau') = -i \langle T_C[\psi_H(\tau)\psi_H^\dagger(\tau')]\rangle$$ where $T_C$ orders operators along the Keldysh time contour γ, which runs from the initial time to infinity (forward branch), then back to the initial time (backward branch), closed with a descent along the imaginary axis for finite temperature (Ridley et al., 2022).

Physical quantities such as particle density, correlation functions, and currents are obtained by projecting $G(\tau,\tau')$ onto the real time branches, yielding the retarded ($G^R$), advanced ($G^A$), lesser ($G^<$), and greater ($G^>$) components. For noninteracting and interacting systems, these have explicit operator definitions (see, e.g., (Cornean et al., 2017)): $$\begin{aligned} G^R(t, t') &= -i\Theta(t-t')\langle[\psi(t),\psi^\dagger(t')]\rangle \ G^<(t, t') &= i\langle \psi^\dagger(t')\psi(t) \rangle \end{aligned}$$ The local density is obtained as $n(x, t) = -i G^<(x, t; x, t)$ (Foldi, 2015).

2. Kadanoff–Baym and Dyson Equations in the Time Domain

The time evolution of the system is governed by the integro-differential Kadanoff–Baym equations (KBEs), which, for the single-particle sector, read: $$\begin{aligned} \left[i \partial_{t_1} - H(t_1)\right] G^R(t_1, t_2) &= \delta(t_1-t_2) + \int_{t_2}^{t_1} \Sigma^R(t_1, t') G^R(t', t_2) dt' \ G^<(t_1, t_2) &= \int_{-\infty}^{t_1} dt \int_{-\infty}^{t_2} dt' G^R(t_1, t) \Sigma^<(t, t') [G^R(t_2, t')]^\dagger \end{aligned}$$ Here, $T_C$0 represents the one-particle part of the Hamiltonian (including time-dependent and interaction terms), and the self-energies $T_C$1 encode the effect of reservoir coupling (embedding) and many-body interactions (Foldi, 2015, Ridley et al., 2022). For interacting systems, the self-energy splits as $T_C$2 with embedding and many-body contributions (Ridley et al., 2022).

In practical partitioning schemes, only the central region is treated explicitly, with the effect of semi-infinite leads entering through boundary self-energies localized to the edges, ensuring exact transparent boundary conditions and nonreflected wavefunctions (Foldi, 2015, Gaury et al., 2013).

3. Embedding Self-Energy, Initial Condition, and Observables

The embedding self-energy $T_C$3 provides a nonperturbative account of the reservoirs: $T_C$4 where $T_C$5 is the tunneling Hamiltonian, and $T_C$6 is the Green’s function of an isolated lead α (Ridley et al., 2022).

Transmission, reflection, and local observables are computed from $T_C$7 or associated wave functions. For example, the local current between sites $T_C$8 and $T_C$9: $G(\tau,\tau')$0 with incoming and outgoing currents yielding the time-resolved transmission $G(\tau,\tau')$1 (Foldi, 2015). Physical charge and energy currents in general open systems follow the Meir–Wingreen formula, which in time domain reads: $G(\tau,\tau')$2 (Ridley et al., 2022, Cornean et al., 2017). The correspondence to the steady-state (Landauer–Büttiker) limit is recovered as $G(\tau,\tau')$3 and drive potentials become stationary (Ridley et al., 2014, Ridley et al., 2022).

4. Numerical Implementation and Computational Schemes

TD-NEGF calculations require efficient handling of memory kernels and matrix-valued propagation:

• Spatial discretization: The device region is discretized into a finite grid or basis $G(\tau,\tau')$4, with kinetic terms represented by finite-difference or tight-binding Hamiltonians (Foldi, 2015).
• Memory kernels: The integral terms arising from lead self-energies are history-dependent and can be re-expressed via auxiliary variables or reduced using pole decompositions (Padé, Lorentzian) for spectral densities and Fermi functions, allowing linear-scaling algorithms in the time domain (Tuovinen et al., 2022, Xie et al., 2013, Ho, 2018, Ho, 2019).
• Wide-band limit (WBL): In many applications, the embedding self-energies are approximated as energy-independent (constant broadening $G(\tau,\tau')$5), greatly simplifying the time-evolution equations to closed forms (Ridley et al., 2014, Ho, 2018).
• ODE formulations: For large-scale simulations, the generalized Kadanoff–Baym ansatz (GKBA) allows the equations to be rewritten as a hierarchy of ODEs for single-time quantities (density matrices and embedding correlators), achieving linear scaling in time (Tuovinen et al., 2022, Ho, 2019).

