Introduction
The long-term rotational evolution of atmosphere-bearing rocky planets located within the habitable zone of their host stars is governed by both gravitational and thermal tides. The former arise from the differential stellar gravitational attraction, whereas the latter result from the day-night contrast in atmospheric heating by stellar radiation. In both cases, the tidal bulge is not aligned with the planet-star axis, leading to exchanges of energy and angular momentum between the orbital motion of the planet-star system and the planetary rotation. In the case of gravitational tides, this lag originates in dissipative mechanisms. By contrast, dissipation is not a prerequisite for thermal tides, which can still induce energy and angular momentum transfer even in its absence. For instance, non-dissipative isothermal models of Earth’s atmosphere predict a semidiurnal surface-pressure oscillation peaking at 09:00 LST (Local Solar Time) on the dayside – three hours before the insolation maximum at 12:00 LST (see e.g. [35 Chapter 3, {xrefext}] Eq. (160))`)1 . This phase lead, illustrated by Figure 1 , gives rise to the long-recognised thermodynamic acceleration of Earth’s rotation ([29 ] ).
Figure 1. Schematic illustrating the competing effects of gravitational and atmospheric thermal tides on the rotation of rocky planets. As long as the pressure bulge (red ellipsoid) leads the planet-star axis, the thermal tide produces a torque that tends to accelerate the rotation, while the torque exerted on the bulge raised by gravitational tidal forces (blue ellipsoid) tends to decelerate it. Lag angles measured from the planet-star axis are denoted by δ T \delta_{\mathrm{T}} δ T for the atmospheric thermal tide and δ G \delta_{\mathrm{G}} δ G for the gravitational tidal tide.
In recent years, the role of thermal tides in the evolution of planetary rotation has attracted renewed attention in the context of Earth’s rotational history. Several studies have revisited the tidal-locking hypothesis proposed by Zahnle & Walker [53 ] , who suggested that the Earth’s length of day (hereafter LOD) may have remained nearly constant for several hundred million years as a result of a balance between thermal and gravitational tidal torques. This hypothesis relies on the resonant excitation of global-scale atmospheric compressibility modes known as Lamb waves (e.g. [20 ] ). During the Precambrian era (540–4570 Ma), one such mode was resonantly excited, thereby amplifying thermal tides and the associated atmospheric torque. The existence of such a tidal-locking episode nevertheless remains a matter of debate, as conflicting interpretations have emerged.
Drawing on geological constraints, several authors argue that the Earth’s LOD remained stalled at approximately 19 hr for nearly one billion years [37 ] (Wu et al. 2023). By contrast, other studies conclude that geological evidence does not support a tidal-locking phase and provide analytical arguments against this scenario based on linear tidal theory [20 , 30 ] . The theoretical framework underpinning these latter studies, however, remains to be further consolidated. Moreover, their analyses focus primarily on Earth, despite the fact that the underlying mechanism is expected to operate on rocky planets more generally. Robust modelling frameworks are therefore required to assess the influence of thermal tides on the rotational evolution of terrestrial planets and to properly account for tidal-orbital coupling in planetary evolution models.
In a previous paper [5 hereafter Paper I] , we developed a general analytical model of atmospheric tides on rocky planets within the framework of linear theory, starting from the primitive equations of fluid dynamics. The model is based on a series of simplifying assumptions consistent with the classical theory (see Paper I, Sects. 2 and 3). In particular, it incorporates the effective damping induced by dissipative processes through Newtonian cooling. The atmosphere is represented as a thin fluid layer at the planet’s surface, with mean flows and horizontal gradients of the background equilibrium fields neglected. Tidal heating is prescribed to decay exponentially with the vertical coordinate. Under these assumptions, we derived a closed-form solution – given by Eqs. (85-92) in Paper I – that applies uniformly to two idealised atmospheric structures: (i) isentropic atmospheres, characterised by a globally uniform potential temperature and neutral stratification, and (ii) isothermal atmospheres, with globally uniform temperature and stable stratification. This solution describes the oscillations
of the three-dimensional dynamical and thermodynamical fields (pressure, density, temperature, and velocities) driven by tidal gravitational forcing and radiative heating in the atmospheric shell. In particular, it captures the resonances of Lamb waves.
In the present study (Paper II), we extend the analytical framework developed in Paper I to investigate the influence of semidiurnal tides on planetary rotation. Our main objective is to quantify the thermotidal torque exerted on the planet’s spin axis and to characterise its dependence on the tidal frequency. This requires examining the behaviour of Lamb modes resonantly excited by thermal forcing, as they play a major role in the rotational evolution of rapidly rotating planets such as Earth. More broadly, we assess the consistency between simplified analytical models and the numerical solutions obtained in previous studies from the simulations performed with general circulation models (GCM). These numerical models capture the complex coupling between atmospheric dynamics, radiative transfer, and surface heat exchanges, which lies beyond the reach of analytical theory. However, they generally do not allow the specific effects of individual physical processes to be disentangled. Owing to these limitations, it is essential to conduct a rigorous comparison between analytical and numerical approaches, and to identify the origins of their agreements and discrepancies. In a third paper (in prep.; hereafter Paper III), it will be shown that a mass-spring-damper model can reproduce the obtained tidal solutions.
The present paper is organised as follows. In Sect. 2, we derive explicit formulations of the thermotidal torque as functions of the atmospheric parameters for both isentropic and isothermal atmospheres. Additionally, we apply the model to the present-day Earth, for which the semidiurnal tidal bulge is determined empirically from barometric measurements. Section 3 characterises the frequency dependence of the tidal response through a quantitative functional analysis. In Sect. 4, the analytical theory is benchmarked against recent GCM simulations addressing the cases of Earth, Venus and dry rocky planets. Finally, our main results are discussed and summarised in Sect. 5. Symbols and acronyms used in the study are listed in the nomenclature of Appendix A.
Theory
General formulation of the thermotidal torque
The small deformations of solid regions due to variations in the atmosphere’s mass distribution are usually ignored in the thin layer approximation, the solid part being able to support most of the atmospheric loading without adjusting hydrodastically like fluid layers (e.g. [18 , 31 , 7 , 17 ] ). However, the atmospheric loading is sometimes taken into account in refined modelling approaches as it reduces the thermotidal torque by approximately 15% in Venus-like configurations (e.g. [12 , 39 ] ). Similarly, using GCM simulations, Musseau et al. [39 , 38 ] find that taking topography into account increases the Venusian thermal torque by a factor of 2.5.
In the present work, we assume that solid regions are rigid and topography is ignored, consistent with the classical theory (e.g. [10 ] ). Thus, the time-averaged tidal torque exerted on the atmosphere about the planet’s spin axis, T z \mathcal{T}_z T z , to be expressed as
T z = lim Δ t → + ∞ 1 Δ t ∫ 0 Δ t ∫ S R 2 δ p s g ∂ φ U T d S d t (1) \mathcal{T}_z = \lim_{\Delta t \to+\infty} \frac{1}{\Delta t} \int_0^{\Delta t} \int_S \frac{R^2 \delta p_s}{g} \partial_\varphi U_{\mathrm{T}} \,\mathrm{d}S\,\mathrm{d}t \tag*{(1)} T z = Δ t → + ∞ lim Δ t 1 ∫ 0 Δ t ∫ S g R 2 δ p s ∂ φ U T d S d t (1)
where R R R denotes the planet radius, g g g the surface gravity, U T U_{\mathrm{T}} U T the tide-raising gravitational potential, ∂ φ \partial_\varphi ∂ φ the partial derivative with respect to longitude, t t t the time, Δ t \Delta t Δ t a time interval, d S \mathrm{d}S d S an infinitesimal surface element on the sphere of unit radius, S S S , and δ p s \delta p_{\mathrm{s}} δ p s the surface pressure anomaly generated by the gravitational and thermal tidal forcings. Owing to linearity and to the time periodicity of the tidal perturbation, the surface pressure anomaly can be expanded as a Fourier series, with the Fourier coefficients being themselves expanded in spherical harmonics. Hence, we express δ p s \delta p_{\mathrm{s}} δ p s as
δ p s ( θ , φ , t ) = p s R { ∑ σ ∑ l = 2 + ∞ ∑ m = − l l δ p ^ s ; l m , σ Y l m ( θ , φ ) e i σ t } (2) \delta p_{\mathrm{s}}(\theta,\varphi,t)=p_{\mathrm{s}}\mathfrak{R}\left\{\sum_{\sigma}\sum_{l=2}^{+\infty}\sum_{m=-l}^{l}\widehat{\delta p}_{\mathrm{s;l}}^{m,\sigma}Y_l^m(\theta,\varphi)e^{i\sigma t}\right\} \tag*{(2)} δ p s ( θ , φ , t ) = p s R { σ ∑ l = 2 ∑ + ∞ m = − l ∑ l δ p s ; l m , σ Y l m ( θ , φ ) e iσ t } (2)
where R \mathfrak{R} R designates the real part of a complex number (I \mathfrak{I} I referring to the imaginary part), i i i the imaginary unit, θ \theta θ the colatitude, p s p_{\mathrm{s}} p s the globally-averaged surface pressure, σ \sigma σ the tidal frequency, Y l m Y_l^m Y l m the degree-l l l , order-m m m spherical harmonic (e.g. Varshalovich et al. [48 ] , Sect. 5.2), and δ p ^ s ; l m , σ \widehat{\delta p}_{\mathrm{s;l}}^{m,\sigma} δ p s ; l m , σ the associated dimensionless expansion coefficient. We emphasise that the term between braces in Eq. (2) is not a real-valued combination of pairs of complex conjugates in the adopted convention but a general complex number with non-zero imaginary part.
Thin atmospheres on rocky planets are generally presumed to rotate rigidly with the planet itself, so that any torque exerted on the atmosphere over long timescales alters the planet’s overall rotation rather than driving large-scale zonal winds. We assume that it is the case here. For convenience, we consider the coplanar-circular configuration, where the planet moves around the host star on a circular orbit of radius a a a , with its spin axis being normal to the orbital plane (no eccentricity, no obliquity). The planet’s mean motion, denoted by n ⋆ n_{\star} n ⋆ , is related to a a a through Kepler’s third law,
a 3 n ⋆ 2 = G ( M ⋆ + M ) (3) a^{3}n_{\star}^{2}=G\left(M_{\star}+M\right) \tag*{(3)} a 3 n ⋆ 2 = G ( M ⋆ + M ) (3)
with G G G being the universal gravitational constant, M M M the mass of the planet, and M ⋆ M_{\star} M ⋆ the mass of the star. In this configuration, using the hydrostatic equilibrium assumption, the time-averaged tidal torque of Eq. (1) simplifies to (e.g. Farhat et al. [20 ] , Eq. (21))2
T z = 6 π 5 M ⋆ M R 6 p s a 3 I { δ p ^ s ; 2 2 , σ } (4) \mathcal{T}_{z}=\sqrt{\frac{6\pi}{5}}\frac{M_{\star}}{M}\frac{R^{6}p_{\mathrm{s}}}{a^{3}}\mathfrak{I}\left\{\widehat{\delta p}_{\mathrm{s;2}}^{2,\sigma}\right\} \tag*{(4)} T z = 5 6 π M M ⋆ a 3 R 6 p s I { δ p s ; 2 2 , σ } (4)
where δ p ^ s ; 2 2 , σ \widehat{\delta p}_{\mathrm{s;2}}^{2,\sigma} δ p s ; 2 2 , σ denotes the degree-2, order-2 spherical expansion coefficient of Eq. (2) and σ = 2 ( Ω − n ⋆ ) \sigma=2(\Omega-n_{\star}) σ = 2 ( Ω − n ⋆ ) is the semidiurnal tidal frequency, with Ω \Omega Ω being the planet’s spin angular velocity. The coefficient δ p ^ s ; 2 2 , σ \widehat{\delta p}_{\mathrm{s;2}}^{2,\sigma} δ p s ; 2 2 , σ is a complex-valued function of tidal frequency, atmospheric parameters, and the multipole expansion components of the tide-raising gravitational potential and stellar irradiance.
The torque can also be expressed as a function of the complex Fourier coefficient of the equatorial semidiurnal surface pressure oscillation, δ p s ; e q ( t , φ ) = R { δ p s ; e q σ e i ( σ t + 2 φ ) } \delta p_{\mathrm{s;eq}}(t,\varphi)=\mathfrak{R}\{\delta p_{\mathrm{s;eq}}^{\sigma}e^{i(\sigma t+2\varphi)}\} δ p s ; eq ( t , φ ) = R { δ p s ; eq σ e i ( σ t + 2 φ ) } . The relationship between the degree-2, order-2 spherical expansion coefficient of tidal surface pressure variations and δ p s ; e q σ \delta p_{\mathrm{s;eq}}^{\sigma} δ p s ; eq σ is given by
δ p ^ s ; 2 2 , σ = 4 2 π 15 c f δ p s ; e q σ p s (5) \widehat{\delta p}_{\mathrm{s;2}}^{2,\sigma}=4\sqrt{\frac{2\pi}{15}}c_{\mathrm{f}}\frac{\delta p_{\mathrm{s;eq}}^{\sigma}}{p_{\mathrm{s}}} \tag*{(5)} δ p s ; 2 2 , σ = 4 15 2 π c f p s δ p s ; eq σ (5)
which yields
T z = 8 π 5 c f M ⋆ M R 6 a 3 I { δ p s ; e q σ } . (6) \mathcal{T}_{z}=\frac{8\pi}{5}c_{\mathrm{f}}\frac{M_{\star}}{M}\frac{R^{6}}{a^{3}}\mathfrak{I}\left\{\delta p_{\mathrm{s;eq}}^{\sigma}\right\}. \tag*{(6)} T z = 5 8 π c f M M ⋆ a 3 R 6 I { δ p s ; eq σ } . (6)
The parameter c f c_{\mathrm{f}} c f introduced in Eqs. (5) and (6) is a dimensionless shape factor that quantifies the degree to which the latitudinal structure of semidiurnal pressure variations resembles a quadrupolar pattern. As defined in Appendix § 7 , c f = 1 c_{\mathrm{f}}=1 c f = 1 corresponds to a purely quadrupolar semidiurnal tidal bulge. In this limit, the classical expression for the tidal torque is recovered (e.g. Dobrovolskis & Ingersoll [18 ] ; Zahnle & Walker [53 ] ).3
Under present-day Earth conditions, the global barometric observations reported by Schindellegger & Ray (2014) indicate that the latitudinal structure of the semidiurnal thermal tide is well approximated by a scaled sin 3 θ \sin^{3}\theta sin 3 θ function, corresponding to a shape factor of c f = 75 π / 256 ≈ 0.92 c_{\mathrm{f}}=75\pi/256\approx0.92 c f = 75 π /256 ≈ 0.92 (see Appendix § 7 ). To ensure consistency with terrestrial observations, this value is adopted throughout the present study. It should be emphasised, however, that the shape factor is intrinsically linked to tidal dynamics, as it implicitly incorporates effects such as rotational scattering. As a consequence, c f c_{\mathrm{f}} c f is not expected to remain constant in general, but rather depend on both the planetary spin rate and the tidal frequency. In particular, substantial variations may arise if a spherical harmonic of degree l ≥ 4 l\ge4 l ≥ 4 were resonantly amplified to amplitudes comparable to that of the degree-2 component at the equator, as discussed in Sect. 2.3. The empirical estimate inferred from Schindellegger & Ray (2014) further indicates that converting δ p ^ s ; 2 2 , σ \widehat{\delta p}_{\mathrm{s;2}}^{2,\sigma} δ p s ; 2 2 , σ into δ p s ; e q σ \delta p_{\mathrm{s;eq}}^{\sigma} δ p s ; eq σ introduces a relative uncertainty of approximately 8 % 8\% 8% in the corresponding thermotidal torque. In the following sections, we derive expressions for δ p ^ s ; 2 2 , σ \delta\hat{p}_{s;2}^{2,\sigma} δ p ^ s ; 2 2 , σ in isentropic and isothermal atmospheres using the analytical framework established in Sect. 3 of Paper I.
Multipole expansion of stellar irradiance
From now on, we ignore the gravitational tidal forcing. The atmospheric tide solely originates in the absorption of the incident stellar flux impinging the top of the atmosphere. Following earlier works (e.g. Auclair-Desrotour et al. [7 ] ; Farhat et al. [20 ] ), this flux is defined as
F = { F ⋆ cos ψ if 0 ≤ ψ ≤ π 2 (dayside) , 0 if π 2 < ψ ≤ π (nightside) . (7) F =
\begin{cases}
F_{\star}\cos\psi& \text{if } 0 \le\psi\le\frac{\pi}{2}\ \text{(dayside)},\\
0 & \text{if } \frac{\pi}{2} < \psi\le\pi\ \text{(nightside)}.
\end{cases} \tag*{(7)} F = { F ⋆ cos ψ 0 if 0 ≤ ψ ≤ 2 π (dayside) , if 2 π < ψ ≤ π (nightside) . (7)
In the above equation, F ⋆ F_{\star} F ⋆ designates the incident stellar flux at the substellar point and ψ \psi ψ the stellar zenithal angle, namely the angle between the planet-star direction and the position vector of any point on the planet’s surface with respect to the planet’s centre of mass. In order to compute the weighting coefficients of the flux associated with Hough functions (Paper I, Eq. (75)), we proceed as detailed in Appendix C.
First, F F F is expanded in series of Legendre polynomials (Abramowitz & Stegun [1 ] , Chapter 22). Second, by virtue of the Legendre polynomial addition theorem (e.g. Arfken & Weber [2 ] , Eq. (9.189), p. 600), we express it in terms of spherical harmonics. Finally, we simplify the obtained results by making use of the coplanar-circular approximation, namely assuming that the perturber’s orbit is a circle within the planet’s equatorial plane. This yields
F ( θ , φ , t ) = F ⋆ R { ∑ l = 0 + ∞ ∑ m = 0 l δ F ^ l m P l m ( cos θ ) e i m ( ω t + φ ) } , (8) F(\theta,\varphi,t)=F_{\star}\mathop{\mathcal{R}}\left\{\sum_{l=0}^{+\infty}\sum_{m=0}^{l}\delta\hat{F}_{l}^{m}P_{l}^{m}(\cos\theta)e^{im(\omega t+\varphi)}\right\}, \tag*{(8)} F ( θ , φ , t ) = F ⋆ R { l = 0 ∑ + ∞ m = 0 ∑ l δ F ^ l m P l m ( cos θ ) e im ( ω t + φ ) } , (8)
where ω = Ω − n ⋆ \omega=\Omega-n_{\star} ω = Ω − n ⋆ designates the diurnal tidal frequency, and P l m P_l^m P l m the degree-l l l , order-m m m normalised associated Legendre function (e.g. Auclair-Desrotour et al. [6 ] , Appendix A). The expansion coefficients of the flux, δ F ^ l m \delta\hat{F}_{l}^{m} δ F ^ l m , are given by
δ F ^ l m = ( 2 − δ m , 0 ) 2 F l 2 l + 1 P l m ( 0 ) , (9) \delta\hat{F}_{l}^{m}=(2-\delta_{m,0})\frac{2F_l}{2l+1}P_l^m(0), \tag*{(9)} δ F ^ l m = ( 2 − δ m , 0 ) 2 l + 1 2 F l P l m ( 0 ) , (9)
where δ m , q \delta_{m,q} δ m , q is the Kronecker delta function, such that δ m , m = 1 \delta_{m,m}=1 δ m , m = 1 and δ m , q = 0 \delta_{m,q}=0 δ m , q = 0 if m ≠ q m\ne q m = q (e.g. Arfken & Weber [2 ] , Eq. (1.20), p. 10), and the coefficients F l F_l F l are functions of the degree-l l l Legendre polynomial, P l P_l P l , expressed as
F l = 2 l + 1 2 ∫ 0 1 μ P l ( μ ) d μ . (10) F_l=\frac{2l+1}{2}\int_0^1\mu P_l(\mu)\,\mathrm{d}\mu. \tag*{(10)} F l = 2 2 l + 1 ∫ 0 1 μ P l ( μ ) d μ . (10)
The values of the latter are listed in Table 1 for degrees ranging from 0 to 15. We note that the order-m m m component of the flux is associated with the tidal frequency σ m = m ω \sigma_m=m\omega σ m = mω . Thus, m = 1 m=1 m = 1 corresponds to the diurnal tide, m = 2 m=2 m = 2 to the semidiurnal tide, m = 3 m=3 m = 3 to the terdiurnal tide, m = 4 m=4 m = 4 to the quaddiurnal tide, and so on.
Table 1. Expansion coefficients of stellar irradiance at the substellar point as series of Legendre polynomials. Notes. The expansion coefficients, given in Eq. (10) , are listed for 0 ≤ l ≤ 15 0 \le l \le15 0 ≤ l ≤ 15 .
