Yan Liu a Abstract We study shear fluctuations in non-relativistic holographic systems coupled to torsional Newton-Cartan geometry, using asymptotically Lifshitz spacetimes in Einstein-Maxwell-dilaton gravity. We identify a universal subdiffusive shear mode characterized by the quartic dispersion relation ω = − i ​ D 4 ​ k 4 \omega=-iD_{4}k^{4} , in sharp contrast to the conventional hydrodynamic diffusion. We derive this result analytically through a systematic higher-order matched asymptotic expansion connecting near-horizon and far-region solutions, and we verify it with direct numerical quasinormal mode calculations. Our numerics demonstrate that the first non-hydrodynamic mode is purely imaginary and gapped, following the dispersion relation ω = − i ​ ω 0 − i ​ D ​ k 2 \omega=-i\omega_{0}-iDk^{2} , and that both the hydrodynamic and the first non-hydrodynamic modes pass through pole-skipping points. These results highlight Lifshitz holography as an efficient framework for anomalous transport in strongly coupled non-relativistic quantum matter. 1 Introduction Diffusion is a fundamental physical process that transports particles, energy, or other quantities from regions of high concentration to low concentration, driving systems toward equilibrium [ 1 ] . Typically, it is described by Fick’s laws \omega=-iDk^{2} , which follows from conservation laws combined with a diffusive flux proportional to the concentration gradient. However, not all conserved quantities obey standard Fickian diffusion at late times. When microscopic dynamics involves additional symmetries, long-lived memory effects or ergodicity breaking, the macroscopic Fickian description breaks down. In such cases, relaxation is governed by subdiffusive hydrodynamic poles, leading to different scaling laws for the relaxation time. Subdiffusive transport plays a crucial role in understanding anomalous dynamical behaviors of many body systems. A canonical example arises in fluids with simultaneous conservation of charge and dipole moment, where enhanced symmetry constraints suppress microscopic kinetics, leading to subdiffusive charge transport [ 4 , 2 , 3 ] . Further hydrodynamic realizations, such as those in tilted lattice systems, further illustrate the generality of subdiffusive phenomena [ 5 ] . These unconventional subdiffusive dispersion relations signal exotic late-time dynamics in non-equilibrium systems, where relaxation rates follow unconventional scaling laws [ 6 ] . While most studies have focused on the subdiffusion of conserved charges, which typically exhibits longitudinal subdiffusive behaviors, — momentum, a conserved quantity that normally follows shear (viscous) diffusion in the transverse direction, can also exhibit slower subdiffusive scaling under sufficiently strong constraints. For example, in the presence of the magnetic field, the Lorentz force can balance the longitudinal pressure gradient against the transverse friction force, causing the net force density to vanish at leading order and thereby inducing subdiffusive momentum transport [ 7 ] . Compared to charge or energy transport, momentum dynamics is often more challenging to study directly, making holographic methods particularly valuable. In this work, we analyze non-relativistic strongly coupled field theories coupled to torsional Newton-Cartan (TNC) geometry using holography. Such field theories widely exist in condensed matter physics and we focus on a class of strongly coupled field theories dual to Lifshitz black holes [ 8 ] . Holographic duality is a powerful tool for studying strongly coupled systems by mapping them to a tractable weakly coupled gravitational system in a higher-dimensional curved spacetime [ 9 , 10 ] . In the asymptotically AdS case, i.e. the celebrated AdS/CFT correspondence, this duality naturally yields standard Fickian diffusion in the hydrodynamic