Introduction
Electron–phonon energy relaxation determines the electronic temperature in many mesoscopic devices based on two-dimensional materials [1 , 2 , 3 ] . It controls how quickly hot electrons return to the lattice temperature and therefore affects the physical mechanisms such as hot-electron transport, bolometric response, calorimetric detection, and the recovery time of thermal detectors. At low temperature, micron-scale graphene devices can have a very small electronic heat capacity, so weak absorbed power can appreciably raise their electronic temperature [4 , 5 , 6 , 7 , 8 , 9 , 10 , 11 ] .
The coupling of graphene electrons to acoustic phonons has been studied extensively both in quasi-particle transport and cooling contexts [12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 ] . To be specific, for long-wavelength longitudinal acoustic phonons, the dominant interaction is commonly described by a deformation potential. This coupling results in two important temperature regimes. At low temperature, only phonons with momenta well below twice the Fermi wave vector are thermally active, and it is known as Bloch–Grüneisen (BG) regime [12 , 16 ] . In the opposite limit of high temperatures, the active acoustic phonons have large classical occupations, and the scattering becomes approximately elastic. This is known as the equipartition (EP) regime [12 , 16 , 19 , 21 ] . The average electron–phonon cooling power has been studied for monolayer graphene (MLG) and Bernal bilayer graphene (BLG) in both regimes [16 , 22 , 23 , 24 ] . In particular, it is worth pointing out that in Ref. [16 ] authors considered a quasiequilibrium state in which the electrons and phonons have separate temperatures. They obtained the limiting analytical cooling laws and evaluated the numerical crossover between them [16 ] . Related experimental and theoretical studies have examined cooling in doped graphene, hot-electron transport, and disorder-assisted energy relaxation [4 , 5 , 6 , 13 , 14 , 15 , 25 , 26 , 27 , 28 , 29 , 30 , 31 , 32 , 33 ] .
The average cooling power, however, gives only the mean rate at which energy is transferred. It is the first moment of the energy-transfer process. A complete description should also determine the fluctuations around this mean value. Microscopically, the energy exchange consists of separate phonon-emission and phonon-absorption events. These events remain random even when the average heat current is stationary. The corresponding heat-current noise is the second moment of the cooling process and is the focus of our work. Heat-current fluctuations have previously been considered in metallic electron–phonon systems and in mesoscopic thermal links [34 , 35 , 36 , 37 ] . These studies show that finite-frequency heat noise contains information that is absent from the average thermal conductance. In particular, a simple frequency-dependent extension of the classical thermal fluctuation–dissipation theorem (FDT) does not generally contain all quantum contributions to heat-current noise [35 , 36 , 37 ] . As a next step, these studies motivate a direct microscopic calculation of the finite-frequency spectrum. As in previous works, we consider the clean MLG and BLG coupled to longitudinal acoustic phonons through an unscreened deformation potential. The electron subsystem is assumed to remain internally thermal at an electron temperature T e T_{\mathrm{e}} T e , while the acoustic phonons remain internally thermal at a phonon temperature T p h T_{\mathrm{ph}} T ph . This two-temperature description is the quasiequilibrium approximation used in the standard cooling-power theory [16 , 34 , 36 ] . The two temperatures are distinct, corresponding to a non-equilibrium state. This non-equilibrium condition results in energy exchange between electron and phonon subsystems. Namely, phonon emission transfers energy from electrons to phonons, whereas phonon absorption transfers energy in the opposite direction. The average heat current measures the imbalance between the two processes, while the heat-current noise measures their combined fluctuations (see Figure 1 ).
Figure 1. Schematic electron–phonon energy exchange in the quasiequilibrium two-temperature description. Phonon emission transfers energy from electrons to phonons, whereas phonon absorption transfers energy in the opposite direction. The average heat current measures the imbalance between the two processes, while the heat-current noise measures their combined fluctuations.
The remainder of the paper is organized as follows. In Sec. II we formulate the problem and discuss the main features of it. Sec. III introduces the microscopic model and heat-current operator. Sec. IV derives the leading in electron-phonon coupling current–current correlator and identifies the physical exchange contribution. Sec. V presents the zero-frequency MLG and BLG results in the BG and EP regimes. Sec. VI develops the finite-frequency spectra for equal and unequal electron and phonon temperatures. Sec. VII presents the numerical crossover calculations, finite-frequency comparisons, and experimental noise scales. Sec. VIII discusses the physical interpretation, applicability, and limitations of the results. Detailed derivations and numerical checks are collected in the appendices.
To study the fluctuations, we define the symmetrized heat-current noise spectrum, which is sum of two unsymmetrized terms
S ( ω ) = 1 2 [ Z ( ω ) + Z ( − ω ) ] . (1) S(\omega) = \frac{1}{2}\left[Z(\omega) + Z(-\omega)\right]. \tag*{(1)} S ( ω ) = 2 1 [ Z ( ω ) + Z ( − ω ) ] . (1)
In case of time invariance under the consideration, each unsymmetrized noise is given by the Fourier transformation of two-point correlation function of heat currents
Z ( ω ) = ∫ d t e i ω t ⟨ δ J ^ ( t ) δ J ^ ( 0 ) ⟩ , (2) Z(\omega) = \int dt\, e^{i\omega t} \langle\delta\hat{J}(t)\delta\hat{J}(0)\rangle, \tag*{(2)} Z ( ω ) = ∫ d t e iω t ⟨ δ J ^ ( t ) δ J ^ ( 0 )⟩ , (2)
where δ J ^ = J ^ − ⟨ J ^ ⟩ \delta\hat{J} = \hat{J} - \langle\hat{J}\rangle δ J ^ = J ^ − ⟨ J ^ ⟩ and J ^ ( t ) \hat{J}(t) J ^ ( t ) is the heat-current operator in the Heisenberg picture, which is defined later in Eqs. (10) and (14) . We choose the sign of J ^ \hat{J} J ^ so that a positive current corresponds to energy flowing from the electrons to the phonons due to the fact that T e ≥ T p h T_{\mathrm{e}} \ge T_{\mathrm{ph}} T e ≥ T ph . The quantity ⟨ J ^ ⟩ \langle\hat{J}\rangle ⟨ J ^ ⟩ is the average heat current, while δ J ^ ( t ) \delta\hat{J}(t) δ J ^ ( t ) can be considered as the instantaneous fluctuation around that average. The average is taken with respect to nonequilibrium density matrix which at t → − ∞ t \to-\infty t → − ∞ is given by the product of equilibrium electron and phonon subsystems ρ ^ e ⊗ ρ ^ p h \hat{\rho}_{\mathrm{e}} \otimes\hat{\rho}_{\mathrm{ph}} ρ ^ e ⊗ ρ ^ ph with corresponding temperatures T e T_{\mathrm{e}} T e and T p h T_{\mathrm{ph}} T ph . The frequency ω \omega ω sets the time scale probed by the measurement. Positive and negative frequencies are physically different in the unsymmetrized spectrum because they describe the ability of the system to supply energy to a detector or absorb energy from it. In contrast, the Eq. (1) treats positive and negative frequencies equally and gives a spectrum that is even in frequency. At zero frequency, the symmetrized and unsymmetrized spectra coincide. Therefore, our work investigates how the spectrum given in Eq. (1) changes when the detector resolves fluctuations at finite frequencies, whether the zero-frequency noise recovers the known equilibrium thermal conductance, and how these results evolve across BG and EP regimes, both at equilibrium and non-equilibrium situations, when the electron and phonon temperatures differ.
The physical distinction between the mean heat current and its noise can be understood from phonon emission
and absorption processes. In an emission event, an electron loses energy and creates a phonon. In an absorption event, an electron gains energy by removing a phonon from the phonon bath. The average heat current assigns opposite signs to these two events because they transfer energy in opposite directions. The noise assigns the same sign to both events because each event contributes positively to the variance of the transferred energy. The average current therefore measures the imbalance between the emission and absorption rates. The zero-frequency noise measures their combined fluctuations and weights each event by the square of the transferred phonon energy. At equilibrium, the emission and absorption rates are equal. Their difference vanishes, so the average current is zero, but their sum remains finite, resulting in non vanishing heat-current noise. This physical picture is summarized in Figure 1 . The electron and phonon subsystems continue to exchange energy even when their temperatures are equal. Equilibrium means that the two opposite energy currents cancel on average, not that the microscopic energy-transfer events stop. Moreover, even at zero temperature due to non-zero electron-phonon coupling the quantum fluctuations of heat current survive at finite frequency.
Our calculations starts from the electron, acoustic-phonon, and electron–phonon Hamiltonians. We construct a symmetric heat-current operator from the rates of change of the electron and phonon subsystem energies. This definition is closely related to the symmetric heat-current operator used in finite-frequency electron–phonon noise calculations in normal metals and low-dimensional systems [35 , 36 , 38 ] . We then evaluate the leading current–current correlation function in electron-phonon coupling by performing the complete Wick expansion. The fermionic part produces a direct contraction and an exchange contraction. The direct contraction forces the initial and final electron states to be identical. Momentum conservation then requires the acoustic phonon momentum to be zero. A zero-momentum acoustic phonon carries no energy, so this contribution vanishes. The exchange contraction describes a genuine electron–phonon scattering event and gives the physical noise spectrum. It contains both phonon emission and phonon absorption. Keeping both processes is essential at zero frequency. Removing either one would give an incomplete rate sum and would prevent the equilibrium noise from reproducing the thermal conductance correctly.
At equilibrium, the zero-frequency noise is related to the linear electron–phonon thermal conductance. More precisely the noise equals twice the Boltzmann constant multiplied by the square of the base temperature and by the conductance. The thermal conductance, G e p ( T ) G_{\mathrm{ep}}(T) G ep ( T ) , is defined as the response of the average heat current to a small temperature difference between electron and phonon subsystems, Δ T \Delta T Δ T , around base temperature, T T T , and the corresponding equilibrium zero-frequency noise is
S ( 0 ) = 2 k B T 2 G e p ( T ) . S(0) = 2 k_{\mathrm{B}} T^2 G_{\mathrm{ep}}(T). S ( 0 ) = 2 k B T 2 G ep ( T ) .
We use this relation only at zero frequency, where it provides a stringent check on the microscopic calculations [35 , 36 , 39 , 40 ] . The zero-frequency calculation reproduces the established cooling-power laws of monolayer and bilayer graphene [12 , 13 , 16 , 31 ] . In the clean BG regime, the MLG conductance is proportional to the absolute chemical potential and to the third power of temperature (∝ T 3 \propto T^3 ∝ T 3 ). Its equilibrium noise is therefore proportional to the absolute chemical potential and to the fifth power of temperature (∝ T 5 \propto T^5 ∝ T 5 ). The BLG conductance has the same third-power temperature dependence in the BG regime, and its equilibrium noise again has a fifth-power temperature dependence. Its doping dependence is different, however. The BLG coefficient is proportional to the inverse square root of the absolute chemical potential. In contrast to MLG, this difference follows from the parabolic low-energy dispersion of BLG and from the chemical-potential dependence of its Fermi velocity [16 , 21 , 22 , 23 , 24 ] . In EP regime, the active acoustic phonons are classically occupied. For MLG, the thermal conductance is controlled by a fourth-order electronic energy moment. This produces separate contributions from the chemical potential and thermal broadening. For BLG, the parabolic electronic dispersion leads to a temperature multiplied by a logarithmic phase-space factor. As a result, MLG and BLG show different high-temperature and near-neutrality behavior [12 , 16 , 21 , 24 ] . In equilibrium but at finite frequency, the detector introduces an additional energy scale, which survives even in case of zero temperature of whole system, which is the result of quantum fluctuations due to finite electron–phonon coupling. The heat-current vertices then contain the square of the detector energy minus twice the phonon energy and the square of the detector energy plus twice the phonon energy. These factors show that finite-frequency noise probes the time dependence of electron–phonon energy exchange rather than only its long-time variance. Previous works have shown that finite-frequency heat-current noise can contain quantum contributions that are not determined by the ordinary dynamic thermal conductance [35 , 36 , 37 ] . Direct evaluation is also necessary away from equilibrium because no single temperature characterizes both subsystems.
Away from equilibrium, at T e ≥ T p h T_{\mathrm{e}} \geq T_{\mathrm{ph}} T e ≥ T ph , the electron temperature controls the available electron–hole phase space, while the phonon temperature controls the acoustic-phonon occupation. The resulting spectrum is therefore an independent non-equilibrium rate-sum observable. It should not be interpreted as a thermal conductance. Therefore, we directly derive the finite-frequency spectra in BG and EP regimes for both MLG and BLG. In each regime, we compare the exact finite-frequency integral with its low-frequency approximation. This comparison identifies the frequency range in which the leading quadratic correction is reliable, following the general strategy used in finite-frequency electron–phonon noise studies [36 , 37 ] . For doped MLG, we additionally evaluate a full intraband on-shell integral without making either BG or EP reduction. This calculation retains the exact scattering geometry, graphene chirality factor, Fermi occupation, Bose occupation, and finite-frequency current vertices within the clean acoustic deformation-potential model. It approaches the BG result at low temperature and EP equipartition result at high temperature, while describing the smooth crossover between them [12 , 16 , 19 , 20 ] .
The numerical analysis includes the temperature dependence of the thermal conductance and zero-frequency noise, the full BG-to-EP crossover, normalized finite-frequency spectra, exact-versus-asymptotic comparisons, nonequilibrium frequency maps, and absolute device-scale noise estimates. For representative graphene parameters, the finite-frequency correction is generally small at ordinary low-gigahertz frequencies but can become measurable in the upper-gigahertz and sub-terahertz ranges. Finite-frequency heat-current fluctuations are particularly relevant to fast thermometry and calorimetric detection [4 , 5 , 7 , 8 , 9 , 10 , 11 , 36 ] . It is worth mentioning, that the present model is restricted to clean graphene coupled to longitudinal acoustic phonons through an unscreened deformation potential. The unscreened model is commonly used to describe intrinsic acoustic-phonon scattering in graphene [12 , 13 , 16 , 17 , 18 , 19 ] . Screening, disorder-assisted supercollisions, flexural phonons, optical phonons, substrate modes, and interface thermal resistance are outside the present scope. Some of these mechanisms can dominate in different temperature, density, disorder, or substrate regimes [16 , 20 , 25 , 26 , 27 , 28 , 32 ] . Keeping the model restricted allows the microscopic structure of the heat-current noise and the relation between its limiting regimes to be identified clearly analytically.
Theoretical model and heat-current operator
Total Hamiltonian
We consider electrons in graphene coupled to in-plane longitudinal acoustic phonons through the deformation potential. The total Hamiltonian consists of three terms
H ^ = H ^ e + H ^ p h + H ^ e − p h . (3) \hat{H} = \hat{H}_{\mathrm{e}} + \hat{H}_{\mathrm{ph}} + \hat{H}_{\mathrm{e-ph}}. \tag*{(3)} H ^ = H ^ e + H ^ ph + H ^ e − ph . (3)
The first term corresponds to free MLG or BLG Hamiltonian and is given by
H ^ e = ∑ k , α ϵ k α c ^ k α † c ^ k α , (4) \hat{H}_{\mathrm{e}}=\sum_{\mathbf{k},\alpha}\epsilon_{\mathbf{k}\alpha}\hat{c}^{\dagger}_{\mathbf{k}\alpha}\hat{c}_{\mathbf{k}\alpha}, \tag*{(4)} H ^ e = k , α ∑ ϵ k α c ^ k α † c ^ k α , (4)
where the electron with momentum k \mathbf{k} k and band index α \alpha α has energy ϵ k α \epsilon_{\mathbf{k}\alpha} ϵ k α . The operators c ^ † \hat{c}^{\dagger} c ^ † and c ^ \hat{c} c ^ create and annihilate electrons, respectively. The band index is α = + 1 \alpha=+1 α = + 1 for the conduction band and α = − 1 \alpha=-1 α = − 1 for the valence band. For MLG the spectrum is linear in momentum, ϵ k α = α ℏ v F k \epsilon_{\mathbf{k}\alpha}=\alpha\hbar v_{\mathrm{F}}k ϵ k α = α ℏ v F k , and v F v_{\mathrm{F}} v F is a Fermi velocity. That is, conduction-band electrons have energy + ℏ v F k +\hbar v_{\mathrm{F}}k + ℏ v F k , while valence-band states have energy − ℏ v F k -\hbar v_{\mathrm{F}}k − ℏ v F k . For BLG, the low-energy dispersion is parabolic: ϵ k α = α ℏ 2 k 2 / 2 m \epsilon_{\mathbf{k}\alpha}=\alpha\hbar^{2}k^{2}/2m ϵ k α = α ℏ 2 k 2 /2 m with effective mass m = γ 1 / 2 v 0 2 m=\gamma_{1}/2v_{0}^{2} m = γ 1 /2 v 0 2 , and v 0 ≃ v F v_{0}\simeq v_{\mathrm{F}} v 0 ≃ v F is the intralayer band velocity of graphene and γ 1 \gamma_{1} γ 1 is the dominant interlayer hopping energy in BLG. It describes electron hopping between the carbon atoms that lie directly above and below one another. This bilayer approximation further is valid when the chemical potential, electron and phonon temperatures, and relevant frequency scale are all much smaller than the interlayer hopping, i.e. max ( ∣ μ ∣ , k B T e , k B T p h , ℏ ∣ ω ∣ ) ≪ γ 1 \max\left(|\mu|,k_{\mathrm{B}}T_{\mathrm{e}},k_{\mathrm{B}}T_{\mathrm{ph}},\hbar|\omega|\right)\ll\gamma_{1} max ( ∣ μ ∣ , k B T e , k B T ph , ℏ∣ ω ∣ ) ≪ γ 1 . The second term is the phonon Hamiltonian
H ^ p h = ∑ q E q b ^ q † b ^ q , (5) \hat{H}_{\mathrm{ph}}=\sum_{\mathbf{q}}E_{\mathbf{q}}\hat{b}^{\dagger}_{\mathbf{q}}\hat{b}_{\mathbf{q}}, \tag*{(5)} H ^ ph = q ∑ E q b ^ q † b ^ q , (5)
where E q = ℏ v s q E_{\mathbf{q} }=\hbar v_{\mathrm{s}}q E q = ℏ v s q is the acoustic phonon energy, v s v_{\mathrm{s}} v s is the sound velocity of the longitudinal acoustic phonons. The third term is the electron–phonon interaction
H ^ e − p h = ∑ k α , p β , q M p k , q β α c ^ p β † c ^ k α ( b ^ q + b ^ − q † ) . (6) \hat{H}_{\mathrm{e-ph}}=\sum_{\mathbf{k}\alpha,\mathbf{p}\beta,\mathbf{q}}M^{\beta\alpha}_{\mathbf{p}\mathbf{k},\mathbf{q}}\hat{c}^{\dagger}_{\mathbf{p}\beta}\hat{c}_{\mathbf{k}\alpha}\left(\hat{b}_{\mathbf{q}}+\hat{b}^{\dagger}_{-\mathbf{q}}\right). \tag*{(6)} H ^ e − ph = k α , p β , q ∑ M pk , q β α c ^ p β † c ^ k α ( b ^ q + b ^ − q † ) . (6)
In the above Eq. (6) the product of electron annihilation and creation operators, c ^ p β † c ^ k α \hat{c}^{\dagger}_{\mathbf{p}\beta}\hat{c}_{\mathbf{k}\alpha} c ^ p β † c ^ k α , removes an electron from the initial state ( k , α ) (\mathbf{k},\alpha) ( k , α ) and places it in the final state ( p , β ) (\mathbf{p},\beta) ( p , β ) . The coefficient M p k , q β α M^{\beta\alpha}_{\mathbf{p}\mathbf{k},\mathbf{q}} M pk , q β α is the electron–phonon coupling matrix element, which contains the deformation-potential strength and the chirality overlap between the two electronic states. The bosonic annihilation operator b ^ q \hat{b}_{\mathbf{q}} b ^ q corresponds to phonon absorption, i.e. an existing phonon is annihilated while the electron changes momentum. In contrast, the bosonic creation operator b ^ − q † \hat{b}^{\dagger}_{-\mathbf{q}} b ^ − q † corresponds to phonon emission, namely the electron loses energy and creates a phonon with momentum − q -\mathbf{q} − q . The translational invariance imposes p = k + q \mathbf{p}=\mathbf{k}+\mathbf{q} p = k + q . The deformation-potential matrix element is known [16 ] and may be written as
∣ M p k , q β α ∣ 2 = D 2 ℏ q 2 A ρ m v s F α β ( ϕ ) , (7) \left|M^{\beta\alpha}_{\mathbf{p}\mathbf{k},\mathbf{q}}\right|^{2}=\frac{D^{2}\hbar q}{2A\rho_{\mathrm{m}}v_{\mathrm{s}}}F_{\alpha\beta}(\phi), \tag*{(7)} M pk , q β α 2 = 2 A ρ m v s D 2 ℏ q F α β ( ϕ ) , (7)
where A A A is the graphene area and ρ m \rho_{\mathrm{m}} ρ m is the areal mass density of the corresponding mono- or bilayer system. The Dirac chirality factors are [12 , 16 ]
F α β M L G ( ϕ ) = 1 + α β cos ϕ 2 , (8) F_{\alpha\beta}^{\mathrm{MLG}}(\phi)=\frac{1+\alpha\beta\cos\phi}{2}, \tag*{(8)} F α β MLG ( ϕ ) = 2 1 + α β cos ϕ , (8)
F α β B L G ( ϕ ) = 1 + α β cos 2 ϕ 2 , (9) F_{\alpha\beta}^{\mathrm{BLG}}(\phi)=\frac{1+\alpha\beta\cos2\phi}{2}, \tag*{(9)} F α β BLG ( ϕ ) = 2 1 + α β cos 2 ϕ , (9)
with ϕ \phi ϕ the angle between the incoming and outgoing electron momenta. Furthermore, the two subsystems are assumed to be internally thermal, thus the distribution functions are given by Fermi-Dirac f e ( ϵ ; T e ) = 1 / [ exp [ ( ϵ − μ ) / ( k B T e ) ] + 1 ] f_{\mathrm{e}}(\epsilon;T_{\mathrm{e}})=1/[\exp[(\epsilon-\mu)/(k_{\mathrm{B}}T_{\mathrm{e}})]+1] f e ( ϵ ; T e ) = 1/ [ exp [( ϵ − μ ) / ( k B T e )] + 1 ] for electrons and Bose-Einstein n p h ( E ; T p h ) = 1 / [ exp ( E / ( k B T p h ) ) − 1 ] n_{\mathrm{ph}}(E;T_{\mathrm{ph}})=1/[\exp(E/(k_{\mathrm{B}}T_{\mathrm{ph}}))-1] n ph ( E ; T ph ) = 1/ [ exp ( E / ( k B T ph )) − 1 ] for phonons. This is the same quasiequilibrium framework used in the cooling-power theory of Ref. [16 ] . It is appropriate when electron–electron scattering maintains a Fermi distribution and phonon relaxation maintains a Bose distribution even though T e T_{\mathrm{e}} T e and T p h T_{\mathrm{ph}} T ph differ.
