betaFeedback

WorksPhysics and Astronomy

Data-driven compression of electron-phonon interactions

Abstract

First-principles calculations of electron interactions in materials have seen rapid progress in recent years, with electron-phonon (e-ph) interactions being a prime example. However, these techniques use large matrices encoding the interactions on dense momentum grids, which reduces computational efficiency and obscures interpretability. For e-ph interactions, existing interpolation techniques leverage locality in real space, but the high dimensionality of the data remains a bottleneck to balance cost and accuracy. Here we show an efficient way to compress e-ph interactions based on singular value decomposition (SVD), a widely used matrix / image compression technique. Leveraging (un)constrained SVD methods, we accurately predict material properties related to e-ph interactions - including charge mobility, spin relaxation times, band renormalization, and superconducting critical temperature - while using only a small fraction (1-2%) of the interaction data. These findings unveil the hidden low-dimensional nature of e-ph interactions. Furthermore, they accelerate state-of-the-art first-principles e-ph calculations by about two orders of magnitudes without sacrificing accuracy. Our Pareto-optimal parametrization of e-ph interactions can be readily generalized to electron-electron and electron-defect interactions, as well as to other couplings, advancing quantitative studies of condensed matter.

Introduction

Electrons in materials are subject to various interactions, including those with phonons, other electrons, and defects. Modeling of these interactions follows two main approaches − analytic treatments that qualitatively capture the main physics with minimal models using only a few parameters, and first-principles calculations aiming at quantitative accuracy but often requiring specialized workflows, high computational cost, and large amounts of data. A middle ground between these extremes would require formulating models of electron interactions that are economical, accurate, and interpretable. Examples of efficient models exist across domains − in quantum chemistry, low-rank approximations [1, 2, 3] can compress two-electron integrals to reduce the computational cost of post-Hartree Fock calculations [4, 5] and extract the critical vibrational modes in a chemical reaction [6, 7]; in correlated-electron physics, efficient parametrization of ee-ee interactions [8] enables the solution of functional renormalization-group flow [9, 10] and the Bethe-Salpeter equation [11, 12]. However, despite these isolated examples, it remains challenging to formulate widely applicable approaches to represent electron interactions both efficiently and accurately.

Focusing on electron-phonon (ee-ph) interactions, analytic treatments such as deformation potential for acoustic phonons [13, 14] and the Fröhlich model for optical phonons [15], which use only a few parameters to describe ee-ph interactions, are still widely utilized [16, 17]. In recent years, first-principles calculations of ee-ph interactions using density functional theory (DFT) [18] and its linear-response variant, density functional perturbation theory (DFPT) [19], have enabled quantitative studies of properties ranging from transport to excited state dynamics to superconductivity [20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33]. Unlike the analytic models, in a typical first-principles calculation one represents the ee-ph interactions using a multi-dimensional matrix with millions or billions of entries. This enormous number of parameters, which are computed rather than assumed, guarantees a faithful description of microscopic details such as the dependence on electronic states and phonon modes of ee-ph interactions. Yet this complexity is also a barrier toward obtaining minimal models and tackling new physics. For example, materials with strong or correlated ee-ph interactions need specialized treatments to capture polaron effects [28, 30, 34, 35] and electron correlations [36, 37]. Reducing the high dimensionality of first-principles ee-ph interactions would allow one to more efficiently describe this physics while retaining quantitative accuracy. The development of data-driven methods to tackle the high-dimensional Hilbert space in the many-electron problem, including neural network states [38, 39] and tensor network methods [40, 41, 42, 43], serve as inspiration.

Here we show a low-rank approximation of first-principles ee-ph interactions which significantly accelerates ee-ph calculations while using only a small fraction (1−2%) of the data and preserving quantitative accuracy. This is achieved by developing SVD calculations of ee-ph matrices in Wannier basis to achieve a minimal representation of ee-ph coupling. We use our compressed ee-ph matrices to compute a range of properties, including charge transport, spin relaxation, band renormalization, and superconductivity, both in metals and semiconductors. Across all benchmarks, the highly compressed ee-ph representation achieves a quantitative accuracy comparable to the standard workflow, while also providing a deeper understanding of the dominant patterns governing ee-ph interactions. Principal component analysis sheds light on the inherent compressibility of ee-ph coupling matrices. Recent interesting work on improving the efficiency of ee-ph calculations [44, 45] is distinct in method and scope from our data-driven approach.

results

compression of e-ph interactions

The key quantities in first-principles ee-ph calculations are the ee-ph matrix elements gmnν(k,q)g_{mn\nu}(\mathbf{k},\mathbf{q}), which represent the probability amplitude for an electron in a band state ∣nk⟩\lvert n\mathbf{k}\rangle, with band index nn and crystal momentum k\mathbf{k}, to scatter into a final state ∣mk+q⟩\lvert m\mathbf{k}+\mathbf{q}\rangle by emitting or absorbing a phonon with mode index ν\nu, wave-vector q\mathbf{q}, energy ℏωνq\hbar\omega_{\nu\mathbf{q}}, and polarization vector eνq\mathbf{e}_{\nu\mathbf{q}} [46]:

gmnν(k,q)=ℏ2ωνq∑καeκανqMκ⟨mk+q∣∂qκαV∣nk⟩(1)g_{mn\nu}(\mathbf{k},\mathbf{q})=\sqrt{\frac{\hbar}{2\omega_{\nu\mathbf{q}}}}\sum_{\kappa\alpha}\frac{e_{\kappa\alpha}^{\nu\mathbf{q}}}{\sqrt{M_{\kappa}}}\langle m\mathbf{k}+\mathbf{q}\lvert\partial_{\mathbf{q}\kappa\alpha}V\lvert n\mathbf{k}\rangle \tag*{(1)}

where ∂qκαV≡∑RpeiqRp∂pκαV\partial_{\mathbf{q}\kappa\alpha}V\equiv\sum_{\mathbf{R}_{p}}e^{i\mathbf{q}\mathbf{R}_{p}}\partial_{\mathbf{p}\kappa\alpha}V is the lattice-periodic ee-ph perturbation potential, given by the change in the DFT Kohn-Sham potential with respect to the position of atom κ\kappa (with mass MκM_{\kappa} and located in unit cell at Rp\mathbf{R}_{p}) in the Cartesian direction α\alpha. The inset in Fig. 1(a) shows schematically such an ee-ph scattering process. We separate the ee-ph interactions into short- and long-ranged [47, 48, 49, 50, 51, 52, 53],

gmnν(k,q)=gmnνL(k,q)+gmnνS(k,q).(2)g_{mn\nu}(\mathbf{k},\mathbf{q})=g_{mn\nu}^{L}(\mathbf{k},\mathbf{q})+g_{mn\nu}^{S}(\mathbf{k},\mathbf{q}). \tag*{(2)}

The long-range part gmnνL(k,q)g_{mn\nu}^{L}(\mathbf{k},\mathbf{q}) includes dipole (Fröhlich) and quadrupole contributions, which can be written analytically, using classical electromagnetism, in terms of Born effective charges and dynamical quadrupoles obtained from DFPT. The short-ranged part gmnνS(k,q)g_{mn\nu}^{S}(\mathbf{k},\mathbf{q}) cannot be written in closed form and needs numerical quantum mechanics to be computed, a consequence of the nearsightedness of electronic matter [54]. Because gmnνS(k,q)g_{mn\nu}^{S}(\mathbf{k},\mathbf{q}) is a smooth function of electron and phonon momenta, it is short-ranged in a real-space representation using a localized basis set such as atomic orbitals [55] or Wannier functions [56] [46].

The short-range ee-ph coupling matrix in Wannier basis, gijκα(Re,Rp)g_{ij}^{\kappa\alpha}(\mathbf{R}_{e},\mathbf{R}_{p}), is obtained by transforming DFPT results computed on a coarse momentum grid (kc,qc)(\mathbf{k}_{c},\mathbf{q}_{c}) [46]:

gijκα(Re,Rp)=1NkcNqc∑mnkc∑qce−i(kcRe+qcRp)Uim†(kc+qc)ΔVmn,καS(kc,qc)Unj(kc).(3)g_{ij}^{\kappa\alpha}(\mathbf{R}_{e},\mathbf{R}_{p})=\frac{1}{N_{\mathbf{k}_{c}}N_{\mathbf{q}_{c}}}\sum_{mn\mathbf{k}_{c}}\sum_{\mathbf{q}_{c}}e^{-i(\mathbf{k}_{c}\mathbf{R}_{e}+\mathbf{q}_{c}\mathbf{R}_{p})}U_{im}^{\dagger}(\mathbf{k}_{c}+\mathbf{q}_{c})\Delta V_{mn,\kappa\alpha}^{S}(\mathbf{k}_{c},\mathbf{q}_{c})U_{nj}(\mathbf{k}_{c}). \tag*{(3)}

where UU is a unitary transformation from Bloch to Wannier basis, and ΔVmn,καS(kc,qc)=⟨mk+q∣∂qκαVS∣nk⟩\Delta V_{mn,\kappa\alpha}^{S}(\mathbf{k}_{c},\mathbf{q}_{c})=\langle m\mathbf{k}+\mathbf{q}\lvert\partial_{\mathbf{q}\kappa\alpha}V^{S}\lvert n\mathbf{k}\rangle is the short-ranged part of the perturbation potential in Bloch basis. To separate acoustic and optical modes, we carry out a rotation in atomic basis:

gijμα(Re,Rp)=∑κAμκgijκα(Re,Rp),(4)g_{ij}^{\mu\alpha}(\mathbf{R}_{e},\mathbf{R}_{p})=\sum_{\kappa}A_{\mu}^{\kappa}g_{ij}^{\kappa\alpha}(\mathbf{R}_{e},\mathbf{R}_{p}), \tag*{(4)}

where Aμκ=exp⁡(i2πNatκμ)A_{\mu}^{\kappa}=\exp\left(i\frac{2\pi}{N_{\mathrm{at}}}\kappa\mu\right) adds a relative phase to different atoms in the unit cell, and μ∈(0,…,Nat−1)\mu\in(0,\ldots,N_{\mathrm{at}}-1) labels phonon modes (NatN_{\mathrm{at}} is the number of atoms in the unit cell). This way, μ=0\mu=0 corresponds to the acoustic subspace, where all the atoms in the unit cell move in phase, and μ≠0\mu\ne0 labels the optical modes. Here and below, we use a collective index F=(ij,μα)F=(ij,\mu\alpha) to label Wannier orbital pairs ijij and phonon mode and direction μα\mu\alpha, simplifying the notation of the Wannier-basis ee-ph matrices to gF(Re,Rp)g^{F}(\mathbf{R}_{e},\mathbf{R}_{p}).