limit [ 11 , 12 ] . This universality highlights the robustness of hydrodynamics but also reflects potential limitations arising from the relativistic UV structure. Realizing exceptional subdiffusion in holography is therefore both challenging and interesting. Subdiffusion has previously been achieved in certain holographic settings, including holographic magnetic hydrodynamics [ 7 , 13 ] , a phenomenological holographic bulk action including higher form fields [ 14 ] and holographic solids [ 15 ] . However, the complexity of these constructions motivates us to explore subdiffusion in a simpler non-relativistic holographic framework which is beyond the conventional AdS/CFT: Lifshitz holography. Lifshitz holography aims to describe the non-relativistic field theories, by introducing a Lifshitz scaling symmetry at the UV boundary [ 16 ] . On the gravity side, Lifshitz spacetime can be constructed in bottom-up approaches, notably via Einstein-Maxwell-dilaton (EMD) theory with a running scalar [ 17 ] , or Einstein-Proca theory with a massive vector field in the bulk [ 16 ] . Due to the breaking of Lorentz boost, the non-relativistic Newton-Cartan structure in the dual field theory is involved. It turns out that the dual field theory Lifshitz black hole is a non-relativistic field theory on a fixed torsional Newton-Cartan geometry — a natural generalization of Newton–Cartan geometry that allows for torsion and correctly encodes the symmetries of non-relativistic field theories coupled to curved backgrounds [ 21 , 20 ] . The definition of the non-relativistic stress-energy tensor complex in Lifshitz duality was proposed from the action principle based on vielbein fields [ 18 ] . Important progress on Lifshitz holography can be found in e.g. [ 22 , 23 , 24 , 25 , 26 , 27 , 28 ] . We exploit Lifshitz holography to study the shear sector of the hydrodynamic response from Einstein-Maxwell dilaton theory. Using linear response theory, we systematically extract the low-energy, long-wavelength excitations and extract the structure of the hydrodynamic pole associated with transverse momentum fluctuations from quasinormal modes in the bulk. The consistent metric and gauge field fluctuations in the bulk are dual to perturbations on the TNC boundary geometry. We employ a combination of analytical matching method (near-horizon and far-region matching) and numerical method to obtain the dispersion relation of the hydrodynamic pole in the shear channel. The organization of this paper is as follows. In Sec. 2 , we briefly introduce the EMD model, and its connection with the effective field theory defined in the Newton-Cartan geometry. Sec. 3 is the main part of this work. In Sec. 3.1 , we do the transverse fluctuation to study the shear dynamics of momentum, and demonstrate the shear subdiffusion of momentum in Sec. 3.2 by the matching method. In Sec. 3.3 , we perform the numerical computations of the lowest quasi-normal modes to confirm our analytical results and show the property of the subdiffusive constant. Finally, in Sec. 4 , we conclude for this study and discuss some open questions for the future. 2 The holographic model and the generating functional We begin by outlining the bulk Einstein-Maxwell-dilaton theory which supports asymptotically Lifshitz black hole solutions. Then we discuss the generating functional on the boundary geometry, which takes the form of a torsional Newton-Cartan spacetime. 2.1 The Einstein-Maxwell-dilaton model We start from the Einstein-Maxwell-dilaton theory [ 17 ] that is described by the action in the bulk S=\frac{1}{16\pi G}\int d^{d+2}x\sqrt{-g}\,\bigg[R-2\Lambda-\frac{1}{4}e^{\lambda\phi}F^{2}-\frac{1}{2}(\nabla\phi)^{2}\bigg]\,, (2.1) where the cosmological constant is \Lambda=-\frac{1}{2}(z+d)(z+d-1)\,. (2.2) The gauge field A_{a} couples to a dilaton field \phi with a special chosen of interaction where the field strength F_{ab}=\partial_{a}A_{b}-\partial_{b}A_{a} . 