Finally, we introduce the order-m m m spin parameter,
ν m ( Ω , n ⋆ ) = 2 Ω m ( Ω − n ⋆ ) , (11) \nu_m(\Omega,n_{\star})=\frac{2\Omega}{m(\Omega-n_{\star})}, \tag*{(11)} ν m ( Ω , n ⋆ ) = m ( Ω − n ⋆ ) 2Ω , (11)
and the coefficients
δ F ^ n m = ∑ l = m + ∞ ⟨ P l m , Θ n m , ν m ⟩ δ F ^ l m . (12) \delta\hat{F}_{n}^{m}=\sum_{l=m}^{+\infty}\left\langle P_l^m,\Theta_n^{m,\nu_m}\right\rangle\delta\hat{F}_{l}^{m}. \tag*{(12)} δ F ^ n m = l = m ∑ + ∞ ⟨ P l m , Θ n m , ν m ⟩ δ F ^ l m . (12)
In the above equations, Θ n m , ν m \Theta_n^{m,\nu_m} Θ n m , ν m denotes the degree-n n n Hough function (see Paper I, Sect. 2.4), and ⟨ ⋅ , ⋅ ⟩ \langle\cdot,\cdot\rangle ⟨ ⋅ , ⋅ ⟩ the dot product
⟨ F , G ⟩ = ∫ − 1 1 F ( μ ) G ( μ ) d μ , (13) \langle\mathcal{F},\mathcal{G}\rangle=\int_{-1}^{1}\mathcal{F}(\mu)\mathcal{G}(\mu)\,\mathrm{d}\mu, \tag*{(13)} ⟨ F , G ⟩ = ∫ − 1 1 F ( μ ) G ( μ ) d μ , (13)
where F \mathcal{F} F and G \mathcal{G} G are two functions defined for − 1 ≤ μ ≤ 1 -1\le\mu\le1 − 1 ≤ μ ≤ 1 . This allows Eq. (8) to be reformulated as
F ( θ , φ , t ) = F ⋆ R { ∑ m ∑ n δ F ^ n m Θ n m , ν m ( cos θ ) e i m ( ω t + φ ) } . (14) F(\theta,\varphi,t)=F_{\star}\mathop{\mathcal{R}}\left\{\sum_m\sum_n\delta\hat{F}_n^m\Theta_n^{m,\nu_m}(\cos\theta)e^{im(\omega t+\varphi)}\right\}. \tag*{(14)} F ( θ , φ , t ) = F ⋆ R { m ∑ n ∑ δ F ^ n m Θ n m , ν m ( cos θ ) e im ( ω t + φ ) } . (14)
Therefore, the coefficient of the flux associated with the degree-n n n Hough function in the thermal forcing (Paper I, Eq. (75)) is expressed, for the semidiurnal tide, as
δ F n = F ⋆ ∑ l = 2 + ∞ ⟨ P l 2 , Θ n 2 , ν 2 ⟩ δ F ^ l 2 . (15) \delta F_n=F_{\star}\sum_{l=2}^{+\infty}\left\langle P_l^2,\Theta_n^{2,\nu_2}\right\rangle\delta\hat{F}_l^2. \tag*{(15)} δ F n = F ⋆ l = 2 ∑ + ∞ ⟨ P l 2 , Θ n 2 , ν 2 ⟩ δ F ^ l 2 . (15)
Thermotidal pressure oscillations
The corresponding pressure variation is obtained by setting to zero the tide-raising potential (U T ; s = 0 U_{\mathrm{T};s}=0 U T ; s = 0 ) in our closed-form solution (Paper I, Eq. (90)), which yields
δ p ~ n = i δ p ~ T ( 1 + b H ) e − ζ x ( X − i ε ) [ k n 2 + ( λ + b H ) 2 ] { ( b H + 1 ) e − b H x + S n ( 1 − γ ) − ( b H + 1 ) X 2 S n ( 1 − γ ) + ( i k n + λ − 1 ) X 2 e ( i k n + λ ) x } . (16) \delta\tilde{p}_n=\frac{i\,\delta\tilde{p}_{\mathrm{T}}(1+b_{\mathrm{H}})e^{-\zeta x}}{(X-i\varepsilon)\left[k_n^2+(\lambda+b_{\mathrm{H}})^2\right]}\left\{(b_{\mathrm{H}}+1)e^{-b_{\mathrm{H}}x}+\frac{S_n(1-\gamma)-(b_{\mathrm{H}}+1)X^2}{S_n(1-\gamma)+(ik_n+\lambda-1)X^2}e^{(ik_n+\lambda)x}\right\}. \tag*{(16)} δ p ~ n = ( X − i ε ) [ k n 2 + ( λ + b H ) 2 ] i δ p ~ T ( 1 + b H ) e − ζ x { ( b H + 1 ) e − b H x + S n ( 1 − γ ) + ( i k n + λ − 1 ) X 2 S n ( 1 − γ ) − ( b H + 1 ) X 2 e ( i k n + λ ) x } . (16)
In the above equation, x x x is the dimensionless vertical coordinate (Paper I, Eq. (8)); γ \gamma γ , ζ \zeta ζ , and λ \lambda λ are dimensionless parameters given by (Paper I, Eqs. (30) and (77))
ζ = { − κ (isentropic) , 0 (isothermal) , γ = ζ + κ , λ = 1 − i α ( 1 + ζ ) 2 ( 1 − i α ) , (17) \zeta=
\begin{cases}
-\kappa& \text{(isentropic)},\\
0 & \text{(isothermal)},
\end{cases}
\qquad
\gamma=\zeta+\kappa,
\qquad
\lambda=\frac{1-i\alpha(1+\zeta)}{2(1-i\alpha)}, \tag*{(17)} ζ = { − κ 0 (isentropic) , (isothermal) , γ = ζ + κ , λ = 2 ( 1 − i α ) 1 − i α ( 1 + ζ ) , (17)
with κ = ( Γ 1 − 1 ) / Γ 1 \kappa=(\Gamma_1-1)/\Gamma_1 κ = ( Γ 1 − 1 ) / Γ 1 being a function of the adiabatic index, Γ 1 \Gamma_1 Γ 1 , and α = σ c / σ \alpha=\sigma_{\mathrm{c}}/\sigma α = σ c / σ denoting the ratio between the Newtonian cooling frequency, σ c \sigma_{\mathrm{c}} σ c , and the tidal frequency, σ \sigma σ . It is worth noting that, in practice, σ c \sigma_{\mathrm{c}} σ c should be regarded as an effective parameter encapsulating all dissipative processes that are not explicitly represented in the model, including radiative transfer, wind-induced broadening, turbulent diffusion within the boundary layer, and spatial inhomogeneities.
The parameter b H b_{\mathrm{H}} b H in Eq. (16) is the vertical variation rate of the considered exponentially decaying tidal heating profile (Paper I, Eq. (74)), S n S_n S n is the sign of the eigenvalue associated with the degree-n n n Hough function,
S n = sign ( Λ n m , ν ) , (18) S_n=\operatorname{sign}\left(\Lambda_n^{m,\nu}\right), \tag*{(18)} S n = sign ( Λ n m , ν ) , (18)
and k n k_n k n is the vertical wavenumber of the degree-n n n mode, such that (Paper I, Eq. (78))
k n 2 = γ β s Λ n m , ν − i α ζ 1 − i α − λ 2 , (19) k_n^2=\frac{\gamma\beta_{\mathrm{s}}\Lambda_n^{m,\nu}-i\alpha\zeta}{1-i\alpha}-\lambda^2, \tag*{(19)} k n 2 = 1 − i α γ β s Λ n m , ν − i α ζ − λ 2 , (19)
where β s = g H s / ( R 2 σ 2 ) \beta_{\mathrm{s}}=gH_{\mathrm{s}}/(R^2\sigma^2) β s = g H s / ( R 2 σ 2 ) is a dimensionless parameter involving the pressure scale height at the planet’s surface, H s H_{\mathrm{s}} H s . A limitation of the present formulation is that b H b_{\mathrm{H}} b H is assumed to be a real-valued parameter, such that the corresponding tidal-heating profile is everywhere in phase within the atmospheric column. The model is therefore unable to discriminate between heating induced by stratospheric ozone absorption and that arising from the surface, although these processes are known to produce tidal responses with different phase signatures (e.g. Chapman & Lindzen [10 ] ). Within the linear regime, this shortcoming can nevertheless be addressed to some extent by representing forcings with distinct phase lags as separate contributions.
Additionally, we have introduced in Eq. (16) the normalised tidal frequency X X X , damping rate ε \varepsilon ε , and thermotidal pressure variation δ p ~ T \delta\tilde{p}_{\mathrm{T}} δ p ~ T , defined as
X = σ σ L ; n , ε = σ C σ L ; n , δ p ~ T = κ η δ F n p s H s σ L ; n , (20) X = \frac{\sigma}{\sigma_{\mathrm{L};n}},\quad\varepsilon= \frac{\sigma^{\mathrm{C}}}{\sigma_{\mathrm{L};n}},\quad\delta\tilde{p}_{\mathrm{T}} = \frac{\kappa\eta\delta F_n}{p_{\mathrm{s}}H_{\mathrm{s}}\sigma_{\mathrm{L};n}}, \tag*{(20)} X = σ L ; n σ , ε = σ L ; n σ C , δ p ~ T = p s H s σ L ; n κ η δ F n , (20)
where η \eta η – such that 0 ≤ η ≤ 1 0 \le\eta\le1 0 ≤ η ≤ 1 – designates the fraction of incident stellar flux absorbed by the atmosphere, and σ L ; n \sigma_{\mathrm{L};n} σ L ; n the resonance frequency of the degree-n n n tidally excited Lamb wave (Paper I, Eq. (97)),
σ L ; n = 1 R g H s ∣ Δ n m , ν ∣ 1 − γ . (21) \sigma_{\mathrm{L};n} = \frac{1}{R}\sqrt{\frac{gH_{\mathrm{s}}\lvert\Delta_n^{m,\nu}\rvert}{1-\gamma}}. \tag*{(21)} σ L ; n = R 1 1 − γ g H s ∣ Δ n m , ν ∣ . (21)
We emphasise that the three nondimensional parameters given by Eq. (20) depend on the degree of the excited tidal mode, like k n k_n k n and S n S_n S n . However, the subscript n n n is omitted in notations in order to lighten expressions, noting that each mode can be considered individually. It should also be noted that the parameter δ p ~ T \delta\tilde{p}_{\mathrm{T}} δ p ~ T , as defined in Eq. (20) , pertains specifically to the regime of direct atmospheric absorption, in which tidal heating arises from the absorption of incident stellar radiation in the visible range (this corresponds to the G = 1 G = 1 G = 1 limit case of Farhat et al. [20 ] ). The analysis presented below can nevertheless be readily extended to cases involving indirect thermal forcing by introducing the transfer function that relates the tidal heat flux to the incident stellar flux and applying it as a multiplicative correction to δ p ~ T \delta\tilde{p}_{\mathrm{T}} δ p ~ T , as discussed in Sect. 2.7.
Similar to stellar irradiance, we obtain the quadrupolar component of the surface pressure variation needed to compute the tidal torque, δ p ^ s ; 2 2 , σ \hat{\delta p}_{\mathrm{s};2}^{2,\sigma} δ p ^ s ; 2 2 , σ (see Eq. (4) ), by expanding Hough functions in series of normalised associated Legendre functions, and noting that
Y l m ( θ , φ ) = 1 2 π P l m ( cos θ ) e i m φ . (22) Y_l^m(\theta,\varphi) = \frac{1}{\sqrt{2\pi}}P_l^m(\cos\theta)e^{im\varphi}. \tag*{(22)} Y l m ( θ , φ ) = 2 π 1 P l m ( cos θ ) e im φ . (22)
It follows
δ p ^ s ; 2 2 , σ = 2 π ∑ n ⟨ Θ n 2 , ν 2 , P 2 2 ⟩ δ p ~ s ; n . (23) \hat{\delta p}_{\mathrm{s};2}^{2,\sigma} = \sqrt{2\pi}\sum_n \langle\Theta_n^{2,\nu_2},P_2^2\rangle\tilde{\delta p}_{\mathrm{s};n}. \tag*{(23)} δ p ^ s ; 2 2 , σ = 2 π n ∑ ⟨ Θ n 2 , ν 2 , P 2 2 ⟩ δ p ~ s ; n . (23)
The Fourier component of the corresponding surface pressure oscillation at the equator, introduced in Eq. (5) , is readily deduced from Eq. (23) ,
δ p s ; e q σ = p s c f 15 16 ∑ n ⟨ Θ n 2 , ν 2 , P 2 2 ⟩ δ p ~ s ; n . (24) \delta p_{\mathrm{s};\mathrm{eq}}^\sigma= \frac{p_{\mathrm{s}}}{c_f}\sqrt{\frac{15}{16}}\sum_n \langle\Theta_n^{2,\nu_2},P_2^2\rangle\tilde{\delta p}_{\mathrm{s};n}. \tag*{(24)} δ p s ; eq σ = c f p s 16 15 n ∑ ⟨ Θ n 2 , ν 2 , P 2 2 ⟩ δ p ~ s ; n . (24)
In the above equations, δ p ~ s ; n = δ p ~ n ∣ x = 0 \tilde{\delta p}_{\mathrm{s};n} = \left.\tilde{\delta p}_n\right|_{x=0} δ p ~ s ; n = δ p ~ n x = 0 designates the surface pressure variation induced by the degree-n n n mode.
The values of the dot products ⟨ Θ n 2 , ν 2 , P 2 2 ⟩ \langle\Theta_n^{2,\nu_2},P_2^2\rangle ⟨ Θ n 2 , ν 2 , P 2 2 ⟩ and ⟨ P 2 2 , Θ n 2 , ν 2 ⟩ \langle P_2^2,\Theta_n^{2,\nu_2}\rangle ⟨ P 2 2 , Θ n 2 , ν 2 ⟩ are given by Table B.1 of Paper I for Earth’s Solar semidiurnal tide, namely m = 2 m = 2 m = 2 and ν 2 = 1 \nu_2 = 1 ν 2 = 1 (neglecting the difference between the Solar and sidereal days). We note that the Hough function associated with the predominating gravity mode of the Earth’s semidiurnal tide, Θ 2 2 , ν 2 \Theta_2^{2,\nu_2} Θ 2 2 , ν 2 , is very similar to the corresponding normalised associated Legendre function, P 2 2 P_2^2 P 2 2 , as their overlap coefficient is ⟨ Θ 2 2 , ν 2 , P 2 2 ⟩ = 0.969613 ≃ 1 \langle\Theta_2^{2,\nu_2},P_2^2\rangle= 0.969613 \simeq1 ⟨ Θ 2 2 , ν 2 , P 2 2 ⟩ = 0.969613 ≃ 1 (Paper I, Table B.1). In the tidal torque exerted about the planet’s spin axis, the contribution of the predominant gravity mode is thus weighted by the factor ⟨ Θ 2 2 , ν 2 , P 2 2 ⟩ 2 = 0.940149 \langle\Theta_2^{2,\nu_2},P_2^2\rangle^2 = 0.940149 ⟨ Θ 2 2 , ν 2 , P 2 2 ⟩ 2 = 0.940149 , meaning that it would represents 94% of the tidally dissipated energy if the modal response were independent of frequency. The associated eigenvalue is given by Λ 2 2 , 1 = 11.12895 \Lambda_2^{2,1} = 11.12895 Λ 2 2 , 1 = 11.12895 (Paper I, Table B.1). As a last step, we shall establish from Eq. (16) the analytical expressions of δ p ~ s ; n \tilde{\delta p}_{\mathrm{s};n} δ p ~ s ; n for isentropic and isothermal atmospheres.
Isentropic atmosphere
The isentropic atmosphere is the simpler of the two configurations described by the closed-form solution of Paper I. Here, the control parameters associated with the atmospheric structure take the values ζ = − κ \zeta= -\kappa ζ = − κ and γ = 0 \gamma= 0 γ = 0 (see Eq. (17) ). Therefore, the vertical wavenumber of a tidal mode is readily deduced from Eq. (19) . As ℑ { k n } > 0 \Im\{k_n\} > 0 ℑ { k n } > 0 by convention, this wavenumber is expressed as
k n = i 2 [ X − i ε ( 1 + κ ) X − i ε ] . (25) k_n = \frac{i}{2}\left[\frac{X-i\varepsilon(1+\kappa)}{X-i\varepsilon}\right]. \tag*{(25)} k n = 2 i [ X − i ε X − i ε ( 1 + κ ) ] . (25)
The vertical profile of the degree-n n n tidal pressure variations, given by Eq. (16) , thus becomes
δ p ~ n = i δ p ~ T e κ x b H ( X − i ε ) + i κ ε { ( b H + 1 ) exp ( − b H x ) + [ S n − ( b H + 1 ) X 2 ( X − i ε ) ( X 2 − S n ) − i κ ε X 2 ] exp ( i κ ε X − i ε x ) } . (26) \tilde{\delta p}_n = \frac{i\delta\tilde{p}_{\mathrm{T}}e^{\kappa x}}{b_{\mathrm{H}}(X-i\varepsilon)+i\kappa\varepsilon}\left\{(b_{\mathrm{H}}+1)\exp(-b_{\mathrm{H}}x)+\left[\frac{S_n-(b_{\mathrm{H}}+1)X^2}{(X-i\varepsilon)(X^2-S_n)-i\kappa\varepsilon X^2}\right]\exp\left(\frac{i\kappa\varepsilon}{X-i\varepsilon}x\right)\right\}. \tag*{(26)} δ p ~ n = b H ( X − i ε ) + iκ ε i δ p ~ T e κ x { ( b H + 1 ) exp ( − b H x ) + [ ( X − i ε ) ( X 2 − S n ) − iκ ε X 2 S n − ( b H + 1 ) X 2 ] exp ( X − i ε iκ ε x ) } . (26)
and the corresponding surface pressure variation simplifies to
δ p ~ s ; n = S n δ p ~ T ε [ ( 1 + κ ) X 2 − S n ] + i X ( X 2 − S n ) . (27) \tilde{\delta p}_{\mathrm{s};n} = \frac{S_n\delta\tilde{p}_{\mathrm{T}}}{\varepsilon\left[(1+\kappa)X^2-S_n\right]+iX\left(X^2-S_n\right)}. \tag*{(27)} δ p ~ s ; n = ε [ ( 1 + κ ) X 2 − S n ] + i X ( X 2 − S n ) S n δ p ~ T . (27)
The key components of this analytical solution can be retrieved in Table 2 . Interestingly, the surface pressure oscillation induced by a tidal mode is independent of b H b_{\mathrm{H}} b H , meaning that the vertical distribution of tidal heating does not affect the tidal torque exerted on the planet. This results from the assumed atmospheric structure. First, the Archimedean forces vanish since the atmosphere is neutrally stratified. Second, we have filtered out Coriolis and compressibility terms in the vertical momentum equation of tidal waves (Paper I, Eq. (14)), which implies that fluid motions are solely driven by the horizontal terms of these two restoring forces. As a consequence, atmospheric tides are analogous to ocean’s barotropic tides in this configuration.
Table 2. Transfer functions relating the surface pressure oscillations to the incident stellar flux in the Fourier domain.
Isothermal atmosphere
For an isothermal atmosphere, the structure control parameters take the values ζ = 0 \zeta= 0 ζ = 0 and γ = κ \gamma= \kappa γ = κ , which implies λ = 1 / 2 \lambda= 1/2 λ = 1/2 (see Eq. (17) ). Due to stable stratification, the dependence of the vertical wavenumber on tidal frequency is significantly more complex than in the isentropic configuration. Derivations yield
k n ( X ) = μ ( X ) 2 2 [ S n + i ε X μ ( X ) ( X 2 + ε 2 ) ( ϑ X ) 2 ] sign ( X ) , (28) k_n(X) = \frac{\sqrt{\mu(X)}}{2\sqrt{2}}\left[S_n+i\frac{\varepsilon X}{\mu(X)(X^2+\varepsilon^2)}\left(\frac{\vartheta}{X}\right)^2\right]\operatorname{sign}(X), \tag*{(28)} k n ( X ) = 2 2 μ ( X ) [ S n + i μ ( X ) ( X 2 + ε 2 ) εX ( X ϑ ) 2 ] sign ( X ) , (28)
Notes. Comparison of the analytical expressions of the quantity in the isentropic (left column) and isothermal (right column) atmospheric solutions. The resonance frequency of the degree-n n n mode, σ L , n \sigma_{\mathrm{L},n} σ L , n , is defined by Eq. (21) , while X X X , ε \varepsilon ε and δ p ~ T \widetilde{\delta p}_{\mathrm{T}} δ p T are defined by Eq. (20) , ϑ \vartheta ϑ by Eq. (29) , μ \mu μ by Eq. (30) , k n k_n k n by Eqs. (25) and (28) , and δ p ~ n \widetilde{\delta p}_n δ p n by Eqs. (26) and (31) . The surface pressure variation is given by Eq. (27) for isentropic atmospheres, and by Eqs. (32) and (33) for isothermal atmospheres.
ϑ = 2 κ ( 1 − κ ) (29) \vartheta=2\sqrt{\kappa(1-\kappa)} \tag*{(29)} ϑ = 2 κ ( 1 − κ ) (29)
where we have introduced the parameter
and the function
μ ( X ) = S n X 2 X 2 + ε 2 ( ϑ X ) 2 − 1 + [ ( ϑ X ) 2 − S n X 2 + ε 2 X 2 + ε 2 ] 1 / 2 . (30) \mu(X)=\dfrac{S_nX^2}{X^2+\varepsilon^2}\left(\dfrac{\vartheta}{X}\right)^2-1+\left[\dfrac{\left(\dfrac{\vartheta}{X}\right)^2-S_n}{X^2+\varepsilon^2}X^2+\varepsilon^2\right]^{1/2}. \tag*{(30)} μ ( X ) = X 2 + ε 2 S n X 2 ( X ϑ ) 2 − 1 + X 2 + ε 2 ( X ϑ ) 2 − S n X 2 + ε 2 1/2 . (30)
The tidal pressure variation associated with the degree-n n n mode is thus expressed as
δ p ~ n = i δ p ~ T ( b H + 1 ) ( X − i ε ) [ k n 2 + ( 1 2 + b H ) 2 ] { ( b H + 1 ) exp ( − b H x ) + [ S n − Γ 1 ( b H + 1 ) X 2 ] ( i k n − 1 2 ) S n + Γ 1 ( i k n − 1 2 ) X 2 exp [ ( i k n + 1 2 ) x ] } . (31) \widetilde{\delta p}_n=\dfrac{i\widetilde{\delta p}_{\mathrm{T}}(b_{\mathrm{H}}+1)}{(X-i\varepsilon)\left[k_n^2+\left(\frac{1}{2}+b_{\mathrm{H}}\right)^2\right]}\left\{(b_{\mathrm{H}}+1)\exp(-b_{\mathrm{H}}x)+\dfrac{[S_n-\Gamma_1(b_{\mathrm{H}}+1)X^2]\left(ik_n-\frac{1}{2}\right)}{S_n+\Gamma_1\left(ik_n-\frac{1}{2}\right)X^2}\exp\left[\left(ik_n+\frac{1}{2}\right)x\right]\right\}. \tag*{(31)} δ p n = ( X − i ε ) [ k n 2 + ( 2 1 + b H ) 2 ] i δ p T ( b H + 1 ) { ( b H + 1 ) exp ( − b H x ) + S n + Γ 1 ( i k n − 2 1 ) X 2 [ S n − Γ 1 ( b H + 1 ) X 2 ] ( i k n − 2 1 ) exp [ ( i k n + 2 1 ) x ] } . (31)
and the corresponding surface pressure variation as
δ p ~ s ; n = i S n δ p ~ T ( 1 + b H ) ( i k n + 1 2 + b H ) X [ 1 4 S n ϑ 2 + b H ( b H + 1 ) X ( X − i ε ) ] [ S n + Γ 1 ( i k n − 1 2 ) X 2 ] . (32) \widetilde{\delta p}_{\mathrm{s};n}=\dfrac{iS_n\widetilde{\delta p}_{\mathrm{T}}(1+b_{\mathrm{H}})\left(ik_n+\frac{1}{2}+b_{\mathrm{H}}\right)X}{\left[\frac{1}{4}S_n\vartheta^2+b_{\mathrm{H}}(b_{\mathrm{H}}+1)X(X-i\varepsilon)\right]\left[S_n+\Gamma_1\left(ik_n-\frac{1}{2}\right)X^2\right]}. \tag*{(32)} δ p s ; n = [ 4 1 S n ϑ 2 + b H ( b H + 1 ) X ( X − i ε ) ] [ S n + Γ 1 ( i k n − 2 1 ) X 2 ] i S n δ p T ( 1 + b H ) ( i k n + 2 1 + b H ) X . (32)
Unlike isentropic atmospheres, isothermal atmospheres permit the propagation of internal waves in the vertical direction. Because of the interplay between the vertical wavelength of these waves and the thickness of the heated region, the surface pressure oscillations explicitly depend on b H b_{\mathrm{H}} b H in Eq. (32) . To remove this dependence and thus reduce the dimensionality of the model, the heated layer must be much thinner than the vertical wavelength of the tidal modes. Since k n k_n k n tends to + ∞ +\infty + ∞ in the zero-frequency limit, the thickness of this layer must consequently vanish, implying b H = + ∞ b_{\mathrm{H}}=+\infty b H = + ∞ . This limiting case corresponds to the Dirac-type heating profile adopted by Dobrovolskis & Ingersoll [18 ] in their model of Venus’ thermal tides. With b H = + ∞ b_{\mathrm{H}}=+\infty b H = + ∞ , Eq. (32) simplifies to
δ p ~ s ; n = i S n δ p ~ T ( X − i ε ) [ S n + Γ 1 ( i k n − 1 2 ) X 2 ] . (33) \widetilde{\delta p}_{\mathrm{s};n}=\dfrac{iS_n\widetilde{\delta p}_{\mathrm{T}}}{(X-i\varepsilon)\left[S_n+\Gamma_1\left(ik_n-\frac{1}{2}\right)X^2\right]}. \tag*{(33)} δ p s ; n = ( X − i ε ) [ S n + Γ 1 ( i k n − 2 1 ) X 2 ] i S n δ p T . (33)
Application to the present-day Earth
We conclude this section by benchmarking the model predictions against observational constraints for semidiurnal tides on the present-day Earth. This case, which corresponds to the G = 1 G=1 G = 1 limit of Farhat et al. [20 ] , is representative of the regime of direct absorption that will be analytically characterised in Sect. 3. From an extensive analysis of global barometric records, Schindelegger & Ray (2014) derived estimates for both the modulus and the phase lag of the semidiurnal surface pressure oscillations driven by Solar thermal tides. They report a value of ∣ δ p s e q σ ∣ = 123.4 P a |\delta p_{\mathrm{seq}}^{\sigma}| = 123.4\ \mathrm{Pa} ∣ δ p seq σ ∣ = 123.4 Pa at the equator, with the maximum amplitude occurring at t max = 09 : 45 L S T t_{\max} = 09{:}45\ \mathrm{LST} t m a x = 09 : 45 LST .
