Introduction
Entanglement dynamics following a global quantum quench provides a useful probe of nonequilibrium phenomena in quantum many-body systems [1 , 2 ] . In integrable theories, the spreading of entanglement can often be understood in terms of entangled quasiparticle pairs [3 ] . This picture, however, does not generally capture the dynamics of strongly coupled systems, which can instead be studied using the gauge/gravity correspondence [4 , 5 , 6 ] . In this setting, the time evolution of entanglement entropy admits a complementary geometric description. In particular, for time-dependent states, the Hubeny–Rangamani–Takayanagi (HRT) prescription relates the entanglement entropy of a boundary region to the area of a bulk extremal surface [7 , 8 ] . A homogeneous global quench is commonly modeled holographically by an infalling null shell in a Vaidya geometry, which interpolates between an initial state and a final black-brane background [9 , 10 , 11 , 12 , 13 ] . In holographic theories, this setup typically gives rise to early-time quadratic growth, an intermediate linear-growth regime, and eventual saturation for sufficiently large regions [14 , 15 , 16 , 17 , 18 ] . The linear regime is associated with extremal surfaces approaching a critical configuration behind the final horizon and is characterized by the entanglement velocity.
In this work, we investigate how these features are modified by spatial noncommutativity. We consider four-dimensional N = 4 \mathcal{N}=4 N = 4 noncommutative Yang–Mills (NCYM) theory, with noncommutativity in the ( x 2 , x 3 ) (x_2,x_3) ( x 2 , x 3 ) plane, [ x 2 , x 3 ] = i θ [x_2,x_3]=i\theta [ x 2 , x 3 ] = i θ . Such theories arise from the worldvolume dynamics of D-branes in a background NS–NS B B B -field [19 ] and exhibit nonlocality and ultraviolet/infrared (UV/IR) mixing [20 , 21 , 22 ] . At large N N N and strong ’t Hooft coupling, their holographic description is provided by the noncommutative D3-brane geometry [23 , 24 ] . The geometry is anisotropic and, in the noncommutative directions, is not asymptotically AdS. Consequently, the entanglement structure depends on the orientation of the boundary region. In particular, for a strip whose finite direction lies along a noncommutative direction, the competing extremal-surface configurations lead to a cutoff-dependent characteristic length scale associated with UV/IR mixing [25 , 26 , 27 , 28 ] . Noncommutative effects on entanglement dynamics and holographic thermalization have also been studied in [29 ] .
We extend these investigations by studying the time evolution of entanglement entropy and mixed-state correlations in a noncommutative Vaidya geometry. We focus on strips whose finite direction is along x 2 x_2 x 2 , and model the quench by an infalling null shell that interpolates between zero- and finite-temperature NCYM backgrounds, keeping the noncommutative scale fixed. Because of the non-AdS asymptotics in the noncommutative directions, we work with a finite radial cutoff and account for the competing extremal-surface branches that arise in this geometry. Our analysis addresses the early-time and intermediate-time scaling of the entanglement entropy, the critical surfaces responsible for linear growth, and the approach to thermal equilibrium. We also examine the resulting dynamics of holographic mutual information and the entanglement wedge cross section (EWCS) for two disjoint strips.
Our results reveal several characteristic effects of noncommutativity. First, the near-boundary behavior of the noncommutative geometry changes the early-time growth of the entanglement entropy: in the regime where the shell-crossing point satisfies z c ≪ a z_c \ll a z c ≪ a , the leading time dependence is logarithmic, with a coefficient proportional to the injected energy density and to a 2 a^2 a 2 . When the shell-crossing point lies beyond the noncommutative scale, a ≪ z c a \ll z_c a ≪ z c , the leading behavior instead becomes quadratic. Thus, the early-time scaling depends on the relation between the time-dependent shell-crossing scale and the noncommutative scale, rather than being universally quadratic.
For sufficiently wide strips, we identify a returning branch of shell-crossing extremal surfaces that approaches a critical configuration in the final black-brane geometry. This branch gives rise to an intermediate linear-growth regime, with an entanglement velocity that increases monotonically with the noncommutative scale and reduces to the AdS5 _5 5 value in the commutative limit. At late times, a branch that connects continuously to the equilibrium extremal surface has a large-strip saturation time controlled by the butterfly velocity along the noncommutative direction. However, competing extremal-surface branches can affect the physical saturation time, which must be determined by comparing their areas.
We further show that the leading, width-independent logarithmic contribution to the early-time entanglement entropy cancels in the connected holographic mutual information of two disjoint strips. Its subsequent evolution is governed by the relative saturation times of the three relevant strip entropies and, in the linear-growth regime, by the same entanglement velocity. Finally, the near-critical connected EWCS exhibits linear growth with a coefficient governed by this velocity. This result applies when the connected entanglement wedge is dominant and the corresponding cross section is the relevant minimal candidate.
The paper is organized as follows. In § 2 , we introduce the noncommutative Vaidya geometry and the HRT setup. § 3 examines the static extremal-surface configurations. In § 4 , we study the time evolution of the entanglement entropy, including its early-time behavior, critical surfaces, linear growth, and saturation. § 5 is devoted to the dynamics of mutual information and the EWCS. Finally, § 6 contains our conclusions and a discussion of the physical implications of our results, along with possible directions for future work.
Setup
We begin by reviewing the noncommutative D3-brane background and its Vaidya extension, followed by the derivation of the HRT equations for a strip whose finite direction lies along a noncommutative direction.
Noncommutative D3-brane background
We consider the bosonic sector of ten-dimensional type IIB supergravity, governed by the following action [30 , 31 , 32 ]
S I I B = 1 2 κ 10 2 ∫ d 10 x − g [ e − 2 Φ ( R ( 10 ) + 4 ( ∂ Φ ) 2 − 1 12 H 3 2 ) − 1 2 F 1 2 − 1 12 F ~ 3 2 − 1 4 ⋅ 5 ! F ~ 5 2 ] + 1 4 κ 10 2 ∫ C 4 ∧ H 3 ∧ F 3 . (1) S_{\mathrm{IIB}} = \frac{1}{2\kappa_{10}^{2}} \int d^{10}x \sqrt{-g} \left[e^{-2\Phi}\left(R^{(10)} + 4(\partial\Phi)^2 - \frac{1}{12}H_3^2\right) - \frac{1}{2}F_1^2 - \frac{1}{12}\widetilde{F}_3^2 - \frac{1}{4\cdot5!}\widetilde{F}_5^2\right] + \frac{1}{4\kappa_{10}^{2}}\int C_4 \wedge H_3 \wedge F_3 . \tag*{(1)} S IIB = 2 κ 10 2 1 ∫ d 10 x − g [ e − 2Φ ( R ( 10 ) + 4 ( ∂ Φ ) 2 − 12 1 H 3 2 ) − 2 1 F 1 2 − 12 1 F 3 2 − 4 ⋅ 5 ! 1 F 5 2 ] + 4 κ 10 2 1 ∫ C 4 ∧ H 3 ∧ F 3 . (1)
Here R ( 10 ) R^{(10)} R ( 10 ) denotes the ten-dimensional Ricci scalar, with 2 κ 10 2 = 16 π G N ( 10 ) 2\kappa_{10}^{2} = 16\pi G_{N}^{(10)} 2 κ 10 2 = 16 π G N ( 10 ) . We work in the axion-free sector, C 0 = F 1 = 0 C_0 = F_1 = 0 C 0 = F 1 = 0 , so that F ~ 3 = F 3 \widetilde{F}_3 = F_3 F 3 = F 3 , and impose the self-duality condition F ~ 5 = ⋆ F ~ 5 \widetilde{F}_5 = \star\widetilde{F}_5 F 5 = ⋆ F 5 after varying the action. The field-strength conventions and equations of motion are given in Appendix § 8 . To accommodate the infalling matter in the Vaidya geometry, we introduce an external source in the string-frame Einstein equation, which takes the form
R μ ν + 2 ∇ μ ∇ ν Φ − 1 4 H μ ρ σ H ν ρ σ − e 2 Φ 4 ( F μ ρ σ F ν ρ σ + 1 24 F ~ μ α β γ δ F ~ ν α β γ δ − 1 6 g μ ν F 3 2 ) = T μ ν e x t . (2) R_{\mu\nu} + 2\nabla_{\mu}\nabla_{\nu}\Phi- \frac{1}{4}H_{\mu\rho\sigma}H_{\nu}^{\ \rho\sigma} - \frac{e^{2\Phi}}{4}\left(F_{\mu\rho\sigma}F_{\nu}^{\ \rho\sigma} + \frac{1}{24}\widetilde{F}_{\mu\alpha\beta\gamma\delta}\widetilde{F}_{\nu}^{\ \alpha\beta\gamma\delta} - \frac{1}{6}g_{\mu\nu}F_3^2\right) = T_{\mu\nu}^{\mathrm{ext}} . \tag*{(2)} R μν + 2 ∇ μ ∇ ν Φ − 4 1 H μ ρ σ H ν ρ σ − 4 e 2Φ ( F μ ρ σ F ν ρ σ + 24 1 F μα β γ δ F ν α β γ δ − 6 1 g μν F 3 2 ) = T μν ext . (2)
The external source T μ ν e x t T_{\mu\nu}^{\mathrm{ext}} T μν ext vanishes in the static backgrounds. The dilaton profile is e 2 Φ ( z ) = g s 2 h ( z ) e^{2\Phi(z)} = g_s^2 h(z) e 2Φ ( z ) = g s 2 h ( z ) , where g s g_s g s is the string coupling. The corresponding finite-temperature noncommutative D3-brane metrics in the string frame are given by [23 , 24 ] 1
d s 2 = R 2 z 2 [ − f ( z ) d t 2 + d x 1 2 + h ( z ) ( d x 2 2 + d x 3 2 ) + d z 2 f ( z ) + z 2 d Ω 5 2 ] , (3) ds^2 = \frac{R^2}{z^2}\left[-f(z)dt^2 + dx_1^2 + h(z)\left(dx_2^2 + dx_3^2\right) + \frac{dz^2}{f(z)} + z^2d\Omega_5^2\right] , \tag*{(3)} d s 2 = z 2 R 2 [ − f ( z ) d t 2 + d x 1 2 + h ( z ) ( d x 2 2 + d x 3 2 ) + f ( z ) d z 2 + z 2 d Ω 5 2 ] , (3)
where the blackening function and the warp factor in the noncommutative plane are
f ( z ) = 1 − ( z z h ) 4 , h ( z ) = z 4 z 4 + a 4 . (4) f(z) = 1 - \left(\frac{z}{z_h}\right)^4,\qquad h(z) = \frac{z^4}{z^4+a^4}. \tag*{(4)} f ( z ) = 1 − ( z h z ) 4 , h ( z ) = z 4 + a 4 z 4 . (4)
Here R R R is the curvature radius, and z h z_h z h denotes the horizon radius. The noncommutative scale a a a is related to θ \theta θ through [23 , 24 ]
a = λ 1 / 4 θ , (5) a = \lambda^{1/4}\sqrt{\theta}, \tag*{(5)} a = λ 1/4 θ , (5)
where λ = g Y M 2 N \lambda= g_{\mathrm{YM}}^2N λ = g YM 2 N is the ’t Hooft coupling. In the commutative limit a → 0 a \to0 a → 0 , the warp factor satisfies h ( z ) → 1 h(z) \to1 h ( z ) → 1 , whereas for finite a a a , its near-boundary behavior is h ( z ) ≃ z 4 / a 4 h(z) \simeq z^4/a^4 h ( z ) ≃ z 4 / a 4 . Consequently, the closed-string metric is not asymptotically AdS in the noncommutative directions. We impose a finite radial cutoff z = δ z = \delta z = δ throughout the analysis [25 , 26 , 27 ] . The Hawking temperature, thermal entropy density, and energy density are given by [24 ]
T = 1 π z h , s t h = R 3 4 G N ( 5 ) z h 3 , E = 3 R 3 16 π G N ( 5 ) z h 4 . (6) T = \frac{1}{\pi z_h},\qquad s_{\mathrm{th}} = \frac{R^3}{4G_N^{(5)}z_h^3},\qquad\mathcal{E} = \frac{3R^3}{16\pi G_N^{(5)}z_h^4}. \tag*{(6)} T = π z h 1 , s th = 4 G N ( 5 ) z h 3 R 3 , E = 16 π G N ( 5 ) z h 4 3 R 3 . (6)
In the following, we express the strip entanglement entropy in units of N ~ = L 2 R 3 / ( 4 G N ( 5 ) ) \widetilde{\mathcal{N}} = L^2R^3/(4G_N^{(5)}) N = L 2 R 3 / ( 4 G N ( 5 ) ) . The corresponding normalization of the thermodynamic densities is specified by Eq. (6) .
Noncommutative Vaidya geometry
We model the global quench by an infalling null shell interpolating between the zero- and finite-temperature noncommutative backgrounds, while keeping the noncommutative scale a a a fixed. In each static region, we introduce an ingoing Eddington–Finkelstein coordinate through
d v = d t − d z f ( z ) , (7) dv = dt - \frac{dz}{f(z)}, \tag*{(7)} d v = d t − f ( z ) d z , (7)
so that the metric becomes
d s 2 = R 2 z 2 [ − f ( v , z ) d v 2 − 2 d v d z + d x 1 2 + h ( z ) ( d x 2 2 + d x 3 2 ) + z 2 d Ω 5 2 ] , (8) ds^2 = \frac{R^2}{z^2}\left[-f(v,z)\,dv^2 - 2\,dv\,dz + dx_1^2 + h(z)\left(dx_2^2+dx_3^2\right) + z^2d\Omega_5^2\right], \tag*{(8)} d s 2 = z 2 R 2 [ − f ( v , z ) d v 2 − 2 d v d z + d x 1 2 + h ( z ) ( d x 2 2 + d x 3 2 ) + z 2 d Ω 5 2 ] , (8)
where
f ( v , z ) = 1 − m ( v ) z 4 . (9) f(v,z) = 1 - m(v)z^4. \tag*{(9)} f ( v , z ) = 1 − m ( v ) z 4 . (9)
The time dependence enters solely through the blackening function,2 while the dilaton and form fields retain their static profiles. In the thin-shell limit, the Vaidya mass function m ( v ) m(v) m ( v ) is given by
m ( v ) = Θ ( v ) z h 4 . (10) m(v) = \frac{\Theta(v)}{z_h^4}. \tag*{(10)} m ( v ) = z h 4 Θ ( v ) . (10)
Here Θ ( v ) \Theta(v) Θ ( v ) is the Heaviside step function. The two static regions are joined across the null shell at v = 0 v = 0 v = 0 , with the time coordinate normalized such that v = t v = t v = t at the asymptotic boundary. For v < 0 v < 0 v < 0 , the geometry reduces to the zero-temperature noncommutative background, with m ( v ) = 0 m(v) = 0 m ( v ) = 0 and f = 1 f = 1 f = 1 . Integrating (7) gives
v = t − z . (11) v = t - z. \tag*{(11)} v = t − z . (11)
For v > 0 v>0 v > 0 , the geometry describes the finite-temperature noncommutative black brane, with
f ( z ) = 1 − g ( z ) , g ( z ) ≡ ( z z h ) 4 (12) f(z)=1-g(z),\qquad g(z)\equiv\left(\frac{z}{z_h}\right)^4 \tag*{(12)} f ( z ) = 1 − g ( z ) , g ( z ) ≡ ( z h z ) 4 (12)
In the exterior region z < z h z<z_h z < z h , the static and Eddington–Finkelstein time coordinates are related by
v = t − ∫ 0 z d ζ f ( ζ ) . (13) v=t-\int_{0}^{z}\frac{d\zeta}{f(\zeta)}. \tag*{(13)} v = t − ∫ 0 z f ( ζ ) d ζ . (13)
Across the ingoing null shell at v = 0 v=0 v = 0 , the external stress tensor associated with a general mass profile takes the form
T μ ν e x t = ϱ ( v , z ) k μ k ν , k μ d x μ = d v , ϱ ( v , z ) = 3 2 z 3 ∂ v m ( v ) , (14) T^{\mathrm{ext}}_{\mu\nu}=\varrho(v,z)k_\mu k_\nu,\qquad k_\mu dx^\mu=dv,\qquad\varrho(v,z)=\frac{3}{2}z^3\partial_v m(v), \tag*{(14)} T μν ext = ϱ ( v , z ) k μ k ν , k μ d x μ = d v , ϱ ( v , z ) = 2 3 z 3 ∂ v m ( v ) , (14)
where k μ k^\mu k μ is null. Thus, for the thin-shell profile, the only nonvanishing component is
T v v e x t = 3 z 3 2 z h 4 δ ( v ) , (15) T^{\mathrm{ext}}_{vv}=\frac{3z^3}{2z_h^4}\delta(v), \tag*{(15)} T v v ext = 2 z h 4 3 z 3 δ ( v ) , (15)
where δ ( v ) \delta(v) δ ( v ) is the Dirac delta distribution. For the numerical analysis, we use the smooth profile
m ( v ) = m 0 2 [ 1 + tanh ( v v 0 ) ] , (16) m(v)=\frac{m_0}{2}\left[1+\tanh\left(\frac{v}{v_0}\right)\right], \tag*{(16)} m ( v ) = 2 m 0 [ 1 + tanh ( v 0 v ) ] , (16)
where m 0 = 1 / z h 4 m_0=1/z_h^4 m 0 = 1/ z h 4 and v 0 > 0 v_0>0 v 0 > 0 controls the shell thickness. The thin-shell profile is recovered in the limit v 0 → 0 v_0\to0 v 0 → 0 .
Extremal surfaces and holographic entanglement entropy
We consider a strip of width ℓ \ell ℓ along the noncommutative direction x 2 x_2 x 2 , with transverse extent L ≫ ℓ L\gg\ell L ≫ ℓ , defined by
− ℓ 2 ≤ x 2 ≡ x ≤ ℓ 2 , − L 2 ≤ x 1 , x 3 ≤ L 2 . (17) -\frac{\ell}{2}\le x_2\equiv x\le\frac{\ell}{2},\qquad-\frac{L}{2}\le x_1,x_3\le\frac{L}{2}. \tag*{(17)} − 2 ℓ ≤ x 2 ≡ x ≤ 2 ℓ , − 2 L ≤ x 1 , x 3 ≤ 2 L . (17)
In the string frame, the holographic entanglement entropy functional includes the dilaton factor [33 ]
S = 1 4 G N ( 10 ) ∫ Γ ℓ d 8 σ e − 2 Φ det g i n d ( S ) . (18) S=\frac{1}{4G_N^{(10)}}\int_{\Gamma_\ell}d^8\sigma\,e^{-2\Phi}\sqrt{\det g_{\mathrm{ind}}^{(S)}}. \tag*{(18)} S = 4 G N ( 10 ) 1 ∫ Γ ℓ d 8 σ e − 2Φ det g ind ( S ) . (18)
Parametrizing the HRT hypersurface by v = v ( x ) v=v(x) v = v ( x ) and z = z ( x ) z=z(x) z = z ( x ) , the area functional takes the form
S = N ~ ∫ − ℓ / 2 ℓ / 2 d x 1 z 3 Q h ( z ) . (19) S=\widetilde{\mathcal{N}}\int_{-\ell/2}^{\ell/2}dx\,\frac{1}{z^3}\sqrt{\frac{Q}{h(z)}}. \tag*{(19)} S = N ∫ − ℓ /2 ℓ /2 d x z 3 1 h ( z ) Q . (19)
where
Q = h ( z ) − f ( v , z ) v ′ 2 − 2 v ′ z ′ . (20) Q=h(z)-f(v,z)v^{\prime2}-2v'z'. \tag*{(20)} Q = h ( z ) − f ( v , z ) v ′2 − 2 v ′ z ′ . (20)
and ′ ≡ d d x \prime\equiv\frac{d}{dx} ′ ≡ d x d . Henceforth, unless stated otherwise, we set the overall normalization factor N ~ = 1 \widetilde{\mathcal{N}}=1 N = 1 to unity. Denoting the tip of the extremal hypersurface by ( z t , v t ) (z_t,v_t) ( z t , v t ) , we impose the following boundary conditions
z ( ℓ 2 ) = δ , v ( ℓ 2 ) = t , z ( 0 ) = z t , v ( 0 ) = v t , z ′ ( 0 ) = v ′ ( 0 ) = 0. (21) z\left(\frac{\ell}{2}\right)=\delta,\qquad v\left(\frac{\ell}{2}\right)=t,\qquad z(0)=z_t,\qquad v(0)=v_t,\qquad z^\prime(0)=v^\prime(0)=0. \tag*{(21)} z ( 2 ℓ ) = δ , v ( 2 ℓ ) = t , z ( 0 ) = z t , v ( 0 ) = v t , z ′ ( 0 ) = v ′ ( 0 ) = 0. (21)
Extremizing the entropy functional (19) yields the equations of motion
z 3 Q h ∂ x [ z ′ + f v ′ z 3 Q h ] = 1 2 ∂ v f v ′ 2 , (22) z^3\sqrt{Qh}\,\partial_x\left[\frac{z^\prime+fv^\prime}{z^3\sqrt{Qh}}\right]=\frac{1}{2}\partial_v f\,v^{\prime2}, \tag*{(22)} z 3 Q h ∂ x [ z 3 Q h z ′ + f v ′ ] = 2 1 ∂ v f v ′2 , (22)
z 3 Q h ∂ x [ v ′ z 3 Q h ] = 3 Q z + 1 2 ∂ z f v ′ 2 + ∂ z h 2 ( Q h − 1 ) . (23) z^3\sqrt{Qh}\,\partial_x\left[\frac{v^\prime}{z^3\sqrt{Qh}}\right]=\frac{3Q}{z}+\frac{1}{2}\partial_z f\,v^{\prime2}+\frac{\partial_z h}{2}\left(\frac{Q}{h}-1\right). \tag*{(23)} z 3 Q h ∂ x [ z 3 Q h v ′ ] = z 3 Q + 2 1 ∂ z f v ′2 + 2 ∂ z h ( h Q − 1 ) . (23)
The absence of explicit x x x -dependence in the Lagrangian leads to the first integral
z 3 Q h ( z ) = z t 3 . (24) z^3\sqrt{\frac{Q}{h(z)}}=z_t^3. \tag*{(24)} z 3 h ( z ) Q = z t 3 . (24)
In each static region, where ∂ v f = 0 \partial_v f=0 ∂ v f = 0 , an additional conserved quantity follows from Eq. (22)
z ′ + f v ′ h ( z ) = E . (25) \frac{z^\prime+fv^\prime}{h(z)}=E. \tag*{(25)} h ( z ) z ′ + f v ′ = E . (25)
In the thin-shell limit, E E E is conserved separately in the two static regions, with its values related by the shell-matching conditions. For the smooth profile Eq. (16) , Eqs. (22) and (23) are solved numerically. We now consider the thin-shell solutions and their matching across v = 0 v=0 v = 0 .