3 3 3 We use the index a,b for the fields in the bulk, and the index \mu,\nu for fields at the boundary. As a result, it deforms the UV boundary to be asymptotic to a Lifshitzian spacetime, where the parameter \displaystyle\lambda=\sqrt{\frac{2d}{z-1}}\, (2.3) is a function of the Lifshitz exponent z with z>1 . It is obvious that with different choice of \lambda , we have different Lifshitz exponent. Here d is the spatial dimension of the dual field theory, and we will specify to d=2 in the following when we consider the hydrodynamics in the dual field theory. Extension to other dimensions is straightforward. We set 16\pi G=1 without loss of generality. The equations of motion can be obtained after performing the variation, \begin{split}R_{ab}-\frac{1}{2}e^{\lambda\phi}F_{ac}F_{b}^{~c}-\frac{1}{2}\partial_{a}\phi\partial_{b}\phi&=\frac{1}{2}g_{ab}\left[R-2\Lambda-\frac{1}{4}e^{\lambda\phi}F^{2}-\frac{1}{2}(\nabla\phi)^{2}\right]\,,\\ \nabla_{a}(e^{\lambda\phi}F^{ab})&=0\,,\\ \nabla^{2}\phi-\frac{\lambda}{4}e^{\lambda\phi}F^{2}&=0\,.\end{split} (2.4) This model has a series of analytic solutions that dual to the finite temperature and finite (particle) density many-body systems, with the line element written as \displaystyle ds^{2}=-r^{2z}f(r)dt^{2}+\frac{dr^{2}}{r^{2}f(r)}+r^{2}d{\vec{x}_{d}}^{2}\,, (2.5) where \displaystyle\begin{split}f(r)=1-\left(\frac{r_{h}}{r}\right)^{z+d}\,,\quad A_{t}=\sqrt{\frac{2(z-1)}{z+d}}\left(r^{z+d}-r_{h}^{z+d}\right)\,,\end{split} (2.6) and \begin{split}e^{\lambda\phi}&=r^{-2d}\,.\\ \end{split} (2.7) The temperature of the black hole is related to the horizon r_{h} as T=\frac{z+d}{4\pi}\,r_{h}^{z}\,. (2.8) Remind that for a pure Lishitz spacetime with Lifshitz symmetry, the geometry should be invariant under the Lifshitz scaling transformation \displaystyle t\rightarrow c^{z}t\,,\quad x^{i}\rightarrow cx^{i}\,,\quad r\rightarrow c^{-1}r\, (2.9) with z>1 as the Lifshitz exponent. However, in this work, we consider the finite temperature and finite mass density systems, temperature and the matter fields together with the scalar field break the Lifshitz symmetry as the relevant deformations. The Lifshitz symmetry Eq.( 2.9 ) naturally breaks the Lorentz boost symmetry between the time and spatial directions. Consequently, the low-energy dynamics of the dual systems should not respect the relativistic symmetry anymore, and the description is replaced by the Galilean invariant non-relativistic field theories. Nevertheless, nonrelativistic diffeomorphism invariance still constrains the low-energy dynamics for the non-relativistic quantum systems [ 29 , 30 ] , especially in the strongly-coupled quantum systems with a holographic dual as our consideration. 2.2 The generating functional In a relativistic quantum field theory, the energy-momentum tensor T^{\mu\nu} and the conserved U(1) current J^{\mu} couple to the external background metric \gamma_{\mu\nu} and gauge field A_{\mu} , respectively. In general, we can always include a scalar operator \mathcal{O} that couples to external scalar source \phi . Therefore, the low-energy dynamics are captured by the generating functional W[\gamma_{\mu\nu}\,,A_{\mu}\,,\phi] , with the variation \delta W=\int d^{3}x\sqrt{-\gamma}\big(T^{\mu\nu}\delta\gamma_{\mu\nu}+J^{\mu}\delta A_{\mu}+\mathcal{O}\delta\phi\big)\,. (2.10) For the non-relativistic theory, there is no underlying non-degenerate metric in the non-relativistic spacetime. It is described by a Newton-Cartan geometry [ 31 , 32 , 33 ] , a (d+1) -dimensional manifold equiped with (n_{\mu},\sigma_{\mu\nu},a_{\mu}) , i.e. a no-where vanishing one-form n_{\mu} which defines local time direction and a symmetric spatial tensor \sigma_{\mu\nu} of rank- d which gives the metric on spatial slices, and a U(1) connection a_{\mu} whose field strength