When viewed as a matrix for each mode and orbital pair, gF(Re,Rp)g^{F}(\mathbf{R}_{e},\mathbf{R}_{p}) decays rapidly with lattice vectors Re\mathbf{R}_{e} and Rp\mathbf{R}_{p} and has a typical size NRe×NRpN_{\mathbf{R}_{e}}\times N_{\mathbf{R}_{p}}, ranging between 102×10210^{2}\times10^{2} and 103×10310^{3}\times10^{3}. After carrying out SVD on gF(Re,Rp)g^{F}(\mathbf{R}_{e},\mathbf{R}_{p}), we obtain

gF(Re,Rp)=∑γsγFuγF(Re)vγF∗(Rp),(5)g^{F}(\mathbf{R}_{e},\mathbf{R}_{p})=\sum_{\gamma}s_{\gamma}^{F}u_{\gamma}^{F}(\mathbf{R}_{e})v_{\gamma}^{F*}(\mathbf{R}_{p}), \tag*{(5)}

where sγFs_{\gamma}^{F} is the singular value (SV) with index γ\gamma, and uγF(Re)u_{\gamma}^{F}(\mathbf{R}_{e}) and vγF∗(Rp)v_{\gamma}^{F*}(\mathbf{R}_{p}) are the left and right singular vectors, respectively. One can interpret sγFs_{\gamma}^{F} as the coupling strength of the generalized electron cloud ∑ReuγF(Re)ci†(Re)cj(0)\sum_{\mathbf{R}_{e}}u_{\gamma}^{F}(\mathbf{R}_{e})c_{i}^{\dagger}(\mathbf{R}_{e})c_{j}(0) and phonon mode ∑RpvγF∗(Rp)(bμα†(Rp)+bμα(Rp))\sum_{\mathbf{R}_{p}}v_{\gamma}^{F*}(\mathbf{R}_{p})(b_{\mu\alpha}^{\dagger}(\mathbf{R}_{p})+b_{\mu\alpha}(\mathbf{R}_{p})), where (c†,c)(c^{\dagger},c) are creation and annihilation operators for electrons and (b†,b)(b^{\dagger},b) for phonons. For each channel FF, there is a total of min⁡(NRe,NRp)\min(N_{\mathbf{R}_{e}},N_{\mathbf{R}_{p}}) SVs; we keep only the NcN_{c} largest ones, resulting in a truncated, low-rank ee-ph matrix g~\tilde{g}:

g~NcF(Re,Rp)=∑γ=1NcsγFuγF(Re)vγF∗(Rp).(6)\tilde{g}_{N_{c}}^{F}(\mathbf{R}_{e},\mathbf{R}_{p})=\sum_{\gamma=1}^{N_{c}}s_{\gamma}^{F}u_{\gamma}^{F}(\mathbf{R}_{e})v_{\gamma}^{F*}(\mathbf{R}_{p}). \tag*{(6)}

This matrix can be conveniently transformed to momentum space using

g~NcF(k,q)=∑Rp,ReeikRe+iqRpg~NcF(Re,Rp)≈∑γ=1NcsγFuγF(k)vγF∗(q),(7)\begin{aligned} \tilde{g}_{N_{c}}^{F}(\mathbf{k},\mathbf{q})=\sum_{\mathbf{R}_{p},\mathbf{R}_{e}}e^{i\mathbf{k}\mathbf{R}_{e}+i\mathbf{q}\mathbf{R}_{p}}\tilde{g}_{N_{c}}^{F}(\mathbf{R}_{e},\mathbf{R}_{p}) \\ &\approx\sum_{\gamma=1}^{N_{c}}s_{\gamma}^{F}u_{\gamma}^{F}(\mathbf{k})v_{\gamma}^{F*}(\mathbf{q}), \tag*{(7)} \end{aligned}

where uγF(k)=∑ReeikReuγF(Re)u_{\gamma}^{F}(\mathbf{k})=\sum_{\mathbf{R}_{e}}e^{i\mathbf{k}\mathbf{R}_{e}}u_{\gamma}^{F}(\mathbf{R}_{e}) and vγF∗(q)=∑RpeiqRpvγF∗(Rp)v_{\gamma}^{F*}(\mathbf{q})=\sum_{\mathbf{R}_{p}}e^{i\mathbf{q}\mathbf{R}_{p}}v_{\gamma}^{F*}(\mathbf{R}_{p}) are singular vectors in momentum space. (Note that the long-range part of the ee-ph matrix is added after interpolation of this short-ranged part.) Eq. (7) provides a generic parametrization of ee-ph interactions, where by increasing the number of SVs one can systematically tune the accuracy and computational cost. According to the Eckart–Young-Mirsky theorem, the truncated matrix g~\tilde{g} obtained from SVD is an optimal low-rank approximation of ee-ph interactions, in the sense that it minimizes the Frobenius-norm distance between the original and low-rank ee-ph matrices [57]. From a computational viewpoint, Eq. (7) can greatly accelerate the calculation of ee-ph interactions and the associated material properties, with a speed up by the inverse fraction of SVs kept in the truncated ee-ph matrix. In most cases, we will keep only 1-2% of SVs, resulting in a 50−100 times speed-up for the key step in ee-ph calculations (see Appendix § 1.1.1 for details).

Error and Pareto-optimal interactions

To test the accuracy of the truncated ee-ph matrix and its convergence with respect to the number of SVs (NcN_c), we define a relative error for the ee-ph matrix averaged over electron bands and momenta, and phonon modes and wave-vectors:

ϵg(Nc)=∑mnν,kq∣gmnν(k,q)−g~mnνNc(k,q)∣2∑mnν,kq∣gmnν(k,q)∣2,(8)\epsilon_g(N_c) = \frac{\sum_{mn\nu,\mathbf{kq}} \left|g_{mn\nu}(\mathbf{k},\mathbf{q})-\widetilde{g}^{N_c}_{mn\nu}(\mathbf{k},\mathbf{q})\right|^2}{\sum_{mn\nu,\mathbf{kq}} \left|g_{mn\nu}(\mathbf{k},\mathbf{q})\right|^2}, \tag*{(8)}

where g~mnνNc(k,q)\widetilde{g}^{N_c}_{mn\nu}(\mathbf{k},\mathbf{q}) is the low-rank (approximate) and gmnν(k,q)g_{mn\nu}(\mathbf{k},\mathbf{q}) is the full first-principles ee-ph matrix. Fig. 1(a) shows this error as a function of the fraction of SVs, Nc/NRpN_c/N_{R_p}, kept in the approximate matrix. In the language of model selection [57], the resulting curve of error versus number of parameters is the Pareto frontier for modeling ee-ph interactions. We find that the error decreases rapidly with the number of SVs − for example, ϵg\epsilon_g is as low as 1% when using only 2% of SVs, which achieves a 98% compression of the original ee-ph matrix. This error curve defines a Pareto-optimal region, highlighted in Fig. 1(a), where ee-ph calculations are both accurate and parsimonious [57]. This region spans 1−4% of SVs in most of our calculations − which corresponds to keeping Nc≈10−50N_c \approx10-50 SVs − and suggests that many materials may possess only ∼\sim10 dominant elementary ee-ph interaction patterns. Accordingly, the ee-ph coupling strength for each phonon mode [58] [46],

Dν(q)=ℏ−12ωνqMuc∑mn∣gmnν(k=Γ,q)∣2/Nb,(9)D_\nu(\mathbf{q}) = \hbar^{-1}\sqrt{2\omega_{\nu\mathbf{q}}M_{\mathrm{uc}}\sum_{mn}\left|g_{mn\nu}(\mathbf{k}=\Gamma,\mathbf{q})\right|^2/N_b}, \tag*{(9)}

(where MucM_{\mathrm{uc}} is the mass of the unit cell and the band indices mm and nn run over NbN_b bands), shown in Fig. 1(b), can be computed accurately using just the largest 1.5% of SVs, matching closely results using the full ee-ph matrix.

Application to transport, spin and superconductivity

We showcase the accuracy of the low-rank approximate ee-ph interactions by computing a wide range of material properties, including charge mobility, spin relaxation, phonon-assisted superconductivity, and phonon-induced band renormalization. Figure 2(a) and Figure 2(b) show

Image not extracted

Figure 2. (a) Mobility of electrons in silicon, and (b) mobility of hole carriers in silicon, computed with the full e-ph matrix (NRe=NRp=1325N_{\mathrm{Re}} = N_{\mathrm{Rp}} = 1325) and compared with standard and constrained SVD, in both cases using 1.5% of the SVs. (c) Spin relaxation times of electrons in silicon, computed with the full e-ph matrix (NRe=NRp=1325N_{\mathrm{Re}} = N_{\mathrm{Rp}} = 1325) and using SVD with 1.5% of the SVs. Experimental data from Refs. [59, 60] are shown for comparison. (d) Eliashberg spectral function α2F(ω)\alpha^{2}F(\omega) for Pb, comparing full e-ph matrix (NRe=NRp=279N_{\mathrm{Re}} = N_{\mathrm{Rp}} = 279) with SVD results using 1.8% of the SVs. (e) Superconducting gap Δ0\Delta_{0} as a function of temperature, comparing full e-ph matrix results with SVD using 1.8% of the SVs. The inset shows the convergence of the critical temperature TcT_{\mathrm{c}} with number of SVs; the darker (lighter) colored regions indicate 5% (10%) error relative to the full calculation. (f) Band renormalization for electronic states near the Dirac cone in graphene at 20 K, comparing the full calculation with SVD using 1.5% of the SVs.