(i) Vacuum region (v < 0 v<0 v < 0 ). The tip conditions (21) imply E − = 0 E_-=0 E − = 0 . Eqs. (24) and (25) then yield
v ′ = − z ′ , z ′ = − h ( z ) z 3 z t 6 − z 6 . (26) v^\prime=-z^\prime,\qquad z^\prime=-\frac{\sqrt{h(z)}}{z^3}\sqrt{z_t^6-z^6}. \tag*{(26)} v ′ = − z ′ , z ′ = − z 3 h ( z ) z t 6 − z 6 . (26)
Choosing the negative root in Eq. (26) for the branch extending from the tip toward the cutoff, we obtain
x ( z ) = ∫ z z t d y y 3 h ( y ) z t 6 − y 6 , v ( z ) = v t + z t − z . (27) x(z)=\int_z^{z_t}dy\,\frac{y^3}{\sqrt{h(y)}\sqrt{z_t^6-y^6}},\qquad v(z)=v_t+z_t-z. \tag*{(27)} x ( z ) = ∫ z z t d y h ( y ) z t 6 − y 6 y 3 , v ( z ) = v t + z t − z . (27)
If the surface intersects the shell at z = z c z=z_c z = z c , the condition v ( z c ) = 0 v(z_c)=0 v ( z c ) = 0 yields
z c = z t + v t . (28) z_c=z_t+v_t. \tag*{(28)} z c = z t + v t . (28)
(ii) Matching at the shell (v = 0 v=0 v = 0 ). The embedding remains continuous across the shell. Integrating (22) and (23) across v = 0 v=0 v = 0 yields the matching conditions
v + ′ = v − ′ , Q + = Q − , z + ′ = [ 1 − g ( z c ) 2 ] z − ′ . (29) v^\prime_+=v^\prime_-,\qquad Q_+=Q_-,\qquad z^\prime_+=\left[1-\frac{g(z_c)}{2}\right]z^\prime_-. \tag*{(29)} v + ′ = v − ′ , Q + = Q − , z + ′ = [ 1 − 2 g ( z c ) ] z − ′ . (29)
The subscripts − - − and + + + denote the vacuum and black-brane sides, respectively. While the first integral (24) is preserved across the shell, the conserved quantity (25) on the black-brane side becomes
E + = g ( z c ) 2 h ( z c ) z − ′ . (30) E_+=\frac{g(z_c)}{2h(z_c)}z^\prime_-. \tag*{(30)} E + = 2 h ( z c ) g ( z c ) z − ′ . (30)
(iii) Black-brane region (v > 0 v>0 v > 0 ). Combining the first integrals (24) and (25) yields
v ′ = h ( z ) E − z ′ f ( z ) , z ′ 2 = h ( z ) f ( z ) ( z t 6 z 6 − 1 ) + h ( z ) 2 E 2 ≡ H ( z ) . (31) v'=\frac{h(z)E-z'}{f(z)},\qquad z'^{2}=h(z)f(z)\left(\frac{z_{t}^{6}}{z^{6}}-1\right)+h(z)^{2}E^{2}\equiv H(z). \tag*{(31)} v ′ = f ( z ) h ( z ) E − z ′ , z ′ 2 = h ( z ) f ( z ) ( z 6 z t 6 − 1 ) + h ( z ) 2 E 2 ≡ H ( z ) . (31)
For the monotonic branch, z ′ = − H ( z ) z'=-\sqrt{H(z)} z ′ = − H ( z ) , giving
d v d z = − 1 f ( z ) [ 1 + h ( z ) E H ( z ) ] . (32) \frac{dv}{dz}=-\frac{1}{f(z)}\left[1+\frac{h(z)E}{\sqrt{H(z)}}\right]. \tag*{(32)} d z d v = − f ( z ) 1 [ 1 + H ( z ) h ( z ) E ] . (32)
The strip width, boundary time, and entanglement entropy are then given by
ℓ 2 = ∫ z c z t d z h ( z ) z t 6 / z 6 − 1 + ∫ δ z c d z H ( z ) , (33) \frac{\ell}{2}=\int_{z_{c}}^{z_{t}}\frac{dz}{\sqrt{h(z)}\sqrt{z_{t}^{6}/z^{6}-1}}+\int_{\delta}^{z_{c}}\frac{dz}{\sqrt{H(z)}}, \tag*{(33)} 2 ℓ = ∫ z c z t h ( z ) z t 6 / z 6 − 1 d z + ∫ δ z c H ( z ) d z , (33)
t = ∫ δ z c d z f ( z ) [ 1 + h ( z ) E H ( z ) ] , (34) t=\int_{\delta}^{z_{c}}\frac{dz}{f(z)}\left[1+\frac{h(z)E}{\sqrt{H(z)}}\right], \tag*{(34)} t = ∫ δ z c f ( z ) d z [ 1 + H ( z ) h ( z ) E ] , (34)
S = 2 z t 3 [ ∫ z c z t d z z 3 h ( z ) z t 6 − z 6 + ∫ δ z c d z z 6 H ( z ) ] . (35) S=2z_{t}^{3}\left[\int_{z_{c}}^{z_{t}}\frac{dz}{z^{3}\sqrt{h(z)}\sqrt{z_{t}^{6}-z^{6}}}+\int_{\delta}^{z_{c}}\frac{dz}{z^{6}\sqrt{H(z)}}\right]. \tag*{(35)} S = 2 z t 3 [ ∫ z c z t z 3 h ( z ) z t 6 − z 6 d z + ∫ δ z c z 6 H ( z ) d z ] . (35)
The above equations provide the basis for our time-dependent analysis. Before examining entanglement evolution, we first study the static extremal hypersurfaces in the vacuum and final black-brane geometries.
Preliminaries: Static backgrounds
We first examine the entanglement entropy in the initial vacuum and final black-brane backgrounds, which provide the reference values for the subsequent time-dependent analysis. At finite radial cutoff, several candidate RT configurations arise [26 ] , as illustrated in Figure 1 . The U-shaped surfaces comprise a deeper branch U 1 U_{1} U 1 and a shallower branch U 2 U_{2} U 2 , while D D D denotes the constant configuration at z = δ z=\delta z = δ . At finite temperature, an additional configuration P P P consists of two radial segments reaching the horizon and a connecting horizon segment required by homology. A further U-shaped branch exists near the cutoff, with z t → δ + z_{t}\to\delta^{+} z t → δ + and ℓ → 0 \ell\to0 ℓ → 0 , distinct from both U 2 U_{2} U 2 and D D D . We denote the local minimum separating the two bulk U-shaped branches by ℓ 0 \ell_{0} ℓ 0 , and the cutoff-dependent scale associated with the competing constant configuration by ℓ c \ell_{c} ℓ c . These definitions apply to both vacuum and thermal backgrounds. Unless stated otherwise, the static numerical results are obtained with z h = 1 z_{h}=1 z h = 1 and δ = 0.05 \delta=0.05 δ = 0.05 .
Figure 1. Candidate static configurations at finite radial cutoff. The surfaces U 1 U_1 U 1 and U 2 U_2 U 2 denote the deeper and shallower U-shaped branches, respectively. The configuration D D D lies at the cutoff surface, while the finite-temperature configuration P P P reaches the horizon and includes the horizon segment required by homology.
Zero-temperature
For the vacuum geometry, the static extremal surface lies on a constant-time slice, so the integration constant in Eq. (25) vanishes. Hence, the strip width and entropy become
ℓ = 2 ∫ δ z t d z z z 4 + a 4 z t 6 − z 6 , S v a c = 2 z t 3 ∫ δ z t d z z 4 + a 4 z 5 z t 6 − z 6 . (36) \ell=2\int_{\delta}^{z_{t}}dz\,\frac{z\sqrt{z^{4}+a^{4}}}{\sqrt{z_{t}^{6}-z^{6}}},\qquad S_{\mathrm{vac}}=2z_{t}^{3}\int_{\delta}^{z_{t}}dz\,\frac{\sqrt{z^{4}+a^{4}}}{z^{5}\sqrt{z_{t}^{6}-z^{6}}}. \tag*{(36)} ℓ = 2 ∫ δ z t d z z t 6 − z 6 z z 4 + a 4 , S vac = 2 z t 3 ∫ δ z t d z z 5 z t 6 − z 6 z 4 + a 4 . (36)
Since the near-cutoff branch extends to arbitrarily small widths, ℓ 0 \ell_0 ℓ 0 is defined as the local minimum separating the two bulk U-shaped branches,
ℓ 0 ≡ ℓ ( z t ) ∣ ∂ z t ℓ = 0 , ∂ z t 2 ℓ > 0 ≃ 1.6 a . (37) \ell_0 \equiv\ell(z_t)\big|_{\partial_{z_t}\ell=0,\partial_{z_t}^2\ell>0} \simeq1.6\,a. \tag*{(37)} ℓ 0 ≡ ℓ ( z t ) ∂ z t ℓ = 0 , ∂ z t 2 ℓ > 0 ≃ 1.6 a . (37)
The coefficient in the above equation is obtained in the small-cutoff limit [26 ] . Thus, for ℓ < ℓ 0 \ell< \ell_0 ℓ < ℓ 0 , there is no bulk U-shaped solution, although the solutions with z t ≃ δ z_t \simeq\delta z t ≃ δ remain. At finite cutoff, U 2 U_2 U 2 joins this branch at a local maximum. This maximum moves to arbitrarily large width as δ / a → 0 \delta/a \to0 δ / a → 0 , and the usual two-branch structure is recovered. Among U 1 U_1 U 1 and U 2 U_2 U 2 , the deeper branch U 1 U_1 U 1 has the smaller area [26 ] . On the other hand, the constant configuration D D D has entropy
S D ( ℓ ) = ℓ δ 3 . (38) S_D(\ell)=\frac{\ell}{\delta^3}. \tag*{(38)} S D ( ℓ ) = δ 3 ℓ . (38)
The linear dependence on ℓ \ell ℓ gives the volume-law behavior of the configuration D D D [25 , 26 , 27 ] . The near-boundary expansion of the U-shaped surface entropy is
S v a c d i v = a 2 2 δ 4 + 1 a 2 log ( a δ ) . (39) S_{\mathrm{vac}}^{\mathrm{div}}=\frac{a^2}{2\delta^4}+\frac{1}{a^2}\log\left(\frac{a}{\delta}\right). \tag*{(39)} S vac div = 2 δ 4 a 2 + a 2 1 log ( δ a ) . (39)
Comparing this cutoff-dependent expression with Eq. (38) gives the following estimate for ℓ c \ell_c ℓ c ,
ℓ c = a 2 2 δ + δ 3 a 2 log ( a δ ) ≃ a 2 2 δ . (40) \ell_c=\frac{a^2}{2\delta}+\frac{\delta^3}{a^2}\log\left(\frac{a}{\delta}\right)\simeq\frac{a^2}{2\delta}. \tag*{(40)} ℓ c = 2 δ a 2 + a 2 δ 3 log ( δ a ) ≃ 2 δ a 2 . (40)
Unlike ℓ 0 \ell_0 ℓ 0 , which follows from the branch structure, ℓ c \ell_c ℓ c depends on the radial cutoff. The cutoff z = δ z=\delta z = δ corresponds to a momentum scale of order δ − 1 \delta^{-1} δ − 1 . Because of UV/IR mixing, this induces a length scale of order a 2 / δ a^2/\delta a 2 / δ along a noncommutative direction. This agrees with the leading term in Eq. (40) [22 , 26 , 27 ] . We therefore restrict our analysis to ℓ > ℓ c \ell>\ell_c ℓ > ℓ c .
Finite-temperature
We now consider the final equilibrium state described by the finite-temperature black brane. The blackening function is given in Eq. (4) , and the turning point satisfies δ < z t < z h \delta< z_t < z_h δ < z t < z h . Again, the minimal surface lies on a constant-time slice, and the integration constant in Eq. (25) vanishes. The strip width and entropy are
ℓ = 2 ∫ δ z t d z z z 4 + a 4 z t 6 − z 6 1 − z 4 / z h 4 , S e q = 2 z t 3 ∫ δ z t d z z 4 + a 4 z 5 z t 6 − z 6 1 − z 4 / z h 4 . (41) \ell= 2 \int_{\delta}^{z_t} dz \frac{z\sqrt{z^4+a^4}}{\sqrt{z_t^6-z^6}\sqrt{1-z^4/z_h^4}}, \qquad S_{\mathrm{eq}} = 2z_t^3 \int_{\delta}^{z_t} dz \frac{\sqrt{z^4+a^4}}{z^5\sqrt{z_t^6-z^6}\sqrt{1-z^4/z_h^4}}. \tag*{(41)} ℓ = 2 ∫ δ z t d z z t 6 − z 6 1 − z 4 / z h 4 z z 4 + a 4 , S eq = 2 z t 3 ∫ δ z t d z z 5 z t 6 − z 6 1 − z 4 / z h 4 z 4 + a 4 . (41)
The thermal U-shaped solutions have the same branch structure as in the vacuum geometry. At low temperature, this approaches the vacuum result in Eq. (37) . For z h ≪ a z_h \ll a z h ≪ a , one finds
ℓ 0 ≃ 2 a 2 z h = 2 π a 2 T , (42) \ell_0 \simeq\frac{2a^2}{z_h} = 2\pi a^2 T, \tag*{(42)} ℓ 0 ≃ z h 2 a 2 = 2 π a 2 T , (42)
which shows a linear dependence on temperature [26 ] . The near-boundary expansion of Eq. (41) gives
S e q d i v = a 2 2 δ 4 + 1 a 2 ( 1 + a 4 z h 4 ) log ( a δ ) . (43) S_{\mathrm{eq}}^{\mathrm{div}} = \frac{a^2}{2\delta^4} + \frac{1}{a^2}\left(1+\frac{a^4}{z_h^4}\right)\log\left(\frac{a}{\delta}\right). \tag*{(43)} S eq div = 2 δ 4 a 2 + a 2 1 ( 1 + z h 4 a 4 ) log ( δ a ) . (43)
The leading power divergence is independent of temperature, while the logarithmic term receives a thermal correction. Comparing with Eq. (38) gives
ℓ c = a 2 2 δ + δ 3 a 2 ( 1 + a 4 z h 4 ) log ( a δ ) ≃ a 2 2 δ . (44) \ell_c = \frac{a^2}{2\delta} + \frac{\delta^3}{a^2}\left(1+\frac{a^4}{z_h^4}\right)\log\left(\frac{a}{\delta}\right) \simeq\frac{a^2}{2\delta}. \tag*{(44)} ℓ c = 2 δ a 2 + a 2 δ 3 ( 1 + z h 4 a 4 ) log ( δ a ) ≃ 2 δ a 2 . (44)
We see that the positive logarithmic correction makes the final-state estimate larger than the vacuum one. As shown in Figure 2 , increasing a a a shifts the local minimum ℓ 0 \ell_0 ℓ 0 toward larger strip widths. For the parameters used below and for sufficiently wide strips, the parallel configuration P P P becomes relevant only when the U-shaped branch is absent [26 ] . At low temperature and small cutoff, where ℓ c ≫ ℓ 0 \ell_c \gg\ell_0 ℓ c ≫ ℓ 0 , the U-shaped solutions may still exist for widths below ℓ c \ell_c ℓ c , although they need not be the globally minimal configurations.
Figure 2. Strip width as a function of turning point for several values of the noncommutativity parameter. The displayed minima define ℓ 0 \ell_0 ℓ 0 . The dashed lines show the asymptotic estimates of ℓ c \ell_c ℓ c from Eq. (44) .
Moreover, Figure 3 shows the final-state estimate of ℓ c \ell_c ℓ c increases with a a a . Using the leading estimates in Eqs. (42) and (44) , the condition ℓ 0 > ℓ c \ell_0 > \ell_c ℓ 0 > ℓ c requires T ≳ 1 / ( 4 π δ ) T \gtrsim1/(4\pi\delta) T ≳ 1/ ( 4 π δ ) . In this range, the temperature is of the same order as the UV cutoff, and the transition between U 1 U_1 U 1 and P P P depends on the cutoff [26 ] . We therefore do not consider this range of parameters. Finally, on the U 1 U_1 U 1 branch, the equilibrium entropy is extensive for sufficiently wide strips. Subtracting the cutoff-dependent terms in Eq. (43) , one finds for ℓ T ≫ 1 \ell T \gg1 ℓ T ≫ 1
Figure 3. Asymptotic final-state estimate of the cutoff scale ℓ c \ell_c ℓ c from Eq. (44) . Left: As a function of a a a at fixed cutoff. Right: As a function of δ \delta δ at fixed a a a .
S e q ( ℓ ) − S e q d i v = ℓ z h p + O ( ℓ 0 ) , (45) S_{\mathrm{eq}}(\ell)-S_{\mathrm{eq}}^{\mathrm{div}}=\frac{\ell}{z_h^p}+\mathcal{O}\left(\ell^0\right), \tag*{(45)} S eq ( ℓ ) − S eq div = z h p ℓ + O ( ℓ 0 ) , (45)
which coincides with the thermal entropy contribution at the leading order.
Time evolution of entanglement entropy
We now investigate the time evolution of holographic entanglement entropy following the global quench. Our analytic treatment is based on the thin-shell geometry, whereas the numerical calculations employ the smooth Vaidya profile introduced in Eq. (16) . We define
Δ S ( ℓ , t ) ≡ S ( ℓ , t ) − S v a c ( ℓ ) , (46) \Delta S(\ell,t)\equiv S(\ell,t)-S_{\mathrm{vac}}(\ell), \tag*{(46)} Δ S ( ℓ , t ) ≡ S ( ℓ , t ) − S vac ( ℓ ) , (46)
where both entropies are evaluated at the same finite radial cutoff.
Early growth
At early times, the HRT surface intersects the infalling shell at a radial position z c z_c z c close to the boundary, satisfying z c ≪ min { z h , z t } z_c \ll\min\{z_h,z_t\} z c ≪ min { z h , z t } . The early-time expansion depends on the relative magnitude of z c z_c z c and the noncommutative scale a a a . We therefore consider the two limits separately.
We first consider the regime δ ≪ z c ≪ min { a , z h , z t } \delta\ll z_c \ll\min\{a,z_h,z_t\} δ ≪ z c ≪ min { a , z h , z t } . The leading time-dependent contribution comes from the portion of the extremal surface in the black-brane region, z < z c z<z_c z < z c . Expanding Eqs. (33) and (35) at fixed strip width yields
Δ S ( ℓ , t ) ≃ a 2 z h 4 log ( z c δ ) + ( 1 8 a 2 z h 4 + 7 a 2 48 z h 8 ) z c 4 − a 2 12 z h 4 z t 6 z c 6 + ⋯ . (47) \Delta S(\ell,t) \simeq\frac{a^2}{z_h^4}\log\left(\frac{z_c}{\delta}\right)+\left(\frac{1}{8a^2z_h^4}+\frac{7a^2}{48z_h^8}\right)z_c^4-\frac{a^2}{12z_h^4z_t^6}z_c^6+\cdots. \tag*{(47)} Δ S ( ℓ , t ) ≃ z h 4 a 2 log ( δ z c ) + ( 8 a 2 z h 4 1 + 48 z h 8 7 a 2 ) z c 4 − 12 z h 4 z t 6 a 2 z c 6 + ⋯ . (47)
Expanding Eq. (34) through order z c 5 z_c^5 z c 5 gives t + δ = z c + 7 z c 5 60 z h 4 + ⋯ t+\delta=z_c+\frac{7z_c^5}{60z_h^4}+\cdots t + δ = z c + 60 z h 4 7 z c 5 + ⋯ . Using the large-strip limit of the vacuum width relation, we obtain, for δ ≪ t ≪ a \delta\ll t\ll a δ ≪ t ≪ a and t ≪ z h , z t t\ll z_h,z_t t ≪ z h , z t ,
Δ S ( ℓ , t ) ≃ 2 π A Σ 3 ε a 2 log ( t δ ) + [ π A Σ 12 a 2 ε + 7 π 2 A Σ 2 a 2 540 ε 2 ] t 4 − 32 π A Σ 9 [ π Γ ( 2 3 ) Γ ( 1 6 ) ] 6 ε a 2 ( t ℓ ) 6 + ⋯ . (48) \Delta S(\ell,t) \simeq\frac{2\pi A_\Sigma}{3}\varepsilon a^2\log\left(\frac{t}{\delta}\right)+\left[\frac{\pi A_\Sigma}{12a^2}\varepsilon+\frac{7\pi^2A_\Sigma^2a^2}{540}\varepsilon^2\right]t^4-\frac{32\pi A_\Sigma}{9}\left[\frac{\sqrt{\pi}\Gamma\left(\frac{2}{3}\right)}{\Gamma\left(\frac{1}{6}\right)}\right]^6\varepsilon a^2\left(\frac{t}{\ell}\right)^6+\cdots. \tag*{(48)} Δ S ( ℓ , t ) ≃ 3 2 π A Σ ε a 2 log ( δ t ) + [ 12 a 2 π A Σ ε + 540 7 π 2 A Σ 2 a 2 ε 2 ] t 4 − 9 32 π A Σ [ Γ ( 6 1 ) π Γ ( 3 2 ) ] 6 ε a 2 ( ℓ t ) 6 + ⋯ . (48)
Here A Σ = 2 L 2 A_\Sigma=2L^2 A Σ = 2 L 2 is the area of the entangling surface. The leading growth is logarithmic and independent of the strip width, while the first explicit width-dependent correction appears at order ( t / ℓ ) 6 (t/\ell)^6 ( t / ℓ ) 6 . To express the logarithmic term as log ( t / δ ) \log(t/\delta) log ( t / δ ) , the additional condition δ ≪ t \delta\ll t δ ≪ t is required, since z c = t + δ + ⋯ z_c=t+\delta+\cdots z c = t + δ + ⋯ . Using Eqs. (39) and (43) , the leading logarithmic result can alternatively be written as
Δ S ( ℓ , t ) ≃ ( S e q d i v − S v a c d i v ) + 2 π A Σ 3 ε a 2 log ( t a ) + ⋯ . (49) \Delta S(\ell,t)\simeq\left(S_{\mathrm{eq}}^{\mathrm{div}}-S_{\mathrm{vac}}^{\mathrm{div}}\right)+\frac{2\pi A_\Sigma}{3}\varepsilon a^2\log\left(\frac{t}{a}\right)+\cdots. \tag*{(49)} Δ S ( ℓ , t ) ≃ ( S eq div − S vac div ) + 3 2 π A Σ ε a 2 log ( a t ) + ⋯ . (49)
This form separates the cutoff-dependent static term from the time-dependent logarithm, which is cutoff independent.