As these experimental estimates agree well with earlier measurements (e.g. Haurwitz [27 ] ; Haurwitz & Cowley [28 ] ; Lindzen [33 ] ; Dai & Wang [16 ] ) and GCM simulations (e.g. Covey et al. [14 , 13 ] ), we adopt them as reference values for experimental constraints. However, these estimates shall not be considered as precise. Particularly, earlier empirical studies end up with lower amplitudes, such as 116.1 Pa (Haurwitz & Cowley [28 ] ) or 107.9 Pa (Dai & Wang [16 ] ), indicating a systematic uncertainty of 10–15% in experimental calibration. Through an exhaustive comparative analysis of climate models for the Earth’s diurnal and semidiurnal tides, Covey et al. [14 ] and Covey et al. [13 ] show that GCM results exhibit a similar variability. Using ∣ δ p s e q σ ∣ = 123.4 P a |\delta p_{\mathrm{seq}}^{\sigma}| = 123.4\ \mathrm{Pa} ∣ δ p seq σ ∣ = 123.4 Pa and t max = 09 : 45 L S T t_{\max} = 09{:}45\ \mathrm{LST} t m a x = 09 : 45 LST in Eq. (6) , together with R = 6371 k m R = 6371\ \mathrm{km} R = 6371 km , a = 1.0 a u a = 1.0\ \mathrm{au} a = 1.0 au , and M ⋆ / M = 3.329 × 10 5 M_{\star}/M = 3.329 \times10^{5} M ⋆ / M = 3.329 × 1 0 5 (Luzum et al. [36 ] ), yields an associated thermotidal torque of T z = 3.51 × 10 15 N m T_z = 3.51 \times10^{15}\ \mathrm{N\,m} T z = 3.51 × 1 0 15 N m . This estimate is affected by the uncertainty of empirical data and GCM simulations, which makes it range between T z = 2.5 × 10 15 N m T_z = 2.5 \times10^{15}\ \mathrm{N\,m} T z = 2.5 × 1 0 15 N m (Zahnle & Walker [53 ] ) and T z = 4.14 × 10 15 N m T_z = 4.14 \times10^{15}\ \mathrm{N\,m} T z = 4.14 × 1 0 15 N m (Wu et al. 2023, Fig. S4 in Supplementary Materials). In what follows, we assess how well the analytical model reproduces the observed modulus and phase of the surface pressure oscillations for both the isentropic and isothermal equilibria.
To facilitate a consistent comparison, we consider atmospheres with the same total thermal energy by fixing the mass-averaged temperature of the air column – hereafter the bulk temperature – as defined in Eq. (71) of Paper I. This temperature, T ‾ \overline{T} T , is given by
T ‾ = T s 1 − ζ = { T s 1 + κ (isentropic) , T s (isothermal) , (34) \overline{T} = \frac{T_s}{1-\zeta} =
\begin{cases}
\frac{T_s}{1+\kappa} & \text{(isentropic)},\\
T_s & \text{(isothermal)},
\end{cases} \tag*{(34)} T = 1 − ζ T s = { 1 + κ T s T s (isentropic) , (isothermal) , (34)
where T s T_s T s denotes the atmospheric temperature at the planet’s surface. Adopting T s = 288 K T_s = 288\ \mathrm{K} T s = 288 K for present-day Earth (U.S. Standard Atmosphere Working Group [47 ] ), we obtain T ‾ = 243 K \overline{T} = 243\ \mathrm{K} T = 243 K for the isentropic configuration via Eq. (34) , which defines the corresponding equivalent isothermal atmosphere.
We use the standard Earth values for the surface gravity and the mean Solar irradiance, namely g = 9.81 m s − 2 g = 9.81\ \mathrm{m\,s^{-2}} g = 9.81 m s − 2 and F ⋆ = 1361 W m − 2 F_{\star} = 1361\ \mathrm{W\,m^{-2}} F ⋆ = 1361 W m − 2 (Coddington et al. [11 ] ). For the atmospheric parameters, we adopt the prescription of the U.S. Standard Atmosphere (U.S. Standard Atmosphere Working Group [47 ] ), adopting p s = 1.013 b a r p_s = 1.013\ \mathrm{bar} p s = 1.013 bar , R s = 287 J k g − 1 K − 1 R_s = 287\ \mathrm{J\,kg^{-1}\,K^{-1}} R s = 287 J k g − 1 K − 1 , and Γ 1 = 1.23 \Gamma_1 = 1.23 Γ 1 = 1.23 (Paper I, Figure 2 ). Finally, the temperature at the planetary surface sets the associated pressure scale height,
Figure 2. Predictions of the isentropic (left) and isothermal (right) models for actual Earth and various values of the damping ratio. From top to bottom: Modulus of surface pressure oscillations at the equator, Solar time at which they are maximal (t max t_{\max} t m a x ), and imaginary component. Solid lines denote the model predictions, which are computed using Eqs. (27) and (33) . The black marker labelled ‘SR2014’ indicates the constraint inferred by Schindelegger & Ray [43 ] for present-day Earth on the basis of barometric measurements. Parameters are g = 9.81 m s − 2 g = 9.81\ \mathrm{m\ s^{-2}} g = 9.81 m s − 2 , R = 6371 k m R = 6371\ \mathrm{km} R = 6371 km , Γ 1 = 1.23 \Gamma_1 = 1.23 Γ 1 = 1.23 , R s = 287 J k g − 1 K − 1 R_s = 287\ \mathrm{J\ kg^{-1}\ K^{-1}} R s = 287 J k g − 1 K − 1 , p s = 1.013 b a r p_s = 1.013\ \mathrm{bar} p s = 1.013 bar , F ⋆ = 1361 W m − 2 F_\star= 1361\ \mathrm{W\ m^{-2}} F ⋆ = 1361 W m − 2 , T ‾ = 243 K \overline{T} = 243\ \mathrm{K} T = 243 K .
H s = R s T s g , (35) H_s = \frac{R_s T_s}{g}, \tag*{(35)} H s = g R s T s , (35)
yielding H s = 8.43 k m H_s = 8.43\ \mathrm{km} H s = 8.43 km for the isentropic equilibrium and H s = 7.10 k m H_s = 7.10\ \mathrm{km} H s = 7.10 km for the isothermal case.
These parameters allow us to evaluate the Lamb-wave resonance frequency of the dominant tidal mode (m = 2 m = 2 m = 2 , n = 2 n = 2 n = 2 , ν = 1 \nu= 1 ν = 1 ), defined in the general case by Eq. (21) and given by
σ L ; 2 = 1 R g H s ∣ Λ 2 2 , 1 ∣ 1 − γ . (36) \sigma_{\mathrm{L;2}} = \frac{1}{R}\sqrt{\frac{gH_s|\Lambda_2^{2,1}|}{1-\gamma}}. \tag*{(36)} σ L ; 2 = R 1 1 − γ g H s ∣ Λ 2 2 , 1 ∣ . (36)
In this expression, we neglect the small difference between the Solar and sidereal day, thereby removing the weak frequency dependence of the eigenvalue associated with the degree-2 mode. The latter takes the value Λ 2 2 , 1 = 11.12895 \Lambda_2^{2,1} = 11.12895 Λ 2 2 , 1 = 11.12895 (Paper I, Table B.1), which is positive, indicating that the mode can be resonantly excited (S n = + 1 S_n = +1 S n = + 1 ) in Eqs. (27) and (33) . For the isentropic configuration, where γ = 0 \gamma= 0 γ = 0 , we obtain σ L ; 2 = 1.51 × 10 − 4 r a d s − 1 \sigma_{\mathrm{L;2}} = 1.51 \times10^{-4}\ \mathrm{rad\,s^{-1}} σ L ; 2 = 1.51 × 1 0 − 4 rad s − 1 . In the isothermal case, γ = κ \gamma= \kappa γ = κ yields σ L ; 2 = 1.53 × 10 − 4 r a d s − 1 \sigma_{\mathrm{L;2}} = 1.53 \times10^{-4}\ \mathrm{rad\,s^{-1}} σ L ; 2 = 1.53 × 1 0 − 4 rad s − 1 . As shown in Paper I (Eq. (100)), the ratio between the isothermal and isentropic values – denoted σ L ; 2 i s o \sigma_{\mathrm{L;2}}^{\mathrm{iso}} σ L ; 2 iso and σ L ; 2 i s e n \sigma_{\mathrm{L;2}}^{\mathrm{isen}} σ L ; 2 isen , respectively – is
σ L ; 2 i s o σ L ; 2 i s e n = Γ 1 1 + κ ≈ 1.02. (37) \frac{\sigma_{\mathrm{L;2}}^{\mathrm{iso}}}{\sigma_{\mathrm{L;2}}^{\mathrm{isen}}}
= \sqrt{\frac{\Gamma_1}{1+\kappa}} \approx1.02. \tag*{(37)} σ L ; 2 isen σ L ; 2 iso = 1 + κ Γ 1 ≈ 1.02. (37)
The associated length of the day,
L O D L ; 2 = 4 π σ L ; 2 , (38) \mathrm{LOD}_{\mathrm{L;2}} = \frac{4\pi}{\sigma_{\mathrm{L;2}}}, \tag*{(38)} LOD L ; 2 = σ L ; 2 4 π , (38)
is therefore 23.2 hr (isentropic) and 22.8 hr (isothermal).
These values closely match the estimate reported by Wu et al. (2023) for present-day Earth, L O D L ; 2 = 22.8 h r \mathrm{LOD}_{\mathrm{L;2}} = 22.8\ \mathrm{hr} LOD L ; 2 = 22.8 hr , which is calculated based on the change in the torque’s sign in their Fig. S4. However, they are 0.8–1.2 hr larger than the 22 hr resonant LOD obtained by fitting the analytical model to the same GCM dataset (see Sect. 4.3) and to the dataset reported by Deitrick & Goldblatt [17 ] for a slightly warmer palaeoatmosphere with a C H 4 \mathrm{CH}_4 CH 4 mixing ratio of 30 ppm (see Sect. 4.2).4 Additionally, our values significantly exceed the standard 21 hr estimate used as a reference for the Earth’s Precambrian resonance (e.g. Zahnle & Walker [53 ] ; Deitrick & Goldblatt [17 ] ). This 21 hr estimate was obtained in the framework of the classical tidal theory (e.g. Lindzen & Blake [34 ] ; Salby [42 ] ), where it can be inferred from the equivalent depth of the mode (Chapman & Lindzen [10 ] ),
h 2 = R 2 σ 2 g Λ 2 2 , 1 . (39) h_2 = \frac{R^2\sigma^2}{g\Lambda_2^{2,1}}. \tag*{(39)} h 2 = g Λ 2 2 , 1 R 2 σ 2 . (39)
Typically, Lindzen & Blake [34 ] find h 2 ≈ 10.5 k m h_2 \approx10.5\ \mathrm{km} h 2 ≈ 10.5 km by assuming an isothermal profile characterised by Γ 1 = 1.4 \Gamma_1 = 1.4 Γ 1 = 1.4 (ideal diatomic gas) and T ‾ = 256 K \overline{T} = 256\ \mathrm{K} T = 256 K . Using Eqs. (36) and (38) together with these values yields L O D L ; 2 = 20.8 h r \mathrm{LOD}_{\mathrm{L;2}} = 20.8\ \mathrm{hr} LOD L ; 2 = 20.8 hr . Similar estimates are reported by several authors solving the tidal equations numerically for realistic profiles of the Earth’s atmosphere (e.g. Lindzen & Blake [34 ] ; Salby [42 ] ; Zahnle & Walker [53 ] ). However, the complexity of these numerical solutions do not allow for elucidating the possible causes of the 2-hour LOD difference between them and our analytical theory.
As a next step, we evaluate the dimensionless factor scaling the response of the degree-2 mode, δ p ~ T \widetilde{\delta p}_T δ p T . Using Eq. (20) , it reads
δ p ~ T = κ η δ F 2 p s H s σ L ; 2 2 , (40) \widetilde{\delta p}_T = \frac{\kappa\eta\delta F_2}{p_s H_s \sigma_{\mathrm{L;2}}^2}, \tag*{(40)} δ p T = p s H s σ L ; 2 2 κ η δ F 2 , (40)
where η \eta η denotes the absorbed-flux fraction and δ F 2 \delta F_2 δ F 2 the expansion coefficient of the incident stellar flux on the degree-2 Hough function. Since tidal heating stems from the direct atmospheric absorption of the incident Solar flux by ozone and water vapour, both δ F 2 \delta F_2 δ F 2 and δ p ~ T \delta\tilde{p}_T δ p ~ T are real-valued parameters. From the absorption profiles of Lindzen [32 ] and Vichare & Rajaram [49 ] , Farhat et al. [20 ] estimate η \eta η to lie between 17.6% and 18.1%. We therefore adopt η = 18 % \eta= 18\% η = 18% . The coefficient δ F 2 \delta F_2 δ F 2 is computed using Eq. (15) , yielding
δ F 2 F ⋆ = ∑ l = 2 + ∞ δ F ^ l 2 ⟨ O 2 2 , ν 2 , P l 2 ⟩ ≈ 0.226502. (41) \frac{\delta F_2}{F_\star} = \sum_{l=2}^{+\infty} \delta\hat{F}_l^2 \langle\mathcal{O}_2^{2,\nu_2}, P_l^2\rangle\approx0.226502. \tag*{(41)} F ⋆ δ F 2 = l = 2 ∑ + ∞ δ F ^ l 2 ⟨ O 2 2 , ν 2 , P l 2 ⟩ ≈ 0.226502. (41)
This yields δ p ~ T = 8.08 × 10 − 5 \delta\tilde{p}_T = 8.08 \times10^{-5} δ p ~ T = 8.08 × 1 0 − 5 for the isentropic configuration and δ p ~ T = 9.42 × 10 − 5 \delta\tilde{p}_T = 9.42 \times10^{-5} δ p ~ T = 9.42 × 1 0 − 5 for the isothermal one. These values allow us to compute the complex Fourier coefficient of surface-pressure oscillations for the degree-2 mode at any tidal frequency, using Eqs. (27) and (33) . Finally, the tidal response at the equator, δ p s ; e q s \delta p_{s;\mathrm{eq}}^{s} δ p s ; eq s , is obtained from the degree-2 surface-pressure component, δ p ~ s ; 2 \delta\tilde{p}_{s;2} δ p ~ s ; 2 , by retaining the dominant contribution in Eq. (24) ,
δ p s ; e q s ≈ c e q p s δ p ~ s ; 2 , (42) \delta p_{s;\mathrm{eq}}^{s} \approx c_{\mathrm{eq}} p_s \delta\tilde{p}_{s;2}, \tag*{(42)} δ p s ; eq s ≈ c eq p s δ p ~ s ; 2 , (42)
where the dimensionless numerical factor c e q c_{\mathrm{eq}} c eq is
c e q = 15 16 ⟨ Θ 2 2 , ν ⟩ 2 ⟨ P 2 2 ⟩ 2 1 c f ≈ 1.019545. (43) c_{\mathrm{eq}}=\sqrt{\frac{15}{16}\frac{\langle\Theta_{2}^{2,\nu}\rangle_{2}}{\langle P_{2}^{2}\rangle_{2}}}\frac{1}{c_{\mathrm{f}}}\approx1.019545. \tag*{(43)} c eq = 16 15 ⟨ P 2 2 ⟩ 2 ⟨ Θ 2 2 , ν ⟩ 2 c f 1 ≈ 1.019545. (43)
Figure 2 displays the frequency dependence of semidiurnal surface-pressure oscillations predicted at the equator by our analytical model for both the isentropic and isothermal equilibria. We plot the oscillation amplitude, the local Solar time of the dayside maximum, and the imaginary component as functions of the LOD for various values of the damping ratio, ε \varepsilon ε . For comparison, we also include the observational constraint derived by Schindelegger & Ray (2014) from global barometric measurements. Interestingly, despite the markedly different functional forms of their responses (Eqs. (27) and (33) ), the isentropic and isothermal models exhibit qualitatively similar behaviour. The isothermal curves appear as slightly shifted versions of their isentropic counterparts, consistent with the frequency ratio given by Eq. (37) .
The isentropic predictions show excellent agreement with the observational data, in line with the fact that the isentropic profile provides a good description of the thermal structure of Earth’s troposphere (see Paper I, Figure 2 ). For ε = 0.25 \varepsilon=0.25 ε = 0.25 , the model reproduces the observed modulus, phase, and imaginary component of the semidiurnal surface-pressure oscillations (∣ δ p s ; e q ∣ = 123.4 |\delta p_{\mathrm{s;eq}}|=123.4 ∣ δ p s ; eq ∣ = 123.4 Pa and t max = 09 : 45 t_{\max}=09{:}45 t m a x = 09 : 45 ; Schindelegger & Ray 2014). The predicted oscillation amplitude is 119.6 Pa, and the Solar time of the dayside maximum is 09:46 LST. By contrast, the isothermal predictions are less accurate owing to the shift in the resonance frequency. For all damping ratio, the oscillation amplitude is ∼ 79 \sim79 ∼ 79 Pa, which is 36% smaller than the empirical value. We note, however, that the discrepancies between theory and measurements in the isothermal case are enhanced by the frequency dependence of δ p s ; e q \delta p_{\mathrm{s;eq}} δ p s ; eq .
Consequences of phase-shifted tidal heating
In Sect. 2.6, the thermal forcing inducing the semidiurnal thermal tides is assumed to be in phase with insolation, which corresponds to the regime of direct absorption: the tidal bulge arises from the absorption of the incident stellar flux. This results in two resonant peaks of the tidal torque on either side of the Lamb-wave resonance frequency, one being negative and the other positive (see Figure 2 , bottom panels). These peaks are symmetric in amplitude in the limit of quasi-adiabatic compression (ε → 0 \varepsilon\to0 ε → 0 ), and the positive peak tends to become larger than the negative one as ε \varepsilon ε increases. However GCM simulations show the opposite. Wu et al. (2023) and Deitrick & Goldblatt [17 ] find the decelerative torque (negative peak) approximately twice larger than the accelerative one (positive peak) in absolute value (see Fig. S4 in supplementary materials in the former, and Figure 1 c in the latter). In Auclair-Desrotour et al. [7 ] the positive peak vanishes and the negative one only remains (see their Fig. 3, top left panel).
Multiple factors could contribute to this predominance of the negative peak. For instance, Zahnle & Walker [53 ] recover a similar behaviour by using a tidal heating with a non-zero phase that accounts for delayed latent heat releases due to cumulonimbus activity (Lindzen [33 ] ; Hamilton [26 ] ). Their analytical solution exhibits a negative peak of larger amplitude than its positive counterpart (see their Fig. 4, dashed line). Similarly, this effect may result from radiative-diffusive heat exchanges at the planet’s surface, as discussed by Auclair-Desrotour et al. [7 ] and Farhat et al. [20 ] . Farhat et al. [20 ] notably
show in their Figs. 3 and 4 that a tidal model incorporating soil dissipative processes in a simplified manner can reproduce, at least qualitatively, the asymmetry observed in numerical solutions, although further refinement of the theory remains necessary (see also Laskar et al. [30 ] , Fig. 7). When soil diffusion — or any physical process changing the phase of the forcing — is included, the latter is delayed with respect to insolation. Consequently, the resulting atmospheric response is also delayed, which generates an offset in the imaginary part of surface pressure oscillations.
We examine here how the phase of the tidal heating affects the dependence of the resulting thermotidal torque on tidal frequency. The physical setup is that used for the present-day Earth in Sect. 2.6 and the plots of Figure 2 in the isentropic configuration. The damping ratio associated with radiative cooling is set to ε = 0.25 \varepsilon= 0.25 ε = 0.25 . For simplicity, we adopt a forcing of the form proposed by Lindzen [33 ] and Hamilton [26 ] to account for delayed latent heat releases. The vertical profile of the heat absorbed per unit mass per unit time for the degree-n n n mode is expressed as
J ~ n ( x ) = J ~ n 0 ( x ) e − i σ t H (44) \widetilde{J}_n(x) = \widetilde{J}_n^0(x)e^{-i\sigma t_{\mathrm{H}}} \tag*{(44)} J n ( x ) = J n 0 ( x ) e − iσ t H (44)
where J ~ n 0 \widetilde{J}_n^0 J n 0 is the real-valued, exponentially decaying heating profile used in the present work, given by Eq. (74) of Paper I, and t H t_{\mathrm{H}} t H a constant time lag. We note that the above model and that given by Eq. (H.4) of Farhat et al. [20 ] for soil diffusion have different frequency dependencies. The former results in larger phase shifts than the latter at high frequencies and smaller phase shifts at low frequencies, which may affect the functional form of analytical solutions and the shape of the associated frequency spectra.
Figure 3 displays the frequency dependence of surface-pressure oscillations at the planet’s equator for values of the time lag lying between 0. hr (no phase) and 2.5 hr (strongly delayed forcing). We plot the oscillation amplitude, the local Solar time of the dayside maximum, and the imaginary component as functions of the LOD. Similar to Figure 2 , the observational constraint derived by Schindelegger & Ray (2014) from barometric measurements is included. As the phase of the tidal forcing does not affect the modulus of the response, the latter is the same for all values of t H t_{\mathrm{H}} t H . Consequently, there is no way to infer the time lag of the forcing from the frequency dependence of the amplitude of surface pressure oscillations. However we retrieve the phase of the forcing as an offset in the phase of the tidal response. Besides, the phase influences the relative predominances of the positive and negative peaks of the resulting thermotidal torque, which is proportional to the imaginary component of δ p s ; e q \delta p_{\mathrm{s;eq}} δ p s ; eq .
Figure 3. Influence of phase-shifted forcings on the tidal response of isentropic atmospheres for various values of the time lag. From top to bottom: Modulus of surface pressure oscillations at the equator, Solar time at which they are maximum (t max t_{\max} t m a x ), and imaginary component. Solid lines denote the model predictions, which are computed using Eq. (27) together with Eq. (44) . The black marker labelled ‘SR2014’ indicates the constraint inferred by Schindelegger & Ray [43 ] for present-day Earth on the basis of barometric measurements. Parameters are g = 9.81 g=9.81 g = 9.81 m s− 2 ^{-2} − 2 , R = 6371 R=6371 R = 6371 km, Γ 1 = 1.23 \Gamma_{1}=1.23 Γ 1 = 1.23 , R = 287 R=287 R = 287 J kg− 1 ^{-1} − 1 K− 1 ^{-1} − 1 , p s = 1013 p_{\mathrm{s}}=1013 p s = 1013 bar, F ⋆ = 1361 F_{\star}=1361 F ⋆ = 1361 W m− 2 ^{-2} − 2 , T s = 288 T_{\mathrm{s}}=288 T s = 288 K, ε = 0.25 \varepsilon=0.25 ε = 0.25 .
As the time lag increases, the positive peak responsible for the thermodynamic acceleration of the Earth’s rotation diminishes, while the negative peak increases in amplitude. We remark that the shape of the numerical solutions obtained from GCM simulations by Deitrick & Goldblatt [17 ] in their fig:1c and Wu et al. (2023) in their fig:S4 , where the braking peak is roughly twice larger than the accelerating one, is approximately reproduced by the analytical model for t H = 1.0 t_{\mathrm{H}} = 1.0 t H = 1.0 hr. Therefore, strong pressure oscillations do not necessarily result in a strong thermodynamic acceleration of the planet’s rotation. Particularly, a tidal lock such as that hypothesised for the Precambrian Earth (Zahnle & Walker [53 ] ; Bartlett & Stevenson [9 ] ; Mitchell & Kirscher [37 ] ; Wu et al. 2023; Deitrick & Goldblatt [17 ] ) turns out to be less likely if the tidal heating is phase-shifted than in the case of direct absorption.