Heat-current operator
We define the heat current to be positive when energy is transferred from the electron subsystem to the phonon subsystem, thus T e ≥ T p h T_{\mathrm{e}}\geq T_{\mathrm{ph}} T e ≥ T ph . The electron and phonon Hamiltonians are denoted by H ^ e \hat{H}_{\mathrm{e}} H ^ e and H ^ p h \hat{H}_{\mathrm{ph}} H ^ ph , respectively. Their Heisenberg time derivatives describe the instantaneous rates at which the energies of the two subsystems change. We use the symmetric heat current definition
J ^ = 1 2 ∂ t ( H ^ p h − H ^ e ) , (10) \hat{J}=\frac{1}{2}\partial_{t}\left(\hat{H}_{\mathrm{ph}}-\hat{H}_{\mathrm{e}}\right), \tag*{(10)} J ^ = 2 1 ∂ t ( H ^ ph − H ^ e ) , (10)
where ∂ t H ^ i = i [ H ^ , H ^ i ] / ℏ \partial_{t}\hat{H}_{i}=i[\hat{H},\hat{H}_{i}]/\hbar ∂ t H ^ i = i [ H ^ , H ^ i ] /ℏ and H ^ \hat{H} H ^ is the total Hamiltonian, including the electron, phonon, and electron–phonon interaction terms from (3) . The sign convention in (10) has a direct physical meaning, namely when electrons lose energy, ∂ t ⟨ H ^ e ⟩ < 0 \partial_{t}\langle\hat{H}_{\mathrm{e}}\rangle<0 ∂ t ⟨ H ^ e ⟩ < 0 , and when phonons gain the same energy, ∂ t ⟨ H ^ p h ⟩ > 0 \partial_{t}\langle\hat{H}_{\mathrm{ph}}\rangle>0 ∂ t ⟨ H ^ ph ⟩ > 0 . Both contributions then make the average of J ^ \hat{J} J ^ positive. Reversing the convention would replace J J J by − J -J − J . The average current would change sign, but the current noise, which contains a product of two current operators, would remain unchanged, which one can see in next sections. The factor 1 / 2 1/2 1/2 treats the changes of the electron and phonon energies symmetrically. This is important because, at a finite observation frequency, the interaction energy cannot in general be assigned completely to either subsystem. In the zero-frequency steady-state limit, the electron energy lost and the phonon energy gained become equal. Throughout this work, the symmetric convention in (10) is used consistently.
Only the electron–phonon interaction contributes to the commutators in (10) . The electron Hamiltonian commutes with the free phonon Hamiltonian, and each free subsystem Hamiltonian commutes with itself. Evaluating the commutators gives
∂ t H ^ e = i ℏ ∑ k α , p β , q M p k , q β α Δ k p α β c ^ p β † c ^ k α ( b ^ q + b ^ − q † ) , (11) \partial_{t}\hat{H}_{\mathrm{e}}=\frac{i}{\hbar}\sum_{\mathbf{k}\alpha,\mathbf{p}\beta,\mathbf{q}}M^{\beta\alpha}_{\mathbf{p}\mathbf{k},\mathbf{q}}\Delta^{\alpha\beta}_{\mathbf{k}\mathbf{p}}\hat{c}^{\dagger}_{\mathbf{p}\beta}\hat{c}_{\mathbf{k}\alpha}\left(\hat{b}_{\mathbf{q}}+\hat{b}^{\dagger}_{-\mathbf{q}}\right), \tag*{(11)} ∂ t H ^ e = ℏ i k α , p β , q ∑ M pk , q β α Δ kp α β c ^ p β † c ^ k α ( b ^ q + b ^ − q † ) , (11)
∂ t H ^ p h = i ℏ ∑ k α , p β , q M p k , q β α E q c ^ p β † c ^ k α ( b ^ q − b ^ − q † ) , (12) \partial_{t}\hat{H}_{\mathrm{ph}}=\frac{i}{\hbar}\sum_{\mathbf{k}\alpha,\mathbf{p}\beta,\mathbf{q}}M^{\beta\alpha}_{\mathbf{p}\mathbf{k},\mathbf{q}}E_{\mathbf{q}}\hat{c}^{\dagger}_{\mathbf{p}\beta}\hat{c}_{\mathbf{k}\alpha}\left(\hat{b}_{\mathbf{q}}-\hat{b}^{\dagger}_{-\mathbf{q}}\right), \tag*{(12)} ∂ t H ^ ph = ℏ i k α , p β , q ∑ M pk , q β α E q c ^ p β † c ^ k α ( b ^ q − b ^ − q † ) , (12)
where to describe an electron transition, we introduce the energy difference
Δ k p α β = ϵ k α − ϵ p β . (13) \Delta_{kp}^{\alpha\beta}=\epsilon_{k\alpha}-\epsilon_{p\beta}. \tag*{(13)} Δ k p α β = ϵ k α − ϵ pβ . (13)
The vector k \mathbf{k} k and band index α \alpha α label the initial electron state, whereas p \mathbf{p} p and β \beta β label the final electron state. The corresponding band energies are ϵ k α \epsilon_{k\alpha} ϵ k α and ϵ p β \epsilon_{p\beta} ϵ pβ . Thus, Δ k p α β > 0 \Delta_{kp}^{\alpha\beta}>0 Δ k p α β > 0 means that the electron loses energy during the transition, while Δ k p α β < 0 \Delta_{kp}^{\alpha\beta}<0 Δ k p α β < 0 means that the electron gains energy. The phonon energy E q E_q E q is given in Eq. (5) and electron–phonon matrix element M p k , q β α M_{pk,q}^{\beta\alpha} M p k , q β α gives the amplitude for the electron–phonon scattering process, Eq. (6) . The different signs of the creation and annihilation operators in Eq. (12) follow from their commutators with the phonon Hamiltonian. An annihilation operator removes the energy E q E_q E q from the phonon subsystem, whereas a creation operator adds the energy E q E_q E q . By contrast, both terms enter Eq. (11) with the same sign inside the parentheses because both are coupled to the same electronic transition operator; their physical distinction is imposed by the corresponding energy shells.
Substituting Eqs. (11) and (12) into Eq. (10) gives the symmetric heat-current operator
J ^ = i 2 ℏ ∑ k α , p β , q M p k , q β α c ^ p β † b ^ q † c ^ k α × [ ( E q − Δ k p α β ) b ^ q − ( E q + Δ k p α β ) b ^ − q † ] . (14) \hat{J}=\frac{i}{2\hbar}\sum_{\mathbf{k}\alpha,\mathbf{p}\beta,q}M_{pk,q}^{\beta\alpha}\hat{c}_{p\beta}^{\dagger}\hat{b}_{q}^{\dagger}\hat{c}_{k\alpha}
\times\left[\left(E_q-\Delta_{kp}^{\alpha\beta}\right)\hat{b}_q-\left(E_q+\Delta_{kp}^{\alpha\beta}\right)\hat{b}_{-q}^{\dagger}\right]. \tag*{(14)} J ^ = 2ℏ i k α , p β , q ∑ M p k , q β α c ^ pβ † b ^ q † c ^ k α × [ ( E q − Δ k p α β ) b ^ q − ( E q + Δ k p α β ) b ^ − q † ] . (14)
The full momentum and band sums, together with the Hermiticity properties of the electron–phonon matrix element, ensure that Eq. (14) represents a Hermitian physical observable. The first term in Eq. (14) contains b q b_q b q and therefore describes a process in which a phonon is annihilated. The electron absorbs the phonon energy and moves to a state of higher energy. On the corresponding zero-frequency energy shell, the electron energy difference is negative and has magnitude E q E_q E q . The coefficient E q − Δ k p α β E_q-\Delta_{kp}^{\alpha\beta} E q − Δ k p α β then has magnitude 2 E q 2E_q 2 E q . The second term contains b − q † b_{-q}^{\dagger} b − q † and describes phonon creation. In this process the electron moves to a lower-energy state and transfers energy to the phonon subsystem. On the corresponding zero-frequency energy shell, the electron energy difference is positive and equal in magnitude to E q E_q E q . The coefficient E q + Δ k p α β E_q+\Delta_{kp}^{\alpha\beta} E q + Δ k p α β again has magnitude 2 E q 2E_q 2 E q .
This observation is important for the noise calculation further. The two current vertices are squared in the heat current–current correlator. At zero frequency, both the emission and absorption contributions therefore acquire the same energy weight, namely four times the square of the phonon energy. The two processes consequently enter the zero-frequency noise with equal energy weights and must both be retained. At finite frequency, the energy provided to or absorbed from a detector shifts the two energy-conservation conditions. The two current vertices then become the square of the detector energy minus twice the phonon energy and the square of the detector energy plus twice the phonon energy. The distinction between the two coefficients in Eq. (14) is therefore the microscopic origin of the nontrivial finite-frequency spectrum.
Heat-current noise
The unsymmetrized spectrum is defined in Eq. (2) . In the leading order of electron-phonon coupling, the fermionic four-point function contains direct and exchange Wick contractions. The direct contraction sets p = k \mathbf{p}=\mathbf{k} p = k . Momentum conservation then forces q = 0 \mathbf{q}=0 q = 0 , and the acoustic phonon energy and both current vertices vanish, thus the direct contribution is identically zero. The exchange contraction sets the two electron lines into a real scattering process and gives the occupation factor f p β ( 1 − f k α ) f_{p\beta}(1-f_{k\alpha}) f pβ ( 1 − f k α ) . After some algebra (see details in Appendix A) we obtain the following expression for noise
Z ( ω ) = π g d 2 ℏ ∑ k α , p β , q ∣ M p k , q β α ∣ 2 f p β ( T e ) [ 1 − f k α ( T e ) ] × { ( Δ k p α β − E q ) 2 [ 1 + n p h ( E q , T p h ) ] δ ( ℏ ω − Δ k p α β − E q ) + ( Δ k p α β + E q ) 2 n p h ( E q , T p h ) δ ( ℏ ω − Δ k p α β + E q ) } . (15) \begin{aligned}
\mathcal{Z}(\omega)=\frac{\pi g_d}{2\hbar}\sum_{\mathbf{k}\alpha,\mathbf{p}\beta,q}\left|M_{pk,q}^{\beta\alpha}\right|^2 f_{p\beta}(T_e)\left[1-f_{k\alpha}(T_e)\right] \\
&\times\left\{\left(\Delta_{kp}^{\alpha\beta}-E_q\right)^2\left[1+n_{\mathrm{ph}}(E_q,T_{\mathrm{ph}})\right]\delta\left(\hbar\omega-\Delta_{kp}^{\alpha\beta}-E_q\right)
+\left(\Delta_{kp}^{\alpha\beta}+E_q\right)^2n_{\mathrm{ph}}(E_q,T_{\mathrm{ph}})\delta\left(\hbar\omega-\Delta_{kp}^{\alpha\beta}+E_q\right)\right\}. \tag*{(15)}
\end{aligned} Z ( ω ) = 2ℏ π g d k α , p β , q ∑ M p k , q β α 2 f pβ ( T e ) [ 1 − f k α ( T e ) ] × { ( Δ k p α β − E q ) 2 [ 1 + n ph ( E q , T ph ) ] δ ( ℏ ω − Δ k p α β − E q ) + ( Δ k p α β + E q ) 2 n ph ( E q , T ph ) δ ( ℏ ω − Δ k p α β + E q ) } . (15)
(15) is the leading contribution, of second order in the electron–phonon coupling, to the unsymmetrized heat-current noise. The electronic occupation factor f p β ( T e ) [ 1 − f k α ( T e ) ] f_{p\beta}(T_e)[1-f_{k\alpha}(T_e)] f pβ ( T e ) [ 1 − f k α ( T e )] requires the initial electron state ( p , β ) (p,\beta) ( p , β ) to be occupied and the final state ( k , α ) (k,\alpha) ( k , α ) to be empty. The first factor is therefore the probability of finding the initial electron, while the second is the Pauli blocking factor for the final state. Both are evaluated at the electron temperature T e T_e T e .
The first term in Eq. (15) is associated with phonon emission. Its statistical factor 1 + n p h ( E q , T p h ) 1+n_{\mathrm{ph}}(E_q,T_{\mathrm{ph}}) 1 + n ph ( E q , T ph ) contains both spontaneous emission and Bose-enhanced stimulated emission. The second term is associated with phonon absorption and is proportional to the phonon occupation n p h ( E q , T p h ) n_{\mathrm{ph}}(E_q,T_{\mathrm{ph}}) n ph ( E q , T ph ) , because absorption is possible only when a phonon is already present in the mode. The phonon distribution is evaluated at the phonon temperature T p h T_{\mathrm{ph}} T ph , which may differ from T e T_e T e . The two Dirac delta functions impose energy conservation for the two processes, including the energy ℏ ω \hbar\omega ℏ ω exchanged with a finite-frequency probe. Consequently, positive and negative frequencies are not generally equivalent in the unsymmetrized spectrum. The factors ( Δ k p α β − E q ) 2 (\Delta_{kp}^{\alpha\beta}-E_q)^2 ( Δ k p α β − E q ) 2 and ( Δ k p α β + E q ) 2 (\Delta_{kp}^{\alpha\beta}+E_q)^2 ( Δ k p α β + E q ) 2 are the squared heat-current vertices generated by the symmetric current operator. They determine how strongly each allowed scattering event contributes to the noise. At zero frequency, the emission shell fixes Δ k p α β = − E q \Delta_{kp}^{\alpha\beta}=-E_q Δ k p α β = − E q , whereas the absorption shell fixes Δ k p α β = + E q \Delta_{kp}^{\alpha\beta}=+E_q Δ k p α β = + E q . The two current vertices then acquire the same value, 4 E q 2 4E_q^2 4 E q 2 . Thus emission and absorption carry equal energy-transfer weights in the zero-frequency noise and must both be retained. At finite frequency, the two weights are different and become functions of both the probe energy and the phonon energy. This difference is the microscopic origin of the nontrivial frequency dependence of the heat-current spectrum. The spectrum in Eq. (15) is not generally even in ω \omega ω . The symmetrized spectrum used below is obtained by averaging the positive- and negative-frequency contributions according to Eq. (1) .
At zero frequency the delta functions in Eq. (15) impose that Δ = ± E q \Delta=\pm E_q Δ = ± E q . In both cases the squared current vertex becomes 4 E q 2 4E_q^2 4 E q 2 . In intraband notations one gets
S ( 0 ) = 2 π g d ℏ ∑ k , q ∣ M k − q , k ∣ 2 E q 2 [ f k ( 1 − f k − q ) ( 1 + n q ) + f k − q ( 1 − f k ) n q ] δ ( ϵ k − ϵ k − q − E q ) . (16) S(0)=\frac{2\pi g_d}{\hbar}\sum_{\mathbf{k},\mathbf{q}}|M_{\mathbf{k}-\mathbf{q},\mathbf{k}}|^2E_q^2\left[f_{\mathbf{k}}(1-f_{\mathbf{k}-\mathbf{q}})(1+n_q)+f_{\mathbf{k}-\mathbf{q}}(1-f_{\mathbf{k}})n_q\right]\delta(\epsilon_{\mathbf{k}}-\epsilon_{\mathbf{k}-\mathbf{q}}-E_q). \tag*{(16)} S ( 0 ) = ℏ 2 π g d k , q ∑ ∣ M k − q , k ∣ 2 E q 2 [ f k ( 1 − f k − q ) ( 1 + n q ) + f k − q ( 1 − f k ) n q ] δ ( ϵ k − ϵ k − q − E q ) . (16)
The corresponding mean heat current, Q ≡ ⟨ J ^ ⟩ Q\equiv\langle\hat{J}\rangle Q ≡ ⟨ J ^ ⟩ , is
Q = 2 π g d ℏ ∑ k , q ∣ M k − q , k ∣ 2 E q [ f k ( 1 − f k − q ) ( 1 + n q ) − f k − q ( 1 − f k ) n q ] δ ( ϵ k − ϵ k − q − E q ) . (17) Q=\frac{2\pi g_d}{\hbar}\sum_{\mathbf{k},\mathbf{q}}|M_{\mathbf{k}-\mathbf{q},\mathbf{k}}|^2E_q\left[f_{\mathbf{k}}(1-f_{\mathbf{k}-\mathbf{q}})(1+n_q)-f_{\mathbf{k}-\mathbf{q}}(1-f_{\mathbf{k}})n_q\right]\delta(\epsilon_{\mathbf{k}}-\epsilon_{\mathbf{k}-\mathbf{q}}-E_q). \tag*{(17)} Q = ℏ 2 π g d k , q ∑ ∣ M k − q , k ∣ 2 E q [ f k ( 1 − f k − q ) ( 1 + n q ) − f k − q ( 1 − f k ) n q ] δ ( ϵ k − ϵ k − q − E q ) . (17)
The only structural change between Eqs. (16) and (17) is the sign between the two rates and one additional transferred-energy factor in the noise.
At equilibrium, T e = T p h = T T_e=T_{\mathrm{ph}}=T T e = T ph = T , detailed balance gives Q = 0 Q=0 Q = 0 . The linear thermal conductance is
G e p ( T ) = ∂ Q ∂ T e ∣ T e = T p h = T , (18) G_{\mathrm{ep}}(T)=\left.\frac{\partial Q}{\partial T_e}\right|_{T_e=T_{\mathrm{ph}}=T}, \tag*{(18)} G ep ( T ) = ∂ T e ∂ Q T e = T ph = T , (18)
and the zero-frequency thermodynamic FDT is satisfied, S ( 0 ) = 2 k B T 2 G e p ( T ) S(0)=2k_{\mathrm{B}}T^2G_{\mathrm{ep}}(T) S ( 0 ) = 2 k B T 2 G ep ( T ) . Below we use this relation as a normalization check in every equilibrium limit. At nonzero frequency we do not infer a dynamic thermal conductance from S ( ω ) S(\omega) S ( ω ) . Finite-frequency heat-current noise can contain extra quantum terms that are not fixed by the static conductance [35 , 36 ] . All finite-frequency results below are therefore obtained directly from Eq. (15) and Eq. (1) .
Zero-frequency results
This section collects the analytic zero-frequency limits. The detailed momentum and energy integrations are given in Appendices B and C. Unless stated otherwise, the spin–valley degeneracy g d = 4 g_d=4 g d = 4 is included.
Bloch–Grüneisen regime
The BG regime describes low-temperature scattering by long-wavelength acoustic phonons. The characteristic momentum available for intraband scattering is set by twice the Fermi wave vector. Namely, we consider doped, degenerate graphene, for which ∣ μ ∣ ≫ k B T |\mu|\gg k_{\mathrm{B}}T ∣ μ ∣ ≫ k B T and scattering occurs near a well-defined Fermi surface. The maximum momentum transfer between states on this Fermi surface is 2 k F 2k_{\mathrm{F}} 2 k F . The corresponding acoustic-phonon energy defines the BG temperature, k B T B G = 2 ℏ v s k F k_{\mathrm{B}}T_{\mathrm{BG}}=2\hbar v_s k_{\mathrm{F}} k B T BG = 2ℏ v s k F . For T ≪ T B G T\ll T_{\mathrm{BG}} T ≪ T BG , the characteristic thermal phonon momentum q T = k B T / ( ℏ v s ) q_T=k_{\mathrm{B}}T/(\hbar v_s) q T = k B T / ( ℏ v s ) is much smaller than 2 k F 2k_{\mathrm{F}} 2 k F , so scattering is dominated by long-wavelength phonons. For monolayer graphene, the Fermi wave vector and Bloch–Grüneisen temperature are
k F M L G = ∣ μ ∣ ℏ v F , T B G M L G = 2 ℏ v s k F M L G k B , (19) k_{\mathrm{F}}^{\mathrm{MLG}}=\frac{|\mu|}{\hbar v_{\mathrm{F}}},\qquad T_{\mathrm{BG}}^{\mathrm{MLG}}=\frac{2\hbar v_s k_{\mathrm{F}}^{\mathrm{MLG}}}{k_{\mathrm{B}}}, \tag*{(19)} k F MLG = ℏ v F ∣ μ ∣ , T BG MLG = k B 2ℏ v s k F MLG , (19)
where μ \mu μ is the chemical potential measured from the charge-neutrality point, v F v_{\mathrm{F}} v F is the monolayer Fermi velocity, and v s v_s v s is the longitudinal sound velocity, the subscript “BG” stands for Bloch–Grüneisen. The use of ∣ μ ∣ |\mu| ∣ μ ∣ makes the result applicable to both electron and hole doping. The typical thermal phonon momentum is of order k B T / ( ℏ v s ) k_{\mathrm{B}}T/(\hbar v_s) k B T / ( ℏ v s ) . Therefore, the condition T ≪ T B G M L G T\ll T_{\mathrm{BG}}^{\mathrm{MLG}} T ≪ T BG MLG implies that the thermally active phonons have momenta much smaller than 2 k F 2k_{\mathrm{F}} 2 k F . Scattering is then dominated by small-momentum processes close to the Fermi surface. The additional condition ∣ μ ∣ ≫ k B T |\mu|\gg k_{\mathrm{B}}T ∣ μ ∣ ≫ k B T ensures that the electronic distribution is strongly degenerate and that only a narrow energy window around the Fermi level contributes. In this regime, the equilibrium zero-frequency heat-current noise and the electron–phonon thermal conductance per unit area are
S B G M L G ( 0 ) A = 8 π 2 15 D 2 ∣ μ ∣ k B 5 T 5 ρ M L G ℏ 5 v F 3 v s 3 , (20) \frac{S_{\mathrm{BG}}^{\mathrm{MLG}}(0)}{A}=\frac{8\pi^2}{15}\frac{D^2|\mu|k_{\mathrm{B}}^5T^5}{\rho_{\mathrm{MLG}}\hbar^5v_{\mathrm{F}}^3v_s^3}, \tag*{(20)} A S BG MLG ( 0 ) = 15 8 π 2 ρ MLG ℏ 5 v F 3 v s 3 D 2 ∣ μ ∣ k B 5 T 5 , (20)
G B G M L G ( T ) A = 4 π 2 15 D 2 ∣ μ ∣ k B 4 T 3 ρ M L G ℏ 5 v F 3 v s 3 , (21) \frac{G_{\mathrm{BG}}^{\mathrm{MLG}}(T)}{A}=\frac{4\pi^2}{15}\frac{D^2|\mu|k_{\mathrm{B}}^4T^3}{\rho_{\mathrm{MLG}}\hbar^5v_{\mathrm{F}}^3v_s^3}, \tag*{(21)} A G BG MLG ( T ) = 15 4 π 2 ρ MLG ℏ 5 v F 3 v s 3 D 2 ∣ μ ∣ k B 4 T 3 , (21)
where D D D is the deformation-potential coupling and ρ M L G \rho_{\mathrm{MLG}} ρ MLG is the monolayer areal mass density.