We next consider the regime δ ≪ a ≪ z c ≪ min { z h , z t } \delta\ll a\ll z_c\ll\min\{z_h,z_t\} δ ≪ a ≪ z c ≪ min { z h , z t } . Expanding Eqs. (33) and (35) at fixed strip width, we obtain the following time-dependent contributions, omitting time-independent terms:
Δ S ( ℓ , t ) ≃ z c 2 2 z h 4 − a 4 4 z h 4 z c 2 + z c 6 16 z h 8 − z c 8 16 z h 6 z t 2 + ⋯ . (50) \Delta S(\ell,t)\simeq\frac{z_c^2}{2z_h^4}-\frac{a^4}{4z_h^4z_c^2}+\frac{z_c^6}{16z_h^8}-\frac{z_c^8}{16z_h^6z_t^2}+\cdots. \tag*{(50)} Δ S ( ℓ , t ) ≃ 2 z h 4 z c 2 − 4 z h 4 z c 2 a 4 + 16 z h 8 z c 6 − 16 z h 6 z t 2 z c 8 + ⋯ . (50)
At fixed strip width, the turning point receives a correction z t → z t + Δ z t z_t\to z_t+\Delta z_t z t → z t + Δ z t , with Δ z t = O ( z c 8 ) \Delta z_t=O(z_c^8) Δ z t = O ( z c 8 ) , as determined by Eq. (33) . Expanding Eq. (34) gives
t + δ = z c + 3 z c 5 40 z h 4 + ⋯ . (51) t+\delta=z_c+\frac{3z_c^5}{40z_h^4}+\cdots. \tag*{(51)} t + δ = z c + 40 z h 4 3 z c 5 + ⋯ . (51)
Combining these results with the large-strip limit of the vacuum width relation in Eq. (36) , we obtain, for δ ≪ a ≪ t ≪ min { z h , z t } \delta\ll a\ll t\ll\min\{z_h,z_t\} δ ≪ a ≪ t ≪ min { z h , z t } ,
Δ S ( ℓ , t ) ≃ π A Σ 3 ε t 2 − π A Σ 6 ε a 4 t 2 − π 2 A Σ 2 180 ε 2 t 6 − 8 π A Σ 3 [ π Γ ( 2 3 ) Γ ( 1 6 ) ] 6 ε t 2 ( t ℓ ) 6 + ⋯ . (52) \Delta S(\ell,t)\simeq\frac{\pi A_\Sigma}{3}\varepsilon t^2-\frac{\pi A_\Sigma}{6}\varepsilon\frac{a^4}{t^2}-\frac{\pi^2A_\Sigma^2}{180}\varepsilon^2t^6-\frac{8\pi A_\Sigma}{3}\left[\frac{\sqrt{\pi}\Gamma\left(\frac{2}{3}\right)}{\Gamma\left(\frac{1}{6}\right)}\right]^6\varepsilon t^2\left(\frac{t}{\ell}\right)^6+\cdots. \tag*{(52)} Δ S ( ℓ , t ) ≃ 3 π A Σ ε t 2 − 6 π A Σ ε t 2 a 4 − 180 π 2 A Σ 2 ε 2 t 6 − 3 8 π A Σ [ Γ ( 6 1 ) π Γ ( 3 2 ) ] 6 ε t 2 ( ℓ t ) 6 + ⋯ . (52)
The leading time-dependent term is quadratic and proportional to the energy density, as in asymptotically AdS–Vaidya geometries [14 ] . The remaining terms give corrections due to the noncommutative scale, higher orders in the black-brane mass, and the finite strip width. In the present geometry, the near-boundary behavior of the warp factor h ( z ) h(z) h ( z ) modifies the quadratic growth found in asymptotically AdS–Vaidya geometries and yields a logarithmic time dependence when z c ≪ a z_c\ll a z c ≪ a . The same near-boundary behavior determines the cutoff dependence of the static entanglement entropy in NCYM [25 , 26 , 27 ] . Thus, the early-time scaling is not universal and depends on the underlying field theory. In the present setup, spatial noncommutativity breaks Lorentz invariance and changes the asymptotic geometry. Finally, the logarithmic expansion is valid only for z c ≪ a z_c \ll a z c ≪ a . The commutative result therefore cannot be obtained by setting a → 0 a \to0 a → 0 in that expansion; one must take this limit in the full equations before expanding near the boundary.
Critical surfaces and linear growth
In AdS–Vaidya geometries, the linear growth regime is associated with shell-crossing extremal surfaces approaching a critical configuration behind the final horizon [15 ] . For parameter values admitting a boundary anchored branch, a critical shell-crossing point z c ∗ ( z t ) z_c^*(z_t) z c ∗ ( z t ) exists. As the shell-crossing point approaches z c ∗ z_c^* z c ∗ , the returning surface develops an extended portion near the critical radius, leading to linear growth of the corresponding entanglement entropy.
From the refraction condition in Eq. (29) , and since z − ′ < 0 z'_- < 0 z − ′ < 0 , the radial derivative immediately after crossing the shell becomes positive when
g ( z c ) > 2 , z c > 2 1 / 4 z h . (53) g(z_c) > 2,\qquad z_c > 2^{1/4}z_h. \tag*{(53)} g ( z c ) > 2 , z c > 2 1/4 z h . (53)
In this case, the surface moves toward larger radial depths rather than toward the boundary. To reach the boundary at z = δ < z c z = \delta< z_c z = δ < z c , it must reverse its radial direction at a turning point z r z_r z r , where z ′ = 0 z' = 0 z ′ = 0 and hence H ( z r ) = 0 H(z_r) = 0 H ( z r ) = 0 . Thus, the above condition is necessary, but not sufficient, for a returning solution to exist. The turning point satisfies z r > z c z_r > z_c z r > z c , with H ( z ) > 0 H(z) > 0 H ( z ) > 0 along the black-brane trajectory for δ ≤ z < z r \delta\le z < z_r δ ≤ z < z r . For an ordinary turning point, we also have H ′ ( z r ) < 0 H'(z_r) < 0 H ′ ( z r ) < 0 . At z r z_r z r , the radial derivative z ′ z' z ′ changes sign from positive to negative. Accordingly, we choose the positive root of Eq. (31) along the outward-in-depth segment from z c z_c z c to z r z_r z r , and the negative root along the returning segment toward the boundary. Hence, taking into account both radial directions of the black-brane trajectory, the strip width, boundary time, and entanglement entropy are given by
ℓ 2 = ∫ z c z t d z h ( z ) z t 6 / z 6 − 1 + ∫ δ z c d z H ( z ) + 2 ∫ z c z r d z H ( z ) , (54) \frac{\ell}{2} = \int_{z_c}^{z_t} \frac{dz}{\sqrt{h(z)}\sqrt{z_t^6/z^6-1}}+\int_{\delta}^{z_c}\frac{dz}{\sqrt{H(z)}}+2\int_{z_c}^{z_r}\frac{dz}{\sqrt{H(z)}}, \tag*{(54)} 2 ℓ = ∫ z c z t h ( z ) z t 6 / z 6 − 1 d z + ∫ δ z c H ( z ) d z + 2 ∫ z c z r H ( z ) d z , (54)
t = ∫ δ z c d z f ( z ) [ 1 + h ( z ) E H ( z ) ] + 2 ∫ z c z r h ( z ) E d z f ( z ) H ( z ) , (55) t = \int_{\delta}^{z_c}\frac{dz}{f(z)}\left[1+\frac{h(z)E}{\sqrt{H(z)}}\right]+2\int_{z_c}^{z_r}\frac{h(z)E\,dz}{f(z)\sqrt{H(z)}}, \tag*{(55)} t = ∫ δ z c f ( z ) d z [ 1 + H ( z ) h ( z ) E ] + 2 ∫ z c z r f ( z ) H ( z ) h ( z ) E d z , (55)
S = 2 z t 3 [ ∫ z c z t d z z 3 h ( z ) z t 6 − z 6 + ∫ δ z c d z z 6 H ( z ) + 2 ∫ z c z r d z z 6 H ( z ) ] . (56) S = 2z_t^3\left[\int_{z_c}^{z_t}\frac{dz}{z^3\sqrt{h(z)}\sqrt{z_t^6-z^6}}+\int_{\delta}^{z_c}\frac{dz}{z^6\sqrt{H(z)}}+2\int_{z_c}^{z_r}\frac{dz}{z^6\sqrt{H(z)}}\right]. \tag*{(56)} S = 2 z t 3 [ ∫ z c z t z 3 h ( z ) z t 6 − z 6 d z + ∫ δ z c z 6 H ( z ) d z + 2 ∫ z c z r z 6 H ( z ) d z ] . (56)
From the matching condition, the integration constant satisfies E < 0 E < 0 E < 0 . At fixed z t z_t z t , the integration constant E E E depends on the shell-crossing point z c z_c z c through Eq. (30) . For a near-critical hypersurface, z r z_r z r is a simple zero of the radial function H ( z ) H(z) H ( z ) and lies on the smaller-z z z side of the critical minimum z m z_m z m , with z c < z r < z m z_c < z_r < z_m z c < z r < z m . As z c z_c z c approaches its critical value z c ∗ ( z t ) z_c^*(z_t) z c ∗ ( z t ) from below, z r z_r z r approaches z m z_m z m from below. At criticality, the two neighboring radial roots coincide at z = z m z = z_m z = z m , where the minimum becomes a double zero, satisfying
H ( z m ) ∣ z c = z c ∗ = 0 , ∂ z H ( z ) ∣ z = z m , z c = z c ∗ = 0. (57) H(z_m)\big|_{z_c=z_c^*}=0,\qquad\partial_z H(z)\big|_{z=z_m,\,z_c=z_c^*}=0. \tag*{(57)} H ( z m ) z c = z c ∗ = 0 , ∂ z H ( z ) z = z m , z c = z c ∗ = 0. (57)
Combining the above equations with Eq. (31) and eliminating E E E gives
z t 6 = z m 7 [ ∂ z f ( z m ) h ( z m ) − f ( z m ) ∂ z h ( z m ) ] z m [ ∂ z f ( z m ) h ( z m ) − f ( z m ) ∂ z h ( z m ) ] − 6 f ( z m ) h ( z m ) . (58) z_t^6 = \frac{z_m^7\left[\partial_z f(z_m)h(z_m)-f(z_m)\partial_z h(z_m)\right]}{z_m\left[\partial_z f(z_m)h(z_m)-f(z_m)\partial_z h(z_m)\right]-6f(z_m)h(z_m)}. \tag*{(58)} z t 6 = z m [ ∂ z f ( z m ) h ( z m ) − f ( z m ) ∂ z h ( z m ) ] − 6 f ( z m ) h ( z m ) z m 7 [ ∂ z f ( z m ) h ( z m ) − f ( z m ) ∂ z h ( z m ) ] . (58)
For fixed z t z_t z t , this equation determines z m z_m z m on the critical branch. The transition between monotonic and returning critical configurations occurs when the critical radius reaches the shell-refraction threshold z s z_s z s , defined by g ( z s ) = 2 g(z_s)=2 g ( z s ) = 2 . At this transition, z m = z c ∗ = z s z_m=z_c^*=z_s z m = z c ∗ = z s . Substituting this condition into Eq. (58) , we find
z t ( s ) = 2 1 / 4 z h [ 2 ( 4 z h 4 + a 4 ) 2 z h 4 − a 4 ] 1 / 6 , a 4 < 2 z h 4 . (59) z_t^{(s)}=2^{1/4}z_h\left[\frac{2(4z_h^4+a^4)}{2z_h^4-a^4}\right]^{1/6},\qquad a^4<2z_h^4. \tag*{(59)} z t ( s ) = 2 1/4 z h [ 2 z h 4 − a 4 2 ( 4 z h 4 + a 4 ) ] 1/6 , a 4 < 2 z h 4 . (59)
For z h < z t < z t ( s ) z_h<z_t<z_t^{(s)} z h < z t < z t ( s ) , the relevant critical surface is monotonic, with its characteristic radii satisfying z m < z c ∗ < z s z_m<z_c^*<z_s z m < z c ∗ < z s . As z t z_t z t increases beyond z t ( s ) z_t^{(s)} z t ( s ) , the ordering changes to z s < z c ∗ < z m z_s<z_c^*<z_m z s < z c ∗ < z m , allowing near-critical returning surfaces for which z s < z c < z r < z m z_s<z_c<z_r<z_m z s < z c < z r < z m and z c < z c ∗ z_c<z_c^* z c < z c ∗ , provided the radial turning point is accessible. Moreover, in the large-strip limit z t → ∞ z_t\to\infty z t → ∞ , the critical equations simplify while z m z_m z m and z c ∗ z_c^* z c ∗ remain finite, yielding
z m = [ 3 ( z h 4 − a 4 ) + 9 ( z h 4 − a 4 ) 2 + 20 a 4 z h 4 2 ] 1 / 4 , (60) z_m=\left[\frac{3\left(z_h^4-a^4\right)+\sqrt{9\left(z_h^4-a^4\right)^2+20a^4z_h^4}}{2}\right]^{1/4}, \tag*{(60)} z m = 2 3 ( z h 4 − a 4 ) + 9 ( z h 4 − a 4 ) 2 + 20 a 4 z h 4 1/4 , (60)
which satisfies z m > z h z_m>z_h z m > z h . The corresponding critical crossing point is determined by combining the matching condition with Eq. (57) which gives
z c ∗ = [ 2 z h 4 ( z m 4 − z h 4 ) ( z m 4 + a 4 ) z m 10 + 4 z h 8 ( z m 4 − z h 4 ) 2 ( z m 4 + a 4 ) 2 z m 20 − a 4 ] 1 / 2 , (61) z_c^*=\left[2z_h^4\frac{\left(z_m^4-z_h^4\right)\left(z_m^4+a^4\right)}{z_m^{10}}+\sqrt{4z_h^8\frac{\left(z_m^4-z_h^4\right)^2\left(z_m^4+a^4\right)^2}{z_m^{20}}-a^4}\right]^{1/2}, \tag*{(61)} z c ∗ = 2 z h 4 z m 10 ( z m 4 − z h 4 ) ( z m 4 + a 4 ) + 4 z h 8 z m 20 ( z m 4 − z h 4 ) 2 ( z m 4 + a 4 ) 2 − a 4 1/2 , (61)
Here z m z_m z m is determined by Eq. (60) . In the large-strip limit, the returning critical branch is characterized by
0 ≤ a < 2 1 / 4 z h , 2 1 / 4 z h < z c ∗ < z m . 0\leq a<2^{1/4}z_h,\qquad2^{1/4}z_h<z_c^*<z_m. 0 ≤ a < 2 1/4 z h , 2 1/4 z h < z c ∗ < z m .
At a = 2 1 / 4 z h a=2^{1/4}z_h a = 2 1/4 z h , the two critical radii coincide, z c ∗ = z m = 2 1 / 4 z h z_c^*=z_m=2^{1/4}z_h z c ∗ = z m = 2 1/4 z h , and the radial interval collapses. At this endpoint, the leading variation of H ( z ) H(z) H ( z ) with the crossing position degenerates in the large-strip limit. Although the algebraic expression for z m z_m z m remains defined beyond this range, it does not establish the existence of a boundary-reaching profile. In the commutative limit, Eqs. (60) and (61) reproduce the AdS5 _5 5 –Vaidya result of [15 ] . Further, both radii decrease with a a a , and their difference also decreases. Near a = 0 a=0 a = 0 , the change is small because the first correction is of order a 4 a^4 a 4 .