Frequency behaviour
Pressure variations spectra
The analytical expressions of tidal surface pressure variations established in Sect. 2 allow the frequency behaviour of the thermotidal torque to be explored exhaustively. Here, we focus on fast rotating Earth-like planets, as the present study builds on a series of works dedicated to the past evolution of the Earth-Moon system (Farhat et al. [21 , 20 , 22 ] ; Auclair-Desrotour et al. [4 , 3 ] ; Laskar et al. [30 ] ).
It is worth noting that we use the expression ‘fast rotation’ to designate configurations in which the central body’s spin angular velocity far exceeds any other characteristic frequency of the orbital system — for example, the perturber’s mean motion (n ⋆ ≪ Ω n_{\star} \ll\Omega n ⋆ ≪ Ω ). In the case of Earth, the fast-rotation approximation is generally well justified for semidiurnal tides, provided that the rotation period remains much shorter than one year, a condition that has been satisfied throughout the planet’s history. By contrast, this criterion is not necessarily met in close-in star-planet systems, where a planet approaching tidal lock may rotate rapidly while its spin rate remains only moderately larger than its orbital mean motion. Nevertheless, in the context of semidiurnal tidal torques, this restriction appears to be primarily a theoretical limitation. Indeed, the torque predicted by the analytical model remains consistent with GCM results in the near-synchronous regime, where the fast-rotation approximation formally breaks down. In particular, the torque decreases linearly to zero as X → 0 X \to0 X → 0 , in agreement with the GCM simulations reported by Leconte et al. [31 ] in the Venus’ case, as discussed in Sect. 3.3.
These findings suggest that, in practice, the frequency analysis developed in this section may remain applicable beyond the formal limits of the fast-rotation regime, despite having been derived under the assumption that n ⋆ ≪ Ω n_{\star} \ll\Omega n ⋆ ≪ Ω . Whether this apparent robustness extends to near-synchronously rotating planets in close-in systems nevertheless remains an open question. To our knowledge, the frequency dependence of the thermotidal torque has not yet been investigated using GCM simulations for such planets. Consequently, the applicability of the analytical model in this regime remains to be established, and its predictions should therefore be interpreted cautiously outside the formal domain of validity of the fast-rotation approximation.
When fast rotation is assumed, the parameters ν \nu ν , Λ n m , ν \Lambda_n^{m,\nu} Λ n m , ν , σ L , n \sigma_{L,n} σ L , n , S n S_n S n , ε \varepsilon ε and δ p T \delta p_{\mathrm{T}} δ p T become independent of the tidal frequency. Consequently, δ p S , n \delta p_{\mathrm{S},n} δ p S , n can be expressed explicitly as a function of X X X in Eqs. (27) and (33) . Also, the asymmetric frequency dependence of the tidal response with respect to synchronisation (X = 0 X = 0 X = 0 ), which is caused by Coriolis terms, vanishes. This implies
∣ δ p S , n ( − X ) ∣ = ∣ δ p S , n ( X ) ∣ , (45) \left|\delta p_{\mathrm{S},n}(-X)\right| = \left|\delta p_{\mathrm{S},n}(X)\right|, \tag*{(45)} ∣ δ p S , n ( − X ) ∣ = ∣ δ p S , n ( X ) ∣ , (45)
ℜ { δ p S , n ( − X ) } = ℜ { δ p S , n ( X ) } , (46) \mathop{\Re}\left\{\delta p_{\mathrm{S},n}(-X)\right\} = \mathop{\Re}\left\{\delta p_{\mathrm{S},n}(X)\right\}, \tag*{(46)} ℜ { δ p S , n ( − X ) } = ℜ { δ p S , n ( X ) } , (46)
ℑ { δ p S , n ( − X ) } = − ℑ { δ p S , n ( X ) } , (47) \mathop{\Im}\left\{\delta p_{\mathrm{S},n}(-X)\right\} = -\mathop{\Im}\left\{\delta p_{\mathrm{S},n}(X)\right\}, \tag*{(47)} ℑ { δ p S , n ( − X ) } = − ℑ { δ p S , n ( X ) } , (47)
and the study can therefore be reduced to positive frequencies (X ≥ 0 X \ge0 X ≥ 0 ). Moreover, in the viscous regime (ε ≫ 1 \varepsilon\gg1 ε ≫ 1 ), radiative cooling damps the tidal response across the entire frequency spectrum, effectively suppressing the thermodynamic acceleration of the planet’s rotation. Thus, we restrict our analysis to the quasi-adiabatic compression regime (ε ≪ 1 \varepsilon\ll1 ε ≪ 1 ). Finally, we note that δ p T \delta p_{\mathrm{T}} δ p T may be assigned any value in the calculations because it serves merely as a scaling factor. We therefore set it to unity in this section.
In Figure 4 , the modulus, real part, and imaginary part of δ p S , n \delta p_{\mathrm{S},n} δ p S , n are plotted as functions of the reduced tidal frequency for isentropic (solid blue line) and isothermal (solid orange line) atmospheres. For each atmospheric structure, the figure shows the behaviours of non-resonant (S n = − 1 S_n = -1 S n = − 1 , left panels) and resonant (S n = + 1 S_n = +1 S n = + 1 , right panels) modes. The plotted quantities were calculated from Eqs. (27) and (33) for diatomic gases and weak dissipative processes, assuming δ p T = 1 \delta p_{\mathrm{T}} = 1 δ p T = 1 , Γ 1 = 7 / 5 \Gamma_1 = 7/5 Γ 1 = 7/5 (i.e. κ = 2 / 7 \kappa= 2/7 κ = 2/7 , and ε = 0.1 \varepsilon= 0.1 ε = 0.1 . Remarkably, both solutions exhibit qualitatively similar dependences of the tidal surface-pressure oscillations on the tidal frequency. In the high-frequency regime, resonant modes display a single peak in ∣ δ p ~ s ; n ∣ |\delta\tilde{p}_{s;n}| ∣ δ p ~ s ; n ∣ and in ℜ { δ p ~ s ; n } \Re\{\delta\tilde{p}_{s;n}\} ℜ { δ p ~ s ; n } , and a double peak in ℑ { δ p ~ s ; n } \Im\{\delta\tilde{p}_{s;n}\} ℑ { δ p ~ s ; n } . The two models, however, predict different peak amplitudes, underscoring the influence of atmospheric structure on the tidal response. For instance, at resonance, the maximum amplitude of ∣ δ p ~ s ; n ∣ |\delta\tilde{p}_{s;n}| ∣ δ p ~ s ; n ∣ in the isothermal case is roughly half that obtained with the isentropic model. In the
Figure 4. Frequency spectra of tidal surface pressure variations for non-resonant (left) and resonant (right) modes in the quasi-adiabatic compression regime. From top to bottom: Modulus, real part and imaginary part of δ p ~ s ; n \delta\tilde{p}_{s;n} δ p ~ s ; n (vertical axis) as functions of the reduced tidal frequency, X X X (horizontal axis). The blue line represents the solution for isentropic atmospheres (Eq. (27) ); the orange line corresponds to the isothermal case with b H = + ∞ b_{\mathrm{H}} = +\infty b H = + ∞ (Eq. (33) ); and the thin dashed purple line shows the empirical model proposed by Leconte et al. [31 ] based on GCM simulations (Eq. (66) ). Spectra are computed using δ p ~ T = 1 \delta\tilde{p}_{\mathrm{T}} = 1 δ p ~ T = 1 , Γ 1 = 7 / 5 \Gamma_1 = 7/5 Γ 1 = 7/5 (i.e. κ = 2 / 7 \kappa= 2/7 κ = 2/7 , diatomic gas), and ε = 0.1 \varepsilon= 0.1 ε = 0.1 . Green labels denote the main spectral features quantified in Table 3 .
Nyquist plot (e.g. [24 ] ) shown in Figure 5 , the resonance produces a circular trajectory in the complex plane (see the positive real part). Similarly, the atmospheric response in the low-frequency range is represented by a semicircle. These patterns correspond to second-order bandpass and first-order lowpass filters, respectively.
Figure 5. Nyquist plot of surface pressure variations induced by resonant modes (S n = + 1 S_n=+1 S n = + 1 ). The trajectory of the atmospheric frequency response is shown in the complex plane, with the real (horizontal axis) and imaginary (vertical axis) parts of δ p ~ s ; n \tilde{\delta p}_{s;n} δ p ~ s ; n plotted for X ≥ 0 X \ge0 X ≥ 0 . The blue line represents the solution for isentropic atmospheres (Eq. (27) ); the orange line corresponds to the isothermal case with b H = + ∞ b_H=+\infty b H = + ∞ (Eq. (33) }); and the thin dashed purple line shows the empirical model proposed by Leconte et al. [31 ] based on GCM simulations ({xref}Eq. (66)`). Parameters are δ p T ~ = 1 \tilde{\delta p_T}=1 δ p T ~ = 1 , Γ 1 = 7 / 5 \Gamma_1=7/5 Γ 1 = 7/5 (i.e. κ = 2 / 7 \kappa=2/7 κ = 2/7 , diatomic gas), and ε = 0.1 \varepsilon=0.1 ε = 0.1 . Arrows indicate the direction of increasing frequency.
Second-order bandpass filter
The circular trajectory observed in the Nyquist plot is characteristic of second-order bandpass filters, whose transfer function is expressed as
H = A 0 1 + i Q ( X − 1 X ) . (48) H=\frac{A_0}{1+iQ\left(X-\frac{1}{X}\right)}. \tag*{(48)} H = 1 + i Q ( X − X 1 ) A 0 . (48)
with A 0 A_0 A 0 being the gain coefficient and Q Q Q the quality factor. For the sake of generality, we include to this expression the phase-shifted tidal heating given by Eq. (44) . Introducing the lag angle at resonance, φ H = σ L ; n t H \varphi_H=\sigma_{L;n}t_H φ H = σ L ; n t H , the latter becomes
J ~ n ( x ) = J ~ n 0 ( x ) e − i φ H X . (49) \tilde{J}_n(x)=\tilde{J}_n^0(x)e^{-i\varphi_H X}. \tag*{(49)} J ~ n ( x ) = J ~ n 0 ( x ) e − i φ H X . (49)
Besides, it is noteworthy that the circular pattern associated with the bandpass filter in the Nyquist plot is not symmetric with respect to the horizontal axis, as described by Eq. (48) , but slightly rotated anticlockwise with respect to the origin. This indicates a positive phase angle inherent to the atmospheric tidal response and depending on Γ 1 \Gamma_1 Γ 1 and ε \varepsilon ε solely. To account for this phase angle, denoted by φ A \varphi_A φ A , we just multiply Eq. (48) by a factor exp ( i φ A X ) \exp(i\varphi_A X) exp ( i φ A X ) similar to that describing phase-shifted tidal heating in Eq. (49) . The total lag angle, φ l a g \varphi_{\rm lag} φ lag , is thus given by
φ l a g = φ H − φ A . (50) \varphi_{\rm lag}=\varphi_H-\varphi_A. \tag*{(50)} φ lag = φ H − φ A . (50)
Since the dependence of the phase on X X X can be neglected in the vicinity of the Lamb-wave resonance (∣ X − 1 ∣ ≪ 1 |X-1|\ll1 ∣ X − 1∣ ≪ 1 ), the isentropic and isothermal solutions given by Eqs. (27) and (33) in this regime are both approximated by
δ p ~ s ; n = A 0 e − i φ l a g 1 + i Q ( X − 1 X ) . (51) \tilde{\delta p}_{s;n}=\frac{A_0e^{-i\varphi_{\rm lag}}}{1+iQ\left(X-\frac{1}{X}\right)}. \tag*{(51)} δ p ~ s ; n = 1 + i Q ( X − X 1 ) A 0 e − i φ lag . (51)
The parameters A 0 A_0 A 0 and Q Q Q in Eq. (51) have different definitions in the two models, which we may establish by expressing the extrema of the modulus, real part and imaginary part of δ p ~ s ; n \tilde{\delta p}_{s;n} δ p ~ s ; n and their associated frequencies as functions of the two parameters. These derivations were carried out by assuming that Q ≫ 1 Q\gg1 Q ≫ 1 , which corresponds to the quasi-adiabatic compression regime (ε ≪ 1 \varepsilon\ll1 ε ≪ 1 ), and retaining the dominant terms only. In order to lighten expressions, we adopt the following notations: subscripts M, R and I refer to quantities associated with the modulus, real part, and imaginary part of δ p ~ s ; n \tilde{\delta p}_{s;n} δ p ~ s ; n , respectively; the plus (+) and minus (−) symbols denote the maxima and minima linked to the resonance. For example, X I , + X_{\mathrm{I},+} X I , + indicates the position of the maximum attained by the resonant peak of ℑ { δ p ~ s ; n } \Im\{\tilde{\delta p}_{s;n}\} ℑ { δ p ~ s ; n } in the frequency spectrum.
The frequency of the maximum of the modulus and its amplitude are simply given by
X M , + = 1 , ∣ δ p ~ s ; n ( X M , + ) ∣ = A 0 . (52) X_{\mathrm{M},+}=1,\qquad\left|\tilde{\delta p}_{s;n}(X_{\mathrm{M},+})\right|=A_0. \tag*{(52)} X M , + = 1 , δ p ~ s ; n ( X M , + ) = A 0 . (52)
As discussed in Sect. 2.7, they are both independent of the phase lag of tidal heating. Similarly, the frequencies at which the real part of δ p ~ s ; n \tilde{\delta p}_{s;n} δ p ~ s ; n is maximal and minimal are expressed as
X R , + = 1 − tan ( φ l a g 2 ) 2 Q , X R , − = 1 + cot ( φ l a g 2 ) 2 Q , (53) X_{\mathrm{R},+}=1-\frac{\tan\left(\frac{\varphi_{\rm lag}}{2}\right)}{2Q},\qquad X_{\mathrm{R},-}=1+\frac{\cot\left(\frac{\varphi_{\rm lag}}{2}\right)}{2Q}, \tag*{(53)} X R , + = 1 − 2 Q tan ( 2 φ lag ) , X R , − = 1 + 2 Q cot ( 2 φ lag ) , (53)
with the corresponding maximum and minimum being given by
ℜ { δ p ~ s ; n ( X R , + ) } = A 0 cos 2 ( φ l a g 2 ) , (54) \Re\{\tilde{\delta p}_{s;n}(X_{\mathrm{R},+})\}=A_0\cos^2\left(\frac{\varphi_{\rm lag}}{2}\right), \tag*{(54)} ℜ { δ p ~ s ; n ( X R , + )} = A 0 cos 2 ( 2 φ lag ) , (54)
ℜ { δ p ~ s ; n ( X R , − ) } = − A 0 sin 2 ( φ l a g 2 ) . (55) \Re\{\tilde{\delta p}_{s;n}(X_{\mathrm{R},-})\}=-A_0\sin^2\left(\frac{\varphi_{\rm lag}}{2}\right). \tag*{(55)} ℜ { δ p ~ s ; n ( X R , − )} = − A 0 sin 2 ( 2 φ lag ) . (55)
The imaginary part becomes zero for
X = 1 − tan φ l a g 2 Q , (56) X=1-\frac{\tan\varphi_{\rm lag}}{2Q}, \tag*{(56)} X = 1 − 2 Q tan φ lag , (56)
and exhibits two extrema in the vicinity of the resonance frequency. The frequencies at which these extrema occur are given by
X I , + = 1 − cot ( π 4 − φ l a g 2 ) 2 Q , X I , − = 1 + tan ( π 4 − φ l a g 2 ) 2 Q . (57) X_{\mathrm{I},+}=1-\frac{\cot\left(\frac{\pi}{4}-\frac{\varphi_{\rm lag}}{2}\right)}{2Q},\qquad X_{\mathrm{I},-}=1+\frac{\tan\left(\frac{\pi}{4}-\frac{\varphi_{\rm lag}}{2}\right)}{2Q}. \tag*{(57)} X I , + = 1 − 2 Q cot ( 4 π − 2 φ lag ) , X I , − = 1 + 2 Q tan ( 4 π − 2 φ lag ) . (57)
The maximum of the positive peak is expressed as
ℑ { δ p ~ s ; n ( X I , + ) } = A 0 sin 2 ( π 4 − φ l a g 2 ) , (58) \Im\{\tilde{\delta p}_{s;n}(X_{\mathrm{I},+})\}=A_0\sin^2\left(\frac{\pi}{4}-\frac{\varphi_{\rm lag}}{2}\right), \tag*{(58)} ℑ { δ p ~ s ; n ( X I , + )} = A 0 sin 2 ( 4 π − 2 φ lag ) , (58)
and the minimum of the negative peak as
ℑ { δ p ~ s ; n ( X I , − ) } = − A 0 cos 2 ( π 4 − φ l a g 2 ) . (59) \Im\{\tilde{\delta p}_{s;n}(X_{\mathrm{I},-})\}=-A_0\cos^2\left(\frac{\pi}{4}-\frac{\varphi_{\rm lag}}{2}\right). \tag*{(59)} ℑ { δ p ~ s ; n ( X I , − )} = − A 0 cos 2 ( 4 π − 2 φ lag ) . (59)
We remark that the phase of the forcing uniformly shifts both the zero and the two extrema of the imaginary part of δ p ~ s ; n \tilde{\delta p}_{s;n} δ p ~ s ; n by tan φ l a g / ( 2 Q ) \tan\varphi_{\rm lag}/(2Q) tan φ lag / ( 2 Q ) towards the lower frequencies. The frequency difference between the two peaks, Δ X I \Delta X_{\mathrm{I}} Δ X I , is expressed as
Δ X I = 1 Q cos φ l a g , (60) \Delta X_{\mathrm{I}}=\frac{1}{Q\cos\varphi_{\rm lag}}, \tag*{(60)} Δ X I = Q cos φ lag 1 , (60)
and the ratio between the maximum and the minimum in absolute value as
Ξ = ∣ ℑ { δ p ~ s ; n ( X I , + ) } ℑ { δ p ~ s ; n ( X I , − ) } ∣ = tan 2 ( π 4 − φ l a g 2 ) , (61) \Xi=\left|\frac{\Im\{\tilde{\delta p}_{s;n}(X_{\mathrm{I},+})\}}{\Im\{\tilde{\delta p}_{s;n}(X_{\mathrm{I},-})\}}\right|=\tan^2\left(\frac{\pi}{4}-\frac{\varphi_{\rm lag}}{2}\right), \tag*{(61)} Ξ = ℑ { δ p ~ s ; n ( X I , − )} ℑ { δ p ~ s ; n ( X I , + )} = tan 2 ( 4 π − 2 φ lag ) , (61)
which explicitly relates the asymmetry observed in Figure 3 to the forcing phase lag. We express the gain coefficient and the quality factor of the bandpass filter in terms of the model parameters by considering the regime of direct absorption treated in Sect. 3.4. In this regime, φ l a g = − φ A \varphi_{\mathrm{lag}}=-\varphi_{\mathrm{A}} φ lag = − φ A , and Eq. (60) thus simplifies to Δ X I = Q − 1 + O ( φ A 2 ) \Delta X_{\mathrm{I}}=Q^{-1}+O(\varphi_{\mathrm{A}}^{2}) Δ X I = Q − 1 + O ( φ A 2 ) assuming that φ A ∝ ε ≪ 1 \varphi_{\mathrm{A}}\propto\varepsilon\ll1 φ A ∝ ε ≪ 1 and neglecting the order-2 term. Using Eqs. (52) and (60) together with the expressions of Table 3 , we obtain
Table 3. Analytical scaling laws characterising the main features of the two solutions for ν = c s t \nu=\mathrm{cst} ν = cst , b H = + ∞ b_{\mathrm{H}=+\infty} b H = + ∞ , S n = + 1 S_{n}=+1 S n = + 1 , and ε ≪ 1 \varepsilon\ll1 ε ≪ 1 . Notes. Comparison of the analytical expressions of the quantity in the isentropic (Eq. (27), left column), and isothermal (Eq. (33), right column) atmospheric solutions. The subscripts M, R, and I refer to the modulus, real part, and imaginary part of the solutions for the surface pressure oscillation, respectively. The plus (+) and minus (−) symbols in subscripts refer to maxima and minima, respectively. Formulae are obtained from Eq. (27) (isentropic case) and Eq. (33) (isothermal case) using the method described in Sect. 3.4.
A 0 = δ p ~ T κ ε , Q = 1 κ ε (isentropic) , A 0 = δ p ~ T ( 2 − Γ 1 ) ( Γ 1 − 1 ) ε , Q = 1 κ Γ 1 ε (isothermal) . (62) \begin{aligned}
A_{0}&=\frac{\widetilde{\delta p}_{\mathrm{T}}}{\kappa\varepsilon}, & Q&=\frac{1}{\kappa\varepsilon} && \text{(isentropic)},\\
A_{0}&=\frac{\widetilde{\delta p}_{\mathrm{T}}(2-\Gamma_{1})}{(\Gamma_{1}-1)\varepsilon}, & Q&=\frac{1}{\kappa\Gamma_{1}\varepsilon} && \text{(isothermal)}. \tag*{(62)}
\end{aligned} A 0 A 0 = κ ε δ p T , = ( Γ 1 − 1 ) ε δ p T ( 2 − Γ 1 ) , Q Q = κ ε 1 = κ Γ 1 ε 1 (isentropic) , (isothermal) . (62)
Similarly, the phase angle inherent to atmospheric tidal dynamics, φ A \varphi_{\mathrm{A}} φ A , is readily deduced from the summation of the two extrema of the imaginary part,
ℑ { δ p ~ s ; n ( X I ; + ) } + ℑ { δ p ~ s ; n ( X I ; − ) } = − A 0 sin ( φ l a g ) , (63) \Im\left\{\widetilde{\delta p}_{\mathrm{s;n}}(X_{\mathrm{I;+}})\right\}+\Im\left\{\widetilde{\delta p}_{\mathrm{s;n}}(X_{\mathrm{I;-}})\right\}=-A_{0}\sin\left(\varphi_{\mathrm{lag}}\right), \tag*{(63)} ℑ { δ p s ; n ( X I ; + ) } + ℑ { δ p s ; n ( X I ; − ) } = − A 0 sin ( φ lag ) , (63)
which, in the regime of direct absorption (φ H = 0 \varphi_{\mathrm{H}}=0 φ H = 0 ) and quasi-adiabatic compression (ε ≪ 1 \varepsilon\ll1 ε ≪ 1 ), yields
φ A = arcsin ( ℑ { δ p ~ s ; n ( X I ; + ) } + ℑ { δ p ~ s ; n ( X I ; − ) } A 0 ) ≈ ℑ { δ p ~ s ; n ( X I ; + ) } + ℑ { δ p ~ s ; n ( X I ; − ) } A 0 . (64) \begin{aligned}
\varphi_{\mathrm{A}}=\arcsin\left(\frac{\Im\left\{\widetilde{\delta p}_{\mathrm{s;n}}(X_{\mathrm{I;+}})\right\}+\Im\left\{\widetilde{\delta p}_{\mathrm{s;n}}(X_{\mathrm{I;-}})\right\}}{A_{0}}\right) \\
&\approx\frac{\Im\left\{\widetilde{\delta p}_{\mathrm{s;n}}(X_{\mathrm{I;+}})\right\}+\Im\left\{\widetilde{\delta p}_{\mathrm{s;n}}(X_{\mathrm{I;-}})\right\}}{A_{0}}. \tag*{(64)}
\end{aligned} φ A = arcsin A 0 ℑ { δ p s ; n ( X I ; + ) } + ℑ { δ p s ; n ( X I ; − ) } ≈ A 0 ℑ { δ p s ; n ( X I ; + ) } + ℑ { δ p s ; n ( X I ; − ) } . (64)
Using the above formula together with the expressions provided by Table 3 , we end up with
φ A = { ( 1 + κ ) ε (isentropic) , 2 − Γ 1 ( 2 − Γ 1 ) 2 ε (isothermal) . (65) \varphi_{\mathrm{A}}=
\begin{cases}
(1+\kappa)\varepsilon& \text{(isentropic)},\\
\dfrac{2-\Gamma_{1}}{(2-\Gamma_{1})^{2}}\varepsilon& \text{(isothermal)}.