Equation (21) shows that the BG conductance grows as the third power of temperature. This strong suppression at low temperature results from the small number of thermally occupied acoustic phonons and the limited electron–phonon scattering phase space. The conductance is proportional to ∣ μ ∣ |\mu| ∣ μ ∣ , mainly because the electronic density of states at the Fermi level in monolayer graphene increases linearly with the absolute chemical potential. The noise in Eq. (20) contains two additional powers of temperature compared with the conductance. In equilibrium, this follows from the zero-frequency FDT, in which the noise is proportional to the conductance multiplied by k B T 2 k_{\mathrm{B}}T^{2} k B T 2 . Therefore, a conductance proportional to T 3 T^{3} T 3 produces an equilibrium noise proportional to T 5 T^{5} T 5 . The finite equilibrium noise reflects continuing phonon-emission and phonon-absorption events even though their average contributions to the heat current cancel.
The corresponding mean cooling power agrees with the Viljas–Heikkilä result obtained from the Boltzmann transport equation [16 ] ,
Q B G M L G A = Σ M L G ( T e 4 − T p h 4 ) , Σ M L G = π 2 D 2 ∣ μ ∣ k B 4 15 ρ M L G ℏ 5 v F 3 v s 3 . (22) \frac{Q_{\mathrm{BG}}^{\mathrm{MLG}}}{A}=\Sigma_{\mathrm{MLG}}\left(T_{\mathrm{e}}^{4}-T_{\mathrm{ph}}^{4}\right),\qquad\Sigma_{\mathrm{MLG}}=\frac{\pi^{2}D^{2}|\mu|k_{\mathrm{B}}^{4}}{15\rho_{\mathrm{MLG}}\hbar^{5}v_{\mathrm{F}}^{3}v_{\mathrm{s}}^{3}}. \tag*{(22)} A Q BG MLG = Σ MLG ( T e 4 − T ph 4 ) , Σ MLG = 15 ρ MLG ℏ 5 v F 3 v s 3 π 2 D 2 ∣ μ ∣ k B 4 . (22)
The coefficient Σ M L G \Sigma_{\mathrm{MLG}} Σ MLG characterizes the strength of the electron–phonon energy coupling. When T e > T p h T_{\mathrm{e}}>T_{\mathrm{ph}} T e > T ph , the expression is positive and energy flows from electrons to phonons. When the two temperatures are equal, the net current vanishes. Differentiating this cooling law with respect to the electron temperature at T e = T p h = T T_{\mathrm{e}}=T_{\mathrm{ph}}=T T e = T ph = T gives Eq. (21) .
We next consider BLG in the low-energy parabolic-band approximation. In this case, the band velocity is not constant but depends on the Fermi wave vector. The relevant electronic scales are
k F B L G = γ 1 ∣ μ ∣ ℏ v 0 , u F = 2 v 0 ∣ μ ∣ γ 1 , (23) k_{\mathrm{F}}^{\mathrm{BLG}}=\frac{\sqrt{\gamma_{1}|\mu|}}{\hbar v_{0}},\qquad u_{\mathrm{F}}=2v_{0}\sqrt{\frac{|\mu|}{\gamma_{1}}}, \tag*{(23)} k F BLG = ℏ v 0 γ 1 ∣ μ ∣ , u F = 2 v 0 γ 1 ∣ μ ∣ , (23)
where γ 1 \gamma_{1} γ 1 is the dominant interlayer hopping energy, v 0 v_{0} v 0 is the monolayer band velocity entering the effective BLG Hamiltonian in Eq. (4) , and u F u_{\mathrm{F}} u F is the physical BLG Fermi velocity. The corresponding BG temperature is T B G B L G = 2 ℏ v s k F B L G / k B T_{\mathrm{BG}}^{\mathrm{BLG}}=2\hbar v_{\mathrm{s}}k_{\mathrm{F}}^{\mathrm{BLG}}/k_{\mathrm{B}} T BG BLG = 2ℏ v s k F BLG / k B . The parabolic-band description requires the relevant electronic energies to remain below the interlayer scale, namely ∣ μ ∣ , k B T ≪ γ 1 |\mu|,k_{\mathrm{B}}T\ll\gamma_{1} ∣ μ ∣ , k B T ≪ γ 1 . The BG result further requires T ≪ T B G B L G T\ll T_{\mathrm{BG}}^{\mathrm{BLG}} T ≪ T BG BLG and ∣ μ ∣ ≫ k B T |\mu|\gg k_{\mathrm{B}}T ∣ μ ∣ ≫ k B T . Under these assumptions, the equilibrium BLG noise and conductance per unit area are
S B G B L G ( 0 ) A = 2 π 2 15 D 2 γ 1 k B 5 T 5 ρ B L G ℏ 5 v 0 3 v s 3 γ 1 ∣ μ ∣ , (24) \frac{S_{\mathrm{BG}}^{\mathrm{BLG}}(0)}{A}=\frac{2\pi^{2}}{15}\frac{D^{2}\gamma_{1}k_{\mathrm{B}}^{5}T^{5}}{\rho_{\mathrm{BLG}}\hbar^{5}v_{0}^{3}v_{\mathrm{s}}^{3}}\sqrt{\frac{\gamma_{1}}{|\mu|}}, \tag*{(24)} A S BG BLG ( 0 ) = 15 2 π 2 ρ BLG ℏ 5 v 0 3 v s 3 D 2 γ 1 k B 5 T 5 ∣ μ ∣ γ 1 , (24)
G B G B L G ( T ) A = π 2 15 D 2 γ 1 k B 4 T 3 ρ B L G ℏ 5 v 0 3 v s 3 γ 1 ∣ μ ∣ , (25) \frac{G_{\mathrm{BG}}^{\mathrm{BLG}}(T)}{A}=\frac{\pi^{2}}{15}\frac{D^{2}\gamma_{1}k_{\mathrm{B}}^{4}T^{3}}{\rho_{\mathrm{BLG}}\hbar^{5}v_{0}^{3}v_{\mathrm{s}}^{3}}\sqrt{\frac{\gamma_{1}}{|\mu|}}, \tag*{(25)} A G BG BLG ( T ) = 15 π 2 ρ BLG ℏ 5 v 0 3 v s 3 D 2 γ 1 k B 4 T 3 ∣ μ ∣ γ 1 , (25)
where ρ B L G \rho_{\mathrm{BLG}} ρ BLG is the bilayer areal mass density. For an ideal bilayer it is approximately twice the monolayer areal mass density. The BLG conductance and noise have the same temperature powers as their MLG counterparts. The conductance again scales as T 3 T^{3} T 3 , while the equilibrium noise scales as T 5 T^{5} T 5 . These common powers reflect the same two-dimensional acoustic-phonon phase space and the same low-temperature BG restriction and FDT.
The BLG cooling power in the BG limit is
Q B G B L G A = Σ B L G ( T e 4 − T p h 4 ) , (26) \frac{Q_{\mathrm{BG}}^{\mathrm{BLG}}}{A}=\Sigma_{\mathrm{BLG}}\left(T_{\mathrm{e}}^{4}-T_{\mathrm{ph}}^{4}\right), \tag*{(26)} A Q BG BLG = Σ BLG ( T e 4 − T ph 4 ) , (26)
where the cooling-power coefficient is
Σ B L G = π 2 D 2 m k B 4 15 ρ B L G ℏ 5 u F v s 3 . (27) \Sigma_{\mathrm{BLG}}=\frac{\pi^{2}D^{2}m k_{\mathrm{B}}^{4}}{15\rho_{\mathrm{BLG}}\hbar^{5}u_{\mathrm{F}}v_{\mathrm{s}}^{3}}. \tag*{(27)} Σ BLG = 15 ρ BLG ℏ 5 u F v s 3 π 2 D 2 m k B 4 . (27)
The important difference is the dependence on carrier density. In MLG, the BG coefficient is proportional to ∣ μ ∣ |\mu| ∣ μ ∣ , whereas in BLG it is proportional to ∣ μ ∣ − 1 / 2 |\mu|^{-1/2} ∣ μ ∣ − 1/2 . The origin of this behavior can be seen from the angular energy-conservation delta function, whose Jacobian supplies a factor proportional to the inverse BLG Fermi velocity (see Appendix B, Eq. (64) ). Because u F ∝ ∣ μ ∣ u_{\mathrm{F}}\propto\sqrt{|\mu|} u F ∝ ∣ μ ∣ , this factor behaves as 1 / u F ∝ ∣ μ ∣ − 1 / 2 1/u_{\mathrm{F}}\propto|\mu|^{-1/2} 1/ u F ∝ ∣ μ ∣ − 1/2 . The apparent increase of the BLG BG result as ∣ μ ∣ |\mu| ∣ μ ∣ approaches zero should not be interpreted as a physical divergence at charge neutrality. The derivation assumes a degenerate Fermi surface, ∣ μ ∣ ≫ k B T |\mu|\gg k_{\mathrm{B}}T ∣ μ ∣ ≫ k B T , as well as a well-defined BG regime. Both assumptions fail sufficiently close to neutrality, where the full finite-temperature expression must be used.
Equipartition regime
In the EP regime the active phonons are classical, thus, n q ≃ k B T / E q ≫ 1 n_{q}\simeq k_{\mathrm{B}}T/E_{q}\gg1 n q ≃ k B T / E q ≫ 1 , and the scattering is quasi-elastic. In the EP regime, the thermal conductances of MLG and BLG are obtained using the deformation-potential matrix element in Eq. (7) and the chirality factors in Eqs. (8) and (9) , as derived in Appendix C. The Fermi-window integrals are evaluated exactly within the classical-phonon and quasi-elastic approximations, yielding asymptotic EP expressions whose validity does not extend across the full BG-to-EP crossover. For the two-temperature extension, we assume E q ≪ k B T p h E_{q}\ll k_{\mathrm{B}}T_{\mathrm{ph}} E q ≪ k B T ph and E q ≪ k B T e E_{q}\ll k_{\mathrm{B}}T_{\mathrm{e}} E q ≪ k B T e for the phonons that contribute appreciably to scattering. The resulting equilibrium thermal conductances per unit area are given by
g M L G ( μ , T ) = G E P M L G ( T ) A = D 2 k B 30 π ρ M L G ℏ 5 v F 6 [ 15 μ 4 + 30 π 2 μ 2 ( k B T ) 2 + 7 π 4 ( k B T ) 4 ] (28) g_{\mathrm{MLG}}(\mu,T)=\frac{G_{\mathrm{EP}}^{\mathrm{MLG}}(T)}{A}=\frac{D^{2}k_{\mathrm{B}}}{30\pi\rho_{\mathrm{MLG}}\hbar^{5}v_{\mathrm{F}}^{6}}\left[15\mu^{4}+30\pi^{2}\mu^{2}(k_{\mathrm{B}}T)^{2}+7\pi^{4}(k_{\mathrm{B}}T)^{4}\right] \tag*{(28)} g MLG ( μ , T ) = A G EP MLG ( T ) = 30 π ρ MLG ℏ 5 v F 6 D 2 k B [ 15 μ 4 + 30 π 2 μ 2 ( k B T ) 2 + 7 π 4 ( k B T ) 4 ] (28)
g B L G ( μ , T ) = G E P B L G ( T ) A = D 2 γ 1 3 k B 2 T 2 π ρ B L G ℏ 5 v 0 6 × ln [ 2 cosh ( μ 2 k B T ) ] . (29) g_{\mathrm{BLG}}(\mu,T)=\frac{G_{\mathrm{EP}}^{\mathrm{BLG}}(T)}{A}=\frac{D^{2}\gamma_{1}^{3}k_{\mathrm{B}}^{2}T}{2\pi\rho_{\mathrm{BLG}}\hbar^{5}v_{0}^{6}}\times\ln\left[2\cosh\left(\frac{\mu}{2k_{\mathrm{B}}T}\right)\right]. \tag*{(29)} g BLG ( μ , T ) = A G EP BLG ( T ) = 2 π ρ BLG ℏ 5 v 0 6 D 2 γ 1 3 k B 2 T × ln [ 2 cosh ( 2 k B T μ ) ] . (29)
In Eq. (29) above, the sound velocity cancels in the EP conductance because the classical Bose factor contributes 1 / ( v s q ) 1/(v_{\mathrm{s}}q) 1/ ( v s q ) , while the matrix element and transferred energy supply the compensating factors. One can again straightforwardly show that the zero-frequency noise again satisfies the FDT, namely S E P l ( 0 ) = 2 k B T 2 G E P l ( T ) S_{\mathrm{EP}}^{l}(0)=2k_{\mathrm{B}}T^{2}G_{\mathrm{EP}}^{l}(T) S EP l ( 0 ) = 2 k B T 2 G EP l ( T ) , where l = M L G , B L G l=\mathrm{MLG},\mathrm{BLG} l = MLG , BLG .
Two-temperature regime
We now consider electrons and phonons maintained at separate temperatures, T e T_{\mathrm{e}} T e and T p h T_{\mathrm{ph}} T ph . In the BG regime, both temperatures are assumed to satisfy T e , T p h ≪ T B G T_{\mathrm{e}},T_{\mathrm{ph}}\ll T_{\mathrm{BG}} T e , T ph ≪ T BG , while the electrons remain degenerate, k B T e ≪ ∣ μ ∣ k_{\mathrm{B}}T_{\mathrm{e}}\ll|\mu| k B T e ≪ ∣ μ ∣ . Within the degenerate Fermi-surface reduction described in Appendix B, the electronic energy integral is extended to the full real axis. The zero-frequency noise is obtained from the sum of the emission and absorption terms in Eq. (16) . Next, using the continuum and angular reductions in the integrations with respect to energy variable (see Appendix B), the MLG noise at zero frequency becomes
S B G M L G ( 0 ; T e , T p h ) A = g d D 2 ∣ μ ∣ 4 π 2 ρ M L G ℏ 5 v F 3 v s 3 ∫ 0 ∞ d E E 4 [ n e + n p h + 2 n e n p h ] (30) \frac{S_{\mathrm{BG}}^{\mathrm{MLG}}(0;T_{\mathrm{e}},T_{\mathrm{ph}})}{A}
=
\frac{g_{\mathrm{d}}D^{2}|\mu|}{4\pi^{2}\rho_{\mathrm{MLG}}\hbar^{5}v_{\mathrm{F}}^{3}v_{\mathrm{s}}^{3}}
\int_{0}^{\infty} dE\,E^{4}\left[n_{\mathrm{e}}+n_{\mathrm{ph}}+2n_{\mathrm{e}}n_{\mathrm{ph}}\right] \tag*{(30)} A S BG MLG ( 0 ; T e , T ph ) = 4 π 2 ρ MLG ℏ 5 v F 3 v s 3 g d D 2 ∣ μ ∣ ∫ 0 ∞ d E E 4 [ n e + n ph + 2 n e n ph ] (30)
where g d = 4 g_{\mathrm{d}}=4 g d = 4 is the physical spin–valley degeneracy, n p h n_{\mathrm{ph}} n ph is the phonon occupation, whereas n e n_{\mathrm{e}} n e is an auxiliary Bose function generated by the electron phase-space integrals,
n i ( E ) = 1 exp [ E / ( k B T i ) ] − 1 , i = e , p h . (31) n_i(E)=\frac{1}{\exp[E/(k_{\mathrm{B}}T_i)]-1},\qquad i=\mathrm{e},\mathrm{ph}. \tag*{(31)} n i ( E ) = exp [ E / ( k B T i )] − 1 1 , i = e , ph . (31)
The BLG result contains the same energy integral, with the diffrent prefactor
S B G B L G ( 0 ; T e , T p h ) A = g d D 2 γ 1 16 π 2 ρ B L G ℏ 5 v F 3 v s 3 γ 1 ∣ μ ∣ ∫ 0 ∞ d E E 4 [ n e + n p h + 2 n e n p h ] . (32) \frac{S_{\mathrm{BG}}^{\mathrm{BLG}}(0;T_{\mathrm{e}},T_{\mathrm{ph}})}{A}
=
\frac{g_{\mathrm{d}}D^{2}\gamma_{1}}{16\pi^{2}\rho_{\mathrm{BLG}}\hbar^{5}v_{\mathrm{F}}^{3}v_{\mathrm{s}}^{3}}\sqrt{\frac{\gamma_{1}}{|\mu|}}
\int_{0}^{\infty} dE\,E^{4}\left[n_{\mathrm{e}}+n_{\mathrm{ph}}+2n_{\mathrm{e}}n_{\mathrm{ph}}\right]. \tag*{(32)} A S BG BLG ( 0 ; T e , T ph ) = 16 π 2 ρ BLG ℏ 5 v F 3 v s 3 g d D 2 γ 1 ∣ μ ∣ γ 1 ∫ 0 ∞ d E E 4 [ n e + n ph + 2 n e n ph ] . (32)
Thus the two temperatures enter separately through the electronic phase space and the phonon occupation. Prefactors in Eqs. (30) and (32) are consistent with the finite-frequency coefficients in Eqs. (37) and (38) of next section. We further parameterize the temperature imbalance by T e / p h = T ± Δ T / 2 T_{\mathrm{e/ph}}=T\pm\Delta T/2 T e/ph = T ± Δ T /2 and thus the noise is even in Δ T \Delta T Δ T . For ∣ Δ T ∣ ≪ T |\Delta T|\ll T ∣Δ T ∣ ≪ T , expansion at fixed T T T and μ \mu μ gives, for both MLG and BLG,
S B G l ( 0 ; T e , T p h ) S B G l ( 0 ; T , T ) = 1 + 5 π 2 21 ( Δ T T ) 2 , l = M L G , B L G . (33) \frac{S_{\mathrm{BG}}^{l}(0;T_{\mathrm{e}},T_{\mathrm{ph}})}{S_{\mathrm{BG}}^{l}(0;T,T)}
=
1+\frac{5\pi^{2}}{21}\left(\frac{\Delta T}{T}\right)^{2},\qquad l=\mathrm{MLG},\mathrm{BLG}. \tag*{(33)} S BG l ( 0 ; T , T ) S BG l ( 0 ; T e , T ph ) = 1 + 21 5 π 2 ( T Δ T ) 2 , l = MLG , BLG . (33)
The sub-leadning correction is propotional to ∝ ( Δ T / T ) 4 \propto(\Delta T/T)^{4} ∝ ( Δ T / T ) 4 .
In the EP regime, we instead use the classical-phonon and quasi-elastic reduction of Appendix C. The phonon occupation is evaluated at T p h T_{\mathrm{ph}} T ph , while the electronic Fermi window is evaluated at T e T_{\mathrm{e}} T e , i.e. n p h ( E q ) ≃ k B T p h / E q n_{\mathrm{ph}}(E_q)\simeq k_{\mathrm{B}}T_{\mathrm{ph}}/E_q n ph ( E q ) ≃ k B T ph / E q and f e ( 1 − f e ) = − k B T e ( ∂ f e / ∂ ϵ ) f_{\mathrm{e}}(1-f_{\mathrm{e}})=-k_{\mathrm{B}}T_{\mathrm{e}}(\partial f_{\mathrm{e}}/\partial\epsilon) f e ( 1 − f e ) = − k B T e ( ∂ f e / ∂ ϵ ) and consequently The microscopic emission-plus-absorption rate sum then yields
S E P l ( 0 ; T e , T p h ) A = 2 k B T e T p h g l ( μ , T e ) , (34) \frac{S_{\mathrm{EP}}^{l}(0;T_{\mathrm{e}},T_{\mathrm{ph}})}{A}
=
2k_{\mathrm{B}}T_{\mathrm{e}}T_{\mathrm{ph}}\,g_l(\mu,T_{\mathrm{e}}), \tag*{(34)} A S EP l ( 0 ; T e , T ph ) = 2 k B T e T ph g l ( μ , T e ) , (34)
where g l g_l g l are given in Eqs. (28) and (29) , respectively. The units of g l g_l g l are W m− 2 ^{-2} − 2 K− 1 ^{-1} − 1 . This expression is derived directly from the rate sum. At T e = T p h = T T_{\mathrm{e}}=T_{\mathrm{ph}}=T T e = T ph = T , it recovers the equilibrium identity S E P l ( 0 ) = 2 k B T 2 G E P l ( T ) S_{\mathrm{EP}}^{l}(0)=2k_{\mathrm{B}}T^{2}G_{\mathrm{EP}}^{l}(T) S EP l ( 0 ) = 2 k B T 2 G EP l ( T ) . Away from equilibrium, a thermal response must be defined from the mean heat current in Eq. (17) , with the held-fixed variables specified; it cannot be inferred from the noise alone.
Finite-frequency spectra
Finite frequency introduces the new energy scale ℏ ω \hbar\omega ℏ ω . The emission and absorption vertices in Eq. (15) become ℏ ω ∓ 2 E q \hbar\omega\mp2E_q ℏ ω ∓ 2 E q after the delta functions are imposed. The symmetrized spectrum is even in ω \omega ω , but its frequency dependence is not fixed by the zero-frequency conductance. We now give the reduced BG and EP spectra used in the numerical analysis.