leads to a unique definition of the covariant derivative, satisfying \displaystyle\gamma_{\mu\nu}=n_{\mu}n_{\nu}+\sigma_{\mu\nu}\,, (2.11) with rank- (d+1) and invertible \gamma^{\mu\nu} . Then a velocity field v^{\mu} and an inverse metric \sigma^{\mu\nu} can be straightforwardly defined via v^{\mu}n_{\mu}=1\,,~~~\sigma_{\mu\nu}v^{\mu}=0\,,~~~\sigma^{\mu\nu}n_{\nu}=0\,,~~~\sigma^{\mu\alpha}\sigma_{\nu\alpha}=\delta^{\mu}_{\nu}-v^{\mu}n_{\nu}\,. (2.12) The one form n_{\mu} effectively define a local time direction and \sigma_{\mu\nu} is a metric on spatial slices. Equivalently, (v^{\mu},\sigma^{\mu\nu}) satisfy v^{\mu}=\gamma^{\mu\nu}n_{\nu}\,,~~~\sigma^{\mu\nu}=\gamma^{\mu\nu}-v^{\mu}v^{\nu}\,. (2.13) In torsionless Newton-Cartan geometry the one form n_{\mu} is exact, i.e. dn=0 . Otherwise it is torsional Newton-Cartan geometry with dn\neq 0 which is the case we focused on. Along with the reduction of the geometry, the energy-momentum tensor is decomposed to the energy density E , the energy flux E^{i} , the momentum density P^{i} , and the symmetric stress tensor \pi^{ij} with only the spatial components, as well a number current J^{\mu} . The generating functional is therefore expressed by the Newton-Cartan data W[n_{\mu},v^{\mu},\sigma^{\mu\nu},a_{\mu},\phi_{0}] , with the variation \delta W=\int d^{3}x\sqrt{-\gamma}\left(J^{\mu}\delta a_{\mu}-P_{\mu}\delta{\bar{v}}^{\mu}-E^{\mu}\delta n_{\mu}-\frac{1}{2}\pi_{\mu\nu}\delta{\bar{\sigma}}^{\mu\nu}+\mathcal{O}\delta\phi_{0}\right)\,, (2.14) where \delta{\bar{v}}^{\mu} is the free variation part of \delta v^{\mu} excluding terms involving \delta n_{\mu} . Imposing the U(1) gauge invariance leads to the particle number conservation \partial_{t}n+\partial_{i}J^{i}=0 . In the special system we studied, it turns out \delta a_{y} is not indepedent of \delta n_{y} which means that we only have an independent \delta a_{t} while there is no independent source for J^{i} [ 8 ] . In the absence of boost invariance, imposing the time and spatial translational symmetry, the energy and momentum are independently conserved (assuming vanishing scalar source) \displaystyle\partial_{t}E+\partial_{i}E^{i}=0\,,\quad\partial_{t}P_{j}+\partial_{i}\pi^{i}_{~j}=0\,. (2.15) 3 Transverse fluctuation and the shear diffusion In this section we perform holographic calculations for the transverse fluctuations. First, we identify the correct gauge invariants with independent sources. Subsequently, we use the near-far matching method to derive the dispersion relation of the hydrodynamic diffusive mode. We also numerically compute the quasinormal modes of the system and compare with the analytical results. 3.1 The linear transverse perturbations To investigate the dynamics of the momentum, we switch on the linear fluctuations of the metric and matter fields in the shear channel. In general, the linear fluctuations of the metric, gauge field and the scalar field are the following \begin{split}\delta g_{ab}&=\int\frac{d\omega dk}{(2\pi)^{2}}e^{-i\omega t+ikx}h_{ab}(r,\omega,k)\,,\\ \delta A_{a}&=\int\frac{d\omega dk}{(2\pi)^{2}}e^{-i\omega t+ikx}a_{a}(r,\omega,k)\,,\\ \delta\phi&=\int\frac{d\omega dk}{(2\pi)^{2}}e^{-i\omega t+ikx}\psi(r,\omega,k)\,,\\ \end{split} (3.1) after the Fourier transformations. These fluctuations can be classified by parity under y\to-y . Consider the parity-odd fluctuations \{h_{ty},h_{xy},a_{y}\} around the background ( 2.7 ) in the radial gauge \delta g_{ri}=\delta a_{r}=0 , they follow the EoMs \displaystyle h_{ty}^{\prime\prime}-\frac{z-1}{r}h_{ty}^{\prime}+\frac{-k^{2}+2r^{2}(z-2)f}{r^{4}f}h_{ty}-\frac{k\omega}{r^{4}f}h_{xy}+e^{\lambda\phi}A_{t}^{\prime}a_{y}^{\prime} \displaystyle=0\,, (3.2a) \displaystyle h_{xy}^{\prime\prime}+\frac{(z-1)f+rf^{\prime}}{rf}h_{xy}^{\prime}+\frac{\omega^{2}-2r^{2z}f(zf+rf^{\prime})}{r^{2(z+1)}f^{2}}h_{xy}+\frac{k\omega}{r^{2(z+1)}f^{2}}h_{ty} \displaystyle=0\,, (3.2b) \displaystyle