A near-critical expansion of the extremal surfaces allows the linear-growth scaling to be extracted for sufficiently wide strips in the intermediate-time regime z h ≪ t ≪ ℓ z_h\ll t\ll\ell z h ≪ t ≪ ℓ . The expansion is parametrized as
z c = z c ∗ ( 1 − ϵ ) , 0 < ϵ ≪ 1 , z c ∗ , z m ≪ z t . (62) z_c=z_c^*(1-\epsilon),\qquad0<\epsilon\ll1,\qquad z_c^*,z_m\ll z_t. \tag*{(62)} z c = z c ∗ ( 1 − ϵ ) , 0 < ϵ ≪ 1 , z c ∗ , z m ≪ z t . (62)
The additional condition z c ∗ / z h ≪ ∣ log ϵ ∣ z_c^*/z_h\ll|\log\epsilon| z c ∗ / z h ≪ ∣ log ϵ ∣ ensures that the near-critical logarithmic contributions dominate over the finite terms. We expand Eqs. (54) –(56) near the critical surface, starting with the radial function H ( z ) H(z) H ( z )
H ( z ) = H 2 ( z − z m ) 2 + b ϵ + ⋯ , H 2 = 1 2 ∂ z 2 H ∣ z = z m , z c = z c ∗ , b = − z c ∗ ∂ H ∂ z c ∣ z = z m , z c = z c ∗ . (63) H(z)=H_2(z-z_m)^2+b\epsilon+\cdots,\qquad H_2=\left.\frac{1}{2}\partial_z^2H\right|_{z=z_m,z_c=z_c^*},\qquad b=-\left.z_c^*\frac{\partial H}{\partial z_c}\right|_{z=z_m,z_c=z_c^*}. \tag*{(63)} H ( z ) = H 2 ( z − z m ) 2 + b ϵ + ⋯ , H 2 = 2 1 ∂ z 2 H z = z m , z c = z c ∗ , b = − z c ∗ ∂ z c ∂ H z = z m , z c = z c ∗ . (63)
Here the derivative with respect to z c z_c z c is evaluated at fixed z t z_t z t . For the near-critical branch at fixed z t > z t ( s ) z_t > z_t^{(s)} z t > z t ( s ) , we have H 2 > 0 H_2 > 0 H 2 > 0 and b < 0 b < 0 b < 0 . Since H ( z r ) = 0 H(z_r) = 0 H ( z r ) = 0 , the smaller root approaches z m z_m z m according to
z r = z m − b H 2 ϵ 1 / 2 + O ( ϵ ) . (64) z_r = z_m - \sqrt{\frac{b}{H_2}} \epsilon^{1/2} + O(\epsilon). \tag*{(64)} z r = z m − H 2 b ϵ 1/2 + O ( ϵ ) . (64)
The logarithmic divergence arises from the returning segment upon using Eq. (64) , since
∫ z c z r d z H ( z ) = − 1 2 H 2 log ϵ + O ( 1 ) . (65) \int_{z_c}^{z_r} \frac{dz}{\sqrt{H(z)}} = -\frac{1}{2\sqrt{H_2}}\log\epsilon+ O(1). \tag*{(65)} ∫ z c z r H ( z ) d z = − 2 H 2 1 log ϵ + O ( 1 ) . (65)
This segment is traversed twice in the parametric expressions, which produces logarithmic contributions to the strip width, boundary time, and entropy. The remainder here is bounded as ϵ → 0 + \epsilon\to0^+ ϵ → 0 + at fixed z t z_t z t . Using Eq. (65) , the leading strip-width expression from Eq. (54) becomes
ℓ 2 ≃ c z t − 1 H 2 log ϵ + ⋯ , (66) \frac{\ell}{2} \simeq c z_t - \frac{1}{\sqrt{H_2}}\log\epsilon+ \cdots, \tag*{(66)} 2 ℓ ≃ c z t − H 2 1 log ϵ + ⋯ , (66)
where
c = π Γ ( 2 3 ) Γ ( 1 6 ) + ( a z t ) 4 [ 1 2 log ( z t a ) + 1 8 + 5 12 log 2 ] + O ( a 8 z t 8 ) . (67) c = \frac{\sqrt{\pi}\,\Gamma\left(\frac{2}{3}\right)}{\Gamma\left(\frac{1}{6}\right)} + \left(\frac{a}{z_t}\right)^4 \left[\frac{1}{2}\log\left(\frac{z_t}{a}\right) + \frac{1}{8} + \frac{5}{12}\log2\right] + O\left(\frac{a^8}{z_t^8}\right). \tag*{(67)} c = Γ ( 6 1 ) π Γ ( 3 2 ) + ( z t a ) 4 [ 2 1 log ( a z t ) + 8 1 + 12 5 log 2 ] + O ( z t 8 a 8 ) . (67)
Further, at the double zero, Eq. (31) fixes the critical value of E E E as follows
E ( z c ∗ ) = − − f ( z m ) h ( z m ) ( z t 6 z m 6 − 1 ) . (68) E(z_c^*) = -\sqrt{-\frac{f(z_m)}{h(z_m)}\left(\frac{z_t^6}{z_m^6}-1\right)}. \tag*{(68)} E ( z c ∗ ) = − − h ( z m ) f ( z m ) ( z m 6 z t 6 − 1 ) . (68)
Expanding Eqs. (55) and (56) near the critical surface, we obtain the leading logarithmic contributions to the boundary time and the entanglement entropy as
t ≃ − E ( z c ∗ ) h ( z m ) f ( z m ) H 2 log ϵ + ⋯ , Δ S ( ℓ , t ) ≃ − 2 z t 3 z m 6 H 2 log ϵ + ⋯ . (69) t \simeq-\frac{E(z_c^*)h(z_m)}{f(z_m)\sqrt{H_2}}\log\epsilon+ \cdots,\qquad\Delta S(\ell,t) \simeq-\frac{2z_t^3}{z_m^6\sqrt{H_2}}\log\epsilon+ \cdots. \tag*{(69)} t ≃ − f ( z m ) H 2 E ( z c ∗ ) h ( z m ) log ϵ + ⋯ , Δ S ( ℓ , t ) ≃ − z m 6 H 2 2 z t 3 log ϵ + ⋯ . (69)
Eliminating log ϵ \log\epsilon log ϵ between the width and time relations gives an implicit relation for z t ( ℓ , t ) z_t(\ell,t) z t ( ℓ , t ) . The leading logarithms in t t t and Δ S \Delta S Δ S originate from the same near-critical returning segment. Combining the two relations in Eq. (69) and using Eq. (68) in the limit z t ≫ z m z_t \gg z_m z t ≫ z m , we obtain
Δ S ( ℓ , t ) = s t h A Σ v E t + ⋯ , (70) \Delta S(\ell,t)=s_{\mathrm{th}}A_{\Sigma}v_Et+\cdots, \tag*{(70)} Δ S ( ℓ , t ) = s th A Σ v E t + ⋯ , (70)
where we identify
v E = z h 3 z m 3 f ( z m ) h ( z m ) . (71) v_E=\frac{z_h^3}{z_m^3}\sqrt{\frac{f(z_m)}{h(z_m)}}. \tag*{(71)} v E = z m 3 z h 3 h ( z m ) f ( z m ) . (71)
The linear-growth result is derived for the near-critical extremal surface. The leading imprint of noncommutativity on the linear-growth regime can be isolated analytically by expanding the entanglement velocity for a / z h ≪ 1 a/z_h\ll1 a / z h ≪ 1 . This gives
v E = 2 3 3 / 4 [ 1 + 1 6 ( a z h ) 4 + 5 216 ( a z h ) 8 + O ( ( a z h ) 12 ) ] , (72) v_E=\frac{\sqrt{2}}{3^{3/4}}\left[1+\frac{1}{6}\left(\frac{a}{z_h}\right)^4+\frac{5}{216}\left(\frac{a}{z_h}\right)^8+\mathcal{O}\left(\left(\frac{a}{z_h}\right)^{12}\right)\right], \tag*{(72)} v E = 3 3/4 2 [ 1 + 6 1 ( z h a ) 4 + 216 5 ( z h a ) 8 + O ( ( z h a ) 12 ) ] , (72)
which shows that noncommutativity enhances the entanglement velocity, with the first correction appearing at order ( a / z h ) 4 (a/z_h)^4 ( a / z h ) 4 . This result agrees with the entanglement velocity obtained from the shock-wave analysis of thermal noncommutative Yang–Mills theory [29 ] . In the commutative limit, v E → 2 / 3 3 / 4 v_E\to\sqrt{2}/3^{3/4} v E → 2 / 3 3/4 , recovering the AdS5 _5 5 result [14 ] .
Saturation
The near-critical branch described above does not necessarily connect continuously to the final equilibrium surface. In particular, the additional radial turning point z r z_r z r is specific to the shell-crossing geometry and has no counterpart in the static equilibrium U-shaped surface. Multiple extremal branches may therefore coexist near saturation, and the physical HRT surface is determined by the branch with the smallest admissible area at each boundary time. In the late-time limit, the physical extremal surface approaches the static black-brane configuration. Let z b z_b z b denote the turning point of the equilibrium U-shaped surface with the same strip width. In the parameter range considered in § 3.2 , the deeper branch U 1 U_1 U 1 has the smallest area. If a dynamical branch approaches this surface continuously, then E = 0 E=0 E = 0 and z c = z t = z b z_c=z_t=z_b z c = z t = z b . Equation (34) then gives
t s ( c o n t ) = ∫ δ z b d z f ( z ) . (73) t_s^{(\mathrm{cont})}=\int_{\delta}^{z_b}\frac{\mathrm{d}z}{f(z)}. \tag*{(73)} t s ( cont ) = ∫ δ z b f ( z ) d z . (73)
This is the earliest boundary time at which the equilibrium surface lies entirely in the final black-brane region. The equilibrium strip width is given by Eq. (41) . In the large entangling region limit, the turning point of the extremal surface lies close to the horizon, so that z b = z h ( 1 − η ) z_b=z_h(1-\eta) z b = z h ( 1 − η ) with 0 < η ≪ 1 0<\eta\ll1 0 < η ≪ 1 . Since the final blackening function has a simple zero at the horizon, the leading terms are
ℓ 2 = − z h 6 ∣ ∂ z f ( z h ) ∣ h ( z h ) log η + ⋯ , t s ( c o n t ) = − log η ∣ ∂ z f ( z h ) ∣ + ⋯ . (74) \frac{\ell}{2}=-\sqrt{\frac{z_h}{6\lvert\partial_z f(z_h)\rvert h(z_h)}}\log\eta+\cdots,\qquad t_s^{(\mathrm{cont})}=-\frac{\log\eta}{\lvert\partial_z f(z_h)\rvert}+\cdots. \tag*{(74)} 2 ℓ = − 6 ∣ ∂ z f ( z h )∣ h ( z h ) z h log η + ⋯ , t s ( cont ) = − ∣ ∂ z f ( z h )∣ log η + ⋯ . (74)
Combining the above results we arrive at
t s ( c o n t ) = ℓ 2 3 z h 4 2 ( z h 4 + a 4 ) + ⋯ . (75) t_s^{(\mathrm{cont})}=\frac{\ell}{2}\sqrt{\frac{3z_h^4}{2(z_h^4+a^4)}}+\cdots. \tag*{(75)} t s ( cont ) = 2 ℓ 2 ( z h 4 + a 4 ) 3 z h 4 + ⋯ . (75)
which is a decreasing function of a a a . Another estimate of the saturation timescale can be obtained by extrapolating the linear-growth expression in Eq. (70) to the leading-order equilibrium entropy. Assuming that the linear-growth regime persists up to saturation, this yields
t l = ℓ 2 v E + ⋯ . (76) t_l = \frac{\ell}{2v_E} + \cdots. \tag*{(76)} t l = 2 v E ℓ + ⋯ . (76)
Let us now turn to the approach to the equilibrium configuration of the extremal surface. Near the saturation regime, both z t z_t z t and z c z_c z c approach the equilibrium turning point z b < z h z_b < z_h z b < z h , so we assume
z c = z t ( 1 − ϵ 2 6 ) , z t = z b ( 1 + σ 6 ) , 0 < ϵ , σ ≪ 1. (77) z_c = z_t \left(1-\frac{\epsilon^2}{6}\right), \qquad z_t = z_b \left(1+\frac{\sigma}{6}\right), \qquad0<\epsilon,\sigma\ll1. \tag*{(77)} z c = z t ( 1 − 6 ϵ 2 ) , z t = z b ( 1 + 6 σ ) , 0 < ϵ , σ ≪ 1. (77)
The parameter ϵ \epsilon ϵ parametrizes the separation between the shell and the tip, while the fixed-width condition gives σ = O ( ϵ ) \sigma= \mathcal{O}(\epsilon) σ = O ( ϵ ) . Substitution into the time integral and expansion to leading order give
t − t s ( c o n t ) = ω ϵ + O ( ϵ 2 ) , ω = g ( z b ) 2 h ( z b ) [ z b 3 f ( z b ) 2 F ′ ( z b ) − J ( z b ) ] , (78) t-t_s^{(\mathrm{cont})}=\omega\epsilon+\mathcal{O}(\epsilon^2), \qquad\omega=\frac{g(z_b)}{2\sqrt{h(z_b)}}\left[\frac{z_b}{3f(z_b)^2F'(z_b)}-\mathcal{J}(z_b)\right], \tag*{(78)} t − t s ( cont ) = ω ϵ + O ( ϵ 2 ) , ω = 2 h ( z b ) g ( z b ) [ 3 f ( z b ) 2 F ′ ( z b ) z b − J ( z b ) ] , (78)
where
F ( q ) = q ∫ 0 1 d y f ( q y ) h ( q y ) ( y − 6 − 1 ) , J ( q ) = q ∫ 0 1 h ( q y ) d y f ( q y ) f ( q y ) h ( q y ) ( y − 6 − 1 ) . (79) \mathcal{F}(q)=q\int_0^1\frac{dy}{\sqrt{f(qy)h(qy)(y^{-6}-1)}}, \qquad\mathcal{J}(q)=q\int_0^1\frac{h(qy)\,dy}{f(qy)\sqrt{f(qy)h(qy)(y^{-6}-1)}}. \tag*{(79)} F ( q ) = q ∫ 0 1 f ( q y ) h ( q y ) ( y − 6 − 1 ) d y , J ( q ) = q ∫ 0 1 f ( q y ) f ( q y ) h ( q y ) ( y − 6 − 1 ) h ( q y ) d y . (79)
At fixed strip width, subtracting the equilibrium contribution from the shell-crossing entropy yields
S ( ℓ , t ) − S e q ( ℓ ) = 2 z t 3 [ ∫ z c z t d z z 3 h ( z ) z t 6 − z 6 + ∫ δ z c d z z 6 H ( z ) ] − 2 z b 3 ∫ δ z b d z z 3 f ( z ) h ( z ) z b 6 − z 6 . (80) S(\ell,t)-S_{\mathrm{eq}}(\ell)=2z_t^3\left[\int_{z_c}^{z_t}\frac{dz}{z^3\sqrt{h(z)}\sqrt{z_t^6-z^6}}+\int_{\delta}^{z_c}\frac{dz}{z^6\sqrt{H(z)}}\right]
-2z_b^3\int_{\delta}^{z_b}\frac{dz}{z^3\sqrt{f(z)h(z)}\sqrt{z_b^6-z^6}}. \tag*{(80)} S ( ℓ , t ) − S eq ( ℓ ) = 2 z t 3 [ ∫ z c z t z 3 h ( z ) z t 6 − z 6 d z + ∫ δ z c z 6 H ( z ) d z ] − 2 z b 3 ∫ δ z b z 3 f ( z ) h ( z ) z b 6 − z 6 d z . (80)
From Eq. (78) we see that the sign of ω \omega ω determines the behavior near the saturation time. For ω < 0 \omega<0 ω < 0 , the HRT surface can approach the equilibrium configuration from earlier times, whereas for ω > 0 \omega>0 ω > 0 it cannot. If ω = 0 \omega=0 ω = 0 , higher orders in ϵ \epsilon ϵ are required [15 ] . For the final static surface, the ingoing time at the tip is t − ∫ δ z b d z / f ( z ) t-\int_{\delta}^{z_b} dz/f(z) t − ∫ δ z b d z / f ( z ) . Hence, the whole surface lies in the final black-brane region only for t ≥ t s ( c o n t ) t\geq t_s^{(\mathrm{cont})} t ≥ t s ( cont ) . Assuming continuous saturation, we have3 ^{3} 3
S e q ( ℓ ) − S ( ℓ , t ) ∝ ϵ 2 ∝ [ t s ( c o n t ) − t ] 2 . (81) S_{\mathrm{eq}}(\ell)-S(\ell,t)\propto\epsilon^2\propto\left[t_s^{(\mathrm{cont})}-t\right]^2. \tag*{(81)} S eq ( ℓ ) − S ( ℓ , t ) ∝ ϵ 2 ∝ [ t s ( cont ) − t ] 2 . (81)
For sufficiently large strips, several extremal surfaces coexist in the saturation regime, producing a characteristic swallow-tail structure in the entropy as a function of boundary time. At saturation, the physical HRT surface may switch between competing branches when their areas cross and the equilibrium branch becomes globally minimal [10 ] . Although the transition between competing branches may be discontinuous at the level of the time derivative, the entropy itself remains continuous across the transition.
The left panel of Figure 5 shows the dependence of t − t s ( c o n t ) t-t_s^{(\mathrm{cont})} t − t s ( cont ) on ϵ \epsilon ϵ for the shell-crossing extremal surfaces. Near ϵ = 0 \epsilon=0 ϵ = 0 , the initial behavior is governed by Eq. (78) . The subsequent nonmonotonic behavior observed in the numerical curves cannot be inferred from this leading-order expansion alone. As the noncommutativity parameter a a a increases, the maximum of the plotted time difference shifts toward larger values of ϵ \epsilon ϵ . The right panel shows the difference between the shell-crossing entropy and the equilibrium entropy at fixed strip width. The intersections indicate candidate transitions between the shell-crossing and equilibrium configurations. Identifying the physical saturation point requires selecting the globally minimal admissible HRT surface at the same boundary time and strip width.
Figure 5. Left: The time difference t − t s ( c o n t ) t-t_s^{(\mathrm{cont})} t − t s ( cont ) . Right: The entropy difference S − S e q S-S_{\mathrm{eq}} S − S eq . The dashed lines mark S = S e q S=S_{\mathrm{eq}} S = S eq .
To conclude this section, we discuss the relation between the entanglement velocity at the saturation point and the butterfly velocity v B v_B v B which characterizes the spatial spread of perturbations [34 ] . The roles of entanglement and butterfly velocities have also been investigated in anisotropic black-brane geometries [35 ] . For noncommutative Yang–Mills theory, v B v_B v B was computed in [29 ] . Because the geometry is anisotropic, the result is direction dependent. In our notation,
v B , x 1 2 = 2 3 , v B , x 2 2 = v B , x 3 2 = z h 4 + a 4 z h 4 v B , x 1 2 . (82) v_{B,x_1}^{2}=\frac{2}{3},\qquad v_{B,x_2}^{2}=v_{B,x_3}^{2}=\frac{z_h^{4}+a^{4}}{z_h^{4}}v_{B,x_1}^{2}. \tag*{(82)} v B , x 1 2 = 3 2 , v B , x 2 2 = v B , x 3 2 = z h 4 z h 4 + a 4 v B , x 1 2 . (82)
In terms of v B , x 2 v_{B,x_2} v B , x 2 from the above result, Eq. (75) becomes
t s ( c o n t ) = ℓ 2 v B , x 2 + ⋯ . (83) t_s^{(\mathrm{cont})}=\frac{\ell}{2v_{B,x_2}}+\cdots. \tag*{(83)} t s ( cont ) = 2 v B , x 2 ℓ + ⋯ . (83)
For sufficiently large a / z h a/z_h a / z h , v B , x 2 > 1 v_{B,x_2}>1 v B , x 2 > 1 . Since NCYM is spatially nonlocal and Lorentz invariance is broken in the noncommutative directions, this does not imply superluminal signaling in a local relativistic theory [29 ] . Note that Equation (83) does not determine the physical saturation time when competing HRT branches are present. The relation between butterfly propagation and entanglement saturation in local holographic theories has been discussed in [36 ] . In the commutative limit, v B , x 2 → 2 / 3 v_{B,x_2}\to\sqrt{2/3} v B , x 2 → 2/3 , in agreement with the AdS5 _5 5 result [15 ] . Figure 6 illustrates the dependence of these velocities on a a a . The left panel shows that the butterfly velocity along the noncommutative directions increases monotonically with a a a , whereas its component along the commutative direction remains fixed. Further, the right panel shows that the entanglement velocity increases monotonically with the noncommutativity parameter a a a .
Figure 6. Left: Butterfly velocity as a function of the noncommutative parameter. The blue curve shows v B , x 2 2 = v B , x 3 2 v_{B,x_2}^{2}=v_{B,x_3}^{2} v B , x 2 2 = v B , x 3 2 , while the dashed line shows v B , x 1 2 = 2 / 3 v_{B,x_1}^{2}=2/3 v B , x 1 2 = 2/3 . Right: Entanglement velocity as a function of a a a . The dashed line shows the commutative value.
Numerical results
In this subsection we examine the time evolution of extremal surfaces using the smooth Vaidya profile of Eq. (16) . Figure 7 displays the HRT configurations for fixed strip parameters at different boundary times. The early-time and intermediate-time surfaces probe both the v < 0 v<0 v < 0 and v > 0 v>0 v > 0 portions of the smooth Vaidya geometry. At sufficiently late times, the extremal surfaces approach the equilibrium configuration of the final black brane. These profiles illustrate the geometric evolution of the extremal surfaces but do not by themselves demonstrate the coexistence of competing HRT branches. Establishing such competition requires comparing the regulated areas of distinct extremal surfaces at the same strip width and boundary time, as discussed in § 4.3 .
Figure 7. Extremal surfaces for the smooth shell at a = 0.3 a = 0.3 a = 0.3 . Left: The radial profiles z ( x ) z(x) z ( x ) ; the dashed horizontal lines mark the cutoff and the final horizon, and z c z_c z c denotes the crossing of v = 0 v = 0 v = 0 . Right: The corresponding ingoing-time profiles v ( x ) v(x) v ( x ) ; the dashed line marks v = 0 v = 0 v = 0 .
Figure 8 shows the time dependence of the entanglement entropy for several values of the parameters. The entropy increases with time and approaches its equilibrium value monotonically. As a a a increases, Δ S \Delta S Δ S grows and the curve shifts slightly toward earlier times. The latter behavior is consistent with the dependence of the entanglement velocity on the noncommutativity scale, as given by Eq. (70) , and with the increase of v E v_{E} v E reported in the right panel of Figure 6 . For large regions, the selected HRT branch can change at saturation; this need not produce a jump in the entropy, although it can produce a kink. In particular, the right panel shows the evolution for a fixed value of a a a in the large ℓ \ell ℓ limit. As ℓ \ell ℓ increases, the linear-growth regime becomes more pronounced. The curves nearly coincide until the smaller strips begin to saturate. Once a strip reaches equilibrium, its entropy becomes constant, while the wider strips continue along nearly the same growing curve. At early times, this behavior is consistent with the width-independent coefficient in Eq. (48) . In the linear regime, it agrees with Eq. (70) , whose leading slope does not depend on ℓ \ell ℓ . Increasing the strip width therefore extends the linear-growth regime and delays saturation without changing the leading growth rate. The numerical curves mainly display the full evolution of the entropy up to saturation. On this scale, a very short initial logarithmic regime may be obscured by the subsequent stages of growth. Furthermore, the temporal resolution of the numerical data may be insufficient to identify such a short regime.
Figure 8. Time evolution of the entanglement entropy for several values of the strip widths and noncommutative scale. The left panel corresponds to short regions, while the middle and right panels correspond to large regions.
The left panel of Figure 9 shows the parametric plots of ( z c ( ℓ , t ) , z t ( ℓ , t ) ) (z_{c}(\ell,t),z_{t}(\ell,t)) ( z c ( ℓ , t ) , z t ( ℓ , t )) for several values of ℓ \ell ℓ . Increasing the strip width shifts the curves toward larger turning points, so that wider strips probe deeper regions of the bulk. For sufficiently wide strips, the shell-crossing positions approach a common limiting region, consistent with the critical scale of the thin-shell analysis in § 4.2 . At finite strip width and nonzero shell thickness, the agreement with z c ∗ ( z t ) z_c^*(z_t) z c ∗ ( z t ) is approximate. The right panel of Figure 9 illustrates the same functions for several values of the noncommutative scale at fixed strip width. At a given z c z_c z c , increasing a a a shifts the curve toward slightly smaller z t z_t z t . The corresponding extremal surfaces therefore become slightly shallower as the noncommutative scale is increased. This effect is small compared with the shift produced by changing the strip width.
Figure 9. z c z_c z c as a function of z t z_t z t . Left: For several values of the strip width. Right: For several values of the noncommutative scale. For wide strips, the shell-crossing positions approach the thin-shell critical scale approximately.
Mixed-state correlations after the quench
Having established the time evolution of the entanglement entropy for a single strip, we now turn to mixed-state correlation measures. We focus on the mutual information and the entanglement wedge cross section, which probe correlations between two disjoint regions. The entanglement wedge cross section has been linked to several mixed-state information measures, including entanglement of purification, reflected entropy, and odd entanglement entropy. Logarithmic negativity provides another measure of mixed-state entanglement [37 ] , and holographic proposals involving entanglement-wedge cross sections have also been explored for this quantity [38 ] . The setup is the same as before: the noncommutative scale a a a is fixed, and the time dependence enters only through the Vaidya mass function. Since both quantities involve extremal surfaces, the techniques developed in the previous section provide a basis for the analysis below.