Hierarchical equations of motion (HEOM) and pole-fitting schemes are leveraged to handle complex and structured spectral densities beyond the WBL (Xie et al., 2013).

5. Extensions: Interactions, Correlations, and Multiphasic Physics

TD-NEGF accommodates many-body interactions via diagrammatic self-energies (Hartree–Fock, second Born, GW, GD, T-matrix, Fan–Migdal) (Ridley et al., 2022, Tuovinen et al., 2022). The collision integral arising from interactions can be closed using small sets of ODEs involving higher-order correlation matrices. The GKBA allows for conserving dynamics at the level of the one-particle density matrix. Electron-electron, electron-phonon, and even electron-photon couplings can be incorporated in this framework.

In multiscale or hybrid schemes, TD-NEGF can be coupled to classical equations of motion for slow variables, e.g., spin dynamics via Landau–Lifshitz–Gilbert equations in spintronics (Petrovic et al., 2018).

Transient charge transfer and Hartree corrections are treated efficiently by coupling the quantum equations to analytical Poisson solvers (e.g., muffin-tin networks), ensuring charge conservation under strong time-dependent driving (Ho, 2018, Ho, 2019).

6. Applications: Ultrafast Scattering, Quantum Pumping, and Device Physics

TD-NEGF is employed to simulate time-resolved quantum transport under arbitrary field protocols, such as voltage pulses, AC or multi-harmonic drives, gate voltages, time-dependent barriers, ultrafast laser fields, and photonic coupling (Foldi, 2015, Ridley et al., 2022, Ho, 2019).

Specific applications include:

• Ultrafast electron scattering by laser pulses: TD-NEGF captures dynamical polarization, induced density modulations, and the outgoing wave packets generated by pulse-driven scattering (Foldi, 2015).
• Photon-assisted and AC transport: Closed-form analytic formulas capture transient and long-time currents, with non-adiabatic features such as ringing, frequency mixing, and sideband generation emerging naturally in the formalism (Ridley et al., 2014, Ho, 2019).
• Correlated excitation transport: Wave-packet dynamics of electron-hole pairs, polaronic evolution, and their decoherence can be tracked in real-time (Tuovinen et al., 2022).
• Spin and charge pumping: TD-NEGF+LLG predicts AC and DC current generation by moving or pulse-driven magnetic domain walls, and quantifies the breakdown of perturbative spin-motive force theory (Petrovic et al., 2018).
• Molecular and nanoscale devices: TD-NEGF implemented with muffin-tin Poisson solvers enables full-scale time-resolved simulation including dynamical charge transfer and screening in metal–semiconductor junctions (Ho, 2018).

7. Efficiency, Scalability, and Limitations

Numerical advances in pole expansion, hierarchical ODE solvers, and memory kernel reduction techniques have enabled TD-NEGF to reach system sizes and time scales inaccessible to brute-force two-time integration (Tuovinen et al., 2022, Gaury et al., 2013, Ho, 2018). For typical nanodevice calculations, runtime is reduced by up to one to two orders of magnitude, and the method remains stable for tens of picoseconds and beyond.

Limitations persist for systems with strong nonequilibrium correlations beyond the single-particle sector, for extremely narrow-band reservoirs (breakdown of WBL), for very large density distortions, or for strong dynamical exchange-correlation effects not captured at the Hartree level (Ho, 2018, Tuovinen et al., 2022).

Potential extensions under active investigation include the incorporation of fully dynamical exchange-correlation kernels, energy-dependent embedding, and time-dependent quantum capacitance, as well as further integration with stochastic or bath models for strongly open quantum systems (Ridley et al., 2022, Ho, 2018).

---

Principal References: (Foldi, 2015, Cornean et al., 2017, Ridley et al., 2022, Tuovinen et al., 2022, Ridley et al., 2014, Gaury et al., 2013, Ho, 2018, Ho, 2019, Petrovic et al., 2018, Xie et al., 2013).