a_{y}^{\prime\prime}+\frac{rf^{\prime}+f(1+z+r\lambda\phi^{\prime})}{rf}a_{y}^{\prime}+\frac{r^{2(1-z)}\omega^{2}-k^{2}f}{r^{4}f^{2}}a_{y}+\frac{A_{t}^{\prime}}{r^{2z}f}h_{ty}^{\prime}-\frac{2A_{t}^{\prime}}{r^{1+2z}f}h_{ty} \displaystyle=0\,, (3.2c) \displaystyle\omega rh_{ty}^{\prime}-2\omega h_{ty}+kr^{2z-1}fh_{xy}^{\prime}-2kr^{2(z-1)}fh_{xy}+\omega re^{\lambda\phi}A_{t}^{\prime}a_{y} \displaystyle=0\,. (3.2d) The EoMs of the parity-odd sector are decoupled from the parity-even sector including \{h_{tt},h_{tx},h_{xx},h_{yy},a_{t},a_{x},\delta\phi\} . With the EoMs of the fluctuations, the standard approach to compute the quasi-normal modes of the system is to solve the ODEs and impose the infalling boundary conditions near the horizon and the vanishing Dirichilet boundary conditions at the UV Lifshitz boundary. As for the AdS scenarios, choosing the gauge invariant variables can make the computations simpler. The gauge invariant variables are combinations of the fluctuations that are invariant under diffeomorphism and gauge transformations. To be concrete, under an infinitesimal diffeomorphism generated by the vector field \xi_{\nu} and an infinitesimal gauge transformation generated by scalar field \lambda , the metric and gauge field fluctuations transform as \begin{split}\delta g_{ab}&\to\delta g_{ab}-\nabla_{a}\xi_{b}-\nabla_{b}\xi_{a}\,,\\ \delta A_{a}&\to\delta A_{a}+\nabla_{a}\lambda-\xi^{b}\nabla_{b}A_{a}-A_{b}\nabla_{a}\xi^{b},\\ \delta\phi&\to\delta\phi-\xi^{a}\nabla_{a}\phi\,,\\ \end{split} (3.3) which constrains the number of independent fields. Therefore, there are only two independent gauge invariant variables in the transverse channel, defined as Z_{1}\equiv\omega h_{xy}+kh_{ty}\,,\quad Z_{2}\equiv a_{y}\,. (3.4) The form of these two gauge invariants is a common choice in the AdS references or the special Lifshitz spacetime with z=1 . However, this choice is problematic when we generalize to the Lifshitz spacetime with z>1 , since the leading order solution of Z_{1} and Z_{2} are linearly dependent where \begin{split}Z_{1}&=r^{2z}\left(Z_{1}^{(0)}+Z_{1}^{(1)}r^{-z-2}+...\right)+r^{2}Z_{3}^{(0)},\\ Z_{2}&=r^{z+2}\left(Z_{2}^{(0)}+Z_{2}^{(1)}r^{-z-2}+...\right),\\ \end{split} (3.5) with Z_{1}^{(0)}=-k\sqrt{\frac{z+2}{2(z-1)}}r^{z-2}Z_{2}^{(0)}\,. (3.6) This is a special property in Lifshitz holography, which is different from the AdS cases. We will show that this is because the energy flux E_{i} and the momentum density P_{i} are independent in the non-relativistic system, while their dynamics are mixed in the fluctuation h_{ty} . Via the holographic dictionary, the leading order solutions Z_{1}^{(0)} and Z_{2}^{(0)} are the external sources in the field theory, while the sub-leading solutions Z_{1}^{(1)} and Z_{2}^{(1)} are the operators that couple to the external sources. In order to extract another independent source that is represented by Z_{3}^{(0)} , we make a linear combination of Z_{1} and Z_{2} to define a new gauge invariant Z_{3}=\omega h_{xy}+kh_{ty}+k\,\sqrt{\frac{z+2}{2(z-1)}}\,r^{z-2}\,a_{y}\,, (3.7) with the asymptotic behaviors of Z_{3} as \begin{split}Z_{3}&=r^{2}Z_{3}^{(0)}+...\,.\\ \end{split} (3.8) As a result, we remove the leading order contribution of Z_{1}^{(0)} and now the source Z_{3}^{(0)} is dominant as the source in Z_{3} . Now we have a consistent choice of gauge invariant variables, i.e. \{Z_{2}\,,Z_{3}\} for further studies of quasi-normal modes as well as the hydrodynamics in this system. However, before the solving the ODEs and read out the hydrodynamic poles, it is necessary to explain the physical meaning of the variables and make a connection with the non-relativistic hydrodynamics. In the non-relativistic systems where the boost symmetry is broken, the momentum density P_{y} is not equal to the energy flux E^{y} and they