Mutual information
The mutual information between two disjoint regions A A A and B B B , which measures their total correlations, is defined as
I ( A : B ) = S ( A ) + S ( B ) − S ( A ∪ B ) . (84) I(A : B) = S(A) + S(B) - S(A \cup B). \tag*{(84)} I ( A : B ) = S ( A ) + S ( B ) − S ( A ∪ B ) . (84)
In local quantum field theories, the mutual information between disjoint regions is ultraviolet finite [39 ] . Holographically, it is determined by the connected and disconnected RT/HRT configurations. The transition between these configurations gives rise to a phase transition in the large-N N N limit [39 , 40 ] . Holographic mutual information in static NCYM was studied in [26 ] .
We consider two disjoint strips A A A and B B B of equal width ℓ \ell ℓ , separated by a distance d d d . Their finite directions lie along the noncommutative coordinate x 2 x_2 x 2 . The two candidate HRT configurations are shown in Figure 10 . We work in the regime { d , ℓ } > ℓ c \{d, \ell\} > \ell_c { d , ℓ } > ℓ c , where ℓ c \ell_c ℓ c is the cutoff-dependent scale of the final black-brane geometry defined in Eq. (44) . For A ∪ B A \cup B A ∪ B , we consider two non-crossing HRT configurations. The disconnected configuration consists of two surfaces of width ℓ \ell ℓ . The connected configuration is formed by the surfaces of widths d d d and 2 ℓ + d 2\ell+ d 2 ℓ + d . Hence,
Figure 10. Schematic connected and disconnected HRT configurations for the two-strip setup. The blue segment denotes the cross section of the connected entanglement wedge and is discussed in § 5.2 .
S ( A ∪ B ; t ) = min { 2 S ( ℓ , t ) , S ( d , t ) + S ( 2 ℓ + d , t ) } . (85) S(A \cup B;t) = \min\{2S(\ell, t), S(d, t) + S(2\ell+ d, t)\}. \tag*{(85)} S ( A ∪ B ; t ) = min { 2 S ( ℓ , t ) , S ( d , t ) + S ( 2 ℓ + d , t )} . (85)
Using the definition in Eq. (84) , the holographic mutual information is therefore
I ( A : B ; t ) = max { 0 , 2 S ( ℓ , t ) − S ( d , t ) − S ( 2 ℓ + d , t ) } . (86) I(A : B;t) = \max\{0, 2S(\ell, t) - S(d, t) - S(2\ell+ d, t)\}. \tag*{(86)} I ( A : B ; t ) = max { 0 , 2 S ( ℓ , t ) − S ( d , t ) − S ( 2 ℓ + d , t )} . (86)
If the connected configuration is dominant, the mutual information is nontrivial. Otherwise, the disconnected configuration is dominant and the leading classical contribution to the mutual information vanishes. At the transition point, the two configurations are degenerate [39 ] . We divide the evolution of the holographic mutual information into three parts and discuss each separately.
Early time. We consider the regime ℓ ≫ d \ell\gg d ℓ ≫ d where the connected HRT configuration is dominant during the time interval under consideration. At early times, the coefficient of the leading logarithmic term in Eq. (48) is independent of the strip width. Consequently, this contribution cancels in the connected combination in Eq. (86) . The leading time dependence of the mutual information is therefore determined by subleading width-dependent terms, which are not fixed by the leading-order result for the entanglement entropy.
Intermediate time. In this regime, Eq. (70) gives the same leading growth rate s t h A Σ v E s_{\mathrm{th}}A_{\Sigma}v_{E} s th A Σ v E for each strip. When all three entropies in Eq. (86) lie in the common linear regime, the leading linear terms cancel and the mutual information is approximately constant. The entropies for regions of length d d d , ℓ \ell ℓ , and 2 ℓ + d 2\ell+d 2 ℓ + d then reach their equilibrium values successively. After S ( d , t ) S(d,t) S ( d , t ) has saturated while the other two entropies continue to grow linearly, the slope of the mutual information becomes s t h A Σ v E s_{\mathrm{th}}A_{\Sigma}v_{E} s th A Σ v E . Once the width-ℓ \ell ℓ entropy also becomes approximately constant while the 2 ℓ + d 2\ell+d 2 ℓ + d entropy continues to grow, the slope becomes − s t h A Σ v E -s_{\mathrm{th}}A_{\Sigma}v_{E} − s th A Σ v E . The mutual information can therefore increase and then decrease, with a maximum between these two regimes. The crossover times depend on the strip widths, their separation, and the parameters of the quench.
Late time. At late times, the HRT surfaces approach their final black-brane configurations. The equilibrium mutual information follows from Eq. (86) after replacing the strip entropies by the corresponding equilibrium values. Here we refer to the leading classical contribution described by the HRT areas; bulk correlations of order N 0 N^{0} N 0 are not included. To characterize the connected-to-disconnected transition in the final equilibrium state, we determine the phase transition point from the condition
2 S e q ( ℓ ) − S e q ( d ) − S e q ( 2 ℓ + d ) = 0. 2S_{\mathrm{eq}}(\ell)-S_{\mathrm{eq}}(d)-S_{\mathrm{eq}}(2\ell+d)=0. 2 S eq ( ℓ ) − S eq ( d ) − S eq ( 2 ℓ + d ) = 0.
Figure 11 shows the resulting phase diagram in the ( T d , d / ℓ ) (Td,d/\ell) ( T d , d / ℓ ) plane for fixed d = 0.2 d=0.2 d = 0.2 and different values of the noncommutative scale a a a . For each curve, the region below the phase boundary corresponds to nonzero equilibrium mutual information, whereas above it the disconnected configuration dominates and the leading holographic mutual information vanishes. Increasing the noncommutativity parameter shifts the phase boundary toward larger values of T d Td T d and d / ℓ d/\ell d / ℓ , thereby enlarging the region of nonzero mutual information within the parameter range considered.
Figure 11. Phase diagram of the holographic mutual information in the (T d Td T d , d / ℓ d/\ell d / ℓ ) plane for fixed d d d . The shaded region corresponds to nonzero mutual information.
In Figure 12 we show the evolution of the mutual information for several values of the parameters. We restrict ourselves to the range where the connected configuration is always dominant.
Figure 12. Evolution of the holographic mutual information I ( A : B ) I(A:B) I ( A : B ) for several values of the parameters.
At sufficiently early times, the mutual information approaches its vacuum value. It then increases, reaches a maximum, and decreases toward its final equilibrium value. For all curves shown, the equilibrium value is smaller than the vacuum value, since the thermal medium weakens the correlations. At fixed separation, the mutual information increases with the noncommutative scale throughout the evolution, with larger vacuum, maximum, and equilibrium values for larger a a a over the range shown. This trend is consistent with the static results of [26 ] . For the configurations shown in this figure, the mutual information remains positive at late times, so the transition to the disconnected configuration does not occur.
Entanglement wedge cross section
Another holographic probe of correlations between two disjoint regions is the entanglement wedge cross section (EWCS). For a connected entanglement wedge in a static geometry, it is defined as the minimal codimension-two cross section separating the portions of the wedge associated with A A A and B B B [41 , 42 ]
E W ( A : B ) = Area ( Σ A ∪ B ) 4 G N . (87) E_W(A:B)=\frac{\operatorname{Area}(\Sigma_{A\cup B})}{4G_N}. \tag*{(87)} E W ( A : B ) = 4 G N Area ( Σ A ∪ B ) . (87)
A notable property of this proposal is that the EWCS undergoes a discontinuous phase transition as the separation between the two subregions increases. As for the mutual information, the transition originates from the competition between the connected and disconnected entanglement-wedge configurations. When the separation is sufficiently large and the disconnected configuration is favored, the entanglement wedge is disconnected and E W E_W E W vanishes. This quantity has been proposed as a holographic dual of the entanglement of purification, a correlation measure for mixed states [41 , 42 , 43 ] . Further geometric and thermal properties of the EWCS have been investigated in a variety of holographic backgrounds [44 , 45 ] .
To find the evolution of EWCS we again consider the symmetric configuration depicted in Figure 10 . In the connected phase, the entanglement wedge is bounded by the HRT surfaces associated with regions of widths d d d and 2 ℓ + d 2\ell+d 2 ℓ + d . We denote the corresponding turning points by z b z_b z b and z u z_u z u , respectively. Reflection symmetry admits a symmetric cross-section candidate on the plane x 2 = 0 x_2=0 x 2 = 0 , whose nontrivial profile lies in the ( v , z ) (v,z) ( v , z ) plane. During the evolution, three geometric configurations are relevant [46 , 47 ] (see Figure 13 ). Before the null shell intersects the cross section, the connected EWCS lies entirely in the vacuum region. At intermediate times, it crosses the shell and contains vacuum and black-brane portions; the latter may itself consist of two radial legs. Finally, if the connected branch remains dominant, the cross section can enter entirely into the final black-brane geometry.
Figure 13. The three geometric configurations relevant for the EWCS evolution. The orange curves denote the HRT surfaces bounding the connected entanglement wedge, the blue curve denotes the EWCS, and the dashed line represents the infalling null shell. In the shell-crossing configuration the EWCS intersects the shell at z = z w z=z_w z = z w .
To determine the connected cross section, we parameterize it by v = v ( z ) v=v(z) v = v ( z ) . Using the same normalization as in the previous sections, the corresponding area functional is given by
E W ( t ) = ∫ z b z u d z z 3 h ( z ) − f ( v , z ) v ′ ( z ) 2 − 2 v ′ ( z ) . (88) E_W(t)=\int_{z_b}^{z_u}\frac{\mathrm{d}z}{z^3\sqrt{h(z)}}\sqrt{-f(v,z)v'(z)^2-2v'(z)}. \tag*{(88)} E W ( t ) = ∫ z b z u z 3 h ( z ) d z − f ( v , z ) v ′ ( z ) 2 − 2 v ′ ( z ) . (88)
where a prime denotes differentiation with respect to z z z . The factor 1 / h ( z ) 1/\sqrt{h(z)} 1/ h ( z ) follows from the noncommutative warp factor together with the string-frame dilaton contribution and distinguishes the present functional from its AdS counterpart. Extremizing Eq. (88) gives
d d z [ 1 + f ( v , z ) v ′ z 3 h ( z ) − f ( v , z ) v ′ 2 − 2 v ′ ] = ∂ v f ( v , z ) v ′ 2 2 z 3 h ( z ) − f ( v , z ) v ′ 2 − 2 v ′ (89) \frac{d}{dz}\left[\frac{1+f(v,z)v'}{z^3\sqrt{h(z)}\sqrt{-f(v,z)v'^2-2v'}}\right]=\frac{\partial_v f(v,z)v'^2}{2z^3\sqrt{h(z)}\sqrt{-f(v,z)v'^2-2v'}} \tag*{(89)} d z d [ z 3 h ( z ) − f ( v , z ) v ′2 − 2 v ′ 1 + f ( v , z ) v ′ ] = 2 z 3 h ( z ) − f ( v , z ) v ′2 − 2 v ′ ∂ v f ( v , z ) v ′2 (89)
The profile is anchored on the two HRT surfaces through
v ( z b ) = v b , v ( z u ) = v u . (90) v(z_b)=v_b,\qquad v(z_u)=v_u. \tag*{(90)} v ( z b ) = v b , v ( z u ) = v u . (90)
Away from the null shell, ∂ v f = 0 \partial_v f=0 ∂ v f = 0 , and hence the momentum conjugate to v v v is conserved separately in the two static regions. We define the corresponding canonical momenta as
π ± = − 1 + f ± v ± ′ z 3 h ( z ) − f ± v ± ′ 2 − 2 v ± ′ . (91) \pi_\pm=-\frac{1+f_\pm v'_\pm}{z^3\sqrt{h(z)}\sqrt{-f_\pm v_\pm'^2-2v'_\pm}}. \tag*{(91)} π ± = − z 3 h ( z ) − f ± v ± ′2 − 2 v ± ′ 1 + f ± v ± ′ . (91)
Here the subscripts − - − and + + + refer to the vacuum region v < 0 v<0 v < 0 and the final black-brane region v > 0 v>0 v > 0 , respectively. Solving Eq. (91) for the profile gives
v ± ′ ( z ) = − 1 f ± ( z ) [ 1 + π ± z 3 h ( z ) V ± ( z ) ] , V ± ( z ) = f ± ( z ) + π ± 2 z 6 h ( z ) . (92) v'_\pm(z)=-\frac{1}{f_\pm(z)}\left[1+\frac{\pi_\pm z^3\sqrt{h(z)}}{\sqrt{\mathcal{V}_\pm(z)}}\right],\qquad\mathcal{V}_\pm(z)=f_\pm(z)+\pi_\pm^2z^6h(z). \tag*{(92)} v ± ′ ( z ) = − f ± ( z ) 1 [ 1 + V ± ( z ) π ± z 3 h ( z ) ] , V ± ( z ) = f ± ( z ) + π ± 2 z 6 h ( z ) . (92)
In the thin-shell limit, the extremal surface can be analyzed in different regions of the Vaidya geometry, as discussed separately below.
Vacuum region (v < 0 v<0 v < 0 ). Using Eq. (92) and imposing the boundary condition v ( z u ) = v u v(z_u)=v_u v ( z u ) = v u , the vacuum profile becomes
v − ( z ) = v u + z u − z + π − ∫ z z u d ζ ζ 5 ζ 4 + a 4 + π − 2 ζ 10 . (93) v_-(z)=v_u+z_u-z+\pi_-\int_z^{z_u}d\zeta\,\frac{\zeta^5}{\sqrt{\zeta^4+a^4+\pi_-^2\zeta^{10}}}. \tag*{(93)} v − ( z ) = v u + z u − z + π − ∫ z z u d ζ ζ 4 + a 4 + π − 2 ζ 10 ζ 5 . (93)
For a shell-crossing configuration, continuity of the profile at z = z w z=z_w z = z w requires v ( z w ) = 0 v(z_w)=0 v ( z w ) = 0 .
Black-brane region (v > 0 v>0 v > 0 ). In this case integrating Eq. (92) and imposing v ( z b ) = v b v(z_b)=v_b v ( z b ) = v b , we find
v + ( z ) = v b − ∫ z b z d ζ V + ( ζ ) [ V + ( ζ ) − π + ζ 3 h ( ζ ) ] . (94) v_+(z)=v_b-\int_{z_b}^{z}\frac{d\zeta}{\sqrt{\mathcal{V}_+(\zeta)}\left[\sqrt{\mathcal{V}_+(\zeta)}-\pi_+\zeta^3\sqrt{h(\zeta)}\right]}. \tag*{(94)} v + ( z ) = v b − ∫ z b z V + ( ζ ) [ V + ( ζ ) − π + ζ 3 h ( ζ ) ] d ζ . (94)
Using the matching condition at the shell the above equation gives
v b = ∫ z b z w d ζ V + ( ζ ) [ V + ( ζ ) − π + ζ 3 h ( ζ ) ] . (95) v_b=\int_{z_b}^{z_w}\frac{d\zeta}{\sqrt{\mathcal{V}_+(\zeta)}\left[\sqrt{\mathcal{V}_+(\zeta)}-\pi_+\zeta^3\sqrt{h(\zeta)}\right]}. \tag*{(95)} v b = ∫ z b z w V + ( ζ ) [ V + ( ζ ) − π + ζ 3 h ( ζ ) ] d ζ . (95)
For a segment entirely outside the horizon, the same condition may also be written in the separated form
0 = v b + G ( z b ) − G ( z w ) − π + ∫ z b z w d ζ ζ 5 f ( ζ ) f ( ζ ) ( ζ 4 + a 4 ) + π + 2 ζ 10 , (96) 0=v_b+G(z_b)-G(z_w)-\pi_+\int_{z_b}^{z_w}d\zeta\,\frac{\zeta^5}{f(\zeta)\sqrt{f(\zeta)(\zeta^4+a^4)+\pi_+^2\zeta^{10}}}, \tag*{(96)} 0 = v b + G ( z b ) − G ( z w ) − π + ∫ z b z w d ζ f ( ζ ) f ( ζ ) ( ζ 4 + a 4 ) + π + 2 ζ 10 ζ 5 , (96)
where
G ( z ) ≡ ∫ δ z d ζ f ( ζ ) . (97) G(z)\equiv\int_{\delta}^{z}\frac{d\zeta}{f(\zeta)}. \tag*{(97)} G ( z ) ≡ ∫ δ z f ( ζ ) d ζ . (97)
Moreover, the endpoint conditions for the vacuum and black-brane segments can also be expressed as
v b = − ∫ z b z w d z v + ′ ( z ) , v u = ∫ z w z u d z v − ′ ( z ) . (98) v_b=-\int_{z_b}^{z_w} dz\,v_+'(z),\qquad v_u=\int_{z_w}^{z_u} dz\,v_-'(z). \tag*{(98)} v b = − ∫ z b z w d z v + ′ ( z ) , v u = ∫ z w z u d z v − ′ ( z ) . (98)
Since the inner endpoint belongs to the final static configuration, the corresponding boundary time satisfies
t = v b + G ( z b ) . (99) t=v_b+G(z_b). \tag*{(99)} t = v b + G ( z b ) . (99)
Further, combining this relation with Eq. (95) gives the boundary time in terms of the shell-crossing point,
t = G ( z b ) + ∫ z b z w d z V + ( z ) [ V + ( z ) − π + z 3 h ( z ) ] . (100) t=G(z_b)+\int_{z_b}^{z_w}\frac{dz}{\sqrt{\mathcal{V}_+(z)}\left[\sqrt{\mathcal{V}_+(z)}-\pi_+z^3\sqrt{h(z)}\right]}. \tag*{(100)} t = G ( z b ) + ∫ z b z w V + ( z ) [ V + ( z ) − π + z 3 h ( z ) ] d z . (100)
Matching at the null shell. Because f ( v , z ) f(v,z) f ( v , z ) jumps at v = 0 v=0 v = 0 , the conserved quantities on the two sides of the shell are in general different. Integrating the corresponding equation of motion through the infinitesimal shell gives
d z d v ∣ + − d z d v ∣ − = g ( z w ) 2 , g ( z w ) = ( z w z h ) 4 . (101) \left.\frac{dz}{dv}\right|_+-\left.\frac{dz}{dv}\right|_-=\frac{g(z_w)}{2},\qquad g(z_w)=\left(\frac{z_w}{z_h}\right)^4. \tag*{(101)} d v d z + − d v d z − = 2 g ( z w ) , g ( z w ) = ( z h z w ) 4 . (101)
For the monotonic branch of Eq. (92) , the above condition can be expressed as
V + ( z w ) − π + z w 3 h ( z w ) = V − ( z w ) − π − z w 3 h ( z w ) . (102) \sqrt{\mathcal{V}_+(z_w)}-\pi_+z_w^3\sqrt{h(z_w)}=\sqrt{\mathcal{V}_-(z_w)}-\pi_-z_w^3\sqrt{h(z_w)}. \tag*{(102)} V + ( z w ) − π + z w 3 h ( z w ) = V − ( z w ) − π − z w 3 h ( z w ) . (102)
Solving this relation for π + \pi_+ π + gives
π + = ( 1 − g ( z w ) 2 ) π − − g ( z w ) 2 z w 3 h ( z w ) 1 + π − 2 z w 6 h ( z w ) . (103) \pi_+=\left(1-\frac{g(z_w)}{2}\right)\pi_--\frac{g(z_w)}{2z_w^3\sqrt{h(z_w)}}\sqrt{1+\pi_-^2z_w^6h(z_w)}. \tag*{(103)} π + = ( 1 − 2 g ( z w ) ) π − − 2 z w 3 h ( z w ) g ( z w ) 1 + π − 2 z w 6 h ( z w ) . (103)
Evaluating the contributions from the vacuum and black-brane regions separately, the entanglement wedge cross section takes the form
E W = ∫ z b z w d z z 3 h ( z ) V + ( z ) + ∫ z w z u d z z 3 h ( z ) 1 + π − 2 z 6 h ( z ) . (104) E_W=\int_{z_b}^{z_w}\frac{dz}{z^3\sqrt{h(z)}\sqrt{\mathcal{V}_+(z)}}+\int_{z_w}^{z_u}\frac{dz}{z^3\sqrt{h(z)}\sqrt{1+\pi_-^2z^6h(z)}}. \tag*{(104)} E W = ∫ z b z w z 3 h ( z ) V + ( z ) d z + ∫ z w z u z 3 h ( z ) 1 + π − 2 z 6 h ( z ) d z . (104)
When the cross section lies entirely in the vacuum region, the extremal surface has π − = 0 \pi_-=0 π − = 0 , and its area becomes
E W v a c = ∫ z b z u d z z 3 h ( z ) . (105) E_W^{\mathrm{vac}}=\int_{z_b}^{z_u}\frac{dz}{z^3\sqrt{h(z)}}. \tag*{(105)} E W vac = ∫ z b z u z 3 h ( z ) d z . (105)
Before turning to the time evolution of the EWCS, it is useful to examine several properties of π − \pi_- π − and π + \pi_+ π + that influence its behavior. In what follows, we mainly consider the large-region regime. We first examine the approximation π − ≃ 0 \pi_-\simeq0 π − ≃ 0 , which considerably simplifies the analytic treatment of E W E_W E W . The null shell intersects the EWCS at z w z_w z w and the outer HRT surface at z c z_c z c , which are generally distinct. Using z c = z u + v u z_c=z_u+v_u z c = z u + v u and Eq. (93) we find
z w − z c = π − ∫ z w z u d ζ ζ 5 ζ 4 + a 4 + π − 2 ζ 10 . (106) z_w-z_c=\pi_-\int_{z_w}^{z_u}d\zeta\,\frac{\zeta^5}{\sqrt{\zeta^4+a^4+\pi_-^2\zeta^{10}}}. \tag*{(106)} z w − z c = π − ∫ z w z u d ζ ζ 4 + a 4 + π − 2 ζ 10 ζ 5 . (106)
In the large-region limit, where z w ≃ z c ≪ z u z_w \simeq z_c \ll z_u z w ≃ z c ≪ z u , the above result allows us to consider π − ≃ 0 \pi_- \simeq0 π − ≃ 0 . Indeed, for fixed a a a and finite z w z_w z w , the integral in Eq. (106) scales as z u 4 z_u^4 z u 4 and hence π − = O ( z u − 4 ) \pi_- = O(z_u^{-4}) π − = O ( z u − 4 ) in this limit. However, this approximation must also satisfy the matching condition. On the monotonic branch specified by Eq. (102) , taking π − → 0 \pi_- \to0 π − → 0 requires g ( z w ) ≤ 2 g(z_w) \le2 g ( z w ) ≤ 2 . On the returning branch the sign of the square root is reversed, allowing π − → 0 \pi_- \to0 π − → 0 when g ( z w ) > 2 g(z_w) > 2 g ( z w ) > 2 . In both cases the algebraic relation for π + \pi_+ π + is Eq. (103) , provided the corresponding geometric root is chosen. In large-region approximation, Eq. (103) gives
π + ≃ − g ( z w ) 2 z w 3 h ( z w ) = − z u 4 + a 4 2 z w z u 4 , ( z w ≪ z u ) . (107) \pi_+ \simeq-\frac{g(z_w)}{2z_w^3\sqrt{h(z_w)}} = -\frac{\sqrt{z_u^4+a^4}}{2z_w z_u^4}, \qquad(z_w \ll z_u). \tag*{(107)} π + ≃ − 2 z w 3 h ( z w ) g ( z w ) = − 2 z w z u 4 z u 4 + a 4 , ( z w ≪ z u ) . (107)
For g ( z w ) > 2 g(z_w)>2 g ( z w ) > 2 , the returning matching sign must be used; the monotonic branch condition cannot be extrapolated into that regime. In fact, in the π − → 0 \pi_- \to0 π − → 0 limit, the matching condition with z ˙ − = − 1 \dot{z}_-=-1 z ˙ − = − 1 gives z ˙ + = g ( z w ) / 2 − 1 \dot{z}_+=g(z_w)/2-1 z ˙ + = g ( z w ) /2 − 1 . If g ( z w ) > 2 g(z_w)>2 g ( z w ) > 2 , a curve that begins at the shell and eventually reaches z b < z h < z w z_b<z_h<z_w z b < z h < z w must turn at some z r > z w z_r>z_w z r > z w ; a globally monotonic z ( v ) z(v) z ( v ) trajectory cannot meet both endpoint conditions. However, when z w ≃ z u z_w \simeq z_u z w ≃ z u , we cannot assume π − ≃ 0 \pi_- \simeq0 π − ≃ 0 . Indeed, in this situation π + ≃ 0 \pi_+ \simeq0 π + ≃ 0 and π − ≠ 0 \pi_- \ne0 π − = 0 . To see this, note that as z w → z u z_w \to z_u z w → z u , the outer HRT surface settles down to its final static configuration as it lies entirely in the black-brane geometry, and hence t ≃ v u + G ( z u ) t \simeq v_u+G(z_u) t ≃ v u + G ( z u ) . On the other hand, this expression must be consistent with Eq. (100) as z w ≃ z u z_w \simeq z_u z w ≃ z u . Since the final static turning point lies outside the horizon, we obtain
v u = π + ∫ z b z u d ζ ζ 5 f ( ζ ) f ( ζ ) ( ζ 4 + a 4 ) + π + 2 ζ 10 . (108) v_u=\pi_+\int_{z_b}^{z_u}d\zeta\frac{\zeta^5}{f(\zeta)\sqrt{f(\zeta)(\zeta^4+a^4)+\pi_+^2\zeta^{10}}}. \tag*{(108)} v u = π + ∫ z b z u d ζ f ( ζ ) f ( ζ ) ( ζ 4 + a 4 ) + π + 2 ζ 10 ζ 5 . (108)
As z w → z u z_w \to z_u z w → z u , the shell-crossing condition implies v u → 0 v_u \to0 v u → 0 . Then the above equation gives π + → 0 \pi_+ \to0 π + → 0 . In this situation, solving Eq. (103) for π + = 0 \pi_+=0 π + = 0 gives
π − 2 ≃ g ( z u ) 2 4 f ( z u ) z u 6 h ( z u ) = z u 4 + a 4 4 z u 2 z h 4 ( z h 4 − z u 4 ) , ( z w ≃ z u ) . (109) \pi_-^2 \simeq\frac{g(z_u)^2}{4f(z_u)z_u^6h(z_u)}=\frac{z_u^4+a^4}{4z_u^2z_h^4(z_h^4-z_u^4)}, \qquad(z_w \simeq z_u). \tag*{(109)} π − 2 ≃ 4 f ( z u ) z u 6 h ( z u ) g ( z u ) 2 = 4 z u 2 z h 4 ( z h 4 − z u 4 ) z u 4 + a 4 , ( z w ≃ z u ) . (109)
As we will argue, this result allows us to explore the late-time behavior of E W E_W E W when the connected cross section lies entirely in the final black-brane region.