Figure 1. (a) Error on the compressed ee-ph matrix, computed using Eq. (8), as a function of the fraction of SVs used in the low-rank approximation. The Pareto-optimal region is shown with a shaded rectangle. (b) Mode-resolved ee-ph coupling strength computed using the full ee-ph matrix (blue) and the low-rank approximate matrix (orange) for silicon. The full ee-ph matrix elements are computed on a real-space grid with size NRe=1325N_{R_e}=1325 and NRp=1325N_{R_p}=1325, the smallest values to achieve convergence, setting the electron momentum to k=Γ\mathbf{k}=\Gamma and using the Nb=3N_b=3 highest valence bands. In the SVD calculation, we keep 20 out of 1325 SVs, corresponding to a ∼\sim1.5% fraction of SVs, as noted in the legend.

the electron and hole mobility in silicon for temperatures between 100−400 K, obtained using the full ee-ph matrix and compared with SVD using 1.5% of the SVs (see Appendix B). The mobility is overestimated for electrons, and underestimated for hole carriers, despite the accuracy of the low-rank ee-ph interactions in silicon (Fig. 1(a)). The error comes from the acoustic phonons, which interact weakly with electrons − and therefore are ignored in the low-rank ee-ph matrix − but carry a considerable contribution to the mobility due to their large thermal occupation. To improve the treatment of acoustic phonons, we develop a constrained SVD (c-SVD) which preserves the deformation potential [61] for long-wavelength acoustic phonons in the compressed ee-ph matrix (see Appendix C). When using c-SVD, the mobility computed using only 1.5% of the SVs is nearly

@page 3/12 starts_mid=1 ends_mid=1 low-rank approximation of ee-ph interactions, in the sense that it minimizes the Frobenius-norm distance between the original and low-rank ee-ph matrices [57]. From a computational viewpoint, Eq. (7) can greatly accelerate the calculation of ee-ph interactions and the associated material properties, with a speed up by the inverse fraction of SVs kept in the truncated ee-ph matrix. In most cases, we will keep only 1-2% of SVs, resulting in a 50−100 times speed-up for the key step in ee-ph calculations (see Appendix § 1.1.1 for details).

Error and Pareto-optimal interactions

To test the accuracy of the truncated ee-ph matrix and its convergence with respect to the number of SVs (NcN_c), we define a relative error for the ee-ph matrix averaged over electron bands and momenta, and phonon modes and wave-vectors:

ϵg(Nc)=∑mnν,kq∣gmnν(k,q)−g~mnνNc(k,q)∣2∑mnν,kq∣gmnν(k,q)∣2,(10)\epsilon_g(N_c) = \frac{\sum_{mn\nu,\mathbf{kq}} \left|g_{mn\nu}(\mathbf{k},\mathbf{q})-\widetilde{g}^{N_c}_{mn\nu}(\mathbf{k},\mathbf{q})\right|^2}{\sum_{mn\nu,\mathbf{kq}} \left|g_{mn\nu}(\mathbf{k},\mathbf{q})\right|^2}, \tag*{(10)}

where g~mnνNc(k,q)\widetilde{g}^{N_c}_{mn\nu}(\mathbf{k},\mathbf{q}) is the low-rank (approximate) and gmnν(k,q)g_{mn\nu}(\mathbf{k},\mathbf{q}) is the full first-principles ee-ph matrix. Fig. 1(a) shows this error as a function of the fraction of SVs, Nc/NRpN_c/N_{R_p}, kept in the approximate matrix. In the language of model selection [57], the resulting curve of error versus number of parameters is the Pareto frontier for modeling ee-ph interactions. We find that the error decreases rapidly with the number of SVs − for example, ϵg\epsilon_g is as low as 1% when using only 2% of SVs, which achieves a 98% compression of the original ee-ph matrix. This error curve defines a Pareto-optimal region, highlighted in Fig. 1(a), where ee-ph calculations are both accurate and parsimonious [57]. This region spans 1−4% of SVs in most of our calculations − which corresponds to keeping Nc≈10−50N_c \approx10-50 SVs − and suggests that many materials may possess only ∼\sim10 dominant elementary ee-ph interaction patterns. Accordingly, the ee-ph coupling strength for each phonon mode [58] [46],

Dν(q)=ℏ−12ωνqMuc∑mn∣gmnν(k=Γ,q)∣2/Nb,(11)D_\nu(\mathbf{q}) = \hbar^{-1}\sqrt{2\omega_{\nu\mathbf{q}}M_{\mathrm{uc}}\sum_{mn}\left|g_{mn\nu}(\mathbf{k}=\Gamma,\mathbf{q})\right|^2/N_b}, \tag*{(11)}

(where MucM_{\mathrm{uc}} is the mass of the unit cell and the band indices mm and nn run over NbN_b bands), shown in Fig. 1(b), can be computed accurately using just the largest 1.5% of SVs, matching closely results using the full ee-ph matrix.

Application to transport, spin and superconductivity

We showcase the accuracy of the low-rank approximate ee-ph interactions by computing a wide range of material properties, including charge mobility, spin relaxation, phonon-assisted superconductivity, and phonon-induced band renormalization. Figure 2(a) and Figure 2(b) show

Figure 1 (PDF p. 3). (a) Error on the compressed ee-ph matrix, computed using Eq. (8), as a function of the fraction of SVs used in the low-rank approximation. The Pareto-optimal region is shown with a shaded rectangle. (b) Mode-resolved ee-ph coupling strength computed using the full ee-ph matrix (blue) and the low-rank approximate matrix (orange) for silicon. The full ee-ph matrix elements are computed on a real-space grid with size NRe=1325N_{R_e}=1325 and NRp=1325N_{R_p}=1325, the smallest values to achieve convergence, setting the electron momentum to k=Γ\mathbf{k}=\Gamma and using the Nb=3N_b=3 highest valence bands. In the SVD calculation, we keep 20 out of 1325 SVs, corresponding to a ∼\sim1.5% fraction of SVs, as noted in the legend.

the electron and hole mobility in silicon for temperatures between 100−400 K, obtained using the full ee-ph matrix and compared with SVD using 1.5% of the SVs (see Appendix B). The mobility is overestimated for electrons, and underestimated for hole carriers, despite the accuracy of the low-rank ee-ph interactions in silicon (Fig. 1(a)). The error comes from the acoustic phonons, which interact weakly with electrons − and therefore are ignored in the low-rank ee-ph matrix − but carry a considerable contribution to the mobility due to their large thermal occupation. To improve the treatment of acoustic phonons, we develop a constrained SVD (c-SVD) which preserves the deformation potential [61] for long-wavelength acoustic phonons in the compressed ee-ph matrix (see Appendix C). When using c-SVD, the mobility computed using only 1.5% of the SVs is nearly identical to the full-matrix result for both carriers.

We also apply the low-rank approximation to spin-dependent e-ph matrices governing spin-flip e-ph interactions; these matrices enable first-principles calculations of spin relaxation times (SRTs) in centrosymmetric materials via the Elliot-Yafet mechanism [62] (see Appendix § 1.1.6). The SRTs for electrons in silicon between 150–400 K are shown in Figure 2(c). Our results from SVD with Nc=20N_{\mathrm{c}} = 20 (corresponding to ∼1.5%\sim1.5\% of the SVs) match closely the full e-ph matrix calculations and agree with experimental results [59, 60]. Different from charge transport, standard SVD gives accurate SRTs in silicon because the optical phonons govern spin-flip processes. For materials where acoustic phonons contribute to spin relaxation, our c-SVD approach can be readily extended to the spin-dependent case.

Calculations of phonon-mediated superconductivity, presented here using lead (Pb) as an example, can also leverage our low-rank approximation (see Appendix § 1.1.7). Figure 2(d) compares the Eliashberg spectral function α2F(ω)\alpha^{2}F(\omega) from full e-ph matrix calculations with SVD results; using only the first Nc=5N_{\mathrm{c}} = 5 SVs (here equal to 1.8% of the SVs) suffices to reproduce the full calculation. We solve the isotropic Eliashberg equation self-consistently (see Methods for details) and compute the superconducting gap Δ0\Delta_{0} as a function of temperature (Figure 2(e)) as well as the critical temperature TcT_{\mathrm{c}} versus number of SVs (inset of Figure 2(e)). The low-rank e-ph matrix with Nc=5N_{\mathrm{c}} = 5 SVs provides a gap function in good agreement with the full-matrix calculation, which can be further improved by using a larger fraction of SVs; the critical temperature for Nc=5N_{\mathrm{c}} = 5 SVs is very accurate, Tc=6.9T_{\mathrm{c}} = 6.9 K, which is within 5% of the Tc=6.6T_{\mathrm{c}} = 6.6 K value obtained using the full e-ph matrix. This result implies that as few as five elementary e-ph interactions determine TcT_{\mathrm{c}} in Pb. Finally, we compute band renormalization from e-ph interactions [64] focusing on the contribution from the Fan-Migdal e-ph self-energy (see Appendix § 1.9). The Debye-Waller term can also be added following Ref. [65]. Results for graphene show that band renormalization near the Dirac cone can be computed keeping only the five largest SVs; this highly compressed e-ph matrix correctly predicts the kinks near the Dirac cone and matches full e-ph matrix results over a 2 eV energy range. We also carry out convergence tests with respect to the fraction of SVs included in the calculation for all the materials studied here (see the Supplemental Material [66]). The rapid convergence with respect to the fraction of SVs guarantees that the full e-ph matrix calculation is not needed, and one can obtain accurate results by converging the desired property with respect to the fraction of SVs with minimal computational overhead.

While all the examples discussed above are for materials with nonpolar bonds, polar materials are even simpler to study with our compression method because the long-range (Fröhlich) dipole contribution is dominant and is well modeled by an analytic formula using Born effective charges [47, 48, 49]. To illustrate this point, we demonstrate accurate mobility calculations in GaAs and PbTiO3_3 using only 1% of the SVs (see the Supplemental Material [66]).