couple to independent source \hat{e}^{(y)}_{~~t} and \hat{e}^{(t)}_{~~y} , as shown in ( 2.14 ). This can be made clear from frame fields defined from Lifshitz geometry [ 18 , 19 ] . We use e^{(0)},e^{(i)} to construct the frame fields in the bulk, while we use \hat{e}^{(0)},\hat{e}^{(i)} with a hat to construct the frame fields in the boundary field theory. The bulk metric can be expressed as ds^{2}=g_{ab}dx^{a}dx^{b}\equiv(e^{(r)})^{2}+\eta_{\mu\nu}e^{(\mu)}e^{(\nu)} in terms of the bulk frame fields e^{(\mu)} . Then the frame field \hat{e}^{\mu} at the boundary can be determined from \displaystyle e^{(0)}=r^{z}\sqrt{f}\hat{e}^{(0)}\,,~~~~e^{(i)}=r\hat{e}^{(i)}\,, (3.9) in the limit r\to\infty . For the background bulk solution Eq. ( 2.5 ) without any fluctuations, the frame fields are defined as \displaystyle e^{(0)}=e^{(0)}_{~~\mu}dx^{\mu}=r^{z}\sqrt{f}dt\,,~~~e^{(i)}=e^{(i)}_{~~\mu}dx^{\mu}=r\left(dx+dy\right) (3.10) which defines the vielbeins \displaystyle e^{(0)}_{~~t}=r^{z}\sqrt{f}\,,~~~e^{(0)}_{~~x}=e^{(0)}_{~~y}=0\,,~~~e^{(i)}_{~~t}=0\,,~~~e^{(i)}_{~~j}=r\delta_{ij}\,. (3.11) When we switch on the transverse sector fluctuations h_{ty},h_{xy},a_{y} , we follow [ 18 ] to reparametrize the fluctuations as 4 4 4 The fluid/gravity duality for five dimensional asymptotically Lifshitz black hole was explored in [ 27 ] , where a Galilean boost was applied to construct perturbative solution. In contrast to the fluctuations considered in the present work, the analysis in [ 27 ] did not include nonzero and non-homogeneous e^{(t)}_{~~y} that sources the energy flux E^{y} . These components, however, plays a crucial role in generating the full TNC structure of the dual boundary field theory [ 18 , 19 , 20 , 21 ] . Consequently, the study [ 27 ] effectively captured only the standard (torsion-free) Newton–Cartan geometry. It would be interesting to perform a proper fluid/gravity duality for our case and make a comparison to those derived here. \displaystyle h_{ty}=-r^{2z}v_{1}+r^{2}v_{2}\,,~~~h_{xy}=r^{2}v_{3}\,,~~~a_{y}=\sqrt{\frac{2(z-1)}{z+2}}r^{z+2}v_{1}\,. (3.12) In Lifshitz holography, the shear stress is encoded in the gravitational fluctuation h_{xy} as for the AdS scenarios. However, the independent external sources time shift and spatial shift are mixed in the gravitational fluctuation h_{ty} and the different conformal dimensions add further complications. Furthermore, note that the variation of a_{\mu} is linked to h_{ty} . This implies that the gauge transformation of the transverse gauge field is not independent of the time shift, which constitutes a constraint on the torsional Newton-Cartan background. This special feature puts a constraint on J^{y} and E^{y} in ( 2.14 ). It is convenient and equivalent to make the following variation in the frame fields as \begin{split}e^{(0)}&=r^{z}\hat{e}^{(0)}=r^{z}\left[dt+v_{1}dy\right]\,,\\ e^{(i)}&=r\hat{e}^{(i)}=r\left[v_{2}dt+\left(\delta^{i}_{~j}+v_{3}\bar{\delta}^{i}_{~j}\right)dx^{j}\right]\,,\end{split} (3.13) where v_{1},v_{2},v_{3} are functions of r,t,x and \bar{\delta}^{i}_{~j}\equiv 1-\delta^{i}_{~j} . When r\to\infty , the leading order of v_{i}~(i=1,2,3) are constants of \mathcal{O}(r^{0}) . The asymptotic solutions near the boundary for 10) are functions of \nu_{a,b}^{(0)} and can be solved order by order. As long as \omega and k are sufficiently small, these series solutions are valid from the horizon to some UV cutoff, which defines the inner region. 