Conserving relaxation-time approximation for electron-electron collisions
Sayak Bhattacharjee, Tatia Kiliptari, Dmitrii L. Maslov · 2026
Read Conserving relaxation-time approximation for electron-electron collisions by Sayak Bhattacharjee, Tatia Kiliptari, Dmitrii L. Maslov (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-26207.txt.
Introduction
Sayak Bhattacharjee, Tatia Kiliptari, and Dmitrii Maslov develop a conserving relaxation-time approximation (2026). It projects electron-electron collisions onto the full space of conserved quantities while keeping energy resolution. Unlike cruder versions, it follows states close to, though not exactly at, the Fermi surface, the boundary between filled and empty electron states. Applied to charge transport in Fermi liquids, it reproduces exact limits and scalings, matching exact answers at the percent level for several observables.
Catalog record
| CREATOR | Sayak Bhattacharjee, Tatia Kiliptari, Dmitrii L. Maslov |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 51622 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.26207
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Conserving relaxation-time approximation for electron-electron collisions Sayak Bhattacharjee, Tatia Kiliptari, Dmitrii L. Maslov Abstract We develop a conserving relaxation-time approximation (cRTA) based on an explicit energy-resolved projection onto the full space of collision invariants. Our cRTA retains the energy dependence of the nonequilibrium quasiparticle distribution, allowing one to describe transport quantities sensitive to states near, but not exactly on, the Fermi surface (FS). We apply the method to several charge-transport problems in both Galilean-invariant and non-Galilean-invariant Fermi liquids. In particular, the cRTA reproduces the low- and high-temperature limits of the dc conductivity of a non-Galilean-invariant Fermi liquid with disorder, the hydrodynamic and collisionless limits of the finite-wavevector longitudinal conductivity of a clean Galilean-invariant Fermi liquid, and the asymptotic scaling forms of the optical conductivity of a clean non-Galilean-invariant Fermi liquid beyond the semiclassical limit. For several observables, the agreement with exact solutions is quantitative at the percent level. These results demonstrate that the cRTA provides a simple and accurate framework for describing transport beyond the FS projection. Introduction. The kinetic equation (KE) is a versatile tool for describing transport phenomena in both classical and quantum systems. However, its exact solutions are rare, even in linear response. The main difficulty arises from the non-locality of the collision integral in momentum space. Consequently, the KE is an integro-differential one. Accordingly, a number of approximate methods for solving the KE were developed in the past. A special category of these methods are relaxation-time approximations (RTAs), which replace the non-local collision integral for binary collisions by a local one. The most common RTA with the collision integral I_{\rm RTA}=-(f_{\mathbf{k}}-n_{\mathbf{k}})/\tau 5 , where n_{\mathbf{k}} is the equilibrium distribution, and \tau is the phenomenological relaxation time, does not enforce conservation laws. Conserving RTAs date