Computational speed-up from compression

We illustrate the computational speed-up achieved by our compression method using the ee-ph coupling constant λ\lambda in doped monolayer MoS2_2 as a case study [67]. Following Ref. [67], we employ a grid size of 2882288^2 kk- and qq-points for numerical integration and a Gaussian smearing of 0.002 Ry. We also leverage the improved Brillouin-zone sampling technique where only electronic states in a small energy window (0.006 Ry) near the Fermi surface are included in the calculation. In Figure 3(a), we show the convergence of λ\lambda with respect to the fraction of SVs. The shaded region corresponds to an accuracy greater than 95% compared to the fully converged result. Similar to other quantities computed in this work, λ\lambda converges rapidly with the fraction of SVs; in particular, using only 1% of the SVs gives λ\lambda within 2% of the converged result. Figure 3(b) compares the computational wall time for the calculation employing the full ee-ph matrices and for our SVD compression method with 1% SVs (note that both calculations use the same improved Brillouin-zone sampling scheme). Our approach achieves a speed-up by 38 times relative to using the full ee-ph matrices; if we count solely the time for ee-ph interpolation, the speed-up is 83 times. This example illustrates the 1–2 orders of magnitude speed-up deriving from compressing the ee-ph matrices with SVD. For a more detailed analysis of computational complexity, see Appendix § 1.1.1.

Image not extracted

Figure 3. (a) ee-ph coupling constant λ\lambda in doped monolayer MoS2_2 computed with our compression method using different fractions of SV. The doping concentration is 0.22 electrons per formula unit. The shaded region corresponds to an accuracy greater than 95% relative to the fully converged calculation. (b) Comparison of the wall time for computing the ee-ph coupling constant λ\lambda with the full ee-ph matrices and with our SVD compression technique using 1% of the SVs. The 38× speed-up achieved by the SVD compression is indicated in red font.

Dominant modes and principal-component analysis

To understand the inherent compressibility of the ee-ph matrices, we analyze the SV spectrum in graphene and silicon (see Figure 4(a) and Figure 4(b)). In both materials, the SVs decay rapidly, dropping by 1–2 orders of magnitude from the largest to the 10th largest SV. In principal component analysis (PCA) [57, 63], this decay can be understood as a consequence of high-variance generalized directions in the ee-ph matrix, gF(Re,Rp)g^{F}(R_e,R_p), which capture the vast majority of the physics, while other principal components can be viewed as noise and neglected [63]. We carry out PCA by treating each row of the matrix g(Re,Rp)g(R_e,R_p) as a feature vector, and find that the variance of the two leading principal components is one order of magnitude greater than for the 10th or following principal components, indicating that most of the physical information is already captured by the first few SVs (see the insets of Figure 4(a) and Figure 4(b)). This analysis reveals that only a few atomic vibrational patterns dominate ee-ph coupling. Although these dominant modes are not known a priori, they can be learned efficiently with SVD. We also apply the PCA to lead (see Fig. S4 in Supplemental Material [66]) and observe a similar rapid decay of the SVs. We remark that the dimensionality reduction is general – it occurs in all the materials studied here, and it is associated to the rapid decay of the SVs, which we view as a consequence of the nearsightedness of electronic interactions.

Image not extracted

Figure 4. (a) Decay of the SVs in graphene and (b) decay of the SVs in silicon, shown by plotting the SVs referenced to the largest SV. Here, s~i\tilde{s}_i refers to the SVs averaged over Wannier orbitals and vibrational modes, s~i=∑F(siF)2\tilde{s}_i = \sqrt{\sum_F (s_i^F)^2}. The respective insets show the real part of the 2nd versus 10th principal components, obtained from the PCA of the ee-ph matrices; in parentheses we give the fraction of explained variance [63], λ~i=σi2/∑iσi2\tilde{\lambda}_i = \sigma_i^2/\sum_i \sigma_i^2, where σi2\sigma_i^2 is the variance of the ii-th principal component. The red oval shows the standard deviation of the corresponding principal components, obtained by dropping feature vectors with norm smaller than 10−3×s~010^{-3} \times\tilde{s}_0. (c) Atomic vibrations associated with the dominant ee-ph interactions in graphene, and (d) the same quantity in silicon, obtained by analyzing the phonon singular vectors, v~(καRp)\tilde{v}(\kappa\alpha R_p) in Eq. (12), for the two largest SVs.

We visualize the atomic vibrations with dominant ee-ph interactions by analyzing the vibrational singular vectors. To that end, we introduce a modified SVD that includes Wannier orbitals and phonon modes in the decomposition:

gijκα(Re,Rp)=∑γsγu~γ(ijRe)v~γ∗(καRp).(12)g_{ij}^{\kappa\alpha}(R_e,R_p)=\sum_{\gamma}s_\gamma\tilde{u}_\gamma(ijR_e)\tilde{v}_\gamma^*(\kappa\alpha R_p). \tag*{(12)}

In this global SVD, the singular vectors u~\tilde{u} depend only on electron variables and v~\tilde{v} only on phonon variables. This way, the phonon singular vectors v~∗(καRp)\tilde{v}^*(\kappa\alpha R_p) can be interpreted as local vibrational modes (in the Wigner-Seitz cell associated with the coarse grid [46]) and visualized to study the dominant ee-ph couplings. We show these singular vectors for the two modes with largest SVs in graphene and silicon in Figure 4(c) and Figure 4(d), using arrows on each atom, with length proportional to the singular vector v~γ∗(καRp)\tilde{v}_\gamma^*(\kappa\alpha R_p), to indicate the atomic displacements in the modes obtained from SVD.

In graphene, where the electronic states consist of pzp_z orbitals centered on each carbon atom, the dominant mode resembles a longitudinal optical phonon that brings the pzp_z orbitals closer together in the unit cell. The second-strongest mode is a shear vibration resembling a transverse optical phonon, which spreads over multiple unit cells (Figure 4(c)).

For the other modes, we observe that the vibrational pattern progressively delocalizes over multiple unit cells for decreasing values of the SVs. In silicon, where the electronic states consist of sp3sp^3-like Wannier orbitals oriented along the chemical bond directions, the two modes with dominant ee-ph coupling are associated with compression and stretching of the bonds (Figure 4(d)). The intuition gained from this mode analysis can aid the formulation of model Hamiltonians in chemically and structurally complex materials, where keeping only the dominant ee-ph interactions can provide effective models of transport and polaron physics informed by first-principles calculations [68, 69, 70, 71, 72].

Discussion and conclusion

The accuracy of the low-rank ee-ph matrices implies that current brute-force first-principles calculations over-parametrize ee-ph interactions, falling too far on the right side of the Pareto-optimal region in Fig. 1(a). Conversely, textbook approaches such as Holstein and Fröhlich models, which use only a handful of ee-ph couplings, may fall short of achieving quantitative accuracy by using too few parameters. Our SVD compression in Wannier basis (followed by interpolation) provides a systematic route to achieve Pareto-optimal calculations. These optimal models enhance interpretability and enable a deeper understanding because they concentrate all the relevant ee-phph physics in just a few parameters − in our case, the leading SVs and singular vectors, which represent dominant ee-phph interactions.

In summary, our results unveil the hidden low-dimensional nature of ee-phph interactions. While accurate, current first-principles calculations overparametrize these interactions due to a lack of a priori knowledge of the dominant atomic vibrational patterns governing ee-phph coupling. We have shown that when this optimal representation is achieved via SVD, using only 10–20 parameters (for each orbital pair and vibrational mode) is sufficient to obtain results with state-of-the-art accuracy. Surprisingly, this is only a small fraction (1–2%) of the typical size of first-principles ee-phph matrices. Compressing ee-phph interactions significantly accelerates calculations of material properties ranging from transport to spin relaxation to superconductivity. Future work will extend these ideas to other electronic interactions, with the goal of advancing “precise but parsimonious” quantum many-body calculations in real materials. Our approach works equally as well for small systems with a few atoms and for large systems with tens of atoms in the unit cell (see Fig. S5 in Supplemental Material [66]). In addition, as we plan to show elsewhere, our method enables many-body ee-phph calculations that are currently inaccessible with standard Wannier interpolation, including the development of first-principles diagrammatic Monte Carlo, which sums ee-phph diagrams to all orders from first principles, thus providing a gold standard for quantitative ee-phph studies.

Acknowledgments

Y.L. thanks Ivan Maliyov and Junjie Yang for fruitful discussions. This work was supported by the National Science Foundation under Grant No. OAC-2209262. This research used resources of the National Energy Research Scientific Computing Center (NERSC), a U.S. Department of Energy Office of Science User Facility located at Lawrence Berkeley National Laboratory, operated under Contract No. DE-AC02-05CH11231.

Computational complexity analysis

From a computational viewpoint, Eq. (7) can greatly accelerate the calculation of ee-phph interactions and the associated material properties. The key bottleneck in these calculations is obtaining the ee-phph matrix elements on fine momentum grids, gmnν(kf,qf)g_{mn\nu}(\mathbf{k}_{f},\mathbf{q}_{f}), starting from the Wannier representation, with a cost scaling as O(NRpNkfNqf)\mathcal{O}(N_{R_{p}}N_{k_{f}}N_{q_{f}}) for an optimal implementation [46] (for a fixed number of Wannier functions and atoms in the unit cell), where NkfN_{k_{f}} and NqfN_{q_{f}} are the number of points in the fine-momentum electron and phonon grids, with typical values of order Nkf≈Nqf≈106N_{k_{f}}\approx N_{q_{f}}\approx10^{6}. In contrast, when using SVD, this interpolation step costs only O(NcNkfNqf)\mathcal{O}(N_{c}N_{k_{f}}N_{q_{f}}), with a speed up by a factor NRp/NcN_{R_{p}}/N_{c}, the inverse fraction of SVs kept in the truncated ee-phph matrix. In most cases, we will keep only 1–2% of SVs, resulting in a 50–100 times speed-up for the key step in ee-phph calculations. We show specific timing comparisons for all the materials studied in this work in Fig. S1 of the Supplemental Material [66]. In all cases, our algorithm achieves a speed up close to the ideal value of NRp/NcN_{R_{p}}/N_{c}. The memory improvement is also dramatic. A converged transport calculation in silicon requires a k\mathbf{k} grid of 1003100^{3} and q\mathbf{q} grid of 50350^{3} points [46]; on these fine grids, the memory required to store the entire ee-phph matrix gmnν(kf,qf)g_{mn\nu}(\mathbf{k}_{f},\mathbf{q}_{f}) is 700 TB, while the memory needed to store the singular vectors uνγ(kf)u^{\gamma}_{\nu}(\mathbf{k}_{f}) and vνγ(qf)v^{\gamma}_{\nu}(\mathbf{q}_{f}) is only 128 GB when we retain 1.5% of SVs, which guarantees accurate results as we show in Figs. 2(a) and 2(b). This efficiency removes the key bottleneck in first-principles ee-phph calculations.