3.2.2 The outer solutions We can solve the equations iteratively in the outer region of the bulk, where the frequency \omega and momentum k are small quantities compared to the radial coordinate r \frac{\omega^{2}}{r^{2z}f^{2}}\ll 1\,,~~~~\frac{k^{2}}{r^{2}f}\ll 1\,. (3.30) One can confirm that for finite \omega and k , \frac{\omega^{2}}{r^{2z}f^{2}} and \frac{k^{2}}{r^{2}f} are negligible in the ODEs as r approaches the UV boundary. As r approaches the IR horizon, these two terms can still remain small in the hydrodynamic limit, which extends the valid range of the far-region that is possible to share overlap with the near-region. Notice that the contributions containing these infinitesimal small quantities are additive and linearly independent in Eq. ( 3.22 ). This indicates that it is not necessary to assume the order relations between these two terms. On the contrary, the scaling relation between \omega and k can be determined when we solve the equations in the full bulk. In other words, the pole condition can be extracted by imposing a specific relation on their respective orders within the matching region. 5 5 5 This differs from the assumption used in [ 34 ] where it is assumed that the frequency \omega and momentum k are of same order from the beginning. Together with an additional assumption that the two variables are of same order \mathcal{H}\sim a_{y} , whereas the correct relation should be \mathcal{H}\sim qa_{y} , that approach [ 34 ] leads to a different structure in the leading-order compared to ( 3.34 ). In our calculation, the orders of \omega and k are initially treated as independent, and their relation is determined through the matching approach in the hydrodynamic limit. The existence of the two small quantities ( 3.30 ) makes it possible to construct the solutions iteratively in the far-region. Taking the first equation in Eq. ( 3.22 ) as an example, in this region the coefficients of the non-derivative terms that are proportional to V_{a} and V_{b} are small quantities and therefore the non-derivative terms in the second line are negligible compared to the total derivative terms in the first line. We use similar observations to simplify Eq. ( 3.22 ) into two coupled ODEs involving only total derivatives. \begin{split}0&=\frac{d}{dr}\Big[r^{1+z}f\Big(\left(z-4\right)r^{2z}f^{2}\Big(\frac{V_{a}^{(0)}}{f}\Big)^{\prime}+\left(z+2\right)r^{2}V_{b}^{(0)^{\prime}}\Big)\Big]\,,\\ 0&=\frac{d}{dr}\Big[\frac{r^{z+3}f}{r^{2}\omega^{2}-k^{2}r^{2z}f}\left(r^{2z}V_{a}^{(0)^{\prime}}-r^{2}V_{b}^{(0)^{\prime}}\right)\Big]\,.\end{split} (3.31) Here we have used the label ``(0)" to represent the solutions to the zeroth order. After integration once over the radial direction r , we obtain \begin{split}\left(z-4\right)r^{2z}f^{2}\Big(\frac{V_{a}^{(0)}}{f}\Big)^{\prime}+\left(z+2\right)r^{2}V_{b}^{(0)^{\prime}}&=\frac{(\pi_{a}-\pi_{b})\omega^{2}}{r^{1+z}f}\,,\\ \frac{r^{z+3}f}{r^{2}\omega^{2}-k^{2}r^{2z}f}\left(r^{2z}V_{a}^{(0)^{\prime}}-r^{2}V_{b}^{(0)^{\prime}}\right)&=\frac{\pi_{b}}{z+2}\,,\end{split} (3.32) where \pi_{a},\pi_{b} are introduced as integration constants that correspond to operators in the effective action after some linear combinations, and they depend on frequency \omega and momentum k . It is simple to decouple V_{a}^{(0)} to obtain a first-order ODE of V_{a}^{(0)} , and the solution of V_{a}^{(0)} is expressed as an integral \displaystyle\begin{split}V_{a}^{(0)}(r)=&\left(z+2+(z-4)f\right)\,\Big(\sigma_{a}+\int_{\infty}^{r}d\rho\frac{\pi_{a}\rho^{2}\omega^{2}-\pi_{b}k^{2}\rho^{2z}f}{\rho^{3z+3}f\left(z+2+(z-4)f\right)^{2}}\Big)\,,\end{split} (3.33) where \sigma_{a} is the integration constant that corresponds to the external source in the field theory. After obtaining the solution of V_{a}^{(0)} , we can obtain V_{b}^{(0)} as \displaystyle V_{b}^{(0)}(r)=\sigma_{b}+\int_{\infty}^{r}d\rho\Big[\pi_{b}\frac{k^{2}\rho^{2z}f-\rho^{2}\omega^{2}}{\rho^{z+5}(z+2)f}+\frac{V_{a}^{(0)^{\prime}}}{\rho^{2-2z}}\Big]\,, (3.34) where \sigma_{b} is the integration constant that corresponds to another external source. Here the asymptotic expansion of the above form is consistent with the expansion in ( 3.21 ) up to a constant coefficient in front of \sigma_{a} . The UV expansions of V_{a,b}^{(0)} are consistent with Eq. ( 