Bloch–Grüneisen spectrum
In doped MLG the BG reduction keeps the intraband Fermi surface, linearizes the electron energy difference, and retains the full transferred phonon energy inside the Fermi functions. In this case, using the Eq. (1) one obtains the following equation for the finite-frequency noise
S B G l ( ω ) A = C B G l 2 ∫ 0 ∞ d E E [ B ( E , ℏ ω ) + B ( E , − ℏ ω ) ] (35) \frac{S_{\mathrm{BG}}^{l}(\omega)}{A}
=
\frac{C_{\mathrm{BG}}^{l}}{2}
\int_{0}^{\infty}dE\,E\left[\mathcal{B}(E,\hbar\omega)+\mathcal{B}(E,-\hbar\omega)\right] \tag*{(35)} A S BG l ( ω ) = 2 C BG l ∫ 0 ∞ d E E [ B ( E , ℏ ω ) + B ( E , − ℏ ω ) ] (35)
where l = M L G , B L G l=\mathrm{MLG},\mathrm{BLG} l = MLG , BLG and the frequency dependent part of integrand is given by
B ( E , ℏ ω ) = ( ℏ ω − 2 E ) 2 [ 1 + n p h ( E , T p h ) ] Φ e ( E − ℏ ω ; T e ) + ( ℏ ω + 2 E ) 2 n p h ( E , T p h ) Ψ e ( E + ℏ ω ; T e ) . (36) \mathcal{B}(E,\hbar\omega)
=
(\hbar\omega-2E)^{2}\left[1+n_{\mathrm{ph}}(E,T_{\mathrm{ph}})\right]\Phi_{\mathrm{e}}(E-\hbar\omega;T_{\mathrm{e}})
+
(\hbar\omega+2E)^{2}n_{\mathrm{ph}}(E,T_{\mathrm{ph}})\Psi_{\mathrm{e}}(E+\hbar\omega;T_{\mathrm{e}}). \tag*{(36)} B ( E , ℏ ω ) = ( ℏ ω − 2 E ) 2 [ 1 + n ph ( E , T ph ) ] Φ e ( E − ℏ ω ; T e ) + ( ℏ ω + 2 E ) 2 n ph ( E , T ph ) Ψ e ( E + ℏ ω ; T e ) . (36)
In the above Eq. (36) the thermal transition factors are related by the symmetry Φ e ( x ; T e ) = Ψ e ( − x ; T e ) \Phi_{\mathrm{e}}(x;T_{\mathrm{e}})=\Psi_{\mathrm{e}}(-x;T_{\mathrm{e}}) Φ e ( x ; T e ) = Ψ e ( − x ; T e ) and Φ e ( x ; T e ) = ∫ d ϵ f e ( ϵ + x ; T e ) [ 1 − f e ( ϵ ; T e ) ] = x / [ exp ( x / k B T e ) − 1 ] \Phi_{\mathrm{e}}(x;T_{\mathrm{e}})=\int d\epsilon\,f_{\mathrm{e}}(\epsilon+x;T_{\mathrm{e}})[1-f_{\mathrm{e}}(\epsilon;T_{\mathrm{e}})]=x/[\exp(x/k_{\mathrm{B}}T_{\mathrm{e}})-1] Φ e ( x ; T e ) = ∫ d ϵ f e ( ϵ + x ; T e ) [ 1 − f e ( ϵ ; T e )] = x / [ exp ( x / k B T e ) − 1 ] . The prefactors in front of the integral which correspond to MLG and BLG have the form
C B G M L G = g d D 2 ∣ μ ∣ 16 π 2 ρ M L G ℏ 5 v F 3 v s 3 . (37) C_{\mathrm{BG}}^{\mathrm{MLG}}
=
\frac{g_{\mathrm{d}}D^{2}|\mu|}{16\pi^{2}\rho_{\mathrm{MLG}}\hbar^{5}v_{\mathrm{F}}^{3}v_{\mathrm{s}}^{3}}. \tag*{(37)} C BG MLG = 16 π 2 ρ MLG ℏ 5 v F 3 v s 3 g d D 2 ∣ μ ∣ . (37)
Table I. Zero-frequency equilibrium power laws and conductances for physical MLG and BLG. All quantities are per area. See Eqs. (22) , (27) , (28) and (29) .
C B G B L G = g d D 2 γ 1 64 π 2 ρ B L G ℏ 5 v 0 3 v s 3 γ 1 ∣ μ ∣ (38) C_{\mathrm{BG}}^{\mathrm{BLG}}=\frac{g_{\mathrm{d}}D^{2}\gamma_{1}}{64\pi^{2}\rho_{\mathrm{BLG}}\hbar^{5}v_{0}^{3}v_{s}^{3}}\sqrt{\frac{\gamma_{1}}{|\mu|}} \tag*{(38)} C BG BLG = 64 π 2 ρ BLG ℏ 5 v 0 3 v s 3 g d D 2 γ 1 ∣ μ ∣ γ 1 (38)
(35) expression is evaluated without expanding in ℏ ω \hbar\omega ℏ ω in the “exact reduced” BG curves below.
At equilibrium and ∣ ℏ ω ∣ ≪ k B T |\hbar\omega|\ll k_{\mathrm{B}}T ∣ℏ ω ∣ ≪ k B T , Eq. (35) gives
S B G ( ω ; T , T ) S B G ( 0 ; T , T ) = 1 + ( 1 12 + 5 16 π 2 ) ( ℏ ω k B T ) 2 (39) \frac{S_{\mathrm{BG}}(\omega;T,T)}{S_{\mathrm{BG}}(0;T,T)}=1+\left(\frac{1}{12}+\frac{5}{16\pi^{2}}\right)\left(\frac{\hbar\omega}{k_{\mathrm{B}}T}\right)^{2} \tag*{(39)} S BG ( 0 ; T , T ) S BG ( ω ; T , T ) = 1 + ( 12 1 + 16 π 2 5 ) ( k B T ℏ ω ) 2 (39)
and the sub-leading correction is proportional to ( ℏ ω / k B T ) 4 (\hbar\omega/k_{\mathrm{B}}T)^{4} ( ℏ ω / k B T ) 4 . In the opposite quantum limit T e = T p h = T = 0 T_{\mathrm{e}}=T_{\mathrm{ph}}=T=0 T e = T ph = T = 0 , the same BG kernel gives
S B G L ( ω ; 0 , 0 ) A = C B G L 60 ∣ ℏ ω ∣ 5 , ∣ ℏ ω ∣ ≪ k B T B G , (40) \frac{S_{\mathrm{BG}}^{\mathrm{L}}(\omega;0,0)}{A}=\frac{C_{\mathrm{BG}}^{\mathrm{L}}}{60}|\hbar\omega|^{5},\qquad|\hbar\omega|\ll k_{\mathrm{B}}T_{\mathrm{BG}}, \tag*{(40)} A S BG L ( ω ; 0 , 0 ) = 60 C BG L ∣ℏ ω ∣ 5 , ∣ℏ ω ∣ ≪ k B T BG , (40)
showing explicitly that finite-frequency heat-current noise survives when the zero-frequency thermal noise vanishes. This is the graphene analogue of the quantum heat-noise effect discussed in Ref. [41 ] .
For the symmetric temperature bias, introducing the dimensionless variables r = Δ T / T r=\Delta T/T r = Δ T / T , w = ℏ ω / k B T w=\hbar\omega/k_{\mathrm{B}}T w = ℏ ω / k B T and x = E / k B T x=E/k_{\mathrm{B}}T x = E / k B T , the Eq. (35) results in
S B G L ( ω ; T e , T p h ) A = C B G L 2 ( k B T ) 5 × ∫ 0 ∞ d x x [ B r ( x , w ) + B r ( x , − w ) ] . (41) \frac{S_{\mathrm{BG}}^{\mathrm{L}}(\omega;T_{\mathrm{e}},T_{\mathrm{ph}})}{A}=\frac{C_{\mathrm{BG}}^{\mathrm{L}}}{2}(k_{\mathrm{B}}T)^{5}\times\int_{0}^{\infty}dx\,x\left[B_{r}(x,w)+B_{r}(x,-w)\right]. \tag*{(41)} A S BG L ( ω ; T e , T ph ) = 2 C BG L ( k B T ) 5 × ∫ 0 ∞ d x x [ B r ( x , w ) + B r ( x , − w ) ] . (41)
with the explicit form of function, B r ( x , w ) B_{r}(x,w) B r ( x , w ) , are given in Appendix D. Eq. (41) is the basis of the nonequilibrium BG ratio and heat-map calculations.
Equipartition spectrum
We obtain the finite-frequency EP spectra by applying the classical-phonon and quasi-elastic approximations to the general finite-frequency kernel in Eq. (15) . For the phonon modes contributing to the integral, n p h ( E q , T p h ) ≃ k B T p h / E q ≫ 1 n_{\mathrm{ph}}(E_{q},T_{\mathrm{ph}})\simeq k_{\mathrm{B}}T_{\mathrm{ph}}/E_{q}\gg1 n ph ( E q , T ph ) ≃ k B T ph / E q ≫ 1 , thus 1 + n p h ≃ n p h 1+n_{\mathrm{ph}}\simeq n_{\mathrm{ph}} 1 + n ph ≃ n ph . In the energy-conservation delta functions, the phonon energy is neglected relative to the electronic energy scales, i.e. δ ( ℏ ω − Δ ∓ E q ) ≃ δ ( ℏ ω − Δ ) \delta(\hbar\omega-\Delta\mp E_{q})\simeq\delta(\hbar\omega-\Delta) δ ( ℏ ω − Δ ∓ E q ) ≃ δ ( ℏ ω − Δ ) , where Δ \Delta Δ denotes the electronic energy difference from Eq. (13) . Thus, at finite frequency the electronic states remain separated by the probe energy ℏ ω \hbar\omega ℏ ω . The phonon energy must nevertheless be retained in the two current vertices. Using these approximations in the two shell contributions of Eq. (15) , the classical-phonon vertex sum becomes
n p h ( ℏ ω − 2 E q ) 2 + n p h ( ℏ ω + 2 E q ) 2 ≃ ( 2 k B T p h E q ) 2 [ ( ℏ ω ) 2 + 4 E q 2 ] . n_{\mathrm{ph}}(\hbar\omega-2E_{q})^{2}+n_{\mathrm{ph}}(\hbar\omega+2E_{q})^{2}\simeq\left(\frac{2k_{\mathrm{B}}T_{\mathrm{ph}}}{E_{q}}\right)^{2}\left[(\hbar\omega)^{2}+4E_{q}^{2}\right]. n ph ( ℏ ω − 2 E q ) 2 + n ph ( ℏ ω + 2 E q ) 2 ≃ ( E q 2 k B T ph ) 2 [ ( ℏ ω ) 2 + 4 E q 2 ] .
The terms linear in ℏ ω E q \hbar\omega E_{q} ℏ ω E q cancel between the two vertices. It is worth pointing out, that the term proportional to ( ℏ ω ) 2 (\hbar\omega)^{2} ( ℏ ω ) 2 vanishes at zero frequency and produces an additional dynamical contribution in case of finite frequency case, under the consideration. For MLG, substituting the above classical-phonon vertex sum into Eq. (15) , performing the angular integrals, and symmetrizing in frequency gives
S E P M L G ( ω ; T e , T p h ) A = D 2 k B T p h 2 π ρ M L G ℏ 5 v s 2 v F 4 × [ ( ℏ ω / 2 ) 2 I 2 ( ∣ ℏ ω ∣ ; T e ) + ( v s / v F ) 2 I 4 ( ∣ ℏ ω ∣ ; T e ) ] . (42) \frac{S_{\mathrm{EP}}^{\mathrm{MLG}}(\omega;T_{\mathrm{e}},T_{\mathrm{ph}})}{A}=\frac{D^{2}k_{\mathrm{B}}T_{\mathrm{ph}}}{2\pi\rho_{\mathrm{MLG}}\hbar^{5}v_{s}^{2}v_{\mathrm{F}}^{4}}\times\left[(\hbar\omega/2)^{2}\mathcal{I}_{2}(|\hbar\omega|;T_{\mathrm{e}})+(v_{s}/v_{\mathrm{F}})^{2}\mathcal{I}_{4}(|\hbar\omega|;T_{\mathrm{e}})\right]. \tag*{(42)} A S EP MLG ( ω ; T e , T ph ) = 2 π ρ MLG ℏ 5 v s 2 v F 4 D 2 k B T ph × [ ( ℏ ω /2 ) 2 I 2 ( ∣ℏ ω ∣ ; T e ) + ( v s / v F ) 2 I 4 ( ∣ℏ ω ∣ ; T e ) ] . (42)
The angular reduction leading to this expression is detailed in Appendix D. The electronic occupation factor entering the symmetrized spectrum is K ( ϵ ; ω , T e ) = f e ( ϵ ; T e ) [ 1 − f e ( ϵ + ℏ ω ; T e ) ] + f e ( ϵ + ℏ ω ; T e ) [ 1 − f e ( ϵ ; T e ) ] K(\epsilon;\omega,T_{\mathrm{e}})=f_{\mathrm{e}}(\epsilon;T_{\mathrm{e}})[1-f_{\mathrm{e}}(\epsilon+\hbar\omega;T_{\mathrm{e}})]+f_{\mathrm{e}}(\epsilon+\hbar\omega;T_{\mathrm{e}})[1-f_{\mathrm{e}}(\epsilon;T_{\mathrm{e}})] K ( ϵ ; ω , T e ) = f e ( ϵ ; T e ) [ 1 − f e ( ϵ + ℏ ω ; T e )] + f e ( ϵ + ℏ ω ; T e ) [ 1 − f e ( ϵ ; T e )] , where f e ( ϵ ; T e ) f_{\mathrm{e}}(\epsilon;T_{\mathrm{e}}) f e ( ϵ ; T e ) is the FD distribution function. This factor adds the occupation probabilities for transitions between electronic states of energies ϵ \epsilon ϵ and ϵ + ℏ ω \epsilon+\hbar\omega ϵ + ℏ ω in both directions. It follows from symmetrizing the electronic rate factors in Eq. (15) . The two electronic integrals appearing in Eq. (42) are
I 2 ( ℏ ω ; T e ) = ∫ d ϵ ∣ ϵ ∣ ∣ ϵ + ℏ ω ∣ Θ ( ϵ ( ϵ + ℏ ω ) ) K ( ϵ ; ω , T e ) , \mathcal{I}_{2}(\hbar\omega;T_{\mathrm{e}})=\int d\epsilon\,|\epsilon||\epsilon+\hbar\omega|\Theta(\epsilon(\epsilon+\hbar\omega))K(\epsilon;\omega,T_{\mathrm{e}}), I 2 ( ℏ ω ; T e ) = ∫ d ϵ ∣ ϵ ∣∣ ϵ + ℏ ω ∣Θ ( ϵ ( ϵ + ℏ ω )) K ( ϵ ; ω , T e ) ,
I 4 ( ℏ ω ; T e ) = ∫ d ϵ ∣ ϵ ∣ ∣ ϵ + ℏ ω ∣ Θ ( ϵ ( ϵ + ℏ ω ) ) K ( ϵ ; ω , T e ) × [ ( ϵ + ℏ ω ) 2 + ϵ 2 − ∣ ϵ ∣ ∣ ϵ + ℏ ω ∣ ] . (43) \mathcal{I}_{4}(\hbar\omega;T_{\mathrm{e}})=\int d\epsilon\,|\epsilon||\epsilon+\hbar\omega|\Theta(\epsilon(\epsilon+\hbar\omega))K(\epsilon;\omega,T_{\mathrm{e}})
\times\left[(\epsilon+\hbar\omega)^{2}+\epsilon^{2}-|\epsilon||\epsilon+\hbar\omega|\right]. \tag*{(43)} I 4 ( ℏ ω ; T e ) = ∫ d ϵ ∣ ϵ ∣∣ ϵ + ℏ ω ∣Θ ( ϵ ( ϵ + ℏ ω )) K ( ϵ ; ω , T e ) × [ ( ϵ + ℏ ω ) 2 + ϵ 2 − ∣ ϵ ∣∣ ϵ + ℏ ω ∣ ] . (43)
In Eq. (43) , the factors ∣ ϵ ∣ ∣ ϵ + ℏ ω ∣ |\epsilon||\epsilon+\hbar\omega| ∣ ϵ ∣∣ ϵ + ℏ ω ∣ arise from the two MLG radial density-of-states measures. The additional polynomial in Eq. (43) comes from the angular average of the momentum-dependent vertex and chirality factor. Consequently, I 2 \mathcal{I}_{2} I 2 weights the ( ℏ ω ) 2 (\hbar\omega)^{2} ( ℏ ω ) 2 contribution in classical-phonon vertex sum, whereas I 4 \mathcal{I}_{4} I 4 weights its 4 E q 2 4E_{q}^{2} 4 E q 2 contribution. The step function restricts both integrals to same-band intraband transitions. Both integrals are even in ω \omega ω , which allows their arguments in Eq. (42) to be written as ∣ ℏ ω ∣ |\hbar\omega| ∣ℏ ω ∣ .
For BLG, the same reduction of Eq. (15) uses the parabolic dispersion and BLG chirality factor, as detailed in Appendix D. The resulting spectrum is
S E P B L G ( ω ; T e , T p h ) A = D 2 γ 1 2 k B T p h 8 π ρ B L G ℏ 5 v s 2 v 0 4 × [ ( ℏ ω 2 ) 2 J 0 ( ∣ ℏ ω ∣ ; T e ) + ( γ 1 v s 2 v 0 2 ) J 1 ( ∣ ℏ ω ∣ ; T e ) ] . (44) \frac{S_{\mathrm{EP}}^{\mathrm{BLG}}(\omega;T_{\mathrm{e}},T_{\mathrm{ph}})}{A}=\frac{D^{2}\gamma_{1}^{2}k_{\mathrm{B}}T_{\mathrm{ph}}}{8\pi\rho_{\mathrm{BLG}}\hbar^{5}v_{\mathrm{s}}^{2}v_{0}^{4}}\times\left[\left(\frac{\hbar\omega}{2}\right)^{2}J_{0}(|\hbar\omega|;T_{\mathrm{e}})+\left(\frac{\gamma_{1}v_{\mathrm{s}}^{2}}{v_{0}^{2}}\right)J_{1}(|\hbar\omega|;T_{\mathrm{e}})\right]. \tag*{(44)} A S EP BLG ( ω ; T e , T ph ) = 8 π ρ BLG ℏ 5 v s 2 v 0 4 D 2 γ 1 2 k B T ph × [ ( 2 ℏ ω ) 2 J 0 ( ∣ℏ ω ∣ ; T e ) + ( v 0 2 γ 1 v s 2 ) J 1 ( ∣ℏ ω ∣ ; T e ) ] . (44)
The electronic integrals in this expression are
J 0 ( ℏ ω ; T e ) = ∫ d ϵ K ( ϵ ; ω ; T e ) Θ [ ϵ ( ϵ + ℏ ω ) ] , J_{0}(\hbar\omega;T_{\mathrm{e}})=\int d\epsilon\,K(\epsilon;\omega;T_{\mathrm{e}})\Theta[\epsilon(\epsilon+\hbar\omega)], J 0 ( ℏ ω ; T e ) = ∫ d ϵ K ( ϵ ; ω ; T e ) Θ [ ϵ ( ϵ + ℏ ω )] ,
J 1 ( ℏ ω ; T e ) = ∫ d ϵ [ ∣ ϵ ∣ + ∣ ϵ + ℏ ω ∣ ] K ( ϵ ; ω ; T e ) Θ [ ϵ ( ϵ + ℏ ω ) ] . (45) J_{1}(\hbar\omega;T_{\mathrm{e}})=\int d\epsilon[|\epsilon|+|\epsilon+\hbar\omega|]K(\epsilon;\omega;T_{\mathrm{e}})\Theta[\epsilon(\epsilon+\hbar\omega)]. \tag*{(45)} J 1 ( ℏ ω ; T e ) = ∫ d ϵ [ ∣ ϵ ∣ + ∣ ϵ + ℏ ω ∣ ] K ( ϵ ; ω ; T e ) Θ [ ϵ ( ϵ + ℏ ω )] . (45)
The constant radial density of states of the parabolic BLG bands gives the unweighted phase-space integral J 0 J_{0} J 0 in Eq. (45) . The angular average of the momentum-dependent vertex introduces the energy weight ∣ ϵ ∣ + ∣ ϵ + ℏ ω ∣ |\epsilon|+|\epsilon+\hbar\omega| ∣ ϵ ∣ + ∣ ϵ + ℏ ω ∣ in J 1 J_{1} J 1 . Thus, J 0 J_{0} J 0 and J 1 J_{1} J 1 describe, respectively, the ( ℏ ω ) 2 (\hbar\omega)^{2} ( ℏ ω ) 2 and 4 E q 2 4E_{q}^{2} 4 E q 2 terms of classical vertex sum. The step functions again correspond to the same-band intraband scattering.