Mobility calculations

The first-principles mobility calculations in silicon follows our previous work [51]. We include the quadrupole contribution analytically for silicon. The quadrupole tensor can be written as

Qsi,αβγ=(−1)κ+1Qsi∣ϵαβγ∣(13)Q_{\mathrm{si},\alpha\beta\gamma}=(-1)^{\kappa+1}Q_{\mathrm{si}}|\epsilon_{\alpha\beta\gamma}| \tag*{(13)}

where ϵαβγ\epsilon_{\alpha\beta\gamma} is the Levi-Civita tensor and the value of Qsi=13.67Q_{\mathrm{si}}=13.67 is taken from Refs. [73, 74].

We compute the phonon-limited mobility at temperature TT using the BTE in the relaxation time approximation (RTA) [46]. We first obtain the ee-phph scattering rate Γnk\Gamma_{n\mathbf{k}} using Fermi’s golden rule, which is equivalent to using the imaginary part of the lowest-order ee-phph self-energy [75]:

Γnk=2πℏ1Nq∑mνq∣gmnν(k,q)∣2×[(Nνq+1−fmk+q)δ(ϵnk−ϵmk+q−ℏωνq)+(Nνq+fmk+q)δ(ϵnk−ϵmk+q+ℏωνq)](14)\begin{aligned} \Gamma_{n\mathbf{k}}=\frac{2\pi}{\hbar}\frac{1}{N_{q}}\sum_{m\nu\mathbf{q}}|g_{mn\nu}(\mathbf{k},\mathbf{q})|^{2} \\ &\times\left[(N_{\nu\mathbf{q}}+1-f_{m\mathbf{k}+\mathbf{q}})\delta(\epsilon_{n\mathbf{k}}-\epsilon_{m\mathbf{k}+\mathbf{q}}-\hbar\omega_{\nu\mathbf{q}}) +(N_{\nu\mathbf{q}}+f_{m\mathbf{k}+\mathbf{q}})\delta(\epsilon_{n\mathbf{k}}-\epsilon_{m\mathbf{k}+\mathbf{q}}+\hbar\omega_{\nu\mathbf{q}})\right] \tag*{(14)} \end{aligned}

where NqN_{q} is the number of q\mathbf{q} points and δ\delta is the Dirac delta function. Then we obtain the mobility from the BTE by summing over contributions from different electronic states and scattering processes [46]:

μαβ(T)=encΩNk∫dE(−∂f∂E)∑nkτnkvnkαvnkβδ(E−ϵnk)(15)\mu_{\alpha\beta}(T)=\frac{e}{n_{c}\Omega N_{k}}\int dE\left(-\frac{\partial f}{\partial E}\right)\sum_{n\mathbf{k}}\tau_{n\mathbf{k}}v^{\alpha}_{n\mathbf{k}}v^{\beta}_{n\mathbf{k}}\delta(E-\epsilon_{n\mathbf{k}}) \tag*{(15)}

where Ω\Omega is the volume of the unit cell, τnk=(Γnk)−1\tau_{n\mathbf{k}}=(\Gamma_{n\mathbf{k}})^{-1} are relaxation times, ncn_{c} is the carrier concentration, $f is the Fermi-Dirac distribution, and NkN_k is the number of kk points; ϵnk\epsilon_{nk} and vnkv_{nk} are electron energies and band velocities, respectively. Our calculations in silicon use a uniform grid with 2003200^3 kk points and a uniform random grid with 10510^5 qq points, where kk and qq are electron and phonon momenta respectively. The delta function is approximated as a Gaussian with a 10 meV smearing [46].

c-SVD algorithm

Let us briefly describe our c-SVD algorithm. Similar to the acoustic sum rule (ASR) for the dynamical matrix [19], we formulate an ASR for the ee-ph matrix elements:

gνA(k,q=0)=0,(16)g_{\nu_A}(\mathbf{k},\mathbf{q}=0)=0, \tag*{(16)}

where νA\nu_A labels the acoustic modes, and we omit band indices for simplicity. The rationale for this ee-ph ASR is that a rigid translation of the lattice will not change the electronic band structure. The real-space version of this ee-ph ASR reads

∑κ,Rpgijκα(Re,Rp)=∑Rpgijμ=0,α(Re,Rp)=0,(17)\sum_{\kappa,\mathbf{R}_p}g_{ij}^{\kappa\alpha}(\mathbf{R}_e,\mathbf{R}_p)=\sum_{\mathbf{R}_p}g_{ij}^{\mu=0,\alpha}(\mathbf{R}_e,\mathbf{R}_p)=0, \tag*{(17)}

where gijμ=0,α(Re,Rp)g_{ij}^{\mu=0,\alpha}(\mathbf{R}_e,\mathbf{R}_p) accounts for the acoustic subspace of the ee-ph matrix defined in Eq. (4). With this ASR, the ee-ph matrix for long-wavelength acoustic phonons can be approximated to first order in q\mathbf{q} as

lim⁡q→0gμ=0,α(Re,q)=lim⁡q→0∑Rpgμ=0,α(Re,Rp)eiqRp≈iq⋅Aα(Re),(18)\lim_{\mathbf{q}\to0}g^{\mu=0,\alpha}(\mathbf{R}_e,\mathbf{q})=\lim_{\mathbf{q}\to0}\sum_{\mathbf{R}_p}g^{\mu=0,\alpha}(\mathbf{R}_e,\mathbf{R}_p)e^{i\mathbf{q}\mathbf{R}_p} \approx i\mathbf{q}\cdot\mathbf{A}^{\alpha}(\mathbf{R}_e), \tag*{(18)}

where we defined a real-space deformation potential, which in general can be anisotropic, as

Aα(Re)≡∑RpRpgμ=0,α(Re,Rp).(19)\mathbf{A}^{\alpha}(\mathbf{R}_e)\equiv\sum_{\mathbf{R}_p}\mathbf{R}_p g^{\mu=0,\alpha}(\mathbf{R}_e,\mathbf{R}_p). \tag*{(19)}

In the limit of ∣q∣→0|\mathbf{q}|\to0, g~μ=0,α(Re,q)\tilde{g}^{\mu=0,\alpha}(\mathbf{R}_e,\mathbf{q}) vanishes linearly in ∣q∣|\mathbf{q}|, but the phonon occupation number diverges as 1∣q∣\frac{1}{|\mathbf{q}|}; therefore in the long-wavelength limit acoustic phonon scattering is often important. This acoustic phonon contribution is challenging to preserve when using the compressed ee-ph matrices because standard SVD primarily captures large entries in the ee-ph matrix. To address this point, we compress the ee-ph matrix while conserving Aα(Re)\mathbf{A}^{\alpha}(\mathbf{R}_e) by imposing the following constraint:

∑RpRpg~Ncμ=0,α(Re,Rp)=Aα(Re).(20)\sum_{\mathbf{R}_p}\mathbf{R}_p\tilde{g}^{\mu=0,\alpha}_{N_c}(\mathbf{R}_e,\mathbf{R}_p)=\mathbf{A}^{\alpha}(\mathbf{R}_e). \tag*{(20)}

Satisfying this set of linear equations leads to a constrained low-rank approximation [76], a more general optimization problem. The c-SVD is applied only to the acoustic subspace, which corresponds to F=(ij,μ=0α)F=(ij,\mu=0\alpha), resulting in a compressed ee-ph matrix of rank NcN_c:

g~NcF(Re,Rp)=g~Nc−3F(Re,Rp)+∑ββ′∈(x,y,z)(δAα(Re))βλββ′(Rp)β′,(21)\tilde{g}^{F}_{N_c}(\mathbf{R}_e,\mathbf{R}_p)=\tilde{g}^{F}_{N_c-3}(\mathbf{R}_e,\mathbf{R}_p) +\sum_{\beta\beta'\in(x,y,z)}\left(\delta\mathbf{A}^{\alpha}(\mathbf{R}_e)\right)_{\beta}\lambda_{\beta\beta'}(\mathbf{R}_p)_{\beta'}, \tag*{(21)}

where g~Nc−3F(Re,Rp)\tilde{g}^{F}_{N_c-3}(\mathbf{R}_e,\mathbf{R}_p) is the truncated SVD of the ee-ph matrix with Nc−3N_c-3 singular values (see Eq. (6)); δAα\delta\mathbf{A}^{\alpha} is the residual term for g~Nc−3F(Re,Rp)\tilde{g}^{F}_{N_c-3}(\mathbf{R}_e,\mathbf{R}_p), defined as

δAα(Re)=Aα(Re)−∑RpRpg~Nc−3μ=0,α(Re,Rp),(22)\delta\mathbf{A}^{\alpha}(\mathbf{R}_e)=\mathbf{A}^{\alpha}(\mathbf{R}_e)-\sum_{\mathbf{R}_p}\mathbf{R}_p\tilde{g}^{\mu=0,\alpha}_{N_c-3}(\mathbf{R}_e,\mathbf{R}_p), \tag*{(22)}

and λββ′\lambda_{\beta\beta'} is the inverse of the overlap matrix between Rp\mathbf{R}_p vectors:

∑β′λββ′∑Rp(Rp)β′(Rp)β′′=δββ′′.(23)\sum_{\beta'}\lambda_{\beta\beta'}\sum_{\mathbf{R}_p}(\mathbf{R}_p)_{\beta'}(\mathbf{R}_p)_{\beta''}=\delta_{\beta\beta''}. \tag*{(23)}

Using this approach, the compressed ee-ph matrix g~NcF(Re,Rp)\tilde{g}^{F}_{N_c}(\mathbf{R}_e,\mathbf{R}_p) gives the same deformation potential as the full ee-ph matrix in Eq. (19), and its rank is smaller than or equal to NcN_c. This c-SVD workflow requires only minimal computational overhead relative to standard SVD.