The numerical results for the time evolution of the EWCS are presented in Figure 14 . The left panel shows that increasing the strip separation suppresses E W E_W E W and can induce a discontinuous transition of the connected configuration to a vanishing EWCS. The middle panel illustrates its dependence on the noncommutativity scale, revealing a transient enhancement as the deformation increases. The right panel compares different strip widths and shows that the EWCS approaches a finite equilibrium value whenever the connected configuration remains dominant.
Figure 14. Time evolution of the entanglement wedge cross section for several values of the parameters.
To further elucidate the time evolution of the EWCS, we examine the radial positions z b z_b z b , z w z_w z w , and z u z_u z u . The left panel of Figure 15 shows their evolution with boundary time. The shell-crossing position z w z_w z w is displayed only when the EWCS intersects the null shell. The inner turning point z b z_b z b remains nearly constant, whereas the outer turning point z u z_u z u decreases toward its final equilibrium value. During the shell-crossing regime, z w z_w z w initially increases and then approaches z u z_u z u , while remaining within the interval z b < z w < z u z_b < z_w < z_u z b < z w < z u . As z w z_w z w approaches z u z_u z u , the vacuum contribution to the EWCS decreases and eventually vanishes. This behavior indicates that, for the configurations considered, the evolution of the EWCS is governed primarily by the outer HRT surface. The right panel of Figure 15 displays the corresponding parametric relation between z w z_w z w and z u z_u z u , with boundary time serving as an implicit parameter. The resulting trajectories are non-monotonic, and increasing the strip width extends them toward larger radial positions.
Figure 15. Left: Time evolution of z b z_b z b , z w z_w z w , and z u z_u z u . Right: Parametric curves ( z u ( t ) , z w ( t ) ) (z_u(t),z_w(t)) ( z u ( t ) , z w ( t )) for several strip widths.
The closed-form expression for E W E_W E W in Eq. (104) allows its scaling behavior to be extracted in different stages of the thermalization process. In the large-ℓ \ell ℓ limit, as in the case of entanglement entropy, the time evolution of the EWCS exhibits three distinct regimes: early-time growth, intermediate linear growth, and late-time saturation. These regimes are examined in turn below.
Early-time growth. At sufficiently early times, the infalling shell does not intersect the EWCS, which therefore retains its vacuum value. The initial evolution begins once the shell reaches the cross section. Following the approach of [46 ] , the shell-crossing position is parametrized as
z w = z b + η , 0 < η ≪ z b , z_w = z_b + \eta,\qquad0 < \eta\ll z_b, z w = z b + η , 0 < η ≪ z b ,
with the large-region hierarchy z b ≪ z h ≪ z u z_b \ll z_h \ll z_u z b ≪ z h ≪ z u . In this regime, the preceding analysis yields π − ≃ 0 \pi_- \simeq0 π − ≃ 0 and π + = O ( z h − 4 ) \pi_+ = \mathcal{O}(z_h^{-4}) π + = O ( z h − 4 ) , at fixed a a a and z w z_w z w . Expanding Eq. (104) then gives the leading thermal correction
Δ E W ( t ) = E W ( t ) − E W v a c ≃ 1 2 z h 4 ∫ z b z w d z z 4 + a 4 z . (110) \Delta E_W(t) = E_W(t) - E_W^{\mathrm{vac}} \simeq\frac{1}{2z_h^4}\int_{z_b}^{z_w} \mathrm{d}z\,\frac{\sqrt{z^4+a^4}}{z}. \tag*{(110)} Δ E W ( t ) = E W ( t ) − E W vac ≃ 2 z h 4 1 ∫ z b z w d z z z 4 + a 4 . (110)
This expression applies near the onset of shell crossing, where z w − z b ≪ z b z_w-z_b\ll z_b z w − z b ≪ z b . Moreover, expanding Eq. (100) at early times gives
t = z w − δ + O ( z w 5 z h 4 ) . (111) t=z_w-\delta+O\left(\frac{z_w^5}{z_h^4}\right). \tag*{(111)} t = z w − δ + O ( z h 4 z w 5 ) . (111)
The inner turning point z b z_b z b is determined by the strip separation d d d . To leading order in z b / z h z_b/z_h z b / z h , the corresponding vacuum width relation reads
d = 2 ∫ δ z b d z z z 4 + a 4 z b 6 − z 6 . (112) d=2\int_{\delta}^{z_b}dz\,\frac{z\sqrt{z^4+a^4}}{\sqrt{z_b^6-z^6}}. \tag*{(112)} d = 2 ∫ δ z b d z z b 6 − z 6 z z 4 + a 4 . (112)
In the limit δ ≪ a ≪ z b \delta\ll a\ll z_b δ ≪ a ≪ z b , the corresponding expansion takes the form
d ≃ B z b + a 4 z b 3 [ log ( z b a ) + 1 4 + 5 6 log 2 ] + O ( a 8 z b 7 ) , (113) d\simeq Bz_b+\frac{a^4}{z_b^3}\left[\log\left(\frac{z_b}{a}\right)+\frac{1}{4}+\frac{5}{6}\log2\right]+O\left(\frac{a^8}{z_b^7}\right), \tag*{(113)} d ≃ B z b + z b 3 a 4 [ log ( a z b ) + 4 1 + 6 5 log 2 ] + O ( z b 7 a 8 ) , (113)
with B = 1 3 B ( 2 3 , 1 2 ) B=\frac{1}{3}B\left(\frac{2}{3},\frac{1}{2}\right) B = 3 1 B ( 3 2 , 2 1 ) and B ( x , y ) B(x,y) B ( x , y ) is the Euler beta function. Considering the limit z b ≲ t ≪ z h z_b\lesssim t\ll z_h z b ≲ t ≪ z h and expanding the integrand of Eq. (110) in powers of a 4 / z 4 a^4/z^4 a 4 / z 4 , at leading order, we obtain
Δ E W ( t , d ) ≃ π A Σ 6 E [ t 2 − a 4 2 t 2 + ⋯ ] . (114) \Delta E_W(t,d)\simeq\frac{\pi A\Sigma}{6}\mathcal{E}\left[t^2-\frac{a^4}{2t^2}+\cdots\right]. \tag*{(114)} Δ E W ( t , d ) ≃ 6 π A Σ E [ t 2 − 2 t 2 a 4 + ⋯ ] . (114)
Therefore at early times EWCS grows quadratically and the rate of growth is a fixed constant proportional to the energy density. Note that the correction at order a 4 a^4 a 4 receives contributions from both the radial integrand and the noncommutative shift of the turning point at fixed separation. In the commutative limit, the above result agrees with the early time scaling of EWCS in relativistic setup [47 ] .4
Intermediate time. We now consider the intermediate-time regime d ≪ z h ≪ t ≪ ℓ d\ll z_h\ll t\ll\ell d ≪ z h ≪ t ≪ ℓ , where z b ≪ z w ≪ z u z_b\ll z_w\ll z_u z b ≪ z w ≪ z u . For the returning branch, the black-brane segment has a radial turning point z r z_r z r , where V + ( z r ) = 0 \mathcal{V}_+(z_r)=0 V + ( z r ) = 0 and z h < z w < z r < z m z_h<z_w<z_r<z_m z h < z w < z r < z m . The EWCS integral and the regular boundary-time relation obtained above are modified by the inclusion of the additional returning segment
E W + = ∫ z b z w d z z 3 h ( z ) V + ( z ) + 2 ∫ z w z r d z z 3 h ( z ) V + ( z ) , t = G ( z b ) + ∫ z b z w d z V + ( z ) [ V + ( z ) − π + z 3 h ( z ) ] + 2 ∫ z w z r π + z 3 h ( z ) d z f ( z ) V + ( z ) . (115) \begin{aligned}
E_W^+&=\int_{z_b}^{z_w}\frac{dz}{z^3\sqrt{h(z)}\sqrt{\mathcal{V}_+(z)}}+2\int_{z_w}^{z_r}\frac{dz}{z^3\sqrt{h(z)}\sqrt{\mathcal{V}_+(z)}},\\
t&=G(z_b)+\int_{z_b}^{z_w}\frac{dz}{\sqrt{\mathcal{V}_+(z)}\left[\sqrt{\mathcal{V}_+(z)}-\pi_+z^3\sqrt{h(z)}\right]}+2\int_{z_w}^{z_r}\frac{\pi_+z^3\sqrt{h(z)}\,dz}{f(z)\sqrt{\mathcal{V}_+(z)}}. \tag*{(115)}
\end{aligned} E W + t = ∫ z b z w z 3 h ( z ) V + ( z ) d z + 2 ∫ z w z r z 3 h ( z ) V + ( z ) d z , = G ( z b ) + ∫ z b z w V + ( z ) [ V + ( z ) − π + z 3 h ( z ) ] d z + 2 ∫ z w z r f ( z ) V + ( z ) π + z 3 h ( z ) d z . (115)
The corresponding expression for a single shell crossing is given in Eq. (100) ; the second integrals above account for the additional returning segment. Along this segment, the local momentum conjugate to v ( z ) v(z) v ( z ) changes sign, from π + \pi_+ π + to − π + -\pi_+ − π + , while π + < 0 \pi_+<0 π + < 0 continues to denote the constant Hamiltonian in the z ( v ) z(v) z ( v ) parametrization. The matching condition for the returning branch reads
V + ( z w ) + π + q ( z w ) = − [ V − ( z w ) − π − q ( z w ) ] , q ( z ) = z 3 h ( z ) . \sqrt{\mathcal{V}_+(z_w)}+\pi_+q(z_w)=-\left[\sqrt{\mathcal{V}_-(z_w)}-\pi_-q(z_w)\right],\qquad q(z)=z^3\sqrt{h(z)}. V + ( z w ) + π + q ( z w ) = − [ V − ( z w ) − π − q ( z w ) ] , q ( z ) = z 3 h ( z ) .
In the large-region limit π − → 0 \pi_- \to0 π − → 0 , it implies
π + q ( z w ) = − 1 2 ( z w z h ) 4 , V + ( z w ) = 1 2 ( z w z h ) 4 − 1. \pi_+q(z_w)=-\frac{1}{2}\left(\frac{z_w}{z_h}\right)^4,\qquad\sqrt{\mathcal{V}_+(z_w)}=\frac{1}{2}\left(\frac{z_w}{z_h}\right)^4-1. π + q ( z w ) = − 2 1 ( z h z w ) 4 , V + ( z w ) = 2 1 ( z h z w ) 4 − 1.
Consequently, the returning shell-crossing configuration requires z w > 2 1 / 4 z h z_w>2^{1/4}z_h z w > 2 1/4 z h . Following the near-critical analysis of [46 , 47 ] , we consider a critical configuration in which V + \mathcal{V}_+ V + develops a double zero at z = z m z=z_m z = z m ,
V + ( z m , π ⋆ ) = 0 , ∂ z V + ∣ z m , π ⋆ = 0. \mathcal{V}_+(z_m,\pi_\star)=0,\qquad\left.\partial_z\mathcal{V}_+\right|_{z_m,\pi_\star}=0. V + ( z m , π ⋆ ) = 0 , ∂ z V + ∣ z m , π ⋆ = 0.
where π ⋆ = π + ∣ z w = z w ⋆ \pi_\star=\left.\pi_+\right|_{z_w=z_w^\star} π ⋆ = π + ∣ z w = z w ⋆ . The returning solutions approach this configuration from π + > π ⋆ \pi_+>\pi_\star π + > π ⋆ , with z r → z m z_r\to z_m z r → z m . Expanding V + \mathcal{V}_+ V + around the critical point gives
V + ( z , π + ) ≃ c 1 ( z − z m ) 2 + c 2 ( π ⋆ − π + ) (116) \mathcal{V}_+(z,\pi_+)\simeq c_1(z-z_m)^2+c_2(\pi_\star-\pi_+) \tag*{(116)} V + ( z , π + ) ≃ c 1 ( z − z m ) 2 + c 2 ( π ⋆ − π + ) (116)
where
c 1 : = 1 2 ∂ z 2 V + ∣ z m , π ⋆ , c 2 : = − ∂ π + V + ∣ z m , π ⋆ = − 2 π ⋆ z m 6 h ( z m ) > 0 , c_1:=\left.\frac{1}{2}\partial_z^2\mathcal{V}_+\right|_{z_m,\pi_\star},\qquad c_2:=-\left.\partial_{\pi_+}\mathcal{V}_+\right|_{z_m,\pi_\star}=-2\pi_\star z_m^6h(z_m)>0, c 1 := 2 1 ∂ z 2 V + z m , π ⋆ , c 2 := − ∂ π + V + ∣ z m , π ⋆ = − 2 π ⋆ z m 6 h ( z m ) > 0 ,
and π ⋆ : = − 1 z m 3 − f ( z m ) h ( z m ) < 0 \pi_\star:=-\frac{1}{z_m^3}\sqrt{-\frac{f(z_m)}{h(z_m)}}<0 π ⋆ := − z m 3 1 − h ( z m ) f ( z m ) < 0 . The critical radius z m z_m z m coincides with that obtained from the HRT analysis in Eq. (60) . At this radius, the double zero of V + \mathcal{V}_+ V + is a nondegenerate minimum, so c 1 > 0 c_1>0 c 1 > 0 . Consequently, as π + → π ⋆ + \pi_+\to\pi_\star^+ π + → π ⋆ + , the returning turning point approaches z m z_m z m as
z r = z m − c 2 ( π + − π ⋆ ) c 1 + ⋯ . z_r=z_m-\sqrt{\frac{c_2(\pi_+-\pi_\star)}{c_1}}+\cdots. z r = z m − c 1 c 2 ( π + − π ⋆ ) + ⋯ .
Combining the shell-matching condition with Eq. (60) , the critical configuration is found to satisfy
2 1 / 4 z h < z w ⋆ < z m . 2^{1/4}z_h<z_w^\star<z_m. 2 1/4 z h < z w ⋆ < z m .
Thus, the single-crossing segment terminates before reaching the critical radius, whereas the returning segment extends toward the double zero, giving rise to the logarithmic divergence and the resulting linear growth of the EWCS. Using Eq. (116) , the leading contribution to the E W E_W E W is
E W ≃ 2 z m 3 h ( z m ) ∫ z w z r d z V + ( z ) + O ( 1 ) . (117) E_W\simeq\frac{2}{z_m^3\sqrt{h(z_m)}}\int_{z_w}^{z_r}\frac{\mathrm{d}z}{\sqrt{\mathcal{V}_+(z)}}+O(1). \tag*{(117)} E W ≃ z m 3 h ( z m ) 2 ∫ z w z r V + ( z ) d z + O ( 1 ) . (117)
The vacuum contribution and the non-returning black-brane segment remain finite in the critical limit, so the logarithmic divergence arises entirely from the returning segment. Similarly, the leading contribution to the boundary time follows from Eq. (115) ,
t ≃ 2 π ⋆ z m 3 h ( z m ) f ( z m ) ∫ z w z r d z V + ( z ) + O ( 1 ) . (118) t\simeq\frac{2\pi_\star z_m^3\sqrt{h(z_m)}}{f(z_m)}\int_{z_w}^{z_r}\frac{\mathrm{d}z}{\sqrt{\mathcal{V}_+(z)}}+O(1). \tag*{(118)} t ≃ f ( z m ) 2 π ⋆ z m 3 h ( z m ) ∫ z w z r V + ( z ) d z + O ( 1 ) . (118)
The EWCS and boundary time share the same leading logarithmic divergence. Taking the ratio of their coefficients gives the reduced growth rate − π ⋆ = v E N C / z h 3 -\pi_\star=v_E^{\mathrm{NC}}/z_h^3 − π ⋆ = v E NC / z h 3 and hence the intermediate scaling of the EWCS becomes
E W ≃ A Σ 2 s t h v E N C t + ⋯ , (119) E_W\simeq\frac{A_\Sigma}{2}s_{\mathrm{th}}v_E^{\mathrm{NC}}t+\cdots, \tag*{(119)} E W ≃ 2 A Σ s th v E NC t + ⋯ , (119)
where s t h s_{\rm th} s th is the thermal entropy density defined in Eq. (6) , A Σ = 2 L 2 A_{\Sigma}=2L^{2} A Σ = 2 L 2 , and
v E N C = z h 3 z m 3 − f ( z m ) h ( z m ) . (120) v_{E}^{\rm NC}=\frac{z_{h}^{3}}{z_{m}^{3}}\sqrt{-\frac{f(z_{m})}{h(z_{m})}}. \tag*{(120)} v E NC = z m 3 z h 3 − h ( z m ) f ( z m ) . (120)
The resulting EWCS growth velocity coincides with the entanglement velocity obtained from the HRT analysis in Eq. (71) . The linear growth is governed by the returning black-brane segment, which probes the critical region near z m z_{m} z m .