The reduced EP curves are obtained by numerical evaluation of Eqs. (42) and (44) within the classical-phonon and quasi-elastic approximations. The low-frequency behavior of heat current noise for MLG in EP regime in the leading order of finite frequencies has quadratic form
S E P M L G ( ω ) S E P M L G ( 0 ) = 1 + ( ℏ ω ) 2 [ 1 12 ( k B T e ) 2 + M 2 M 4 ( 1 + v 0 2 4 v s 2 ) ] . (46) \frac{S_{\mathrm{EP}}^{\mathrm{MLG}}(\omega)}{S_{\mathrm{EP}}^{\mathrm{MLG}}(0)}=1+(\hbar\omega)^{2}\left[\frac{1}{12(k_{\mathrm{B}}T_{\mathrm{e}})^{2}}+\frac{M_{2}}{M_{4}}\left(1+\frac{v_{0}^{2}}{4v_{\mathrm{s}}^{2}}\right)\right]. \tag*{(46)} S EP MLG ( 0 ) S EP MLG ( ω ) = 1 + ( ℏ ω ) 2 [ 12 ( k B T e ) 2 1 + M 4 M 2 ( 1 + 4 v s 2 v 0 2 ) ] . (46)
where the Fermi-window moments follow from M j = ∫ − ∞ ∞ d ϵ ϵ j ( − ∂ f e / ∂ ϵ ) M_{j}=\int_{-\infty}^{\infty}d\epsilon\,\epsilon^{j}(-\partial f_{\mathrm{e}}/\partial\epsilon) M j = ∫ − ∞ ∞ d ϵ ϵ j ( − ∂ f e / ∂ ϵ ) for j = 2 , 4 j=2,4 j = 2 , 4 and have the form
M 2 ( μ , T e ) = μ 2 + π 2 3 ( k B T e ) 2 , (47) M_{2}(\mu,T_{\mathrm{e}})=\mu^{2}+\frac{\pi^{2}}{3}(k_{\mathrm{B}}T_{\mathrm{e}})^{2}, \tag*{(47)} M 2 ( μ , T e ) = μ 2 + 3 π 2 ( k B T e ) 2 , (47)
M 4 ( μ , T e ) = μ 4 + 2 π 2 μ 2 ( k B T e ) 2 + 7 π 4 15 ( k B T e ) 4 . (48) M_{4}(\mu,T_{\mathrm{e}})=\mu^{4}+2\pi^{2}\mu^{2}(k_{\mathrm{B}}T_{\mathrm{e}})^{2}+\frac{7\pi^{4}}{15}(k_{\mathrm{B}}T_{\mathrm{e}})^{4}. \tag*{(48)} M 4 ( μ , T e ) = μ 4 + 2 π 2 μ 2 ( k B T e ) 2 + 15 7 π 4 ( k B T e ) 4 . (48)
In the degenerate BLG EP limit, the low-frequency dependence is given by
S E P B L G ( ω ) S E P B L G ( 0 ) = 1 + ( ℏ ω ) 2 [ 1 12 ( k B T e ) 2 + 1 16 m v s 2 ∣ μ ∣ ] . (49) \frac{S_{\mathrm{EP}}^{\mathrm{BLG}}(\omega)}{S_{\mathrm{EP}}^{\mathrm{BLG}}(0)}=1+(\hbar\omega)^{2}\left[\frac{1}{12(k_{\mathrm{B}}T_{\mathrm{e}})^{2}}+\frac{1}{16mv_{\mathrm{s}}^{2}|\mu|}\right]. \tag*{(49)} S EP BLG ( 0 ) S EP BLG ( ω ) = 1 + ( ℏ ω ) 2 [ 12 ( k B T e ) 2 1 + 16 m v s 2 ∣ μ ∣ 1 ] . (49)
The sub-leading corrections to Eq. (46) and Eq. (49) proportional to ∝ ( ℏ ω ) 4 \propto(\hbar\omega)^{4} ∝ ( ℏ ω ) 4 . The detailed calculations are provided in Appendix D.
Full on-shell MLG benchmark
To connect the two limiting theories, we also evaluate the doped intraband MLG spectrum without the BG small-q q q or EP classical-phonon reductions. Momentum conservation gives p = k + q \mathbf{p}=\mathbf{k}+\mathbf{q} p = k + q . The angular integration over the energy-conservation delta function is performed analytically (see Appendix E). The resulting shell variables are
p λ s ( k , q ) = k + s v s v F q − λ ω v F , (50) p_{\lambda s}(k,q)=k+s\frac{v_{\mathrm{s}}}{v_{\mathrm{F}}}q-\lambda\frac{\omega}{v_{\mathrm{F}}}, \tag*{(50)} p λ s ( k , q ) = k + s v F v s q − λ v F ω , (50)
μ λ s ( k , q ) = p λ s 2 − k 2 − q 2 2 k q , (51) \mu_{\lambda s}(k,q)=\frac{p_{\lambda s}^{2}-k^{2}-q^{2}}{2kq}, \tag*{(51)} μ λ s ( k , q ) = 2 k q p λ s 2 − k 2 − q 2 , (51)
where λ = ± 1 \lambda=\pm1 λ = ± 1 labels the two frequency signs and s = + 1 s=+1 s = + 1 (− 1 -1 − 1 ) labels emission (absorption). The scattering event is allowed only if p λ s > 0 p_{\lambda s}>0 p λ s > 0 and ∣ μ λ s ∣ < 1 |\mu_{\lambda s}|<1 ∣ μ λ s ∣ < 1 . The exact chirality factor is
F + + = 1 + cos ϕ 2 , cos ϕ = k + q μ λ s p λ s . (52) F_{++}=\frac{1+\cos\phi}{2},\qquad\cos\phi=\frac{k+q\mu_{\lambda s}}{p_{\lambda s}}. \tag*{(52)} F ++ = 2 1 + cos ϕ , cos ϕ = p λ s k + q μ λ s . (52)
The resulting two-dimensional on-shell integral is written in Appendix E. It retains the exact Bose and Fermi functions and the vertex ( λ ℏ ω − 2 s E q ) 2 (\lambda\hbar\omega-2sE_{q})^{2} ( λ ℏ ω − 2 s E q ) 2 . We refer to this result as the “full on-shell” intraband calculation. This numerical calculation retains the exact intraband scattering geometry and distribution functions of the chosen doped MLG model. Its low- and high-temperature limits give the BG and EP reductions.
Numerical Results
We now evaluate the analytic formulas and the full doped-intraband MLG integral. The purpose of the numerical analysis is threefold. First, it verifies the zero-frequency FDT normalization and the Viljas–Heikkilä limiting laws [16 ] . Second, it determines the range over which the low-frequency expansions describe the full reduced spectra. Third, it shows how the non-equilibrium temperature bias changes the finite-frequency correction.
Unless noted otherwise, we use the parameters listed in Table II . The chosen deformation-potential value D D D is representative rather than universal; all conductances and heat-current noise spectra scale quadratically with D D D . For BLG, we take the areal mass density to be twice the monolayer value. The figures are presented in terms of the ordinary frequency f f f , while the analytical expressions use the angular frequency ω \omega ω . The corresponding probe energy may therefore be written equivalently as ℏ ω \hbar\omega ℏ ω or h f hf h f , with ω = 2 π f \omega=2\pi f ω = 2 π f . The numerical curves are evaluated from the zero-frequency BG and EP expressions in Eqs. (20) , (21) , (24) , (25) , (28) , and (29) , and from the reduced finite-frequency spectra in Eqs. (37) , (51) , and (54) . The full doped-intraband MLG benchmark is evaluated from Eq. (E6), while the quadratic approximations follow Eqs. (43) , (55), and (D18) given in Secs. V and VI.
Table II. Parameters used in the numerical figures. Material parameters follow representative graphene values discussed in Refs. [16 , 23 ] .
Figure 2 shows the equilibrium linear electron–phonon thermal conductance per unit area, G e p / A G_{\mathrm{ep}}/A G ep / A , for both MLG and BLG in the BG and EP limits. Here G e p ( T ) = ∂ Q / ∂ T ∣ T e = T p h = T G_{\mathrm{ep}}(T)=\left.\partial Q/\partial T\right|_{T_{\mathrm{e}}=T_{\mathrm{ph}}=T} G ep ( T ) = ∂ Q / ∂ T ∣ T e = T ph = T , as defined in Eq. (18) . The corresponding zero-frequency noise curves are constructed using the equilibrium relation S ( 0 ) / A = 2 k B T 2 [ G e p ( T ) / A ] S(0)/A=2k_{\mathrm{B}}T^{2}[G_{\mathrm{ep}}(T)/A] S ( 0 ) / A = 2 k B T 2 [ G ep ( T ) / A ] . The conductance expressions are asymptotic limits; their extrapolations outside the respective BG and EP regimes do not describe the full crossover. These curves reproduce the first-moment limiting laws of Ref. [16 ] and provide the equilibrium normalization of the second moment.
Figure 2. Equilibrium thermal conductance and zero-frequency heat-current noise in the BG and EP limits. The BG conductances follow Eqs. (21) and (25) ; the EP conductances follow Eqs. (28) and (29) . The temperature is normalized by the appropriate BG temperature of MLG or BLG. Dashed curves show the BG laws and solid curves the EP laws. The lower panels follow from the upper panels through S ( 0 ) = 2 k B T 2 G e p ( T ) S(0)=2k_{\mathrm{B}}T^2G_{\mathrm{ep}}(T) S ( 0 ) = 2 k B T 2 G ep ( T ) .
The limiting formulas do not by themselves describe the crossover. Figure 3 therefore evaluates the full zero-frequency MLG on-shell integral. The low-temperature numerical curve approaches the T 5 T^5 T 5 BG law, whereas the high-temperature curve approaches the EP result. The ratio panel gives a direct validity check, and the local logarithmic slope evolves from approximately 5 toward the degenerate EP value 2. The crossover is smooth and occurs over a broad interval around T B G T_{\mathrm{BG}} T BG , rather than at a sharp boundary.
Figure 3. Full zero-frequency MLG crossover. (a) Full on-shell noise compared with BG and EP limits obtained from Eq. (108) at ω = 0 \omega= 0 ω = 0 , compared with the BG expression, Eq. (20) , and the EP noise S E P ( 0 ) / A = 2 k B T 2 g M L G ( μ , T ) S_{\mathrm{EP}}(0)/A = 2 k_{\mathrm{B}} T^{2} g_{\mathrm{MLG}}(\mu,T) S EP ( 0 ) / A = 2 k B T 2 g MLG ( μ , T ) , with g M L G g_{\mathrm{MLG}} g MLG given by Eq. (28) . The BG and EP limits correspond to T ≪ T B G T \ll T_{\mathrm{BG}} T ≪ T BG and T ≫ T B G T \gg T_{\mathrm{BG}} T ≫ T BG , respectively. The BG noise scales as T 5 T^{5} T 5 , whereas the MLG EP noise is proportional to T 2 M 4 ( μ , T ) T^{2}M_{4}(\mu,T) T 2 M 4 ( μ , T ) , with M 4 M_{4} M 4 defined in Eq. (61). The latter reduces to T 2 T^{2} T 2 for k B T ≪ ∣ μ ∣ k_{\mathrm{B}}T \ll|\mu| k B T ≪ ∣ μ ∣ . Limiting curves outside their respective regimes are extrapolations. (b) Ratios S f u l l ( 0 ) / S l i m ( 0 ) S_{\mathrm{full}}(0)/S_{\mathrm{lim}}(0) S full ( 0 ) / S lim ( 0 ) , where S l i m ( 0 ) S_{\mathrm{lim}}(0) S lim ( 0 ) denotes the BG or EP limiting expression. (c) Local logarithmic slope d ln S / d ln T \mathrm{d}\ln S/\mathrm{d}\ln T d ln S / d ln T . The full result approaches the BG law at low temperature and the EP law at high temperature.
In order to investigate the exact curves it is practical to introduce the dimensionless observable [36 ]
R ( ω ; T e , T p h ) = S ( ω ; T e , T p h ) S ( 0 ; T e , T p h ) . (53) R(\omega; T_{\mathrm{e}}, T_{\mathrm{ph}}) = \frac{S(\omega; T_{\mathrm{e}}, T_{\mathrm{ph}})}{S(0; T_{\mathrm{e}}, T_{\mathrm{ph}})}. \tag*{(53)} R ( ω ; T e , T ph ) = S ( 0 ; T e , T ph ) S ( ω ; T e , T ph ) . (53)
It removes the overall D 2 D^2 D 2 and area dependence and highlights the dynamical part of the fluctuations. Figure 4 shows R R R for the exact reduced BG and EP spectra. The equilibrium and nonequilibrium curves are displayed together with the quadratic low-frequency approximation. At small frequency the exact and asymptotic curves coincide. Their separation at higher frequency is caused by the O ( ω 4 ) O(\omega^4) O ( ω 4 ) and higher terms omitted from the expansion.
Figure 4. Finite-frequency noise ratio R = S ( ω ) / S ( 0 ) R=S(\omega)/S(0) R = S ( ω ) / S ( 0 ) in the reduced BG and EP theories. The equilibrium curves use T e = T p h T_{\mathrm{e}}=T_{\mathrm{ph}} T e = T ph ; the nonequilibrium curves use the symmetric bias with Δ T / T = 0.4 \Delta T/T=0.4 Δ T / T = 0.4 . Dashed lines show the quadratic low-frequency approximations. Horizontal guides mark 1% and 10% corrections.
The BG ratio is the same for MLG and BLG within the leading Fermi-surface reduction because their different material factors cancel in Eq. (53) . The EP ratios differ because the two electronic dispersions give different energy moments. For the parameters of Table II , finite-frequency corrections are negligible at ordinary low-GHz frequencies, but reach the percent level in the high-GHz to sub-THz window. The characteristic frequency scale depends on the electronic and phononic temperature scales. In the BG regime, the normalized spectrum depends on both T e T_{\mathrm{e}} T e and T p h T_{\mathrm{ph}} T ph through the electronic phase-space and phonon-occupation factors in Eq. (38) . Within the classical-phonon EP reduction, Eqs. (51) and (54) contain an overall factor T p h T_{\mathrm{ph}} T ph , which cancels in R ( ω ) = S ( ω ) / S ( 0 ) R(\omega) = S(\omega)/S(0) R ( ω ) = S ( ω ) / S ( 0 ) . The normalized EP frequency dependence on T p h T_{\mathrm{ph}} T ph .
The non-equilibrium dependence is summarized by the percentage correction [36 ]
P ( f , Δ T / T ) = 100 [ R ( ω ; T e , T p h ) − 1 ] , (54) \mathcal{P}(f,\Delta T/T) = 100\left[R(\omega; T_{\mathrm{e}}, T_{\mathrm{ph}})-1\right], \tag*{(54)} P ( f , Δ T / T ) = 100 [ R ( ω ; T e , T ph ) − 1 ] , (54)
with symmetric temperature bias T e / p h = T ± Δ T / 2 T_{\mathrm{e/ph}} = T \pm\Delta T/2 T e/ph = T ± Δ T /2 . Figure 5 plots this quantity for MLG and BLG in both limiting regimes. The BG maps are obtained from Eq. (35) ; the EP maps use Eqs. (42) and (44) . These are direct non-equilibrium rate-sum calculations.
Figure 5. Nonequilibrium finite-frequency percentage correction P \mathcal{P} P from Eq. (54) . The horizontal axis is ordinary frequency and the vertical axis is the symmetric temperature bias. Contours identify 1%, 5%, and 10% corrections. The maps demonstrate that finite-frequency noise outside equilibrium must be evaluated directly rather than inferred from a conductance.
The frequency correction is generally not symmetric in Δ T \Delta T Δ T in the EP regime because the electronic phase-space factor is evaluated at T e T_{\mathrm{e}} T e , while the classical phonon amplitude is proportional to T p h T_{\mathrm{ph}} T ph . In the BG zero-frequency limit the symmetric rate sum is even in Δ T \Delta T Δ T , as shown by Eq. (33) , and at finite frequency the detailed curvature is determined by the shifted electron phase-space functions.
The strongest crossover test is obtained by evaluating the full doped-intraband MLG on-shell spectrum described in Sec. VI. This calculation keeps the exact triangle geometry, the MLG chirality factor, the Bose and Fermi distributions, and both frequency signs and phonon processes. It does not impose q ≪ k F q \ll k_{\mathrm{F}} q ≪ k F , replace the Bose function by its classical form, or make the scattering quasi-elastic.
Figure 6 compares the normalized full result with the reduced BG and EP spectra at T / T B G = 0.1 T/T_{\mathrm{BG}} = 0.1 T / T BG = 0.1 , 1, and 5, for equilibrium and for Δ T / T = 0.4 \Delta T/T = 0.4 Δ T / T = 0.4 . Deep in the BG regime the full curve follows the BG prediction. At high temperature it follows the EP prediction. Around T B G T_{\mathrm{BG}} T BG neither limiting expression is uniformly accurate, and the full integral is required. The same trend survives at finite temperature bias. Thus, the BG and EP formulas are controlled asymptotic limits of one microscopic on-shell spectrum rather than unrelated models.
Figure 6. Full doped-intraband MLG finite-frequency ratio with the BG and EP reductions from Eqs. (108) and (53) , compared with the reduced BG and EP ratios obtained from Eqs. (35) and (42) . The upper row is equilibrium; the lower row uses Δ T / T = 0.4 \Delta T/T=0.4 Δ T / T = 0.4 . Columns show T / T B G = 0.1 T/T_{\mathrm{BG}}=0.1 T / T BG = 0.1 , 1 1 1 , and 5 5 5 . The full result tends to BG at low temperature and to EP at high temperature, while the crossover requires the unreduced on-shell integral.
The noise spectral density per area has dimensions [ S / A ] = W 2 s m − 2 [S/A] = \mathrm{W}^2\,\mathrm{s\,m}^{-2} [ S / A ] = W 2 s m − 2 . For a device of area A d e v A_{\mathrm{dev}} A dev , the total spectrum is S t o t = A d e v ( S / A ) S_{\mathrm{tot}} = A_{\mathrm{dev}}(S/A) S tot = A dev ( S / A ) . Figure 7 gives both the per-area spectrum and the square-root device scale for A d e v = 1 μ m 2 A_{\mathrm{dev}} = 1\,\mu\mathrm{m}^2 A dev = 1 μ m 2 . The absolute values are small in SI units because individual phonon energies are microscopic and the BG noise is strongly suppressed as T 5 T^5 T 5 . Nevertheless, the corresponding amplitude can enter the femtowatt-per-root-hertz range in micron-scale devices, which is relevant to fast calorimetry and sensitive thermal detection [4 , 7 , 8 ] .
Figure 7. Absolute equilibrium zero-frequency noise. The BG curves follow Eqs. (20) and (24) . The EP curves are obtained from Eqs. (28) and (29) using S ( 0 ) / A = 2 k B T 2 [ G e p ( T ) / A ] S(0)/A = 2k_{\mathrm{B}}T^{2}[G_{\mathrm{ep}}(T)/A] S ( 0 ) / A = 2 k B T 2 [ G ep ( T ) / A ] . (a) Spectral density per area. (b) Square-root total noise for a device area A = 1 μ m 2 A = 1\,\mu\mathrm{m}^{2} A = 1 μ m 2 . The curves represent limiting BG and EP expressions; extensions outside their respective validity regimes are extrapolations.
Discussion
The average cooling power and the heat-current noise are two moments of the same microscopic energy-transfer process. The first moment reproduces the quasiequilibrium laws of Viljas and Heikkilä [16 ] ; the second moment contains additional information about fluctuations and associated time (energy) scales. This relation is particularly transparent at zero frequency. The current is proportional to the difference of emission and absorption rates, whereas the noise is proportional to their sum. Detailed balance therefore sets the average current to zero at equilibrium without suppressing the exchange noise.
The EP limits are controlled mainly by electronic phase space. The sound velocity cancels from the zero-frequency EP conductance. MLG has a fourth-order electronic moment and therefore a strong neutral-regime temperature dependence. BLG has a constant low-energy density of states and instead produces the weaker factor T ln [ 2 cosh ( μ / 2 k B T ) ] T\ln[2\cosh(\mu/2k_{\mathrm{B}}T)] T ln [ 2 cosh ( μ /2 k B T )] . These differences provide the possibility to distinguish mono- and bilayer cooling through their temperature and doping dependence.
The finite-frequency spectrum probes the correlation time of the energy exchange. The detector energy ℏ ω \hbar\omega ℏ ω enters both the energy-conservation shells and the current vertices. The resulting factors ( ℏ ω ∓ 2 E q ) 2 (\hbar\omega\mp2E_q)^2 ( ℏ ω ∓ 2 E q ) 2 are absent from the static cooling-power problem. They generate nonzero quantum noise even when the classical zero-frequency noise vanishes. In the BG limit, Eq. (40) gives S ( ω ) ∝ ∣ ω ∣ 5 S(\omega)\propto|\omega|^5 S ( ω ) ∝ ∣ ω ∣ 5 at zero temperature. This is consistent with the broader observation that finite-frequency heat noise contains quantum terms beyond the classical zero-frequency FDT [35 , 36 , 41 ] . For this reason, the equilibrium identity S ( 0 ) = 2 k B T 2 G e p ( T ) S(0)=2k_{\mathrm{B}}T^2G_{\mathrm{ep}}(T) S ( 0 ) = 2 k B T 2 G ep ( T ) should not be extended automatically to nonzero frequency. In particular, a temperature difference is not a microscopic dynamical force in the same way as an electric field. A carefully defined dynamic thermal response may be introduced for a specified protocol, but its relation to the full heat-current spectrum is not universal. The present work therefore treats the finite-frequency spectrum as the primary object and derives it directly from the current operator.
The reduced BG and EP spectra give useful analytic understanding. The full on-shell MLG calculation demonstrates where those approximations are valid and how they connect. The numerical comparison also shows that the finite-frequency crossover depends on both T / T B G T/T_{\mathrm{BG}} T / T BG and h f / k B T hf/k_{\mathrm{B}}T h f / k B T . A system may be in the static EP regime while still requiring the full frequency dependence when h f hf h f is no longer small compared with the electronic thermal scale.
When T e ≠ T p h T_{\mathrm{e}} \ne T_{\mathrm{ph}} T e = T ph , the electron and phonon factors are governed by different temperatures. In the BG regime the zero-frequency rate sum is n e + n p h + 2 n e n p h n_{\mathrm{e}} + n_{\mathrm{ph}} + 2n_{\mathrm{e}}n_{\mathrm{ph}} n e + n ph + 2 n e n ph . In the EP regime the classical phonon population supplies an overall T p h T_{\mathrm{ph}} T ph factor while the electronic moments are evaluated at T e T_{\mathrm{e}} T e . These structures explain why the noise is not determined by a single effective temperature.
A natural observable is the ratio R ( ω ) = S ( ω ) / S ( 0 ) R(\omega) = S(\omega)/S(0) R ( ω ) = S ( ω ) / S ( 0 ) , analogous to the normalized spectrum used in electron–phonon heat-noise studies of metals [36 ] . In the parameter range considered here, the correction is usually very small at a few gigahertz and becomes percent-level at tens to hundreds of gigahertz, with the largest relative correction occurring at low temperature. The required frequency range is challenging but is moving toward the bandwidth of radio-frequency thermometry and circuit-based calorimetry [7 , 8 ] . The absolute scale depends strongly on D 2 D^{2} D 2 , device area, chemical potential, and the phonon environment.