Spin relaxation times

The first-principles calculation of SRTs in silicon follows our recent work [62]. The spin-flip relaxation time τnkflip\tau^{\mathrm{flip}}_{nk}, for a band electron in state ∣nk⟩|nk\rangle, accounts for the Elliott-Yafet spin relaxation mechanism and is computed using [62, 77, 78]

1τnkflip=4πℏ∑mνq∣gmnνflip(k,q)∣2×[(Nνq+1−fmk+q)δ(ϵnk−ϵmk+q−ℏωνq)+(Nνq+fmk+q)δ(ϵnk−ϵmk+q+ℏωνq)].(24)\frac{1}{\tau^{\mathrm{flip}}_{nk}}=\frac{4\pi}{\hbar}\sum_{m\nu\mathbf{q}}\left|g^{\mathrm{flip}}_{mn\nu}(\mathbf{k},\mathbf{q})\right|^2 \times\left[(N_{\nu\mathbf{q}}+1-f_{m\mathbf{k}+\mathbf{q}})\delta(\epsilon_{nk}-\epsilon_{m\mathbf{k}+\mathbf{q}}-\hbar\omega_{\nu\mathbf{q}}) +(N_{\nu\mathbf{q}}+f_{m\mathbf{k}+\mathbf{q}})\delta(\epsilon_{nk}-\epsilon_{m\mathbf{k}+\mathbf{q}}+\hbar\omega_{\nu\mathbf{q}})\right]. \tag*{(24)}

The key ingredients in this equation are the spin-flip ee-ph matrix elements,

gmnνflip(k,q)=⟨mk+q↓∣ΔV^νq∣nk↑⟩,(25)g^{\mathrm{flip}}_{mn\nu}(\mathbf{k},\mathbf{q})=\langle m\mathbf{k}+\mathbf{q}\mathord{\downarrow}|\Delta\hat{V}_{\nu\mathbf{q}}|n\mathbf{k}\mathord{\uparrow}\rangle, \tag*{(25)}

with ↓\downarrow and ↑\uparrow denoting nearly spin-down and nearly spin-up states, respectively. These matrix elements describe the probability amplitude to flip the spin of a band electron due to a particular phonon mode νq\nu\mathbf{q}. The macroscopic spin relaxation time, τs(T)\tau_s(T), is a thermal average over electronic states of the spin-flip scattering rates [62]:

τs(T)=⟨1τnkflip⟩T−1=(∑nk1τnkflip(−dfnkdE)∑nk(−dfnkdE))−1.(26)\tau_s(T)=\left\langle\frac{1}{\tau^{\mathrm{flip}}_{nk}}\right\rangle_T^{-1}=\left(\frac{\sum_{nk}\frac{1}{\tau^{\mathrm{flip}}_{nk}}\left(-\frac{df_{nk}}{dE}\right)}{\sum_{nk}\left(-\frac{df_{nk}}{dE}\right)}\right)^{-1}. \tag*{(26)}

The SRT calculations employ a uniform grid with up to 1403140^3 kk points and a 5 meV Gaussian smearing for the delta functions.

Eliashberg spectral function and superconducting gap

We carry out DFT calculations on lead (Pb) using the generalized gradient approximation [79] in the Quantum ESPRESSO code [80]. The ground state and electron wave functions are computed on a 14×14×1414 \times14 \times14 kk-point grid with a kinetic energy cutoff of 90 Ry, and the lattice constant is set to 4.88 Å. We use DFPT to calculate the phonon frequencies and eigenvectors, and the e-ph matrix elements gmnν(k,q)g_{mn\nu}(\mathbf{k},\mathbf{q}), on coarse 6×6×66 \times6 \times6 kk- and qq-point grids. We wannierize the 4 bands near the Fermi surface using the Wannier90 code [81], and obtain the e-ph matrices in Wannier basis using Perturbo [46]. The Eliashberg spectral function is computed as

α2F(ω)=12∑νqωνqλνqδ(ω−ωνq),with\alpha^{2}F(\omega) = \frac{1}{2}\sum_{\nu q}\omega_{\nu q}\lambda_{\nu q}\delta(\omega-\omega_{\nu q}), \qquad\text{with}
λνq=1N(ϵF)ωνq∑mnk∣gmnν(k,q)∣2(27)\lambda_{\nu q} = \frac{1}{N(\epsilon_{\mathrm{F}})\omega_{\nu q}}\sum_{mn\mathbf{k}} \lvert g_{mn\nu}(\mathbf{k},\mathbf{q})\rvert^{2} \tag*{(27)}
×δ(ϵnk−ϵF)δ(ϵmk+q−ϵF).{}\times\delta(\epsilon_{n\mathbf{k}}-\epsilon_{\mathrm{F}})\delta(\epsilon_{m\mathbf{k}+\mathbf{q}}-\epsilon_{\mathrm{F}}).

where N(ϵF)N(\epsilon_{\mathrm{F}}) is the density of states at the Fermi energy (ϵF\epsilon_{\mathrm{F}}). The Eliashberg function α2F(ω)\alpha^{2}F(\omega) encodes the isotropic and retarded effective attraction between electronic states on the Fermi surface. Using α2F(ω)\alpha^{2}F(\omega), we obtain the gap function by solving the isotropic Migdal-Eliashberg Eq. [82]:

Z(iωj)=1+πkBTωj∑j′ωj′ωj′2+Δ2(iωj′)λ(ωj−ωj′),Z(i\omega_j) = 1 + \frac{\pi k_{\mathrm{B}}T}{\omega_j}\sum_{j'}\frac{\omega_{j'}}{\sqrt{\omega_{j'}^{2}+\Delta^{2}(i\omega_{j'})}}\lambda(\omega_j-\omega_{j'}),
Z(iωj)Δ(iωj)=πkBT∑j′Δ(iωj′)ωj′2+Δ2(iωj′)[λ(ωj−ωj′)−μc∗](28)Z(i\omega_j)\Delta(i\omega_j) = \pi k_{\mathrm{B}}T\sum_{j'}\frac{\Delta(i\omega_{j'})}{\sqrt{\omega_{j'}^{2}+\Delta^{2}(i\omega_{j'})}}\left[\lambda(\omega_j-\omega_{j'})-\mu_c^{*}\right] \tag*{(28)}

where TT is the temperature, ωj=(2j+1)πkBT\omega_j=(2j+1)\pi k_{\mathrm{B}}T is the Matsubara frequency, μc∗\mu_c^{*} is the screened Coulomb potential, Z(iωj)Z(i\omega_j) is the mass renormalization function, Δ(iωj)\Delta(i\omega_j) is the superconducting gap function, and

λ(ωj)=∫0∞dωα2F(ω)2ωωj2+ω2\lambda(\omega_j)=\int_{0}^{\infty}d\omega\alpha^{2}F(\omega)\frac{2\omega}{\omega_j^{2}+\omega^{2}}

is the isotropic e-ph coupling strength.

For the numerical integrations in Eq. (27), we employ a kk-point grid consisting of 400,000 quasi-random Sobol points (generated using SciPy[83]) and a qq-point grid with 30,000 uniformly distributed random points; the delta functions are approximated as Gaussians with a 30 meV smearing for electrons and 0.1 meV smearing for phonons. Using the converged α2F(ω)\alpha^{2}F(\omega) function, we set μc∗=0.1\mu_c^{*}=0.1 and solve Eq. (28) iteratively for a range of temperatures. The critical temperature TcT_c is obtained as the temperature where Δo=Δ(iωj=iπkBT)\Delta_o=\Delta(i\omega_j=i\pi k_{\mathrm{B}}T) extrapolates to zero.

Band structure renormalization

The DFT ground state calculation in graphene uses the local density approximation with a norm-conserving pseudopotential from Pseudo DOJO[84]. We employ a 90 Ry plane-wave kinetic energy cutoff, a 60×60×160 \times60 \times1 kk-point grid, and a 2.46 Å lattice constant. For the DFPT calculation, we use coarse grids with 36×36×136 \times36 \times1 kk-points for electrons and 18×18×118 \times18 \times1 qq-points for phonons. The band structure renormalized by e-ph interactions, ϵ~nk\tilde{\epsilon}_{n\mathbf{k}}, is obtained as the DFT band structure plus the real part of the e-ph self-energy evaluated on-shell,

ϵ~nk=ϵnk+Re⁡Σnk(E=ϵnk,T).(29)\tilde{\epsilon}_{n\mathbf{k}}=\epsilon_{n\mathbf{k}}+\operatorname{Re}\Sigma_{n\mathbf{k}}(E=\epsilon_{n\mathbf{k}},T). \tag*{(29)}

We use the lowest-order (Fan-Migdal) e-ph self-energy,

Σnk(E,T)=1Nq∑qν∣gmnν(k,q)∣2×[Nνq+1−fmk+qE−ϵmk+q−ℏωνq−iη+Nνq+fmk+qE−ϵmk+q+ℏωνq−iη],(30)\begin{aligned} \Sigma_{n\mathbf{k}}(E,T)=\frac{1}{N_q}\sum_{q\nu}\lvert g_{mn\nu}(\mathbf{k},\mathbf{q})\rvert^{2} \\ &\times\left[\frac{N_{\nu q}+1-f_{m\mathbf{k}+\mathbf{q}}}{E-\epsilon_{m\mathbf{k}+\mathbf{q}}-\hbar\omega_{\nu q}-i\eta}+\frac{N_{\nu q}+f_{m\mathbf{k}+\mathbf{q}}}{E-\epsilon_{m\mathbf{k}+\mathbf{q}}+\hbar\omega_{\nu q}-i\eta}\right], \tag*{(30)} \end{aligned}

where η\eta is a Lorentzian smearing [85]. We employ 10710^{7} uniform random qq-points for the numerical integration in Eq. (30) and set the Lorentzian smearing to 15 meV. For more accurate band renormalization calculations, one could use Wannier function perturbation theory (WFPT) to overcome errors resulting from finite number of Wannier functions [86].