back to the seminal paper by Bhatnagar, Gross, and Krook on mono-atomic classical gases 7 . In condensed matter physics, partially and fully conserving RTAs for binary collision integrals were developed in Refs. 9 ; 23 ; 28 ; 12 ; 13 ; 11 ; 10 ; 6 . Earlier conserving RTAs were formulated in terms of local thermodynamic variables and applied mainly to density response and hydrodynamics of electron gases with a parabolic dispersion 6 . In this Letter, we formulate the conserving approximation as an explicit projection onto the hydrodynamic manifold, yielding a systematic scheme for describing transport in generic non-Galilean-invariant Fermi liquids. Our cRTA also differs from the previous methods in that it does not rely on a Fermi-surface projection but operates with an energy-resolved projection onto the full space of collision invariants. This makes it applicable to transport quantities sensitive to the energy dependence of the nonequilibrium distribution near the Fermi surface (FS), e.g., the thermal conductivity of a generic Fermi liquid (FL) and the charge conductivity of non-Galilean–invariant FLs. We show that our cRTA reproduces several exact and asymptotically exact results. General construction. We consider a single-band system of electrons with arbitrary dispersion \varepsilon=\varepsilon(\mathbf{k}) , measured from the Fermi energy. The interaction between electrons is parameterized by a phenomenological scattering time \tau_{\mathrm{ee}}(\mathbf{k},T) which, in general, depends on the momentum \mathbf{k} and temperature T . Beyond the semiclassical limit, i.e., for arbitrary \Omega/T , where \Omega is the frequency of the applied electric field, one also needs to include the dependence of \tau_{\mathrm{ee}} on \Omega . Electron-electron collisions conserve particle number, momentum (in the absence of umklapp scattering), and energy, such that a vector in the d+2 “hydrodynamic” space, \displaystyle\vec{\psi}_{\mathbf{k}}=\left(1,\varepsilon,\mathbf{k}\right), (1) is conserved. Here and thereafter, \vec{a} denotes a vector in this d+2 space, whereas \mathbf{a} denotes a vector in real space. Our goal is to construct an algebraic form of the collision integral which satisfies all the conservation laws. A deviation from equilibrium is parameterized as \displaystyle\delta f_{\mathbf{k}} \displaystyle= \displaystyle f_{\mathbf{k}}-n_{\mathbf{k}}=(-n_{\mathbf{k}}^{\prime})\chi_{\mathbf{k}}, (2) where n^{\prime}_{\mathbf{k}}\equiv\partial_{\varepsilon}n(\varepsilon) . Introduce another deviation projected onto the hydrodynamic space: \displaystyle\delta f^{\mathrm{h}}_{\mathbf{k}}=(-n_{\mathbf{k}}^{\prime})\chi^{\mathrm{h}}_{\mathbf{k}},\quad\chi^{\mathrm{h}}_{\mathbf{k}}=\vec{c}\cdot\vec{\psi}_{\mathbf{k}}, (3) where \vec{c} is a \mathbf{k} -independent d+2 vector. Consider the collision integral: \displaystyle I_{\mathrm{ee}}[f_{\mathbf{k}}]=-\frac{\delta f_{\mathbf{k}}-\delta f^{\mathrm{h}}_{\mathbf{k}}}{\tau_{\mathrm{ee}}}=-(-n_{\mathbf{k}}^{\prime})\frac{\chi_{\mathbf{k}}-\chi_{\mathbf{k}}^{\mathrm{h}}}{\tau_{\mathrm{ee}}}. (4) If the vector \vec{c} in Eq. ( 3 ) is chosen as \displaystyle\vec{c}=\hat{M}^{-1}\cdot\left\langle\frac{\chi_{\mathbf{k}}\vec{\psi}_{\mathbf{k}}}{\tau_{\mathrm{ee}}}\right\rangle, (5) with \displaystyle\langle