An experimental comparison should therefore determine or independently constrain the deformation potential and account for the actual mass loading and acoustic mode structure. The present results describe the intrinsic clean LA-phonon channel and provide a baseline against which additional cooling mechanisms can be identified.
It is worth mentioning, that the model omits several effects that can be important in specific samples. Screening modifies the small-q q q matrix element and may change BG power laws. Disorder-assisted supercollisions can dominate the cooling over an intermediate temperature range [25 , 26 , 27 ] . Suspended devices can couple to flexural modes, while supported devices can couple to substrate and remote surface phonons. Optical phonons dominate at sufficiently high temperature [16 ] . Electron–electron interactions are included only indirectly through the assumption of a local Fermi distribution. The full on-shell benchmark is presently restricted to the doped MLG intraband channel. A complete calculation near charge neutrality would require both bands and a consistent ultraviolet cutoff for the low-energy model. The analogous full BLG crossover should include the exact parabolic kinematics and, at sufficiently small momentum, possible interband processes. These extensions are numerically more demanding and are left for future work.
Conclusion
We developed a microscopic theory of electron–phonon heat-current noise in monolayer and bilayer graphene. It is worth mentioning that the equilibrium FDT identity,
S ( 0 ) = 2 k B T 2 G e p ( T ) S(0)=2k_{\mathrm{B}}T^{2}G_{\mathrm{ep}}(T) S ( 0 ) = 2 k B T 2 G ep ( T ) , can not be applied straightforwardly for finite frequencies as in case of electric current. To be more precise, the temperature difference is not a microscopic dynamical force in the same way as an electric field. Thus, the primary focus of present work is the finite-frequency spectrum which we calculated using the direct of heat current operator between electron and phonon subsystems.
Starting from the deformation-potential Hamiltonian, we derived a symmetric subsystem heat current operator and evaluated its leading current–current correlator. The direct Wick contraction vanishes because momentum conservation forces an acoustic phonon with zero momentum and zero energy. The exchange contraction produces the physical emission and absorption spectrum.
At zero frequency, both processes carry the same transferred-energy weight and must be retained. The resulting equilibrium spectra satisfy the thermodynamic relation S ( 0 ) = 2 k B T 2 G e p ( T ) S(0)=2k_{\mathrm{B}}T^{2}G_{\mathrm{ep}}(T) S ( 0 ) = 2 k B T 2 G ep ( T ) . This reproduces the Viljas–Heikkilä conductance laws from Ref. [16 ] and yields the noise scalings S M L G , B G ( 0 ) ∝ ∣ μ ∣ T 5 S^{\mathrm{MLG,BG}}(0)\propto|\mu|T^{5} S MLG , BG ( 0 ) ∝ ∣ μ ∣ T 5 and S B L G , B G ( 0 ) ∝ ∣ μ ∣ − 1 / 2 T 5 S^{\mathrm{BLG,BG}}(0)\propto|\mu|^{-1/2}T^{5} S BLG , BG ( 0 ) ∝ ∣ μ ∣ − 1/2 T 5 . The EP results reflect the different electronic phase spaces of MLG and BLG.
At finite frequency, the current vertices become ( ℏ ω ∓ 2 E q ) (\hbar\omega\mp2E_q) ( ℏ ω ∓ 2 E q ) . We derived reduced BG and EP spectra for separate electron and phonon temperatures and evaluated them without a low-frequency expansion. The corresponding quadratic asymptotics agree at small frequency with numerics. A full doped-intraband MLG on-shell calculation connects the two regimes and shows explicitly that the BG and EP spectra are the low- and high-temperature limits of one microscopic result.
Out of equilibrium, the noise is a rate sum governed separately by T e T_{\mathrm{e}} T e and T p h T_{\mathrm{ph}} T ph . It is not fixed by FDT and should not be interpreted as a unique nonlinear thermal conductance. The finite-frequency ratios and heat maps proposed here provide direct signatures of this nonequilibrium energy-exchange statistics.
The present theory supplies an intrinsic acoustic-phonon baseline for graphene calorimetry and thermal-noise spectroscopy.
ACKNOWLEDGMENTS
This work is supported by United Arab Emirates University under the Startup Grant No. G00004974, UPAR Grant G00005479 and UAEU-AUA Grant G00005522.
Heat-current operator and Wick expansion
This appendix gives the algebra behind Eqs. (14) and (15) .
Subsystem energy derivatives
Only H ^ e − p h \hat{H}_{\mathrm{e-ph}} H ^ e − ph contributes to the subsystem energy derivatives. The fermionic commutator is [ c ^ p β † c ^ k α , H ^ e ] = ( ϵ k α − ϵ p β ) c ^ p β † c ^ k α [\hat{c}^{\dagger}_{p\beta}\hat{c}_{k\alpha},\hat{H}_{\mathrm{e}}] = (\epsilon_{k\alpha}-\epsilon_{p\beta})\hat{c}^{\dagger}_{p\beta}\hat{c}_{k\alpha} [ c ^ pβ † c ^ k α , H ^ e ] = ( ϵ k α − ϵ pβ ) c ^ pβ † c ^ k α and the bosonic commutators are [ b ^ q , H ^ p h ] = E q b ^ q [\hat{b}_{q},\hat{H}_{\mathrm{ph}}] = E_{q}\hat{b}_{q} [ b ^ q , H ^ ph ] = E q b ^ q , [ b ^ − q † , H ^ p h ] = − E q b ^ − q † [\hat{b}^{\dagger}_{-q},\hat{H}_{\mathrm{ph}}] = -E_{q}\hat{b}^{\dagger}_{-q} [ b ^ − q † , H ^ ph ] = − E q b ^ − q † . Substitution into Eq. (10) gives Eq. (14) from main text.
Free evolution and contractions
To second order in the electron-phonon matrix element, all operators may be evolved with H ^ 0 = H ^ e + H ^ p h \hat{H}_0=\hat{H}_{\mathrm{e}}+\hat{H}_{\mathrm{ph}} H ^ 0 = H ^ e + H ^ ph ,
c ^ k α ( t ) = c ^ k α e − i ϵ k α t / ℏ , c ^ k α † ( t ) = c ^ k α † e i ϵ k α t / ℏ (55) \hat{c}_{k\alpha}(t)=\hat{c}_{k\alpha}e^{-i\epsilon_{k\alpha}t/\hbar},\qquad\hat{c}^{\dagger}_{k\alpha}(t)=\hat{c}^{\dagger}_{k\alpha}e^{i\epsilon_{k\alpha}t/\hbar} \tag*{(55)} c ^ k α ( t ) = c ^ k α e − i ϵ k α t /ℏ , c ^ k α † ( t ) = c ^ k α † e i ϵ k α t /ℏ (55)
b ^ q ( t ) = b ^ q e − i E q t / ℏ , b ^ − q † ( t ) = b ^ − q † e i E q t / ℏ . (56) \hat{b}_{q}(t)=\hat{b}_{q}e^{-iE_{q}t/\hbar},\qquad\hat{b}^{\dagger}_{-q}(t)=\hat{b}^{\dagger}_{-q}e^{iE_{q}t/\hbar}. \tag*{(56)} b ^ q ( t ) = b ^ q e − i E q t /ℏ , b ^ − q † ( t ) = b ^ − q † e i E q t /ℏ . (56)
The nonzero phonon contractions are
⟨ b ^ q ( t ) b ^ − q ′ † ( 0 ) ⟩ = [ 1 + n q ] e − i E q t / ℏ δ q , − q ′ (57) \langle\hat{b}_{q}(t)\hat{b}^{\dagger}_{-q'}(0)\rangle=[1+n_q]e^{-iE_qt/\hbar}\delta_{q,-q'} \tag*{(57)} ⟨ b ^ q ( t ) b ^ − q ′ † ( 0 )⟩ = [ 1 + n q ] e − i E q t /ℏ δ q , − q ′ (57)
⟨ b ^ − q † ( t ) b ^ q ′ ( 0 ) ⟩ = n q e i E q t / ℏ δ q , − q ′ . (58) \langle\hat{b}^{\dagger}_{-q}(t)\hat{b}_{q'}(0)\rangle=n_qe^{iE_qt/\hbar}\delta_{q,-q'}. \tag*{(58)} ⟨ b ^ − q † ( t ) b ^ q ′ ( 0 )⟩ = n q e i E q t /ℏ δ q , − q ′ . (58)
The electron four-point correlation function is
⟨ c ^ p β † c ^ k α c ^ p ′ β ′ † c ^ k ′ α ′ ⟩ = f p β f p ′ β ′ δ p k δ β α δ p ′ k ′ δ β ′ α ′ + f p β [ 1 − f k α ] δ p k ′ δ β α ′ δ k p ′ δ α β ′ . (59) \langle\hat{c}^{\dagger}_{p\beta}\hat{c}_{k\alpha}\hat{c}^{\dagger}_{p'\beta'}\hat{c}_{k'\alpha'}\rangle
=f_{p\beta}f_{p'\beta'}\delta_{pk}\delta_{\beta\alpha}\delta_{p'k'}\delta_{\beta'\alpha'}+f_{p\beta}[1-f_{k\alpha}]\delta_{pk'}\delta_{\beta\alpha'}\delta_{kp'}\delta_{\alpha\beta'}. \tag*{(59)} ⟨ c ^ pβ † c ^ k α c ^ p ′ β ′ † c ^ k ′ α ′ ⟩ = f pβ f p ′ β ′ δ p k δ β α δ p ′ k ′ δ β ′ α ′ + f pβ [ 1 − f k α ] δ p k ′ δ β α ′ δ k p ′ δ α β ′ . (59)
The first term is the direct contraction. It imposes p = k p=k p = k , and translational invariance then imposes q = 0 q=0 q = 0 . Because E q = 0 = 0 E_{q=0}=0 E q = 0 = 0 and Δ k k = 0 \Delta_{kk}=0 Δ k k = 0 , both current vertices vanish. Thus the direct term is zero before any frequency integration.
The exchange contraction in Eq. (59) pairs the creation operator in each electronic bilinear with the annihilation operator in the other. Its Kronecker deltas impose k ′ = p k'=p k ′ = p , p ′ = k p'=k p ′ = k , α ′ = β \alpha'=\beta α ′ = β , and β ′ = α \beta'=\alpha β ′ = α . Thus, the two current vertices connect the same pair of electronic states in opposite directions. Momentum conservation at the vertices then gives q ′ = − q q'=-q q ′ = − q .
Hermiticity of the electron–phonon interaction implies
M k p , − q α β = ( M p k , q β α ) ∗ , M^{\alpha\beta}_{kp,-q}=\left(M^{\beta\alpha}_{pk,q}\right)^*, M k p , − q α β = ( M p k , q β α ) ∗ ,
which means that the matrix element for the reverse transition is the complex conjugate of that for the forward transition. Consequently, the product of matrix elements from the two current vertices becomes
M p k , q β α M k p , − q α β = ∣ M p k , q β α ∣ 2 . (60) M^{\beta\alpha}_{pk,q}M^{\alpha\beta}_{kp,-q}=\left|M^{\beta\alpha}_{pk,q}\right|^2. \tag*{(60)} M p k , q β α M k p , − q α β = M p k , q β α 2 . (60)
The exchange contribution therefore weights each allowed electronic transition by its squared coupling amplitude and the occupation factor f p β [ 1 − f k α ] f_{p\beta}[1-f_{k\alpha}] f pβ [ 1 − f k α ] . This is the transition weight entering the emission and absorption terms of Eq. (15) . Using the Fourier identity ∫ d t exp [ i ( ℏ ω − X ) t / ℏ ] = 2 π ℏ δ ( ℏ ω − X ) \int dt\,\exp[i(\hbar\omega-X)t/\hbar] = 2\pi\hbar\,\delta(\hbar\omega-X) ∫ d t exp [ i ( ℏ ω − X ) t /ℏ ] = 2 π ℏ δ ( ℏ ω − X ) with X = Δ k p α β ± E q X = \Delta^{\alpha\beta}_{\mathbf{k p}} \pm E_q X = Δ kp α β ± E q and combining the electronic evolution with the two phonon contractions gives the time-dependent factors exp [ − i ( Δ k p α β + E q ) t / ℏ ] \exp[-i(\Delta^{\alpha\beta}_{\mathbf{k p}}+E_q)t/\hbar] exp [ − i ( Δ kp α β + E q ) t /ℏ ] and exp [ − i ( Δ k p α β − E q ) t / ℏ ] \exp[-i(\Delta^{\alpha\beta}_{\mathbf{k p}}-E_q)t/\hbar] exp [ − i ( Δ kp α β − E q ) t /ℏ ] and therefore results δ ( ℏ ω − Δ k p α β − E q ) \delta(\hbar\omega-\Delta^{\alpha\beta}_{\mathbf{k p}}-E_q) δ ( ℏ ω − Δ kp α β − E q ) and δ ( ℏ ω − Δ k p α β + E q ) \delta(\hbar\omega-\Delta^{\alpha\beta}_{\mathbf{k p}}+E_q) δ ( ℏ ω − Δ kp α β + E q ) , respectively, yielding the two energy-conservation delta functions in Eq. (15) from main text.
Zero-frequency shell weights
The two energy-conservation delta functions derived above determine the current-vertex weights in the zero-frequency limit. Setting ω = 0 \omega= 0 ω = 0 in Eq. (15) gives
δ ( − Δ − E q ) ⇒ Δ = − E q , ( Δ − E q ) 2 = 4 E q 2 , (61) \delta(-\Delta-E_q)\Rightarrow\Delta=-E_q,\qquad(\Delta-E_q)^2=4E_q^2, \tag*{(61)} δ ( − Δ − E q ) ⇒ Δ = − E q , ( Δ − E q ) 2 = 4 E q 2 , (61)
δ ( − Δ + E q ) ⇒ Δ = E q , ( Δ + E q ) 2 = 4 E q 2 , (62) \delta(-\Delta+E_q)\Rightarrow\Delta=E_q,\qquad(\Delta+E_q)^2=4E_q^2, \tag*{(62)} δ ( − Δ + E q ) ⇒ Δ = E q , ( Δ + E q ) 2 = 4 E q 2 , (62)
where Δ ≡ Δ k p α β \Delta\equiv\Delta^{\alpha\beta}_{\mathbf{k p}} Δ ≡ Δ kp α β . Thus, both phonon emission and absorption acquire the same squared current-vertex weight, 4 E q 2 4E_q^2 4 E q 2 , although their occupation factors generally differ. Multiplying this weight by the prefactor π / ( 2 ℏ ) \pi/(2\hbar) π / ( 2ℏ ) in Eq. (15) gives 2 π / ℏ 2\pi/\hbar 2 π /ℏ . Since the symmetrized and unsymmetrized spectra coincide at ω = 0 \omega=0 ω = 0 , the resulting expression is the emission–absorption rate sum in Eq. (16) . At equilibrium and finite temperature, both rates contribute to the noise; omitting absorption would give an incomplete rate sum and violate the zero-frequency FDT.
Bloch–Grüneisen derivations
Monolayer graphene
For a doped conduction band, let θ \theta θ denote the angle between the electron wave vector k \mathbf{k} k and the phonon wave vector q \mathbf{q} q . In the BG regime q ≪ k F q\ll k_{\mathrm{F}} q ≪ k F , thus
ϵ k − ϵ k − q ≃ ℏ v F q cos θ , F + + M L G ≃ 1. (63) \epsilon_{\mathbf{k}}-\epsilon_{\mathbf{k-q}}\simeq\hbar v_{\mathrm{F}}q\cos\theta,\qquad F^{\mathrm{MLG}}_{++}\simeq1. \tag*{(63)} ϵ k − ϵ k − q ≃ ℏ v F q cos θ , F ++ MLG ≃ 1. (63)
The angular delta function is
∫ 0 2 π d θ δ [ ℏ q ( v F cos θ − v s ) ] = 2 ℏ v F q 1 − ( v s / v F ) 2 ≃ 2 ℏ v F q . (64) \int_{0}^{2\pi}d\theta\,\delta[\hbar q(v_{\mathrm{F}}\cos\theta-v_{\mathrm{s}})]
=\frac{2}{\hbar v_{\mathrm{F}}q\sqrt{1-(v_{\mathrm{s}}/v_{\mathrm{F}})^2}}
\simeq\frac{2}{\hbar v_{\mathrm{F}}q}. \tag*{(64)} ∫ 0 2 π d θ δ [ ℏ q ( v F cos θ − v s )] = ℏ v F q 1 − ( v s / v F ) 2 2 ≃ ℏ v F q 2 . (64)
After converting the momentum sums to integrals, Eq. (16) becomes
S B G M L G ( 0 ) = g d A D 2 ℏ v s 4 π ρ M L G v F ∫ 0 ∞ d k k ∫ 0 ∞ d q q 3 F ( k , q ) , (65) S_{\mathrm{BG}}^{\mathrm{MLG}}(0)=\frac{g_{\mathrm{d}}AD^2\hbar v_{\mathrm{s}}}{4\pi\rho_{\mathrm{MLG}}v_{\mathrm{F}}}\int_{0}^{\infty}dk\,k\int_{0}^{\infty}dq\,q^3\mathcal{F}(k,q), \tag*{(65)} S BG MLG ( 0 ) = 4 π ρ MLG v F g d A D 2 ℏ v s ∫ 0 ∞ d k k ∫ 0 ∞ d q q 3 F ( k , q ) , (65)
where F \mathcal{F} F contains the two Fermi–Bose rate factors
F ( k , q ) = f e ( ϵ k ) [ 1 − f e ( ϵ k − E q ) ] ( 1 + n q ) + f e ( ϵ k − E q ) [ 1 − f e ( ϵ k ) ] n q , (66) \mathcal{F}(k,q)=f_{\mathrm{e}}(\epsilon_k)[1-f_{\mathrm{e}}(\epsilon_k-E_q)](1+n_q)
+f_{\mathrm{e}}(\epsilon_k-E_q)[1-f_{\mathrm{e}}(\epsilon_k)]n_q, \tag*{(66)} F ( k , q ) = f e ( ϵ k ) [ 1 − f e ( ϵ k − E q )] ( 1 + n q ) + f e ( ϵ k − E q ) [ 1 − f e ( ϵ k )] n q , (66)
with E q = ℏ v s q E_q=\hbar v_{\mathrm{s}}q E q = ℏ v s q , f e ( ϵ ) = [ exp [ ( ϵ − μ ) / ( k B T e ) ] + 1 ] − 1 f_{\mathrm{e}}(\epsilon)=[\exp[(\epsilon-\mu)/(k_{\mathrm{B}}T_{\mathrm{e}})]+1]^{-1} f e ( ϵ ) = [ exp [( ϵ − μ ) / ( k B T e )] + 1 ] − 1 , and n q = [ exp [ E q / ( k B T p h ) ] − 1 ] − 1 n_q=[\exp[E_q/(k_{\mathrm{B}}T_{\mathrm{ph}})]-1]^{-1} n q = [ exp [ E q / ( k B T ph )] − 1 ] − 1 . The first term describes phonon emission and the second describes the reverse absorption process.
Near the Fermi surface, k d k ≃ ( k F / ℏ v F ) d ϵ k\,dk\simeq(k_{\mathrm{F}}/\hbar v_{\mathrm{F}})\,d\epsilon k d k ≃ ( k F /ℏ v F ) d ϵ . Thus, keeping the finite phonon energy E q E_q E q inside the Fermi functions gives
∫ d ϵ f e ( ϵ + E q ) [ 1 − f e ( ϵ ) ] = E q n B ( E q ) . (67) \int d\epsilon\, f_e(\epsilon+E_q)[1-f_e(\epsilon)] = E_q n_B(E_q). \tag*{(67)} ∫ d ϵ f e ( ϵ + E q ) [ 1 − f e ( ϵ )] = E q n B ( E q ) . (67)
∫ d ϵ f e ( ϵ ) [ 1 − f e ( ϵ + E q ) ] = E q [ 1 + n B ( E q ) ] . (68) \int d\epsilon\, f_e(\epsilon)[1-f_e(\epsilon+E_q)] = E_q[1+n_B(E_q)]. \tag*{(68)} ∫ d ϵ f e ( ϵ ) [ 1 − f e ( ϵ + E q )] = E q [ 1 + n B ( E q )] . (68)
Hence ∫ d ϵ d F = 2 E q n q ( 1 + n q ) \int d\epsilon\,d\mathcal{F}=2E_qn_q(1+n_q) ∫ d ϵ d F = 2 E q n q ( 1 + n q ) and using E q = ℏ v s q E_q=\hbar v_s q E q = ℏ v s q in Eq. (65) gives
S B G M L G ( 0 ) = g d A D 2 k F ℏ v s 2 2 π 2 ρ M L G v F 2 ∫ 0 ∞ d q q 4 n q ( 1 + n q ) . (69) S_{\mathrm{BG}}^{\mathrm{MLG}}(0)=\frac{g_dAD^2k_F\hbar v_s^2}{2\pi^2\rho_{\mathrm{MLG}}v_F^2}\int_0^\infty dq\,q^4n_q(1+n_q). \tag*{(69)} S BG MLG ( 0 ) = 2 π 2 ρ MLG v F 2 g d A D 2 k F ℏ v s 2 ∫ 0 ∞ d q q 4 n q ( 1 + n q ) . (69)
After performing the integration by parts we have
∫ 0 ∞ d q q 4 n q ( 1 + n q ) = 4 π 4 15 ( k B T ℏ v s ) 5 . (70) \int_0^\infty dq\,q^4n_q(1+n_q)=\frac{4\pi^4}{15}\left(\frac{k_BT}{\hbar v_s}\right)^5. \tag*{(70)} ∫ 0 ∞ d q q 4 n q ( 1 + n q ) = 15 4 π 4 ( ℏ v s k B T ) 5 . (70)
Thus, the final result is
S B G M L G ( 0 ) = 2 g d π 2 A D 2 ∣ μ ∣ k B 5 T 5 15 ρ M L G ℏ 5 v F 3 v s 3 , (71) S_{\mathrm{BG}}^{\mathrm{MLG}}(0)=\frac{2g_d\pi^2AD^2|\mu|k_B^5T^5}{15\rho_{\mathrm{MLG}}\hbar^5v_F^3v_s^3}, \tag*{(71)} S BG MLG ( 0 ) = 15 ρ MLG ℏ 5 v F 3 v s 3 2 g d π 2 A D 2 ∣ μ ∣ k B 5 T 5 , (71)
which reduces to Eq. (20) for g d = 4 g_d=4 g d = 4 .