References

  1. [1]B. I. Dunlap, Robust and variational fitting, Phys. Chem. Chem. Phys. 2, 2113 (2000).
  2. [2]B. Q. Pham and M. S. Gordon, Compressing the four-index two-electron repulsion integral matrix using the resolution-of-the-identity approximation combined with the rank factorization approximation, J. Chem. Theory Comput. 15, 2254 (2019).DOI
  3. [3]J. Lu and L. Ying, Compression of the electron repulsion integral tensor in tensor hypercontraction format with cubic scaling cost, J. Comput. Phys. 302, 329 (2015).arxiv.org/abs/1410.7757
  4. [4]E. G. Hohenstein, R. M. Parrish, and T. J. Martínez, Tensor hypercontraction density fitting. i. quartic scaling second- and third-order Møller-Plesset perturbation theory, J. Chem. Phys. 137, 044103 (2012).pubmed.ncbi.nlm.nih.gov/22852593
  5. [5]A. E. DePrince III and C. D. Sherrill, Accuracy and efficiency of coupled-cluster theory using density fitting/Cholesky decomposition, frozen natural orbitals, and at 1-transformed Hamiltonian, J. Chem. Theory Comput. 9, 2687 (2013).DOI
  6. [6]L. Riccardi, P. H. Nguyen, and G. Stock, Free-Energy Landscape of RNA Hairpins Constructed via Dihedral Angle Principal Component Analysis, J. Phys. Chem. B 113, 16660 (2009).DOI
  7. [7]Y. Yuan, Q. Zhu, R. Song, J. Ma, and H. Dong, A two-end data-driven accelerated sampling method for exploring the transition pathways between two known states of protein, J. Chem. Theory Comput. 16, 4631 (2020).DOI
  8. [8]C. Husemann and M. Salmhofer, Efficient parametrization of the vertex function, Ω scheme, and the t, t′ Hub- bard model at van Hove filling, Phys. Rev. B 79, 195125 (2009).
  9. [9]W.-S. Wang, Y.-Y. Xiang, Q.-H. Wang, F. Wang, F. Yang, and D.-H. Lee, Functional renormalization group and variational Monte Carlo studies of the electronic instabilities in graphene near 1/4 doping, Phys. Rev. B 85, 035414 (2012).DOI
  10. [10]W.-S. Wang, Z.-Z. Li, Y.-Y. Xiang, and Q.-H. Wang, Competing electronic orders on kagome lattices at van Hove filling, Phys. Rev. B 87, 115135 (2013).DOI
  11. [11]C. J. Eckhardt, C. Honerkamp, K. Held, and A. Kauch, Truncated unity parquet solver, Phys. Rev. B 101, 155104 (2020).DOI
  12. [12]M. Wallerberger, H. Shinaoka, and A. Kauch, Solving the Bethe-Salpeter equation with exponential convergence, Phys. Rev. Research 3, 033168 (2021).
  13. [13]J. Bardeen and W. Shockley, Deformation potentials and mobilities in non-polar crystals, Phys. Rev. 80, 72 (1950).DOI
  14. [14]C. Herring and E. Vogt, Transport and deformation-potential theory for many-valley semiconductors with anisotropic scattering, Phys. Rev. 101, 944 (1956).
  15. [15]H. Fröhlich, Electrons in lattice fields, Adv. Phys. 3, 325 (1954).DOI
  16. [16]G. D. Mahan, Many-Particle Physics, 3rd ed. (Springer, 2000).
  17. [17]A. M. Ganose, J. Park, A. Faghaninia, R. Woods-Robinson, K. A. Persson, and A. Jain, Efficient calculation of carrier scattering rates from first principles, Nat. Commun. 12, 2222 (2021).DOI
  18. [18]R. M. Martin, Electronic Structure: Basic Theory and Practical Methods (Cambridge University Press, 2004).
  19. [19]S. Baroni, S. de Gironcoli, A. Dal Corso, and P. Giannozzi, Phonons and related crystal properties from density-functional perturbation theory, Rev. Mod. Phys. 73, 515 (2001).DOI
  20. [20]M. Bernardi, D. Vigil-Fowler, J. Lischner, J. B. Neaton, and S. G. Louie, Ab initio study of hot carriers in the first picosecond after sunlight absorption in silicon, Phys. Rev. Lett. 112, 257402 (2014).pubmed.ncbi.nlm.nih.gov/25014830
  21. [21]W. Li, Electrical transport limited by electron-phonon coupling from Boltzmann transport equation: An ab initio study of Si, Al, and MoS2, Phys. Rev. B 92, 075405 (2015).
  22. [22]J.-J. Zhou and M. Bernardi, Ab initio electron mobility and polar phonon scattering in GaAs, Phys. Rev. B 94, 201201 (2016).DOI
  23. [23]T.-H. Liu, J. Zhou, B. Liao, D. J. Singh, and G. Chen, First-principles mode-by-mode analysis for electron-phonon scattering channels and mean free path spectra in GaAs, Phys. Rev. B 95, 075206 (2017).DOI
  24. [24]V. A. Jhalani, J.-J. Zhou, and M. Bernardi, Ultrafast hot carrier dynamics in GaN and its impact on the efficiency droop, Nano Lett. 17, 5012 (2017).DOI
  25. [25]J. Ma, A. S. Nissimagoudar, and W. Li, First-principles study of electron and hole mobilities of Si and GaAs, Phys. Rev. B 97, 045201 (2018).DOI
  26. [26]J.-J. Zhou, O. Hellman, and M. Bernardi, Electron-phonon scattering in the presence of soft modes and electron mobility in SrTiO3 perovskite from first principles, Phys. Rev. Lett. 121, 226603 (2018).DOI
  27. [27]T. Sohier, D. Campi, N. Marzari, and M. Gibertini, Mobility of two-dimensional materials from first principles in an accurate and automated framework, Phys. Rev. Mater. 2, 114010 (2018).DOI
  28. [28]J.-J. Zhou and M. Bernardi, Predicting charge transport in the presence of polarons: The beyond-quasiparticle regime in SrTiO3, Phys. Rev. Res. 1, 033138 (2019).DOI
  29. [29]S. Poncé, W. Li, S. Reichardt, and F. Giustino, First-principles calculations of large carrier mobility and conductivity in bulk semiconductors and two-dimensional materials, Rep. Prog. Phys. 83, 036501 (2020).DOI
  30. [30]N.-E. Lee, J.-J. Zhou, H.-Y. Chen, and M. Bernardi, Ab initio electron-two-phonon scattering in GaAs from next-to-leading order perturbation theory, Nat. Commun. 11, 1607 (2020).DOI
  31. [31]D. C. Desai, B. Zviazhynski, J.-J. Zhou, and M. Bernardi, Magnetotransport in semiconductors and two-dimensional materials from first principles, Phys. Rev. B 103, L161103 (2021).DOI
  32. [32]B. K. Chang, J.-J. Zhou, N.-E. Lee, and M. Bernardi, Intermediate polaronic charge transport in organic crystals from a many-body first-principles approach, NPJ Comput. Mater. 8, 63 (2022).DOI
  33. [33]Z. Zheng, Y. Shi, J.-J. Zhou, O. V. Prezhdo, Q. Zheng, and J. Zhao, Ab initio real-time quantum dynamics of charge carriers in momentum space, Nat. Comput. Sci 3, 532 (2023).DOI
  34. [34]Y. Luo, B. K. Chang, and M. Bernardi, Comparison of the canonical transformation and energy functional formalisms for ab initio calculations of self-localized polarons, Phys. Rev. B 105, 155132 (2022).DOI
  35. [35]J. Lafuente-Bartolome, C. Lian, W. H. Sio, I. G. Gurtubay, A. Eiguren, and F. Giustino, Unified approach to polarons and phonon-induced band structure renormalization, Phys. Rev. Lett. 129, 076402 (2022).
  36. [36]J.-J. Zhou, J. Park, I. Timrov, A. Floris, M. Cococcioni, N. Marzari, and M. Bernardi, Ab initio electron-phonon interactions in correlated electron systems, Phys. Rev. Lett. 127, 126404 (2021).
  37. [37]Z. Li, G. Antonius, M. Wu, F. H. da Jornada, and S. G. Louie, Electron-phonon coupling from ab initio linear-response theory within the GW method: Correlation-enhanced interactions and superconductivity in Ba1−xKxBiO3, Phys. Rev. Lett. 122, 186402 (2019).DOI
  38. [38]D. Pfau, J. S. Spencer, A. G. D. G. Matthews, and W. M. C. Foulkes, Ab initio solution of the many-electron Schrödinger equation with deep neural networks, Phys. Rev. Res. 2, 033429 (2020).
  39. [39]J. Hermann, Z. Schätzle, and F. Noé, Deep-neural-network solution of the electronic Schrödinger equation, Nat. Chem. 12, 891 (2020).
  40. [40]U. Schollwöck, The density-matrix renormalization group, Rev. Mod. Phys. 77, 259 (2005).DOI
  41. [41]G. K.-L. Chan and S. Sharma, The density matrix renormalization group in quantum chemistry, Annu. Rev. Phys. Chem. 62, 465 (2011).DOI
  42. [42]R. Orús, A practical introduction to tensor networks: Matrix product states and projected entangled pair states, Ann. Phys. 349, 117 (2014).DOI
  43. [43]F. Verstraete, T. Nishino, U. Schollwöck, M. C. Bańuls, G. K. Chan, and M. E. Stoudenmire, Density matrix renormalization group, 30 years on, Nat. Rev. Phys. 5, 273 (2023).DOI
  44. [44]A. S. Chaves, A. Antonelli, D. T. Larson, and E. Kaxiras, Boosting the efficiency of ab initio electron-phonon coupling calculations through dual interpolation, Phys. Rev. B 102, 125116 (2020).DOI
  45. [45]T. Deng, G. Wu, M. B. Sullivan, Z. M. Wong, J.-J. Zhou, J. Park, I.-T. Lu, I. Maliyev, X. Tong, and M. Bernardi, PERTURBO: A software package for ab initio electron–phonon interactions, charge transport and ultrafast dynamics, Comput. Phys. Commun. 264, 107970 (2021).
  46. [47]P. Vogl, Microscopic theory of electron-phonon interaction in insulators or semiconductors, Phys. Rev. B 13, 694 (1976).DOI