Saturation. At late times, provided the connected entanglement wedge remains dominant, the EWCS approaches its equilibrium value in the final black-brane geometry. As the shell-crossing position approaches its static limit, z w → z u z_{w}\to z_{u} z w → z u , the vacuum contribution vanishes, E W − → 0 E_{W}^{-}\to0 E W − → 0 . In this limit, the static configuration satisfies v ′ ( z ) = − 1 / f ( z ) v'(z)=-1/f(z) v ′ ( z ) = − 1/ f ( z ) , implying that the conserved momentum vanishes, π + = 0 \pi_{+}=0 π + = 0 , and hence V + ( z ) = f ( z ) V_{+}(z)=f(z) V + ( z ) = f ( z ) . Restoring the transverse and gravitational normalization, Eq. (104) yields
E W = A Σ 2 s t h z h 3 ∫ z b z u d z z 3 h ( z ) f ( z ) . (121) E_{W}=\frac{A_{\Sigma}}{2}s_{\rm th}z_{h}^{3}\int_{z_{b}}^{z_{u}}\frac{\mathrm{d}z}{z^{3}\sqrt{h(z)f(z)}}. \tag*{(121)} E W = 2 A Σ s th z h 3 ∫ z b z u z 3 h ( z ) f ( z ) d z . (121)
which coincides with the results first reported in [48 ] . In particular, in the low-temperature, small a a a limit, the EWCS becomes
E W ( ℓ , d ) ≃ A Σ β 2 4 s t h z h 3 [ 1 d 2 − 1 ( 2 ℓ + d ) 2 + a 4 β 4 6 ( 1 d 6 − 1 ( 2 ℓ + d ) 6 ) + ⋯ ] . (122) E_{W}(\ell,d)\simeq\frac{A_{\Sigma}\beta^{2}}{4}s_{\rm th}z_{h}^{3}\left[\frac{1}{d^{2}}-\frac{1}{(2\ell+d)^{2}}+\frac{a^{4}\beta^{4}}{6}\left(\frac{1}{d^{6}}-\frac{1}{(2\ell+d)^{6}}\right)+\cdots\right]. \tag*{(122)} E W ( ℓ , d ) ≃ 4 A Σ β 2 s th z h 3 [ d 2 1 − ( 2 ℓ + d ) 2 1 + 6 a 4 β 4 ( d 6 1 − ( 2 ℓ + d ) 6 1 ) + ⋯ ] . (122)
If the disconnected HRT configuration instead becomes dominant, the entanglement wedge disconnects and the physical EWCS vanishes. The transition to this configuration is generically discontinuous [47 ] .
Conclusions and discussions
We have studied the time evolution of holographic information measures following a homogeneous global quench in four-dimensional N = 4 \mathcal{N}=4 N = 4 noncommutative Yang–Mills theory. The quench is modeled by an infalling null shell that changes the energy density at fixed noncommutative scale. We consider strip regions whose finite direction lies along the noncommutative coordinate x 2 x_{2} x 2 , combining analytic results in the thin-shell and large-region limits with numerical calculations for a smooth shell. The corresponding noncommutative Vaidya geometry is supported by an external null source and satisfies the sourced type IIB equations, as demonstrated in Appendix § 8 .
Noncommutativity affects the evolution through the non-AdS near-boundary geometry and its associated UV/IR mixing. Already in the static problem, the U-shaped extremal surfaces exhibit a local minimum width ℓ 0 \ell_{0} ℓ 0 , separating two branches, as well as a cutoff-dependent scale ℓ c \ell_{c} ℓ c associated with UV/IR mixing [26 ] . For δ ≪ a \delta\ll a δ ≪ a , the latter behaves as ℓ c ≃ a 2 2 δ \ell_{c}\simeq\frac{a^{2}}{2\delta} ℓ c ≃ 2 δ a 2 . We therefore restrict our dynamical analysis to the resolvable regime above ℓ c \ell_{c} ℓ c . In the parameter range considered, the selected vacuum and equilibrium surfaces lie on the deeper U-shaped branch.
At early times, the leading near-boundary contribution to the entanglement entropy is logarithmic, with a coefficient proportional to the injected energy density and to a 2 a^{2} a 2 . At this order, the contribution is independent of the strip width. This result holds within the hierarchy specified in § 4.1 ; width-dependent terms enter beyond the leading near-boundary expansion. Moreover, the commutative limit of this expansion is nonuniform, and the commutative result must instead be recovered from the full expressions.
For sufficiently wide strips, the evolution is governed by the critical geometry behind the final horizon, as in the standard holographic quench scenario [14 ] . In particular, an intermediate linear-growth regime emerges from the double zero of the effective potential at z = z m z=z_m z = z m . At fixed z h z_h z h , both z m z_m z m and the critical shell-crossing radius z c ∗ z_c^* z c ∗ move toward the horizon as a a a increases over the range shown in Figure 4 , while the ordering z h < z c ∗ < z m z_h<z_c^*<z_m z h < z c ∗ < z m is maintained. The corresponding entanglement velocity v E N C v_{\mathrm{E}}^{\mathrm{NC}} v E NC approaches the A d S 5 \mathrm{AdS}_5 AdS 5 result in the commutative limit and increases with a a a over the range studied, in line with the previously observed effects of noncommutativity on entanglement propagation [29 ] .
Figure 4. Critical radii z m z_m z m and z c ∗ z_c^* z c ∗ as functions of the noncommutativity parameter a a a . The returning branch satisfies 2 1 / 4 z h < z c ∗ < z m 2^{1/4}z_h < z_c^* < z_m 2 1/4 z h < z c ∗ < z m and ends when the two radii coincide.
The late-time analysis reveals a distinction between the entanglement velocity and the butterfly velocity. As shown in Eq. (83) , the leading large-strip behavior of the continuous-completion time t s ( c o n t ) t_s^{(\mathrm{cont})} t s ( cont ) is controlled by v B , x 2 v_{\mathrm{B},x2} v B , x 2 , whereas the intermediate linear growth is governed by v E N C v_{\mathrm{E}}^{\mathrm{NC}} v E NC . The continuous-completion scale should not, in general, be identified with the physical saturation time. Different HRT branches may compete in the noncommutative theory, and the physical entropy is determined by the globally minimal branch.
The numerical results are consistent with this picture. At fixed cutoff, increasing the noncommutative scale raises the equilibrium value of the entropy difference and modestly increases the slope in the intermediate growth regime; the change in the equilibrium value is more pronounced than that in the growth rate. Increasing the strip width primarily extends the growth regime and delays saturation, while leaving the leading growth rate nearly width independent. For wider strips, the change in slope near the plateau is consistent with a change in the dominant HRT branch. The numerical profiles also indicate that, at fixed width, varying the noncommutative scale affects the turning point less strongly than varying the strip width.
To further characterize the time dependence, we consider the instantaneous entropy growth rate, following the holographic quench analysis of [15 ] . With the normalization of Eq. (19) , we define
R ( t ; ℓ ) = 1 s t h A Σ d S ( ℓ , t ) d t , (123) \mathcal{R}(t;\ell)=\frac{1}{s_{\mathrm{th}}A_{\Sigma}}\frac{dS(\ell,t)}{dt}, \tag*{(123)} R ( t ; ℓ ) = s th A Σ 1 d t d S ( ℓ , t ) , (123)
where in the intermediate linear-growth regime reduces to the entanglement velocity. Figure 16 shows the instantaneous rate for a strip of width ℓ = 1 \ell=1 ℓ = 1 . For all three values of the noncommutative scale, the rate rises to a single maximum and subsequently decreases toward zero as the entropy approaches equilibrium. Increasing a a a shifts the maximum to earlier times and modestly increases its height.
Figure 16. R ( t ) \mathcal{R}(t) R ( t ) for different values of the noncommutative scale.
We have also investigated the holographic mutual information between two disjoint strips. The leading width-independent logarithmic contribution at early times cancels in the connected combination. Consequently, determining the leading time-dependent correction requires the subleading, width-dependent terms. For sufficiently large regions whose linear-growth regimes overlap, the mutual-information growth rates are inherited from the physical single-strip entropies and are governed by the same entanglement velocity v E N C v_{\mathrm{E}}^{\mathrm{NC}} v E NC . This is consistent with previous holographic studies of mutual information following a quench [49 , 50 , 51 ] . The numerical HMI curves begin at their vacuum values, rise to a maximum, and then decrease toward equilibrium. Increasing the noncommutative scale raises the mutual information throughout the evolution which is consistent with the static
NCYM result of [26 ] .
We next considered the entanglement wedge cross section. Its vacuum and equilibrium expressions reproduce the known static NCYM results [48 ] . On the near-critical returning branch, the EWCS extremal candidate probes the same critical radius z m z_m z m as the HRT surface and exhibits linear growth governed by the same entanglement velocity v E N C v_{\mathrm{E}}^{\mathrm{NC}} v E NC . The numerical results exhibit two possible late-time behaviors. If the connected entanglement wedge becomes subdominant, the physical EWCS vanishes; increasing the separation then leads to earlier disentangling. If the connected phase remains dominant, the EWCS develops a transient maximum and subsequently approaches a finite equilibrium value. Again, we observed that increasing the noncommutative scale raises the vacuum, maximum, and equilibrium values.
Although our analysis has focused on the entanglement entropy, mutual information, and EWCS, the latter is conjectured to be dual to several information-theoretic measures of mixed-state correlations. It is therefore interesting to briefly examine what our results imply for these quantities. In particular, the time evolution of the EWCS, together with the entanglement entropies calculated above, allows us to infer the behavior of reflected entropy, the Markov gap, and odd entanglement entropy. These quantities provide complementary information-theoretic perspectives on the results obtained in this work.
The reflected entropy S R ( A : B ) S_R(A:B) S R ( A : B ) is defined through the canonical purification of the reduced density matrix ρ A B \rho_{AB} ρ A B [52 ] . At leading order in holography, it is related to the EWCS by
S R ( A : B ; t ) = 2 E W ( A : B ; t ) . (124) S_R(A:B;t)=2EW(A:B;t). \tag*{(124)} S R ( A : B ; t ) = 2 E W ( A : B ; t ) . (124)
Reflected entropy has also been studied in anisotropic holographic theories [53 ] , in covariant time-dependent settings [54 ] , and following global quenches in two-dimensional conformal field theories [55 ] .
The Markov gap M g ( A : B ; t ) \mathcal{M}_g(A:B;t) M g ( A : B ; t ) is defined as the difference between reflected entropy and mutual information. It is nonnegative and is associated with quantum Markov recovery [56 ] . Using
Eq. (124) , we obtain
M g ( A : B ; t ) ≡ S R ( A : B ; t ) − I ( A : B ; t ) = 2 E W ( A : B ; t ) − I ( A : B ; t ) ≥ 0. (125) \mathcal{M}_g(A:B;t)\equiv S_R(A:B;t)-I(A:B;t)=2E_W(A:B;t)-I(A:B;t)\geq0. \tag*{(125)} M g ( A : B ; t ) ≡ S R ( A : B ; t ) − I ( A : B ; t ) = 2 E W ( A : B ; t ) − I ( A : B ; t ) ≥ 0. (125)
Thus, the Markov gap provides a measure of the difference between the correlations captured by reflected entropy and those quantified by mutual information.
We also consider the odd entanglement entropy S O ( A : B ) S_O(A:B) S O ( A : B ) , defined through the odd-replica continuation of the partial transpose of ρ A B \rho_{AB} ρ A B [57 ] . Adopting the proposed leading-order holographic relation, we write
S O ( A : B ; t ) = S ( A ∪ B ; t ) + E W ( A : B ; t ) . (126) S_O(A:B;t)=S(A\cup B;t)+E_W(A:B;t). \tag*{(126)} S O ( A : B ; t ) = S ( A ∪ B ; t ) + E W ( A : B ; t ) . (126)
Here, S ( A ∪ B ; t ) S(A\cup B;t) S ( A ∪ B ; t ) is determined by the competing connected and disconnected HRT configurations discussed in Eq. (85) . In evaluating S O S_O S O , we subtract the UV-divergent contribution from S ( A ∪ B ; t ) S(A\cup B;t) S ( A ∪ B ; t ) , while leaving the EWCS contribution unchanged.
Figure 17 illustrates the time evolution of these three quantities for several parameters. As shown in the left panel the reflected entropy directly inherits the time dependence of the EWCS. The middle panel shows that the Markov gap develops a transient maximum. As the strip width increases, this maximum becomes larger and shifts to later times, reflecting the different time dependences of the EWCS and mutual information. The right panel shows that the odd entanglement entropy has a more involved evolution, since it receives contributions from both the EWCS and the entropy of the union A ∪ B A\cup B A ∪ B . An interesting feature emerges when the disconnected HRT configuration becomes dominant. In this regime, the entanglement wedge disconnects, and both E W E_W E W and I ( A : B ) I(A:B) I ( A : B ) vanish at leading order. Consequently, the reflected entropy and Markov gap also vanish, whereas the odd entanglement entropy reduces to S ( A ∪ B ; t ) S(A\cup B;t) S ( A ∪ B ; t ) . These observations illustrate that, although all three quantities are related to the EWCS, they capture different aspects of mixed-state correlations and provide complementary perspectives on the nonequilibrium dynamics.
Figure 17. Time evolution of three boundary-theory measures of mixed-state correlations for different strip widths at fixed d = 0.2 d=0.2 d = 0.2 and a = 0.1 a=0.1 a = 0.1 . Left: reflected entropy S R S_R S R . Middle: Markov gap M g \mathcal{M}_g M g . Right: UV-subtracted odd entanglement entropy S O S_O S O .
Several extensions of this work would be worthwhile. First, it would be interesting to investigate the dynamics of small entangling regions, along the lines of [58 ] , and to determine whether the corresponding behavior can be reproduced directly through field-theory calculations in noncommutative gauge theories. More generally, deriving the early-time logarithmic response and the entanglement velocity v E N C v_E^{\mathrm{NC}} v E NC from the boundary theory would provide an independent test of their holographic interpretation and clarify the role of noncommutativity. Another interesting direction is to investigate whether the large-region dynamics admits a membrane-like description adapted to NCYM theory, following the approach of [59 ] . Finally, studying the evolution of other boundary information measures, including those related to the EWCS, may help identify universal features of entanglement growth and the dynamics of mixed-state correlations in noncommutative field theories.
Acknowledgements
We are very grateful to Mohammad Hasan Vahidinia for careful reading of the manuscript and his valuable comments. The work of M. Reza Mohammadi Mozaffar is supported by the Iran National Science Foundation (INSF) under project No. 40403689. Portions of the text were revised with the assistance of ChatGPT (OpenAI) to improve English expression.
Consistency of the noncommutative Vaidya background
We verify the sourced type IIB supergravity equations for the noncommutative Vaidya metric (8) . The dilaton and form fields are kept at their static noncommutative D3-brane profiles, with the time dependence entering only through the blackening function in Eq. (9) . We use the string-frame action and equations of motion in Eqs. (1) and (2) . The signs of the modified Ramond–Ramond (RR) field strengths follow [30 ] , while the general type IIB framework and the imposition of five-form self-duality after variation follow Refs. [31 , 32 ] . Considering the action given by Eq. (1) , we have the following field strengths in terms of the p p p -form potentials
H 3 = d B 2 , F { 1 , 3 , 5 } = d C { 0 , 2 , 4 } , F ~ { 3 , 5 } = F { 3 , 5 } + C { 0 , 2 } ∧ H 3 . (127) H_3=dB_2,\qquad F_{\{1,3,5\}}=dC_{\{0,2,4\}},\qquad\widetilde{F}_{\{3,5\}}=F_{\{3,5\}}+C_{\{0,2\}}\wedge H_3. \tag*{(127)} H 3 = d B 2 , F { 1 , 3 , 5 } = d C { 0 , 2 , 4 } , F { 3 , 5 } = F { 3 , 5 } + C { 0 , 2 } ∧ H 3 . (127)
We work in the axion-free sector and impose self-duality only after varying the action.
C 0 = 0 , F 1 = 0 , F ~ 3 = F 3 , ⋆ F ~ 5 = F ~ 5 . (128) C_0=0,\qquad F_1=0,\qquad\widetilde{F}_3=F_3,\qquad\star\widetilde{F}_5=\widetilde{F}_5. \tag*{(128)} C 0 = 0 , F 1 = 0 , F 3 = F 3 , ⋆ F 5 = F 5 . (128)
The dilaton profile is the one given in § 2.1 . The remaining potentials of the static noncommutative D3-brane solution may be chosen as [23 , 24 ]
B 2 = R 2 a 2 z 4 + a 4 d x 2 ∧ d x 3 , C 2 = R 2 a 2 g s z 4 d v ∧ d x 1 . (129) B_2=\frac{R^2a^2}{z^4+a^4}\,dx_2\wedge dx_3,\qquad C_2=\frac{R^2a^2}{g_s z^4}\,dv\wedge dx_1. \tag*{(129)} B 2 = z 4 + a 4 R 2 a 2 d x 2 ∧ d x 3 , C 2 = g s z 4 R 2 a 2 d v ∧ d x 1 . (129)
In each static region, the second expression is gauge-equivalent to the standard RR two-form in static coordinates. We retain the same radial profile, with no explicit v v v -dependence, across the Vaidya geometry. The corresponding three-form field strengths are
H 3 = − 4 R 2 a 2 z 3 ( z 4 + a 4 ) 2 d z ∧ d x 2 ∧ d x 3 , F 3 = − 4 R 2 a 2 g s z 5 d z ∧ d v ∧ d x 1 . (130) H_3=-\frac{4R^2a^2z^3}{(z^4+a^4)^2}\,dz\wedge dx_2\wedge dx_3,\qquad F_3=-\frac{4R^2a^2}{g_s z^5}\,dz\wedge dv\wedge dx_1. \tag*{(130)} H 3 = − ( z 4 + a 4 ) 2 4 R 2 a 2 z 3 d z ∧ d x 2 ∧ d x 3 , F 3 = − g s z 5 4 R 2 a 2 d z ∧ d v ∧ d x 1 . (130)
Let ρ 5 \rho_5 ρ 5 denote the volume form of the unit S 5 S^5 S 5 and choose the orientation d v ∧ d z ∧ d x 1 ∧ d x 2 ∧ d x 3 ∧ ρ 5 > 0 dv\wedge dz\wedge dx_1\wedge dx_2\wedge dx_3\wedge\rho_5>0 d v ∧ d z ∧ d x 1 ∧ d x 2 ∧ d x 3 ∧ ρ 5 > 0 . The self-dual five-form is
F ~ 5 = 4 R 4 g s [ h ( z ) z 5 d v ∧ d z ∧ d x 1 ∧ d x 2 ∧ d x 3 + ρ 5 ] . (131) \widetilde{F}_5=\frac{4R^4}{g_s}\left[\frac{h(z)}{z^5}\,dv\wedge dz\wedge dx_1\wedge dx_2\wedge dx_3+\rho_5\right]. \tag*{(131)} F 5 = g s 4 R 4 [ z 5 h ( z ) d v ∧ d z ∧ d x 1 ∧ d x 2 ∧ d x 3 + ρ 5 ] . (131)
The Bianchi identities are satisfied directly
d H 3 = 0 , d F 3 = 0 , F 3 ∧ H 3 = 0 , d F ~ 5 = 0. (132) dH_3 = 0,\qquad dF_3 = 0,\qquad F_3 \wedge H_3 = 0,\qquad d\widetilde{F}_5 = 0. \tag*{(132)} d H 3 = 0 , d F 3 = 0 , F 3 ∧ H 3 = 0 , d F 5 = 0. (132)
With the conventions of Eq. (A.1), the last identity and five-form self-duality also imply the five-form equation of motion. Moreover, the axion equation reduces to
H M N P F 3 M N P = 0. (133) H_{MNP}F_3^{MNP} = 0. \tag*{(133)} H M N P F 3 M N P = 0. (133)
Every nonzero component of H 3 H_3 H 3 contains both x 2 x_2 x 2 and x 3 x_3 x 3 , whereas every nonzero component of F 3 F_3 F 3 contains x 1 x_1 x 1 . Raising the v v v and z z z indices with the ingoing metric does not change this index mismatch. Therefore the contraction in Eq. (A.7) vanishes.