\end{cases} \tag*{(65)} φ A = ⎩ ⎨ ⎧ ( 1 + κ ) ε ( 2 − Γ 1 ) 2 2 − Γ 1 ε (isentropic) , (isothermal) . (65)
First-order lowpass filter
In the low-frequency range, the imaginary part of δ p ~ s ; n \widetilde{\delta p}_{\mathrm{s;n}} δ p s ; n behaves similarly to its numerical counterpart obtained from GCM simulations, with ℑ { δ p ~ s ; n } \Im\left\{\widetilde{\delta p}_{\mathrm{s;n}}\right\} ℑ { δ p s ; n } tending to zero at X = 0 X=0 X = 0 and reaching a maximum near synchronous rotation (see Figure 1 and Figure 3 in Leconte et al. [31 ] ; Auclair-Desrotour et al. [7 ] , respectively). It is worth noting that ℑ { δ p ~ s ; n } \Im\left\{\widetilde{\delta p}_{\mathrm{s;n}}\right\} ℑ { δ p s ; n } is an odd function of σ \sigma σ in these numerical results, which mitigates the impact of Coriolis effects on the tidal torque for X ≪ 1 X\ll1 X ≪ 1 . In particular, our ab initio analytical solutions both converge towards the empirical model proposed by Leconte et al. [31 ] from GCM simulations (see their Eq. (2)),
δ p ~ s ; n = − δ p ~ T ε + i X , (66) \widetilde{\delta p}_{\mathrm{s;n}}=-\frac{\widetilde{\delta p}_{\mathrm{T}}}{\varepsilon+iX}, \tag*{(66)} δ p s ; n = − ε + i X δ p T , (66)
represented by the dashed purple line in Figure 4 , which corresponds to the characteristic transfer function of first-order lowpass filters. In Eq. (66) , the imaginary part of δ p ~ s ; n \widetilde{\delta p}_{\mathrm{s;n}} δ p s ; n linearly tends to zero in the zero-frequency limit and decays as ℑ ( δ p ~ s ; n ) ∝ X − 1 \Im\left(\widetilde{\delta p}_{\mathrm{s;n}}\right)\propto X^{-1} ℑ ( δ p s ; n ) ∝ X − 1 as X → ∞ X\to\infty X → ∞ . It reaches a maximum for X = X I ; L F X=X_{\mathrm{I;LF}} X = X I ; LF . The frequency at which the maximum occurs and the corresponding value are given by
X I ; L F = ε , ℑ { δ p ~ s ; n ( X I ; L F ) } = δ p ~ T 2 ε . (67) X_{\mathrm{I;LF}}=\varepsilon,\qquad\Im\left\{\widetilde{\delta p}_{\mathrm{s;n}}(X_{\mathrm{I;LF}})\right\}=\frac{\widetilde{\delta p}_{\mathrm{T}}}{2\varepsilon}. \tag*{(67)} X I ; LF = ε , ℑ { δ p s ; n ( X I ; LF ) } = 2 ε δ p T . (67)
This asymptotic behaviour indicates that the assumed fast-rotation approximation remains valid in practice at low frequencies. Consequently, the solutions computed here with frequency-independent spin parameters can be applied over the entire frequency range, even though ν \nu ν , Λ n m , ν \Lambda_{n}^{m,\nu} Λ n m , ν , and Θ n m , ν \Theta_{n}^{m,\nu} Θ n m , ν are, in principle, expected to vary with σ \sigma σ in the zero-frequency limit. Furthermore, in the fast-rotation limit, the stellar semidiurnal forcing primarily excites the degree-2 mode, implying that δ p ~ 2 \widetilde{\delta p}_{2} δ p 2 largely determines the frequency spectrum of the thermotidal torque for X ≳ 1 X\gtrsim1 X ≳ 1 . One may reasonably assume that this predominance extends to lower frequencies. Under this assumption we neglect the summation in Eq. (23) and retain only the degree-2 term. The simplified analytical solutions therefore yield the pressure coefficient δ p ^ s ; 2 2 , σ \widehat{\delta p}_{\mathrm{s;2}}^{2,\sigma} δ p s ; 2 2 , σ appearing in Eq. (4) ,
δ p ^ s ; 2 2 , σ ≈ 2 π ⟨ P 2 2 , Θ 2 2 , ν ⟩ δ p ~ s ; 2 . (68) \widehat{\delta p}_{\mathrm{s;2}}^{2,\sigma}\approx\sqrt{2\pi}\left\langle P_{2}^{2},\Theta_{2}^{2,\nu}\right\rangle\widetilde{\delta p}_{\mathrm{s;2}}. \tag*{(68)} δ p s ; 2 2 , σ ≈ 2 π ⟨ P 2 2 , Θ 2 2 , ν ⟩ δ p s ; 2 . (68)
Quantitative functional analysis
We end this theoretical section by carrying out a quantitative functional analysis of the two solutions given by Eqs. (27) and (33) for resonant tidal modes (s n = − 1 s_{n}=-1 s n = − 1 ) in the regime of quasi-adiabatic compression (ε ≪ 1 \varepsilon\ll1 ε ≪ 1 ). For convenience, we assume that tidal heating arises from direct absorption (φ H = 0 \varphi_{\mathrm{H}}=0 φ H = 0 ), which amounts to setting G = 1 \mathcal{G}=1 G = 1 in Appendix H of Farhat et al. [20 ] . In this framework, we progressively explore the transition from low to high tidal frequencies. The obtained results are summarised in Table 3 .
To find the extrema of the functions, we expand the frequencies at which they occur, X ± X_{\pm} X ± , as power series in ε \varepsilon ε ,
X ± = X 0 + ∑ j = 1 N a j ε j + o ( ε N ) , (69) X_{\pm}=X_{0}+\sum_{j=1}^{N}a_{j}\varepsilon^{j}+o\left(\varepsilon^{N}\right), \tag*{(69)} X ± = X 0 + j = 1 ∑ N a j ε j + o ( ε N ) , (69)
where N N N is the truncation order of the expansion, X 0 X_{0} X 0 is the position of the extremum in the adiabatic limit (ε → 0 \varepsilon\to0 ε → 0 ), and a j a_{j} a j is a real coefficient depending solely on Γ 1 \Gamma_{1} Γ 1 . The coefficients a j a_{j} a j are obtained analytically by substituting Eq. (69) into the derivatives of the functions and identifying the zeros of the resulting equations. Computations were performed both with Mathematica (Wolfram Research [51 ] ) and with the TRIP dedicated computer algebra system (Gastineau & Laskar [25 ] ).
We first investigate the asymptotic behaviour of δ p ~ s ; n \widetilde{\delta p}_{\mathrm{s;n}} δ p s ; n near synchronisation (X ≪ ε X\ll\varepsilon X ≪ ε , see Table 3 , top section). Consistent with Figure 4 , there is no discrepancy between the two models in the zero-frequency limit. Besides, the tendencies at X = 0 X=0 X = 0 only depends on δ p ~ T \widetilde{\delta p}_{\mathrm{T}} δ p T and ε \varepsilon ε . In the low-frequency range, the local maximum of ℑ { δ p ~ s ; n } \Im\left\{\widetilde{\delta p}_{\mathrm{s;n}}\right\} ℑ { δ p s ; n } is asymptotic to δ p ~ T / ( 2 ε ) \widetilde{\delta p}_{\mathrm{T}}/(2\varepsilon) δ p T / ( 2 ε ) as ε → 0 \varepsilon\to0 ε → 0 for both isentropic and isothermal atmospheres. Similarly, the frequency at which this maximum is reached, X I ; L F X_{\mathrm{I;LF}} X I ; LF , is asymptotic to ε \varepsilon ε in the two models. Then we move to the high-frequency range and examine the local maxima and minima associated with the Lamb wave resonance near X = 1 X=1 X = 1 . Peak positions in frequency spectra are not affected much by dissipation, while peak heights are inversely proportional to ε \varepsilon ε , which is a classical outcome of linear tidal theory (e.g. Ogilvie & Lin [40 ] ; Auclair Desrotour et al. [8 ] ).
Interestingly, the difference in peak heights that can be observed in Figure 4 between isentropic and isothermal atmospheres is quantified by an explicit function of Γ 1 \Gamma_{1} Γ 1 in Table 3 . The ratio of amplitudes between the resonant peak and the maximum reached at low frequencies, ξ = ℑ { δ p ~ s ; n ( X I ; + ) } / ℑ { δ p ~ s ; n ( X I ; L F ) } \xi=\Im\left\{\widetilde{\delta p}_{\mathrm{s;n}}(X_{\mathrm{I;+}})\right\}/\Im\left\{\widetilde{\delta p}_{\mathrm{s;n}}(X_{\mathrm{I;LF}})\right\} ξ = ℑ { δ p s ; n ( X I ; + ) } /ℑ { δ p s ; n ( X I ; LF ) } , is asymptotic to κ − 1 \kappa^{-1} κ − 1 as ε → 0 \varepsilon\to0 ε → 0 for isentropic atmospheres and to ( 2 − Γ 1 ) / ( Γ 1 − 1 ) (2-\Gamma_{1})/(\Gamma_{1}-1) ( 2 − Γ 1 ) / ( Γ 1 − 1 ) for isothermal atmospheres. With Γ 1 = 7 / 5 \Gamma_{1}=7/5 Γ 1 = 7/5 , this yields ξ ≈ 3.5 \xi\approx3.5 ξ ≈ 3.5 in the former case, and ξ ≈ 1.5 \xi\approx1.5 ξ ≈ 1.5 in the latter, consistent with the spectra of Figure 4 . Also, the full width at half maximum (FWHM) of the modulus’ peak, Δ X 1 / 2 M \Delta X_{1/2\mathrm{M}} Δ X 1/2 M , and the frequency separation between two peaks of the imaginary part, Δ X I \Delta X_{\mathrm{I}} Δ X I , are simple functions of Γ 1 \Gamma_{1} Γ 1 and ε \varepsilon ε in both cases. Finally, calculations highlight the rapid decay of the atmospheric response beyond the resonance frequency. Both the modulus and imaginary part of δ p ~ s ; n \widetilde{\delta p}_{s;n} δ p s ; n are asymptotic to X − 3 X^{-3} X − 3 as X → + ∞ X\to+\infty X → + ∞ .
Benchmark against GCM simulations
The quantitative analysis performed in Sect. 3.4 yields explicit relationships between spectral features of thermotidal surface pressure oscillations in the Fourier domain and physical parameters (see Table 3 ). These relationships can be used to benchmark the analytical solutions of the present work against numerical solutions obtained from GCM simulations in former studies.
To our knowledge, such studies remain relatively scarce owing to their substantial computational cost. Evaluating the semidiurnal tidal torque at a given tidal frequency requires running a GCM simulation until a steady state is reached. The integration must then be extended over a large number of tidal cycles, after which the tidal fields are phase-folded over a stellar day in order to filter out meteorological variability. The resulting surface-pressure distribution is subsequently decomposed into spherical harmonics, and the torque is computed from the degree-2, order-2 component using Eq. (4) .5 This procedure must be repeated for each sampled tidal frequency, thereby requiring a large ensemble of simulations.
Leconte et al. [31 ] brought this approach to prominence by applying it to the case of Venus’ asynchronous rotation. Since then, it has been employed to investigate the equilibrium rotation states of Venus-like rocky exoplanets (e.g. Auclair-Desrotour et al. [7 ] ; Salazar & Wordsworth [41 ] ), as well as the past evolution of Earth’s rotation (e.g. Wu et al. 2023; Deitrick & Goldblatt [17 ] ). In this section, we place our analytical model in the context of these numerical studies, showing that the free parameters of the former can be constrained by comparison with the latter. The resulting estimates are summarised in Table 4 .
Table 4. Estimates of analytical solutions’ parameters deduced from GCM simulations in various cases. Notes. From top to bottom: the fixed parameters and the corresponding estimates of free parameters in analytical solutions for the two studied atmospheric structures (isentropic and isothermal atmospheres). Calculations are detailed in Sect. 4.
Leconte2015
To start with, we consider the Venus case in Leconte et al. [31 ] . In this study, GCM simulations were performed in the low-frequency range. As a consequence, they do not explore the resonant excitation of the thermally forced Lamb wave that occurs at higher tidal frequencies. Nevertheless, the resonance frequency can be predicted to a certain extent from the planet’s thermal state, which is known. As the deep layers of Venus’ atmosphere are neutrally stratified due to strong convection (e.g. Seiff et al. [44 ] ), the isentropic profile represents well Venus’ atmospheric structure (see Paper I, Figure 2 ). In this configuration, we set the surface temperature to the planet’s average surface temperature, namely T s = 737 T_{\mathrm{s}} = 737 T s = 737 K (e.g. Seiff et al. [44 ] ). We deduce the bulk temperature from T s T_{\mathrm{s}} T s using Eq. (34) , which yields T ‾ = 632 \overline{T} = 632 T = 632 K. This value of T ‾ \overline{T} T is also adopted for the isothermal configuration as we consider model atmospheres with the same heat content. Pressure scale heights at the planet’s surface (H s H_{\mathrm{s}} H s ) are then evaluated from T s T_{\mathrm{s}} T s using Eq. (35) .
Recalling that Λ 2 2 = 11.12895 \Lambda_{2}^{2} = 11.12895 Λ 2 2 = 11.12895 (Paper I, Table B.1), Eq. (36) yields the resonance frequency σ L ; 2 \sigma_{\mathrm{L};2} σ L ; 2 for the isentropic and isothermal atmospheres. We also obtain the associated length of the day, L O D L ; 2 \mathrm{LOD}_{\mathrm{L};2} LOD L ; 2 , given by Eq. (38) , which takes the values 16.6 d (isothermal) and 16.9 h (isentropic). In the low-frequency range, Leconte et al. [31 ] fit to GCM simulations a model similar to that given by Eq. (66) . This model is parametrised by the diurnal tidal frequency at which the maximum near synchronisation occurs, ω 0 \omega_{0} ω 0 , and a factor dimensionally consistent with a pressure, q 0 q_{0} q 0 , which take the values ω 0 = 3.77 × 10 − 7 r a d s − 1 \omega_{0} = 3.77 \times10^{-7}\ \mathrm{rad\ s^{-1}} ω 0 = 3.77 × 1 0 − 7 rad s − 1 and q 0 = 201 q_{0} = 201 q 0 = 201 Pa for Venus (see Leconte et al. [31 ] , Table 1). By comparing the formulation of the torque given by Eqs. (1-2) of Leconte et al. [31 ] with ours (Eqs. (4) and (66)), we observe that the maximum is reached for σ = σ C \sigma= \sigma_{\mathrm{C}} σ = σ C in our model and we obtain the correspondence relations linking ω 0 \omega_{0} ω 0 and q 0 q_{0} q 0 to the parameters of this work,
σ C = 2 ω 0 , J { δ p s ; 2 2 , σ } ∣ σ = σ C = q 0 p s . (70) \sigma_{\mathrm{C}} = 2\omega_{0},
\qquad
\left.\mathcal{J}\left\{\delta p_{\mathrm{s};2}^{2,\sigma}\right\}\right|_{\sigma=\sigma_{\mathrm{C}}} = \frac{q_{0}}{p_{\mathrm{s}}}. \tag*{(70)} σ C = 2 ω 0 , J { δ p s ; 2 2 , σ } σ = σ C = p s q 0 . (70)
The LOD at the maximum is thus given by
L O D 0 = 4 π σ C , (71) \mathrm{LOD}_{0} = \frac{4\pi}{\sigma_{\mathrm{C}}}, \tag*{(71)} LOD 0 = σ C 4 π , (71)
yielding L O D 0 = 193 \mathrm{LOD}_{0} = 193 LOD 0 = 193 days, and the value of the dissipation parameter is deduced from those calculated for σ C \sigma_{\mathrm{C}} σ C and σ L ; 2 \sigma_{\mathrm{L};2} σ L ; 2 in both configurations. We get ε = 3.64 × 10 − 3 \varepsilon= 3.64 \times10^{-3} ε = 3.64 × 1 0 − 3 for the isentropic profile and ε = 3.59 × 10 − 3 \varepsilon= 3.59 \times10^{-3} ε = 3.59 × 1 0 − 3 for the isothermal one.
Following Leconte et al. [31 ] , we note that ω 0 \omega_{0} ω 0 is inversely proportional to the radiative cooling time scale of the atmosphere. Therefore, this parameter should scale as
ω 0 ∝ g T e 3 C p p s , (72) \omega_{0} \propto\frac{gT_{\mathrm{e}}^{3}}{C_{p}p_{\mathrm{s}}}, \tag*{(72)} ω 0 ∝ C p p s g T e 3 , (72)
where C p C_{p} C p designates the atmospheric heat capacity per unit mass and T e T_{\mathrm{e}} T e the emission temperature of the atmosphere, namely the temperature of the deepest layers radiating energy towards space. This scaling law highlights the strong dependence of the radiative timescale on temperature and pressure, implying a corresponding sensitivity of the damping ratio, ε \varepsilon ε , and, by extension, of the amplitude of the Lamb-wave resonance. However, it should be borne in mind that the dependence of ω 0 \omega_{0} ω 0 on atmospheric parameters is in reality more complex than implied by Eq. (72) . In particular, ω 0 \omega_{0} ω 0 is governed by the atmospheric physics and composition, neither of which is explicitly accounted for in Eq. (72) . Besides, this effective parameter encapsulates the coupled contributions of all dissipative mechanisms, as discussed in Sect. 2.3, and not only radiative cooling.
To calculate δ p ~ T \delta\widetilde{p}_{\mathrm{T}} δ p T , we need to consider the relationship between spherical harmonics and Hough functions. In the zero-frequency limit, the phase of the forcing in Eq. (49) becomes zero. Consequently, we assume that the direct-absorption approximation holds in the considered frequency range and set φ H = 0 \varphi_{\mathrm{H}} = 0 φ H = 0 . For this configuration, the coefficient δ p ~ s ; 2 2 \widetilde{\delta p}_{\mathrm{s};2}^{2} δ p s ; 2 2 given by Eq. (68) is expanded in Table 3 as a power series of ε \varepsilon ε near the maximum of J { δ p s ; 2 2 } \mathcal{J}\left\{\delta p_{\mathrm{s};2}^{2}\right\} J { δ p s ; 2 2 } . Retaining only the dominant term of this series, we get
J { δ p ~ s ; 2 ( X l ; L F ) } ≈ δ p ~ T 2 ε , (73) \mathcal{J}\left\{\widetilde{\delta p}_{\mathrm{s};2}\left(X_{\mathrm{l;LF}}\right)\right\} \approx\frac{\widetilde{\delta p}_{\mathrm{T}}}{2\varepsilon}, \tag*{(73)} J { δ p s ; 2 ( X l ; LF ) } ≈ 2 ε δ p T , (73)
which, combined with Eq. (70) , yields
δ p ~ T = 2 π ε q 0 p s ( P 2 2 , Θ 2 2 ν 2 2 ) , (74) \widetilde{\delta p}_{\mathrm{T}} = \sqrt{\frac{2}{\pi}}\frac{\varepsilon q_{0}}{p_{\mathrm{s}}\sqrt{\left(P_{2}^{2},\Theta_{2}^{2}\nu_{2}^{2}\right)}}, \tag*{(74)} δ p T = π 2 p s ( P 2 2 , Θ 2 2 ν 2 2 ) ε q 0 , (74)
It should be noted that retaining only the dominant term of the series expansion introduce a relative uncertainty of order ∼ ε 2 \sim\varepsilon^{2} ∼ ε 2 in the inferred value of δ p ~ T \widetilde{\delta p}_{\mathrm{T}} δ p T . While this approximation error is negligible in the present case (ε 2 ∼ 10 − 5 \varepsilon^{2} \sim10^{-5} ε 2 ∼ 1 0 − 5 ), it may no longer be insignificant when ε \varepsilon ε approaches unity, as is the case in Sections 4.2 and § 4.3 .
The gain and quality factor characterising the equivalent bandpass filter introduced in Sect. 3.2 are computed from Γ \Gamma Γ , δ p ~ T \widetilde{\delta p}_{\mathrm{T}} δ p T , and ε \varepsilon ε using Eq. (62) . The resulting quality factors are found to be large, indicating a pronounced Lamb-wave resonance. Although this outcome is consistent with the long radiative cooling timescale of Venus’ atmosphere, the corresponding values are derived within the framework of linear theory. Consequently, the continuation
References. (a) Leconte et al. [31 ] , Table 1, Venus; (b) Deitrick & Goldblatt [17 ] , Fig. 1c, ExoCAM, 30 ppm CH4 _4 4 ; (c) Wu et al. (2023), Supplementary Materials, Fig. S4; (d) Auclair-Desrotour et al. [7 ] , Table 1 and Fig. 3.
† Quantities constrained by GCM data in case (a), from which the remaining quantities are inferred.
‡ Quantities affected by the torque renormalisation applied by Wu et al. (2023) in their Fig. S4, where the torque scale is assumed rather than derived from the GCM simulations (see Footnote 8 in Sect. 4.3).
resonance strength may be substantially reduced in reality by nonlinear dissipative processes.
We deduce the absorbed flux fraction from δ p ~ T \widetilde{\delta p}_{\mathrm{T}} δ p T using Eq. (40) ,
η = p s H s σ L ; 2 δ p ~ T κ δ F 2 , (75) \eta= \frac{p_s H_s \sigma_{\mathrm{L};2}\widetilde{\delta p}_{\mathrm{T}}}{\kappa\delta F_2}, \tag*{(75)} η = κ δ F 2 p s H s σ L ; 2 δ p T , (75)
where the resonance frequency, σ L ; 2 \sigma_{\mathrm{L};2} σ L ; 2 , and the expansion coefficient δ F 2 \delta F_2 δ F 2 are given by Eqs. (36) and (41) , respectively. Interestingly, the values obtained for the absorbed flux fraction, η ≈ 2 % \eta\approx2\% η ≈ 2% , are close to estimates of the average flux impinging Venus’ surface, which represents nearly 2.5% of the total incident Solar flux (e.g. Tomasko et al. [46 ] ). This is consistent with findings of former theoretical works showing that the effect of atmospheric tides on Venus’ spin evolution mainly originates in semidiurnal oscillations due to ground heating ([18 , 45 ] ).
As a final step, we compute the LOD maximising the tidal torque in the high-frequency range,
L O D I , + = L O D L ; 2 X I , + , (76) \mathrm{LOD}_{I,+}=\frac{\mathrm{LOD}_{L;2}}{X_{I,+}}, \tag*{(76)} LOD I , + = X I , + LOD L ; 2 , (76)
which is slightly greater than the LOD at which the resonance of the tidally forced Lamb wave occurs, given by Eq. (38) . The normalised frequency of the positive peak of the imaginary part, X I , + X_{I,+} X I , + , is provided by Table 3 as a function of Γ 1 \Gamma_1 Γ 1 and ε \varepsilon ε . For simplicity, we truncate the power series of ε \varepsilon ε to first order, which yields
X I , + ≈ 1 − κ ε 2 (isentropic), (77) X_{I,+}\approx1-\frac{\kappa\varepsilon}{2}
\qquad\text{(isentropic),} \tag*{(77)} X I , + ≈ 1 − 2 κ ε (isentropic), (77)
X I , + ≈ 1 − κ Γ 1 ε 2 (isothermal). (78) X_{I,+}\approx1-\frac{\kappa\Gamma_1\varepsilon}{2}
\qquad\text{(isothermal).} \tag*{(78)} X I , + ≈ 1 − 2 κ Γ 1 ε (isothermal). (78)
It is noteworthy that L O D I , + \mathrm{LOD}_{I,+} LOD I , + does not differ much from L O D L ; 2 \mathrm{LOD}_{L;2} LOD L ; 2 in this case since ε \varepsilon ε is very small.