X_{\mathbf{k}}\rangle\equiv\int\frac{d^{d}k}{(2\pi)^{d}}(-n^{\prime}_{\mathbf{k}})X_{\mathbf{k}} (6) and \displaystyle M_{\alpha\beta} \displaystyle= \displaystyle M_{\beta\alpha}=\left\langle\frac{\psi_{\alpha\mathbf{k}}\psi_{\beta\mathbf{k}}}{\tau_{\mathrm{ee}}}\right\rangle, (7) then I_{\mathrm{ee}}[f_{\mathbf{k}}] in Eq. ( 4 ) satisfies all the necessary conservation laws, i.e., \displaystyle\int\frac{d^{d}k}{(2\pi)^{d}}\,\psi_{\alpha\mathbf{k}}\,I_{\mathrm{ee}}[f_{\mathbf{k}}]=0,\quad\alpha=1\dots d+2. (8) The above statement is proven as follows. Since \int\frac{d^{d}k}{(2\pi)^{d}}\psi_{\alpha\mathbf{k}}I_{\mathrm{ee}}[f_{\mathbf{k}}]=\left\langle\psi_{\alpha\mathbf{k}}(\chi_{\mathbf{k}}^{\mathrm{h}}-\chi_{\mathbf{k}})/\tau_{\mathrm{ee}}\right\rangle , Eq. ( 8 ) is equivalent to \displaystyle\left\langle\frac{\psi_{\alpha\mathbf{k}}\,\chi_{\mathbf{k}}^{\mathrm{h}}}{\tau_{\mathrm{ee}}}\right\rangle=\left\langle\frac{\psi_{\alpha\mathbf{k}}\,\chi_{\mathbf{k}}}{\tau_{\mathrm{ee}}}\right\rangle. (9) Substituting Eqs. ( 5 ) and ( 7 ) into Eq. ( 9 ) yields \displaystyle\left\langle\frac{\psi_{\alpha\mathbf{k}}\,\chi_{\mathbf{k}}^{\mathrm{h}}}{\tau_{\mathrm{ee}}}\right\rangle=\left\langle\frac{\psi_{\alpha\mathbf{k}}}{\tau_{\mathrm{ee}}}\sum_{\beta}c_{\beta}\psi_{\beta\mathbf{k}}\right\rangle=\sum_{\beta}c_{\beta}M_{\alpha\beta}= (10) \displaystyle\sum_{\beta,\gamma}(M^{-1})_{\beta\gamma}\left\langle\frac{\chi_{\mathbf{k}}\psi_{\gamma\mathbf{k}}}{\tau_{\mathrm{ee}}}\right\rangle M_{\alpha\beta}=\left\langle\frac{\chi_{\mathbf{k}}\psi_{\alpha\mathbf{k}}}{\tau_{\mathrm{ee}}}\right\rangle, as required. In what follows, we demonstrate that this method reproduces some known results. dc conductivity of a non-Galilean–invariant Fermi liquid with disorder. By “non-Galilean invariant” we mean a system with a non-parabolic \varepsilon that may otherwise be an arbitrary function of k\equiv|\mathbf{k}| , so that \tau_{\mathrm{ee}} depends only on \varepsilon and T . Previous work has shown that the electron-electron (ee) contribution to the conductivity depends on the ratio of the impurity scattering time, \tau_{\mathrm{i}} , and the quasiparticle scattering time due to the ee interaction, \tau_{\mathrm{ee}}(T)\propto T^{-2} , with a \ln T factor in two dimensions (2D) 17 . At low temperatures ( \tau_{\mathrm{ee}}\gg\tau_{\mathrm{i}} ), the conductivity exhibits a correction of order T^{4} (again, with a \ln T factor in 2D) 24 ; 26 ; 19 ; 16 , and saturates at high T ( \tau_{\mathrm{ee}}(T)\ll\tau_{\mathrm{i}} ) 24 ; 16 . To mimic Coulomb interaction in 2D, we choose (in units where \hbar=k_{B}=1 ) \displaystyle\frac{1}{\tau_{\mathrm{ee}}}=a\frac{(b\varepsilon)^{2}+\pi^{2}T^{2}}{\varepsilon_{\rm F}}\ln\frac{\Omega_{\rm p}}{T},\qquad T,\varepsilon\ll\Omega_{\rm p}, (11) where \varepsilon_{\textrm{F}} is the Fermi energy, \Omega_{\rm p}=v_{\mathrm{F}}\kappa is the characteristic plasma frequency, v_{\textrm{F}} is the Fermi velocity, \kappa is the screening wavevector, and a and b are numerical coefficients. Although b=1 for the quasiparticle rate, we treat both a and b as fitting parameters. We assume isotropic impurity scattering with an energy-independent \tau_{\mathrm{i}} , and consider a 2D system with Dirac dispersion, as a particular