The source in Eq. (14) has vanishing trace, since it is proportional to k M k N k_Mk_N k M k N with k M k M = 0 k^Mk_M = 0 k M k M = 0 . The trace of the Einstein equation may therefore be combined with the dilaton equation obtained from Eq. (1) , giving
∇ 2 Φ − 2 ( ∇ Φ ) 2 + 1 12 ( H 3 2 − e 2 Φ F 3 2 ) = 0. (134) \nabla^2\Phi- 2(\nabla\Phi)^2 + \frac{1}{12}\left(H_3^2 - e^{2\Phi}F_3^2\right) = 0. \tag*{(134)} ∇ 2 Φ − 2 ( ∇Φ ) 2 + 12 1 ( H 3 2 − e 2Φ F 3 2 ) = 0. (134)
The left-hand side vanishes identically using Eq. (A.4) and the dilaton profile. In particular, no term proportional to ∂ v m ( v ) \partial_v m(v) ∂ v m ( v ) is generated, so the Vaidya deformation requires no external dilaton source. For the axion-free field content in Eq. (A.2), the RR and NS–NS three-form equations are
∇ L F 3 L M N − 1 3 ! H A B C F ~ 5 M N A B C = 0 , \nabla_L F_3^{LMN} - \frac{1}{3!}H_{ABC}\widetilde{F}_5^{MNABC} = 0, ∇ L F 3 L M N − 3 ! 1 H A B C F 5 M N A B C = 0 ,
∇ L ( e − 2 Φ H 3 L M N ) + 1 3 ! F A B C F ~ 5 M N A B C = 0. (135) \nabla_L\left(e^{-2\Phi}H_3^{LMN}\right) + \frac{1}{3!}F_{ABC}\widetilde{F}_5^{MNABC} = 0. \tag*{(135)} ∇ L ( e − 2Φ H 3 L M N ) + 3 ! 1 F A B C F 5 M N A B C = 0. (135)
Up to antisymmetry, the only RR equation with nonzero terms is ( M , N ) = ( v , x 1 ) (M,N)=(v,x_1) ( M , N ) = ( v , x 1 ) . The two terms satisfy
∇ L F 3 L v x 1 = 1 3 ! H A B C F ~ 5 v x 1 A B C = − 16 a 2 z 4 g s R 4 ( z 4 + a 4 ) . (136) \nabla_L F_3^{Lvx_1} = \frac{1}{3!}H_{ABC}\widetilde{F}_5^{vx_1ABC} = -\frac{16a^2z^4}{g_sR^4(z^4+a^4)}. \tag*{(136)} ∇ L F 3 Lv x 1 = 3 ! 1 H A B C F 5 v x 1 A B C = − g s R 4 ( z 4 + a 4 ) 16 a 2 z 4 . (136)
For the NS–NS equation, the only equation with nonzero terms is ( M , N ) = ( x 2 , x 3 ) (M,N)=(x_2,x_3) ( M , N ) = ( x 2 , x 3 ) , for which
∇ L ( e − 2 Φ H 3 L x 2 x 3 ) = − 1 3 ! F A B C F ~ 5 x 2 x 3 A B C = 16 a 2 ( z 4 + a 4 ) g s 2 R 4 z 4 . (137) \nabla_L\left(e^{-2\Phi}H_3^{Lx_2x_3}\right) = -\frac{1}{3!}F_{ABC}\widetilde{F}_5^{x_2x_3ABC} = \frac{16a^2(z^4+a^4)}{g_s^2R^4z^4}. \tag*{(137)} ∇ L ( e − 2Φ H 3 L x 2 x 3 ) = − 3 ! 1 F A B C F 5 x 2 x 3 A B C = g s 2 R 4 z 4 16 a 2 ( z 4 + a 4 ) . (137)
All remaining components vanish due to antisymmetry, the index structure of the three-forms, and the coordinate dependence of the background fields. Equations (A.10) and (A.11) verify the two equations in Eq. (A.9); no external NS–NS or RR two-form current is required.
For constant m m m , the fields introduced above reproduce the static noncommutative D3-brane solution [23 , 24 ] , and the source-free Einstein equation is satisfied componentwise. For a general mass profile m ( v ) m(v) m ( v ) , all terms that do not contain ∂ v m ( v ) \partial_v m(v) ∂ v m ( v ) cancel as in the static solution. The only new derivative-dependent contribution appears in the ( v , v ) (v,v) ( v , v ) component.
The contributions proportional to ∂ v m ( v ) \partial_v m(v) ∂ v m ( v ) in the Ricci tensor and dilaton Hessian are
R v v ∣ ∂ v m = z 3 ( 3 z 4 − a 4 ) 2 ( z 4 + a 4 ) ∂ v m ( v ) , 2 ∇ v ∇ v Φ ∣ ∂ v m = 2 a 4 z 3 z 4 + a 4 ∂ v m ( v ) . (138) \left.R_{vv}\right|_{\partial_v m} = \frac{z^3(3z^4-a^4)}{2(z^4+a^4)}\partial_v m(v),\qquad
\left.2\nabla_v\nabla_v\Phi\right|_{\partial_v m} = \frac{2a^4z^3}{z^4+a^4}\partial_v m(v). \tag*{(138)} R v v ∣ ∂ v m = 2 ( z 4 + a 4 ) z 3 ( 3 z 4 − a 4 ) ∂ v m ( v ) , 2 ∇ v ∇ v Φ ∣ ∂ v m = z 4 + a 4 2 a 4 z 3 ∂ v m ( v ) . (138)
The form-field contributions are algebraic in f ( v , z ) f(v,z) f ( v , z ) and contain no derivatives of the mass profile. Adding the two terms in Eq. (A.12), all explicit dependence on the noncommutative scale cancels, and one obtains 3 2 z 3 ∂ v m ( v ) \frac{3}{2}z^{3}\partial_{v}m(v) 2 3 z 3 ∂ v m ( v ) . This is precisely the ( v , v ) (v,v) ( v , v ) component of the external source defined in Eq. (14) . All other ∂ v m ( v ) \partial_{v}m(v) ∂ v m ( v ) -dependent components vanish. The null energy condition is equivalent to ∂ v m ( v ) ≥ 0 \partial_{v}m(v) \ge0 ∂ v m ( v ) ≥ 0 , which is satisfied by the smooth mass profile in Eq. (16) . With the normalization adopted in Eq. (2) , the source in the trace-reversed string-frame equation obeys the dilaton-weighted identity
∇ M ( e − 2 Φ T M N e x t ) = 0. (139) \nabla^{M}\left(e^{-2\Phi}T^{\mathrm{ext}}_{MN}\right)=0. \tag*{(139)} ∇ M ( e − 2Φ T M N ext ) = 0. (139)
This is the conservation law compatible with the sourced string-frame field equations. Thus, the metric (8) with the static dilaton and form-field profiles satisfies the sourced type IIB equations for arbitrary smooth m ( v ) m(v) m ( v ) , with the null source in Eq. (14) . No external scalar or form-field current is generated, and the null energy condition requires a nondecreasing mass profile.
Details of the saturation regime
In this appendix, we give the details of the saturation analysis in § 4.3 . We work in the thin-shell limit and consider the monotonic shell-crossing branch described by Eqs. (33) –(35) . In the parameter range considered in § 3.2 , the equilibrium surface is the deeper U 1 U_{1} U 1 branch selected by the minimum-area condition. We consider two limiting regimes: the near-horizon large-width expansion leading to Eq. (75) , and the fixed-z b z_b z b near-saturation expansion used in Eqs. (78) and (81) .
We first consider the near-horizon large-width limit of § 4.3 . The final blackening function has a simple zero, f ( z h ) = 0 f(z_h)=0 f ( z h ) = 0 with f ′ ( z h ) ≠ 0 f'(z_h)\ne0 f ′ ( z h ) = 0 , while the warp factor is finite and positive, 0 < h ( z h ) < ∞ 0<h(z_h)<\infty 0 < h ( z h ) < ∞ . The endpoint of the time integral in Eq. (73) therefore gives the second logarithm in Eq. (74) . For the width, write y = z / z b y=z/z_b y = z / z b in the static integral ℓ / 2 = F ( z b ) \ell/2=\mathcal{F}(z_b) ℓ /2 = F ( z b ) and set y = 1 − s y=1-s y = 1 − s near its upper endpoint. For s , η ≪ 1 s,\eta\ll1 s , η ≪ 1 ,
f ( z b y ) = − z h f ′ ( z h ) ( s + η ) + O ( s 2 , s η , η 2 ) , h ( z b y ) = h ( z h ) + O ( s , η ) . (140) f(z_b y)=-z_h f'(z_h)(s+\eta)+O(s^2,s\eta,\eta^2),\qquad h(z_b y)=h(z_h)+O(s,\eta). \tag*{(140)} f ( z b y ) = − z h f ′ ( z h ) ( s + η ) + O ( s 2 , sη , η 2 ) , h ( z b y ) = h ( z h ) + O ( s , η ) . (140)
The singular part of the width integral is
ℓ 2 ∣ s i n g = z h 6 ∣ f ′ ( z h ) ∣ h ( z h ) ∫ 0 Λ d s s ( s + η ) = z h 6 ∣ f ′ ( z h ) ∣ h ( z h ) log ( 1 η ) + O ( 1 ) , (141) \left.\frac{\ell}{2}\right|_{\mathrm{sing}}=\sqrt{\frac{z_h}{6\lvert f'(z_h)\rvert h(z_h)}}\int_{0}^{\Lambda}\frac{ds}{\sqrt{s(s+\eta)}}=\sqrt{\frac{z_h}{6\lvert f'(z_h)\rvert h(z_h)}}\log\left(\frac{1}{\eta}\right)+O(1), \tag*{(141)} 2 ℓ sing = 6 ∣ f ′ ( z h )∣ h ( z h ) z h ∫ 0 Λ s ( s + η ) d s = 6 ∣ f ′ ( z h )∣ h ( z h ) z h log ( η 1 ) + O ( 1 ) , (141)
where Λ \Lambda Λ is fixed and small. Here O ( 1 ) O(1) O ( 1 ) denotes terms finite as η → 0 \eta\to0 η → 0 ; such terms may depend on a a a , z h z_h z h , δ \delta δ , and the fixed upper limit Λ \Lambda Λ . The above equation reproduces the first logarithm in Eq. (74) . Eliminating η \eta η between the two logarithms gives Eq. (75) .
We next use the parametrization introduced in Eq. (77) . The limit ϵ → 0 \epsilon\to0 ϵ → 0 is taken at fixed z b < z h z_b<z_h z b < z h on a regular monotonic branch, and we assume F ′ ( z b ) ≠ 0 \mathcal{F}'(z_b)\ne0 F ′ ( z b ) = 0 . This condition is required for the fixed-width implicit-function expansion and fails at a turning point or cusp of the width–turning-point curve. The functions F \mathcal{F} F and J \mathcal{J} J are defined in Eq. (79) , while H ( z ) H(z) H ( z ) is defined in Eq. (31) . Using Eqs. (26) and (30) , the black-brane integration constant and its expansion near ϵ = 0 \epsilon=0 ϵ = 0 are
E = − g ( z c ) 2 h ( z c ) ( z t z c ) 3 1 − ( z c z t ) 6 , E = − g ( z t ) 2 h ( z t ) ϵ + O ( ϵ 3 ) . (142) E=-\frac{g(z_c)}{2\sqrt{h(z_c)}}\left(\frac{z_t}{z_c}\right)^3\sqrt{1-\left(\frac{z_c}{z_t}\right)^6},\qquad E=-\frac{g(z_t)}{2\sqrt{h(z_t)}}\epsilon+O(\epsilon^3). \tag*{(142)} E = − 2 h ( z c ) g ( z c ) ( z c z t ) 3 1 − ( z t z c ) 6 , E = − 2 h ( z t ) g ( z t ) ϵ + O ( ϵ 3 ) . (142)
Further, adding and subtracting the static black-brane integrand in Eq. (33) , we write ℓ / 2 = A 1 − A 2 + A 3 + F ( z t ) \ell/2 = A_1 - A_2 + A_3 + \mathcal{F}(z_t) ℓ /2 = A 1 − A 2 + A 3 + F ( z t ) where
A 1 = ∫ z c z t d z h ( z ) z t 6 / z 6 − 1 , A 2 = ∫ z c z t d z H ( z ) . (143) A_1 = \int_{z_c}^{z_t} \frac{dz}{\sqrt{h(z)}\sqrt{z_t^6/z^6-1}}, \qquad A_2 = \int_{z_c}^{z_t} \frac{dz}{\sqrt{H(z)}}. \tag*{(143)} A 1 = ∫ z c z t h ( z ) z t 6 / z 6 − 1 d z , A 2 = ∫ z c z t H ( z ) d z . (143)
A 3 = ∫ δ z t d z [ 1 H ( z ) − 1 f ( z ) h ( z ) z t 6 / z 6 − 1 ] . (144) A_3 = \int_{\delta}^{z_t} dz \left[\frac{1}{\sqrt{H(z)}}-\frac{1}{\sqrt{f(z)h(z)}\sqrt{z_t^6/z^6-1}}\right]. \tag*{(144)} A 3 = ∫ δ z t d z [ H ( z ) 1 − f ( z ) h ( z ) z t 6 / z 6 − 1 1 ] . (144)
Expanding near z = z t z=z_t z = z t gives
A 1 = A 2 = z t 3 h ( z t ) ϵ + O ( ϵ 3 ) , A 3 = − z t g ( z t ) 6 f ( z t ) h ( z t ) ϵ + O ( ϵ 2 ) . (145) A_1=A_2=\frac{z_t}{3\sqrt{h(z_t)}}\epsilon+\mathcal{O}(\epsilon^3), \qquad A_3=-\frac{z_tg(z_t)}{6f(z_t)\sqrt{h(z_t)}}\epsilon+\mathcal{O}(\epsilon^2). \tag*{(145)} A 1 = A 2 = 3 h ( z t ) z t ϵ + O ( ϵ 3 ) , A 3 = − 6 f ( z t ) h ( z t ) z t g ( z t ) ϵ + O ( ϵ 2 ) . (145)
The linear terms in A 1 A_1 A 1 and A 2 A_2 A 2 cancel. At fixed width, ℓ / 2 = F ( z b ) \ell/2=\mathcal{F}(z_b) ℓ /2 = F ( z b ) , while
F ( z t ) = F ( z b ) + z b σ 6 F ′ ( z b ) + O ( ϵ 2 ) . (146) \mathcal{F}(z_t)=\mathcal{F}(z_b)+\frac{z_b\sigma}{6}\mathcal{F}'(z_b)+\mathcal{O}(\epsilon^2). \tag*{(146)} F ( z t ) = F ( z b ) + 6 z b σ F ′ ( z b ) + O ( ϵ 2 ) . (146)
The fixed-width constraint then yields
σ = g ( z b ) f ( z b ) h ( z b ) F ′ ( z b ) ϵ + O ( ϵ 2 ) , (147) \sigma=\frac{g(z_b)}{f(z_b)\sqrt{h(z_b)}\mathcal{F}'(z_b)}\epsilon+\mathcal{O}(\epsilon^2), \tag*{(147)} σ = f ( z b ) h ( z b ) F ′ ( z b ) g ( z b ) ϵ + O ( ϵ 2 ) , (147)
where the derivative is evaluated at fixed cutoff δ \delta δ . Subtracting Eq. (73) from Eq. (34) , we obtain t − t s ( c o n t ) = B 1 + B 2 − B 3 t-t_s^{(\mathrm{cont})}=B_1+B_2-B_3 t − t s ( cont ) = B 1 + B 2 − B 3 where
B 1 = ∫ z b z c d z f ( z ) , B 2 = E ∫ δ z t h ( z ) d z f ( z ) H ( z ) , B 3 = E ∫ z c z t h ( z ) d z f ( z ) H ( z ) . (148) B_1=\int_{z_b}^{z_c}\frac{dz}{f(z)}, \qquad B_2=E\int_{\delta}^{z_t}\frac{h(z)\,dz}{f(z)\sqrt{H(z)}}, \qquad B_3=E\int_{z_c}^{z_t}\frac{h(z)\,dz}{f(z)\sqrt{H(z)}}. \tag*{(148)} B 1 = ∫ z b z c f ( z ) d z , B 2 = E ∫ δ z t f ( z ) H ( z ) h ( z ) d z , B 3 = E ∫ z c z t f ( z ) H ( z ) h ( z ) d z . (148)
To first order in ϵ \epsilon ϵ we have
B 1 = z b σ 6 f ( z b ) + O ( ϵ 2 ) , B 2 = − g ( z b ) 2 h ( z b ) J ( z b ) ϵ + O ( ϵ 2 ) , B 3 = O ( ϵ 2 ) . (149) B_1=\frac{z_b\sigma}{6f(z_b)}+\mathcal{O}(\epsilon^2), \qquad B_2=-\frac{g(z_b)}{2\sqrt{h(z_b)}}\mathcal{J}(z_b)\epsilon+\mathcal{O}(\epsilon^2), \qquad B_3=\mathcal{O}(\epsilon^2). \tag*{(149)} B 1 = 6 f ( z b ) z b σ + O ( ϵ 2 ) , B 2 = − 2 h ( z b ) g ( z b ) J ( z b ) ϵ + O ( ϵ 2 ) , B 3 = O ( ϵ 2 ) . (149)
Using Eq. (B.8) in these expressions reproduces Eq. (78) , including the coefficient ω \omega ω . Further, considering S e q S_{\mathrm{eq}} S eq in Eq. (41) as a function of its turning point, we write the nearby dynamical entropy as S ( ℓ , t ) = C 1 − C 2 + C 3 + S e q ( z t ) S(\ell,t)=C_1-C_2+C_3+S_{\mathrm{eq}}(z_t) S ( ℓ , t ) = C 1 − C 2 + C 3 + S eq ( z t ) with
C 1 = 2 z t 3 ∫ z c z t d z z 6 h ( z ) z t 6 / z 6 − 1 , C 2 = 2 z t 3 ∫ z c z t d z z 6 H ( z ) , C_1=2z_t^3\int_{z_c}^{z_t}\frac{dz}{z^6\sqrt{h(z)}\sqrt{z_t^6/z^6-1}}, \qquad C_2=2z_t^3\int_{z_c}^{z_t}\frac{dz}{z^6\sqrt{H(z)}}, C 1 = 2 z t 3 ∫ z c z t z 6 h ( z ) z t 6 / z 6 − 1 d z , C 2 = 2 z t 3 ∫ z c z t z 6 H ( z ) d z ,
C 3 = 2 z t 3 ∫ δ z t d z z 6 [ 1 H ( z ) − 1 f ( z ) h ( z ) z t 6 / z 6 − 1 ] . (150) C_3=2z_t^3\int_{\delta}^{z_t}\frac{dz}{z^6}\left[\frac{1}{\sqrt{H(z)}}-\frac{1}{\sqrt{f(z)h(z)}\sqrt{z_t^6/z^6-1}}\right]. \tag*{(150)} C 3 = 2 z t 3 ∫ δ z t z 6 d z [ H ( z ) 1 − f ( z ) h ( z ) z t 6 / z 6 − 1 1 ] . (150)
The corresponding leading terms are
C 1 = C 2 = 2 ϵ 3 z t 2 h ( z t ) + O ( ϵ 3 ) , C 3 = − g ( z t ) 3 z t 2 f ( z t ) h ( z t ) ϵ + O ( ϵ 2 ) . (151) C_1=C_2=\frac{2\epsilon}{3z_t^2\sqrt{h(z_t)}}+\mathcal{O}(\epsilon^3), \qquad C_3=-\frac{g(z_t)}{3z_t^2f(z_t)\sqrt{h(z_t)}}\epsilon+\mathcal{O}(\epsilon^2). \tag*{(151)} C 1 = C 2 = 3 z t 2 h ( z t ) 2 ϵ + O ( ϵ 3 ) , C 3 = − 3 z t 2 f ( z t ) h ( z t ) g ( z t ) ϵ + O ( ϵ 2 ) . (151)
The linear terms in C 1 C_1 C 1 and C 2 C_2 C 2 cancel. The remaining linear contribution cancels against the expansion of S e q ( z t ) S_{\mathrm{eq}}(z_t) S eq ( z t ) , as follows from the relation between the static entropy and the strip width. The first integral (24) implies d S e q / d ℓ = 1 / q 3 dS_{\mathrm{eq}}/d\ell= 1/q^3 d S eq / d ℓ = 1/ q 3 along the static family, while ℓ = 2 F ( q ) \ell= 2F(q) ℓ = 2 F ( q ) . Hence,
q 3 S e q ′ ( q ) = 2 F ′ ( q ) , (152) q^3 S'_{\mathrm{eq}}(q) = 2F'(q), \tag*{(152)} q 3 S eq ′ ( q ) = 2 F ′ ( q ) , (152)
where both derivatives are evaluated at fixed δ \delta δ . Expanding S e q ( z t ) S_{\mathrm{eq}}(z_t) S eq ( z t ) about z b z_b z b and using Eq. (B.8), the coefficient linear in ϵ \epsilon ϵ becomes
[ S ( ℓ , t ) − S e q ( ℓ ) ] O ( ϵ ) = g ( z b ) 3 z b 2 f ( z b ) h ( z b ) [ z b 3 S e q ′ ( z b ) 2 F ′ ( z b ) − 1 ] ϵ = 0. (153) \left[S(\ell,t)-S_{\mathrm{eq}}(\ell)\right]_{O(\epsilon)} = \frac{g(z_b)}{3z_b^2 f(z_b)\sqrt{h(z_b)}}\left[\frac{z_b^3 S'_{\mathrm{eq}}(z_b)}{2F'(z_b)}-1\right]\epsilon= 0. \tag*{(153)} [ S ( ℓ , t ) − S eq ( ℓ ) ] O ( ϵ ) = 3 z b 2 f ( z b ) h ( z b ) g ( z b ) [ 2 F ′ ( z b ) z b 3 S eq ′ ( z b ) − 1 ] ϵ = 0. (153)
Thus, the entropy difference starts at O ( ϵ 2 ) O(\epsilon^2) O ( ϵ 2 ) . If the quadratic coefficient is nonzero and the physical continuous branch approaches equilibrium from below, the coefficient of the leading ϵ 2 \epsilon^2 ϵ 2 term in S ( ℓ , t ) − S e q ( ℓ ) S(\ell,t)-S_{\mathrm{eq}}(\ell) S ( ℓ , t ) − S eq ( ℓ ) is negative. For ω ≠ 0 \omega\ne0 ω = 0 , this gives the quadratic time dependence in Eq. (81) . This completes the fixed-cutoff derivation of the near-saturation time expansion and the cancellation of the linear entropy correction used in § 4.3 .