In addition to Venus, Leconte et al. [31 ] report values of q 0 q_0 q 0 and ω 0 \omega_0 ω 0 for a range of idealised configurations intended to represent the inner and outer boundaries of the habitable zone (see Table 1 of their study). These scenarios encompass Venus-like planets with either N 2 \mathrm{N}_2 N 2 - or C O 2 \mathrm{CO}_2 CO 2 -dominated atmospheres, surface pressures of 1 bar or 10 bar, and incident stellar fluxes of 450 W m− 2 ^{-2} − 2 or 1366 W m− 2 ^{-2} − 2 . While these configurations constitute compelling cases for further investigation, the corresponding surface temperatures are not provided. As a result, the resonance frequency of the tidally forced Lamb wave cannot be constrained, preventing the inference of the model parameters except in the case of Venus, where the temperature profile is well established.
DG2024
The influence of atmospheric tides on the evolution of planetary spin has also been investigated in the context of Earth’s geological history. Recent studies by Wu et al. (2023) and Deitrick & Goldblatt [17 ] address this problem using GCM simulations. Although the latter was published more recently, we begin by discussing it first, as it provides a more detailed analysis of the numerical results. Deitrick & Goldblatt [17 ] explore the role of tidally induced Lamb-wave resonances in shaping Earth’s climate and surface conditions during the Proterozoic Eon (539–2500 Ma). Their calculations were performed with a range of GCMs and atmospheric compositions, with particular emphasis on the effects of varying the methane volume mixing ratio in the simulated palaeo-atmospheres. We focus here on the modulus and imaginary part of the semidiurnal surface-pressure Fourier coefficient at the equator reported by Deitrick & Goldblatt [17 ] in their 30 ppm C H 4 \mathrm{CH}_4 CH 4 configuration (Figure 3 of the article, bottom-left panel, and fig:1c , respectively), for which R. Deitrick kindly provided the data tables (Deitrick 2025, priv. comm.).6 These GCM results are reproduced in Figure 6 (left panels; grey markers).
Figure 6. Frequency dependence of the semidiurnal pressure bulge at the equator. Left: Amplitude (top) and imaginary part (bottom) of surface pressure oscillations as a function of the LOD (2 π / ∣ Ω − n ⋆ ∣ 2\pi/ |\Omega- n_{\star}| 2 π /∣Ω − n ⋆ ∣ ) in the case labelled ‘ExoCAM, 30 ppm CH4 _4 4 ’ of Deitrick & Goldblatt [17 ] , Figure 1 c and Figure 3 . Right: Thermotidal torque as a function of the LOD in the Earth’s case treated by Wu et al. (2023) in their Fig. S4 (top), and imaginary part of surface pressure oscillations as a function of the normalised frequency ω ~ = ( Ω − n ⋆ ) / n ⋆ \widetilde{\omega} = (\Omega- n_{\star}) / n_{\star} ω = ( Ω − n ⋆ ) / n ⋆ in the reference case of Auclair-Desrotour et al. [7 ] , bottom. In the three cases, plots show the numerical solutions obtained using GCM simulations (gray markers), and the closed-form solutions given by our analytical model for isentropic (solid blue line) and isothermal (solid orange line) atmospheres. The two analytical solutions, given by Eqs. (27) and (33) , are fitted to GCM data points using the parameter values of Table 4 , as detailed in Sects 4.2, 4.3 and 4.4.
As in the Venus case treated by Leconte et al. [31 ] , the simulations of Deitrick & Goldblatt [17 ] do not cover the entire frequency range, but only a narrow interval around the resonance. Consequently, they do not constrain the near-synchronous maximum. Nevertheless, we can still infer all the model parameters, this time from the features of the resonant peaks characterising the imaginary part of the surface-pressure Fourier coefficient. This analysis may be carried out using the relations established in Sect. 3, which connect peak extrema and widths to model parameters. In Appendix § 9 , we show that good fits of Deitrick & Goldblatt [17 ] ’s GCM datasets are obtained by retaining only the leading-order terms of the power-series expansions in ε \varepsilon ε reported in Table 3 . However, for a quantitative purpose, we shall adopt an alternative approach here, as well as in Sects. 4.3 and 4.4.
In this approach, the model parameters are inferred by fitting the imaginary part of the isentropic and isothermal solutions given by Eqs. (27) and (33) to data points numerically with a minimisation method. A delayed tidal heating of the form given by Eq. (49) is included in both fitted models. The four adjusted free parameters are the resonant length of the day, L O D L ; 2 \mathrm{LOD}_{L;2} LOD L ; 2 ; the normalised pressure coefficient, δ p ~ T \delta\tilde{p}_{T} δ p ~ T , the damping ratio, ε \varepsilon ε , and the phase angle of tidal heating at resonance, φ H \varphi_H φ H . We emphasise that Deitrick & Goldblatt [17 ] report values for the Fourier component of equatorial surface-pressure oscillations (δ p s e q σ \delta p^{\sigma}_{\mathrm{seq}} δ p seq σ ) while our closed-form solutions characterise the Hough modes forming the global surface-pressure anomaly, especially the degree-2 component (δ p ~ S ; 2 \delta\tilde{p}_{S;2} δ p ~ S ; 2 ), which is predominant in semidiurnal tides. To relate the two, one uses Eq. (42) , yielding
δ p ~ S ; 2 ≈ 16 15 c f δ p e q σ Θ 2 2 , 1 , P 2 2 p s . (79) \delta\tilde{p}_{S;2}\approx\sqrt{\frac{16}{15}}\frac{c_f\,\delta p^{\sigma}_{\mathrm{eq}}}{\sqrt{\Theta_2^{2,1},P_2^2}p_s}. \tag*{(79)} δ p ~ S ; 2 ≈ 15 16 Θ 2 2 , 1 , P 2 2 p s c f δ p eq σ . (79)
Calculations were conducted using the Levenberg–Marquardt algorithm, as implemented in the LsqFit.jl library for the Julia programming language (White & Mogensen [50 ] ).7
For the isentropic model, the best-fitting parameter values with 2 σ 2\sigma 2 σ -errors are L O D L ; 2 = 22.018 ± 0.027 h r \mathrm{LOD}_{L;2}=22.018\pm0.027\ \mathrm{hr} LOD L ; 2 = 22.018 ± 0.027 hr , δ p ~ T = ( 1.503 ± 0.074 ) × 10 − 4 \delta\tilde{p}_{T}=(1.503\pm0.074)\times10^{-4} δ p ~ T = ( 1.503 ± 0.074 ) × 1 0 − 4 , ε = 0.257 ± 0.015 \varepsilon=0.257\pm0.015 ε = 0.257 ± 0.015 , and φ H = 0.533 ± 0.040 \varphi_H=0.533\pm0.040 φ H = 0.533 ± 0.040 rad. The associated coefficient of determination is R 2 = 99.9 % R^2=99.9\% R 2 = 99.9% . For the isothermal model, the best-fitting parameters are L O D L ; 2 = 22.049 ± 0.028 \mathrm{LOD}_{L;2}=22.049\pm0.028 LOD L ; 2 = 22.049 ± 0.028 hr, δ p ~ T = ( 1.907 ± 0.099 ) × 10 − 4 \delta\tilde{p}_{T}=(1.907\pm0.099)\times10^{-4} δ p ~ T = ( 1.907 ± 0.099 ) × 1 0 − 4 , ε = 0.204 ± 0.012 \varepsilon=0.204\pm0.012 ε = 0.204 ± 0.012 , and φ H = 0.399 ± 0.042 \varphi_H=0.399\pm0.042 φ H = 0.399 ± 0.042 rad with R 2 = 99.9 % R^2=99.9\% R 2 = 99.9% . It is worth noting that the error margins provided above are just formal fit errors. Consequently, these errors do not account for modelling uncertainties or uncertainties associated with the frequency sampling of GCM data, which may be substantially larger, as discussed in Sect. 4.4.
The two fitted models are shown by Figure 6 both for the modulus (top left panel) and the imaginary part (bottom left panel) of the equatorial surface-pressure component. Both match the GCM data with an excellent accuracy, demonstrating that our analytical solutions successfully reproduce the frequency dependence of semidiurnal tidal surface-pressure oscillations and the associated tidal torque for Earth. In particular, the curves agree well with the GCM dataset for the modulus, whilst only the imaginary part was used to constrain the parameters. The plots also illustrate the fact that the frequency dependence of the tidal bulge near the Lamb-wave resonance does not allow the isentropic and isothermal structures to be distinguished from each other, as both lead to the very same behaviour in the considered frequency range.
The resonance occurs for a daylength of 22.0 hr, which corresponds to σ L ; 2 = 1.59 × 10 − 4 r a d s − 1 \sigma_{\mathrm{L};2} = 1.59 \times10^{-4}\ \mathrm{rad}\ \mathrm{s}^{-1} σ L ; 2 = 1.59 × 1 0 − 4 rad s − 1 . This frequency is lower than that evaluated for Venus mainly because of the temperature difference between the two planets and the scaling law σ L ; 2 ∝ T s \sigma_{\mathrm{L};2} \propto\sqrt{T_{\mathrm{s}}} σ L ; 2 ∝ T s (see Eq. (36) ). The values found for the damping ratio – ε = 0.257 \varepsilon= 0.257 ε = 0.257 (isentropic) and ε = 0.204 \varepsilon= 0.204 ε = 0.204 (isothermal) – are close to that inferred from barometric measurements for the present-day Earth (ε = 0.25 \varepsilon= 0.25 ε = 0.25 , see Figure 2 ), suggesting dissipative mechanisms of similar efficiencies in the present-day atmosphere and the palaeoatmosphere modelled by Deitrick & Goldblatt [17 ] in their GCM simulations.
From the damping ratio, we infer the Newtonian cooling frequency, σ C = ε σ L ; 2 \sigma_{\mathrm{C}} = \varepsilon\sigma_{\mathrm{L};2} σ C = ε σ L ; 2 (see Eq. (20) ), the diurnal tidal frequency at which the torque reaches its maximum near synchronisation, ω 0 = σ C / 2 \omega_{0} = \sigma_{\mathrm{C}}/2 ω 0 = σ C /2 , and the associated daylength, given by Eq. (71) . The estimated values of σ C \sigma_{\mathrm{C}} σ C are approximately two orders of magnitude greater than those found for the Venus’ case treated in Sect. 4.1, indicating a dependence on surface pressure qualitatively consistent with the rough scaling relation given by Eq. (72) for the radiative cooling timescale. Although this consistency might suggest that σ C \sigma_{\mathrm{C}} σ C is predominantly controlled by radiative cooling, one shall bear in mind that it is an effective parameter depending on atmospheric pressure, temperature and composition in a complex way. Characterising the dependence of σ C \sigma_{\mathrm{C}} σ C to the atmospheric physics thus requires further investigation, which lies beyond the scope of this study.
The pressure scale height at the planet’s surface is then determined from Eq. (36) ,
H s = ( 1 − γ ) ( R σ L ; 2 ) 2 g ∣ Λ 2 2 , 1 ∣ , (80) H_{\mathrm{s}} = \frac{(1-\gamma)(R\sigma_{\mathrm{L};2})^{2}}{g\left|\Lambda_{2}^{2,1}\right|}, \tag*{(80)} H s = g Λ 2 2 , 1 ( 1 − γ ) ( R σ L ; 2 ) 2 , (80)
which in turn allows us to compute the atmospheric surface and bulk temperatures (see Eq. (34) ). The surface temperature obtained in GCM simulations (T s ≈ 295 K T_{\mathrm{s}} \approx295\ \mathrm{K} T s ≈ 295 K ; see Deitrick & Goldblatt [17 ] , Figure 1 a) lies between the two values predicted by the analytical models (259 K and 319 K), even though one would expect it to lie closer to the isentropic estimate given the excellent agreement between the isentropic model and present-day barometric measurements shown in Figure 2 . This discrepancy may arise either from differences in the vertical atmospheric structure or from qualitative differences in the thermotidal response between the present-day Earth and the palaeoatmosphere modelled by Deitrick & Goldblatt [17 ] .
The absorbed flux fraction is finally inferred from δ p T \delta p_{\mathrm{T}} δ p T using Eqs. (41) and (75) , following the same procedure as in Sect. 4.1. The resulting values, η = 42.9 \eta= 42.9 η = 42.9 –44.1 % 44.1\% 44.1% , are roughly 2.4 times greater than the estimates derived from absorption profiles in the visible for the present-day Earth (η ≈ 17.6 \eta\approx17.6 η ≈ 17.6 –18.1 % 18.1\% 18.1% ; see Farhat et al. [20 ] , Appendix J). Although multiple factors could contribute to this discrepancy, the significant phase angle found for tidal heating suggests that the energy excess primarily stems from another heating source delayed with respect to the incident Solar flux, as discussed in Sect. 2.7. Typically, the opacity parameter should not be interpreted as a physical absorbed flux fraction but rather as an effective parameter that compensates for the omission of boundary layer physics in the model. The estimated values for η \eta η may therefore account for the infrared radiative tidal heating of the planet’s surface, which emits part of the incident flux reaching it upwards, as hypothesised by Farhat et al. [20 ] . They may also reflect cumulonimbus activity and the associated latent heat transport (e.g. Lindzen [33 ] ). The latter mechanism seems more likely in the studied case, as Deitrick & Goldblatt [17 ] show that clouds and precipitation are organised by the semidiurnal tide in their GCM simulations, thus suggesting a strong coupling between tidal dynamics and latent heat release.
Wu et al. (2023)
Like Deitrick & Goldblatt [17 ] , Wu et al. (2023) investigate the thermotidal response of Earth’s palaeoatmosphere and its influence on the planet’s rotation. In their study, GCM simulations serve mainly as a secondary tool for exploring the parameter space, as the authors primarily adopt a simplified ad hoc model that reproduces the frequency dependence of the tidal torque predicted by linear theory (Wu et al. 2023, Eqs. (3–4)). The free parameters of this model are calibrated by fitting it to geological data. Nonetheless, the tidal torque is also quantified directly from GCM simulations, as presented in the Supplementary Materials of the article (Sect. S2). In particular, we focus on the frequency spectrum of the tidal torque shown in their Fig. S4, reconstructed from GCM data points near the Lamb-wave resonance. These data are plotted in the top right panel of Figure 6 .
It is worth noting that the numerical solution obtained by Wu et al. (2023) from GCM simulations behaves similarly to that shown in Figure 1 c of Deitrick & Goldblatt [17 ] in the vicinity of the resonance. In both cases, the torque exhibits a negative peak whose amplitude significantly exceeds that of the corresponding positive peak. As discussed in Sect. 2.7, such behaviour is indicative of delayed tidal heating, which may arise either from latent heat transport (Lindzen [33 ] ; Hamilton [26 ] ) or from soil diffusion (Farhat et al. [20 ] ). In particular, Laskar et al. [30 ] approximately reproduce, in their Fig. 7, the numerical solution displayed in Fig. S4 of Wu et al. (2023) by including an indirect thermotidal forcing of the atmosphere through the planet’s surface that incorporates radiative-conductive heat transfer.
To fit our models to the GCM data point of Wu et al. (2023), we must first establish the relationship between the torque exerted on the planet with respect to its spin axis, T z \mathcal{T}_z T z , and the degree-2 Hough coefficient of the normalized semidiurnal surface pressure component, δ p s ; 2 \delta p_{\mathrm{s};2} δ p s ; 2 . This relationship is obtained by combining Eqs. (4) and (68) together, yielding
T z ≈ π 12 5 ⟨ P 2 2 , Θ 2 2 ⟩ M ⋆ M R 6 p s a 3 ℑ { δ p s ; 2 } . (81) \mathcal{T}_z \approx\pi\sqrt{\frac{12}{5}} \left\langle P_2^2,\Theta_2^2 \right\rangle\frac{M_\star}{M}\frac{R^6 p_{\mathrm{s}}}{a^3}\Im\left\{\delta p_{\mathrm{s};2}\right\}. \tag*{(81)} T z ≈ π 5 12 ⟨ P 2 2 , Θ 2 2 ⟩ M M ⋆ a 3 R 6 p s ℑ { δ p s ; 2 } . (81)
For the isentropic model, the best-fitting parameter values with 2 σ 2\sigma 2 σ -errors are L O D L ; 2 = 21.995 ± 0.089 \mathrm{LOD}_{\mathrm{L};2}=21.995\pm0.089 LOD L ; 2 = 21.995 ± 0.089 hr, δ p T = ( 2.852 ± 0.163 ) × 10 − 4 \delta p_{\mathrm{T}}=(2.852\pm0.163)\times10^{-4} δ p T = ( 2.852 ± 0.163 ) × 1 0 − 4 , ϵ = 0.508 ± 0.054 \epsilon=0.508\pm0.054 ϵ = 0.508 ± 0.054 , and φ H = 0.948 ± 0.066 \varphi_{\mathrm{H}}=0.948\pm0.066 φ H = 0.948 ± 0.066 rad. The associated coefficient of determination is R 2 = 98.1 % R^2=98.1\% R 2 = 98.1% . For the isothermal model, the best-fitting parameters are L O D L ; 2 = 22.137 ± 0.060 \mathrm{LOD}_{\mathrm{L};2}=22.137\pm0.060 LOD L ; 2 = 22.137 ± 0.060 hr, δ p T = ( 3.333 ± 0.135 ) × 10 − 4 \delta p_{\mathrm{T}}=(3.333\pm0.135)\times10^{-4} δ p T = ( 3.333 ± 0.135 ) × 1 0 − 4 , ϵ = 0.359 ± 0.026 \epsilon=0.359\pm0.026 ϵ = 0.359 ± 0.026 , and φ H = 0.691 ± 0.045 \varphi_{\mathrm{H}}=0.691\pm0.045 φ H = 0.691 ± 0.045 rad with R 2 = 98.2 % R^2=98.2\% R 2 = 98.2% . The fit reveals a noticeable difference between the resonant LOD and the LOD of the change in the torque’s sign, which are sometimes confused with each other. The resonance actually occurs near the minimum of the negative peak, at a 0.8 hr shorter LOD than the change in the torque’s sign. This difference is not negligible, as it approximately corresponds to a 100 Myr gap in geological time for Earth (e.g. Daher et al. [15 ] ; Farhat et al. [23 ] ).
Using Eq. (38) , we deduce the resonance frequency from the resonant length of the day, L O D L ; 2 \mathrm{LOD}_{\mathrm{L};2} LOD L ; 2 , which yields σ L ; 2 = 1.59 × 10 − 4 r a d s − 1 \sigma_{\mathrm{L};2}=1.59\times10^{-4}\ \mathrm{rad\ s^{-1}} σ L ; 2 = 1.59 × 1 0 − 4 rad s − 1 for the isentropic atmosphere and σ L ; 2 = 1.58 × 10 − 4 r a d s − 1 \sigma_{\mathrm{L};2}=1.58\times10^{-4}\ \mathrm{rad\ s^{-1}} σ L ; 2 = 1.58 × 1 0 − 4 rad s − 1 for the isothermal one. The associated values of σ C \sigma_{\mathrm{C}} σ C , ω 0 \omega_0 ω 0 , H s H_{\mathrm{s}} H s , T s T_{\mathrm{s}} T s , and T ‾ \overline{T} T are then obtained following the same steps as in Sect. 4.2. They are close to those calculated for the GCM simulations of Deitrick & Goldblatt [17 ] in order of magnitude (see Table 4 ). However, we remark that the values of σ C \sigma_{\mathrm{C}} σ C derived from Wu et al. (2023) for present-day Earth and from Deitrick & Goldblatt [17 ] for ancient Earth differ by a factor of ∼ 2 \sim2 ∼ 2 , despite modelling similar conditions. This discrepancy indicates that radiative cooling is not the only dissipative process controlling σ C \sigma_{\mathrm{C}} σ C , as discussed in Sect. 2.3.
The absorbed-flux fraction, η \eta η , is inferred from δ p T \delta p_{\mathrm{T}} δ p T using Eq. (75) . The estimated values are of order of ∼ 70 % \sim70\% ∼ 70% , exceeding those obtained from the GCM simulations of Deitrick & Goldblatt [17 ] for the ancient Earth. This difference, however, cannot lead to physical interpretation since the absorbed flux fraction inherits the scale of the torque assumed by Wu et al. (2023) in their Fig. S4.8 The substantial phase lag of tidal heating derived from the fit, for its part, is consistent with that found for case (b) in Sect. 4.2, supporting the interpretation that thermal tides in the simulations of Wu et al. (2023) are driven predominantly by indirect thermal forcing rather than by direct atmospheric absorption alone, as discussed in Sect. 2.7. When interpreted through the lens of analytical models, the GCM results therefore appear to challenge the classical assumption underlying linear tidal theory, according to which thermal tides on Earth are generated by the absorption of incident Solar radiation in the visible range (e.g. Chapman & Lindzen [10 ] ).
In particular, we note that the amplitude of the equatorial surface-pressure oscillations inferred from the fit for present-day Earth, ∣ δ p s ; e q ∣ \left|\delta p_{\mathrm{s};\mathrm{eq}}\right| ∣ δ p s ; eq ∣ , is substantially larger than the standard empirical estimates derived from barometric observations (e.g. Haurwitz & Cowley [28 ] ; Dai & Wang [16 ] ; Schindelegger & Ray 2014). This amplitude can be readily inferred from Eq. (6) , by writing ℑ { δ p s ; e q ∘ } = ∣ δ p s ; e q ∘ ∣ sin ( φ s ; e q ) \Im\left\{\delta p_{\mathrm{s};\mathrm{eq}}^\circ\right\}=\left|\delta p_{\mathrm{s};\mathrm{eq}}^\circ\right|\sin\left(\varphi_{\mathrm{s};\mathrm{eq}}\right) ℑ { δ p s ; eq ∘ } = δ p s ; eq ∘ sin ( φ s ; eq ) . In the fitted isentropic model, the phase angle of the surface-pressure oscillations, φ s ; e q \varphi_{\mathrm{s};\mathrm{eq}} φ s ; eq , is 0.7125 0.7125 0.7125 rad, corresponding to a pressure maximum at 10:38 LST, approximately one hour later than the observed value (09:45 LST; e.g. Chapman & Lindzen [10 ] ; Schindelegger & Ray 2014). This gives ∣ δ p s ; e q ∘ ∣ = 205.8 \left|\delta p_{\mathrm{s};\mathrm{eq}}^\circ\right|=205.8 δ p s ; eq ∘ = 205.8 Pa, almost twice the empirical estimate of 123.4 Pa reported by Schindelegger & Ray (2014). Although part of this discrepancy comes from Wu et al. (2023)’s normalisation exceeding the empirical constraint derived in Sect. 2.6, it also arises from the strong sensitivity of the thermotidal torque to the phase lag of tidal heating. The latter reduces the torque by shifting the tidal bulge towards the zero-torque configuration, in which it is aligned with the planet-star direction. Consequently, a larger amplitude of the surface-pressure oscillations is required to compensate for this reduction in torque.
Finally, we determine the LOD corresponding to the maximum resonant thermotidal torque, L O D I , + \mathrm{LOD}_{\mathrm{I},+} LOD I , + , by maximising the two fitted models. In both cases, the GCM results of Wu et al. (2023) yield L O D I , + = 23.9 \mathrm{LOD}_{\mathrm{I},+}=23.9 LOD I , + = 23.9 hr (see Figure 6 ), implying that the Solar thermotidal torque acting on Earth should currently be close to its maximum value. This theoretical prediction emphasises the difference between the resonant LOD and the LOD maximising the tidal torque, which should be clearly distinguished from each other. Here, this difference in LOD almost reaches two hours given that L O D L ; 2 ≈ 22 \mathrm{LOD}_{\mathrm{L};2}\approx22 LOD L ; 2 ≈ 22 hr in both fitted models. This two-hour gap is particularly significant in the context of the Earth’s rotational evolution, as it corresponds to a 500–1000 Myr time separation between the resonant maximum and the induced tidal locking (see e.g. Farhat et al. [23 ] , Figure 5 ). This effect adds to the increase of the resonant LOD induced by the tidal heating lag, both of them tending to bring the accelerating peak closer to the present day.
It is noteworthy that a stalled day length of approximately 19 hr, as inferred by Mitchell & Kirscher [37 ] and Wu et al. (2023) from their interpretation of geological data, would require a resonant LOD ranging between 17 hr and 18 hr, depending whether tidal heating is delayed or in phase (see Sect. 2.6) with insolation, respectively. This value is substantially shorter than the 21 hr reference value of Zahnle & Walker [53 ] , and than the smallest values obtained by Wu et al. (2023) and Deitrick & Goldblatt [17 ] from GCM simulations. Typically, in the GCM results shown by Wu et al. (2023) in their Fig. S2, the shortest resonant LOD, 21.525 hr, is obtained in the limiting case of a 2 bar-atmosphere with a 100 mbar-C O 2 \mathrm{CO}_2 CO 2 partial pressure. Such extreme conditions are consistent with the 60°C Earth found by these authors for a 18 hr resonant LOD in their reconstruction of the resonant LOD as a function of mean temperature (see Figure 5 in Wu et al. (2023)). The gap observed between model predictions and the scenario proposed by Mitchell & Kirscher [37 ] and Wu et al. (2023) tends to corroborate the conclusions by Laskar et al. [30 ] , who note that there is no geological evidence for a stalled day length of 19 hr. Together with the discrepancy identified in the phase of tidal heating, the above features of the Lamb-wave resonance highlight the need for further GCM investigations of the frequency dependence of the thermotidal torque for both present-day and early Earth conditions.
ADL2019
As a last example, we consider the reference case of Auclair-Desrotour et al. [7 ] . In this scenario, a dry, bare rock Venus-sized planet orbits a Sun-like star circularly with an orbital radius equal to the present-day Venus-Sun distance. The physical setup of this last case is the same as that adopted by Leconte et al. [31 ] to study the inner and outer edges of the habitable zone. Particularly, the 10-bar N 2 \mathrm{N}_2 N 2 atmosphere of our idealised planet contains a C O 2 \mathrm{CO}_2 CO 2 mixing ratio equivalent to that of the Earth at the turn of the millennium (∼ \sim ∼ 370 ppmv; e.g. Etheridge et al. [19 ] ), which represents the only greenhouse gas taken into account in GCM simulations.
The GCM data of Auclair-Desrotour et al. [7 ] are reproduced in the bottom right panel of Figure 6 . The figure displays the imaginary part of the semidiurnal surface pressure component at the equator, ℑ { δ p s ; e q σ } \Im\left\{\delta p_{\mathrm{s;eq}}^{\sigma}\right\} ℑ { δ p s ; eq σ } , plotted as a function of the normalised tidal frequency, ω ~ = ( Ω − n ⋆ ) / n ⋆ \widetilde{\omega}=(\Omega-n_{\star})/n_{\star} ω = ( Ω − n ⋆ ) / n ⋆ , with n ⋆ = 3.237 × 10 − 7 r a d s − 1 n_{\star}=3.237\times10^{-7}\ \mathrm{rad\ s}^{-1} n ⋆ = 3.237 × 1 0 − 7 rad s − 1 . We note that Auclair-Desrotour et al. [7 ] provide values for the degree-2, order-2 coefficient of the surface pressure anomaly, δ p s ; 2 2 , σ \delta p_{\mathrm{s};2}^{2,\sigma} δ p s ; 2 2 , σ . For convenience, this coefficient is converted into δ p s ; e q σ \delta p_{\mathrm{s;eq}}^{\sigma} δ p s ; eq σ using Eq. (68) and taking into account the factor of two difference between the conventions adopted in the two papers for spherical harmonic expansions (see Footnote 2 in Sect. 2.1).
In Auclair-Desrotour et al. [7 ] , the thermotidal torque was computed over a wide range of frequencies, from spin-orbit synchronisation to rapid rotation (see Auclair-Desrotour et al. [7 ] , Figure 3 ). Consequently, the GCM data capture both the near-synchronous maximum and the resonance associated with the tidally forced Lamb wave. The intermediate frequency dependence agrees well with the scaling law derived from linear theory, T a ∝ σ − 1 \mathcal{T}_{\mathrm{a}}\propto\sigma^{-1} T a ∝ σ − 1 . However, in the low-frequency regime, the torque obtained from GCM simulations departs substantially from the radiating-slab model adopted by Leconte et al. [31 ] to fit their GCM results (Eq. (66) ). In particular, it exhibits sharp transitions in the vicinity of the near-synchronous maximum (see Auclair-Desrotour et al. [7 ] , Figure 3 , top right panel).
Moreover, the frequency dependence of the torque near resonance deviates markedly from analytical predictions derived under the assumption of direct atmospheric absorption akin to that observed on present-day Earth. Instead of displaying two peaks of comparable amplitude and opposite sign, the spectrum is characterised solely by the negative peak, while the positive counterpart is absent. This behaviour appears to correspond to an asymptotic limit of the symmetry breaking mechanism discussed in Sect. 2.7 and identified in the numerical solutions of Deitrick & Goldblatt [17 ] and Wu et al. (2023).
In this section, we conduct a quantitative analysis of the GCM results reported by Auclair-Desrotour et al. [7 ] , following the same methodology as that applied in Sects. 4.2 and 4.3 to the GCM data of Deitrick & Goldblatt [17 ] and Wu et al. (2023). As a first step, the imaginary components of the analytical isentropic and isothermal solutions were fitted to the corresponding data points. They are plotted in Figure 6 (bottom right panel). For the isentropic model, the best-fitting parameters are L O D L ; 2 = 20.675 ± 0.00066 \mathrm{LOD}_{\mathrm{L};2}=20.675\pm0.00066 LOD L ; 2 = 20.675 ± 0.00066 hr, δ p ~ T = ( 6.7027 ± 0.0051 ) × 10 − 5 \tilde{\delta p}_{\mathrm{T}}=(6.7027\pm0.0051)\times10^{-5} δ p ~ T = ( 6.7027 ± 0.0051 ) × 1 0 − 5 , ε = ( 1.8719 ± 0.0025 ) × 10 − 2 \varepsilon=(1.8719\pm0.0025)\times10^{-2} ε = ( 1.8719 ± 0.0025 ) × 1 0 − 2 , and φ H = 1.5100 ± 0.0011 \varphi_{\mathrm{H}}=1.5100\pm0.0011 φ H = 1.5100 ± 0.0011 rad, yielding a coefficient of determination of R 2 = 94.0 % R^2=94.0\% R 2 = 94.0% . For the isothermal model, the corresponding values are L O D L ; 2 = 20.742 ± 0.00061 \mathrm{LOD}_{\mathrm{L};2}=20.742\pm0.00061 LOD L ; 2 = 20.742 ± 0.00061 hr, δ p ~ T = ( 6.9775 ± 0.0061 ) × 10 − 5 \tilde{\delta p}_{\mathrm{T}}=(6.9775\pm0.0061)\times10^{-5} δ p ~ T = ( 6.9775 ± 0.0061 ) × 1 0 − 5 , ε = ( 1.9879 ± 0.0031 ) × 10 − 2 \varepsilon=(1.9879\pm0.0031)\times10^{-2} ε = ( 1.9879 ± 0.0031 ) × 1 0 − 2 , φ H = 1.2891 ± 0.0018 \varphi_{\mathrm{H}}=1.2891\pm0.0018 φ H = 1.2891 ± 0.0018 rad, with a coefficient of determination of R 2 = 94.3 % R^2=94.3\% R 2 = 94.3% . The parameter estimates reported in Table 4 are derived directly from these fitted values.
Both models successfully reproduce the near-synchronous maximum and the resonant peak, although the quality of the fit remains inferior to that obtained for the GCM results reported by Deitrick & Goldblatt [17 ] (see Sect. 4.2). The inferred absorption fraction and phase lag of tidal heating at resonance are consistent with the estimates derived in Sect. 4.2 and Sect. 4.3. These results support the interpretation that thermal tides are driven predominantly by an indirect heating mechanism rather than by direct absorption of stellar radiation in the visible range. In particular, the phase lag of tidal heating is sufficiently large to entirely offset the resonant advance of the thermotidal bulge, thereby suppressing the positive torque peak and leaving only its negative counterpart in the spectrum.
Owing to the coarse frequency resolution of the GCM dataset, the maximum amplitude of the resonant peak is considerably less well constrained than in the preceding cases. As a result, the corresponding estimates of the quality factor (Q ≃ 126 Q \simeq126 Q ≃ 126 –187 187 187 ) and damping ratio (ε ≃ 1.87 \varepsilon\simeq1.87 ε ≃ 1.87 –1.99 × 10 − 2 1.99 \times10^{-2} 1.99 × 1 0 − 2 ) are subject to substantial uncertainties and should therefore be interpreted with caution. Similar uncertainties affect the model prediction for the near-synchronous peak, which is poorly resolved by GCM data. The normalised diurnal frequency obtained from the fit in the isentropic model, ω ~ = 4.88 \tilde{\omega}=4.88 ω ~ = 4.88 , is less than that of the maximum GCM value, ω ~ = 6.0 \tilde{\omega}=6.0 ω ~ = 6.0 , which also corresponds to the frequency interval separating two GCM data points (the difference is less in the isothermal model, for which the inferred value is ω ~ = 5.16 \tilde{\omega}=5.16 ω ~ = 5.16 ). As a consequence, the uncertainty associated with ω 0 \omega_0 ω 0 and L O D 0 \mathrm{LOD}_0 LOD 0 is greater than 23% in these calculations, and propagate to other fitted parameters. Typically, we observe a 52% relative difference between the prediction of the fitted model (ω ~ = 4.88 \tilde{\omega}=4.88 ω ~ = 4.88 ) and that obtained by Auclair-Desrotour et al. [7 ] using a cubic-spline interpolation of a larger set of GCM data (ω ~ = 7.4 \tilde{\omega}=7.4 ω ~ = 7.4 ). Furthermore, it is worth noting that the near-synchronous peak described by GCM simulations is flatter than that predicted by the analytical theory in the vicinity of the maximum, resulting in a ill-defined peak frequency (see Auclair-Desrotour et al. [7 ] , Fig. 3, top right panel).
Despite these limitations, the inferred parameters remain consistent with the scaling relation given by Eq. (72) . In particular, Eq. (72) predicts the radiative-cooling frequency of the 10-bar atmosphere to be approximately one order of magnitude lower than that of present-day Earth, consistent with the inferred range, ω 0 = 1.58 \omega_0=1.58 ω 0 = 1.58 –1.67 × 10 − 6 r a d s − 1 1.67 \times10^{-6}\,\mathrm{rad}\,\mathrm{s}^{-1} 1.67 × 1 0 − 6 rad s − 1 . However, as discussed in Sect. 4.2, the dependence of ω 0 \omega_0 ω 0 on surface pressure warrants further investigation, as this parameter is governed by atmospheric physics in a complex manner. For instance, Auclair-Desrotour et al. [6 ] inferred from GCM simulations of idealised Venus-like planets a power-law dependence ω 0 ∝ p s − 0.30 \omega_0 \propto p_s^{-0.30} ω 0 ∝ p s − 0.30 , which departs substantially from a linear scaling.
Finally, it is worth noting that Farhat et al. [20 ] achieved a comparably good fit to the same dataset by assuming that atmospheric tides are forced through radiative-diffusive heat exchange with the planetary surface (see Farhat et al. [20 ] , Fig. K.1). Within the framework developed here, the corresponding ab initio model can be incorporated by replacing the factor exp ( − i ω H X ) \exp(-\mathrm{i}\omega H_X) exp ( − i ω H X ) in Eq. (49) with the transfer function given by Eq. (H.4) of Farhat et al. [20 ] . Comparable results may likewise be obtained using an equivalent mass-spring-damper formulation, as will be discussed in Paper III. Regardless of the specific modelling approach adopted, the inclusion of delayed tidal heating appears essential for accurately reproducing GCM results, which departs significantly from the conventional assumption of direct atmospheric absorption. Further work is therefore required to revisit and strengthen the numerical modelling framework applied to dry rocky planets.
Conclusions
This study continues the theoretical work initiated in Paper I, which aimed to analytically characterise the influence of atmospheric thermal tides on the rotation of rocky planets. In Paper I, starting from the primitive equations of fluid dynamics, we derived a closed-form solution describing the three-dimensional atmospheric response to gravitational and thermal tidal forcings
(Paper I, Eqs. 85-92). This solution, obtained within linear theory, incorporates dissipative effects through Newtonian cooling and assumes a vertically decaying distribution of tidal heating. It applies uniformly to two atmospheric structures of practical relevance: isentropic atmospheres (uniform potential temperature) and isothermal atmospheres (uniform temperature). In particular, we explicitly identified the forcing frequencies at which the main tidal modes – namely Lamb waves – are resonantly excited.
In the present work, we have extended this framework by focusing on semidiurnal thermal tides and their specific role in the long-term evolution of planetary rotation. Using the solution from Paper I, we have expressed the thermotidal torque as a function of tidal frequency and a reduced set of physical parameters defining the star-planet system. This has been done for both the convective and isothermal equilibria, assuming a Dirac distribution of tidal heating for the latter (in line with Dobrovolskis & Ingersoll [18 ] ). The resulting expressions (Eqs. (27) and (33) and Table 2 ) enable a detailed characterisation of the frequency dependence of the torque in the regime of quasi-adiabatic compression, where dissipative processes remain weak. We have derived scaling laws describing key properties of this dependence – such as the frequencies, amplitudes, and widths of resonant peaks – as functions of dimensionless control parameters (Table 3 ). Finally, we have benchmarked the analytical solutions against numerical results from GCM simulations for the Earth and Venus (Table 4 ; Figure 6 ).
Although the analytical expressions of the tidal torque differ markedly between isentropic and isothermal atmospheres, they exhibit several robust similarities. Both predict the coexistence of resonant modes, which grow in amplitude near specific forcing frequencies, and non-resonant modes, which gradually diminish as frequency increases. Despite their simplicity, both formulations reproduces the main features of numerical solutions with notable accuracy. In particular, they asymptotically converge towards the functional behaviour empirically inferred in the zero-frequency limit from GCM studies of Venus-like, near-synchronous planets. This agreement holds for both resonant and non-resonant modes. The most significant results, however, concern the high-frequency regime.
The analytical model reproduces with high accuracy the Precambrian resonant peak obtained in Earth GCM simulations for surface pressure oscillations, thereby demonstrating its ability to capture the influence of Lamb-wave resonance on the planet’s past rotational evolution. The agreement between analytical and numerical results provides explicit relationships between resonance characteristics and key atmospheric parameters, including the adiabatic index, and enables the estimation of free parameters – such as the radiative cooling time or the absorbed flux fraction – from numerical solutions. Moreover, this consistency indicates that the thermal tides produced in simulations of Earth’s palaeoatmospheres operate within the quasi-adiabatic regime, where the scaling laws derived here remain valid, albeit near the critical limit. Typically, our analysis suggests that the past Earth’s effective radiative cooling frequency lies between 3.23 × 10 − 5 r a d s − 1 3.23 \times10^{-5}\,\mathrm{rad}\,\mathrm{s}^{-1} 3.23 × 1 0 − 5 rad s − 1 and 8.06 × 10 − 5 r a d s − 1 8.06 \times10^{-5}\,\mathrm{rad}\,\mathrm{s}^{-1} 8.06 × 1 0 − 5 rad s − 1 . Besides, the amplitude and lag of semidiurnal surface pressure oscillations predicted by the isentropic direct-absorption model for present-day Earth conditions are in excellent agreement with estimates derived from barometric observations.
Similarly, by fitting the analytical model to numerical simulations, we have recovered previous estimates of Venusian semidiurnal tidal heating and derived quantitative constraints on both the frequency and amplitude of the associated thermotidal resonance. The same analysis also yields physically consis-
[assistant/analysisেন্স] tent results for the idealised 10-bar rocky-planet configuration considered in this study. Across most configurations examined, the inferred tidal heating is characterised by substantial phase lags and large effective absorption fractions, departing markedly from the commonly adopted assumption of direct atmospheric absorption. This systematic discrepancy suggests that semidiurnal thermal tides in the analysed GCM simulations are driven predominantly by indirect thermal forcing mechanisms, such as heat diffusion within the soil or latent heat transport, as discussed in previous studies.
The phase lag of tidal heating has important implications for the rotational evolution of rocky planets, as it can strongly attenuate – or even suppress entirely – the accelerative tidal torque that may counteract tidal deceleration in the vicinity of the Lamb-wave resonance. Delayed tidal heating offers less favourable configurations for resonant torque amplification than the direct-absorption regime, which characterises semidiurnal tides on present-day Earth. This effect notably weakens the Precambrian tidal-locking hypothesis in scenarios where the Precambrian atmosphere differed from the actual one. More generally, the results demonstrate that the amplitude of pressure oscillations alone is insufficient to quantify the thermotidal torque: pronounced resonant peaks in the pressure response do not necessarily translate into strong thermodynamical accelerations of planetary rotation. These conclusions should nevertheless be interpreted with appropriate caution, as the comparative analysis presented here relies on a limited set of benchmark simulations. Additional GCM studies are therefore needed to provide a more comprehensive characterisation of the thermotidal response of planetary atmospheres.
Because the analytical solution developed here can be readily incorporated into planetary evolution models, it represents a valuable tool for investigating equilibrium states and the long-term rotational evolution of rocky planets. Its excellent agreement with Earth GCM results makes it particularly suitable for studying the role of thermal tides in Earth’s rotational length-of-day history. In particular, it offers a versatile framework for quantifying the influence of Lamb-wave resonance during the Precambrian Era and for exploring the associated parameter space. More broadly, the cases of Earth and Venus highlight the intimate coupling between planetary rotation and thermal tides and underscore the potential importance of the latter in shaping the rotational evolution of rocky exoplanets in the habitable zone of their host stars. In the forthcoming Paper III, we will show that the complex frequency behaviour of thermotidal surface pressure oscillations described by our analytical solutions and GCM simulations is accurately reproduced by a simple mass-spring-damper model.
Acknowledgements. The authors are grateful to the anonymous referee for their insightful comments, which substantially improved the quality and clarity of the manuscript. M. Farhat is supported by the Miller Institute for Basic Research in Science at UC Berkeley through a Miller Research Fellowship. This work was supported by the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (Advanced Grant AstroGeo-885250). This research has made use of NASA’s Astrophysics Data System.
Nomenclature
The notations introduced in the main text are listed below in order of appearance.
L O D \mathrm{LOD} LOD length of the day p. 1
G C M \mathrm{GCM} GCM general circulation model p. 1
T z \mathcal{T}_{z} T z thermotidal torque about the planet’s spin axis p. 2
R R R planet radius p. 3
g g g planet’s surface gravity p. 3
U T U_{\mathrm{T}} U T tide-raising gravitational potential p. 3
φ \varphi φ longitude p. 3
∂ φ \partial_{\varphi} ∂ φ partial derivative with respect to longitude p. 3
t t t time p. 3
Δ t \Delta t Δ t time interval p. 3
S S S sphere of unit radius p. 3
δ p s \delta p_{\mathrm{s}} δ p s tidally induced surface pressure anomaly p. 3
R \mathfrak{R} R real part of a complex number p. 3
I \mathfrak{I} I imaginary part of a complex number p. 3
i i i imaginary unit p. 3
θ \theta θ colatitude p. 3
p s p_{\mathrm{s}} p s globally-averaged surface pressure p. 3
σ \sigma σ tidal frequency p. 3
Y l m Y_{l}^{m} Y l m degree-l l l , order-m m m spherical harmonic p. 3
δ p ^ s ; l m , σ \delta\hat{p}_{\mathrm{s};l}^{m,\sigma} δ p ^ s ; l m , σ nondimensional degree-l l l , order-m m m spherical harmonic expansion coefficient of surface pressure variations p. 3
a a a planet’s orbital radius p. 3
n ⋆ n_{\star} n ⋆ planet’s anomalistic mean motion p. 3
G G G universal gravitational constant p. 3
M M M planet mass p. 3
M ⋆ M_{\star} M ⋆ mass of the star p. 3
δ p ^ s ; 2 2 , σ \delta\hat{p}_{\mathrm{s};2}^{2,\sigma} δ p ^ s ; 2 2 , σ degree-2, order-2 spherical expansion coefficient of surface pressure variations p. 3
Ω \Omega Ω planet’s spin angular velocity p. 3
δ p s ; e q \delta p_{\mathrm{s};\mathrm{eq}} δ p s ; eq equatorial semidiurnal surface pressure oscillation p. 3
δ p ^ s ; e q σ \delta\hat{p}_{\mathrm{s};\mathrm{eq}}^{\sigma} δ p ^ s ; eq σ Fourier coefficient of the equatorial semidiurnal surface pressure oscillation p. 3
c f c_{\mathrm{f}} c f shape factor characterising the latitudinal distribution of surface pressure variations p. 3
F F F incident stellar flux impinging the top of the atmosphere p. 4
F ⋆ F_{\star} F ⋆ incident stellar flux at the substellar point p. 4
ψ \psi ψ stellar zenithal angle p. 4
ω \omega ω diurnal tidal frequency p. 4
P l m P_{l}^{m} P l m degree-l l l , order-m m m normalised associated Legendre function p. 4
δ F ^ l m \delta\hat{F}_{l}^{m} δ F ^ l m spherical harmonic coefficients of the flux p. 4
F l F_{l} F l degree-l l l expansion coefficient of the flux in Legendre polynomials p. 4
P l P_{l} P l degree-l l l Legendre polynomial p. 4
ν m \nu_{m} ν m order-m m m spin parameter p. 4
δ F ^ n m \delta\hat{F}_{n}^{m} δ F ^ n m degree-n n n , order-m m m Hough function expansion coefficient of the incident stellar flux p. 4
Θ n m , ν \Theta_{n}^{m,\nu} Θ n m , ν degree-n n n Hough function p. 4
δ F n \delta F_{n} δ F n degree-n n n , order-2 expansion coefficient of the flux in terms of Hough functions p. 4
U T ; s U_{\mathrm{T};\mathrm{s}} U T ; s tide-raising potential at the planet’s surface p. 4
δ p ~ n \delta\tilde{p}_{n} δ p ~ n degree-n n n expansion coefficient of the Fourier coefficient of the normalised tidal pressure oscillation p. 4
x x x dimensionless vertical coordinate p. 4
γ \gamma γ reduced stratification number p. 4
ζ \zeta ζ reduced lapse rate p. 4
λ \lambda λ dimensionless atmospheric parameter p. 4
Γ 1 \Gamma_{1} Γ 1 adiabatic exponent p. 4
κ \kappa κ κ = ( Γ 1 − 1 ) / Γ 1 \kappa=(\Gamma_{1}-1)/\Gamma_{1} κ = ( Γ 1 − 1 ) / Γ 1 p. 4
α \alpha α ratio between the Newtonian and tidal frequencies p. 4
σ C \sigma_{\mathrm{C}} σ C Newtonian cooling frequency p. 4
b H b_{\mathrm{H}} b H vertical variation rate of the tidal heating profile p. 4
S n S_{n} S n sign of the eigenvalue associated with the degree-n n n Hough function p. 4
k n k_{n} k n vertical wavenumber of the degree-n n n mode p. 4
β s \beta_{\mathrm{s}} β s dimensionless atmospheric parameter p. 4
H s H_{\mathrm{s}} H s pressure scale height at the planet’s surface p. 4
X X X normalised tidal frequency p. 5
ε \varepsilon ε normalised damping rate p. 5
δ p ^ T \delta\hat{p}_{\mathrm{T}} δ p ^ T normalised scaling factor of thermotidal pressure variations p. 5
η \eta η absorbed flux fraction p. 5
σ L ; n \sigma_{\mathrm{L};n} σ L ; n resonance frequency of the degree-n n n tidally excited Lamb wave p. 5
δ p ~ s ; n \delta\tilde{p}_{\mathrm{s};n} δ p ~ s ; n δ p ~ n \delta\tilde{p}_{n} δ p ~ n evaluated at the planet’s surface p. 5
ϑ \vartheta ϑ κ \kappa κ -dependent dimensionless parameter of the isothermal case p. 6
μ \mu μ calculation quantity introduced in the isothermal case p. 6
T ‾ \overline{T} T atmospheric bulk temperature p. 7
T s T_{\mathrm{s}} T s atmospheric temperature at the planet’s surface p. 7
σ L ; 2 \sigma_{\mathrm{L};2} σ L ; 2 resonance frequency of the main mode excited by the semidiurnal tidal heating p. 7
L O D L ; 2 \mathrm{LOD}_{\mathrm{L};2} LOD L ; 2 LOD at which the resonance of the main semidiurnal tidal mode occurs p. 7
c e q c_{\mathrm{eq}} c eq dimensionless numerical factor in Eq. (42) p. 9
J ~ n 0 \tilde{J}_{n}^{0} J ~ n 0 real-valued vertical profile of tidal heating in Eq. (44) p. 10
t H t_{\mathrm{H}} t H constant time lag of tidal heating in Eq. (44) p. 10
A 0 A_{0} A 0 gain coefficient of the equivalent second-order bandpass filter p. 12
Q Q Q quality factor of the equivalent second-order bandpass filter p. 12
φ H \varphi_{\mathrm{H}} φ H lag angle of tidal heating at resonance p. 12
φ A \varphi_{\mathrm{A}} φ A phase angle of surface pressure variations at resonance in the regime of direct absorption p. 12
φ l a g \varphi_{\mathrm{lag}} φ lag total lag angle of the equivalent bandpass filter at resonance p. 12
Ξ \Xi Ξ ratio between the torque’s maximum and minimum near resonance p. 12
δ p ^ s ; 2 2 , σ \delta\hat{p}_{\mathrm{s};2}^{2,\sigma} δ p ^ s ; 2 2 , σ semidiurnal Fourier coefficient of the tidal surface-pressure oscillation p. 13
X I ; L F X_{\mathrm{I};\mathrm{LF}} X I ; LF frequency at which the maximum of I { δ p ~ s ; n } \mathfrak{I}\{\delta\tilde{p}_{\mathrm{s};n}\} I { δ p ~ s ; n } is reached in the near-synchronous regime p. 14
X R ; + X_{\mathrm{R};+} X R ; + frequency at which the maximum of R { δ p ~ s ; n } \mathfrak{R}\{\delta\tilde{p}_{\mathrm{s};n}\} R { δ p ~ s ; n } associated with the Lamb-wave resonance is reached p. 14
X I ; ± X_{\mathrm{I};\pm} X I ; ± frequencies at which the minimum ( − ) (-) ( − ) and maximum ( + ) (+) ( + ) of I { δ p ~ s ; n } \mathfrak{I}\{\delta\tilde{p}_{\mathrm{s};n}\} I { δ p ~ s ; n } associated with the Lamb-wave resonance are attained p. 14
Δ X I \Delta X_{\mathrm{I}} Δ X I frequency interval between the minimum and maximum of I { δ p ~ s ; n } \mathfrak{I}\{\delta\tilde{p}_{\mathrm{s};n}\} I { δ p ~ s ; n } associated with the Lamb-wave resonance p. 14
N N N truncation order of the power series in Eq. (69) p. 13
a j a_{j} a j j j j -th coefficient of the power series in Eq. (69) p. 13
Shape factor characterising the latitudinal structure of the tidal bulge
In Eq. (5) , the amplitude of semidiurnal surface pressure oscillations at the equator is connected to the corresponding degree-2, order-2 spherical harmonic by introducing the shape factor c f c_f c f . This parameter quantifies the deviation of the latitudinal structure of semidiurnal surface pressure variations from the quadrupole, the latter being shaped by the degree-2, order-2 normalised associated Legendre function. The latitudinal distribution of the semidiurnal component of surface pressure is assumed to be described by a function f ( μ ) f(\mu) f ( μ ) , with μ = cos θ \mu= \cos\theta μ = cos θ . The shape factor associated with f f f is formulated using the dot product defined in Eq. (13) as
c f = ⟨ P 2 2 , f ⟩ P 2 2 ( 0 ) f ( 0 ) . (82) c_f = \langle P_2^2,f\rangle\frac{P_2^2(0)}{f(0)}. \tag*{(82)} c f = ⟨ P 2 2 , f ⟩ f ( 0 ) P 2 2 ( 0 ) . (82)
where P 2 2 P_2^2 P 2 2 designates the degree-2, order-2 normalised associated Legendre function introduced in Eq. (8) , given by (e.g. Auclair-Desrotour et al. [6 ] , Appendix A)
P 2 2 = 3 5 48 ( 1 − μ 2 ) . (83) P_2^2 = 3\sqrt{\frac{5}{48}}\left(1-\mu^2\right). \tag*{(83)} P 2 2 = 3 48 5 ( 1 − μ 2 ) . (83)
The scaling function sin 3 θ \sin^3\theta sin 3 θ reported by Schindelegger & Ray (2014) for the Earth’s semidiurnal thermal tides corresponds to
f ( μ ) = ( 1 − μ 2 ) 3 / 2 . (84) f(\mu)=\left(1-\mu^2\right)^{3/2}. \tag*{(84)} f ( μ ) = ( 1 − μ 2 ) 3/2 . (84)
Using this expression in Eq. (B.1), we obtain
c f = 75 π 256 ≈ 0.92 , (85) c_f = \frac{75\pi}{256}\approx0.92, \tag*{(85)} c f = 256 75 π ≈ 0.92 , (85)
which is the value adopted in calculations throughout the study.
Multipole expansion of stellar irradiance
We detail here the derivations leading to Eq. (8) . As a first step, the spatial distribution of the incident stellar flux, given by Eq. (7) , is expanded as a series of Legendre polynomials,
F = F ⋆ ∑ l = 0 + ∞ F l P l ( cos ψ ) , (86) F = F_\star\sum_{l=0}^{+\infty} F_l P_l(\cos\psi), \tag*{(86)} F = F ⋆ l = 0 ∑ + ∞ F l P l ( cos ψ ) , (86)
where the expansion coefficients are expressed as
F l = 2 l + 1 2 ∫ 0 1 μ P l ( μ ) d μ . (87) F_l = \frac{2l+1}{2}\int_0^1 \mu P_l(\mu)\,\mathrm{d}\mu. \tag*{(87)} F l = 2 2 l + 1 ∫ 0 1 μ P l ( μ ) d μ . (87)
The Legendre polynomial addition theorem (e.g. Arfken & Weber [2 ] , Eq. (9.189), p. 600) allows for separating the local coordinates ( θ , φ ) (\theta,\varphi) ( θ , φ ) from the coordinates of the star ( θ ⋆ , φ ⋆ ) (\theta_\star,\varphi_\star) ( θ ⋆ , φ ⋆ ) in P l ( cos ψ ) P_l(\cos\psi) P l ( cos ψ ) , which is expressed as
P l ( cos ψ ) = 4 π 2 l + 1 ∑ m = − l l Y l m ( θ , φ ) Y l m ( θ ⋆ , φ ⋆ ) ‾ . (88) P_l(\cos\psi)=\frac{4\pi}{2l+1}\sum_{m=-l}^{l}Y_l^m(\theta,\varphi)\overline{Y_l^m(\theta_\star,\varphi_\star)}. \tag*{(88)} P l ( cos ψ ) = 2 l + 1 4 π m = − l ∑ l Y l m ( θ , φ ) Y l m ( θ ⋆ , φ ⋆ ) . (88)
In the above equation, Y l m ‾ \overline{Y_l^m} Y l m designates the complex conjugate of the degree-l l l , order-m m m spherical harmonic, Y l m Y_l^m Y l m (e.g. Varshalovich et al. [48 ] , Sect. 5.2). Writing the spherical harmonics in terms of the normalised associated Legendre functions, P l m P_l^m P l m , we obtain
Y l m ( θ , φ ) = 1 2 π P l m ( cos θ ) e i m φ , (89) Y_l^m(\theta,\varphi)=\frac{1}{\sqrt{2\pi}}P_l^m(\cos\theta)e^{im\varphi}, \tag*{(89)} Y l m ( θ , φ ) = 2 π 1 P l m ( cos θ ) e im φ , (89)
and Eq. (C.1) thus becomes
F = F ⋆ ∑ l = 0 + ∞ ∑ m = − l l 2 F l 2 l + 1 P l m ( cos θ ⋆ ) P l m ( cos θ ) e i m ( φ − φ ⋆ ) . (90) F=F_\star\sum_{l=0}^{+\infty}\sum_{m=-l}^{l}\frac{2F_l}{2l+1}P_l^m(\cos\theta_\star)P_l^m(\cos\theta)e^{im(\varphi-\varphi_\star)}. \tag*{(90)} F = F ⋆ l = 0 ∑ + ∞ m = − l ∑ l 2 l + 1 2 F l P l m ( cos θ ⋆ ) P l m ( cos θ ) e im ( φ − φ ⋆ ) . (90)
In the coplanar-circular configuration, where a zero-obliquity planet orbits its host star circularly, the colatitude of the star is θ ⋆ = π / 2 \theta_\star=\pi/2 θ ⋆ = π /2 and its longitude is φ ⋆ = − ω t \varphi_\star=-\omega t φ ⋆ = − ω t , with ω = Ω − n ⋆ \omega=\Omega-n_\star ω = Ω − n ⋆ denoting the diurnal tidal frequency. Besides, the normalised associated Legendre functions of positive and negative orders are linked together by the symmetry relation
P l − m ( cos θ ) = ( − 1 ) m P l m ( cos θ ) , (91) P_l^{-m}(\cos\theta)=(-1)^mP_l^m(\cos\theta), \tag*{(91)} P l − m ( cos θ ) = ( − 1 ) m P l m ( cos θ ) , (91)
which allows for gathering pairs of terms of orders m m m and − m -m − m in Eq. (C.5). It follows
F ( θ , φ , t ) = F ⋆ ℜ { ∑ l = 0 + ∞ ∑ m = 0 l δ F ^ l m P l m ( cos θ ) e i m ( ω t + φ ) } , (92) F(\theta,\varphi,t)=F_\star\Re\left\{\sum_{l=0}^{+\infty}\sum_{m=0}^{l}\delta\widehat{F}_l^mP_l^m(\cos\theta)e^{im(\omega t+\varphi)}\right\}, \tag*{(92)} F ( θ , φ , t ) = F ⋆ ℜ { l = 0 ∑ + ∞ m = 0 ∑ l δ F l m P l m ( cos θ ) e im ( ω t + φ ) } , (92)
where the expansion coefficients are defined as
δ F ^ l m = ( 2 − δ m , 0 ) 2 F l 2 l + 1 P l m ( 0 ) . (93) \delta\widehat{F}_l^m=(2-\delta_{m,0})\frac{2F_l}{2l+1}P_l^m(0). \tag*{(93)} δ F l m = ( 2 − δ m , 0 ) 2 l + 1 2 F l P l m ( 0 ) . (93)
It is noteworthy that P l m ( 0 ) P_l^m(0) P l m ( 0 ) is a special value of the normalised associated Legendre function given by (e.g. Varshalovich et al. [48 ] , Sect. 5.14, Eq. (2))
P l m ( 0 ) = { ( − 1 ) l + m 2 ( 2 l + 1 ) ( l + m − 1 ) ! ! ( l − m − 1 ) ! ! 2 ( l + m ) ! ! ( l − m ) ! ! if l + m is even , 0 if l + m is odd . (94) P_l^m(0)=
\begin{cases}
(-1)^{\frac{l+m}{2}}\sqrt{\frac{(2l+1)(l+m-1)!!(l-m-1)!!}{2(l+m)!!(l-m)!!}} & \text{if } l+m \text{ is even},\\
0 & \text{if } l+m \text{ is odd}.
\end{cases} \tag*{(94)} P l m ( 0 ) = { ( − 1 ) 2 l + m 2 ( l + m )!! ( l − m )!! ( 2 l + 1 ) ( l + m − 1 )!! ( l − m − 1 )!! 0 if l + m is even , if l + m is odd . (94)
In the above equation, the double factorial of a positive integer n n n is defined as
n ! ! = ∏ k = 0 ⌈ n / 2 ⌉ − 1 ( n − 2 k ) = n ( n − 2 ) ( n − 4 ) ⋯ , (95) n!!=\prod_{k=0}^{\left\lceil n/2\right\rceil-1}(n-2k)=n(n-2)(n-4)\cdots, \tag*{(95)} n !! = k = 0 ∏ ⌈ n /2 ⌉ − 1 ( n − 2 k ) = n ( n − 2 ) ( n − 4 ) ⋯ , (95)
with ⌈ n / 2 ⌉ \lceil n/2\rceil ⌈ n /2 ⌉ denoting the ceiling of n / 2 n/2 n /2 .
Table D.1. Table D.1. Estimates of analytical solutions’ parameters deduced from GCM simulations using analytical (Ana.) or numerical (Num.) methods for the case treated in Sect. 4.2.
Notes. From top to bottom: the fixed parameters and the corresponding estimates of free parameters in analytical solutions for the two studied atmospheric structures (isentropic and isothermal atmospheres). Calculations are detailed in Appendix D (Ana.) and Sect. 4 (Num.).
Figure D.1. Fig. D.1. Modulus (top) and imaginary part (bottom) of surface pressure oscillations as a function of the LOD (2 π / ∣ Ω − n ⋆ ∣ 2\pi/|\Omega-n_{\star}| 2 π /∣Ω − n ⋆ ∣ ) in the case labelled ‘ExoCAM, 30 ppm CH4 _4 4 ’, Deitrick & Goldblatt [17 ] , Figs. 1c and 3). Plots show the numerical solutions obtained using GCM simulations (gray markers), and the closed-form solutions given by our analytical model for isentropic (solid blue line) and isothermal (solid orange line) atmospheres. The two analytical solutions, given by Eqs. (27) and (33) , are fitted to GCM data points using the formulae provided by Table 3 and the parameter values of Table 4 , as detailed in Appendix § 9 .
Benchmark of analytical scalings against DG2024's GCM data
In Sects. 4.2 to 4.4, the isentropic and isothermal models are fitted to GCM data numerically using a minimisation algorithm. However, this is not the only way to proceed. Alternatively, one may fit the models analytically using the formulae established in Sect. 3 and reported in Table 3 . In this appendix, we detail this analytical method for the case of Deitrick & Goldblatt [17 ] , which is treated in Sect. 4.2. Analytical results are benchmarked against numerical ones in Table D.1. The corresponding solutions are plotted in Fig. D.1 together with GCM data, analogous to Figure 6 .
To begin with, we derive the parameters σ L ; 2 \sigma_{\mathrm{L};2} σ L ; 2 , δ p ~ T \tilde{\delta p}_{\mathrm{T}} δ p ~ T , and ε \varepsilon ε from the resonant peak LOD (L O D L ; 2 \mathrm{LOD}_{\mathrm{L};2} LOD L ; 2 ), amplitude (∣ δ p s ; e q σ ∣ max |\delta p_{\mathrm{s;eq}}^{\sigma}|_{\max} ∣ δ p s ; eq σ ∣ m a x ), and FWHM (Δ X 1 / 2 ; M \Delta X_{1/2;\mathrm{M}} Δ X 1/2 ; M ). These three quantities are obtained by performing a cubic-spline interpolation of GCM data for the modulus of δ p s ; e q σ \delta p_{\mathrm{s;eq}}^{\sigma} δ p s ; eq σ , which are shown by Fig. D.1 (top panel).
In particular, the resonant LOD is approximated to the LOD at which the amplitude of semidiurnal tidal surface-pressure oscillations is maximal, and the normalised FWHM is expressed as
Δ X 1 / 2 ; M = L O D L ; 2 ( L O D 2 − L O D 1 ) L O D 1 L O D 2 , (96) \Delta X_{1/2;\mathrm{M}} = \frac{\mathrm{LOD}_{\mathrm{L};2}(\mathrm{LOD}_2-\mathrm{LOD}_1)}{\mathrm{LOD}_1\mathrm{LOD}_2}, \tag*{(96)} Δ X 1/2 ; M = LOD 1 LOD 2 LOD L ; 2 ( LOD 2 − LOD 1 ) , (96)
with L O D 1 \mathrm{LOD}_1 LOD 1 and L O D 2 \mathrm{LOD}_2 LOD 2 being the two LODs at which the modulus is half its maximum value.
For the resonant wavelength, we find 22.1 hr, which is just 0.5% larger than best-fitting values in Sect. 4.2. For the maximal pressure amplitude at the equator and the normalised FWHM of the peak, we get ∣ δ p s ; e q σ ∣ max = 337 |\delta p_{\mathrm{s;eq}}^{\sigma}|_{\max}=337 ∣ δ p s ; eq σ ∣ m a x = 337 Pa and Δ X 1 / 2 ; M = 7.88 × 10 − 2 \Delta X_{1/2;\mathrm{M}}=7.88\times10^{-2} Δ X 1/2 ; M = 7.88 × 1 0 − 2 , respectively. We deduce ε \varepsilon ε from Δ X 1 / 2 ; M \Delta X_{1/2;\mathrm{M}} Δ X 1/2 ; M in the two models by retaining only the leading-order terms of the power-series expansions in ε \varepsilon ε reported in Table 3 ,
ε ≈ Δ X 1 / 2 ; M 3 κ \varepsilon\approx\frac{\Delta X_{1/2;\mathrm{M}}}{\sqrt{3}\kappa} ε ≈ 3 κ Δ X 1/2 ; M
(isentropic),
ε ≈ Δ X 1 / 2 ; M 3 κ Γ 1 \varepsilon\approx\frac{\Delta X_{1/2;\mathrm{M}}}{\sqrt{3}\kappa\Gamma_1} ε ≈ 3 κ Γ 1 Δ X 1/2 ; M
(isothermal).
The estimated values – ε = 0.243 \varepsilon= 0.243 ε = 0.243 (isentropic) and ε = 0.198 \varepsilon= 0.198 ε = 0.198 (isothermal) – are 5% smaller than those obtained numerically using the minimisation method, consistent with the error of order ∼ ε 2 \sim\varepsilon^2 ∼ ε 2 introduced by reducing series to their leading-order terms. The Newtonian cooling frequency and the pressure scale height at the planet’s surface are then deduced from ε \varepsilon ε and LODL ; 2 _{\mathrm{L};2} L ; 2 using Eqs. (20) and (80) , respectively.
We note that the accuracy of the inferred parameters is inherently limited by the frequency resolution of the GCM datasets. Their relative uncertainty may be approximated by the ratio of the LOD spacing between adjacent data points to the resonant LOD. For the dataset considered here, this ratio is 0.023, corresponding to an uncertainty of approximately 2.3% in the determination of the resonant LOD. The resulting error propagates to all inferred quantities and adds to the aforementioned ε 2 \varepsilon^2 ε 2 -uncertainty.
As discussed in Sect. 2.7, the phase lag of tidal heating (see Eq. (49) ) does not affect the modulus of surface pressure oscillations. Therefore, to compute this phase lag, we use the imaginary component of the numerical solution of Deitrick & Goldblatt [17 ] (Fig. 1c; see the grey markers in the bottom panel of Fig. D.1). The phase angle φ H \varphi_{\mathrm{H}} φ H is readily deduced from the ratio between the two extrema of the imaginary part using Eqs. (61) and (65) . Interpolating GCM data with a cubic spline, we find ℑ { δ p s ; e q σ } + = 137.17 \Im\{\delta p_{\mathrm{s;eq}}^{\sigma}\}_{+} = 137.17 ℑ { δ p s ; eq σ } + = 137.17 Pa for the positive peak and ℑ { δ p s ; e q σ } − = − 215.62 \Im\{\delta p_{\mathrm{s;eq}}^{\sigma}\}_{-} = -215.62 ℑ { δ p s ; eq σ } − = − 215.62 Pa for the negative one, which yields φ l a g = 0.224 \varphi_{\mathrm{lag}} = 0.224 φ lag = 0.224 rad. As a result, φ H = 0.513 \varphi_{\mathrm{H}} = 0.513 φ H = 0.513 rad in the isentropic model and φ H = 0.387 \varphi_{\mathrm{H}} = 0.387 φ H = 0.387 rad in the isothermal one.
To estimate δ p ~ T \widetilde{\delta p}_{\mathrm{T}} δ p T , we exploit its relationship with the height of the resonant peak (Table 3 ), retaining only the leading-order term in ε \varepsilon ε ,
δ p ~ T ≈ κ ε ∣ δ p ~ s ; 2 ( X M ; + ) ∣ (isentropic), (97) \widetilde{\delta p}_{\mathrm{T}} \approx\kappa\varepsilon\left|\widetilde{\delta p}_{\mathrm{s};2}(\mathrm{X}_{\mathrm{M};+})\right|
\qquad\text{(isentropic),} \tag*{(97)} δ p T ≈ κ ε δ p s ; 2 ( X M ; + ) (isentropic), (97)
δ p ~ T ≈ Γ 1 − 1 2 − Γ 1 ε ∣ δ p ~ s ; 2 ( X M ; + ) ∣ (isothermal). (98) \widetilde{\delta p}_{\mathrm{T}} \approx\frac{\Gamma_1-1}{2-\Gamma_1}\varepsilon\left|\widetilde{\delta p}_{\mathrm{s};2}(\mathrm{X}_{\mathrm{M};+})\right|
\qquad\text{(isothermal).} \tag*{(98)} δ p T ≈ 2 − Γ 1 Γ 1 − 1 ε δ p s ; 2 ( X M ; + ) (isothermal). (98)
As in Sect. 4.2, we express δ p ~ s ; 2 \widetilde{\delta p}_{\mathrm{s};2} δ p s ; 2 as a function of δ p s ; e q σ \delta p_{\mathrm{s;eq}}^{\sigma} δ p s ; eq σ using Eq. (79) , which yields
δ p ~ T ≈ κ ε 16 c f 15 ∣ δ p s ; e q σ ∣ max ( Θ 2 2 , 1 , P 2 2 ) p s (isentropic), (99) \widetilde{\delta p}_{\mathrm{T}} \approx\kappa\varepsilon\sqrt{\frac{16c_{\mathrm{f}}}{15}\frac{\left|\delta p_{\mathrm{s;eq}}^{\sigma}\right|_{\max}}{\left(\Theta_2^{2,1},P_2^2\right)p_{\mathrm{s}}}}
\qquad\text{(isentropic),} \tag*{(99)} δ p T ≈ κ ε 15 16 c f ( Θ 2 2 , 1 , P 2 2 ) p s δ p s ; eq σ m a x (isentropic), (99)
δ p ~ T ≈ Γ 1 − 1 2 − Γ 1 ε 16 c f 15 ∣ δ p s ; e q σ ∣ max ( Θ 2 2 , 1 , P 2 2 ) p s (isothermal). (100) \widetilde{\delta p}_{\mathrm{T}} \approx\frac{\Gamma_1-1}{2-\Gamma_1}\varepsilon\sqrt{\frac{16c_{\mathrm{f}}}{15}\frac{\left|\delta p_{\mathrm{s;eq}}^{\sigma}\right|_{\max}}{\left(\Theta_2^{2,1},P_2^2\right)p_{\mathrm{s}}}}
\qquad\text{(isothermal).} \tag*{(100)} δ p T ≈ 2 − Γ 1 Γ 1 − 1 ε 15 16 c f ( Θ 2 2 , 1 , P 2 2 ) p s δ p s ; eq σ m a x (isothermal). (100)
The analytical fits, plotted in Fig. D.1, agree very well with the GCM outputs. Although not as good as the numerical fits, they do not differ much from the latter, indicating that analytical scalings can provide accurate estimates of the inferred model parameters even if they are limited to leading-order terms.