Bilayer graphene
For BLG, the spectrum and Fermi velocity ϵ k = ℏ 2 k 2 / 2 m \epsilon_k=\hbar^2k^2/2m ϵ k = ℏ 2 k 2 /2 m and u F = ℏ k F / m u_F=\hbar k_F/m u F = ℏ k F / m , thus at small q q q ,
ϵ k − ϵ k − q = ℏ 2 2 m ( 2 k q cos θ − q 2 ) ≃ ℏ u F q cos θ , (72) \epsilon_k-\epsilon_{k-q}=\frac{\hbar^2}{2m}(2kq\cos\theta-q^2)\simeq\hbar u_Fq\cos\theta, \tag*{(72)} ϵ k − ϵ k − q = 2 m ℏ 2 ( 2 k q cos θ − q 2 ) ≃ ℏ u F q cos θ , (72)
and F + + B L G = cos 2 ϕ ≃ 1 F_{++}^{\mathrm{BLG}}=\cos^2\phi\simeq1 F ++ BLG = cos 2 ϕ ≃ 1 . The angular factor is therefore 2 / ( ℏ u F q ) 2/(\hbar u_Fq) 2/ ( ℏ u F q ) and the radial measure is k d k = ( m / ℏ 2 ) d ϵ k\,dk=(m/\hbar^2)d\epsilon k d k = ( m / ℏ 2 ) d ϵ . Repeating the MLG steps gives
S B G B L G ( 0 ) = 2 g d π 2 A D 2 m 15 ρ B L G u F k B 5 T 5 ℏ 5 v s 3 . (73) S_{\mathrm{BG}}^{\mathrm{BLG}}(0)=\frac{2g_d\pi^2AD^2m}{15\rho_{\mathrm{BLG}}u_F}\frac{k_B^5T^5}{\hbar^5v_s^3}. \tag*{(73)} S BG BLG ( 0 ) = 15 ρ BLG u F 2 g d π 2 A D 2 m ℏ 5 v s 3 k B 5 T 5 . (73)
Since m / u F = ( γ 1 / 4 v 0 2 ) γ 1 / ∣ μ ∣ m/u_F=(\gamma_1/4v_0^2)\sqrt{\gamma_1/|\mu|} m / u F = ( γ 1 /4 v 0 2 ) γ 1 /∣ μ ∣ , we obtain Eq. (24) for g d = 4 g_d=4 g d = 4 .
Nonequilibrium Bose integral
For T e ≠ T p h T_e\ne T_{\mathrm{ph}} T e = T ph we evaluate the electronic energy integrals in the emission–absorption rate sum of Eq. (16) , using the BG Fermi-surface reduction described above. The emission integral gives E n e ( E ) E n_e(E) E n e ( E ) , whereas the absorption integral gives E [ 1 + n e ( E ) ] E[1+n_e(E)] E [ 1 + n e ( E )] . Including the corresponding phonon occupation factors therefore yields E [ n e ( 1 + n p h ) + ( 1 + n e ) n p h ] = E [ n e + n p h + 2 n e n p h ] E[n_e(1+n_{\mathrm{ph}})+(1+n_e)n_{\mathrm{ph}}]=E[n_e+n_{\mathrm{ph}}+2n_en_{\mathrm{ph}}] E [ n e ( 1 + n ph ) + ( 1 + n e ) n ph ] = E [ n e + n ph + 2 n e n ph ] and using the symmetric temperatures T e = T ( 1 + r ) T_e=T(1+r) T e = T ( 1 + r ) and T p h = T ( 1 − r ) T_{\mathrm{ph}}=T(1-r) T ph = T ( 1 − r ) , with r = Δ T / ( 2 T ) r=\Delta T/(2T) r = Δ T / ( 2 T ) , and introducing x = E / k B T x=E/k_BT x = E / k B T , we define the dimensionless phonon-energy integral
L B G ( r ) = ∫ 0 ∞ d x x 4 [ n ( x 1 + r ) + n ( x 1 − r ) + 2 n ( x 1 + r ) n ( x 1 − r ) ] . (74) \mathcal{L}_{\mathrm{BG}}(r)=\int_0^\infty dx\,x^4\left[n\left(\frac{x}{1+r}\right)+n\left(\frac{x}{1-r}\right)+2n\left(\frac{x}{1+r}\right)n\left(\frac{x}{1-r}\right)\right]. \tag*{(74)} L BG ( r ) = ∫ 0 ∞ d x x 4 [ n ( 1 + r x ) + n ( 1 − r x ) + 2 n ( 1 + r x ) n ( 1 − r x ) ] . (74)
where n ( y ) = ( e y − 1 ) − 1 n(y)=(e^y-1)^{-1} n ( y ) = ( e y − 1 ) − 1 . Expanding with the Bose series n ( y ) = ∑ l ≥ 1 e − l y n(y)=\sum_{l\geq1}e^{-ly} n ( y ) = ∑ l ≥ 1 e − l y gives
L B G ( r ) = 8 π 4 15 + 32 π 6 63 r 2 + O ( r 4 ) . (75) \mathcal{L}_{\mathrm{BG}}(r)=\frac{8\pi^4}{15}+\frac{32\pi^6}{63}r^2+O(r^4). \tag*{(75)} L BG ( r ) = 15 8 π 4 + 63 32 π 6 r 2 + O ( r 4 ) . (75)
Within the BG approximation, the material-dependent prefactor cancels when the nonequilibrium noise is normalized to its equilibrium value at the same mean temperature T T T and fixed chemical potential:
S B G ( 0 ; T e , T p h ) S B G ( 0 ; T , T ) = L B G ( r ) L B G ( 0 ) . (76) \frac{S_{\mathrm{BG}}(0;T_e,T_{\mathrm{ph}})}{S_{\mathrm{BG}}(0;T,T)}=\frac{\mathcal{L}_{\mathrm{BG}}(r)}{\mathcal{L}_{\mathrm{BG}}(0)}. \tag*{(76)} S BG ( 0 ; T , T ) S BG ( 0 ; T e , T ph ) = L BG ( 0 ) L BG ( r ) . (76)
Using L B G ( 0 ) = 8 π 4 / 15 \mathcal{L}_{\mathrm{BG}}(0)=8\pi^4/15 L BG ( 0 ) = 8 π 4 /15 and r = Δ T / ( 2 T ) r=\Delta T/(2T) r = Δ T / ( 2 T ) recovers Eq. (33) in the main text. The absence of a linear term follows from the invariance of the BG rate sum under T e ↔ T p h T_e\leftrightarrow T_{\mathrm{ph}} T e ↔ T ph .
Equipartition derivations
Common quasi-elastic reduction
Starting from the zero-frequency rate sum in Eq. (16) , we consider equilibrium, T e = T p h = T T_e=T_{\mathrm{ph}}=T T e = T ph = T , in the EP regime. The phonon occupations satisfy n q ≃ k B T / E q ≫ 1 n_q\simeq k_BT/E_q\gg1 n q ≃ k B T / E q ≫ 1 and 1 + n q ≃ n q 1+n_q\simeq n_q 1 + n q ≃ n q . In the quasi-elastic approximation, we neglect E q E_q E q in both electronic energy-conservation delta functions, while retaining it in the heat-current weights:
δ ( ϵ k α − ϵ p α ± E q ) ≃ δ ( ϵ k α − ϵ p α ) . (77) \delta(\epsilon_{k\alpha}-\epsilon_{p\alpha}\pm E_q)\simeq\delta(\epsilon_{k\alpha}-\epsilon_{p\alpha}). \tag*{(77)} δ ( ϵ k α − ϵ p α ± E q ) ≃ δ ( ϵ k α − ϵ p α ) . (77)
Both electronic occupation products therefore reduce to f e ( 1 − f e ) f_e(1-f_e) f e ( 1 − f e ) and
E q 2 f e ( 1 − f e ) ( 1 + 2 n q ) ≃ 2 k B T E q f e ( 1 − f e ) = 2 k B 2 T 2 E q ( − ∂ f e ∂ ϵ ) . (78) E_q^2f_e(1-f_e)(1+2n_q)\simeq2k_BT E_qf_e(1-f_e)
=2k_B^2T^2E_q\left(-\frac{\partial f_e}{\partial\epsilon}\right). \tag*{(78)} E q 2 f e ( 1 − f e ) ( 1 + 2 n q ) ≃ 2 k B T E q f e ( 1 − f e ) = 2 k B 2 T 2 E q ( − ∂ ϵ ∂ f e ) . (78)
This produces the microscopic FDT pair
S E P ( 0 ) = 4 π k B 2 T 2 ℏ g d ∑ k p α ∣ M p k ∣ 2 E q ( − ∂ f e ∂ ϵ ) δ ( ϵ k α − ϵ p α ) , (79) S_{\mathrm{EP}}(0)=\frac{4\pi k_B^2T^2}{\hbar}g_d\sum_{kp\alpha}|M_{pk}|^2E_q\left(-\frac{\partial f_e}{\partial\epsilon}\right)\delta(\epsilon_{k\alpha}-\epsilon_{p\alpha}), \tag*{(79)} S EP ( 0 ) = ℏ 4 π k B 2 T 2 g d k p α ∑ ∣ M p k ∣ 2 E q ( − ∂ ϵ ∂ f e ) δ ( ϵ k α − ϵ p α ) , (79)
G E P ( T ) = 2 π k B ℏ g d ∑ k p α ∣ M p k ∣ 2 E q ( − ∂ f e ∂ ϵ ) δ ( ϵ k α − ϵ p α ) . (80) G_{\mathrm{EP}}(T)=\frac{2\pi k_B}{\hbar}g_d\sum_{kp\alpha}|M_{pk}|^2E_q\left(-\frac{\partial f_e}{\partial\epsilon}\right)\delta(\epsilon_{k\alpha}-\epsilon_{p\alpha}). \tag*{(80)} G EP ( T ) = ℏ 2 π k B g d k p α ∑ ∣ M p k ∣ 2 E q ( − ∂ ϵ ∂ f e ) δ ( ϵ k α − ϵ p α ) . (80)
Their ratio is 2 k B T 2 2k_BT^2 2 k B T 2 before the remaining integrals are evaluated.
Monolayer graphene
We now evaluate the remaining momentum integral in Eq. (80) for MLG to obtain Eq. (28) , namely
∣ M p k ∣ 2 E q = D 2 ℏ 2 q 2 2 A ρ M L G F ( ϕ ) , (81) |M_{p k}|^2 E_q = \frac{D^2 \hbar^2 q^2}{2 A \rho_{\mathrm{MLG}}} F(\phi), \tag*{(81)} ∣ M p k ∣ 2 E q = 2 A ρ MLG D 2 ℏ 2 q 2 F ( ϕ ) , (81)
so v s v_s v s cancels. Elastic intraband scattering gives
δ ( ϵ k α − ϵ p α ) = 1 ℏ v F δ ( k − p ) . (82) \delta(\epsilon_{k\alpha} - \epsilon_{p\alpha}) = \frac{1}{\hbar v_{\mathrm{F}}} \delta(k-p). \tag*{(82)} δ ( ϵ k α − ϵ p α ) = ℏ v F 1 δ ( k − p ) . (82)
On the shell p = k p=k p = k ,
q 2 F + + M L G = 2 k 2 ( 1 − cos ϕ ) 1 + cos ϕ 2 = k 2 sin 2 ϕ . (83) q^2 F_{++}^{\mathrm{MLG}} = 2 k^2 (1-\cos\phi) \frac{1+\cos\phi}{2} = k^2 \sin^2\phi. \tag*{(83)} q 2 F ++ MLG = 2 k 2 ( 1 − cos ϕ ) 2 1 + cos ϕ = k 2 sin 2 ϕ . (83)
The angular part of the momentum integral in Eq. (80) is therefore
∫ 0 2 π d θ k ∫ 0 2 π d θ p q 2 F + + M L G = 2 π 2 k 2 , (84) \int_0^{2\pi} d\theta_k \int_0^{2\pi} d\theta_p q^2 F_{++}^{\mathrm{MLG}} = 2\pi^2 k^2, \tag*{(84)} ∫ 0 2 π d θ k ∫ 0 2 π d θ p q 2 F ++ MLG = 2 π 2 k 2 , (84)
where ϕ = θ k − θ p \phi=\theta_k-\theta_p ϕ = θ k − θ p , and thus
G E P M L G A = g d D 2 k B 8 π ρ M L G ℏ v F 6 ∫ d ϵ ϵ 4 ( − ∂ f e ∂ ϵ ) . (85) \frac{G_{\mathrm{EP}}^{\mathrm{MLG}}}{A} = \frac{g_d D^2 k_{\mathrm{B}}}{8\pi\rho_{\mathrm{MLG}} \hbar v_{\mathrm{F}}^6} \int d\epsilon\epsilon^4 \left(-\frac{\partial f_e}{\partial\epsilon}\right). \tag*{(85)} A G EP MLG = 8 π ρ MLG ℏ v F 6 g d D 2 k B ∫ d ϵ ϵ 4 ( − ∂ ϵ ∂ f e ) . (85)
The exact Fermi-window moment is
∫ d ϵ ϵ 4 ( − f e ′ ) = μ 4 + 2 π 2 μ 2 ( k B T ) 2 + 7 π 4 15 ( k B T ) 4 , (86) \int d\epsilon\,\epsilon^4(-f_e') = \mu^4 + 2\pi^2\mu^2(k_{\mathrm{B}}T)^2 + \frac{7\pi^4}{15}(k_{\mathrm{B}}T)^4, \tag*{(86)} ∫ d ϵ ϵ 4 ( − f e ′ ) = μ 4 + 2 π 2 μ 2 ( k B T ) 2 + 15 7 π 4 ( k B T ) 4 , (86)
leading to Eq. (28) .
Bilayer graphene
To obtain the BLG equipartition conductance in Eq. (29) , we start from the general conductance expression, Eq. (80) . We evaluate its momentum integrals using the parabolic BLG dispersion and the intraband chirality factor F = cos 2 ϕ F=\cos^2\phi F = cos 2 ϕ , where ϕ \phi ϕ is the angle between k \mathbf{k} k and p \mathbf{p} p . Elastic energy conservation fixes p = k p=k p = k , so the radial delta function becomes
δ ( ϵ k − ϵ p ) = m ℏ 2 p δ ( p − k ) . (87) \delta(\epsilon_k-\epsilon_p) = \frac{m}{\hbar^2 p}\delta(p-k). \tag*{(87)} δ ( ϵ k − ϵ p ) = ℏ 2 p m δ ( p − k ) . (87)
The intraband chirality factor is F = cos 2 ϕ F=\cos^2\phi F = cos 2 ϕ . On p = k p=k p = k ,
q 2 F = 2 k 2 ( 1 − cos ϕ ) cos 2 ϕ , (88) q^2 F = 2k^2(1-\cos\phi)\cos^2\phi, \tag*{(88)} q 2 F = 2 k 2 ( 1 − cos ϕ ) cos 2 ϕ , (88)
and
∫ d θ k d θ p q 2 F = 4 π 2 k 2 . (89) \int d\theta_k d\theta_p q^2 F = 4\pi^2 k^2. \tag*{(89)} ∫ d θ k d θ p q 2 F = 4 π 2 k 2 . (89)
The conductance then becomes
G E P B L G A = g d D 2 k B m 3 2 π ρ B L G ℏ 5 ∫ d E ∣ E ∣ ( − ∂ f e ∂ E ) . (90) \frac{G_{\mathrm{EP}}^{\mathrm{BLG}}}{A} = \frac{g_d D^2 k_{\mathrm{B}}m^3}{2\pi\rho_{\mathrm{BLG}}\hbar^5} \int dE\,|E|\left(-\frac{\partial f_e}{\partial E}\right). \tag*{(90)} A G EP BLG = 2 π ρ BLG ℏ 5 g d D 2 k B m 3 ∫ d E ∣ E ∣ ( − ∂ E ∂ f e ) . (90)
Splitting the signed-energy integral into electron and hole parts gives
∫ d E ∣ E ∣ ( − f e ′ ) = 2 k B T ln [ 2 cosh ( μ 2 k B T ) ] . (91) \int dE\,|E|(-f_e') = 2 k_{\mathrm{B}}T \ln\left[2\cosh\left(\frac{\mu}{2k_{\mathrm{B}}T}\right)\right]. \tag*{(91)} ∫ d E ∣ E ∣ ( − f e ′ ) = 2 k B T ln [ 2 cosh ( 2 k B T μ ) ] . (91)
Using the above two equations and m = γ 1 / ( 2 v 0 2 ) m=\gamma_1/(2v_0^2) m = γ 1 / ( 2 v 0 2 ) and g d = 4 g_d=4 g d = 4 gives Eq. (29) from main text.
Two-temperature EP rate sum
At different temperatures, the classical phonon factor is evaluated at T p h T_{\mathrm{ph}} T ph while f e ( 1 − f e ) = k B T e ( − ∂ ϵ f e ) f_e(1-f_e)=k_{\mathrm{B}}T_e(-\partial_\epsilon f_e) f e ( 1 − f e ) = k B T e ( − ∂ ϵ f e ) . Consequently, the same momentum integral that defines g ( μ , T e ) g(\mu,T_e) g ( μ , T e ) is multiplied by 2 k B T e T p h 2k_{\mathrm{B}}T_eT_{\mathrm{ph}} 2 k B T e T ph , yielding Eq. (34) . This direct calculation is the origin of that formula; it is not obtained by applying FDT away from equilibrium.
Finite-frequency reductions and asymptotics
BG continuum reduction
In the doped BG regime, the electronic energy difference Δ ≡ ϵ k α − ϵ p α \Delta\equiv\epsilon_{k\alpha}-\epsilon_{p\alpha} Δ ≡ ϵ k α − ϵ p α , with p = k − q \mathbf{p}=\mathbf{k}-\mathbf{q} p = k − q , is linearized as Δ ≃ ℏ v F q cos θ \Delta\simeq\hbar v_{\mathrm{F}}q\cos\theta Δ ≃ ℏ v F q cos θ for MLG. For the two finite-frequency shells, define c ∓ = ( ℏ ω ∓ E q ) / ( ℏ v F q ) c_{\mp}=(\hbar\omega\mp E_q)/(\hbar v_{\mathrm{F}}q) c ∓ = ( ℏ ω ∓ E q ) / ( ℏ v F q ) . The angular integral is
∫ 0 2 π d θ δ ( ℏ ω − ℏ v F q cos θ ∓ E q ) = 2 Θ ( 1 − ∣ c ∓ ∣ ) ℏ v F q 1 − c ∓ 2 . (92) \int_0^{2\pi}d\theta\,\delta(\hbar\omega-\hbar v_{\mathrm{F}}q\cos\theta\mp E_q)=\frac{2\Theta(1-|c_{\mp}|)}{\hbar v_{\mathrm{F}}q\sqrt{1-c_{\mp}^2}}. \tag*{(92)} ∫ 0 2 π d θ δ ( ℏ ω − ℏ v F q cos θ ∓ E q ) = ℏ v F q 1 − c ∓ 2 2Θ ( 1 − ∣ c ∓ ∣ ) . (92)
The leading BG continuum reduction replaces this factor by 2 / ( ℏ v F q ) 2/(\hbar v_{\mathrm{F}}q) 2/ ( ℏ v F q ) when ∣ c ∓ ∣ ≪ 1 |c_{\mp}|\ll1 ∣ c ∓ ∣ ≪ 1 . Thus, the shell sign affects the exact Jacobian and kinematic constraint, but drops out of its leading approximation. The Fermi-surface and phonon-energy measures are k d k ≃ ( k F / ℏ v F ) d ϵ k\,dk\simeq(k_{\mathrm{F}}/\hbar v_{\mathrm{F}})d\epsilon k d k ≃ ( k F /ℏ v F ) d ϵ and q d q = ( E / ℏ 2 v s 2 ) d E q\,dq=(E/\hbar^2v_s^2)dE q d q = ( E / ℏ 2 v s 2 ) d E . The first frequency shell imposes Δ = ℏ ω − E \Delta=\hbar\omega-E Δ = ℏ ω − E and gives the vertex ( ℏ ω − 2 E ) 2 (\hbar\omega-2E)^2 ( ℏ ω − 2 E ) 2 together with
∫ d ϵ f e ( ϵ + E − ℏ ω ) [ 1 − f e ( ϵ ) ] = Φ e ( E − ℏ ω ; T e ) (93) \int d\epsilon\,f_e(\epsilon+E-\hbar\omega)[1-f_e(\epsilon)] = \Phi_e(E-\hbar\omega;T_e) \tag*{(93)} ∫ d ϵ f e ( ϵ + E − ℏ ω ) [ 1 − f e ( ϵ )] = Φ e ( E − ℏ ω ; T e ) (93)
The second shell imposes Δ = ℏ ω + E \Delta=\hbar\omega+E Δ = ℏ ω + E and gives ( ℏ ω + 2 E ) 2 Ψ e ( E + ℏ ω ; T e ) (\hbar\omega+2E)^2\Psi_e(E+\hbar\omega;T_e) ( ℏ ω + 2 E ) 2 Ψ e ( E + ℏ ω ; T e ) . Substituting these shell weights and electronic energy integrals into Eq. (15) gives Eq. (93) . Symmetrizing in ω \omega ω then gives Eq. (35) . BLG has the same dimensionless kernel; its parabolic radial measure and angular Jacobian change only the prefactor.