  47. [48]J. Sjakste, N. Vast, M. Calandra, and F. Mauri, Wannier interpolation of the electron-phonon matrix elements in polar semiconductors: Polar-optical coupling in GaAs, Phys. Rev. B 92, 054307 (2015).DOI
  48. [49]C. Verdi and F. Giustino, Fröhlich electron-phonon vertex from first principles, Phys. Rev. Lett. 115, 176401 (2015).DOI
  49. [50]V. A. Jhalani, J.-J. Zhou, J. Park, C. E. Dreyer, and M. Bernardi, Piezoelectric electron-phonon interaction from ab initio dynamical quadrupoles: Impact on charge transport in wurtzite GaN, Phys. Rev. Lett. 125, 136602 (2020).DOI
  50. [51]J. Park, J.-J. Zhou, V. A. Jhalani, C. E. Dreyer, and M. Bernardi, Long-range quadrupole electron-phonon interaction from first principles, Phys. Rev. B 102, 125203 (2020).DOI
  51. [52]G. Brunin, H. P. C. Miranda, M. Giantomassi, M. Royo, M. Stengel, M. J. Verstraete, X. Gonze, G.-M. Rignanese, and G. Hautier, Electron-phonon beyond Fröhlich: Dynamical quadrupoles in polar and covalent solids, Phys. Rev. Lett. 125, 136601 (2020).DOI
  52. [53]G. Brunin, H. P. C. Miranda, M. Giantomassi, M. Royo, M. Stengel, M. J. Verstraete, X. Gonze, G.-M. Rignanese, and G. Hautier, Phonon-limited electron mobility in Si, GaAs, and GaP with exact treatment of dynamical quadrupoles, Phys. Rev. B 102, 094308 (2020).DOI
  53. [54]E. Prodan and W. Kohn, Nearsightedness of electronic matter, Proc. Natl. Acad. Sci. U.S.A. 102, 11635 (2005).pubmed.ncbi.nlm.nih.gov/16087868
  54. [55]L. A. Agapito and M. Bernardi, Ab initio electron-phonon interactions using atomic orbital wave functions, Phys. Rev. B 97, 235146 (2018).DOI
  55. [56]F. Giustino, M. L. Cohen, and S. G. Louie, Electron-phonon interaction using wannier functions, Phys. Rev. B 76, 165108 (2007).
  56. [57]S. L. Brunton and J. N. Kutz, Data-driven science and engineering: Machine learning, dynamical systems, and control (Cambridge University Press, Cambridge, 2022).
  57. [58]J. Sjakste, N. Vast, M. Calandra, and F. Mauri, Wannier interpolation of the electron-phonon matrix elements in polar semiconductors: Polar-optical coupling in GaAs, Phys. Rev. B 92, 054307 (2015).DOI
  58. [59]D. J. Lépine, Spin resonance of localized and delocalized electrons in phosphorus-doped silicon between 20 and 30 K, Phys. Rev. B 2, 2429 (1970).DOI
  59. [60]G. Lancaster, J. A. van Wyk, and E. E. Schneider, Spin-lattice relaxation of conduction electrons in silicon, Proc. Phys. Soc. 84, 19 (1964).DOI
  60. [61]C. Herring and E. Vogt, Transport and deformation-potential theory for many-valley semiconductors with anisotropic scattering, Phys. Rev. 101, 944 (1956).
  61. [62]J. Park, J.-J. Zhou, and M. Bernardi, Spin-phonon relaxation times in centrosymmetric materials from first principlesDOI
  62. [63]P. Mehta, M. Bukov, C.-H. Wang, A. G. Day, C. Richardson, C. K. Fisher, and D. J. Schwab, A high-bias, low-variance introduction to machine learning for physicists, Phys. Rep. 810, 1 (2019).DOI
  63. [64]G. Grimvall, The electron-phonon interaction in metals (North-Holland, New York, 1981).
  64. [65]H. Shang and J. Yang, The electron–phonon renormalization in the electronic structure calculation: Fundamentals, current status, and challenges, J. Chem. Phys. 158, 130901 (2023).DOI
  65. [66]See Supplemental Material for specific timing comparisons, c-SVD mobility in polar materials, convergence tests of physical quantities for all the materials studied here, principal component analysis for lead and the validation of our compression method for large systems .
  66. [67]Y. Ge and A. Y. Liu, Phonon-mediated superconductivity in electron-doped single-layer moS₂: A first-principles prediction, Phys. Rev. B 87, 241408 (2013).DOI
  67. [68]N.-E. Lee, H.-Y. Chen, J.-J. Zhou, and M. Bernardi, First principles approach for self-localized polarons from canonical transformations, Phys. Rev. Mater. 5, 063805 (2021).
  68. [69]Y. Luo, B. K. Chang, and M. Bernardi, Comparison of the canonical transformation and energy functional formalisms for ab initio calculations of self-localized polarons, Phys. Rev. B 105, 155132 (2022).DOI
  69. [70]W. H. Sio, C. Verdi, S. Poncé, and F. Giustino, Polarons from first principles, without supercells, Phys. Rev. Lett. 122, 246403 (2019).DOI
  70. [71]M. Setvin, C. Franchini, X. Hao, M. Schmid, A. Janotti, M. Kaltak, C. G. Van de Walle, G. Kresse, and U. Diebold, Direct view at excess electrons in TiO₂ rutile and anatase, Phys. Rev. Lett. 113, 086402 (2014).pubmed.ncbi.nlm.nih.gov/25192111
  71. [72]A. Janotti, J. B. Varley, M. Choi, and C. G. Van de Walle, Vacancies and small polarons in SrTiO₃, Phys. Rev. B 90, 085202 (2014).DOI
  72. [73]M. Royo and M. Stengel, First-principles theory of spatial dispersion: Dynamical quadrupoles and flexoelectricity, Phys. Rev. X 9, 021050 (2019).
  73. [74]X. Gonze, B. Amadon, G. Antonius, F. Arnardi, L. Baguet, J.-M. Beuken, J. Bieder, F. Bottin, J. Bouchet, E. Bousquet, N. Brouwer, F. Bruneval, G. Brunin, T. Cavignac, J.-B. Charraud, W. Chen, M. Côté, S. Cottier, J. Denier, G. Geneste, P. Ghosez, M. Giantomassi, Y. Gillet, O. Gingras, D. R. Hamann, G. Hautier, X. He, N. Helbig, N. Holzwarth, Y. Jia, F. Jollet, W. Lafargue-Dit-Hauret, K. Lejaeghere, M. A. Marques, A. Martin, C. Martins, H. P. Miranda, F. Naccarato, K. Persson, G. Petretto, V. Planes, Y. Pouillon, S. Prokhorenko, F. Ricci, G.-M. Rignanese, A. H. Romero, M. M. Schmitt, M. Torrent, M. J. van Setten, B. Van Troeye, M. J. Verstraete, G. Zérah, and J. W. Zwanziger, The abinitproject: Impact, environment and recent developments, Comput. Phys. Commun. 248, 107042 (2020).DOI
  74. [75]M. Bernardi, First-principles dynamics of electrons and phonons, Eur. Phys. J. B 89, 239 (2016).DOI
  75. [76]M. T. C. Huh, R. E. Funde, and R. J. Plemmons, Structured low rank approximation, Linear Algebra Its Appl. 366, 157 (2003).
  76. [77]R. J. Elliott, Theory of the effect of spin-orbit coupling on magnetic resonance in some semiconductors, Phys. Rev. 96, 266 (1954).DOI
  77. [78]Y. Yafet, Solid State Physics (Academic Press, New York, 1963).
  78. [79]J. P. Perdew, K. Burke, and M. Ernzerhof, Generalized gradient approximation made simple, Phys. Rev. Lett. 77, 3865 (1996).pubmed.ncbi.nlm.nih.gov/10062328
  79. [80]P. Giannozzi, S. Baroni, N. Bonini, M. Calandra, R. Car, C. Cavazzoni, D. Ceresoli, G. L. Chiarotti, M. Cococioni, I. Dabo, A. D. Corso, S. de Gironcoli, S. Fabris, G. Fratesi, R. Gebauer, U. Gerstmann, C. Gougoussis, A. Kokalj, M. Lazzeri, L. Martin-Samos, N. Marzari, F. Mauri, R. Mazzarello, S. Paolini, A. Pasquarello, L. Paulatto, C. Sbraccia, S. Scandolo, G. Schlauser, A. P. Seitonen, A. Smogunov, P. Umari, and R. M. Wentzcovitch, QUANTUM ESPRESSO: a modular and open-source software project for quantum simulations of materials, J. Phys. Condens. Matter 21, 395502 (2009).DOI
  80. [81]A. A. Mostofi, J. R. Yates, G. Pizzi, Y.-S. Lee, I. Souza, D. Vanderbilt, and N. Marzari, An updated version of wannier90: A tool for obtaining maximally-localised Wannier functions, Comput. Phys. Commun. 185, 2309 (2014).
  81. [83]P. Virtanen, R. Gommers, T. E. Oliphant, M. Haberland, T. Reddy, D. Cournapeau, E. Burovski, P. Peterson, W. Weckesser, J. Bright, S. J. van der Walt, M. Brett, J. Wilson, K. Millman, N. Mayorov, A. R. J. Nelson, E. Jones, R. Kern, E. Larson, C. J. Carey, I. Polat, Y. Feng, E. W. Moore, J. VanderPlas, D. Laxalde, J. Perktold, R. C. Rimmel, I. Henrikesen, E. A. Quintero, C. R. Harris, A. M. Archibald, A. H. Ribeiro, F. Pedregosa, P. van Mulbregt, and SciPy 1.0 Contributors, SciPy 1.0: Fundamental Algorithms for Scientific Computing in Python, Nat. Methods 17, 261 (2020).
  82. [84]M. van Setten, M. Giantomassi, E. Bousquet, M. Verstraete, D. Hamann, X. Gonze, and G.-M. Rignanese, The PseudoDojo: training and grading a 85 element optimized norm-conserving pseudopotential table, Comput. Phys. Commun. 226, 39 (2018).DOI
  83. [85]C.-H. Park, F. Giustino, M. L. Cohen, and S. G. Louie, Velocity renormalization and carrier lifetime in graphene from the electron–phonon interaction, Phys. Rev. Lett. 99, 086804 (2007).pubmed.ncbi.nlm.nih.gov/17930972
  84. [86]J.-M. Lihm and C.-H. Park, Wannier function perturbation theory: Localized representation and interpolation of wave function perturbation, Phys. Rev. X 11, 041053 (2021).DOI

Paper details

vlnc:W_h7dqndmt
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents