Introduction
Holography provides a useful way to study strongly coupled systems in which perturbative methods are not available [1 ] . In its simplest form, the gravitational description is given by an asymptotically anti-de Sitter (AdS) spacetime with a planar boundary. However, spatial anisotropy and nontrivial scaling behavior arise in a variety of physical systems, including strongly correlated matter. Examples include electronic nematic phases, where rotational symmetry is spontaneously broken and the low-energy excitations exhibit direction-dependent behavior [2 , 3 ] , and quantum critical metals near spin-density-wave ordering transitions [4 ] . Lifshitz critical points can exhibit anisotropic scaling and sensitivity to spatial anisotropy [5 , 6 ] , while semi-Dirac spectra with linear and quadratic dispersion along different directions have been predicted in oxide nanostructures and modeled theoretically [7 , 8 ] . In a separate setting, non-Hermitian models can exhibit anisotropic, system-size-dependent scaling of localization lengths [9 ] . Spatially modulated phases also appear in strongly correlated electron systems, including electronic stripe and charge-density-wave phases [10 , 11 , 12 ] , as well as in chiral magnets supporting helical and spiral magnetic configurations [13 , 14 ] . Related spatially modulated structures arise in chiral soliton lattices under strong magnetic fields [15 ] . Although their microscopic mechanisms differ, these examples motivate the study of gravitational backgrounds with homogeneous but anisotropic spatial geometries and nontrivial scaling exponents. In particular, Einstein–Maxwell–dilaton (EMD) gravity provides a tractable framework for investigating the thermodynamic and dynamical properties of strongly coupled systems exhibiting such behavior.
Lifshitz and hyperscaling-violating backgrounds are familiar examples of gravitational geometries with nonrelativistic scaling [16 , 17 , 18 , 19 ] ; anisotropic spacetimes were also introduced in Hořava gravity [20 , 21 ] and as anisotropic conformal infinity [22 , 23 ] . The holographic dictionary for Lifshitz-type backgrounds is developed in [24 , 25 , 26 , 27 ] ; see [28 ] for a review of Lifshitz holography. More general anisotropic spacetimes can be obtained if the spatial sections are homogeneous but not necessarily isotropic. In five bulk dimensions, the three-dimensional spatial geometries can be classified in terms of Bianchi groups, closely related to the Thurston geometries [29 , 30 ] . Homogeneous spacetimes carrying Lifshitz symmetry are classified by algebraic methods in [31 ] . These provide a convenient setting in which one can preserve homogeneity without requiring ordinary spatial translation invariance [32 , 33 ] . Such geometries have also been used to describe helical lattices and spatially modulated states holographically [34 , 35 , 36 , 37 , 38 ] .
Homogeneous black holes with Nil and Solv horizons were constructed in Einstein gravity by Cadeau and Woolgar [39 ] . These were subsequently generalized to scaling and hyperscaling-violating solutions [40 ] . Higher-dimensional examples based on Einstein solvmanifolds and Ricci nilsolitons were considered in [41 , 42 ] ; more recent work constructs black holes with nilpotent horizons in higher dimensions [43 ] . In a previous paper [44 ] , we constructed a family of five-dimensional vacuum black branes with a negative cosmological constant and Bianchi type VIh _h h spatial geometry, characterized by a parameter h h h . We also obtained a Ricci-flat branch when the cosmological constant vanishes, which exhibits hyperscaling violation. The Bianchi VIh _h h solutions include the known Solv case as a special point and show that the Einstein equations can admit nontrivial homogeneous anisotropic scaling even in the absence of matter fields.
A dilaton and electromagnetic fields enlarge the set of allowed scaling geometries. An electric field can change the dynamical exponent, while a magnetic field distinguishes spatial planes and must satisfy the Maxwell equation and Bianchi identity on the chosen homogeneous space. Matter-supported black holes with Thurston horizon geometries, including dyonic solutions, have been studied in [45 , 46 , 47 ] . Magnetic backgrounds also provide a setting for anisotropic holographic transport, and other anisotropic holographic backgrounds are constructed from axion or dilaton deformations and from top-down setups [48 , 49 , 50 , 51 ] ; anisotropic critical points from holography are studied in [52 ] . The distinction between a field that supports the asymptotic geometry and an independently variable charge is essential here: the explicit EMD families considered below belong to the former class.
We formulate the neutral and EMD equations in arbitrary spacetime dimension using a diagonal metric in a left-invariant spatial coframe. Let r r r denote the radial coordinate and α i \alpha_i α i the scaling exponent of the i i i th spatial direction; the corresponding spatial metric coefficient is proportional to r 2 α i r^{2\alpha_i} r 2 α i . The spatial Lie algebra restricts the radial exponents through scaling compatibility and the mixed Einstein equations. The remaining equations determine the intrinsic Ricci tensor, blackening function, and matter amplitudes. For the neutral solutions, we distinguish a branch with radially constant intrinsic spatial curvature and a negative cosmological constant from one with radially varying intrinsic curvature and a vanishing cosmological constant, and explain the relation to algebraic Ricci solitons. The matter equations are organized into electric, magnetic, and dyonic sectors, retaining the geometric restrictions on homogeneous magnetic fields. The examples include five-dimensional Bianchi geometries, a dyonic Nil solution with a single Abelian gauge field, flat higher-dimensional dyonic solutions, and a construction that adjoins a magnetic two-plane to a Ricci nilsoliton.
Let r h > 0 r_h>0 r h > 0 be the horizon radius and f 0 > 0 f_0>0 f 0 > 0 the asymptotic value of the blackening function. For a positive exponent Δ \Delta Δ equal to the dynamical exponent plus the sum of the spatial scaling exponents, we consider solutions with blackening function
f ( r ) = f 0 [ 1 − ( r h r ) Δ ] . (1) f(r) = f_0 \left[1 - \left(\frac{r_h}{r}\right)^{\Delta}\right]. \tag*{(1)} f ( r ) = f 0 [ 1 − ( r r h ) Δ ] . (1)
The neutral planar AdS black brane with n n n spatial directions and AdS radius L L L has f 0 = L − 2 f_0 = L^{-2} f 0 = L − 2 and Δ = n + 1 \Delta= n + 1 Δ = n + 1 in the metric convention used here [53 ] . Thus, (1) generalizes its blackening function by allowing a different exponent and normalization, without requiring the geometry to be asymptotically AdS. For these solutions, we calculate the temperature, entropy, and Euclidean action with the supporting sources held fixed. Including the electric Legendre term gives a fixed-flux free energy relative to the corresponding horizonless scaling background. The resulting energy agrees with the fixed-source Hamiltonian variation. In this family, the entropy scales as a power of temperature, with exponent given by the sum of the spatial scaling exponents divided by the dynamical exponent. The heat capacity is positive when both the sum of the spatial exponents and the dynamical exponent are positive. This calculation determines relative thermodynamic quantities within the specified family; a variational problem with arbitrary boundary sources requires a more general renormalization prescription [54 ] .
We then examine a massive probe scalar to investigate how fluctuations propagate and relax on these anisotropic branes. The momentum dependence of the scalar quasinormal frequencies provides dispersion relations and a first view of the dynamical response relevant to transport. It does not, by itself, determine hydrodynamic modes or transport coefficients of the dual theory. The scalar also provides a tractable probe of stability against disturbances in this sector.
In the spatially homogeneous sector, the asymptotic equation gives a mass bound analogous to the Breitenlohner–Freedman bound, and the quasinormal frequencies scale linearly with temperature at fixed background parameters. When the dynamical exponent equals the sum of the spatial scaling exponents and the blackening function has no logarithmic term, the radial equation becomes hypergeometric. For masses satisfying the asymptotic bound, it yields an exact, purely damped spectrum under source-free boundary conditions. We use this result as a benchmark for a pseudospectral calculation of the electric Nil, Solv, and Bianchi VIh _h h families. The numerical study follows the dependence on probe mass and dynamical exponent and examines finite-momentum sectors of Nil and Solv. Resolution comparisons and a separate shooting calculation assess the accuracy and limitations of the verified modes (Appendix § 10 ). The exact spectrum is consistent with stability in the homogeneous scalar sector under the stated boundary conditions, but the numerical selection keeps only damped modes and cannot rule out growing modes. Neither calculation establishes stability of the coupled metric, gauge field, and dilaton system.
The diagonal metric ansatz with these radial scaling coefficients restricts the class of backgrounds considered here. The solutions are generally not asymptotically locally AdS, although they may serve as scaling regions in more general holographic geometries. Their explicit form allows us to separate the restrictions imposed by spatial geometry from those imposed by matter and to study how both enter the thermodynamics and scalar response.
The organization of the paper is as follows. In § 2 we derive the neutral Einstein equations and examine the two principal anisotropic branches, including their relation to earlier homogeneous black holes. In § 3 we introduce a dilaton and an Abelian gauge field, and discuss the electric, magnetic, and dyonic cases. § 4 contains representative five-dimensional and higher-dimensional examples. In § 5 we derive the horizon thermodynamics and evaluate the fixed-flux Euclidean on-shell action. § 6 studies massive probe scalar perturbations, including their asymptotic falloffs, quasinormal boundary conditions, an analytic spectrum when the dynamical exponent equals the sum of the spatial scaling exponents, and a pseudospectral computation of the spectra with its validation (Appendix § 10 ). We conclude in § 7 with a summary and possible extensions. Appendix § 8 summarizes the radial integration, including the case where a source term has the same radial power as a homogeneous mode, Appendix § 9 gives the fixed-source Hamiltonian variation underlying the relative mass, and Appendix § 10 documents the convergence and reliability checks of the numerical spectra.
Neutral homogeneous anisotropic black branes
Einstein gravity and homogeneous geometry
We first consider Einstein gravity with a cosmological constant in D : = n + 2 D := n + 2 D := n + 2 dimensions. We work in units 8 π G N = 1 8\pi G_N = 1 8 π G N = 1 . The action is
I : = 1 2 ∫ d n + 2 x − g ( R − 2 Λ ) + I b d y , (2) I := \frac{1}{2} \int\mathrm{d}^{n+2}x \sqrt{-g}\left(R - 2\Lambda\right) + I_{\mathrm{bdy}}, \tag*{(2)} I := 2 1 ∫ d n + 2 x − g ( R − 2Λ ) + I bdy , (2)
where I b d y I_{\mathrm{bdy}} I bdy contains the boundary terms required for a suitable variational principle. The equations of motion are
R μ ν − 1 2 R g μ ν + Λ g μ ν = 0 , R μ ν = 2 Λ n g μ ν . (3) R_{\mu\nu} - \frac{1}{2}R g_{\mu\nu} + \Lambda g_{\mu\nu} = 0,\qquad R_{\mu\nu} = \frac{2\Lambda}{n}g_{\mu\nu}. \tag*{(3)} R μν − 2 1 R g μν + Λ g μν = 0 , R μν = n 2Λ g μν . (3)
Let G G G be an n n n -dimensional Lie group with Lie algebra g 1 \mathfrak{g}^{1} g 1 with a left-invariant coframe θ i \theta^{i} θ i satisfying
d θ k = − 1 2 c i j k θ i ∧ θ j . (4) \mathrm{d}\theta^{k} = -\frac{1}{2}c_{ij}^{k}\theta^{i}\wedge\theta^{j}. \tag*{(4)} d θ k = − 2 1 c ij k θ i ∧ θ j . (4)
The structure constants c i j k c_{ij}^{k} c ij k are constant in this invariant coframe. The metric
d s 2 = − G t ( r ) d t 2 + G r ( r ) d r 2 + ∑ i = 1 n G i ( r ) ( θ i ) 2 (5) \mathrm{d}s^{2} = -G_{t}(r)\mathrm{d}t^{2} + G_{r}(r)\mathrm{d}r^{2} + \sum_{i=1}^{n}G_{i}(r)(\theta^{i})^{2} \tag*{(5)} d s 2 = − G t ( r ) d t 2 + G r ( r ) d r 2 + i = 1 ∑ n G i ( r ) ( θ i ) 2 (5)
is homogeneous on the spatial slices of constant r r r , but not necessarily isotropic. Unless stated otherwise, frame indices t t t , r r r , i i i , j j j , k k k , … \ldots … denote components in the orthonormal frame of the full metric. Thus, the field equations reduce to ordinary differential equations in the radial coordinate r r r . We assume that the intrinsic Ricci tensor of the reference spatial metric can be made diagonal in the chosen coframe.
Following the constructions in [40 , 39 , 44 ] , we choose the following diagonal metric ansatz
d s 2 = − r 2 z f ( r ) d t 2 + d r 2 r 2 f ( r ) + ∑ i = 1 n b i 2 r 2 α i ( θ i ) 2 . (6) \mathrm{d}s^{2} = -r^{2z}f(r)\mathrm{d}t^{2} + \frac{\mathrm{d}r^{2}}{r^{2}f(r)} + \sum_{i=1}^{n}b_{i}^{2}r^{2\alpha_{i}}(\theta^{i})^{2}. \tag*{(6)} d s 2 = − r 2 z f ( r ) d t 2 + r 2 f ( r ) d r 2 + i = 1 ∑ n b i 2 r 2 α i ( θ i ) 2 . (6)
The parameters z z z and α i \alpha_i α i control the radial scaling of time and the spatial directions, respectively, while the constants b i > 0 b_i > 0 b i > 0 determine the spatial curvature scales. The function f ( r ) f(r) f ( r ) is the blackening factor. Let X i X_i X i denote the spatial vector fields dual to the invariant coframe, θ i ( X j ) = δ j i \theta^i(X_j) = \delta^i_j θ i ( X j ) = δ j i . In the exterior region f > 0 f > 0 f > 0 , the orthonormal frame vectors of the full metric are
e t : = 1 r z f ∂ t , e r : = r f ∂ r , e i : = 1 b i r α i X i . (7) e_t := \frac{1}{r^z\sqrt{f}}\partial_t,\qquad e_r := r\sqrt{f}\,\partial_r,\qquad e_i := \frac{1}{b_i r^{\alpha_i}}X_i. \tag*{(7)} e t := r z f 1 ∂ t , e r := r f ∂ r , e i := b i r α i 1 X i . (7)
Thus g ( e t , e t ) = − 1 g(e_t,e_t) = -1 g ( e t , e t ) = − 1 and g ( e r , e r ) = g ( e i , e i ) = 1 g(e_r,e_r) = g(e_i,e_i) = 1 g ( e r , e r ) = g ( e i , e i ) = 1 (no sum). For notational simplicity we define
S : = ∑ i = 1 n α i , Q : = ∑ i = 1 n α i 2 , Δ : = z + S , D : = r d d r . (8) S := \sum_{i=1}^{n}\alpha_i,\qquad Q := \sum_{i=1}^{n}\alpha_i^2,\qquad\Delta:= z + S,\qquad\mathcal{D} := r\frac{\mathrm{d}}{\mathrm{d}r}. \tag*{(8)} S := i = 1 ∑ n α i , Q := i = 1 ∑ n α i 2 , Δ := z + S , D := r d r d . (8)
In most of this paper we take r > r h r > r_h r > r h , z > 0 z > 0 z > 0 , S > 0 S > 0 S > 0 , and f > 0 f > 0 f > 0 outside the horizon where r h r_h r h is the horizon radius. In addition to constraints from the Einstein equations, we will also impose additional constraints on α i \alpha_i α i .1
Scaling compatibility and mixed Einstein equations
The metric ansatz (6) assigns a radial scaling exponent to each spatial direction. This is straightforward for a flat horizon, but a non-Abelian homogeneous space has an additional feature: two spatial directions can be mixed by the geometry itself. To obtain solutions of the form (6) , we impose
c i j k ≠ 0 ⟹ α i + α j − α k = κ . (9) c_{ij}^{k} \ne0 \quad\Longrightarrow\quad\alpha_i+\alpha_j-\alpha_k=\kappa. \tag*{(9)} c ij k = 0 ⟹ α i + α j − α k = κ . (9)
As a consequence the connection associated with the spatial geometry then scales as r − κ r^{-\kappa} r − κ , and its intrinsic Ricci tensor carries a common factor r − 2 κ r^{-2\kappa} r − 2 κ . In particular, κ = 0 \kappa=0 κ = 0 means that the intrinsic curvature contributes at the same radial order as the cosmological constant. This is the branch that admits the asymptotically constant blackening factors studied below. We write the reference spatial metric and its Ricci eigenvalues as
h b : = ∑ i b i 2 ( θ i ) 2 , Ric ( h b ) i ^ j ^ = : R i ^ j ^ . (10) h_b:=\sum_i b_i^2(\theta^i)^2,\qquad\operatorname{Ric}(h_b)_{\hat i\hat j}=:R_{\hat i\hat j}. \tag*{(10)} h b := i ∑ b i 2 ( θ i ) 2 , Ric ( h b ) i ^ j ^ =: R i ^ j ^ . (10)
Here hatted indices refer to the orthonormal frame e ^ i : = b i − 1 X i \hat e_i:=b_i^{-1}X_i e ^ i := b i − 1 X i of h b h_b h b .
For κ = 0 \kappa=0 κ = 0 , (9) has a simple scaling interpretation. A radial dilatation r → ℓ r r\to\ell r r → ℓ r can be accompanied by a transformation of the invariant one-forms θ i → ℓ − α i θ i \theta^i\to\ell^{-\alpha_i}\theta^i θ i → ℓ − α i θ i , so that the zero-temperature metric retains its form. The condition on the structure constants ensures that this transformation preserves the relations between the one-forms. Equivalently, the map
A ( X i ) : = α i X i (11) A(X_i):=\alpha_iX_i \tag*{(11)} A ( X i ) := α i X i (11)
on the spatial Lie algebra satisfies
A ( [ X i , X j ] ) = [ A X i , X j ] + [ X i , A X j ] . (12) A([X_i,X_j])=[AX_i,X_j]+[X_i,AX_j]. \tag*{(12)} A ([ X i , X j ]) = [ A X i , X j ] + [ X i , A X j ] . (12)
This is called a derivation; it need not be an inner derivation. An associated dilatation generator X A X_A X A can be adjoined to g \mathfrak{g} g , with its action defined by the bracket [ X A , X i ] : = A ( X i ) = α i X i [X_A,X_i]:=A(X_i)=\alpha_iX_i [ X A , X i ] := A ( X i ) = α i X i . For κ ≠ 0 \kappa\ne0 κ = 0 , the corresponding extension is instead generated by the derivation A − κ 1 A-\kappa1 A − κ 1 . Thus the anisotropic radial scaling must preserve the homogeneous spatial geometry. Such structures also appear in earlier homogeneous Einstein constructions [41 , 42 ] .
Nevertheless, the parametrization of the solutions carries a residual redundancy. Performing the radial diffeomorphism r → r γ r\to r^\gamma r → r γ , together with t → t / γ t\to t/\gamma t → t / γ so that the ansatz keeps its canonical form, rescales the parameters as
( z , α i , S , Δ , σ ) → 1 γ ( z , α i , S , Δ , σ ) , f 0 → γ 2 f 0 , r h → r h γ . (13) (z,\alpha_i,S,\Delta,\sigma)\to\frac{1}{\gamma}(z,\alpha_i,S,\Delta,\sigma),\qquad f_0\to\gamma^2f_0,\qquad r_h\to r_h^\gamma. \tag*{(13)} ( z , α i , S , Δ , σ ) → γ 1 ( z , α i , S , Δ , σ ) , f 0 → γ 2 f 0 , r h → r h γ . (13)
while the constants b i b_i b i , ϕ 0 \phi_0 ϕ 0 , E \mathcal{E} E , B B B , and Λ \Lambda Λ are unchanged. The overall scale of the exponent vector is therefore a coordinate artifact. If κ ≠ 0 \kappa\ne0 κ = 0 , this freedom can be used to set κ = 1 \kappa=1 κ = 1 . If κ = 0 \kappa=0 κ = 0 , the value of κ \kappa κ cannot be changed, and the freedom is instead used to fix the overall scale of the exponents, which we denote by a a a , to unity. The cases κ = 0 \kappa=0 κ = 0 and κ ≠ 0 \kappa\ne0 κ = 0 therefore define distinct scaling branches.
The mixed radial–spatial Einstein equations impose a further compatibility condition. The diagonal metric and radial matter fields carry no momentum flux between the radial direction and a spatial direction. The geometry must therefore produce no such off-diagonal curvature either. For the ansatz (6) , this gives
R r i = f b i r α i ∑ j c i j j ( α j − α i ) = 0 ⟹ ∑ j c i j j ( α j − α i ) = 0. (14) R_{ri}=\frac{\sqrt{f}}{b_i r^{\alpha_i}}\sum_j c_{ij}^{j}(\alpha_j-\alpha_i)=0
\quad\Longrightarrow\quad
\sum_j c_{ij}^{j}(\alpha_j-\alpha_i)=0. \tag*{(14)} R r i = b i r α i f j ∑ c ij j ( α j − α i ) = 0 ⟹ j ∑ c ij j ( α j − α i ) = 0. (14)
This condition is automatic for an Abelian group and for isotropic spatial scaling, but can be restrictive for a non-Abelian geometry. For example, in Bianchi V I h \mathrm{VI}_h VI h it gives α 1 + h α 2 = ( 1 + h ) α 3 \alpha_1+h\alpha_2=(1+h)\alpha_3 α 1 + h α 2 = ( 1 + h ) α 3 , or α 1 = − h α 2 \alpha_1=-h\alpha_2 α 1 = − h α 2 for α 3 = 0 \alpha_3=0 α 3 = 0 [44 ] .
Trace-reversed Einstein equations
Using an orthonormal frame, the nontrivial diagonal components of the Ricci tensor are
R t t = 1 2 r 2 f ′ ′ + 1 2 ( 1 + 3 z + S ) r f ′ + z ( z + S ) f , (15) R_{tt} = \frac{1}{2}r^2 f'' + \frac{1}{2}(1+3z+S)rf' + z(z+S)f, \tag*{(15)} R tt = 2 1 r 2 f ′′ + 2 1 ( 1 + 3 z + S ) r f ′ + z ( z + S ) f , (15)
R r r = − 1 2 r 2 f ′ ′ − 1 2 ( 1 + 3 z + S ) r f ′ − ( z 2 + Q ) f , (16) R_{rr} = -\frac{1}{2}r^2 f'' - \frac{1}{2}(1+3z+S)rf' - (z^2+Q)f, \tag*{(16)} R r r = − 2 1 r 2 f ′′ − 2 1 ( 1 + 3 z + S ) r f ′ − ( z 2 + Q ) f , (16)
R i i = R ^ i ^ i ^ r − 2 κ − α i [ r f ′ + ( z + S ) f ] . (17) R_{ii} = \widehat{R}_{\hat{i}\hat{i}}r^{-2\kappa} - \alpha_i\left[rf' + (z+S)f\right]. \tag*{(17)} R ii = R i ^ i ^ r − 2 κ − α i [ r f ′ + ( z + S ) f ] . (17)
In our orthonormal signature the trace-reversed vacuum equations are
R t t = − 2 Λ / n , R r r = 2 Λ / n , R i i = 2 Λ / n . (18) R_{tt}=-2\Lambda/n,\quad R_{rr}=2\Lambda/n,\quad R_{ii}=2\Lambda/n. \tag*{(18)} R tt = − 2Λ/ n , R r r = 2Λ/ n , R ii = 2Λ/ n . (18)
Adding the first two equations immediately yields
z S = Q . (19) zS=Q. \tag*{(19)} z S = Q . (19)
This equation is useful because it fixes the time scaling exponent before solving for f f f . For nonzero spatial scaling, S ≠ 0 S\ne0 S = 0 and z = Q / S z=Q/S z = Q / S .
Substituting (19) into the temporal equation gives
( D + 2 z ) ( D + Δ ) f = − 4 Λ n . (20) (\mathcal{D}+2z)(\mathcal{D}+\Delta)f=-\frac{4\Lambda}{n}. \tag*{(20)} ( D + 2 z ) ( D + Δ ) f = − n 4Λ . (20)
For Q ≠ S 2 Q\ne S^2 Q = S 2 , the solution is
f ( r ) = − 2 Λ n z Δ + C 1 r 2 z + C 2 r Δ . (21) f(r)=-\frac{2\Lambda}{nz\Delta}+\frac{C_1}{r^{2z}}+\frac{C_2}{r^\Delta}. \tag*{(21)} f ( r ) = − n z Δ 2Λ + r 2 z C 1 + r Δ C 2 . (21)
The homogeneous radial modes have powers 2 z 2z 2 z and Δ \Delta Δ , while the cosmological constant contributes a constant particular solution. The spatial equations further constrain the integration constants.
The spatial equations can be written as
α i ( D + Δ ) f = R ^ i ^ i ^ r − 2 κ − 2 Λ n \alpha_i(\mathcal{D}+\Delta)f=\widehat{R}_{\hat{i}\hat{i}}r^{-2\kappa}-\frac{2\Lambda}{n} α i ( D + Δ ) f = R i ^ i ^ r − 2 κ − n 2Λ
⟹ R ^ i ^ i ^ r − 2 κ − α i ( S − z ) C 1 r − 2 z = 2 Λ n ( 1 − α i z ) . (22) \Longrightarrow\ \widehat{R}_{\hat{i}\hat{i}}r^{-2\kappa}-\alpha_i(S-z)C_1r^{-2z}=\frac{2\Lambda}{n}\left(1-\frac{\alpha_i}{z}\right). \tag*{(22)} ⟹ R i ^ i ^ r − 2 κ − α i ( S − z ) C 1 r − 2 z = n 2Λ ( 1 − z α i ) . (22)
Note that ( D + Δ ) r − Δ = 0 (\mathcal{D}+\Delta)r^{-\Delta}=0 ( D + Δ ) r − Δ = 0 , so only the C 1 C_1 C 1 mode ∝ r − 2 z \propto r^{-2z} ∝ r − 2 z contributes ( D + Δ ) ( r − 2 z ) = ( S − z ) r − 2 z (\mathcal{D}+\Delta)(r^{-2z})=(S-z)r^{-2z} ( D + Δ ) ( r − 2 z ) = ( S − z ) r − 2 z and thus appears in (22) .
For S ≠ 0 S\ne0 S = 0 , the case Q = S 2 Q=S^2 Q = S 2 corresponds to a repeated radial root. Here z = S z=S z = S and
f ( r ) = − Λ n S 2 + r − 2 S ( C 1 + C 2 log r ) , (23) f(r)=-\frac{\Lambda}{nS^2}+r^{-2S}(C_1+C_2\log r), \tag*{(23)} f ( r ) = − n S 2 Λ + r − 2 S ( C 1 + C 2 log r ) , (23)
⟹ R ^ i ^ i ^ r − 2 κ − α i C 2 r − 2 S = 2 Λ n ( 1 − α i S ) . (24) \Longrightarrow\ \widehat{R}_{\hat{i}\hat{i}}r^{-2\kappa}-\alpha_i C_2r^{-2S}=\frac{2\Lambda}{n}\left(1-\frac{\alpha_i}{S}\right). \tag*{(24)} ⟹ R i ^ i ^ r − 2 κ − α i C 2 r − 2 S = n 2Λ ( 1 − S α i ) . (24)
The logarithmic term cannot be obtained by substituting z = S z=S z = S into the distinct-root solution. As shown below, the spatial equations eliminate it when κ = 0 \kappa=0 κ = 0 and S ≠ 0 S\ne0 S = 0 , but it may appear in special Λ = 0 \Lambda=0 Λ = 0 geometries.
Neutral solution branches
**Anisotropic solutions for κ = 0 \kappa=0 κ = 0 and Λ < 0 ∗ ∗ \Lambda<0** Λ < 0 ∗ ∗ We first take κ = 0 \kappa=0 κ = 0 and Λ < 0 \Lambda<0 Λ < 0 . For a genuinely anisotropic solution with S ≠ 0 S\ne0 S = 0 , the spatial equations fix C 1 = 0 C_1=0 C 1 = 0 in the case where the radial equation has distinct roots, Q ≠ S 2 Q\ne S^2 Q = S 2 , and require
R ^ i ^ i ^ = 2 Λ n ( 1 − α i z ) . (25) \widehat{R}_{\hat{i}\hat{i}}=\frac{2\Lambda}{n}\left(1-\frac{\alpha_i}{z}\right). \tag*{(25)} R i ^ i ^ = n 2Λ ( 1 − z α i ) . (25)
For the repeated-root case Q = S 2 Q=S^2 Q = S 2 with S ≠ 0 S\ne0 S = 0 , (19) gives z = S z=S z = S and Δ = 2 S \Delta=2S Δ = 2 S . Setting κ = 0 \kappa=0 κ = 0 in (24) shows that α i C 2 = 0 \alpha_i C_2=0 α i C 2 = 0 for every i i i , since the remaining terms are independent of r r r . At least one α i \alpha_i α i is nonzero, so C 2 = 0 C_2=0 C 2 = 0 and the logarithmic mode is excluded. The remaining spatial condition is again (25) , now with z = S z=S z = S , and (23) reduces to
f ( r ) = − Λ n S 2 + C 1 r 2 S . (26) f(r)=-\frac{\Lambda}{nS^2}+\frac{C_1}{r^{2S}}. \tag*{(26)} f ( r ) = − n S 2 Λ + r 2 S C 1 . (26)
Here C 1 C_1 C 1 multiplies the surviving repeated-root mode; it is not the coefficient set to zero in the distinct-root case. The constant part of f f f is f 0 = − Λ / ( n S 2 ) > 0 f_0=-\Lambda/(nS^2)>0 f 0 = − Λ/ ( n S 2 ) > 0 . For S > 0 S>0 S > 0 , choosing C 1 < 0 C_1<0 C 1 < 0 gives a horizon at r h = ( − C 1 / f 0 ) 1 / ( 2 S ) r_h=(-C_1/f_0)^{1/(2S)} r h = ( − C 1 / f 0 ) 1/ ( 2 S ) .
Thus, in either the distinct-root or repeated-root case, the blackening function of the horizon solution can be written as
f ( r ) = f 0 [ 1 − ( r h r ) Δ ] , f 0 = 2 Λ n z Δ . (27) f(r)=f_0\left[1-\left(\frac{r_h}{r}\right)^\Delta\right],\qquad f_0=\frac{2\Lambda}{n z\Delta}. \tag*{(27)} f ( r ) = f 0 [ 1 − ( r r h ) Δ ] , f 0 = n z Δ 2Λ . (27)
For z > 0 z>0 z > 0 , S > 0 S>0 S > 0 , the normalization f 0 f_0 f 0 is positive and f f f has a simple zero at r = r h r=r_h r = r h . In particular, the repeated-root case has Δ = 2 S \Delta=2S Δ = 2 S and f 0 = − Λ / ( n S 2 ) f_0=-\Lambda/(nS^2) f 0 = − Λ/ ( n S 2 ) . The h = 0 h=0 h = 0 member of the Bianchi VIh _h h family in § 4.1 realizes this case.
The spatial equations have a useful interpretation beyond the algebraic restriction on R ^ i i \widehat{R}_{ii} R ii . In vacuum there is no anisotropic matter stress to distinguish the spatial directions. Nevertheless, the intrinsic curvature of the homogeneous slices is direction dependent and can balance their different radial expansion rates. (25) specifies precisely how much curvature each direction must carry. In particular, when Λ < 0 \Lambda<0 Λ < 0 , a direction with α i > z \alpha_i>z α i > z requires a positive intrinsic Ricci eigenvalue, while directions with α i < z \alpha_i<z α i < z require a negative one. The Nil example in section 4.1 shows this clearly.
In geometric terms, using the scaling map (11) , the same condition reads
Ric ( h b ) = 2 Λ n ( 1 − 1 z A ) . (28) \operatorname{Ric}(h_b)=\frac{2\Lambda}{n}\left(1-\frac{1}{z}A\right). \tag*{(28)} Ric ( h b ) = n 2Λ ( 1 − z 1 A ) . (28)
This is a form of algebraic Ricci soliton condition, as in [42 , 55 ] . For the present calculation, it provides a convenient way to identify homogeneous spatial geometries whose directional curvature can support an anisotropic black brane without additional matter fields.
Isotropic Einstein horizons. The isotropic Einstein horizon case is a separate specialization of the metric ansatz (6) . Take n > 1 n>1 n > 1 and equal nonzero spatial exponents α i = a > 0 \alpha_i=a>0 α i = a > 0 . The neutral constraint (19) then gives z = a z=a z = a . For a non-Abelian spatial coframe, the scaling condition (9) requires κ = a \kappa=a κ = a , so a curved isotropic horizon does not belong to the κ = 0 \kappa=0 κ = 0 branch above. For an Abelian coframe the condition is vacuous, and the spatial metric is flat.
Let the reference spatial metric be Einstein, with Ric ( h b ) = ( n − 1 ) k h b \operatorname{Ric}(h_b)=(n-1)k h_b Ric ( h b ) = ( n − 1 ) k h b , where k k k is its normalized Ricci curvature. For the Lie-group ansatz, the spatial equal exponents give k = a 2 C 1 k=a^2 C_1 k = a 2 C 1 ; in the Abelian case both vanish. Writing the remaining integration constant as μ : = − a 2 C 2 \mu:=-a^2 C_2 μ := − a 2 C 2 , and defining R : = r a R:=r^a R := r a and t ~ : = t / a \tilde{t}:=t/a t ~ := t / a , the metric takes the familiar form
d s 2 = − F ( R ) d t ~ 2 + d R 2 F ( R ) + R 2 h b , F ( R ) = k − μ R n − 1 − 2 Λ R 2 n ( n + 1 ) . (29) \mathrm{d}s^2=-F(R)\mathrm{d}\tilde{t}^{\,2}+\frac{\mathrm{d}R^2}{F(R)}+R^2h_b,\qquad F(R)=k-\frac{\mu}{R^{n-1}}-\frac{2\Lambda R^2}{n(n+1)}. \tag*{(29)} d s 2 = − F ( R ) d t ~ 2 + F ( R ) d R 2 + R 2 h b , F ( R ) = k − R n − 1 μ − n ( n + 1 ) 2Λ R 2 . (29)
For Λ < 0 \Lambda<0 Λ < 0 and suitable μ \mu μ , this includes topological AdS black holes with flat, spherical or hyperbolic horizons [56 , 57 , 58 , 53 ] ; the stability of these black holes is analyzed in [59 ] . The expression also applies to Einstein horizons beyond the Lie-group coframes assumed in the rest of this paper.
Vanishing cosmological constant and κ ≠ 0 \kappa\ne0 κ = 0 . Another possibility is obtained by setting Λ = 0 \Lambda= 0 Λ = 0 . If Q ≠ S 2 Q \ne S^2 Q = S 2 , the temporal equation becomes
f ( r ) = C 1 r − 2 z + C 2 r − Δ . (30) f(r) = C_1 r^{-2z} + C_2 r^{-\Delta}. \tag*{(30)} f ( r ) = C 1 r − 2 z + C 2 r − Δ . (30)
For nonzero intrinsic spatial curvature, the spatial Einstein equations require
κ = z , R ı ^ ı ^ = α i ( S − z ) C 1 . (31) \kappa= z,\qquad R_{\hat{\imath}\hat{\imath}} = \alpha_i(S-z)C_1. \tag*{(31)} κ = z , R ^ ^ = α i ( S − z ) C 1 . (31)
This gives an anisotropic Ricci-flat spacetime even though the individual spatial slices are curved. The spatial curvature is balanced by the radial warp factors. These Λ = 0 \Lambda= 0 Λ = 0 solutions are hyperscaling-violating geometries: the spatial warp factors carry unequal exponents α i \alpha_i α i , so the effective number of spatial dimensions differs from n n n even though the full spacetime is Ricci flat. Vacuum black holes of this type were constructed for the Nil geometry in [40 ] and generalized to the Bianchi V I h \mathrm{VI}_h VI h family in [44 ] ; hyperscaling-violating geometries were introduced holographically in [60 , 17 ] . In the case of a repeated radial root, Q = S 2 Q = S^2 Q = S 2 , the corresponding nonflat branch is
κ = S , R ı ^ ı ^ = α i C 2 , f ( r ) = r − 2 S ( C 1 + C 2 log r ) . (32) \kappa= S,\qquad R_{\hat{\imath}\hat{\imath}} = \alpha_i C_2,\qquad f(r) = r^{-2S}(C_1 + C_2 \log r). \tag*{(32)} κ = S , R ^ ^ = α i C 2 , f ( r ) = r − 2 S ( C 1 + C 2 log r ) . (32)
Zero spatial scaling and other special cases. When S = 0 S = 0 S = 0 , (19) requires Q = 0 Q = 0 Q = 0 for real neutral exponents. Hence α i = 0 \alpha_i = 0 α i = 0 in every spatial direction. The spatial metric must be Einstein, with R ı ^ ı ^ = 2 Λ / n R_{\hat{\imath}\hat{\imath}} = 2\Lambda/n R ^ ^ = 2Λ/ n . Using ρ : = log r \rho:= \log r ρ := log r , the metric can be written as a two-dimensional constant curvature factor times h b h_b h b ,
d s 2 = − U ( ρ ) d t 2 + d ρ 2 U ( ρ ) + h b , U ( ρ ) = − 2 Λ n ρ 2 + c 1 ρ + c 2 . (33) ds^2 = -U(\rho)dt^2 + \frac{d\rho^2}{U(\rho)} + h_b,\qquad U(\rho) = -\frac{2\Lambda}{n}\rho^2 + c_1\rho+ c_2. \tag*{(33)} d s 2 = − U ( ρ ) d t 2 + U ( ρ ) d ρ 2 + h b , U ( ρ ) = − n 2Λ ρ 2 + c 1 ρ + c 2 . (33)
For Λ < 0 \Lambda< 0 Λ < 0 , this includes A d S 2 \mathrm{AdS}_2 AdS 2 product geometries. Such branches are related to generalized (anti-)Nariai constructions [61 , 62 , 63 , 64 ] . Appendix § 8 gives the radial integration rule for the cases of distinct and repeated roots, including the case where a source term has the same radial power as a homogeneous mode. The cases where S S S or z z z vanishes, or where Q = S 2 Q = S^2 Q = S 2 , must be treated separately, because the general formulas divide by S S S , z z z , or Q − S 2 Q-S^2 Q − S 2 and can otherwise remove valid solutions.
Einstein–Maxwell–dilaton black branes
In the preceding section we saw that homogeneous spatial curvature can support anisotropic scaling even in vacuum. We now extend the construction by adding a U ( 1 ) U(1) U ( 1 ) gauge field and a dilaton. The electromagnetic field changes the spatial Einstein equations and, consequently, the conditions on the scaling exponents. Our main interest is to understand which homogeneous geometries admit electric and magnetic fields without losing the cohomogeneity-one form of the metric.
Action and equations of motion
We take the Einstein–Maxwell–dilaton (EMD) action
I : = 1 2 ∫ d n + 2 x − g [ R − 2 Λ − ( ∂ ϕ ) 2 − 1 2 e λ ϕ F μ ν F μ ν ] + I b d y , F : = d A . (34) I := \frac{1}{2}\int d^{n+2}x\sqrt{-g}\left[R-2\Lambda-(\partial\phi)^2-\frac{1}{2}e^{\lambda\phi}F_{\mu\nu}F^{\mu\nu}\right]+I_{\mathrm{bdy}},\qquad F := dA. \tag*{(34)} I := 2 1 ∫ d n + 2 x − g [ R − 2Λ − ( ∂ ϕ ) 2 − 2 1 e λ ϕ F μν F μν ] + I bdy , F := d A . (34)
Here λ \lambda λ is a constant coupling and ϕ \phi ϕ is a real scalar. We use the convention where 8 π G N = 1 8\pi G_N = 1 8 π G N = 1 . The field equations, in trace-reversed form, are
R μ ν = 2 Λ n g μ ν + ∂ μ ϕ ∂ ν ϕ + e λ ϕ ( F μ ρ F ν ρ − 1 2 n g μ ν F 2 ) , (35) R_{\mu\nu} = \frac{2\Lambda}{n}g_{\mu\nu}+\partial_\mu\phi\partial_\nu\phi+e^{\lambda\phi}\left(F_{\mu\rho}F_\nu{}^\rho-\frac{1}{2n}g_{\mu\nu}F^2\right), \tag*{(35)} R μν = n 2Λ g μν + ∂ μ ϕ ∂ ν ϕ + e λ ϕ ( F μ ρ F ν ρ − 2 n 1 g μν F 2 ) , (35)
∇ μ ( e λ ϕ F μ ν ) = 0 , d F = 0 , (36) \nabla_\mu\left(e^{\lambda\phi}F^{\mu\nu}\right)=0,\qquad dF=0, \tag*{(36)} ∇ μ ( e λ ϕ F μν ) = 0 , d F = 0 , (36)
□ ϕ = λ 4 e λ ϕ F 2 . (37) \Box\phi= \frac{\lambda}{4}e^{\lambda\phi}F^2. \tag*{(37)} □ ϕ = 4 λ e λ ϕ F 2 . (37)
We continue to use the metric (6) , and set
ϕ ( r ) = ϕ 0 + σ log r . (38) \phi(r) = \phi_0 + \sigma\log r. \tag*{(38)} ϕ ( r ) = ϕ 0 + σ log r . (38)
Purely electric solutions
We first take a homogeneous electric field,
A = A t ( r ) d t , F = E ( r ) d r ∧ d t . (39) A = A_t(r)\mathrm{d}t,\qquad F = E(r)\mathrm{d}r \wedge\mathrm{d}t. \tag*{(39)} A = A t ( r ) d t , F = E ( r ) d r ∧ d t . (39)
The Maxwell equation is readily integrated:
d d r ( r S + 1 − z + λ σ E ) = 0 , E ( r ) = q e − λ ϕ 0 r z − S − 1 − λ σ . (40) \frac{\mathrm{d}}{\mathrm{d}r}\left(r^{S+1-z+\lambda\sigma}E\right) = 0,\qquad E(r) = q e^{-\lambda\phi_0}r^{z-S-1-\lambda\sigma}. \tag*{(40)} d r d ( r S + 1 − z + λσ E ) = 0 , E ( r ) = q e − λ ϕ 0 r z − S − 1 − λσ . (40)
Defining E : = q 2 e − λ ϕ 0 \mathcal{E} := q^2 e^{-\lambda\phi_0} E := q 2 e − λ ϕ 0 and p : = 2 S + λ σ p := 2S + \lambda\sigma p := 2 S + λσ , the electric energy in a local orthonormal frame is proportional to E r − p \mathcal{E}r^{-p} E r − p . Adding the temporal and radial Einstein equations gives
z S = Q + σ 2 . (41) zS = Q + \sigma^2. \tag*{(41)} z S = Q + σ 2 . (41)
Therefore the dilaton increases the effective time exponent relative to its neutral value. For S ≠ 0 S \ne0 S = 0 we have z = ( Q + σ 2 ) / S z = (Q + \sigma^2)/S z = ( Q + σ 2 ) / S . The independent equations can be written as
( D + 2 z ) ( D + Δ ) f = − 4 Λ n + 2 ( n − 1 ) n E r − p , (42) (\mathcal{D} + 2z)(\mathcal{D} + \Delta)f = -\frac{4\Lambda}{n} + \frac{2(n-1)}{n}\mathcal{E}r^{-p}, \tag*{(42)} ( D + 2 z ) ( D + Δ ) f = − n 4Λ + n 2 ( n − 1 ) E r − p , (42)
α i ( D + Δ ) f = R i ^ i ^ r − 2 κ − 2 Λ n − E n r − p , (43) \alpha_i(\mathcal{D} + \Delta)f = R_{\hat{i}\hat{i}}r^{-2\kappa} - \frac{2\Lambda}{n} - \frac{\mathcal{E}}{n}r^{-p}, \tag*{(43)} α i ( D + Δ ) f = R i ^ i ^ r − 2 κ − n 2Λ − n E r − p , (43)
σ ( D + Δ ) f = − λ E 2 r − p . (44) \sigma(\mathcal{D} + \Delta)f = -\frac{\lambda\mathcal{E}}{2}r^{-p}. \tag*{(44)} σ ( D + Δ ) f = − 2 λ E r − p . (44)
As in the neutral case, the temporal equation determines the possible radial powers, whereas the spatial and scalar equations constrain their coefficients.
Nonzero cosmological constant. For a running dilaton, σ ≠ 0 \sigma\ne0 σ = 0 , and Λ ≠ 0 \Lambda\ne0 Λ = 0 , the scalar equation and the constant term of the temporal equation require a constant electric source. Hence,
p = 0 ⟺ λ σ = − 2 S . (45) p = 0 \Longleftrightarrow\lambda\sigma= -2S. \tag*{(45)} p = 0 ⟺ λσ = − 2 S . (45)
For a curved homogeneous spatial metric, the common scaling ansatz further gives κ = 0 \kappa= 0 κ = 0 . The blackening factor then takes the form
f ( r ) = f 0 [ 1 − ( r h r ) Δ ] , f 0 = − 2 Λ + ( n − 1 ) E n z Δ . (46) f(r) = f_0\left[1 - \left(\frac{r_h}{r}\right)^\Delta\right],\qquad f_0 = \frac{-2\Lambda+ (n-1)\mathcal{E}}{n z\Delta}. \tag*{(46)} f ( r ) = f 0 [ 1 − ( r r h ) Δ ] , f 0 = n z Δ − 2Λ + ( n − 1 ) E . (46)
The scalar and spatial equations imply
z Δ f 0 = − 2 Λ + ( n − 1 ) E n , (47) z\Delta f_0 = \frac{-2\Lambda+ (n-1)\mathcal{E}}{n}, \tag*{(47)} z Δ f 0 = n − 2Λ + ( n − 1 ) E , (47)
σ Δ f 0 = − λ E 2 , (48) \sigma\Delta f_0 = -\frac{\lambda\mathcal{E}}{2}, \tag*{(48)} σ Δ f 0 = − 2 λ E , (48)
R i ^ i ^ = α i Δ f 0 + 2 Λ + E n . (49) R_{\hat{i}\hat{i}} = \alpha_i\Delta f_0 + \frac{2\Lambda+ \mathcal{E}}{n}. \tag*{(49)} R i ^ i ^ = α i Δ f 0 + n 2Λ + E . (49)
Combining the first two relations with (41) and (45) yields
E = 2 Λ σ 2 n Q + σ 2 , f 0 = − 2 Λ S 2 ( n Q + σ 2 ) ( Q + σ 2 + S 2 ) . (50) \mathcal{E} = \frac{2\Lambda\sigma^2}{nQ + \sigma^2},\qquad f_0 = -\frac{2\Lambda S^2}{(nQ + \sigma^2)(Q + \sigma^2 + S^2)}. \tag*{(50)} E = n Q + σ 2 2Λ σ 2 , f 0 = − ( n Q + σ 2 ) ( Q + σ 2 + S 2 ) 2Λ S 2 . (50)
For Λ < 0 \Lambda< 0 Λ < 0 , these amplitudes are positive whenever S > 0 S > 0 S > 0 and σ ≠ 0 \sigma\ne0 σ = 0 . The electric amplitude is fixed by the scaling exponents, λ \lambda λ , and Λ \Lambda Λ and supports the asymptotic geometry. It does not introduce an independent charge term in the blackening function.
Vanishing cosmological constant and nonconstant source powers. For Λ = 0 \Lambda= 0 Λ = 0 , an electric source need not be constant. If p ≠ 2 z , Δ p \ne2z, \Delta p = 2 z , Δ , the temporal equation admits a particular solution proportional to r − p r^{-p} r − p , in addition to the homogeneous modes r − 2 z r^{-2z} r − 2 z and r − Δ r^{-\Delta} r − Δ . The remaining equations constrain which of these modes survive. The cases where p = 2 z p = 2z p = 2 z or p = Δ p = \Delta p = Δ give logarithmic particular solutions, and the case of a repeated radial root (z = S z = S z = S ) must be integrated separately. We give a unified rule for these possibilities in Appendix § 8 . The spatial and scalar equations must also be satisfied.
Homogeneous magnetic fields
Next, we replace the electric field by a homogeneous magnetic field. A magnetic two-form can be expressed in terms of the invariant coframe as
F m = 1 2 B i j θ i ∧ θ j , B i j = − B j i , (51) F_{\mathrm{m}} = \frac{1}{2}B_{ij}\theta^{i} \wedge\theta^{j}, \qquad B_{ij} = -B_{ji}, \tag*{(51)} F m = 2 1 B ij θ i ∧ θ j , B ij = − B j i , (51)
where B i j B_{ij} B ij are constants. Unlike a Cartesian magnetic field, the invariant two-form need not be closed for a non-Abelian spatial group. Thus the Bianchi identity and Maxwell equations impose restrictions on the allowed orientations.
A homogeneous magnetic field is not necessarily a magnetic field with constant Cartesian components. Instead, it has constant components in the invariant frame of the spatial geometry. This distinction matters because the invariant one-forms can depend on the spatial coordinates. The Bianchi identity d F = 0 \mathrm{d}F = 0 d F = 0 requires the magnetic flux to have no local sources, while the Maxwell equation requires it to be supported without an external current. Together, these requirements select the allowed magnetic planes. In five bulk dimensions, writing M 1 : = B 23 M_1 := B_{23} M 1 := B 23 , M 2 : = B 31 M_2 := B_{31} M 2 := B 31 , and M 3 : = B 12 M_3 := B_{12} M 3 := B 12 , they take the form
∑ i < j B i j d ( θ i ∧ θ j ) = 0 , (52) \sum_{i<j} B_{ij}\mathrm{d}(\theta^{i} \wedge\theta^{j}) = 0, \tag*{(52)} i < j ∑ B ij d ( θ i ∧ θ j ) = 0 , (52)
∑ i b i 2 M i r 2 α i d θ i = 0. (53) \sum_i b_i^2 M_i r^{2\alpha_i}\mathrm{d}\theta^i = 0. \tag*{(53)} i ∑ b i 2 M i r 2 α i d θ i = 0. (53)
For distinct spatial exponents, terms with different radial powers cannot cancel, so a nonzero M i M_i M i requires d θ i = 0 \mathrm{d}\theta^i = 0 d θ i = 0 . Equal exponents may allow cancellations. These conditions are best viewed as the magnetic analogue of the scaling-compatibility equations: they ensure that the proposed magnetic background respects the same spatial homogeneity as the metric.
For a single magnetic plane ( i j ) (ij) ( ij ) , let χ ℓ \chi_\ell χ ℓ be one in the magnetic plane and zero in the remaining directions. Define
B : = B i j 2 e λ ϕ 0 b i 2 b j 2 , m B : = λ σ − 2 ( α i + α j ) , v ℓ : = χ ℓ − 1 n . (54) \mathcal{B} := \frac{B_{ij}^{2}e^{\lambda\phi_0}}{b_i^2b_j^2}, \qquad m_{\mathcal{B}} := \lambda\sigma- 2(\alpha_i + \alpha_j), \qquad v_\ell:= \chi_\ell- \frac{1}{n}. \tag*{(54)} B := b i 2 b j 2 B ij 2 e λ ϕ 0 , m B := λσ − 2 ( α i + α j ) , v ℓ := χ ℓ − n 1 . (54)
We assume the magnetic stress tensor is diagonal in the same frame as the spatial Ricci tensor. The field equations are
( D + 2 z ) ( D + Δ ) f = − 4 Λ n + 2 B n r m B , (55) (\mathcal{D} + 2z)(\mathcal{D} + \Delta)f = -\frac{4\Lambda}{n} + \frac{2\mathcal{B}}{n}r^{m_{\mathcal{B}}}, \tag*{(55)} ( D + 2 z ) ( D + Δ ) f = − n 4Λ + n 2 B r m B , (55)
α ℓ ( D + Δ ) f = R ℓ ℓ ^ ℓ ^ r − 2 κ − 2 Λ n − v ℓ B r m B , (56) \alpha_\ell(\mathcal{D} + \Delta)f = R_{\ell\hat{\ell}\hat{\ell}}r^{-2\kappa} - \frac{2\Lambda}{n} - v_\ell\mathcal{B}r^{m_{\mathcal{B}}}, \tag*{(56)} α ℓ ( D + Δ ) f = R ℓ ℓ ^ ℓ ^ r − 2 κ − n 2Λ − v ℓ B r m B , (56)
σ ( D + Δ ) f = λ B 2 r m B . (57) \sigma(\mathcal{D} + \Delta)f = \frac{\lambda\mathcal{B}}{2}r^{m_{\mathcal{B}}}. \tag*{(57)} σ ( D + Δ ) f = 2 λ B r m B . (57)
The t t + r r tt + rr tt + r r equation again gives z S = Q + σ 2 zS = Q + \sigma^2 z S = Q + σ 2 .
For Λ ≠ 0 \Lambda\ne0 Λ = 0 , λ σ ≠ 0 \lambda\sigma\ne0 λσ = 0 , and a curved spatial geometry, the equations require
m B = 0 , κ = 0 , λ σ = 2 ( α i + α j ) . (58) m_{\mathcal{B}} = 0, \qquad\kappa= 0, \qquad\lambda\sigma= 2(\alpha_i + \alpha_j). \tag*{(58)} m B = 0 , κ = 0 , λσ = 2 ( α i + α j ) . (58)
Writing Δ f 0 \Delta f_0 Δ f 0 for the common left-hand side, the remaining equations reduce to
f 0 = − 2 Λ + B z Δ n , R ℓ ^ ℓ ^ = α ℓ Δ f 0 + 2 Λ n + v ℓ B , σ = λ B 2 Δ f 0 , z S = Q + σ 2 . (59) \begin{aligned}
f_0 &= \frac{-2\Lambda+\mathcal{B}}{z\Delta n},\\
R_{\hat{\ell}\hat{\ell}} &= \alpha_\ell\Delta f_0+\frac{2\Lambda}{n}+v_\ell\mathcal{B},\\
\sigma&= \frac{\lambda\mathcal{B}}{2\Delta f_0},\qquad zS=Q+\sigma^2. \tag*{(59)}
\end{aligned} f 0 R ℓ ^ ℓ ^ σ = z Δ n − 2Λ + B , = α ℓ Δ f 0 + n 2Λ + v ℓ B , = 2Δ f 0 λ B , z S = Q + σ 2 . (59)
For Δ > 0 \Delta>0 Δ > 0 the blackening factor is
f ( r ) = f 0 [ 1 − ( r h r ) Δ ] . (60) f(r)=f_0\left[1-\left(\frac{r_h}{r}\right)^\Delta\right]. \tag*{(60)} f ( r ) = f 0 [ 1 − ( r r h ) Δ ] . (60)
The important difference from the electric case is the directional magnetic stress v ℓ B v_\ell\mathcal{B} v ℓ B . It can support anisotropic radial scaling even when the spatial group is Abelian, and it singles out the magnetic plane. For a constant dilaton and a nonzero isolated magnetic field, the scalar equation requires λ = 0 \lambda=0 λ = 0 . The cases with Λ = 0 \Lambda=0 Λ = 0 , with a magnetic field whose radial power is nonzero (m B ≠ 0 m_B\ne0 m B = 0 ), and with a source term sharing the radial power of a homogeneous mode are summarized in Appendix § 8 .
Dyonic configurations
We now include both an electric field and a single homogeneous magnetic plane. The field strength is
F = q e − λ ϕ 0 r z − S − 1 − λ σ d r ∧ d t + B i j θ i ∧ θ j . (61) F=qe^{-\lambda\phi_0}r^{z-S-1-\lambda\sigma}\mathrm{d}r\wedge\mathrm{d}t+B_{ij}\theta^i\wedge\theta^j. \tag*{(61)} F = q e − λ ϕ 0 r z − S − 1 − λσ d r ∧ d t + B ij θ i ∧ θ j . (61)
The magnetic constraints (3.19)–(3.20) are unchanged, while the electric Maxwell equation remains (40) . Using E \mathcal{E} E , B \mathcal{B} B , p p p and m B m_B m B defined above, we obtain
( D + 2 z ) ( D + Δ ) f = − 4 Λ n + 2 ( n − 1 ) E n r − p + 2 B n r m B . (62) (\mathcal{D}+2z)(\mathcal{D}+\Delta)f=-\frac{4\Lambda}{n}+\frac{2(n-1)\mathcal{E}}{n}r^{-p}+\frac{2\mathcal{B}}{n}r^{m_B}. \tag*{(62)} ( D + 2 z ) ( D + Δ ) f = − n 4Λ + n 2 ( n − 1 ) E r − p + n 2 B r m B . (62)
α ℓ ( D + Δ ) f = R ℓ ^ ℓ ^ r − 2 κ − 2 Λ n − E n r − p − v ℓ B r m B . (63) \alpha_\ell(\mathcal{D}+\Delta)f=R_{\hat{\ell}\hat{\ell}}r^{-2\kappa}-\frac{2\Lambda}{n}-\frac{\mathcal{E}}{n}r^{-p}-v_\ell\mathcal{B}r^{m_B}. \tag*{(63)} α ℓ ( D + Δ ) f = R ℓ ^ ℓ ^ r − 2 κ − n 2Λ − n E r − p − v ℓ B r m B . (63)
σ ( D + Δ ) f = λ 2 ( B r m B − E r − p ) . (64) \sigma(\mathcal{D}+\Delta)f=\frac{\lambda}{2}\left(\mathcal{B}r^{m_B}-\mathcal{E}r^{-p}\right). \tag*{(64)} σ ( D + Δ ) f = 2 λ ( B r m B − E r − p ) . (64)
z S = Q + σ 2 . (65) zS=Q+\sigma^2. \tag*{(65)} z S = Q + σ 2 . (65)
The electric and magnetic contributions generally scale differently with radius. In the particularly simple branch
p = m B = 0 ⟹ λ σ = − 2 S = 2 ( α i + α j ) , (66) p=m_B=0\quad\Longrightarrow\quad\lambda\sigma=-2S=2(\alpha_i+\alpha_j), \tag*{(66)} p = m B = 0 ⟹ λσ = − 2 S = 2 ( α i + α j ) , (66)
both local electromagnetic energy densities have the same radial weight as the cosmological-constant term. Thus the fields support the asymptotic anisotropic geometry rather than acting as independent charge deformations of an otherwise fixed background. This is why their amplitudes are determined algebraically together with the scaling exponents. The running dilaton compensates for the radial dilution of the electric and magnetic fields.
If Λ ≠ 0 \Lambda\ne0 Λ = 0 and λ σ ≠ 0 \lambda\sigma\ne0 λσ = 0 , a nonzero spatial curvature then requires κ = 0 \kappa=0 κ = 0 . Taking ( D + Δ ) f = Δ f 0 (\mathcal{D}+\Delta)f=\Delta f_0 ( D + Δ ) f = Δ f 0 , the algebraic equations become
f 0 = − 2 Λ + ( n − 1 ) E + B z Δ n , (67) f_0 = \frac{-2\Lambda+ (n-1)\mathcal{E}+\mathcal{B}}{z\Delta n}, \tag*{(67)} f 0 = z Δ n − 2Λ + ( n − 1 ) E + B , (67)
R ℓ = α ℓ Δ f 0 + 2 Λ + E n + ν ℓ B , (68) R_\ell= \alpha_\ell\Delta f_0 + \frac{2\Lambda+\mathcal{E}}{n}+\nu_\ell\mathcal{B}, \tag*{(68)} R ℓ = α ℓ Δ f 0 + n 2Λ + E + ν ℓ B , (68)
σ = λ 2 Δ f 0 ( B − E ) , z S = Q + σ 2 , (69) \sigma= \frac{\lambda}{2\Delta f_0}(\mathcal{B}-\mathcal{E}), \qquad zS = Q+\sigma^2, \tag*{(69)} σ = 2Δ f 0 λ ( B − E ) , z S = Q + σ 2 , (69)
f ( r ) = f 0 [ 1 − ( r h r ) Δ ] . (70) f(r)=f_0\left[1-\left(\frac{r_h}{r}\right)^\Delta\right]. \tag*{(70)} f ( r ) = f 0 [ 1 − ( r r h ) Δ ] . (70)
For a specified spatial Lie algebra and magnetic orientation, these equations constrain the exponents and matter amplitudes. The five-dimensional Nil solution and higher-dimensional examples are given in § 4 .
The Λ = 0 \Lambda=0 Λ = 0 dyonic branches require a separate analysis, analogous to the neutral case in § 2.4 .
Remark on charged branches. The constant source configurations above use the electromagnetic fields to maintain the scaling background, so their amplitudes are not independent thermodynamic charges. One may instead seek solutions closer to a Reissner–Nordström black brane, in which a constant dilaton and balanced electric and magnetic fields introduce an additional, faster-decaying charge term in the blackening function, together with extremal horizons. Charged black holes of this type in Lifshitz spacetimes were constructed in [18 ] , and on Thurston horizon geometries, including balanced dyonic configurations and their thermodynamics, in [45 , 46 , 47 ] . Extending the present anisotropic scaling backgrounds with such independently variable charge deformations is left for future work.
Examples of homogeneous black branes
We now apply the reduced field equations to five-dimensional Bianchi geometries and higher-dimensional Ricci nilsolitons, beginning with the neutral solutions and then their matter-supported extensions.
Neutral solutions in five dimensions
We take n = 3 n=3 n = 3 , with Λ < 0 \Lambda<0 Λ < 0 , and use
d s 2 = − r 2 z f ( r ) d t 2 + d r 2 r 2 f ( r ) + ∑ i = 1 3 b i 2 r 2 α i ( θ i ) 2 . (71) ds^2=-r^{2z}f(r)dt^2+\frac{dr^2}{r^2f(r)}+\sum_{i=1}^{3}b_i^2r^{2\alpha_i}(\theta^i)^2. \tag*{(71)} d s 2 = − r 2 z f ( r ) d t 2 + r 2 f ( r ) d r 2 + i = 1 ∑ 3 b i 2 r 2 α i ( θ i ) 2 . (71)
The spatial one-forms and corresponding exponents are summarized below. Except where indicated, f ( r ) = f 0 [ 1 − ( r h / r ) Δ ] f(r)=f_0[1-(r_h/r)^\Delta] f ( r ) = f 0 [ 1 − ( r h / r ) Δ ] with f 0 = − 2 Λ / ( 3 z Δ ) f_0=-2\Lambda/(3z\Delta) f 0 = − 2Λ/ ( 3 z Δ ) and Δ = z + S \Delta=z+S Δ = z + S .
Bianchi II (Nil). The Nil coframe and scaling exponents are
θ 1 = d x , θ 2 = d y , θ 3 = d w − x d y , ( α 1 , α 2 , α 3 ) = ( 1 , 1 , 2 ) , (72) \theta^1=dx,\qquad\theta^2=dy,\qquad\theta^3=dw-x\,dy,\qquad(\alpha_1,\alpha_2,\alpha_3)=(1,1,2), \tag*{(72)} θ 1 = d x , θ 2 = d y , θ 3 = d w − x d y , ( α 1 , α 2 , α 3 ) = ( 1 , 1 , 2 ) , (72)
so that S = 4 S=4 S = 4 and Q = 6 Q=6 Q = 6 . We take Λ < 0 \Lambda<0 Λ < 0 , b 1 = b 2 = 1 b_1=b_2=1 b 1 = b 2 = 1 , and b 3 2 = − 4 Λ / 9 b_3^2=-4\Lambda/9 b 3 2 = − 4Λ/9 .
The exponents satisfy the derivation condition (9) with κ = 0 \kappa=0 κ = 0 and the mixed condition (14) , and (19) gives z = Q / S = 3 / 2 z=Q/S=3/2 z = Q / S = 3/2 . The reference metric h b = ( θ 1 ) 2 + ( θ 2 ) 2 + b 3 2 ( θ 3 ) 2 h_b=(\theta^1)^2+(\theta^2)^2+b_3^2(\theta^3)^2 h b = ( θ 1 ) 2 + ( θ 2 ) 2 + b 3 2 ( θ 3 ) 2 has orthonormal Ricci eigenvalues R 1 = R 2 = − b 3 2 / 2 R_1=R_2=-b_3^2/2 R 1 = R 2 = − b 3 2 /2 and R 3 = b 3 2 / 2 R_3=b_3^2/2 R 3 = b 3 2 /2 , so the constraint (25) fixes b 3 2 = − 4 Λ / 9 b_3^2=-4\Lambda/9 b 3 2 = − 4Λ/9 . The blackening function follows from (27) with Δ = z + S = 11 / 2 \Delta=z+S=11/2 Δ = z + S = 11/2 .
d s 2 = − r 3 f ( r ) d t 2 + d r 2 r 2 f ( r ) + r 2 [ ( θ 1 ) 2 + ( θ 2 ) 2 ] − 4 Λ 9 r 4 ( θ 3 ) 2 , (73) \mathrm{d}s^2=-r^3 f(r)\mathrm{d}t^2+\frac{\mathrm{d}r^2}{r^2f(r)}+r^2\left[(\theta^1)^2+(\theta^2)^2\right]-\frac{4\Lambda}{9}r^4(\theta^3)^2, \tag*{(73)} d s 2 = − r 3 f ( r ) d t 2 + r 2 f ( r ) d r 2 + r 2 [ ( θ 1 ) 2 + ( θ 2 ) 2 ] − 9 4Λ r 4 ( θ 3 ) 2 , (73)
f ( r ) = − 8 Λ 99 [ 1 − ( r h r ) 11 / 2 ] . f(r)=-\frac{8\Lambda}{99}\left[1-\left(\frac{r_h}{r}\right)^{11/2}\right]. f ( r ) = − 99 8Λ [ 1 − ( r r h ) 11/2 ] .
This is the Nil-horizon black brane already considered in [40 , 39 ] . The positive eigenvalue of Ric ( h b ) \operatorname{Ric}(h_b) Ric ( h b ) in the third direction is compensated by the larger radial scaling exponent α 3 \alpha_3 α 3 .
Solv (Bianchi VI− 1 _{-1} − 1 ). The Solv coframe and scaling exponents are
θ 1 = e w d x , θ 2 = e − w d y , θ 3 = d w , ( α 1 , α 2 , α 3 ) = ( 1 , 1 , 0 ) , (74) \theta^1=e^w\mathrm{d}x,\qquad\theta^2=e^{-w}\mathrm{d}y,\qquad\theta^3=\mathrm{d}w,\qquad(\alpha_1,\alpha_2,\alpha_3)=(1,1,0), \tag*{(74)} θ 1 = e w d x , θ 2 = e − w d y , θ 3 = d w , ( α 1 , α 2 , α 3 ) = ( 1 , 1 , 0 ) , (74)
so that S = Q = 2 S=Q=2 S = Q = 2 . We take Λ < 0 \Lambda<0 Λ < 0 and b 1 = b 2 = 1 b_1=b_2=1 b 1 = b 2 = 1 .
As in the Nil example, (9) and (14) hold with κ = 0 \kappa=0 κ = 0 , and (19) gives z = 1 z=1 z = 1 . The reference metric has orthonormal Ricci eigenvalues R 1 = R 2 = 0 R_1=R_2=0 R 1 = R 2 = 0 and R 3 = − 2 / b 3 2 R_3=-2/b_3^2 R 3 = − 2/ b 3 2 , so (25) fixes b 3 2 = − 3 / Λ b_3^2=-3/\Lambda b 3 2 = − 3/Λ . The blackening function is (27) with Δ = z + S = 3 \Delta=z+S=3 Δ = z + S = 3 .
d s 2 = − r 2 f ( r ) d t 2 + d r 2 r 2 f ( r ) + r 2 [ ( θ 1 ) 2 + ( θ 2 ) 2 ] − 3 Λ ( θ 3 ) 2 , (75) \mathrm{d}s^2=-r^2f(r)\mathrm{d}t^2+\frac{\mathrm{d}r^2}{r^2f(r)}+r^2\left[(\theta^1)^2+(\theta^2)^2\right]-\frac{3}{\Lambda}(\theta^3)^2, \tag*{(75)} d s 2 = − r 2 f ( r ) d t 2 + r 2 f ( r ) d r 2 + r 2 [ ( θ 1 ) 2 + ( θ 2 ) 2 ] − Λ 3 ( θ 3 ) 2 , (75)
f ( r ) = − 2 Λ 9 [ 1 − ( r h r ) 3 ] . f(r)=-\frac{2\Lambda}{9}\left[1-\left(\frac{r_h}{r}\right)^3\right]. f ( r ) = − 9 2Λ [ 1 − ( r r h ) 3 ] .
Here R 1 = R 2 = 0 R_1=R_2=0 R 1 = R 2 = 0 and R 3 = 2 Λ / 3 R_3=2\Lambda/3 R 3 = 2Λ/3 . This familiar solution is the special Bianchi VI− 1 _{-1} − 1 case of the family below [39 , 40 ] .
Bianchi VIh _h h . In our previous work [44 ] we considered the one-parameter family of coframes
θ 1 = e w d x , θ 2 = e h w d y , θ 3 = d w , (76) \theta^1=e^w\mathrm{d}x,\qquad\theta^2=e^{hw}\mathrm{d}y,\qquad\theta^3=\mathrm{d}w, \tag*{(76)} θ 1 = e w d x , θ 2 = e h w d y , θ 3 = d w , (76)
with exponents
( α 1 , α 2 , α 3 ) = ( − h , 1 , 0 ) , z = 1 + h 2 1 − h , Δ = 2 ( h 2 − h + 1 ) 1 − h , (77) (\alpha_1,\alpha_2,\alpha_3)=(-h,1,0),\qquad z=\frac{1+h^2}{1-h},\qquad\Delta=\frac{2(h^2-h+1)}{1-h}, \tag*{(77)} ( α 1 , α 2 , α 3 ) = ( − h , 1 , 0 ) , z = 1 − h 1 + h 2 , Δ = 1 − h 2 ( h 2 − h + 1 ) , (77)
so that S = 1 − h S=1-h S = 1 − h and Q = 1 + h 2 Q=1+h^2 Q = 1 + h 2 . We take h < 1 h<1 h < 1 , Λ < 0 \Lambda<0 Λ < 0 , and b 1 = b 2 = 1 b_1=b_2=1 b 1 = b 2 = 1 .
The mixed condition (14) reduces to α 1 = − h α 2 \alpha_1=-h\alpha_2 α 1 = − h α 2 for α 3 = 0 \alpha_3=0 α 3 = 0 and is satisfied by these exponents. The constraint (25) applied to h b = ( θ 1 ) 2 + ( θ 2 ) 2 + b 3 2 ( θ 3 ) 2 h_b=(\theta^1)^2+(\theta^2)^2+b_3^2(\theta^3)^2 h b = ( θ 1 ) 2 + ( θ 2 ) 2 + b 3 2 ( θ 3 ) 2 fixes b 3 2 = − 3 ( 1 + h 2 ) / ( 2 Λ ) b_3^2=-3(1+h^2)/(2\Lambda) b 3 2 = − 3 ( 1 + h 2 ) / ( 2Λ ) and is then satisfied for every h < 1 h<1 h < 1 . The blackening function is (27) with f 0 = − Λ ( 1 − h ) 2 / [ 3 ( 1 + h 2 ) ( h 2 − h + 1 ) ] f_0=-\Lambda(1-h)^2/[3(1+h^2)(h^2-h+1)] f 0 = − Λ ( 1 − h ) 2 / [ 3 ( 1 + h 2 ) ( h 2 − h + 1 )] .
d s 2 = − r 2 z f ( r ) d t 2 + d r 2 r 2 f ( r ) + r − 2 h ( θ 1 ) 2 + r 2 ( θ 2 ) 2 − 3 ( 1 + h 2 ) 2 Λ ( θ 3 ) 2 , (78) \mathrm{d}s^2=-r^{2z}f(r)\mathrm{d}t^2+\frac{\mathrm{d}r^2}{r^2f(r)}+r^{-2h}(\theta^1)^2+r^2(\theta^2)^2-\frac{3(1+h^2)}{2\Lambda}(\theta^3)^2, \tag*{(78)} d s 2 = − r 2 z f ( r ) d t 2 + r 2 f ( r ) d r 2 + r − 2 h ( θ 1 ) 2 + r 2 ( θ 2 ) 2 − 2Λ 3 ( 1 + h 2 ) ( θ 3 ) 2 , (78)
f ( r ) = − Λ ( 1 − h ) 2 3 ( 1 + h 2 ) ( h 2 − h + 1 ) [ 1 − ( r h r ) Δ ] . f(r)=-\frac{\Lambda(1-h)^2}{3(1+h^2)(h^2-h+1)}\left[1-\left(\frac{r_h}{r}\right)^\Delta\right]. f ( r ) = − 3 ( 1 + h 2 ) ( h 2 − h + 1 ) Λ ( 1 − h ) 2 [ 1 − ( r r h ) Δ ] .
The limiting value h = − 1 h=-1 h = − 1 gives (75) . The h = 0 h=0 h = 0 case has Q = S 2 Q=S^2 Q = S 2 , where the radial root is repeated, and must be interpreted with the corresponding equations; it still admits the displayed black brane. The point h = 1 h=1 h = 1 is a separate Bianchi V branch, for which the radial ansatz must be modified. These cases and the associated Ricci-flat, hyperscaling-violating family were derived in Ref. [44 ] .
Electric five-dimensional solutions
We next turn to EMD gravity. For concreteness, we use the constant source electric branch (45) . In each case the dilaton takes the form ϕ = ϕ 0 + σ log r \phi=\phi_0+\sigma\log r ϕ = ϕ 0 + σ log r and the gauge field is given by (40) . The field strength therefore shares the spatial homogeneity of the metric.
Bianchi I. For a flat spatial metric we take θ i = d x i \theta^i=dx^i θ i = d x i with ( α 1 , α 2 , α 3 ) = ( 1 , 1 , 1 ) (\alpha_1,\alpha_2,\alpha_3)=(1,1,1) ( α 1 , α 2 , α 3 ) = ( 1 , 1 , 1 ) , so that S = Q = 3 S=Q=3 S = Q = 3 . We take Λ < 0 \Lambda<0 Λ < 0 , λ ≠ 0 \lambda\ne0 λ = 0 , and b i = 1 b_i=1 b i = 1 .
The constant source branch (45) gives λ σ = − 6 \lambda\sigma=-6 λσ = − 6 , and (41) gives z = 1 + 12 / λ 2 z=1+12/\lambda^2 z = 1 + 12/ λ 2 . The amplitudes (50) give E = − 8 Λ / ( λ 2 + 4 ) \mathcal{E}=-8\Lambda/(\lambda^2+4) E = − 8Λ/ ( λ 2 + 4 ) , and (46) gives Δ = 4 + 12 / λ 2 \Delta=4+12/\lambda^2 Δ = 4 + 12/ λ 2 . Since the spatial geometry is flat, R ^ i ^ i ^ = 0 \widehat{R}_{\hat{i}\hat{i}}=0 R i ^ i ^ = 0 , the spatial equations (48) are satisfied identically by this data.
d s 2 = − r 2 z f ( r ) d t 2 + d r 2 r 2 f ( r ) + r 2 ( d x 2 + d y 2 + d w 2 ) , f ( r ) = − Λ λ 4 6 ( λ 2 + 4 ) ( λ 2 + 3 ) [ 1 − ( r h r ) Δ ] , ϕ = ϕ 0 − 6 λ log r , F = q e − λ ϕ 0 r 3 + 12 / λ 2 d r ∧ d t , q 2 e − λ ϕ 0 = E = − 8 Λ λ 2 + 4 . (79) \begin{aligned}
ds^2&=-r^{2z}f(r)dt^2+\frac{dr^2}{r^2f(r)}+r^2\left(dx^2+dy^2+dw^2\right),\\
f(r)&=-\frac{\Lambda\lambda^4}{6(\lambda^2+4)(\lambda^2+3)}
\left[1-\left(\frac{r_h}{r}\right)^\Delta\right],
\qquad
\phi=\phi_0-\frac{6}{\lambda}\log r,\\
F&=q e^{-\lambda\phi_0}r^{3+12/\lambda^2}dr\wedge dt,
\qquad
q^2e^{-\lambda\phi_0}=\mathcal{E}=-\frac{8\Lambda}{\lambda^2+4}. \tag*{(79)}
\end{aligned} d s 2 f ( r ) F = − r 2 z f ( r ) d t 2 + r 2 f ( r ) d r 2 + r 2 ( d x 2 + d y 2 + d w 2 ) , = − 6 ( λ 2 + 4 ) ( λ 2 + 3 ) Λ λ 4 [ 1 − ( r r h ) Δ ] , ϕ = ϕ 0 − λ 6 log r , = q e − λ ϕ 0 r 3 + 12/ λ 2 d r ∧ d t , q 2 e − λ ϕ 0 = E = − λ 2 + 4 8Λ . (79)
This is a homogeneous Lifshitz-type solution supported by a running dilaton and an electric field. It is useful as a check on the normalization of our equations.
Bianchi II. We use the Nil coframe above with ( α 1 , α 2 , α 3 ) = ( 1 , 1 , 2 ) (\alpha_1,\alpha_2,\alpha_3)=(1,1,2) ( α 1 , α 2 , α 3 ) = ( 1 , 1 , 2 ) , so that S = 4 S=4 S = 4 and Q = 6 Q=6 Q = 6 . We take Λ < 0 \Lambda<0 Λ < 0 , λ ≠ 0 \lambda\ne0 λ = 0 , and b 1 = b 2 = 1 b_1=b_2=1 b 1 = b 2 = 1 .
The constant source branch (45) gives λ σ = − 8 \lambda\sigma=-8 λσ = − 8 , and (41) gives z = 3 2 + 16 λ 2 z=\frac{3}{2}+\frac{16}{\lambda^2} z = 2 3 + λ 2 16 . The amplitudes (50) give E = − 64 Λ 32 + 9 λ 2 \mathcal{E}=-\frac{64\Lambda}{32+9\lambda^2} E = − 32 + 9 λ 2 64Λ and Δ = 11 2 + 16 λ 2 \Delta=\frac{11}{2}+\frac{16}{\lambda^2} Δ = 2 11 + λ 2 16 . The Nil Ricci data R 1 = R 2 = − b 3 2 / 2 R_1=R_2=-b_3^2/2 R 1 = R 2 = − b 3 2 /2 and R 3 = b 3 2 / 2 R_3=b_3^2/2 R 3 = b 3 2 /2 in the spatial equations (48) fix b 3 2 = λ 2 E / 16 b_3^2=\lambda^2\mathcal{E}/16 b 3 2 = λ 2 E /16 .
d s 2 = − r 2 z f ( r ) d t 2 + d r 2 r 2 f ( r ) + r 2 [ ( θ 1 ) 2 + ( θ 2 ) 2 ] + b 3 2 r 4 ( θ 3 ) 2 , f ( r ) = − 8 Λ λ 4 ( 9 λ 2 + 32 ) ( 11 λ 2 + 32 ) [ 1 − ( r h r ) Δ ] , ϕ = ϕ 0 − 8 λ log r , F = q e − λ ϕ 0 r 9 / 2 + 16 / λ 2 d r ∧ d t , q 2 e − λ ϕ 0 = E = − 64 Λ 9 λ 2 + 32 . (80) \begin{aligned}
ds^2&=-r^{2z}f(r)dt^2+\frac{dr^2}{r^2f(r)}
+r^2\left[(\theta^1)^2+(\theta^2)^2\right]+b_3^2r^4(\theta^3)^2,\\
f(r)&=-\frac{8\Lambda\lambda^4}{(9\lambda^2+32)(11\lambda^2+32)}
\left[1-\left(\frac{r_h}{r}\right)^\Delta\right],
\qquad
\phi=\phi_0-\frac{8}{\lambda}\log r,\\
F&=q e^{-\lambda\phi_0}r^{9/2+16/\lambda^2}dr\wedge dt,
\qquad
q^2e^{-\lambda\phi_0}=\mathcal{E}=-\frac{64\Lambda}{9\lambda^2+32}. \tag*{(80)}
\end{aligned} d s 2 f ( r ) F = − r 2 z f ( r ) d t 2 + r 2 f ( r ) d r 2 + r 2 [ ( θ 1 ) 2 + ( θ 2 ) 2 ] + b 3 2 r 4 ( θ 3 ) 2 , = − ( 9 λ 2 + 32 ) ( 11 λ 2 + 32 ) 8Λ λ 4 [ 1 − ( r r h ) Δ ] , ϕ = ϕ 0 − λ 8 log r , = q e − λ ϕ 0 r 9/2 + 16/ λ 2 d r ∧ d t , q 2 e − λ ϕ 0 = E = − 9 λ 2 + 32 64Λ . (80)
with b 3 2 = − 4 Λ λ 2 / ( 9 λ 2 + 32 ) b_3^2=-4\Lambda\lambda^2/(9\lambda^2+32) b 3 2 = − 4Λ λ 2 / ( 9 λ 2 + 32 ) . The scalar and electric field modify z z z while preserving the underlying Nil spatial geometry. In the limit λ → ∞ \lambda\to\infty λ → ∞ at fixed Λ \Lambda Λ the neutral solution (73) is recovered.
Bianchi VIh _h h . We may extend the neutral family (78) with the same coframe (76) and exponents ( − h , 1 , 0 ) (-h,1,0) ( − h , 1 , 0 ) , so that S = 1 − h S=1-h S = 1 − h and Q = 1 + h 2 Q=1+h^2 Q = 1 + h 2 . We take h < 1 h<1 h < 1 , Λ < 0 \Lambda<0 Λ < 0 , λ ≠ 0 \lambda\ne0 λ = 0 , and b 1 = b 2 = 1 b_1=b_2=1 b 1 = b 2 = 1 .
The constant source branch (45) gives λ σ = − 2 ( 1 − h ) \lambda\sigma=-2(1-h) λσ = − 2 ( 1 − h ) , and (41) gives z = 1 + h 2 1 − h + 4 ( 1 − h ) λ 2 z=\frac{1+h^2}{1-h}+\frac{4(1-h)}{\lambda^2} z = 1 − h 1 + h 2 + λ 2 4 ( 1 − h ) . The amplitudes (50) give
E = − 8 Λ ( 1 − h ) 2 4 ( 1 − h ) 2 + 3 λ 2 ( 1 + h 2 ) , b 3 2 = − 4 ( 1 − h ) 2 + 3 λ 2 ( 1 + h 2 ) 2 Λ λ 2 . (81) \mathcal{E}=-\frac{8\Lambda(1-h)^2}{4(1-h)^2+3\lambda^2(1+h^2)},
\qquad
b_3^2=-\frac{4(1-h)^2+3\lambda^2(1+h^2)}{2\Lambda\lambda^2}. \tag*{(81)} E = − 4 ( 1 − h ) 2 + 3 λ 2 ( 1 + h 2 ) 8Λ ( 1 − h ) 2 , b 3 2 = − 2Λ λ 2 4 ( 1 − h ) 2 + 3 λ 2 ( 1 + h 2 ) . (81)
where b 3 2 b_3^2 b 3 2 follows from the spatial equations (48) . The blackening function is (46) with Δ = z + 1 − h \Delta=z+1-h Δ = z + 1 − h .
d s 2 = − r 2 z f ( r ) d t 2 + d r 2 r 2 f ( r ) + r − 2 h ( θ 1 ) 2 + r 2 ( θ 2 ) 2 + b 3 2 ( θ 3 ) 2 , (82) \mathrm{d}s^2=-r^{2z}f(r)\mathrm{d}t^2+\frac{\mathrm{d}r^2}{r^2f(r)}+r^{-2h}(\theta^1)^2+r^2(\theta^2)^2+b_3^2(\theta^3)^2, \tag*{(82)} d s 2 = − r 2 z f ( r ) d t 2 + r 2 f ( r ) d r 2 + r − 2 h ( θ 1 ) 2 + r 2 ( θ 2 ) 2 + b 3 2 ( θ 3 ) 2 , (82)
f ( r ) = f 0 [ 1 − ( r h r ) Δ ] , ϕ = ϕ 0 − 2 ( 1 − h ) λ log r , f(r)=f_0\left[1-\left(\frac{r_h}{r}\right)^\Delta\right],\qquad\phi=\phi_0-\frac{2(1-h)}{\lambda}\log r, f ( r ) = f 0 [ 1 − ( r r h ) Δ ] , ϕ = ϕ 0 − λ 2 ( 1 − h ) log r ,
F = q e − λ ϕ 0 r Δ − 1 d r ∧ d t , q 2 e − λ ϕ 0 = E , F=q e^{-\lambda\phi_0}r^{\Delta-1}\mathrm{d}r\wedge\mathrm{d}t,\qquad q^2e^{-\lambda\phi_0}=\mathcal{E}, F = q e − λ ϕ 0 r Δ − 1 d r ∧ d t , q 2 e − λ ϕ 0 = E ,
f 0 = − 2 Λ ( 1 − h ) 2 λ 4 [ 3 λ 2 ( 1 + h 2 ) + 4 ( 1 − h ) 2 ] [ λ 2 ( 2 − 2 h + 2 h 2 ) + 4 ( 1 − h ) 2 ] . f_0=-\frac{2\Lambda(1-h)^2\lambda^4}{\left[3\lambda^2(1+h^2)+4(1-h)^2\right]\left[\lambda^2(2-2h+2h^2)+4(1-h)^2\right]}. f 0 = − [ 3 λ 2 ( 1 + h 2 ) + 4 ( 1 − h ) 2 ] [ λ 2 ( 2 − 2 h + 2 h 2 ) + 4 ( 1 − h ) 2 ] 2Λ ( 1 − h ) 2 λ 4 .
The radial electric field modifies the time exponent and the curvature scale, while the continuous Bianchi parameter h h h is retained. In the limit of vanishing matter this is not necessarily a continuous limit at fixed λ \lambda λ , because the relation λ σ = − 2 S \lambda\sigma=-2S λσ = − 2 S selects a specific charged branch.
Magnetic and dyonic five-dimensional solutions
Magnetic Bianchi I. We consider a five-dimensional spacetime with flat spatial sections, θ i = d x i \theta^i=\mathrm{d}x^i θ i = d x i , and a magnetic field in the (12)-plane,
F = B d x ∧ d y . (83) F=B\mathrm{d}x\wedge\mathrm{d}y. \tag*{(83)} F = B d x ∧ d y . (83)
Although the spatial geometry is flat, the magnetic field introduces anisotropy between the magnetic plane and the remaining direction. We take λ ≠ 0 \lambda\ne0 λ = 0 and Λ < 0 \Lambda<0 Λ < 0 ; the radial diffeomorphism (13) fixes the exponent scale to a = 1 a=1 a = 1 .
The magnetic Bianchi identity and Maxwell equation (52) –(53) hold trivially since d θ i = 0 \mathrm{d}\theta^i=0 d θ i = 0 . The conditions m B = 0 m_B=0 m B = 0 and κ = 0 \kappa=0 κ = 0 give λ σ = 4 \lambda\sigma=4 λσ = 4 , and the spatial equations give α 3 = z \alpha_3=z α 3 = z . The constraint z S = Q + σ 2 zS=Q+\sigma^2 z S = Q + σ 2 then yields z = 1 + 8 / λ 2 z=1+8/\lambda^2 z = 1 + 8/ λ 2 . Thus the exponents are ( α 1 , α 2 , α 3 ) = ( 1 , 1 , z ) (\alpha_1,\alpha_2,\alpha_3)=(1,1,z) ( α 1 , α 2 , α 3 ) = ( 1 , 1 , z ) and Δ = 2 ( 1 + z ) \Delta=2(1+z) Δ = 2 ( 1 + z ) . The algebraic system (59) with R ^ i ^ i ^ = 0 \widehat{R}_{\hat{i}\hat{i}}=0 R i ^ i ^ = 0 then gives B = − 2 Λ ( z − 1 ) / ( 1 + 2 z ) = − 16 Λ / ( 3 λ 2 + 16 ) B=-2\Lambda(z-1)/(1+2z)=-16\Lambda/(3\lambda^2+16) B = − 2Λ ( z − 1 ) / ( 1 + 2 z ) = − 16Λ/ ( 3 λ 2 + 16 ) and f 0 = − 2 Λ / [ ( 1 + 2 z ) Δ ] f_0=-2\Lambda/[(1+2z)\Delta] f 0 = − 2Λ/ [( 1 + 2 z ) Δ ] .
d s 2 = − r 2 z f ( r ) d t 2 + d r 2 r 2 f ( r ) + r 2 ( d x 2 + d y 2 ) + r 2 z d w 2 , (84) \mathrm{d}s^2=-r^{2z}f(r)\mathrm{d}t^2+\frac{\mathrm{d}r^2}{r^2f(r)}+r^2(\mathrm{d}x^2+\mathrm{d}y^2)+r^{2z}\mathrm{d}w^2, \tag*{(84)} d s 2 = − r 2 z f ( r ) d t 2 + r 2 f ( r ) d r 2 + r 2 ( d x 2 + d y 2 ) + r 2 z d w 2 , (84)
f ( r ) = − 2 Λ ( 1 + 2 z ) Δ [ 1 − ( r h r ) Δ ] , ϕ = ϕ 0 + 4 λ log r , f(r)=-\frac{2\Lambda}{(1+2z)\Delta}\left[1-\left(\frac{r_h}{r}\right)^\Delta\right],\qquad\phi=\phi_0+\frac{4}{\lambda}\log r, f ( r ) = − ( 1 + 2 z ) Δ 2Λ [ 1 − ( r r h ) Δ ] , ϕ = ϕ 0 + λ 4 log r ,
F = B d x ∧ d y , B 2 e λ ϕ 0 = B = − 16 Λ 3 λ 2 + 16 . F=B\mathrm{d}x\wedge\mathrm{d}y,\qquad B^2e^{\lambda\phi_0}=\mathcal{B}=-\frac{16\Lambda}{3\lambda^2+16}. F = B d x ∧ d y , B 2 e λ ϕ 0 = B = − 3 λ 2 + 16 16Λ .
For Λ < 0 \Lambda<0 Λ < 0 , the magnetic amplitude and blackening function are positive outside the horizon. The magnetic field supports different radial scaling in the magnetic plane and the remaining spatial direction, despite the vanishing intrinsic spatial curvature. The solution approaches the neutral planar AdS black brane as λ 2 → ∞ \lambda^2\to\infty λ 2 → ∞ : the exponents become isotropic, and the scalar gradient and electromagnetic stress vanish.
Magnetic Solv. Consider the Solv coframe above with ( α 1 , α 2 , α 3 ) = ( 1 , 1 , 0 ) (\alpha_1,\alpha_2,\alpha_3)=(1,1,0) ( α 1 , α 2 , α 3 ) = ( 1 , 1 , 0 ) and a magnetic field in the (12)-plane,
F = B θ 1 ∧ θ 2 . (85) F=B\theta^1\wedge\theta^2. \tag*{(85)} F = B θ 1 ∧ θ 2 . (85)
We take Λ < 0 \Lambda<0 Λ < 0 , λ ≠ 0 \lambda\ne0 λ = 0 , and b 1 = b 2 = 1 b_1=b_2=1 b 1 = b 2 = 1 .
Since d ( θ 1 ∧ θ 2 ) = 0 \mathrm{d}(\theta^1\wedge\theta^2)=0 d ( θ 1 ∧ θ 2 ) = 0 and d θ 3 = 0 \mathrm{d}\theta^3=0 d θ 3 = 0 , both the Bianchi identity (52) and the Maxwell equation (53) are satisfied. The conditions m B = 0 m_B=0 m B = 0 and κ = 0 \kappa=0 κ = 0 give λ σ = 4 \lambda\sigma=4 λσ = 4 , and z S = Q + σ 2 zS=Q+\sigma^2 z S = Q + σ 2 gives z = 1 + 8 / λ 2 z = 1 + 8/\lambda^2 z = 1 + 8/ λ 2 ; the radial diffeomorphism (13) fixes the exponent scale to a = 1 a = 1 a = 1 . The algebraic system (59) , with the Solv Ricci data R 1 = R 2 = 0 R_1 = R_2 = 0 R 1 = R 2 = 0 and R 3 = − 2 / b 3 2 R_3 = -2/b_3^2 R 3 = − 2/ b 3 2 , gives
σ = 4 λ , z = 1 + 8 λ 2 , B = − 2 Λ ( z − 1 ) 1 + 2 z , b 3 2 = − 1 + 2 z z Λ . (86) \sigma= \frac{4}{\lambda}, \qquad z = 1 + \frac{8}{\lambda^2}, \qquad\mathcal{B} = -\frac{2\Lambda(z-1)}{1+2z}, \qquad b_3^2 = -\frac{1+2z}{z\Lambda}. \tag*{(86)} σ = λ 4 , z = 1 + λ 2 8 , B = − 1 + 2 z 2Λ ( z − 1 ) , b 3 2 = − z Λ 1 + 2 z . (86)
d s 2 = − r 2 z f ( r ) d t 2 + d r 2 r 2 f ( r ) + r 2 [ ( θ 1 ) 2 + ( θ 2 ) 2 ] + b 3 2 ( θ 3 ) 2 , (87) ds^2 = -r^{2z}f(r)dt^2 + \frac{dr^2}{r^2f(r)} + r^2\left[(\theta^1)^2 + (\theta^2)^2\right] + b_3^2(\theta^3)^2, \tag*{(87)} d s 2 = − r 2 z f ( r ) d t 2 + r 2 f ( r ) d r 2 + r 2 [ ( θ 1 ) 2 + ( θ 2 ) 2 ] + b 3 2 ( θ 3 ) 2 , (87)
f ( r ) = − 2 Λ ( 1 + 2 z ) ( z + 2 ) [ 1 − ( r h r ) z + 2 ] , ϕ = ϕ 0 + 4 λ log r , f(r) = -\frac{2\Lambda}{(1+2z)(z+2)}\left[1-\left(\frac{r_h}{r}\right)^{z+2}\right], \qquad\phi= \phi_0 + \frac{4}{\lambda}\log r, f ( r ) = − ( 1 + 2 z ) ( z + 2 ) 2Λ [ 1 − ( r r h ) z + 2 ] , ϕ = ϕ 0 + λ 4 log r ,
F = B θ 1 ∧ θ 2 , B 2 e λ ϕ 0 = B . F = \mathcal{B}\theta^1\wedge\theta^2, \qquad\mathcal{B}^2e^{\lambda\phi_0}=\mathcal{B}. F = B θ 1 ∧ θ 2 , B 2 e λ ϕ 0 = B .
For Λ < 0 \Lambda< 0 Λ < 0 , the solution has a simple horizon and positive magnetic amplitude. The different signs of the electric and magnetic contributions to the scalar equation explain the opposite sign of σ \sigma σ compared with the electric examples.
Dyonic Nil. The Nil group also admits a configuration with both electric and magnetic fields. We use the Nil coframe with ( α 1 , α 2 , α 3 ) = ( 1 , − 3 / 4 , 1 / 4 ) (\alpha_1,\alpha_2,\alpha_3)=(1,-3/4,1/4) ( α 1 , α 2 , α 3 ) = ( 1 , − 3/4 , 1/4 ) , so that S = 1 / 2 S=1/2 S = 1/2 and Q = 13 / 8 Q=13/8 Q = 13/8 , and choose the magnetic plane ( 23 ) (23) ( 23 ) . The field strength is
F = E ( r ) d r ∧ d t + B 23 θ 2 ∧ θ 3 . (88) F = E(r)dr\wedge dt+B_{23}\theta^2\wedge\theta^3. \tag*{(88)} F = E ( r ) d r ∧ d t + B 23 θ 2 ∧ θ 3 . (88)
We take λ ≠ 0 \lambda\ne0 λ = 0 and Λ < 0 \Lambda<0 Λ < 0 ; the radial diffeomorphism (13) fixes the exponent scale to a = 1 a=1 a = 1 .
The magnetic form is closed, and its spatial dual is proportional to the closed form θ 1 \theta^1 θ 1 , so the constraints (52) –(53) hold. The constant source conditions (66) give λ σ = − 1 \lambda\sigma=-1 λσ = − 1 , and (65) gives z = 13 4 + 2 λ 2 z=\frac{13}{4}+\frac{2}{\lambda^2} z = 4 13 + λ 2 2 and Δ = 15 4 + 2 λ 2 \Delta=\frac{15}{4}+\frac{2}{\lambda^2} Δ = 4 15 + λ 2 2 . The algebraic equations (67) –(69) , with the Nil Ricci data R 1 = R 2 = − b 3 2 / 2 R_1=R_2=-b_3^2/2 R 1 = R 2 = − b 3 2 /2 and R 3 = b 3 2 / 2 R_3=b_3^2/2 R 3 = b 3 2 /2 , have the solution
σ = − 1 / λ , z = 13 4 + 2 λ 2 . (89) \sigma=-1/\lambda, \qquad z=\frac{13}{4}+\frac{2}{\lambda^2}. \tag*{(89)} σ = − 1/ λ , z = 4 13 + λ 2 2 . (89)
Δ = 15 4 + 2 λ 2 , B = − 7 λ 2 9 λ 2 + 4 , (90) \Delta=\frac{15}{4}+\frac{2}{\lambda^2}, \qquad\mathcal{B}=-\frac{7\lambda^2}{9\lambda^2+4}, \tag*{(90)} Δ = 4 15 + λ 2 2 , B = − 9 λ 2 + 4 7 λ 2 , (90)
E = − Λ ( 7 λ 2 + 8 ) 9 λ 2 + 4 , b 3 2 = − 4 Λ λ 2 9 λ 2 + 4 b 1 2 b 2 2 . \mathcal{E}=-\frac{\Lambda(7\lambda^2+8)}{9\lambda^2+4}, \qquad b_3^2=-\frac{4\Lambda\lambda^2}{9\lambda^2+4}b_1^2b_2^2. E = − 9 λ 2 + 4 Λ ( 7 λ 2 + 8 ) , b 3 2 = − 9 λ 2 + 4 4Λ λ 2 b 1 2 b 2 2 .
d s 2 = − r 2 z f ( r ) d t 2 + d r 2 r 2 f ( r ) + b 1 2 r 2 ( θ 1 ) 2 + b 2 2 r − 3 / 2 ( θ 2 ) 2 + b 3 2 r 1 / 2 ( θ 3 ) 2 , (91) ds^2=-r^{2z}f(r)dt^2+\frac{dr^2}{r^2f(r)}+b_1^2r^2(\theta^1)^2+b_2^2r^{-3/2}(\theta^2)^2+b_3^2r^{1/2}(\theta^3)^2, \tag*{(91)} d s 2 = − r 2 z f ( r ) d t 2 + r 2 f ( r ) d r 2 + b 1 2 r 2 ( θ 1 ) 2 + b 2 2 r − 3/2 ( θ 2 ) 2 + b 3 2 r 1/2 ( θ 3 ) 2 , (91)
f ( r ) = − 4 Λ λ 2 ( 9 λ 2 + 4 ) Δ [ 1 − ( r h r ) Δ ] , ϕ = ϕ 0 − 1 λ log r , f(r)=-\frac{4\Lambda\lambda^2}{(9\lambda^2+4)\Delta}\left[1-\left(\frac{r_h}{r}\right)^\Delta\right], \qquad\phi=\phi_0-\frac{1}{\lambda}\log r, f ( r ) = − ( 9 λ 2 + 4 ) Δ 4Λ λ 2 [ 1 − ( r r h ) Δ ] , ϕ = ϕ 0 − λ 1 log r ,
F = q e − λ ϕ 0 r z + S − 1 d r ∧ d t + B 23 θ 2 ∧ θ 3 , F=qe^{-\lambda\phi_0}r^{z+S-1}dr\wedge dt+B_{23}\theta^2\wedge\theta^3, F = q e − λ ϕ 0 r z + S − 1 d r ∧ d t + B 23 θ 2 ∧ θ 3 ,
where
q 2 e − λ ϕ 0 = − Λ ( 7 λ 2 + 8 ) 9 λ 2 + 4 , B 23 2 e λ ϕ 0 b 1 2 b 2 2 = − 7 λ 2 9 λ 2 + 4 . (92) q^2e^{-\lambda\phi_0}=-\frac{\Lambda(7\lambda^2+8)}{9\lambda^2+4}, \qquad\frac{B_{23}^2e^{\lambda\phi_0}}{b_1^2b_2^2}=-\frac{7\lambda^2}{9\lambda^2+4}. \tag*{(92)} q 2 e − λ ϕ 0 = − 9 λ 2 + 4 Λ ( 7 λ 2 + 8 ) , b 1 2 b 2 2 B 23 2 e λ ϕ 0 = − 9 λ 2 + 4 7 λ 2 . (92)
For Λ < 0 \Lambda<0 Λ < 0 , both electric and magnetic amplitudes are positive and the blackening function has a single horizon. The running dilaton supports a homogeneous anisotropic dyonic configuration in which both sources enter the same constant part of the reduced Einstein equations.
Higher-dimensional families
Dyonic Bianchi I in general dimensions. We now consider a flat spatial geometry in D = n + 2 D=n+2 D = n + 2 dimensions, with n ≥ 3 n\ge3 n ≥ 3 , and introduce electric and magnetic fields,
F = E ( r ) d r ∧ d t + B d x 1 ∧ d x 2 . (93) F=E(r)\mathrm{d}r\wedge\mathrm{d}t+B\,\mathrm{d}x^{1}\wedge\mathrm{d}x^{2}. \tag*{(93)} F = E ( r ) d r ∧ d t + B d x 1 ∧ d x 2 . (93)
We take λ ≠ 0 \lambda\ne0 λ = 0 and Λ < 0 \Lambda<0 Λ < 0 ; the radial diffeomorphism (13) fixes the exponent scale to a = 1 a=1 a = 1 .
For the constant source branch, the conditions p = m B = 0 p=m_{B}=0 p = m B = 0 in (66) require λ σ = − 2 S = 2 ( α 1 + α 2 ) \lambda\sigma=-2S=2(\alpha_{1}+\alpha_{2}) λσ = − 2 S = 2 ( α 1 + α 2 ) . Since the intrinsic spatial Ricci tensor vanishes, the spatial Einstein equations require equal scaling exponents within the magnetic plane and within its orthogonal complement. The algebraic equations (3.34)–(3.36) then give the exponents and amplitudes
α 1 = α 2 = − 1 , α A = 4 n − 2 , A = 3 , … , n , (94) \alpha_{1}=\alpha_{2}=-1,\qquad\alpha_{A}=\frac{4}{n-2},\qquad A=3,\ldots,n, \tag*{(94)} α 1 = α 2 = − 1 , α A = n − 2 4 , A = 3 , … , n , (94)
S = 2 , σ = − 4 λ , (95) S=2,\qquad\sigma=-\frac{4}{\lambda}, \tag*{(95)} S = 2 , σ = − λ 4 , (95)
z = 1 + 8 n − 2 + 8 λ 2 , Δ = z + 2. (96) z=1+\frac{8}{n-2}+\frac{8}{\lambda^{2}},\qquad\Delta=z+2. \tag*{(96)} z = 1 + n − 2 8 + λ 2 8 , Δ = z + 2. (96)
The remaining algebraic equations give
B = ( 1 + 4 n − 2 ) Δ f 0 , (97) \mathcal{B}=\left(1+\frac{4}{n-2}\right)\Delta f_{0}, \tag*{(97)} B = ( 1 + n − 2 4 ) Δ f 0 , (97)
E = ( z − 4 n − 2 ) Δ f 0 . (98) \mathcal{E}=\left(z-\frac{4}{n-2}\right)\Delta f_{0}. \tag*{(98)} E = ( z − n − 2 4 ) Δ f 0 . (98)
d s 2 = − r 2 z f ( r ) d t 2 + d r 2 r 2 f ( r ) + r − 2 [ ( d x 1 ) 2 + ( d x 2 ) 2 ] + r 8 n − 2 ∑ A = 3 n ( d x A ) 2 , (99) \mathrm{d}s^{2}=-r^{2z}f(r)\mathrm{d}t^{2}+\frac{\mathrm{d}r^{2}}{r^{2}f(r)}+r^{-2}\left[(\mathrm{d}x^{1})^{2}+(\mathrm{d}x^{2})^{2}\right]+r^{\frac{8}{n-2}}\sum_{A=3}^{n}(\mathrm{d}x^{A})^{2}, \tag*{(99)} d s 2 = − r 2 z f ( r ) d t 2 + r 2 f ( r ) d r 2 + r − 2 [ ( d x 1 ) 2 + ( d x 2 ) 2 ] + r n − 2 8 A = 3 ∑ n ( d x A ) 2 , (99)
f ( r ) = − 2 Λ ( z + 3 ) ( z + 2 ) [ 1 − ( r h r ) Δ ] , ϕ = ϕ 0 − 4 λ log r , f(r)=-\frac{2\Lambda}{(z+3)(z+2)}\left[1-\left(\frac{r_{h}}{r}\right)^{\Delta}\right],\qquad\phi=\phi_{0}-\frac{4}{\lambda}\log r, f ( r ) = − ( z + 3 ) ( z + 2 ) 2Λ [ 1 − ( r r h ) Δ ] , ϕ = ϕ 0 − λ 4 log r ,
F = q e − λ ϕ 0 r z + 1 d r ∧ d t + B d x 1 ∧ d x 2 , q 2 e − λ ϕ 0 = E , B 2 e λ ϕ 0 = B . F=q\mathrm{e}^{-\lambda\phi_{0}}r^{z+1}\mathrm{d}r\wedge\mathrm{d}t+B\mathrm{d}x^{1}\wedge\mathrm{d}x^{2},\qquad q^{2}\mathrm{e}^{-\lambda\phi_{0}}=\mathcal{E},\qquad B^{2}\mathrm{e}^{\lambda\phi_{0}}=\mathcal{B}. F = q e − λ ϕ 0 r z + 1 d r ∧ d t + B d x 1 ∧ d x 2 , q 2 e − λ ϕ 0 = E , B 2 e λ ϕ 0 = B .
For Λ < 0 \Lambda<0 Λ < 0 , both E \mathcal{E} E and B \mathcal{B} B are positive, and f ( r ) f(r) f ( r ) has a simple horizon at r = r h r=r_{h} r = r h . In contrast to the neutral and purely electric flat solutions, the magnetic directions have negative scaling exponents, while the remaining directions have positive exponents. Both electromagnetic sources contribute to the constant part of the Einstein equations, rather than introducing an independent charge-dependent radial mode in the blackening function.
Neutral Ricci nilsolitons. The neutral construction of § 2 extends to arbitrary dimensions, as first considered in [42 ] . The input is an n n n -dimensional nilpotent Lie algebra with a left-invariant metric h b h_{b} h b satisfying
Ric ( h b ) = − c 1 + D , c > 0 , D ( X i ) = d i X i , (100) \operatorname{Ric}(h_{b})=-c\,1+D,\qquad c>0,\qquad D(X_{i})=d_{i}X_{i}, \tag*{(100)} Ric ( h b ) = − c 1 + D , c > 0 , D ( X i ) = d i X i , (100)
where D D D is a symmetric derivation. We take Λ < 0 \Lambda<0 Λ < 0 and a free time exponent z > 0 z>0 z > 0 , which plays the role of the overall exponent scale a a a and can likewise be set to unity by (13) .
Rescale h b h_{b} h b so that c = 2 ( − Λ ) / n c=2(-\Lambda)/n c = 2 ( − Λ ) / n . The Ricci constraint (25) is then solved by the soliton data
α i = z c d i , S = z c Tr D , f 0 = c z ( z + S ) . (101) \alpha_{i}=\frac{z}{c}d_{i},\qquad S=\frac{z}{c}\operatorname{Tr}D,\qquad f_{0}=\frac{c}{z(z+S)}. \tag*{(101)} α i = c z d i , S = c z Tr D , f 0 = z ( z + S ) c . (101)
For a nonflat Ricci nilsoliton, the identity Tr D 2 = c Tr D \operatorname{Tr}D^{2}=c\operatorname{Tr}D Tr D 2 = c Tr D gives Q = z S Q=zS Q = z S , so (19) holds.
d s 2 = − r 2 z f ( r ) d t 2 + d r 2 r 2 f ( r ) + ∑ i b i 2 r 2 z d i / c ( θ i ) 2 , f ( r ) = f 0 [ 1 − ( r h r ) z + S ] (102) \mathrm{d}s^2=-r^{2z}f(r)\mathrm{d}t^2+\frac{\mathrm{d}r^2}{r^2f(r)}+\sum_i b_i^2r^{2z d_i/c}(\theta^i)^2,\qquad f(r)=f_0\left[1-\left(\frac{r_h}{r}\right)^{z+S}\right] \tag*{(102)} d s 2 = − r 2 z f ( r ) d t 2 + r 2 f ( r ) d r 2 + i ∑ b i 2 r 2 z d i / c ( θ i ) 2 , f ( r ) = f 0 [ 1 − ( r r h ) z + S ] (102)
The metric solves the neutral Einstein equations, subject to the mixed constraints. This is the Ricci nilsoliton black hole construction of Ref. [42 ] .
Dyonic extension of a Ricci nilsoliton. We add two flat spatial directions to an m m m -dimensional nonflat Ricci nilsoliton. Let
Ric ( h 0 ) = − c 1 + D , D ( X i ) = d i X i , ∑ i d i 2 = c ∑ i d i . (103) \operatorname{Ric}(h_0)=-c\mathbf{1}+\mathrm{D},\qquad\mathrm{D}(X_i)=d_iX_i,\qquad\sum_i d_i^2=c\sum_i d_i. \tag*{(103)} Ric ( h 0 ) = − c 1 + D , D ( X i ) = d i X i , i ∑ d i 2 = c i ∑ d i . (103)
We use g = n m ⊕ R 2 \mathfrak{g}=\mathfrak{n}_m\oplus\mathbb{R}^2 g = n m ⊕ R 2 with coordinates ( x , y ) (x,y) ( x , y ) along the additional flat directions. The total spatial dimension is n = m + 2 n=m+2 n = m + 2 and the spacetime dimension is D = m + 4 D=m+4 D = m + 4 . We take λ ≠ 0 \lambda\ne0 λ = 0 and Λ < 0 \Lambda<0 Λ < 0 , and a magnetic field in the additional flat plane, F = E ( r ) d r ∧ d t + B d x ∧ d y F=E(r)\mathrm{d}r\wedge\mathrm{d}t+B\mathrm{d}x\wedge\mathrm{d}y F = E ( r ) d r ∧ d t + B d x ∧ d y .
The radial diffeomorphism (13) fixes the exponent scale to a = 1 a=1 a = 1 . The dyonic constant source conditions (66) are met by the soliton data below, and the algebraic equations (67) –(69) give
α i = d i ( i = 1 , … , m ) , α x = α y = − 1 4 ∑ i d i (104) \alpha_i=d_i\quad(i=1,\ldots,m),\qquad\alpha_x=\alpha_y=-\frac{1}{4}\sum_i d_i \tag*{(104)} α i = d i ( i = 1 , … , m ) , α x = α y = − 4 1 i ∑ d i (104)
S = 1 2 ∑ i d i , σ = − 1 λ ∑ i d i (105) S=\frac{1}{2}\sum_i d_i,\qquad\sigma=-\frac{1}{\lambda}\sum_i d_i \tag*{(105)} S = 2 1 i ∑ d i , σ = − λ 1 i ∑ d i (105)
z = 2 c + ( 1 4 + 2 λ 2 ) ∑ i d i , Δ = z + S (106) z=2c+\left(\frac{1}{4}+\frac{2}{\lambda^2}\right)\sum_i d_i,\qquad\Delta=z+S \tag*{(106)} z = 2 c + ( 4 1 + λ 2 2 ) i ∑ d i , Δ = z + S (106)
L 2 = − n c − 2 λ 2 ∑ i d i 2 Λ , B = c + 1 4 ∑ i d i L 2 (107) L^2=\frac{-nc-\frac{2}{\lambda^2}\sum_i d_i}{2\Lambda},\qquad\mathcal{B}=\frac{c+\frac{1}{4}\sum_i d_i}{L^2} \tag*{(107)} L 2 = 2Λ − n c − λ 2 2 ∑ i d i , B = L 2 c + 4 1 ∑ i d i (107)
E = c + ( 1 4 + 2 λ 2 ) ∑ i d i L 2 \mathcal{E}=\frac{c+\left(\frac{1}{4}+\frac{2}{\lambda^2}\right)\sum_i d_i}{L^2} E = L 2 c + ( 4 1 + λ 2 2 ) ∑ i d i
We take a representative nilsoliton metric h 0 = ∑ i b ˉ i 2 ( θ i ) 2 h_0=\sum_i\bar b_i^2(\theta^i)^2 h 0 = ∑ i b ˉ i 2 ( θ i ) 2 and obtain
d s 2 = − r 2 z f ( r ) d t 2 + d r 2 r 2 f ( r ) + L 2 ∑ i = 1 m b ˉ i 2 r 2 d i ( θ i ) 2 + L 2 r − 1 2 ∑ i d i ( d x 2 + d y 2 ) (108) \mathrm{d}s^2=-r^{2z}f(r)\mathrm{d}t^2+\frac{\mathrm{d}r^2}{r^2f(r)}+L^2\sum_{i=1}^{m}\bar b_i^2r^{2d_i}(\theta^i)^2+L^2r^{-\frac{1}{2}\sum_i d_i}(\mathrm{d}x^2+\mathrm{d}y^2) \tag*{(108)} d s 2 = − r 2 z f ( r ) d t 2 + r 2 f ( r ) d r 2 + L 2 i = 1 ∑ m b ˉ i 2 r 2 d i ( θ i ) 2 + L 2 r − 2 1 ∑ i d i ( d x 2 + d y 2 ) (108)
f ( r ) = 1 L 2 Δ [ 1 − ( r h r ) Δ ] , ϕ = ϕ 0 − 1 λ ∑ i d i log r , f(r)=\frac{1}{L^2\Delta}\left[1-\left(\frac{r_h}{r}\right)^\Delta\right],\qquad\phi=\phi_0-\frac{1}{\lambda}\sum_i d_i\log r, f ( r ) = L 2 Δ 1 [ 1 − ( r r h ) Δ ] , ϕ = ϕ 0 − λ 1 i ∑ d i log r ,
F = q e − λ ϕ 0 r z + S − 1 d r ∧ d t + B d x ∧ d y , F=qe^{-\lambda\phi_0}r^{z+S-1}\mathrm{d}r\wedge\mathrm{d}t+B\mathrm{d}x\wedge\mathrm{d}y, F = q e − λ ϕ 0 r z + S − 1 d r ∧ d t + B d x ∧ d y ,
q 2 e − λ ϕ 0 = E , B 2 e λ ϕ 0 L 4 = B . q^2e^{-\lambda\phi_0}=\mathcal{E},\qquad\frac{B^2e^{\lambda\phi_0}}{L^4}=\mathcal{B}. q 2 e − λ ϕ 0 = E , L 4 B 2 e λ ϕ 0 = B .
The two matter source powers are constant, p = m B = 0 p=m_B=0 p = m B = 0 . The construction also obeys the scalar constraint and the temporal equation because ∑ i d i 2 = c ∑ i d i \sum_i d_i^2=c\sum_i d_i ∑ i d i 2 = c ∑ i d i . The radial momentum constraint follows from the divergence-free property of the soliton derivation. The magnetic form is closed, while its dual is proportional to the volume form on the nilpotent factor; hence the Maxwell equation also holds. The additional magnetic plane extends the neutral Ricci nilsoliton construction to a family of higher-dimensional dyonic solutions.
As one example, take the three-dimensional Nil soliton with c = 3 / 2 c=3/2 c = 3/2 and ( d 1 , d 2 , d 3 ) = ( 1 , 1 , 2 ) (d_1,d_2,d_3)=(1,1,2) ( d 1 , d 2 , d 3 ) = ( 1 , 1 , 2 ) , so that n = 5 n=5 n = 5 and D = 7 D=7 D = 7 . In this case,
α x = α y = − 1 , S = 2 , σ = − 4 λ , z = 4 + 8 λ 2 , L 2 = − 15 + 16 / λ 2 4 Λ (109) \alpha_x=\alpha_y=-1,\qquad S=2,\qquad\sigma=-\frac{4}{\lambda},\qquad z=4+\frac{8}{\lambda^2},\qquad L^2=-\frac{15+16/\lambda^2}{4\Lambda} \tag*{(109)} α x = α y = − 1 , S = 2 , σ = − λ 4 , z = 4 + λ 2 8 , L 2 = − 4Λ 15 + 16/ λ 2 (109)
with Δ = 6 + 8 / λ 2 \Delta= 6 + 8/\lambda^2 Δ = 6 + 8/ λ 2 , B = 5 / ( 2 L 2 ) \mathcal{B} = 5/(2L^2) B = 5/ ( 2 L 2 ) , and E = ( 5 / 2 + 8 / λ 2 ) / L 2 \mathcal{E} = (5/2 + 8/\lambda^2)/L^2 E = ( 5/2 + 8/ λ 2 ) / L 2 , so that
d s 2 = − r 2 z f ( r ) d t 2 + d r 2 r 2 f ( r ) + L 2 [ b ˉ 1 2 r 2 ( θ 1 ) 2 + b ˉ 2 2 r 2 ( θ 2 ) 2 + b ˉ 3 2 r 4 ( θ 3 ) 2 + r − 2 ( d x 2 + d y 2 ) ] (110) ds^2 = -r^{2z}f(r)dt^2 + \frac{dr^2}{r^2f(r)} + L^2\left[\bar b_1^2r^2(\theta^1)^2+\bar b_2^2r^2(\theta^2)^2+\bar b_3^2r^4(\theta^3)^2+r^{-2}(dx^2+dy^2)\right] \tag*{(110)} d s 2 = − r 2 z f ( r ) d t 2 + r 2 f ( r ) d r 2 + L 2 [ b ˉ 1 2 r 2 ( θ 1 ) 2 + b ˉ 2 2 r 2 ( θ 2 ) 2 + b ˉ 3 2 r 4 ( θ 3 ) 2 + r − 2 ( d x 2 + d y 2 ) ] (110)
f ( r ) = 1 L 2 Δ [ 1 − ( r h r ) Δ ] , ϕ = ϕ 0 − 4 λ log r , f(r)=\frac{1}{L^2\Delta}\left[1-\left(\frac{r_h}{r}\right)^\Delta\right],\qquad\phi=\phi_0-\frac{4}{\lambda}\log r, f ( r ) = L 2 Δ 1 [ 1 − ( r r h ) Δ ] , ϕ = ϕ 0 − λ 4 log r ,
F = q e − λ ϕ 0 r z + 1 d r ∧ d t + B d x ∧ d y , q 2 e − λ ϕ 0 = E , B 2 e λ ϕ 0 L 4 = B . F=qe^{-\lambda\phi_0}r^{z+1}dr\wedge dt+\mathcal{B}\,dx\wedge dy,\qquad q^2e^{-\lambda\phi_0}=\mathcal{E},\qquad\frac{\mathcal{B}^2e^{\lambda\phi_0}}{L^4}=\mathcal{B}. F = q e − λ ϕ 0 r z + 1 d r ∧ d t + B d x ∧ d y , q 2 e − λ ϕ 0 = E , L 4 B 2 e λ ϕ 0 = B .
The transverse magnetic directions shrink with increasing r r r , whereas the Nil directions grow. This illustrates how additional spatial dimensions can produce strongly anisotropic exact black brane geometries.
Thermodynamics
We next examine the physical properties of solutions with the blackening function (1) . These include the κ = 0 \kappa=0 κ = 0 , Λ < 0 \Lambda<0 Λ < 0 neutral branch of § 2 and the constant-source electric, magnetic, and dyonic solutions in § 3 .
Temperature and entropy
Let
d s 2 = − r 2 z f ( r ) d t 2 + d r 2 r 2 f ( r ) + ∑ i b i 2 r 2 α i ( θ i ) 2 (111) ds^2=-r^{2z}f(r)dt^2+\frac{dr^2}{r^2f(r)}+\sum_i b_i^2r^{2\alpha_i}(\theta^i)^2 \tag*{(111)} d s 2 = − r 2 z f ( r ) d t 2 + r 2 f ( r ) d r 2 + i ∑ b i 2 r 2 α i ( θ i ) 2 (111)
f ( r ) = f 0 − μ r Δ = f 0 [ 1 − ( r h r ) Δ ] , μ = f 0 r h Δ . f(r)=f_0-\frac{\mu}{r^\Delta}=f_0\left[1-\left(\frac{r_h}{r}\right)^\Delta\right],\qquad\mu=f_0r_h^\Delta. f ( r ) = f 0 − r Δ μ = f 0 [ 1 − ( r r h ) Δ ] , μ = f 0 r h Δ .
We take f 0 > 0 f_0>0 f 0 > 0 , z > 0 z>0 z > 0 , S > 0 S>0 S > 0 , and Δ = z + S > 0 \Delta=z+S>0 Δ = z + S > 0 . The horizon is at r = r h > 0 r=r_h>0 r = r h > 0 and the exterior region is r > r h r>r_h r > r h . The Euclidean regularity condition gives the Hawking temperature
T = r h z + 1 f ′ ( r h ) 4 π = f 0 Δ 4 π r h z . (112) T=\frac{r_h^{z+1}f'(r_h)}{4\pi}=\frac{f_0\Delta}{4\pi}r_h^z. \tag*{(112)} T = 4 π r h z + 1 f ′ ( r h ) = 4 π f 0 Δ r h z . (112)
Let Σ \Sigma Σ be a horizon spatial section. Its area is proportional to the volume of the reference left-invariant spatial metric. Define
V b : = ∫ Σ b 1 ⋯ b n θ 1 ∧ ⋯ ∧ θ n . (113) \mathcal{V}_b:=\int_\Sigma b_1\cdots b_n\,\theta^1\wedge\cdots\wedge\theta^n. \tag*{(113)} V b := ∫ Σ b 1 ⋯ b n θ 1 ∧ ⋯ ∧ θ n . (113)
If the spatial geometry has no compact quotient we instead use a fixed regulated cell and interpret the results as densities. The Bekenstein–Hawking entropy is
S B H = 2 π V b r h S . (114) S_{\mathrm{BH}}=2\pi\mathcal{V}_b r_h^S. \tag*{(114)} S BH = 2 π V b r h S . (114)
The relation between entropy and temperature is therefore
S B H = 2 π V b ( 4 π T f 0 Δ ) S / z . (115) S_{\mathrm{BH}}=2\pi\mathcal{V}_b\left(\frac{4\pi T}{f_0\Delta}\right)^{S/z}. \tag*{(115)} S BH = 2 π V b ( f 0 Δ 4 π T ) S / z . (115)
We see that the scaling exponent S / z S/z S / z controls the entropy. In the isotropic AdS case, α i = z = 1 \alpha_i=z=1 α i = z = 1 and S = n S=n S = n , so the familiar result S B H ∼ T n S_{\mathrm{BH}}\sim T^n S BH ∼ T n is recovered. For an anisotropic geometry, S S S is the sum of the individual spatial scaling exponents, rather than the number of spatial dimensions. This temperature scaling is analogous to that found in Lifshitz black holes [65 ] , and the dilatonic Lifshitz families of [66 ] exhibit the corresponding extended thermodynamics.
Euclidean on-shell action and free energy at fixed flux
The fixed-flux free energy follows from the Euclidean on-shell action. Because the geometries are generally not asymptotically locally AdS, the action must be supplemented by boundary terms appropriate to the chosen asymptotic data. The radial Hamilton–Jacobi approach to these boundary terms was developed in [67 , 68 , 54 ] and extended to nonrelativistic asymptotics in [69 ] . Here we perform an explicit background subtraction within the restricted family (111) , with all supporting sources and fluxes held fixed. This subtraction determines the relative on-shell action for variations within the fixed-source family.
Under t = − i τ t=-i\tau t = − i τ , the electric potential is continued as A τ = − i A t A_\tau=-iA_t A τ = − i A t , and the Euclidean time period is β : = 1 / T \beta:=1/T β := 1/ T . At a large radial cutoff r = R r=R r = R , let γ a b \gamma_{ab} γ ab be the induced metric, γ : = det γ a b \gamma:=\det\gamma_{ab} γ := det γ ab , and K K K the trace of the extrinsic curvature with the outward normal. We take
I E : = I b u l k E + I G H Y E + I N E + I c t , I G H Y E : = − ∫ r = R d n + 1 x γ K . (116) I_{\mathrm{E}}:=I_{\mathrm{bulk}}^{\mathrm{E}}+I_{\mathrm{GHY}}^{\mathrm{E}}+I_{\mathrm{N}}^{\mathrm{E}}+I_{\mathrm{ct}},\qquad I_{\mathrm{GHY}}^{\mathrm{E}}:=-\int_{r=R}\mathrm{d}^{n+1}x\sqrt{\gamma}\,K. \tag*{(116)} I E := I bulk E + I GHY E + I N E + I ct , I GHY E := − ∫ r = R d n + 1 x γ K . (116)
Let n μ n_\mu n μ be the outward unit normal and A E A_{\mathrm{E}} A E the Euclidean gauge potential, with field strength F E : = d A E F_{\mathrm{E}:=\mathrm{d}A_{\mathrm{E}}} F E := d A E and boundary components A a A_a A a . Fixing the radial electric flux, rather than the boundary potential, amounts to adding the Euclidean Legendre term
I N E : = − ∫ r = R d n + 1 x γ e λ ϕ n μ F E μ a A a . (117) I_{\mathrm{N}}^{\mathrm{E}}:=-\int_{r=R}\mathrm{d}^{n+1}x\sqrt{\gamma}\,\mathrm{e}^{\lambda\phi}n_\mu F_{\mathrm{E}}^{\mu a}A_a. \tag*{(117)} I N E := − ∫ r = R d n + 1 x γ e λ ϕ n μ F E μ a A a . (117)
This distinction between electric ensembles is familiar from black-hole thermodynamics [70 ] . We set I N E = 0 I_{\mathrm{N}}^{\mathrm{E}}=0 I N E = 0 in the neutral and purely magnetic cases.
Taking the trace of (35) , the expression in square brackets in the bulk action (34) simplifies on shell to
[ R − 2 Λ − ( ∂ ϕ ) 2 − 1 2 e λ ϕ F 2 ] o s = 4 Λ n − 1 n e λ ϕ F 2 . (118) \left[R-2\Lambda-(\partial\phi)^2-\frac{1}{2}\mathrm{e}^{\lambda\phi}F^2\right]_{\mathrm{os}}=\frac{4\Lambda}{n}-\frac{1}{n}\mathrm{e}^{\lambda\phi}F^2. \tag*{(118)} [ R − 2Λ − ( ∂ ϕ ) 2 − 2 1 e λ ϕ F 2 ] os = n 4Λ − n 1 e λ ϕ F 2 . (118)
We now specialize to the case where κ = 0 \kappa=0 κ = 0 and the local electric and magnetic energy densities are constant in r r r (including the cases in which one or both fields are absent). With the conventions of Sec. 3, e λ ϕ F 2 = 2 ( B − E ) \mathrm{e}^{\lambda\phi}F^2=2(\mathcal{B}-\mathcal{E}) e λ ϕ F 2 = 2 ( B − E ) after the Euclidean continuation, and the temporal Einstein equation yields
4 Λ + 2 E − 2 B n = 2 ( E − z Δ f 0 ) . (119) \frac{4\Lambda+2\mathcal{E}-2\mathcal{B}}{n}=2(\mathcal{E}-z\Delta f_0). \tag*{(119)} n 4Λ + 2 E − 2 B = 2 ( E − z Δ f 0 ) . (119)
For the ansatz (111) , the volume element is independent of f f f and the extrinsic curvature at r = R r=R r = R is
g E d n + 2 x = d τ d r r Δ − 1 d vol h b , K = R f ( R ) [ Δ + R f ′ ( R ) 2 f ( R ) ] . (120) \sqrt{g_{\mathrm{E}}}\,\mathrm{d}^{n+2}x=\mathrm{d}\tau\,\mathrm{d}r\,r^{\Delta-1}\mathrm{d}\operatorname{vol}_{h_b},\qquad K=R\sqrt{f(R)}\left[\Delta+\frac{Rf'(R)}{2f(R)}\right]. \tag*{(120)} g E d n + 2 x = d τ d r r Δ − 1 d vol h b , K = R f ( R ) [ Δ + 2 f ( R ) R f ′ ( R ) ] . (120)
The Euclidean bulk and Gibbons–Hawking terms therefore evaluate to
I b u l k E β V b = ( z f 0 − E Δ ) ( R Δ − r h Δ ) , (121) \frac{I_{\mathrm{bulk}}^{\mathrm{E}}}{\beta\mathcal{V}_b}=\left(zf_0-\frac{\mathcal{E}}{\Delta}\right)\left(R^\Delta-r_h^\Delta\right), \tag*{(121)} β V b I bulk E = ( z f 0 − Δ E ) ( R Δ − r h Δ ) , (121)
I G H Y E β V b = − ( Δ f 0 R Δ − Δ μ 2 ) , μ = f 0 r h Δ . (122) \frac{I_{\mathrm{GHY}}^{\mathrm{E}}}{\beta\mathcal{V}_b}=-\left(\Delta f_0R^\Delta-\frac{\Delta\mu}{2}\right),\qquad\mu=f_0r_h^\Delta. \tag*{(122)} β V b I GHY E = − ( Δ f 0 R Δ − 2 Δ μ ) , μ = f 0 r h Δ . (122)
For p = 0 p=0 p = 0 the horizon-regular Lorentzian potential is A t = q e − λ ϕ 0 ( r Δ − r h Δ ) / Δ A_t=qe^{-\lambda\phi_0}(r^\Delta-r_h^\Delta)/\Delta A t = q e − λ ϕ 0 ( r Δ − r h Δ ) /Δ . With E = q 2 e − λ ϕ 0 \mathcal{E}=q^2e^{-\lambda\phi_0} E = q 2 e − λ ϕ 0 , (117) gives
I N E β V b = E Δ ( R Δ − r h Δ ) . (123) \frac{I_{\mathrm{N}}^{\mathrm{E}}}{\beta\mathcal{V}_b}=\frac{\mathcal{E}}{\Delta}(R^\Delta-r_h^\Delta). \tag*{(123)} β V b I N E = Δ E ( R Δ − r h Δ ) . (123)
The E \mathcal{E} E -dependent terms in the bulk action and the electric Legendre term cancel. Hence, prior to counterterms, the regulated canonical action is
I E , r e g ( N ) β V b = 1 2 [ − 2 S f 0 R Δ + ( S − z ) μ ] . (124) \frac{I_{\mathrm{E,reg}}^{(\mathrm{N})}}{\beta\mathcal{V}_b}=\frac{1}{2}\left[-2Sf_0R^\Delta+(S-z)\mu\right]. \tag*{(124)} β V b I E , reg ( N ) = 2 1 [ − 2 S f 0 R Δ + ( S − z ) μ ] . (124)
One way to obtain the finite relative action is to subtract the μ = 0 \mu= 0 μ = 0 scaling geometry with precisely the same Λ \Lambda Λ , b i b_i b i , α i \alpha_i α i , ϕ 0 \phi_0 ϕ 0 , electric flux, and magnetic flux. The reference fixes the zero of the relative action within the family with fixed asymptotic sources. To match the proper length of the Euclidean time circle at the cutoff, its period is
β 0 = β f ( R ) f 0 = β [ 1 − μ 2 f 0 R Δ + O ( R − 2 Δ ) ] . (125) \beta_0 = \beta\sqrt{\frac{f(R)}{f_0}} = \beta\left[1 - \frac{\mu}{2f_0R^\Delta} + O\left(R^{-2\Delta}\right)\right]. \tag*{(125)} β 0 = β f 0 f ( R ) = β [ 1 − 2 f 0 R Δ μ + O ( R − 2Δ ) ] . (125)
Since the reference action is I E , 0 ( N ) = − β 0 V b S f 0 R Δ I^{(N)}_{E,0}=-\beta_0\mathcal{V}_b S f_0 R^\Delta I E , 0 ( N ) = − β 0 V b S f 0 R Δ , we obtain
lim R → ∞ I E , r e g ( N ) [ μ ] − I E , r e g ( N ) [ 0 ] β V b = 1 2 lim R → ∞ [ − 2 S f 0 R Δ ( 1 − 1 − μ f 0 R Δ ) + ( S − z ) μ ] = − z μ 2 . (126) \lim_{R\to\infty}\frac{I^{(N)}_{E,\mathrm{reg}}[\mu]-I^{(N)}_{E,\mathrm{reg}}[0]}{\beta\mathcal{V}_b}
=\frac{1}{2}\lim_{R\to\infty}\left[-2S f_0R^\Delta\left(1-\sqrt{1-\frac{\mu}{f_0R^\Delta}}\right)+(S-z)\mu\right]
=-\frac{z\mu}{2}. \tag*{(126)} R → ∞ lim β V b I E , reg ( N ) [ μ ] − I E , reg ( N ) [ 0 ] = 2 1 R → ∞ lim [ − 2 S f 0 R Δ ( 1 − 1 − f 0 R Δ μ ) + ( S − z ) μ ] = − 2 z μ . (126)
Alternatively, the on-shell divergence of this family is removed by
I c t f a m : = S f 0 ∫ r = R d n + 1 x γ . (127) I_{\mathrm{ct}}^{\mathrm{fam}}:=S\sqrt{f_0}\int_{r=R}\mathrm{d}^{n+1}x\sqrt{\gamma}. \tag*{(127)} I ct fam := S f 0 ∫ r = R d n + 1 x γ . (127)
which gives the same finite result for the black brane. In planar AdS, S = n S=n S = n and f 0 = L − 2 f_0=L^{-2} f 0 = L − 2 , and this reduces to the usual leading n / L n/L n / L counterterm. In a general anisotropic background, however, (127) is only a counterterm evaluated on the restricted solution family. It need not provide finite one-point functions or a well-posed variational problem under arbitrary variations of the anisotropic metric and matter sources. A complete Hamilton–Jacobi counterterm functional [68 ] may include intrinsic-curvature and matter invariants at the same asymptotic order.
We may therefore identify the fixed-flux free energy relative to the scaling background as
F N − F 0 = I E , r e n ( N ) β = − V b z 2 μ = − V b z f 0 2 r h Δ . (128) F_N-F_0=\frac{I^{(N)}_{E,\mathrm{ren}}}{\beta}=-\frac{\mathcal{V}_b z}{2}\mu=-\frac{\mathcal{V}_b z f_0}{2}r_h^\Delta. \tag*{(128)} F N − F 0 = β I E , ren ( N ) = − 2 V b z μ = − 2 V b z f 0 r h Δ . (128)
Combining this result with the independently obtained horizon entropy and temperature, the associated thermodynamic energy difference is
M − M 0 = ( F N − F 0 ) + T S B H = V b S 2 μ , Δ ( M − M 0 ) = S T S B H . (129) M-M_0=(F_N-F_0)+TS_{\mathrm{BH}}=\frac{\mathcal{V}_b S}{2}\mu,\qquad\Delta(M-M_0)=ST S_{\mathrm{BH}}. \tag*{(129)} M − M 0 = ( F N − F 0 ) + T S BH = 2 V b S μ , Δ ( M − M 0 ) = S T S BH . (129)
The resulting energy agrees with the restricted first law and the Hamiltonian variation in Appendix § 9 . At fixed supporting sources, the heat capacity is
C f i x e d : = T ( ∂ S B H ∂ T ) s o u r c e s = S z S B H . (130) C_{\mathrm{fixed}}:=T\left(\frac{\partial S_{\mathrm{BH}}}{\partial T}\right)_{\mathrm{sources}}=\frac{S}{z}S_{\mathrm{BH}}. \tag*{(130)} C fixed := T ( ∂ T ∂ S BH ) sources = z S S BH . (130)
The positive sign holds throughout the family (111) with S , z > 0 S,z>0 S , z > 0 . Thus the heat capacity is positive for variations of the horizon radius at fixed supporting sources. This establishes local thermal stability along this one-parameter family. Stability under more general thermodynamic or dynamical perturbations requires a separate analysis.
Massive probe scalar perturbations
We consider a free massive probe scalar on the backgrounds constructed above, with the metric, gauge field, and supporting dilaton ϕ \phi ϕ held fixed. Its quasinormal spectrum describes scalar relaxation and, in separable finite-momentum sectors, dispersion on the anisotropic geometry. This gives a first probe of dynamical response relevant to transport and of stability against scalar disturbances. Denoting the scalar mass by m m m , we use the dimensionless combination M 2 : = m 2 / f 0 M^2:=m^2/f_0 M 2 := m 2 / f 0 . Where a holographic description is available, the quasinormal frequencies correspond to poles of a scalar retarded correlation function [71 , 72 ] .
Klein–Gordon equation on a homogeneous spatial geometry
We introduce a real scalar Ψ \Psi Ψ with probe action
I Ψ : = − 1 2 ∫ d n + 2 x − g [ g μ ν ∂ μ Ψ ∂ ν Ψ + m 2 Ψ 2 ] , ( □ − m 2 ) Ψ = 0. (131) I_{\Psi}:=-\frac{1}{2}\int d^{n+2}x\sqrt{-g}\left[g^{\mu\nu}\partial_{\mu}\Psi\partial_{\nu}\Psi+m^2\Psi^2\right],\qquad(\Box-m^2)\Psi=0. \tag*{(131)} I Ψ := − 2 1 ∫ d n + 2 x − g [ g μν ∂ μ Ψ ∂ ν Ψ + m 2 Ψ 2 ] , ( □ − m 2 ) Ψ = 0. (131)
We work with the metric (111) , although most expressions below only require the general metric ansatz (6) . Its determinant is independent of the blackening factor and obeys − g = r Δ − 1 h b \sqrt{-g}=r^{\Delta-1}\sqrt{h_b} − g = r Δ − 1 h b , in the invariant spatial frame. Thus,
[ r − Δ + 1 ∂ r ( r Δ + 1 f ∂ r ) − ∂ t 2 r 2 z f + ∇ h r 2 − m 2 ] Ψ = 0 , h r : = ∑ i b i 2 r − 2 α i ( θ i ) 2 . (132) \left[r^{-\Delta+1}\partial_r\left(r^{\Delta+1}f\partial_r\right)-\frac{\partial_t^2}{r^{2z}f}+\nabla^2_{h_r}-m^2\right]\Psi=0,\qquad h_r:=\sum_i b_i^2r^{-2\alpha_i}(\theta^i)^2. \tag*{(132)} [ r − Δ + 1 ∂ r ( r Δ + 1 f ∂ r ) − r 2 z f ∂ t 2 + ∇ h r 2 − m 2 ] Ψ = 0 , h r := i ∑ b i 2 r − 2 α i ( θ i ) 2 . (132)
This expression is valid for any of the neutral or EMD backgrounds considered in this paper; the supporting matter fields enter only through the metric.
The spatial part requires some care. The invariant vector fields X i X_i X i introduced above satisfy [ X i , X j ] = c i j k X k [X_i,X_j]=c_{ij}^{k}X_k [ X i , X j ] = c ij k X k . The spatial Laplacian takes the form
∇ h r 2 = ∑ i r − 2 α i b i 2 ( X i 2 − ∑ j c i j j X j ) . (133) \nabla^2_{h_r}=\sum_i\frac{r^{-2\alpha_i}}{b_i^2}\left(X_i^2-\sum_j c_{ij}^{j}X_j\right). \tag*{(133)} ∇ h r 2 = i ∑ b i 2 r − 2 α i ( X i 2 − j ∑ c ij j X j ) . (133)
The terms linear in X i X_i X i arise from the invariant spatial volume form and may be nonzero for non-unimodular Bianchi groups. On a flat horizon, X i = ∂ i X_i=\partial_i X i = ∂ i and this is the usual Fourier-space Laplacian. For a non-Abelian spatial geometry, however, the operators belonging to different scaling exponents do not necessarily commute. Therefore a generic spatial perturbation cannot always be reduced to a single radial ordinary differential equation. One must instead solve a coupled system of spatial harmonics, or retain the corresponding radial–spatial partial differential equation.
A universal sector is the spatially homogeneous mode, Ψ ( t , r , x ) = e − i ω t ψ ( r ) \Psi(t,r,x)=e^{-i\omega t}\psi(r) Ψ ( t , r , x ) = e − iω t ψ ( r ) . At nonzero spatial momentum, separation requires an additional condition. Let A \mathcal{A} A be the set of distinct values among the spatial exponents α i \alpha_i α i , and define
L a : = ∑ i : α i = a 1 b i 2 ( X i 2 − ∑ j c i j j X j ) , ∇ h r 2 = ∑ a ∈ A r − 2 a L a . (134) L_a:=\sum_{i:\alpha_i=a}\frac{1}{b_i^2}\left(X_i^2-\sum_j c_{ij}^{j}X_j\right),\qquad\nabla^2_{h_r}=\sum_{a\in\mathcal{A}}r^{-2a}L_a. \tag*{(134)} L a := i : α i = a ∑ b i 2 1 ( X i 2 − j ∑ c ij j X j ) , ∇ h r 2 = a ∈ A ∑ r − 2 a L a . (134)
For a product Ψ = e − i ω t ψ ( r ) Y ( x ) \Psi=e^{-i\omega t}\psi(r)Y(x) Ψ = e − iω t ψ ( r ) Y ( x ) to reduce the equation to a single radial ODE, the same spatial harmonic must obey
L a Y = − κ a Y for every a ∈ A . (135) L_aY=-\kappa_aY\qquad\text{for every }a\in\mathcal{A}. \tag*{(135)} L a Y = − κ a Y for every a ∈ A . (135)
The distinct radial powers are independent functions of r r r ; diagonalizing the spatial Laplacian at one radius is therefore insufficient. Operators carrying the same radial power need only be diagonalized as a sum, whereas the groups with different powers must share the chosen eigenfunction. The κ a \kappa_a κ a are spatial spectral labels, distinct from the background parameter κ \kappa κ . With this condition, the equation of motion becomes
r − Δ + 1 ∂ r ( r Δ + 1 f ψ ′ ) + [ ω 2 r 2 z f − m 2 − ∑ a ∈ A κ a r − 2 a ] ψ = 0 , [ f D 2 + ( Δ f + D f ) D + ω 2 r 2 z f − m 2 − ∑ a ∈ A κ a r − 2 a ] ψ = 0. (136) \begin{aligned}
r^{-\Delta+1}\partial_r\left(r^{\Delta+1}f\psi'\right)+\left[\frac{\omega^2}{r^{2z}f}-m^2-\sum_{a\in\mathcal{A}}\kappa_a r^{-2a}\right]\psi&=0,\\
\left[f\mathcal{D}^2+(\Delta f+\mathcal{D}f)\mathcal{D}+\frac{\omega^2}{r^{2z}f}-m^2-\sum_{a\in\mathcal{A}}\kappa_a r^{-2a}\right]\psi&=0. \tag*{(136)}
\end{aligned} r − Δ + 1 ∂ r ( r Δ + 1 f ψ ′ ) + [ r 2 z f ω 2 − m 2 − a ∈ A ∑ κ a r − 2 a ] ψ [ f D 2 + ( Δ f + D f ) D + r 2 z f ω 2 − m 2 − a ∈ A ∑ κ a r − 2 a ] ψ = 0 , = 0. (136)
Here D = r ∂ r \mathcal{D}=r\partial_r D = r ∂ r . The homogeneous harmonic has κ a = 0 \kappa_a=0 κ a = 0 for all a a a and always satisfies it.
On a flat horizon, plane waves diagonalize every ∂ i 2 \partial_i^2 ∂ i 2 , and κ a = ∑ i : α i = a k i 2 \kappa_a = \sum_{i:\alpha_i=a} k_i^2 κ a = ∑ i : α i = a k i 2 , where k i k_i k i includes the factor b i − 1 b_i^{-1} b i − 1 . Only in this case, or another sector admitting the corresponding individual eigenfunctions, may the potential be interpreted as a sum of independent directional momentum squares. On Nil, the equal-scaling horizontal operators instead combine into an oscillator operator; its eigenvalue and the central momentum yield the separated potential described in § 6.5 . They are not three independent Cartesian momenta.
Solv illustrates the obstruction explicitly:
∇ h r 2 = r − 2 ( e − 2 w ∂ x 2 + e 2 w ∂ y 2 ) + b 3 − 2 ∂ w 2 . (137) \nabla^2_{h_r}=r^{-2}\left(e^{-2w}\partial_x^2+e^{2w}\partial_y^2\right)+b_3^{-2}\partial_w^2. \tag*{(137)} ∇ h r 2 = r − 2 ( e − 2 w ∂ x 2 + e 2 w ∂ y 2 ) + b 3 − 2 ∂ w 2 . (137)
After Fourier transforming x , y x,y x , y , the first group is multiplication by − k x 2 e − 2 w − k y 2 e 2 w -k_x^2e^{-2w}-k_y^2e^{2w} − k x 2 e − 2 w − k y 2 e 2 w , while the second differentiates in w w w . For generic nonzero transverse momenta these do not share a nonzero smooth eigenfunction, so the product ansatz fails. Setting k x = k y = 0 k_x=k_y=0 k x = k y = 0 leaves separable waves along w w w .
Asymptotic solutions and the scalar mass bound
We first consider solutions with the blackening function (111) , f ( r ) = f 0 − μ r − Δ f(r)=f_0-\mu r^{-\Delta} f ( r ) = f 0 − μ r − Δ , for which f ( r ) → f 0 > 0 f(r)\to f_0>0 f ( r ) → f 0 > 0 as r → ∞ r\to\infty r → ∞ . For the homogeneous mode, the frequency-dependent term is subleading when z > 0 z>0 z > 0 and the asymptotic equation becomes
[ f 0 D ( D + Δ ) − m 2 ] ψ = 0. (138) \left[f_0\mathcal{D}(\mathcal{D}+\Delta)-m^2\right]\psi=0. \tag*{(138)} [ f 0 D ( D + Δ ) − m 2 ] ψ = 0. (138)
Define the radial falloff exponents by
ν : = Δ 2 4 + M 2 , δ ± : = Δ 2 ± ν . (139) \nu:=\sqrt{\frac{\Delta^2}{4}+M^2},\qquad\delta_\pm:=\frac{\Delta}{2}\pm\nu. \tag*{(139)} ν := 4 Δ 2 + M 2 , δ ± := 2 Δ ± ν . (139)
Away from coincident roots, write J ( ω ) J(\omega) J ( ω ) and V ( ω ) V(\omega) V ( ω ) for the coefficients of the minus and plus falloffs, respectively. Then
ψ ( r ) ≃ J ( ω ) r − δ − + V ( ω ) r − δ + . (140) \psi(r)\simeq J(\omega)r^{-\delta_-}+V(\omega)r^{-\delta_+}. \tag*{(140)} ψ ( r ) ≃ J ( ω ) r − δ − + V ( ω ) r − δ + . (140)
This is the direct generalization of the familiar AdS scalar falloffs, with the AdS boundary dimension replaced by the effective radial scaling exponent z + S z+S z + S . In standard Dirichlet quantization, J J J is the prescribed leading boundary datum and V V V is the response. In a putative dual field theory, δ + \delta_+ δ + is the corresponding formal scaling dimension in the chosen radial convention. The invariant comparison between solutions is provided by ratios such as δ + / z \delta_+/z δ + / z . The holographic identification of J J J and V V V requires a choice of asymptotic boundary conditions and scalar counterterms; it does not follow from the metric alone.
The roots remain real provided
M 2 ≥ M B F 2 : = − Δ 2 4 . (141) M^2\geq M_{\mathrm{BF}}^2:=-\frac{\Delta^2}{4}. \tag*{(141)} M 2 ≥ M BF 2 := − 4 Δ 2 . (141)
This is a BF-like asymptotic mass bound for the homogeneous sector of these scaling geometries, analogous to the stability bound at a local maximum of AdS [73 ] . For planar A d S n + 2 \mathrm{AdS}_{n+2} AdS n + 2 , z = α i = 1 z=\alpha_i=1 z = α i = 1 , Δ = n + 1 \Delta=n+1 Δ = n + 1 , and f 0 = L − 2 f_0=L^{-2} f 0 = L − 2 , recovering M B F 2 = − ( n + 1 ) 2 / 4 M_{\mathrm{BF}}^2=-(n+1)^2/4 M BF 2 = − ( n + 1 ) 2 /4 . For the five-dimensional neutral Solv solution, one has ( z , S , Δ ) = ( 1 , 2 , 3 ) (z,S,\Delta)=(1,2,3) ( z , S , Δ ) = ( 1 , 2 , 3 ) and therefore δ ± = 3 / 2 ± 9 / 4 + M 2 \delta_\pm=3/2\pm\sqrt{9/4+M^2} δ ± = 3/2 ± 9/4 + M 2 ; for Nil, ( z , S , Δ ) = ( 3 / 2 , 4 , 11 / 2 ) (z,S,\Delta)=(3/2,4,11/2) ( z , S , Δ ) = ( 3/2 , 4 , 11/2 ) gives δ ± = 11 / 4 ± 121 / 16 + M 2 \delta_\pm=11/4\pm\sqrt{121/16+M^2} δ ± = 11/4 ± 121/16 + M 2 . The radial falloff exponents are thus sensitive to the anisotropic scaling, even for a probe that carries no spatial momentum. Similar massive-scalar asymptotics and quasinormal boundary conditions have been examined in Lifshitz black holes [74 ] .
There are two conditions for spatially varying modes. First, momentum along an unscaled direction, α i = 0 \alpha_i=0 α i = 0 , is equivalent to a mass shift: the associated term κ 0 r 0 \kappa_0 r^0 κ 0 r 0 in (136) survives as r → ∞ r\to\infty r → ∞ and enters the indicial equation at the same order as m 2 m^2 m 2 . For such a separable mode, with − L 0 Y = k 0 2 Y -L_0Y = k_0^2Y − L 0 Y = k 0 2 Y and k 0 2 ≥ 0 k_0^2 \ge0 k 0 2 ≥ 0 , the replacement m 2 → m 2 + k 0 2 m^2 \to m^2 + k_0^2 m 2 → m 2 + k 0 2 must be made in (140) . The neutral Solv background provides an example: the third spatial direction is unscaled, so its momentum shifts the asymptotic scalar mass. Second, if any α i < 0 \alpha_i < 0 α i < 0 , a nonzero spatial derivative in that direction grows toward large r r r . In this case momentum may dominate the asymptotic equation, and the solutions need not have the radial falloffs of the form (140) . This issue is relevant, for example, to the Bianchi VIh _h h family with 0 < h < 1 0 < h < 1 0 < h < 1 , where α 1 = − h < 0 \alpha_1 = -h < 0 α 1 = − h < 0 . The finite-momentum asymptotics must therefore be determined separately in these sectors.
The radial Klein–Gordon norm of the slow falloff behaves asymptotically as ∫ ∞ d r r − 2 z − 1 + 2 ν \int^\infty\mathrm{d}r\,r^{-2z-1+2\nu} ∫ ∞ d r r − 2 z − 1 + 2 ν . Hence both falloffs are normalizable under this elementary norm criterion when 0 ≤ ν < z 0 \le\nu< z 0 ≤ ν < z . Normalizability alone does not guarantee that alternative scalar boundary conditions define a stable theory. For ordinary Lifshitz backgrounds, additional short-wavelength instabilities impose a stronger restriction [75 ] ; the analogous analysis for the more general anisotropic spatial geometries is outside the scope of this paper.
Quasinormal modes
For a nonextremal horizon at r = r h r = r_h r = r h , write f ( r ) = f ′ ( r h ) ( r − r h ) + ⋯ f(r) = f'(r_h)(r-r_h)+\cdots f ( r ) = f ′ ( r h ) ( r − r h ) + ⋯ . The radial equation gives two local solutions, corresponding to ingoing and outgoing waves:
ψ ( r ) ∼ ( r − r h ) ∓ i ω / ( 4 π T ) , T = r h z + 1 f ′ ( r h ) 4 π . (142) \psi(r) \sim(r-r_h)^{\mp i\omega/(4\pi T)}, \qquad T = \frac{r_h^{z+1}f'(r_h)}{4\pi}. \tag*{(142)} ψ ( r ) ∼ ( r − r h ) ∓ iω / ( 4 π T ) , T = 4 π r h z + 1 f ′ ( r h ) . (142)
With the time dependence e − i ω t \mathrm{e}^{-i\omega t} e − iω t , the minus sign is ingoing at the future horizon. Equivalently, the full field is regular in the ingoing Eddington–Finkelstein coordinate v : = t + ∫ d r / ( r z + 1 f ) v := t+\int\mathrm{d}r/(r^{z+1}f) v := t + ∫ d r / ( r z + 1 f ) . The quasinormal frequencies are those complex ω \omega ω for which the ingoing solution also obeys the chosen source-free boundary condition, J ( ω ) = 0 J(\omega)=0 J ( ω ) = 0 . Under the standard holographic prescription these frequencies are poles of the scalar retarded Green function, G R ∝ V / J G_R \propto V/J G R ∝ V / J , up to contact terms and normalization [71 , 72 ] . They describe the relaxation of an external scalar operator; since Ψ \Psi Ψ is not a conserved density, there is no reason to expect a hydrodynamic mode with ω → 0 \omega\to0 ω → 0 in this probe sector.
For the blackening function (111) , it is useful to introduce u : = r h / r u := r_h/r u := r h / r and w : = ω / ( 2 π T ) \mathfrak{w} := \omega/(2\pi T) w := ω / ( 2 π T ) . (136) becomes
u 2 ( 1 − u Δ ) ψ u u + u ( 1 − Δ − u Δ ) ψ u + [ Δ 2 w 2 u 2 z 4 ( 1 − u Δ ) − M 2 − ∑ a ∈ A κ a f 0 r h 2 a u 2 a ] ψ = 0. (143) u^2(1-u^\Delta)\psi_{uu}+u(1-\Delta-u^\Delta)\psi_u+\left[\frac{\Delta^2\mathfrak{w}^2u^{2z}}{4(1-u^\Delta)}-M^2-\sum_{a\in A}\frac{\kappa_a}{f_0r_h^{2a}}u^{2a}\right]\psi=0. \tag*{(143)} u 2 ( 1 − u Δ ) ψ uu + u ( 1 − Δ − u Δ ) ψ u + [ 4 ( 1 − u Δ ) Δ 2 w 2 u 2 z − M 2 − a ∈ A ∑ f 0 r h 2 a κ a u 2 a ] ψ = 0. (143)
For the spatially homogeneous mode, label the quasinormal overtones by the nonnegative integer N N N . Their frequencies have the scaling form
ω N = 2 π T w N ( z , S , M 2 ) . (144) \omega_N=2\pi T\,\mathfrak{w}_N(z,S,M^2). \tag*{(144)} ω N = 2 π T w N ( z , S , M 2 ) . (144)
Here w N \mathfrak{w}_N w N is determined by the radial boundary-value problem. Thus ω N / T \omega_N/T ω N / T is independent of the horizon radius within a fixed background of the form (111) . For separable finite-momentum modes, the additional arguments are κ a / ( f 0 r h 2 a ) \kappa_a/(f_0r_h^{2a}) κ a / ( f 0 r h 2 a ) , subject to the common-harmonic condition (135) . In general the complex values of w N \mathfrak{w}_N w N require a numerical solution, for example by imposing the ingoing factor ( 1 − u ) − i ω / ( 4 π T ) (1-u)^{-i\omega/(4\pi T)} ( 1 − u ) − iω / ( 4 π T ) and solving for a vanishing leading boundary coefficient.
An analytic quasinormal spectrum for z = S z=S z = S
A useful simplification occurs for the homogeneous mode when z = S z=S z = S , so that Δ = 2 z \Delta=2z Δ = 2 z , provided the blackening factor has the form f = f 0 [ 1 − ( r h / r ) 2 z ] f=f_0[1-(r_h/r)^{2z}] f = f 0 [ 1 − ( r h / r ) 2 z ] , with no logarithmic term. This includes special members of the neutral family and does not require the spacetime to be AdS. Define
y : = ( r h r ) 2 z , a : = δ + 2 z = 1 2 + 1 2 z z 2 + M 2 . (145) y:=\left(\frac{r_h}{r}\right)^{2z}, \qquad a:=\frac{\delta_+}{2z}=\frac{1}{2}+\frac{1}{2z}\sqrt{z^2+M^2}. \tag*{(145)} y := ( r r h ) 2 z , a := 2 z δ + = 2 1 + 2 z 1 z 2 + M 2 . (145)
The radial equation then becomes
y 2 ( 1 − y ) ψ y y − y 2 ψ y + [ m 2 y 4 ( 1 − y ) − m 2 4 z 2 f 0 ] ψ = 0. (146) y^2(1-y)\psi_{yy}-y^2\psi_y+\left[\frac{m^2y}{4(1-y)}-\frac{m^2}{4z^2f_0}\right]\psi=0. \tag*{(146)} y 2 ( 1 − y ) ψ y y − y 2 ψ y + [ 4 ( 1 − y ) m 2 y − 4 z 2 f 0 m 2 ] ψ = 0. (146)
This is a hypergeometric equation after extracting its asymptotic powers. A solution ingoing at y = 1 y=1 y = 1 is
ψ i n = y a ( 1 − y ) − i ω / 2 2 F 1 ( a − i ω 2 , a − i ω 2 ; 1 − i ω ; 1 − y ) . (147) \psi_{\mathrm{in}}=y^a(1-y)^{-i\omega/2}{}_2F_1\left(a-\frac{i\omega}{2},a-\frac{i\omega}{2};1-i\omega;1-y\right). \tag*{(147)} ψ in = y a ( 1 − y ) − iω /2 2 F 1 ( a − 2 iω , a − 2 iω ; 1 − iω ; 1 − y ) . (147)
For ν > 0 \nu>0 ν > 0 , the coefficient of the slow falloff y 1 − a y^{1-a} y 1 − a is proportional to
J ( ω ) ∝ Γ ( 1 − i ω ) Γ ( 2 a − 1 ) Γ ( a − i ω / 2 ) 2 . (148) J(\omega)\propto\frac{\Gamma(1-i\omega)\Gamma(2a-1)}{\Gamma\left(a-i\omega/2\right)^2}. \tag*{(148)} J ( ω ) ∝ Γ ( a − iω /2 ) 2 Γ ( 1 − iω ) Γ ( 2 a − 1 ) . (148)
Vanishing of the source coefficient gives the exact homogeneous quasinormal spectrum, with the usual limiting prescription in the cases where the arguments of the gamma functions take integer values,
ω N = − i 4 π T ( N + δ + Δ ) , N = 0 , 1 , 2 , … , z = S . (149) \omega_N=-i4\pi T\left(N+\frac{\delta_+}{\Delta}\right),\qquad N=0,1,2,\ldots,\qquad z=S. \tag*{(149)} ω N = − i 4 π T ( N + Δ δ + ) , N = 0 , 1 , 2 , … , z = S . (149)
In particular, for M 2 = 0 M^2=0 M 2 = 0 one has δ + = Δ \delta_+=\Delta δ + = Δ , and ω N = − i 4 π T ( N + 1 ) \omega_N=-i4\pi T(N+1) ω N = − i 4 π T ( N + 1 ) . These scalar modes are purely damped. Since δ + \delta_+ δ + is real for any M 2 M^2 M 2 obeying the bound (141) , the frequencies (149) are purely imaginary: at z = S z=S z = S the homogeneous spectrum lies entirely on the negative imaginary axis, so the modes are overdamped, with no oscillation frequency. Consistently, the time-reversal partners ω → − ω ˉ \omega\to-\bar{\omega} ω → − ω ˉ degenerate to self-pairings on the axis. For real asymptotic exponents, the spectrum contains no growing homogeneous quasinormal modes under the chosen source-free boundary condition. This is evidence for stability within this probe sector, not for the coupled gravitational and matter perturbations. Away from this homogeneous z = S z=S z = S sector, the spectrum is determined by the corresponding radial boundary-value problem.
Numerical comparison of probe mass and supporting matter
We probe the linear response of the anisotropic scaling geometries of § 2 with a simple minimally coupled scalar field, computing its spectra in the homogeneous and separable finite-momentum sectors and varying the probe mass and background parameters under standard source-free boundary conditions. Gravitational quasinormal modes of Lifshitz black branes have been computed in [76 ] ; the scalar sector isolates the linear response of the background before the coupled metric, gauge-field, and dilaton perturbations are included.
Background parameters. For fixed spatial exponents, let z N : = Q / S z_{\mathrm{N}}:=Q/S z N := Q / S , with S > 0 S>0 S > 0 . The electric extensions of the neutral Nil and Bianchi V I h \mathrm{VI}_h VI h solutions can be parameterized by
σ 2 = S ( z − z N ) , z ≥ z N , (150) \sigma^2=S(z-z_{\mathrm{N}}),\qquad z\ge z_{\mathrm{N}}, \tag*{(150)} σ 2 = S ( z − z N ) , z ≥ z N , (150)
f 0 = − 2 Λ ( z + S ) [ z + ( n − 1 ) z N ] , E = ( z − z N ) ( z + S ) f 0 , (151) f_0=\frac{-2\Lambda}{(z+S)[z+(n-1)z_{\mathrm{N}}]},\qquad\mathcal{E}=(z-z_{\mathrm{N}})(z+S)f_0, \tag*{(151)} f 0 = ( z + S ) [ z + ( n − 1 ) z N ] − 2Λ , E = ( z − z N ) ( z + S ) f 0 , (151)
λ σ = − 2 S , R ^ ı ^ ı ^ = ( α i − z N ) ( z + S ) f 0 . (152) \lambda\sigma=-2S,\qquad\widehat{R}_{\hat{\imath}\hat{\imath}}=(\alpha_i-z_{\mathrm{N}})(z+S)f_0. \tag*{(152)} λσ = − 2 S , R ^ ^ = ( α i − z N ) ( z + S ) f 0 . (152)
At the endpoint z = z N z=z_{\mathrm{N}} z = z N we use the neutral solution directly. Away from it, the spatial curvature scales adjust according to the last equation. The scan changes the EMD coupling λ \lambda λ and therefore compares backgrounds across a family of theories, rather than varying an independent charge at fixed coupling. The three families and their parameters are listed in Table 1 . Magnetic and dyonic solutions admit the same probe equation below when their blackening factor has the same form, but their allowed parameters must separately satisfy the magnetic and spatial constraints.
Table 1. Background families for the homogeneous numerical comparison. In the last row h < 1 h<1 h < 1 ; Solv is the case h = − 1 h=-1 h = − 1 .
Equation and boundary conditions. We use Fourier time dependence e − i w v e^{-i\mathfrak{w}v} e − i w v in ingoing Eddington–Finkelstein coordinates, v = t + ∫ d r r z + 1 f v=t+\int\frac{\mathrm{d}r}{r^{z+1}f} v = t + ∫ r z + 1 f d r , with u = r h / r u=r_h/r u = r h / r . Frequencies are quoted in thermal units,
w : = ω 2 π T , 2 π T = Δ f 0 r h z 2 . (153) \mathfrak{w}:=\frac{\omega}{2\pi T}, \qquad2\pi T=\frac{\Delta f_0 r_h^z}{2}. \tag*{(153)} w := 2 π T ω , 2 π T = 2 Δ f 0 r h z . (153)
Let F ( u ) : = f ( r h / u ) / f 0 = 1 − u Δ F(u):=f(r_h/u)/f_0=1-u^\Delta F ( u ) := f ( r h / u ) / f 0 = 1 − u Δ . For each separable harmonic, define the momentum potential
K ( u ) : = ∑ a ∈ A κ a f 0 r h 2 a u 2 a . (154) \mathcal{K}(u):=\sum_{a\in A}\frac{\kappa_a}{f_0 r_h^{2a}}u^{2a}. \tag*{(154)} K ( u ) := a ∈ A ∑ f 0 r h 2 a κ a u 2 a . (154)
It vanishes for the homogeneous mode. When its boundary limit is finite, we use the momentum-dependent exponent
δ + : = Δ 2 + Δ 2 4 + M 2 + K ( 0 ) . (155) \delta_+:=\frac{\Delta}{2}+\sqrt{\frac{\Delta^2}{4}+M^2+\mathcal{K}(0)}. \tag*{(155)} δ + := 2 Δ + 4 Δ 2 + M 2 + K ( 0 ) . (155)
This reduces to the homogeneous exponent defined above when K ( 0 ) = 0 \mathcal{K}(0)=0 K ( 0 ) = 0 . For each harmonic write Ψ = e − i w v u δ + y ( u ) Y ( x ) \Psi=e^{-i\mathfrak{w}v}u^{\delta_+}y(u)Y(x) Ψ = e − i w v u δ + y ( u ) Y ( x ) . The single numerical equation, including momentum, is
0 = u 2 F y ′ ′ + { u [ ( 2 δ + + 1 − Δ ) F + u F ′ ] + i Δ w u z + 1 } y ′ + { δ + [ ( δ + − Δ ) F + u F ′ ] − M 2 − K ( u ) + i Δ w 2 ( 2 δ + − S ) u z } y . (156) 0=u^2Fy''+\left\{u\left[(2\delta_++1-\Delta)F+uF'\right]+i\Delta\mathfrak{w}u^{z+1}\right\}y'
+\left\{\delta_+\left[(\delta_+-\Delta)F+uF'\right]-M^2-\mathcal{K}(u)+\frac{i\Delta\mathfrak{w}}{2}(2\delta_+-S)u^z\right\}y. \tag*{(156)} 0 = u 2 F y ′′ + { u [ ( 2 δ + + 1 − Δ ) F + u F ′ ] + i Δ w u z + 1 } y ′ + { δ + [ ( δ + − Δ ) F + u F ′ ] − M 2 − K ( u ) + 2 i Δ w ( 2 δ + − S ) u z } y . (156)
Primes denote u u u derivatives. Source-free boundary conditions select the finite branch of y y y at u = 0 u=0 u = 0 , excluding the branch u Δ − 2 δ + u^{\Delta-2\delta_+} u Δ − 2 δ + . In particular y ( 0 ) y(0) y ( 0 ) need not vanish. At the horizon select the regular ingoing Frobenius branch, satisfying
0 = Δ ( − 1 + i w ) y ′ ( 1 ) + [ − δ + Δ − M 2 − K ( 1 ) + i Δ w 2 ( 2 δ + − S ) ] y ( 1 ) . (157) 0=\Delta(-1+i\mathfrak{w})y'(1)+\left[-\delta_+\Delta-M^2-\mathcal{K}(1)+\frac{i\Delta\mathfrak{w}}{2}(2\delta_+-S)\right]y(1). \tag*{(157)} 0 = Δ ( − 1 + i w ) y ′ ( 1 ) + [ − δ + Δ − M 2 − K ( 1 ) + 2 i Δ w ( 2 δ + − S ) ] y ( 1 ) . (157)
For the numerical calculation, we discretize the time-even equation (143) in the factorized variable
ψ = u δ + ( 1 − u Δ ) − i w / 2 q ( u ) , F ( u ) = 1 − u Δ , (158) \psi=u^{\delta_+}(1-u^\Delta)^{-i\mathfrak{w}/2}q(u), \qquad F(u)=1-u^\Delta, \tag*{(158)} ψ = u δ + ( 1 − u Δ ) − i w /2 q ( u ) , F ( u ) = 1 − u Δ , (158)
for which the horizon singularity of the w 2 / ( 1 − u Δ ) \mathfrak{w}^2/(1-u^\Delta) w 2 / ( 1 − u Δ ) term is removed and the equation becomes a regular nonlinear eigenvalue problem on u ∈ [ 0 , 1 ] u\in[0,1] u ∈ [ 0 , 1 ] , with quadratic dependence on w \mathfrak{w} w ,
0 = u 2 F q ′ ′ + { u [ ( 2 δ + + 1 − Δ ) − ( 2 δ + + 1 ) u Δ ] + i Δ w u Δ + 1 } q ′ + { K ( 0 ) − δ + 2 u Δ − K ( u ) + i Δ w δ + u Δ + w 2 Δ 2 u 2 z ( 1 − u 2 S ) 4 ( 1 − u Δ ) } q . (159) 0=u^2Fq''+\left\{u\left[(2\delta_++1-\Delta)-(2\delta_++1)u^\Delta\right]+i\Delta\mathfrak{w}u^{\Delta+1}\right\}q'
+\left\{\mathcal{K}(0)-\delta_+^2u^\Delta-\mathcal{K}(u)+i\Delta\mathfrak{w}\delta_+u^\Delta+\frac{\mathfrak{w}^2\Delta^2u^{2z}(1-u^{2S})}{4(1-u^\Delta)}\right\}q. \tag*{(159)} 0 = u 2 F q ′′ + { u [ ( 2 δ + + 1 − Δ ) − ( 2 δ + + 1 ) u Δ ] + i Δ w u Δ + 1 } q ′ + { K ( 0 ) − δ + 2 u Δ − K ( u ) + i Δ w δ + u Δ + 4 ( 1 − u Δ ) w 2 Δ 2 u 2 z ( 1 − u 2 S ) } q . (159)
This equation is equivalent to (143) and was checked against the exact hypergeometric solution of § 6.4 at generic complex w \mathfrak{w} w . It is collocated on a Chebyshev–Lobatto grid of N + 1 N+1 N + 1 points in u u u , where N N N now denotes the collocation resolution rather than an overtone number (with the second-derivative matrix built as D 1 2 D_1^2 D 1 2 ). The horizon condition is the row obtained by evaluating (159) at u = 1 u=1 u = 1 ,
Δ ( − 1 + i w ) q ′ ( 1 ) + [ K ( 0 ) − δ + 2 − K ( 1 ) + i Δ w δ + + 1 2 w 2 Δ S ] q ( 1 ) = 0. (160) \Delta(-1+i\mathfrak{w})q'(1)+\left[\mathcal{K}(0)-\delta_+^2-\mathcal{K}(1)+i\Delta\mathfrak{w}\delta_++\frac{1}{2}\mathfrak{w}^2\Delta S\right]q(1)=0. \tag*{(160)} Δ ( − 1 + i w ) q ′ ( 1 ) + [ K ( 0 ) − δ + 2 − K ( 1 ) + i Δ w δ + + 2 1 w 2 Δ S ] q ( 1 ) = 0. (160)
which selects the same regular ingoing branch as (157) . The source-free condition is imposed as q ′ ( 0 ) = 0 q'(0)=0 q ′ ( 0 ) = 0 : the homogeneous regular branch behaves as q = q 0 + O ( u min ( 2 z , Δ ) ) q=q_0+O(u^{\min(2z,\Delta)}) q = q 0 + O ( u m i n ( 2 z , Δ ) ) away from resonant expansions, and min ( 2 z , Δ ) > 1 \min(2z,\Delta)>1 min ( 2 z , Δ ) > 1 for the backgrounds scanned here, so the divergent branch u Δ − 2 δ + ( 1 − u Δ ) i ω / 2 u^{\Delta-2\delta_+}(1-u^\Delta)^{i\omega/2} u Δ − 2 δ + ( 1 − u Δ ) iω /2 is excluded. The resulting nonlinear eigenvalue problem ( A + ω B + ω 2 C ) q = 0 (\mathbf{A}+\omega\mathbf{B}+\omega^2\mathbf{C})\mathbf{q}=0 ( A + ω B + ω 2 C ) q = 0 is linearized in the standard companion form and solved with the QZ algorithm. Because the factorized equation (159) encodes the ingoing branch at the horizon by construction, the outgoing (anti-quasinormal) partner of a mode is not an eigenvalue of the same problem. For real background parameters, the physical spectral symmetry is ω ↔ − ω ˉ \omega\leftrightarrow-\bar{\omega} ω ↔ − ω ˉ , which preserves the sign of the imaginary part.
Spurious collocation eigenvalues are filtered out by comparing the spectra at several resolutions N N N . We retain eigenvalues whose relative variation across N = 14 , 16 , 18 N=14,16,18 N = 14 , 16 , 18 is less than 6 × 10 − 2 6\times10^{-2} 6 × 1 0 − 2 , and report their values at N = 16 N=16 N = 16 . Only damped modes with Im ω < 0 \operatorname{Im}\omega<0 Im ω < 0 and ∣ ω ∣ ≤ 12 |\omega|\leq12 ∣ ω ∣ ≤ 12 are included. We call modes passing this resolution-stability test verified . The filtering tolerance is a selection criterion rather than an error estimate; accuracy is checked separately against the analytic spectrum and the shooting calculation in Appendix § 10 .
Parameter scans. We scan M 2 M^2 M 2 and z z z directly, displaying frequencies as w = ω / ( 2 π T ) \mathbf{w}=\omega/(2\pi T) w = ω / ( 2 π T ) . Holding M 2 M^2 M 2 fixed keeps the probe mass fixed in radial curvature units. At fixed Λ \Lambda Λ , a fixed dimensionful mass instead requires using M 2 ( z ) M^2(z) M 2 ( z ) . The initial comparison uses M 2 = 0 , 1 M^2=0,1 M 2 = 0 , 1 and z / z N = 1 , 3 / 2 z/z_{\mathrm{N}}=1,3/2 z / z N = 1 , 3/2 , followed by continuous scans 0 ≤ M 2 ≤ 4 0\leq M^2\leq4 0 ≤ M 2 ≤ 4 and 1 ≤ ( z / z N ) ≤ 4 1\leq(z/z_{\mathrm{N}})\leq4 1 ≤ ( z / z N ) ≤ 4 ; the M 2 = 4 M^2=4 M 2 = 4 values appear in the benchmark table of Appendix § 10 . The Nil and Solv families provide the principal examples; a fixed h = − 1 / 2 h=-1/2 h = − 1/2 provides a distinct Bianchi V I h \mathrm{VI}_h VI h comparison. All these probe masses lie above the homogeneous BF-like bound.
For each family we compare the massless neutral endpoint with a massive neutral probe, a massless probe on the electric extension, and a massive probe on that extension. We display both Re w \operatorname{Re}\mathbf{w} Re w and − Im w -\operatorname{Im}\mathbf{w} − Im w . The latter measures damping for a stable mode, and τ T = [ 2 π ( − Im w ) ] − 1 \tau T=[2\pi(-\operatorname{Im}\mathbf{w})]^{-1} τ T = [ 2 π ( − Im w ) ] − 1 . We track the least damped branches continuously as the parameters vary to resolve crossings and changes in their damping order.
Finite momentum and spatial harmonics. For − ∇ h r 2 Y = V k ( r ) Y -\nabla_{h_r}^2Y=V_k(r)Y − ∇ h r 2 Y = V k ( r ) Y , the momentum potential defined above is K ( u ) = V k ( r h / u ) / f 0 \mathcal{K}(u)=V_k(r_h/u)/f_0 K ( u ) = V k ( r h / u ) / f 0 . This definition groups all spatial labels into one radial function; its finite boundary value is already included in δ + \delta_+ δ + above. The spatial boundary conditions and harmonic labels must be fixed before this reduction; independent Cartesian momenta are generally unavailable on a non-Abelian spatial geometry, and a stationary brane may carry momentum along the symmetry-breaking directions [77 ] . For compact quotients of these spaces, the momenta must belong to the discrete spectrum of the Laplacian on the quotient.
For Nil, the coframe in § 4.1 gives
∇ h b 2 = r − 2 [ ∂ x 2 + ( ∂ y + x ∂ w ) 2 ] + b 3 2 r − 4 ∂ w 2 . (161) \nabla_{h_b}^2=r^{-2}\left[\partial_x^2+\left(\partial_y+x\partial_w\right)^2\right]+b_3^2r^{-4}\partial_w^2. \tag*{(161)} ∇ h b 2 = r − 2 [ ∂ x 2 + ( ∂ y + x ∂ w ) 2 ] + b 3 2 r − 4 ∂ w 2 . (161)
There are two convenient separable sectors. For k w = 0 k_w=0 k w = 0 , a plane wave in ( x , y ) (x,y) ( x , y ) has
k ^ ⊥ 2 : = k x 2 + k y 2 f 0 r h 2 , K ( u ) = k ^ ⊥ 2 u 2 . (162) \hat{k}_{\perp}^{\,2}:=\frac{k_x^2+k_y^2}{f_0r_h^2},\qquad\mathcal{K}(u)=\hat{k}_{\perp}^{\,2}u^2. \tag*{(162)} k ^ ⊥ 2 := f 0 r h 2 k x 2 + k y 2 , K ( u ) = k ^ ⊥ 2 u 2 . (162)
For k w ≠ 0 k_w\neq0 k w = 0 , take Y = e i ( k y y + k w w ) X ℓ ( x ) Y=e^{i(k_y y+k_w w)}X_\ell(x) Y = e i ( k y y + k w w ) X ℓ ( x ) , where
[ − ∂ x 2 + ( k y + k w x ) 2 ] X ℓ = ( 2 ℓ + 1 ) ∣ k w ∣ X ℓ , ℓ = 0 , 1 , … (163) \left[-\partial_x^2+(k_y+k_wx)^2\right]X_\ell=(2\ell+1)|k_w|X_\ell,\qquad\ell=0,1,\ldots \tag*{(163)} [ − ∂ x 2 + ( k y + k w x ) 2 ] X ℓ = ( 2 ℓ + 1 ) ∣ k w ∣ X ℓ , ℓ = 0 , 1 , … (163)
These oscillator harmonics give
K ( u ) = ( 2 ℓ + 1 ) ∣ k w ∣ f 0 r h 2 u 2 + k w 2 f 0 b 3 2 r h 4 u 4 . (164) \mathcal{K}(u)=\frac{(2\ell+1)|k_w|}{f_0r_h^2}u^2+\frac{k_w^2}{f_0b_3^2r_h^4}u^4. \tag*{(164)} K ( u ) = f 0 r h 2 ( 2 ℓ + 1 ) ∣ k w ∣ u 2 + f 0 b 3 2 r h 4 k w 2 u 4 . (164)
Both momentum terms vanish at the boundary, so the homogeneous exponent δ + \delta_{+} δ + remains valid. The oscillator level is a spatial label and is distinct from the radial QNM overtone. At fixed level, the limit k w → 0 k_{\mathrm{w}}\to0 k w → 0 does not reproduce an arbitrary nonzero transverse plane-wave momentum. For the electric Nil family, b 3 2 = Δ f 0 b_3^2=\Delta f_0 b 3 2 = Δ f 0 ; defining ρ : = ∣ k w ∣ / ( f 0 r h 2 ) \rho:=|k_{\mathrm{w}}|/(f_0r_{\mathrm{h}}^2) ρ := ∣ k w ∣/ ( f 0 r h 2 ) yields K = ( 2 ℓ + 1 ) ρ 2 u 2 + ρ 2 u 4 / Δ \mathcal{K}=(2\ell+1)\rho^2u^2+\rho^2u^4/\Delta K = ( 2 ℓ + 1 ) ρ 2 u 2 + ρ 2 u 4 /Δ .
For Bianchi V I h \mathrm{VI}_h VI h , including Solv at h = − 1 h=-1 h = − 1 , the spatial operator is
∇ h 2 = r 2 h e − 2 w ∂ x 2 + r − 2 e − 2 h w ∂ y 2 + b 3 − 2 [ ∂ w 2 + ( 1 + h ) ∂ w ] . (165) \nabla^2_{\mathrm{h}}=r^{2h}e^{-2w}\partial_x^2+r^{-2}e^{-2hw}\partial_y^2+b_3^{-2}\left[\partial_w^2+(1+h)\partial_w\right]. \tag*{(165)} ∇ h 2 = r 2 h e − 2 w ∂ x 2 + r − 2 e − 2 h w ∂ y 2 + b 3 − 2 [ ∂ w 2 + ( 1 + h ) ∂ w ] . (165)
A separable nonzero-momentum sector is obtained by setting k x = k y = 0 k_x=k_y=0 k x = k y = 0 and taking Y = e − ( 1 + h ) w / 2 e i p w Y=e^{-(1+h)w/2}e^{ipw} Y = e − ( 1 + h ) w /2 e i pw , p ∈ R p\in\mathbb{R} p ∈ R . The prefactor accounts for the measure e ( 1 + h ) w d w e^{(1+h)w}\mathrm{d}w e ( 1 + h ) w d w . In this sector
K 0 : = p 2 + ( 1 + h ) 2 / 4 f 0 b 3 2 , K ( u ) = K 0 . (166) \mathcal{K}_0:=\frac{p^2+(1+h)^2/4}{f_0b_3^2},\qquad\mathcal{K}(u)=\mathcal{K}_0. \tag*{(166)} K 0 := f 0 b 3 2 p 2 + ( 1 + h ) 2 /4 , K ( u ) = K 0 . (166)
The radial spectrum therefore depends on M 2 + K 0 M^2+\mathcal{K}_0 M 2 + K 0 exactly as the homogeneous spectrum depends on M 2 M^2 M 2 . This is a useful numerical check. For Solv the measure prefactor disappears and p = 0 p=0 p = 0 is the homogeneous mode. For h ≠ − 1 h\neq-1 h = − 1 , the constant homogeneous harmonic is a separate sector and is not the p = 0 p=0 p = 0 member of this real-p p p continuum.
We restrict the numerical analysis to these separable sectors: homogeneous modes, Nil transverse plane waves, Nil central oscillator harmonics, and V I h \mathrm{VI}_h VI h modes along the unscaled direction. Generic transverse momentum on V I h \mathrm{VI}_h VI h is outside the scope of the present numerical study. Every calculation below therefore reduces to a single radial ordinary differential equation.
Momentum comparisons. We compute dispersion curves at M 2 = 0 , 1 M^2=0,1 M 2 = 0 , 1 and z / z N = 1 , 3 / 2 z/z_{\mathrm{N}}=1,3/2 z / z N = 1 , 3/2 for Nil transverse momenta 0 ≤ k ^ ⊥ ≤ 2 0\leq\widehat{k}_{\perp}\leq2 0 ≤ k ⊥ ≤ 2 and Solv unscaled momenta p ^ : = p / ( f 0 b 3 ) \widehat{p}:=p/(\sqrt{f_0}b_3) p := p / ( f 0 b 3 ) in the range 0 ≤ p ^ ≤ 2 0\leq\widehat{p}\leq2 0 ≤ p ≤ 2 . For Nil central momentum, we scan 0 ≤ ρ ≤ 2 0\leq\rho\leq2 0 ≤ ρ ≤ 2 separately at ℓ = 0 , 1 \ell=0,1 ℓ = 0 , 1 . These dimensionless momenta are held fixed as the background parameters vary. Figure 5 shows the oscillation frequencies and damping rates of the continuously tracked branches, with separate panels for the different spatial sectors.
Figure 5. Finite-momentum dispersion of the least damped level (the member with Re ϖ ≥ 0 \operatorname{Re}\boldsymbol{\varpi} \ge0 Re ϖ ≥ 0 ; the partner follows by ϖ → − ϖ ‾ \boldsymbol{\varpi} \to-\overline{\boldsymbol{\varpi}} ϖ → − ϖ ). Top row: Re ϖ \operatorname{Re}\boldsymbol{\varpi} Re ϖ ; bottom row: − Im ϖ -\operatorname{Im}\boldsymbol{\varpi} − Im ϖ . Columns: Nil transverse plane waves (162) , Nil central oscillator harmonics (164) at ℓ = 0 \ell=0 ℓ = 0 and ℓ = 1 \ell=1 ℓ = 1 , and Solv modes along the unscaled direction (166) . Solid, dashed, dotted and dash-dotted lines: ( z / z N , M 2 ) = ( 1 , 0 ) , ( 1 , 1 ) , ( 3 2 , 0 ) , ( 3 2 , 1 ) (z/z_{\mathrm N},\mathcal{M}^2)=(1,0),(1,1),(\frac{3}{2},0),(\frac{3}{2},1) ( z / z N , M 2 ) = ( 1 , 0 ) , ( 1 , 1 ) , ( 2 3 , 0 ) , ( 2 3 , 1 ) . The Solv curves obey the effective-mass identification M 2 ↦ M 2 + p ^ 2 \mathcal{M}^2\mapsto\mathcal{M}^2+\hat{p}^{\,2} M 2 ↦ M 2 + p ^ 2 exactly (verified to machine precision; Appendix § 10 ).
Validation. The collocation spectra are validated by three independent checks, collected in Appendix § 10 : comparison with the exact analytic spectrum at the z = S z=S z = S points of the three families, stability of the verified modes under changes of the collocation resolution, and an independent shooting calculation. The resolution filter that defines the verified modes is a selection criterion rather than an error estimate.
Interpretation. At fixed ( z , S , M 2 ) (z,S,\mathcal{M}^2) ( z , S , M 2 ) the homogeneous equation contains no additional dependence on the Lie algebra or supporting flux. Any agreement of dimensionless spectra at these matched parameters is therefore built into the probe equation. The numerical comparison tests how the allowed background parameters and probe mass change relaxation; it does not independently resolve electric or magnetic charge at fixed metric data. Finite spatial momentum and coupled perturbations are needed to probe further geometric and matter dependence.
The computed spectra show the following behavior (Figure 1 –Figure 5 ); the spectra, benchmark, and momentum results refer to the verified modes, while the scans in Figure 2 and Figure 3 track the raw N = 16 N=16 N = 16 spectra to keep the branches continuous through the pinch, and the anisotropy scan of Figure 4 tracks the raw N = 32 N=32 N = 32 spectra, since the overtones at h h h close to 1 require the higher resolution.
Figure 1. Homogeneous scalar spectra in the complex w \mathsf{w} w plane, one panel per family (left: Nil; right: Solv), showing all four comparison points ( M 2 , z / z N ) = ( 0 , 1 ) (M^2,z/z_{\mathrm{N}})=(0,1) ( M 2 , z / z N ) = ( 0 , 1 ) , ( 1 , 1 ) (1,1) ( 1 , 1 ) , ( 0 , 3 / 2 ) (0,3/2) ( 0 , 3/2 ) and ( 1 , 3 / 2 ) (1,3/2) ( 1 , 3/2 ) . Circles: M 2 = 0 M^2=0 M 2 = 0 ; squares: M 2 = 1 M^2=1 M 2 = 1 ; open markers: the neutral endpoint (z = z N z=z_{\mathrm{N}} z = z N ), filled markers: the electrically supported extension. Thin horizontal lines join the two members of each w ↔ − w ‾ \mathsf{w}\leftrightarrow-\overline{\mathsf{w}} w ↔ − w pair. Modes are filtered using N = 22 , 24 , 26 N=22,24,26 N = 22 , 24 , 26 with the same relative tolerance (6 × 10 − 2 6\times10^{-2} 6 × 1 0 − 2 ) and reported at N = 24 N=24 N = 24 ; the panel is restricted to the window ∣ Re w ∣ ≤ 5 |\operatorname{Re}\mathsf{w}|\leq5 ∣ Re w ∣ ≤ 5 , − Im w ≤ 5.8 -\operatorname{Im}\mathsf{w}\leq5.8 − Im w ≤ 5.8 . All numbers quoted in the text use N = 14 , 16 , 18 N=14,16,18 N = 14 , 16 , 18 .
Figure 2. Mass and dynamical-exponent scans of the least damped homogeneous scalar branches, for the neutral Nil, Solv and Bianchi VIh _h h (h = − 1 / 2 h = -1/2 h = − 1/2 ) families and their electric extensions. Left column: scan in M 2 M^2 M 2 at fixed z / z N = 3 / 2 z/z_{\mathrm{N}} = 3/2 z / z N = 3/2 ; right column: scan in z / z N z/z_{\mathrm{N}} z / z N at fixed M 2 = 1 M^2 = 1 M 2 = 1 . In each panel black curves show Re ω \operatorname{Re}\,\omega Re ω and gray curves − Im ω -\operatorname{Im}\,\omega − Im ω ; solid, dashed and dash-dotted lines are the fundamental and the first two overtones, tracked continuously through the scans. For each level the member with Re ω ≥ 0 \operatorname{Re}\,\omega\ge0 Re ω ≥ 0 is plotted; the partner follows by ω → − ω ˉ \omega\to-\bar{\omega} ω → − ω ˉ . Where the scan crosses z = S z = S z = S (Solv at z / z N = 2 z/z_{\mathrm{N}} = 2 z / z N = 2 , Bianchi VIh _h h at z / z N = 9 / 5 z/z_{\mathrm{N}} = 9/5 z / z N = 9/5 , Nil at z / z N = 8 / 3 z/z_{\mathrm{N}} = 8/3 z / z N = 8/3 ), the real parts of a pair meet and the modes continue as purely damped levels with split damping, in agreement with the analytic spectrum (149) .
Figure 3. Complex-plane paths of the least damped homogeneous scalar branches under the z / z N z/z_{\mathrm{N}} z / z N scan at fixed M 2 = 1 M^2 = 1 M 2 = 1 , for the neutral Nil, Solv and Bianchi VIh _h h (h = − 1 / 2 h = -1/2 h = − 1/2 ) families; the same data as the right column of Figure 2 . Each branch traces a path as z / z N z/z_{\mathrm{N}} z / z N increases from 1 (open circle) to 4 (filled dot); solid, dashed and dash-dotted lines are the fundamental and the first two overtones. The tracked path follows the member with Re ω ≥ 0 \operatorname{Re}\omega\ge0 Re ω ≥ 0 ; small dots along each path mark equal path-length intervals and do not indicate direction, the open circle marks the start of the scan (z / z N = 1 z/z_{\mathrm{N}} = 1 z / z N = 1 ) and the filled dot its end (z / z N = 4 z/z_{\mathrm{N}} = 4 z / z N = 4 ), and the mirror path of the ω → − ω ˉ \omega\to-\bar{\omega} ω → − ω ˉ partner is drawn as a thin light line. Along each path the real part decreases toward the pinch (Solv at z / z N = 2 z/z_{\mathrm{N}} = 2 z / z N = 2 , Bianchi VIh _h h at z / z N = 9 / 5 z/z_{\mathrm{N}} = 9/5 z / z N = 9/5 , Nil at z / z N = 8 / 3 z/z_{\mathrm{N}} = 8/3 z / z N = 8/3 ) while the damping rate grows, in agreement with the analytic spectrum (149) .
Figure 4. Anisotropy scan of the least damped homogeneous scalar branches of the Bianchi VIh _h h family, at fixed z / z N = 3 / 2 z/z_{\mathrm{N}} = 3/2 z / z N = 3/2 and M 2 = 1 M^2 = 1 M 2 = 1 with zero momentum, as h h h varies from − 1 -1 − 1 to 0.9; the endpoint h = − 1 h = -1 h = − 1 is the Solv point of the family, and h = 1 h = 1 h = 1 , where z N → ∞ z_{\mathrm{N}} \to\infty z N → ∞ and the family meets a separate Bianchi V branch, is excluded. Black curves show Re ω \operatorname{Re}\omega Re ω and gray curves − Im ω -\operatorname{Im}\omega − Im ω ; solid, dashed, dash-dotted and dotted lines are the fundamental and the first three overtones, tracked continuously through the scan from h = 0 h = 0 h = 0 in both directions. For the oscillatory branches the member with Re ω ≥ 0 \operatorname{Re}\omega\ge0 Re ω ≥ 0 is plotted; the partner follows by ω → − ω ˉ \omega\to-\bar{\omega} ω → − ω ˉ . The solid and dashed branches meet at the pinch h ≃ − 0.27 h \simeq-0.27 h ≃ − 0.27 , where the scan crosses the z = S z = S z = S point of the family, and the dashed branch terminates there; the kink in the dotted branch at h ≃ 0.28 h \simeq0.28 h ≃ 0.28 is the pinch where its oscillatory pair merges on the imaginary axis.
Mass dependence. At fixed z / z N = 3 / 2 z/z_{\mathrm N}=3/2 z / z N = 3/2 , raising M 2 \mathcal{M}^2 M 2 from 0 to 4 increases the damping of the least damped branch for all three families while the oscillation frequency changes mildly: for the fundamental, ( Re ϖ , − Im ϖ ) (\operatorname{Re}\boldsymbol{\varpi},-\operatorname{Im}\boldsymbol{\varpi}) ( Re ϖ , − Im ϖ ) moves from ( 1.09 , 1.87 ) (1.09,1.87) ( 1.09 , 1.87 ) to ( 1.19 , 2.09 ) (1.19,2.09) ( 1.19 , 2.09 ) for Nil, from ( 0.68 , 2.01 ) (0.68,2.01) ( 0.68 , 2.01 ) to ( 0.81 , 2.60 ) (0.81,2.60) ( 0.81 , 2.60 ) for Solv, and from ( 0.50 , 2.02 ) (0.50,2.02) ( 0.50 , 2.02 ) to ( 0.56 , 2.86 ) (0.56,2.86) ( 0.56 , 2.86 ) for Bianchi VIh _h h . The overtones behave in the same direction, so within this family a heavier probe relaxes faster; we do not extend this to a stability statement, since the BF-like bound (141) is only an asymptotic criterion.
Dynamical-exponent dependence. At fixed M 2 = 1 \mathcal{M}^2=1 M 2 = 1 , raising z / z N z/z_{\mathrm N} z / z N from 1 toward the point z = S z=S z = S of each family increases the damping substantially and, for Nil and Solv, reduces ∣ Re ϖ ∣ |\operatorname{Re}\boldsymbol{\varpi}| ∣ Re ϖ ∣ of the fundamental: from ( 1.54 , 1.56 ) (1.54,1.56) ( 1.54 , 1.56 ) at z / z N = 1 z/z_{\mathrm N}=1 z / z N = 1 to ( 0 , 2.32 ) (0,2.32) ( 0 , 2.32 ) at z / z N = 4 z/z_{\mathrm N}=4 z / z N = 4 for Nil, and from ( 1.36 , 2.01 ) (1.36,2.01) ( 1.36 , 2.01 ) to ( 0 , 2.118 ) (0,2.118) ( 0 , 2.118 ) for Solv. The Solv endpoint z / z N = 2 z/z_{\mathrm N}=2 z / z N = 2 is the point z = S = 2 z=S=2 z = S = 2 , and the fundamental becomes purely damped exactly at the analytic value ω 0 = − 2 i δ + / Δ = − 2.11803 i \omega_0=-2i\,\delta_+/\Delta=-2.11803i ω 0 = − 2 i δ + /Δ = − 2.11803 i ; the same pinch occurs for Bianchi VIh _h h at z / z N = 9 / 5 z/z_{\mathrm{N}}=9/5 z / z N = 9/5 and for Nil at z / z N = 8 / 3 z/z_{\mathrm{N}}=8/3 z / z N = 8/3 , where the two members of the least damped pair meet at Re ω = 0 \operatorname{Re}\omega=0 Re ω = 0 at the frequencies predicted by (149) (for Bianchi VIh _h h : ω = − 2.20184 i \omega=-2.20184i ω = − 2.20184 i , − 4.20184 i -4.20184i − 4.20184 i , … \ldots … ). This transition of a quasinormal pair from oscillatory to purely damped as a state parameter is varied, with the pair merging on the imaginary axis and splitting into two damped levels, is a familiar feature of quasinormal spectra, documented for Schwarzschild–de Sitter black holes [78 ] and for holographic Einstein–Maxwell–dilaton plasmas, where purely damped modes take over as the dominant equilibration channels [79 ] . Beyond the pinch the two modes remain purely damped with split damping that moves in opposite directions: for Bianchi VIh _h h at z / z N = 1.9 z/z_{\mathrm{N}}=1.9 z / z N = 1.9 one finds − Im ω = 1.94 -\operatorname{Im}\omega=1.94 − Im ω = 1.94 and 2.40 2.40 2.40 , for Solv at z / z N = 5 / 2 z/z_{\mathrm{N}}=5/2 z / z N = 5/2 the values 1.62 1.62 1.62 and 2.40 2.40 2.40 bracket the pre-pinch damping 2.01 2.01 2.01 , and for Nil at z / z N = 4 z/z_{\mathrm{N}}=4 z / z N = 4 the values 1.43 1.43 1.43 and 2.32 2.32 2.32 bracket the pinch value 2.03 2.03 2.03 . The tracked oscillation frequency approaches zero at z = S z=S z = S , and the branches were followed continuously through this crossing; across z = S z=S z = S each branch continues from the corresponding analytic frequency. The same scans are shown as paths in the complex plane in Figure 3 .
Anisotropy dependence. So far the spatial geometry was held fixed within each family. Figure 4 scans the anisotropy parameter of the Bianchi VIh _h h family over − 1 ≤ h ≤ 0.9 -1\le h\le0.9 − 1 ≤ h ≤ 0.9 at fixed z / z N = 3 / 2 z/z_{\mathrm{N}}=3/2 z / z N = 3/2 and M 2 = 1 M^2=1 M 2 = 1 , zero momentum; the family parameters S = 1 − h S=1-h S = 1 − h and z N = ( 1 + h 2 ) / ( 1 − h ) z_{\mathrm{N}}=(1+h^2)/(1-h) z N = ( 1 + h 2 ) / ( 1 − h ) then vary along the scan. The endpoint h = − 1 h=-1 h = − 1 is the Solv point of the family, and the fundamental found there, ω = ± 0.71 − 2.18 i \omega=\pm0.71-2.18i ω = ± 0.71 − 2.18 i , is the Solv value of Figure 1 ; h = 1 h=1 h = 1 , where z N → ∞ z_{\mathrm{N}}\to\infty z N → ∞ , is excluded. The z = S z=S z = S point of the family moves with h h h as z / z N = ( 1 − h ) 2 / ( 1 + h 2 ) z/z_{\mathrm{N}}=(1-h)^2/(1+h^2) z / z N = ( 1 − h ) 2 / ( 1 + h 2 ) and crosses the scan value 3 / 2 3/2 3/2 at h = 3 − 2 ≃ − 0.268 h=\sqrt{3}-2\simeq-0.268 h = 3 − 2 ≃ − 0.268 . For h h h above the crossing the scan point lies beyond the family’s pinch and the least damped levels are purely damped, with a damping that decreases monotonically with h h h : the fundamental moves from ω = − 1.62 i \omega=-1.62i ω = − 1.62 i at h = 0 h=0 h = 0 to − 1.00 i -1.00i − 1.00 i at h = 0.9 h=0.9 h = 0.9 , so at these fixed ratios the brane relaxes more slowly as h h h grows. At the crossing itself the fundamental pair pinches on the imaginary axis exactly at the analytic value (149) , ω = − 2.27358 i \omega=-2.27358i ω = − 2.27358 i (with z = S = 3 − 3 z=S=3-\sqrt{3} z = S = 3 − 3 ), and splits into two purely damped levels whose damping moves in opposite directions, bracketing the pinch value (− 1.97 i -1.97i − 1.97 i and − 2.53 i -2.53i − 2.53 i at h = − 0.2 h=-0.2 h = − 0.2 ); the tracked partner of the split pair terminates at the pinch. For h h h below the crossing the least damped modes are oscillatory, as at the Solv endpoint. The overtones remain oscillatory through the crossing and pinch successively at larger h h h (the dash-dotted branch at h ≃ − 0.24 h\simeq-0.24 h ≃ − 0.24 , the dotted branch at h ≃ 0.28 h\simeq0.28 h ≃ 0.28 ); the remaining oscillatory pairs of the window ∣ ω ∣ < 12 |\omega|<12 ∣ ω ∣ < 12 pinch at still larger h h h , the last ones at h ≃ 0.55 h\simeq0.55 h ≃ 0.55 , beyond which the entire window spectrum is purely damped. As h → 1 − h\to1^- h → 1 − the levels arrange themselves into the equally spaced ladder − Im ω = 1 , 2 , 3 , … -\operatorname{Im}\omega=1,2,3,\ldots − Im ω = 1 , 2 , 3 , … ; at h = 0.9 h=0.9 h = 0.9 the first four levels lie at 1.003 1.003 1.003 , 2.003 2.003 2.003 , 3.003 3.003 3.003 and 4.007 4.007 4.007 in these units.
Neutral endpoint versus electric extension. Comparing the neutral endpoint with the electric extension within a family (Figure 1 ), the extension changes the frequencies appreciably already at z / z N = 3 / 2 z/z_{\mathrm{N}}=3/2 z / z N = 3/2 ; since the homogeneous equation depends only on ( z , S , M 2 , K ) (z,S,M^2,\mathcal{K}) ( z , S , M 2 , K ) , this change is controlled by the shift of the dynamical exponent z z z rather than by the supporting flux as such.
Overtone structure and symmetry. The eigenvalues occur in pairs ω \omega ω and − ω ˉ -\bar{\omega} − ω ˉ , visible as the symmetric patterns of Figure 1 ; at the z = S z=S z = S points the pairs collapse onto the imaginary axis and the analytic spectrum (149) is equally spaced in damping, Δ η = 2 \Delta\eta=2 Δ η = 2 . Away from z = S z=S z = S the spacing is no longer uniform.
Momentum dependence. For a simple isolated transverse Nil mode, the potential depends analytically on k ⊥ 2 k_\perp^2 k ⊥ 2 , so the local expansion has the form ω ( k ⊥ ) = ω ( 0 ) + c 2 k ⊥ 2 + O ( k ⊥ 4 ) \omega(k_\perp)=\omega(0)+c_2k_\perp^2+O(k_\perp^4) ω ( k ⊥ ) = ω ( 0 ) + c 2 k ⊥ 2 + O ( k ⊥ 4 ) , with generally complex c 2 c_2 c 2 . This expansion need not apply at a mode degeneracy. In the central oscillator sector the potential instead contains a term linear in ρ ∝ ∣ k ω ∣ \rho\propto|k_\omega| ρ ∝ ∣ k ω ∣ at fixed ℓ \ell ℓ , together with a term quadratic in ρ \rho ρ ; its small-momentum dependence therefore differs from that of transverse plane waves. For the unscaled Solv direction, the dispersion is determined by the exact effective-mass replacement M 2 → M 2 + p ~ 2 M^2\to M^2+\tilde{p}^{\,2} M 2 → M 2 + p ~ 2 .
Conclusions and discussion
We have constructed neutral and matter-supported homogeneous anisotropic black branes in arbitrary dimension using the diagonal metric ansatz (6) , whose spatial coefficients have the form b i 2 r 2 α i b_i^2 r^{2\alpha_i} b i 2 r 2 α i . The spatial Lie algebra enters through scaling compatibility, intrinsic curvature, and the mixed Einstein equations. Together with the radial equations, these conditions determine which homogeneous geometries admit the ansatz. The neutral construction includes the negative-cosmological-constant branch with κ = 0 \kappa=0 κ = 0 and a Ricci-flat branch with κ ≠ 0 \kappa\ne0 κ = 0 , and relates the former to earlier homogeneous black holes and algebraic Ricci solitons. The Ricci-flat branch realizes the hyperscaling-violating vacuum geometries constructed in [40 , 44 ] .
In EMD theory, the electric, magnetic, and dyonic equations retain this separation between radial evolution and geometric constraints. Homogeneous magnetic fields must also satisfy the Maxwell equation and Bianchi identity, so an admissible magnetic orientation cannot be inferred from the Einstein equations alone. We gave electric extensions of five-dimensional Bianchi backgrounds, magnetic Bianchi I and Solv solutions, and a single-U ( 1 ) U(1) U ( 1 ) dyonic Nil solution. Higher-dimensional examples include flat dyonic geometries and a family obtained by adjoining a magnetic two-plane to a Ricci nilsoliton. In these solutions, the electromagnetic amplitudes are tied to the scaling background. Independently charged deformations require additional branches beyond those constructed here.
For f ( r ) = f 0 [ 1 − ( r h / r ) z + S ] f(r)=f_0[1-(r_{\mathrm h}/r)^{z+S}] f ( r ) = f 0 [ 1 − ( r h / r ) z + S ] , with f 0 , z , S > 0 f_0,z,S>0 f 0 , z , S > 0 , the horizon quantities give S B H ∝ T S / z S_{\mathrm{BH}}\propto T^{S/z} S BH ∝ T S / z . The Euclidean action at fixed supporting fluxes determines the relative free energy, and the associated mass agrees with the fixed-source Hamiltonian variation. The heat capacity is positive along this one-parameter family. These results do not establish stability under arbitrary source variations or a complete phase diagram. We also do not consider the interior structure of the anisotropic geometries; related static anisotropic black holes have recently been shown to exclude inner horizons and to develop spacelike Kasner-like singularities [80 ] . In particular, the electric potential and the exterior electromagnetic energies require a subtraction or renormalization prescription before they can be interpreted as finite boundary quantities.
The probe scalar gives a first view of wave propagation, dispersion, and relaxation on these backgrounds, as well as a restricted test of stability. Its homogeneous asymptotics give the BF-like bound M 2 ≥ − ( z + S ) 2 / 4 M^2\ge-(z+S)^2/4 M 2 ≥ − ( z + S ) 2 /4 , which constrains the asymptotic exponents but does not prove global stability. For fixed background parameters, homogeneous quasinormal frequencies are proportional to temperature. At z = S z=S z = S , with no logarithmic term in the blackening function, the exact spectrum is ω N = − i 4 π T ( N + δ + / Δ ) \omega_N=-i4\pi T(N+\delta_+/\Delta) ω N = − i 4 π T ( N + δ + /Δ ) . For real asymptotic exponents, the purely damped analytic spectrum (149) supports stability in the homogeneous scalar sector under the chosen boundary conditions and provides an analytic benchmark for the numerical calculation.
The numerical study extends the homogeneous spectrum to electric Nil, Solv, and Bianchi V I h \mathrm{VI}_h VI h backgrounds away from z = S z=S z = S . In the Nil, Solv, and Bianchi V I h \mathrm{VI}_h VI h scans, the tracked low-lying oscillatory branches meet on the imaginary axis near z = S z=S z = S and continue as purely damped levels. The damping increases with probe mass at the background parameters examined. An anisotropy scan of the Bianchi V I h \mathrm{VI}_h VI h family from its Solv point (h = − 1 h=-1 h = − 1 ) toward h = 1 h=1 h = 1 crosses the z = S z=S z = S point of the family, where the fundamental pair pinches exactly at the analytic frequency; beyond the crossing the least damped levels are purely damped with a damping that decreases as h h h grows, the oscillatory overtones pinch away successively, and the spectrum approaches an equally spaced purely damped ladder as h → 1 − h\to1^- h → 1 − . Finite-momentum calculations also distinguish the spatial geometries: transverse Nil modes have a potential quadratic in transverse momentum, central Nil harmonics involve an oscillator spectrum depending on the absolute central momentum, and momentum along the unscaled Solv direction gives an effective-mass shift. The analytic benchmark, resolution comparisons, and shooting calculation assess the numerical accuracy, while the higher overtones remain more sensitive to resolution. The numerical procedure selects damped branches and therefore cannot rule out growing scalar modes. These results describe scalar dispersion and relaxation in selected sectors, rather than a complete stability analysis or a determination of transport coefficients.
A natural next step is to study more general, coupled gravitational, dilaton, and U ( 1 ) U(1) U ( 1 ) gauge-field perturbations, including inhomogeneous sectors. These would test the dynamical stability of the backgrounds beyond the probe scalar and, where a suitable holographic description is available, reveal the dispersion of conserved-current and stress-tensor fluctuations. The master-function methods of [81 , 82 ] and the gravitational quasinormal-mode literature [83 ] provide useful tools where applicable. Such an analysis could also identify hydrodynamic modes and anisotropic transport coefficients.
Another direction is to relax the purely radial electric ansatz and allow transverse electric-field components along the homogeneous spatial directions. Solving their Maxwell and Bianchi constraints together with the mixed Einstein equations, possibly with a more general stationary metric, would allow us to test whether the enlarged ansatz admits branches carrying nonzero, independently variable electric charge. Such a charge must be identified through the radial electric flux; a transverse electric field does not by itself imply a nonzero conserved charge. A systematic classification of admissible spatial Lie algebras and magnetic two-forms could extend the families found here, while stationary metrics would allow rotation to be included, as in rotating Nil and Solv solutions [84 , 85 ] .
For holographic applications, two additional steps are needed: a renormalized variational problem allowing general boundary sources, as developed for Lifshitz asymptotics in [86 ] , and solutions connecting the homogeneous scaling regions to an appropriate ultraviolet geometry. These would provide the basis for calculating boundary stress tensors, comparing phases in specified ensembles, and studying anisotropic transport.
Radial integration and logarithmic cases
For completeness, consider the operator
P ( D ) : = ( D + 2 z ) ( D + Δ ) , D = r d d r . (167) P(D) := (D + 2z)(D + \Delta), \qquad D = r\frac{\mathrm{d}}{\mathrm{d}r}. \tag*{(167)} P ( D ) := ( D + 2 z ) ( D + Δ ) , D = r d r d . (167)
The neutral and EMD temporal equations have the common form P ( D ) f = ∑ s J s r s P(D)f = \sum_s J_s r^s P ( D ) f = ∑ s J s r s . Source terms with the same power must first be combined. For a single source the particular solution is
f s ( r ) = { J s r s / P ( s ) , P ( s ) ≠ 0 , J s r s log r / P ′ ( s ) , P ( s ) = 0 , P ′ ( s ) ≠ 0 , 1 2 J s r s ( log r ) 2 , P ( s ) = P ′ ( s ) = 0. (168) f_s(r) =
\begin{cases}
J_s r^s/P(s), & P(s) \ne0,\\
J_s r^s \log r/P'(s), & P(s) = 0,\ P'(s) \ne0,\\
\frac{1}{2}J_s r^s(\log r)^2, & P(s) = P'(s) = 0.
\end{cases} \tag*{(168)} f s ( r ) = ⎩ ⎨ ⎧ J s r s / P ( s ) , J s r s log r / P ′ ( s ) , 2 1 J s r s ( log r ) 2 , P ( s ) = 0 , P ( s ) = 0 , P ′ ( s ) = 0 , P ( s ) = P ′ ( s ) = 0. (168)
The homogeneous solutions are r − 2 z r^{-2z} r − 2 z and r − Δ r^{-\Delta} r − Δ when z ≠ S z \ne S z = S . For z = S z = S z = S the second homogeneous solution is r − 2 z log r r^{-2z}\log r r − 2 z log r . In the dyonic equations the source pairs ( s , J s ) (s,J_s) ( s , J s ) are
( 0 , − 4 Λ / n ) , ( − p , 2 ( n − 1 ) E / n ) , ( m B , 2 B / n ) . (169) (0,-4\Lambda/n), \qquad(-p,2(n-1)\mathcal{E}/n), \qquad(m_B,2\mathcal{B}/n). \tag*{(169)} ( 0 , − 4Λ/ n ) , ( − p , 2 ( n − 1 ) E / n ) , ( m B , 2 B / n ) . (169)
This covers every case in which a source power coincides with a homogeneous power of the temporal equation, including the case where the electric and magnetic source powers coincide with each other. After integrating, the spatial and dilaton equations must be imposed separately. They can remove an integration constant or rule out a putative logarithmic solution. This is particularly important in the Λ = 0 \Lambda= 0 Λ = 0 cases.
Fixed-source Hamiltonian variation
We briefly explain the relative mass used in § 5 . For the metric {(5.1)}, the radial logarithmic derivative of the volume of a constant-r r r spatial section is S / r S/r S / r . A variation of the blackening mode at fixed spatial scales and fixed EMD supporting sources gives the gravitational Hamiltonian surface variation
δ H t g r a v = − ν b S 2 r z + S δ f . (170) \delta H_t^{\mathrm{grav}}=-\frac{\nu_b S}{2}r^{z+S}\delta f. \tag*{(170)} δ H t grav = − 2 ν b S r z + S δ f . (170)
Using δ f = − r − Δ δ μ \delta f=-r^{-\Delta}\delta\mu δ f = − r − Δ δ μ gives δ M = ν b S δ μ / 2 \delta M=\nu_b S\delta\mu/2 δ M = ν b S δ μ /2 , consistent with (129) . The scalar contribution vanishes because δ ϕ = 0 \delta\phi=0 δ ϕ = 0 , and the electric variation vanishes in the fixed-flux ensemble. This derivation specifies the relative variation in the restricted family. It does not determine a universal mass for more general asymptotically homogeneous boundary conditions, nor does it provide the counterterms needed when the matter sources are allowed to vary.
Numerical convergence and reliability
This appendix documents the validation of the collocation spectra of § 6.5 . The accuracy is assessed by comparison with the analytic spectrum, by variation of the collocation resolution, and by an independent shooting calculation.
(i) Analytic benchmark at z = S z=S z = S . Each electric family admits a point z = S z=S z = S , where the spectrum (149) is exact. Table 2 compares the verified collocation modes with the analytic values at the three z = S z=S z = S points (Nil at z / z N = 8 / 3 z/z_{\mathrm{N}}=8/3 z / z N = 8/3 , Solv at z / z N = 2 z/z_{\mathrm{N}}=2 z / z N = 2 , Bianchi VIh _h h at z / z N = 9 / 5 z/z_{\mathrm{N}}=9/5 z / z N = 9/5 ) for M 2 = 0 , 1 , 4 M^2=0,1,4 M 2 = 0 , 1 , 4 . For each overtone, let ω N n u m \omega_N^{\mathrm{num}} ω N num be the numerical dimensionless frequency and define its damping rate by η N n u m : = − Im ω N n u m \eta_N^{\mathrm{num}}:=-\operatorname{Im}\omega_N^{\mathrm{num}} η N num := − Im ω N num . The exact damping rates are η N : = − Im ω N = 2 ( N + δ + / Δ ) \eta_N:=-\operatorname{Im}\omega_N=2(N+\delta_+/\Delta) η N := − Im ω N = 2 ( N + δ + /Δ ) . The damping-rate errors are at most 1.4 × 10 − 5 1.4\times10^{-5} 1.4 × 1 0 − 5 for the fundamental, 7.3 × 10 − 4 7.3\times10^{-4} 7.3 × 1 0 − 4 for the first overtone, and 1.2 × 10 − 2 1.2\times10^{-2} 1.2 × 1 0 − 2 for the second overtone; the deviations are displayed in Figure 6 . The table separately reports the maximum absolute error in the full complex frequency. The coincident gamma-function poles do not alter the spectrum in this range: for M 2 ≥ 0 M^2\geq0 M 2 ≥ 0 the numerator pole of the source coefficient (148) is of lower order than the denominator poles, so the zeros are unaffected; the massless case (M 2 = 0 M^2=0 M 2 = 0 ), where the coincident poles occur, is reproduced by the numerics at the level quoted above.
Table 2. Benchmark at the z = S z=S z = S points: verified collocation frequencies η N n u m \eta_N^{\mathrm{num}} η N num compared with the exact damping rates η N = 2 ( N + δ + / Δ ) \eta_N=2(N+\delta_+/\Delta) η N = 2 ( N + δ + /Δ ) of (149) . The reported mode is the value at N c o l = 16 N_{\mathrm{col}}=16 N col = 16 . The column “max dev.” is the maximum over levels N = 0 , 1 , 2 N=0,1,2 N = 0 , 1 , 2 of the absolute complex deviation ∣ ω n u m − ω e x a c t ∣ \lvert\omega^{\mathrm{num}}-\omega^{\mathrm{exact}}\rvert ∣ ω num − ω exact ∣ at the middle resolution, computed from the unrounded benchmark data and includes any residual real part of the numerical frequency. Deviations of the damping rates are ≤ 1.4 × 10 − 5 \leq1.4\times10^{-5} ≤ 1.4 × 1 0 − 5 (N = 0 N=0 N = 0 ), ≤ 7.3 × 10 − 4 \leq7.3\times10^{-4} ≤ 7.3 × 1 0 − 4 (N = 1 N=1 N = 1 ) and ≤ 1.2 × 10 − 2 \leq1.2\times10^{-2} ≤ 1.2 × 1 0 − 2 (N = 2 N=2 N = 2 ); all levels shown are verified by the filter.
Figure 6. Deviations of the verified collocation modes from the analytic spectrum (149) at the three z = S z=S z = S benchmark points (open circles: Nil; open squares: Solv; open triangles: Bianchi VIh _h h ; horizontally offset for readability). All levels shown are verified by the resolution comparison.
(ii) Resolution stability. Verified modes were required to be stable across the three resolutions used in the filter; at a representative non-analytic point (Nil electric, z / z N = 3 / 2 z/z_{\mathrm{N}}=3/2 z / z N = 3/2 , M 2 = 1 M^2=1 M 2 = 1 ) the verified frequencies agree across the resolution sets ( 12 , 14 , 16 ) (12,14,16) ( 12 , 14 , 16 ) , ( 14 , 16 , 18 ) (14,16,18) ( 14 , 16 , 18 ) , ( 16 , 18 , 20 ) (16,18,20) ( 16 , 18 , 20 ) and ( 20 , 22 , 24 ) (20,22,24) ( 20 , 22 , 24 ) to the quoted digits (identifying the two members of each ω ↔ − ω ˉ \omega\leftrightarrow-\bar{\omega} ω ↔ − ω ˉ pair, Table 3 ): the fundamental is stable to 3 × 10 − 6 3\times10^{-6} 3 × 1 0 − 6 across the sets, the first overtone to 1.4 × 10 − 3 1.4\times10^{-3} 1.4 × 1 0 − 3 (Table 3 ).
Table 3. Resolution stability of the verified modes at the Nil electric point z / z N = 3 / 2 z/z_{\rm N}=3/2 z / z N = 3/2 , M 2 = 1 M^2=1 M 2 = 1 : the positive-Re member of each ω ↔ − ω ˉ \omega\leftrightarrow-\bar{\omega} ω ↔ − ω ˉ pair at the four resolution sets used in the check (the partner follows by ω → − ω ˉ \omega\to-\bar{\omega} ω → − ω ˉ ). The fundamental varies by 3 × 10 − 6 3 \times10^{-6} 3 × 1 0 − 6 across the sets, the first overtone by 1.4 × 10 − 3 1.4 \times10^{-3} 1.4 × 1 0 − 3 .
The tracked branches in Figure 2 were also compared with the filtered spectra at representative points on both sides of z = S z=S z = S , using N c o l = 14 , 16 , 18 N_{\rm col}=14,16,18 N col = 14 , 16 , 18 . At coalescence and beyond, comparison with N c o l = 22 , 24 , 26 N_{\rm col}=22,24,26 N col = 22 , 24 , 26 gives agreement within 10 − 4 10^{-4} 1 0 − 4 for the fundamental. The pre-coalescence overtones remain resolution-limited at the tracking resolution. Adjacent to z = S z=S z = S , the tracked frequencies approach the analytic values in (149) within 2 × 10 − 5 2 \times10^{-5} 2 × 1 0 − 5 for the fundamental and 2 × 10 − 2 2 \times10^{-2} 2 × 1 0 − 2 for the overtones. These comparisons test resolution dependence and branch assignment within the collocation method. The branches in Figure 2 were followed continuously by proximity matching on the raw collocation spectra, with adaptive step refinement as the overtones rapidly approach the analytic spectrum (149) ; across z = S z=S z = S each branch continues from the corresponding analytic frequency.
(iii) Shooting comparison. A shooting calculation integrates the ingoing horizon Frobenius solution and matches it to the source-free boundary branch. Table 4 compares the imaginary parts of the fundamental frequencies. For the five entries other than the VIh _h h neutral endpoint, the differences calculated from the displayed values range from 1.0 × 10 − 5 1.0 \times10^{-5} 1.0 × 1 0 − 5 to 2.45 × 10 − 3 2.45 \times10^{-3} 2.45 × 1 0 − 3 . The shooting root search at the VIh _h h neutral endpoint did not satisfy its convergence tolerance; that entry is reported for reference and is excluded from the validation range. Agreement of the imaginary parts alone does not determine the error in the full complex frequency.
Table 4. Imaginary parts of the fundamental frequencies at M 2 = 1 M^{2}=1 M 2 = 1 from collocation and shooting. Negative imaginary parts indicate damping; the real parts are not displayed. The V I h \mathrm{VI}_{h} VI h neutral-endpoint shooting result (marked † \dagger † ) did not converge within the root-finding tolerance and is excluded from the accuracy comparison.
An additional exact check uses the Solv unscaled-direction sector: since its potential is constant, K 0 = p ^ 2 \mathcal{K}_0=\hat{p}^2 K 0 = p ^ 2 , the radial spectrum at ( p ^ , M 2 ) (\hat{p},M^2) ( p ^ , M 2 ) must coincide with the homogeneous spectrum at ( 0 , M 2 + p ^ 2 ) (0, M^{2}+\hat{p}^{2}) ( 0 , M 2 + p ^ 2 ) . The collocation modes satisfy this identification to machine precision for all tested combinations.