For the symmetric temperature bias, T e = T ( 1 + r ) T_e=T(1+r) T e = T ( 1 + r ) and T p h = T ( 1 − r ) T_{\mathrm{ph}}=T(1-r) T ph = T ( 1 − r ) , where r = Δ T / ( 2 T ) r=\Delta T/(2T) r = Δ T / ( 2 T ) , introduce x = E / ( k B T ) x=E/(k_{\mathrm{B}}T) x = E / ( k B T ) and Ω = ℏ ω / ( k B T ) \Omega=\hbar\omega/(k_{\mathrm{B}}T) Ω = ℏ ω / ( k B T ) . The dimensionless occupation and electronic phase-space functions are
N p h ( x ; r ) = 1 e x / ( 1 − r ) − 1 , (94) N_{\mathrm{ph}}(x;r)=\frac{1}{e^{x/(1-r)}-1}, \tag*{(94)} N ph ( x ; r ) = e x / ( 1 − r ) − 1 1 , (94)
Φ r ( z ) = z e z / ( 1 + r ) − 1 , (95) \Phi_r(z)=\frac{z}{e^{z/(1+r)}-1}, \tag*{(95)} Φ r ( z ) = e z / ( 1 + r ) − 1 z , (95)
Ψ r ( z ) = z 1 − e − z / ( 1 + r ) . (96) \Psi_r(z)=\frac{z}{1-e^{-z/(1+r)}}. \tag*{(96)} Ψ r ( z ) = 1 − e − z / ( 1 + r ) z . (96)
Under this substitution, Φ e ( E − ℏ ω ; T e ) = k B T Φ r ( x − Ω ) \Phi_{\mathrm{e}}(E-\hbar\omega;T_{\mathrm{e}})=k_{\mathrm{B}}T\Phi_{\mathrm{r}}(x-\Omega) Φ e ( E − ℏ ω ; T e ) = k B T Φ r ( x − Ω ) and Ψ e ( E + ℏ ω ; T e ) = k B T Ψ r ( x + Ω ) \Psi_{\mathrm{e}}(E+\hbar\omega;T_{\mathrm{e}})=k_{\mathrm{B}}T\Psi_{\mathrm{r}}(x+\Omega) Ψ e ( E + ℏ ω ; T e ) = k B T Ψ r ( x + Ω ) . (93) therefore contains the dimensionless kernel
B r ( x , Ω ) = ( Ω − 2 x ) 2 [ 1 + N p h ( x ; r ) ] Φ r ( x − Ω ) + ( Ω + 2 x ) 2 N p h ( x ; r ) Ψ r ( x + Ω ) . (97) B_{\mathrm{r}}(x,\Omega)=(\Omega-2x)^2[1+N_{\mathrm{ph}}(x;r)]\Phi_{\mathrm{r}}(x-\Omega)
+(\Omega+2x)^2N_{\mathrm{ph}}(x;r)\Psi_{\mathrm{r}}(x+\Omega). \tag*{(97)} B r ( x , Ω ) = ( Ω − 2 x ) 2 [ 1 + N ph ( x ; r )] Φ r ( x − Ω ) + ( Ω + 2 x ) 2 N ph ( x ; r ) Ψ r ( x + Ω ) . (97)
Combining the dimensionless kernel B r ( x , Ω ) B_{\mathrm{r}}(x,\Omega) B r ( x , Ω ) in Eq. (97) with the measure x d x x\,dx x d x and symmetrizing through [ B r ( x , Ω ) + B r ( x , − Ω ) ] / 2 [B_{\mathrm{r}}(x,\Omega)+B_{\mathrm{r}}(x,-\Omega)]/2 [ B r ( x , Ω ) + B r ( x , − Ω )] /2 gives Eq. (41) . The overall ( k B T ) 5 (k_{\mathrm{B}}T)^5 ( k B T ) 5 factor follows from the energy measure E d E E\,dE E d E , the squared current vertex, and the electronic phase-space function, which contribute two, two, and one powers of k B T k_{\mathrm{B}}T k B T , respectively.
BG low-frequency coefficient
To obtain the quadratic low-frequency correction, we expand the dimensional BG kernel entering Eq. (93) about frequency, at fixed T e T_{\mathrm{e}} T e and T p h T_{\mathrm{ph}} T ph . For compactness, write Φ ( E ) = Φ e ( E ; T e ) \Phi(E)=\Phi_{\mathrm{e}}(E;T_{\mathrm{e}}) Φ ( E ) = Φ e ( E ; T e ) , Ψ ( E ) = Ψ e ( E ; T e ) \Psi(E)=\Psi_{\mathrm{e}}(E;T_{\mathrm{e}}) Ψ ( E ) = Ψ e ( E ; T e ) , and N ( E ) = n p h ( E , T p h ) N(E)=n_{\mathrm{ph}}(E,T_{\mathrm{ph}}) N ( E ) = n ph ( E , T ph ) . The kernel is B ( E , ℏ ω ) = ( ℏ ω − 2 E ) 2 [ 1 + N ( E ) ] Φ ( E − ℏ ω ) + ( ℏ ω + 2 E ) 2 N ( E ) Ψ ( E + ℏ ω ) B(E,\hbar\omega)=(\hbar\omega-2E)^2[1+N(E)]\Phi(E-\hbar\omega)+(\hbar\omega+2E)^2N(E)\Psi(E+\hbar\omega) B ( E , ℏ ω ) = ( ℏ ω − 2 E ) 2 [ 1 + N ( E )] Φ ( E − ℏ ω ) + ( ℏ ω + 2 E ) 2 N ( E ) Ψ ( E + ℏ ω ) . Symmetrization cancels the odd powers of ω \omega ω , giving B ( E , ℏ ω ) + B ( E , − ℏ ω ) = 2 B ( E , 0 ) + ( ℏ ω ) 2 B 2 ( E ) + O ( ( ℏ ω ) 4 ) B(E,\hbar\omega)+B(E,-\hbar\omega)=2B(E,0)+(\hbar\omega)^2B_2(E)+O((\hbar\omega)^4) B ( E , ℏ ω ) + B ( E , − ℏ ω ) = 2 B ( E , 0 ) + ( ℏ ω ) 2 B 2 ( E ) + O (( ℏ ω ) 4 ) , where B 2 ( E ) = ( 1 + N ) [ 2 Φ + 8 E Φ ′ + 4 E 2 Φ ′ ′ ] + N [ 2 Ψ + 8 E Ψ ′ + 4 E 2 Ψ ′ ′ ] B_2(E)=(1+N)\left[2\Phi+8E\Phi'+4E^2\Phi''\right]+N\left[2\Psi+8E\Psi'+4E^2\Psi''\right] B 2 ( E ) = ( 1 + N ) [ 2Φ + 8 E Φ ′ + 4 E 2 Φ ′′ ] + N [ 2Ψ + 8 E Ψ ′ + 4 E 2 Ψ ′′ ] . At equilibrium, Bose-series integration or repeated integration by parts gives
1 2 ∫ 0 ∞ d E E B 2 ( E ) ∫ 0 ∞ d E E B ( E , 0 ) = 1 ( k B T ) 2 ( 1 12 + 5 16 π 2 ) , (98) \frac{1}{2}\frac{\int_0^\infty dE\,E B_2(E)}{\int_0^\infty dE\,E B(E,0)}
=\frac{1}{(k_{\mathrm{B}}T)^2}\left(\frac{1}{12}+\frac{5}{16\pi^2}\right), \tag*{(98)} 2 1 ∫ 0 ∞ d E E B ( E , 0 ) ∫ 0 ∞ d E E B 2 ( E ) = ( k B T ) 2 1 ( 12 1 + 16 π 2 5 ) , (98)
which yields Eq. (39) . Figure 8 compares the complete reduced integrals with the quadratic approximations. The inset magnifies the range in which the curves begin to separate. The same comparison for BLG is shown in Figure 9 . The BG normalized kernel is common to MLG and BLG, while the EP electronic phase space is different.
Figure 8. Exact reduced MLG finite-frequency ratios compared with their quadratic low-frequency approximations. The upper panels show BG results at mean temperature T = 5 T = 5 T = 5 K, calculated from Eq. (41) . The equilibrium BG approximation follows Eq. (39) ; the nonequilibrium quadratic coefficient is obtained from the small-frequency expansion of the same BG integral, as described in Appendix D. The lower panels show EP results at T = 100 T = 100 T = 100 K from Eq. (42) , with quadratic approximations from Eq. (46) . The left column is at equilibrium; the right column uses T e / p h = T ± Δ T / 2 T_{\mathrm{e/ph}} = T \pm\Delta T/2 T e/ph = T ± Δ T /2 with Δ T / T = 0.4 \Delta T/T = 0.4 Δ T / T = 0.4 . All ratios use Eq. (53) . Insets magnify the high-frequency separation.
Figure 9. Exact reduced BLG finite-frequency ratios and quadratic approximations. The upper panels show BG results at mean temperature T = 5 T=5 T = 5 K from Eq. (41) . The equilibrium quadratic approximation follows Eq. (39) ; the nonequilibrium coefficient is obtained from the small-frequency expansion described in Appendix D. The BG ratio has the same dimensionless form as in MLG. The BLG EP panels use the J 0 J_0 J 0 and J 1 J_1 J 1 energy integrals.
Zero-temperature BG spectrum
At T e = T p h = 0 T_{\mathrm{e}}=T_{\mathrm{ph}}=0 T e = T ph = 0 and ℏ ω > 0 \hbar\omega>0 ℏ ω > 0 , the only nonvanishing interval in the symmetrized BG integral is 0 < E < ℏ ω 0<E<\hbar\omega 0 < E < ℏ ω . Direct evaluation gives
1 2 ∫ 0 ℏ ω d E E ( ℏ ω − E ) ( ℏ ω − 2 E ) 2 = ( ℏ ω ) 5 60 , (99) \frac{1}{2}\int_0^{\hbar\omega}dE\,E(\hbar\omega-E)(\hbar\omega-2E)^2=\frac{(\hbar\omega)^5}{60}, \tag*{(99)} 2 1 ∫ 0 ℏ ω d E E ( ℏ ω − E ) ( ℏ ω − 2 E ) 2 = 60 ( ℏ ω ) 5 , (99)
which yields Eq. (40) . The ∣ ω ∣ 5 |\omega|^5 ∣ ω ∣ 5 law is the graphene BG counterpart of the finite-frequency quantum heat noise found in metallic systems [36 ] .
MLG EP angular factors
In the quasi-elastic EP reduction, the finite electron energies are ϵ \epsilon ϵ and ϵ + ℏ ω \epsilon+\hbar\omega ϵ + ℏ ω . The magnitudes of the two electron momenta are ∣ ϵ ∣ / ( ℏ v F ) |\epsilon|/(\hbar v_{\mathrm{F}}) ∣ ϵ ∣/ ( ℏ v F ) and ∣ ϵ + ℏ ω ∣ / ( ℏ v F ) |\epsilon+\hbar\omega|/(\hbar v_{\mathrm{F}}) ∣ ϵ + ℏ ω ∣/ ( ℏ v F ) . The angular averages are
∫ d θ k d θ p F M L G ( ϕ ) = 2 π 2 , (100) \int d\theta_kd\theta_p\,F^{\mathrm{MLG}}(\phi)=2\pi^2, \tag*{(100)} ∫ d θ k d θ p F MLG ( ϕ ) = 2 π 2 , (100)
∫ d θ k d θ p q 2 F M L G ( ϕ ) = 2 π 2 ( k 2 + p 2 − k p ) . (101) \int d\theta_kd\theta_p\,q^2F^{\mathrm{MLG}}(\phi)=2\pi^2(k^2+p^2-kp). \tag*{(101)} ∫ d θ k d θ p q 2 F MLG ( ϕ ) = 2 π 2 ( k 2 + p 2 − k p ) . (101)
They produce the two energy integrals I 2 \mathcal{I}_2 I 2 and I 4 \mathcal{I}_4 I 4 in Eqs. (43) . At ℏ ω = 0 \hbar\omega=0 ℏ ω = 0 , I 4 ( 0 ; T e ) = 2 k B T e M 4 ( μ , T e ) \mathcal{I}_4(0;T_{\mathrm{e}})=2k_{\mathrm{B}}T_{\mathrm{e}}M_4(\mu,T_{\mathrm{e}}) I 4 ( 0 ; T e ) = 2 k B T e M 4 ( μ , T e ) , which reduces Eq. (42) to Eq. (34) .
BLG EP angular factors
For BLG, F B L G = cos 2 ϕ F^{\mathrm{BLG}}=\cos^2\phi F BLG = cos 2 ϕ and
∫ d θ k d θ p F B L G = 2 π 2 , (102) \int d\theta_kd\theta_p\,F^{\mathrm{BLG}}=2\pi^2, \tag*{(102)} ∫ d θ k d θ p F BLG = 2 π 2 , (102)
∫ d θ k d θ p q 2 F B L G = 2 π 2 ( k 2 + p 2 ) . (103) \int d\theta_kd\theta_p\,q^2F^{\mathrm{BLG}}=2\pi^2(k^2+p^2). \tag*{(103)} ∫ d θ k d θ p q 2 F BLG = 2 π 2 ( k 2 + p 2 ) . (103)
These averages lead to J 0 J_0 J 0 and J 1 J_1 J 1 . At ℏ ω = 0 \hbar\omega=0 ℏ ω = 0 ,
J 1 ( 0 ; T e ) = 8 ( k B T e ) 2 ln [ 2 cosh ( μ 2 k B T e ) ] , (104) J_1(0;T_{\mathrm{e}})=8(k_{\mathrm{B}}T_{\mathrm{e}})^2\ln\left[2\cosh\left(\frac{\mu}{2k_{\mathrm{B}}T_{\mathrm{e}}}\right)\right], \tag*{(104)} J 1 ( 0 ; T e ) = 8 ( k B T e ) 2 ln [ 2 cosh ( 2 k B T e μ ) ] , (104)
which reproduces the two-temperature BLG EP zero-frequency noise. In the degenerate BLG EP limit, ∣ μ ∣ ≫ k B T e |\mu|\gg k_{\mathrm{B}}T_{\mathrm{e}} ∣ μ ∣ ≫ k B T e , the low-frequency ratio is
S E P B L G ( ω ) S E P B L G ( 0 ) ≃ 1 + 1 12 ( ℏ ω k B T e ) 2 + ( ℏ ω ) 2 16 π v s 2 ∣ μ ∣ + O ( ( ℏ ω ) 4 ) . (105) \frac{S_{\mathrm{EP}}^{\mathrm{BLG}}(\omega)}{S_{\mathrm{EP}}^{\mathrm{BLG}}(0)}
\simeq1+\frac{1}{12}\left(\frac{\hbar\omega}{k_{\mathrm{B}}T_{\mathrm{e}}}\right)^2+\frac{(\hbar\omega)^2}{16\pi v_{\mathrm{s}}^2|\mu|}+O((\hbar\omega)^4). \tag*{(105)} S EP BLG ( 0 ) S EP BLG ( ω ) ≃ 1 + 12 1 ( k B T e ℏ ω ) 2 + 16 π v s 2 ∣ μ ∣ ( ℏ ω ) 2 + O (( ℏ ω ) 4 ) . (105)
Full on-shell MLG integral and numerical implementation
The full MLG benchmark keeps exact intraband kinematics for μ > 0 \mu>0 μ > 0 . Introduce x = k / k F x=k/k_{\mathrm{F}} x = k / k F , y = q / k F y=q/k_{\mathrm{F}} y = q / k F , η = v s / v F \eta=v_{\mathrm{s}}/v_{\mathrm{F}} η = v s / v F , z = ℏ ω / μ z=\hbar\omega/\mu z = ℏ ω / μ . For frequency sign λ = ± 1 \lambda=\pm1 λ = ± 1 and process sign s = ± 1 s=\pm1 s = ± 1 ,
u λ s = p ˉ λ s 2 − x 2 − y 2 2 x y , p ˉ λ s = x + s η y − λ z , (106) u_{\lambda s}=\frac{\bar{p}_{\lambda s}^{\,2}-x^2-y^2}{2xy},\qquad\bar{p}_{\lambda s}=x+s\eta y-\lambda z, \tag*{(106)} u λ s = 2 x y p ˉ λ s 2 − x 2 − y 2 , p ˉ λ s = x + sη y − λ z , (106)
F λ s = 1 + cos ϕ λ s 2 , cos ϕ λ s = x + y u λ s p ˉ λ s . (107) F_{\lambda s}=\frac{1+\cos\phi_{\lambda s}}{2},\qquad\cos\phi_{\lambda s}=\frac{x+yu_{\lambda s}}{\bar{p}_{\lambda s}}. \tag*{(107)} F λ s = 2 1 + cos ϕ λ s , cos ϕ λ s = p ˉ λ s x + y u λ s . (107)
The shell is allowed if p ˉ λ s > 0 \bar{p}_{\lambda s}>0 p ˉ λ s > 0 and ∣ u λ s ∣ < 1 |u_{\lambda s}|<1 ∣ u λ s ∣ < 1 . The phonon energy is E q = η μ y E_q=\eta\mu y E q = η μ y . The phonon weight is W + = 1 + n p h W_+=1+n_{\mathrm{ph}} W + = 1 + n ph for emission and W − = n p h W_-=n_{\mathrm{ph}} W − = n ph for absorption. The exact spectrum implemented in the code is
S ( ω ; T e , T p h ) A = g d D 2 k F 4 32 π 2 ρ M L G v s ℏ v F ∑ λ , s = ± 1 ∫ 0 ∞ d x ∫ 0 ∞ d y p ˉ λ s y F λ s 1 − u λ s 2 f e ( μ p ˉ λ s ; T e ) [ 1 − f e ( μ x ; T e ) ] × W s ( η μ y ; T p h ) [ λ ℏ ω − 2 s η μ y ] 2 Θ ( p ˉ λ s ) Θ ( 1 − ∣ u λ s ∣ ) . (108) \frac{S(\omega;T_{\mathrm{e}},T_{\mathrm{ph}})}{A}
=
\frac{g_{\mathrm{d}}D^{2}k_{\mathrm{F}}^{4}}{32\pi^{2}\rho_{\mathrm{MLG}}v_{\mathrm{s}}\hbar v_{\mathrm{F}}}
\sum_{\lambda,s=\pm1}
\int_{0}^{\infty} dx
\int_{0}^{\infty} dy\,
\frac{\bar{p}_{\lambda s}yF_{\lambda s}}{\sqrt{1-u_{\lambda s}^{2}}}
f_{\mathrm{e}}(\mu\bar{p}_{\lambda s};T_{\mathrm{e}})
\left[1-f_{\mathrm{e}}(\mu x;T_{\mathrm{e}})\right]
\times W_{s}(\eta\mu y;T_{\mathrm{ph}})
\left[\lambda\hbar\omega-2s\eta\mu y\right]^{2}
\Theta(\bar{p}_{\lambda s})\Theta(1-\lvert u_{\lambda s}\rvert). \tag*{(108)} A S ( ω ; T e , T ph ) = 32 π 2 ρ MLG v s ℏ v F g d D 2 k F 4 λ , s = ± 1 ∑ ∫ 0 ∞ d x ∫ 0 ∞ d y 1 − u λ s 2 p ˉ λ s y F λ s f e ( μ p ˉ λ s ; T e ) [ 1 − f e ( μx ; T e ) ] × W s ( η μ y ; T ph ) [ λ ℏ ω − 2 sη μ y ] 2 Θ ( p ˉ λ s ) Θ ( 1 − ∣ u λ s ∣) . (108)
No BG small-q q q , EP classical-Bose, or quasi-elastic approximation enters Eq. (108) . The remaining restrictions are the linear MLG band, linear LA phonons, unscreened deformation potential, leading order in D 2 D^{2} D 2 , and the doped intraband channel.
Numerical quadrature
The reduced BG, MLG EP, and BLG EP one-dimensional integrals are evaluated with Gauss–Legendre quadrature on finite intervals chosen so that the distribution functions suppress the neglected tails. Stable series are used near zero argument for the Bose and electron phase-space functions. The full two-dimensional integral uses composite nonuniform grids. The k k k grid is refined around k F k_{\mathrm{F}} k F over the thermal width δ k ∼ k B T e / ℏ v F \delta k \sim k_{\mathrm{B}}T_{\mathrm{e}}/\hbar v_{\mathrm{F}} δ k ∼ k B T e /ℏ v F while the q q q grid resolves both the thermal phonon scale q T = k B T p h / ( ℏ v s ) q_{\mathrm{T}} = k_{\mathrm{B}}T_{\mathrm{ph}}/(\hbar v_{\mathrm{s}}) q T = k B T ph / ( ℏ v s ) and the large-q q q EP region. Kinematically forbidden points are removed by the two step functions in Eq. (108) . The square-root edge is regulated only at machine-level distance from the boundary and convergence is checked by increasing N k N_k N k and N q N_q N q , the numbers of integration points in the electron and phonon wave-vector magnitudes, respectively. Figure 10 shows two representative tests. Panel (a) compares the converged full zero-frequency noise at T = 0.02 T B G T = 0.02T_{\mathrm{BG}} T = 0.02 T BG with the analytic BG expression. Panel (b) tests the stability of R ( ω ) = S ( ω ) / S ( 0 ) R(\omega) = S(\omega)/S(0) R ( ω ) = S ( ω ) / S ( 0 ) at T = T B G T = T_{\mathrm{BG}} T = T BG and f = 100 f = 100 f = 100 GHz by plotting its relative deviation from the finest-grid result in parts per million. The validation script also verifies the zero-frequency normalizations and the analytic quadratic coefficients.
Figure 10. Grid-convergence tests for the full MLG on-shell integral, Eq. (108) . Here N k N_k N k and N q N_q N q are the numbers of integration points for the electron and phonon wave-vector magnitudes, respectively. Here N k N_k N k and N q N_q N q denote the numbers of integration points for the electron and phonon wave-vector magnitudes, respectively. (a) Zero-frequency ratio S f u l l ( 0 ) / S B G ( 0 ) S_{\mathrm{full}}(0)/S_{\mathrm{BG}}(0) S full ( 0 ) / S BG ( 0 ) at T e = T p h = 0.02 T B G T_e=T_{\mathrm{ph}}=0.02T_{\mathrm{BG}} T e = T ph = 0.02 T BG , using Eq. (20) for the BG reference. (b) Relative deviation 10 6 [ R N k , N q ( ω ) / R f i n e s t ( ω ) − 1 ] 10^6[R_{N_k,N_q}(\omega)/R_{\mathrm{finest}}(\omega)-1] 1 0 6 [ R N k , N q ( ω ) / R finest ( ω ) − 1 ] at T e = T p h = T B G T_e=T_{\mathrm{ph}}=T_{\mathrm{BG}} T e = T ph = T BG and f = 100 f=100 f = 100 GHz, expressed in parts per million. The ratio R R R is defined in Eq. (53) , and R f i n e s t R_{\mathrm{finest}} R finest uses the finest grid shown, ( N k , N q ) = ( 650 , 920 ) (N_k,N_q)=(650,920) ( N k , N q ) = ( 650 , 920 ) .