example of a non-Galilean FL. With \mathbf{E} along the x -axis, the linearized semiclassical KE reads \displaystyle ev_{x}E(-n_{\mathbf{k}}^{\prime})=I_{\mathrm{ee}}[f_{\mathbf{k}}]-\frac{f_{\mathbf{k}}-\bar{f}_{\mathbf{k}}}{\tau_{\mathrm{i}}}, (12) where \bm{v}=\partial_{\mathbf{k}}\varepsilon and \bar{O}_{\mathbf{k}} is the average of O_{\mathbf{k}} over the directions of \mathbf{k} . The impurity part of the collision integral is exact for point-like disorder. Since the non-equilibrium distribution is proportional to v_{x}E , \langle\chi_{\mathbf{k}}/\tau_{\mathrm{ee}}\rangle=\langle\chi_{\mathbf{k}}\varepsilon/\tau_{\mathrm{ee}}\rangle=\langle\chi_{k}k_{y}/\tau_{\mathrm{ee}}\rangle=0 . From Eq. ( 5 ), c_{1}=c_{\varepsilon}=c_{y}=0 , and also \bar{\chi}_{\mathbf{k}}=0 . Substituting Eq. ( 4 ) with c_{1}=c_{\varepsilon}=0 into Eq. ( 12 ) yields \displaystyle\chi_{\mathbf{k}}=\frac{\tau_{\mathrm{tot}}}{\tau_{\mathrm{ee}}}c_{x}k_{x}-\tau_{\mathrm{tot}}ev_{x}E, (13) where \tau_{\mathrm{tot}}=\tau_{\mathrm{ee}}\tau_{\mathrm{i}}/(\tau_{\mathrm{ee}}+\tau_{\mathrm{i}}) , whereas Eq. ( 5 ) gives, \displaystyle c_{x}=\left(\hat{M}^{-1}\right)_{xx}\left\langle\frac{\chi_{\mathbf{k}}k_{x}}{\tau_{\mathrm{ee}}}\right\rangle. (14) For an isotropic system, the matrix \hat{M} is block-diagonal in the scalar (1,\varepsilon) and vector ( \mathbf{k} ) subspaces, with \left(\hat{M}^{-1}\right)_{xx}=1/\langle k_{x}^{2}/\tau_{\mathrm{ee}}\rangle . Equations ( 13 ) and ( 14 ) then yield \displaystyle c_{x}=-\frac{eE\left\langle\frac{\tau_{\mathrm{tot}}}{\tau_{\mathrm{ee}}}v_{x}k_{x}\right\rangle}{\left\langle\frac{1}{\tau_{\mathrm{ee}}+\tau_{\mathrm{i}}}k_{x}^{2}\right\rangle}. (15) The conductivity is thus \displaystyle\sigma=2e^{2}\tau_{\mathrm{i}}\left(\left\langle\frac{\tau_{\mathrm{ee}}}{\tau_{\mathrm{ee}}+\tau_{\mathrm{i}}}v_{x}^{2}\right\rangle+\tau_{\mathrm{i}}\frac{\left\langle\frac{1}{\tau_{\mathrm{ee}}+\tau_{\mathrm{i}}}k_{x}v_{x}\right\rangle^{2}}{\left\langle\frac{1}{\tau_{\mathrm{ee}}+\tau_{\mathrm{i}}}k_{x}^{2}\right\rangle}\right). (16) For a parabolic dispersion with band mass m (k_{x}=mv_{x} ), \tau_{\mathrm{ee}} drops out from the result, and the conductivity is controlled entirely by impurities: \sigma=\sigma_{\mathrm{i}}=2e^{2}\tau_{\mathrm{i}}\left\langle v_{x}^{2}\right\rangle . The same holds after FS projection, i.e., for v=v_{\mathrm{F}} and k=k_{\rm F} (Fermi momentum). A finite ee contribution thus requires expanding k and v near the FS. For \tau_{\mathrm{ee}}\gg\tau_{\mathrm{i}} , expanding Eq. ( 16 ) in \tau_{\mathrm{i}}/\tau_{\mathrm{ee}} yields a correction \delta\sigma=\sigma-\sigma_{i0} to the residual conductivity, \sigma_{\mathrm{i}0}=\sigma_{\rm i}|_{T=0} , \displaystyle\delta\sigma=-2e^{2}\tau_{\mathrm{i}}^{2}\frac{\left\langle\frac{v_{x}^{2}}{\tau_{\mathrm{ee}}}\right\rangle\left\langle\frac{k_{x}^{2}}{\tau_{\mathrm{ee}}}\right\rangle-\left\langle\frac{v_{x}k_{x}}{\tau_{\mathrm{ee}}}\right\rangle^{2}}{\left\langle\frac{k_{x}^{2}}{\tau_{\mathrm{ee}}}\right\rangle}. (17) For Dirac spectrum, v_{F}={\rm constant} , k=k_{\rm F}(1+\varepsilon/v_{\mathrm{F}}) , and the density of states \nu(\varepsilon)=\nu_{\mathrm{F}}(1+\varepsilon/\varepsilon_{\rm F}) , where \nu_{\mathrm{F}}=\varepsilon_{\rm F}/2\pi v_{\mathrm{F}}^{2} .