3.21 ). One can check the integrals are convergent in the UV for z<4 . However, we will encounter divergence when we consider higher order corrections, as discussed in Eq. ( 3.38 ) for z\geq 2 . Before our calculations of the ODEs to the next order, it is an appropriate time to make a comparison between the non-relativistic hydrodynamics and the relativistic hydrodynamics. In relativistic hydrodynamics, calculations to this order are sufficient to determine the hydrodynamic pole structure [ 35 ] . However, as demonstrated in Subsection 3.2.3 , a higher-order analysis is crucial for obtaining the correct pole structure in our case. This parallels the findings of recent studies on low-temperature matching for RN black holes [ 36 ] . For latter convenience in calculations, we assume V_{a,b}=V_{a,b}^{(0)}+V_{a,b}^{(c)}+\cdots, (3.35) where V_{a,b}^{(c)} represent the next-to-leading order corrections in terms of parameters ( 3.30 ) to V_{a,b}^{(0)} . The dots represent higher order of \epsilon corrections that are not important in this work. V_{a}^{(0)} and V_{b}^{(0)} play the role of the source terms in the inhomogeneous ODEs of V_{a,b}^{(c)} as follows \begin{split}r^{1+z}f\Big(\left(z-4\right)r^{2z}f^{2}\Big(\frac{V_{a}^{(c)}}{f}\Big)^{\prime}+\left(z+2\right)r^{2}V_{b}^{(c)^{\prime}}\Big)&=-h(r)\,,\\ \frac{r^{z+3}f}{r^{2}\omega^{2}-k^{2}r^{2z}f}\left(r^{2z}V_{a}^{(c)^{\prime}}-r^{2}V_{b}^{(c)^{\prime}}\right)&=-g(r)\,,\end{split} (3.36) where \begin{split}h(r)\equiv&\int_{r_{h}}^{r}dl\frac{l^{2}\omega^{2}-k^{2}l^{2z}f}{l^{z+3}f}\Big[\left(z-4\right)l^{2z}fV_{a}^{(0)}+\left(z+2\right)l^{2}V_{b}^{(0)}\Big]\,,\\ g(r)\equiv&\int_{r_{h}}^{r}dl\frac{l^{2z}V_{a}^{(0)}-l^{2}V_{b}^{(0)}}{l^{z+1}f}\,.\end{split} (3.37) The lower bound of integration is chosen so that the subleading solution remains regular in the matching region. The solutions of V_{a,b}^{(c)} contain the homogeneous part and the inhomogeneous part. The homogeneous part represent the higher order corrections in \omega and k to the integration constants \sigma_{a,b} and \pi_{a,b} , and can be absorbed into V_{a,b}^{(0)} . As a result, we can expand \sigma_{a,b} and \pi_{a,b} order by order in \omega and k . The inhomogeneous part of the correction order solution can be expressed as, \begin{split}V_{a}^{(c)}(r)=&\left(z+2+(z-4)f\right)\,\int_{\infty}^{r}d\rho\frac{\left(z+2\right)\left(k^{2}\rho^{2z}f-\rho^{2}\omega^{2}\right)g(\rho)-\rho^{2}h(\rho)}{\rho^{3z+3}f\left(z+2+(z-4)f\right)^{2}}\,,\\ V_{b}^{(c)}(r)=&\int_{\infty}^{r}d\rho\Big[\frac{\rho^{2}\omega^{2}-k^{2}\rho^{2z}f}{\rho^{z+5}f}g(\rho)+\frac{V_{a}^{(c)^{\prime}}}{\rho^{2-2z}}\Big]\,.\end{split} (3.38) These integrals contribute to the corrections terms in the asymptotic UV expansions. They are convergent only if 11 ) at the ultraviolet fixed point naturally breaks Lorentz boost invariance, yielding a boundary field theory whose geometry is described by torsional Newton–Cartan spacetime. This setup provides a powerful holographic laboratory for exploring non-relativistic hydrodynamics. Focusing on the shear sector, we analyzed the dynamics of momentum in the hydrodynamic limit. Unlike relativistic hydrodynamics, where energy flux and momentum density are tightly coupled and shear diffusion is universal ( \omega=-iDk^{2} ), non-relativistic systems treat the energy flux E_{i} and momentum density P_{i} as independent observables that both couple to the shear stress tensor \pi_{ij} . This additional structure fundamentally modifies the low-energy pole structure. Our key result is the breakdown of standard shear diffusion in favor of subdiffusive momentum transport characterized by the quartic dispersion relation \omega=-iD_{4}k^{4} , where D_{4} is a subdiffusive constant that depends on an integral over the entire bulk geometry and z (see Eq. ( 3.51 )). Remarkably, this quartic scaling is universal for all 1