This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

aainstitutetext: Graduate School of China Academy of Engineering Physics , P. R. China

Entropy Linear Response Theory with Non-Markovian Bath

Yu Chen a [email protected]
Abstract

We developed a perturbative calculation for entropy dynamics considering a sudden coupling between a system and a bath. The theory we developed can work in general environment without Markovian approximation. A perturbative formula is given for bosonic environment and fermionic environment, respectively. We find the Rényi entropy response is only related to the spectral functions of the system and the environment, together with a specific statistical kernel distribution function. We find a t2t^{2} growth/decay in the short time limit and a tt linear growth/decay in longer time scale for second Rényi entropy. A non-monotonic behavior of Rényi entropy for fermionic systems is found to be quite general when the environment’s temperature is lower. A Fourier’s law in heat transport is obtained when two systems’ temperature are close to each other. A consistency check is made for Sachdev-Ye-Kitaev model coupling to free fermions, a Page curve alike dynamics is found in a process dual to black hole evaporation. An oscillation of entanglement entropy is found for a gapped spectrum of environment.

Keywords:
Rényi entropy, non-Markovian, SYK model

1 Introduction

Linear response theory (LRT) is essencial for studying quantum matters, it lays the foundation for physical interpretations of the dynamical response after a perturbative probe to the systemMahan . LRT is widely used in all kinds of condensed matter experiments, cold atoms experiments, such as APRES, conductivity, neutron scattering, etc. In original linear response theory, the dynamics of physical observables are considered in response to external sources. These perturbations are hermitian couplings to the systems. However, when we consider a more general case for dissipative couplings, a new response theory is needed and established as a Non-hermitian linear response theory (NHLRT) Hui20 . A relation between dissipative dynamics DissBH and the correlation function in initial equilibrium state can be then established with the help of NHLRT.

However, previous response theories are mostly focused on physical observables, or for dissipative couplings with Markovian approximations, entropies response theories for a general environment is still not fully established. In this paper, we are going to establish a general perturbative response theory for entropy with a generic bath. Nowadays, thermal entropies as well as entanglement entropies become increasingly important in studying eigenstate thermalizationDeutsch91 ; Srednicki94 ; Cardy04 ; Grover18 , information scramblingQi16 ; Ruihua17 and quantum gravityTBook . It is in the core of holographic dualityMaldacena98 which bridges the conformal field theory and the gravity through Ryu-Takayanagi formulaRT06 ; T07 ; Wall15 . In addition to these theoretical interests, new technical advances are also achieved recently in experimental measurement of Rényi entropy (RE) Greiner15 ; Greiner16 ; Xinhua17 . All these advances in both the theoretical and experimental studies make a general response theory for entropy a necessity.

Following the spirit of taking the coupling between the environment and the system as a perturbation, we can establish a linear response theory of Rényi entropy for a general system with a sudden coupling to a general environment. Here we assume a general non-Markovian bath. The term non-Markovian means the temperature and spectrum of the environment can be arbitrary. Now we summarize our results as follows. We find that to the lowest order of coupling strength, the Rényi entropy response is only related to the spectral functions of the system and the environment, as well as a specific statistical kernel function. General properties of bosonic systems and fermionic systems are discussed and we predict a short time quadratic growth (decay) and long time linear growth (decay).

Further, we apply the linear response theory to Sachdev-Ye-Kitaev (SYK) modelKitaev15 ; SY ; Maldacena16 ; Kitaev17 with a free fermion bath with arbitrary different temperatures. SYK model is recently discovered as an example of holographic dualityKitaev17 ; Maldacena16a ; Gross17 . It is dual to a Jackiw-Teitelboim (JT) gravityJackiw ; Teitelboim in AdS2 spacetimeKitaev17 ; Maldacena16 . Since SYK model is a solvable model in large N limit, quite a few results for SYK model’s entropy dynamics are obtained for various SYK-like modelsGu17a ; Gu17b ; Balents17 ; Pengfei19 ; Yiming20 ; Pengfei20a ; Pengfei20b ; Banerjee20 ; Pengfei20c and for Maldacena-Qi modelMQ which is dual to a traversable wormhole. These entropy dynamics are also calculated in gravity sidePenington19 ; Almheiri19 ; Maldacena20 ; Pedraza17 ; Yiming19 . Here we try to ask, to what kind of extent could our perturbative theory apply to a general case. Can we recover the Page curve in “ black hole ” evaporation casePage01 ; Page02 where SYK model’s initial temperature is higher than the environment’s temperature. A Page curve is an entanglement entropy curve of the environment which follows a linear decay after an initial linear growth. Even through we are calculating black hole’s entropy, still, we wonder if we can see signals for Page curve. Eventually, we observed this Page curve like behavior in long time dynamics. Further we can change a gapless fermion to a gapped fermion to change the environment’s spectrum dramatically. an interesting phenomenon is observed for RE oscillation at some special temperature window. It seems that the entangled information can travel back and forth between the system and the bath in this situation, which is quite curious.

Our paper is orgarnized as follows. In the next section, we first establish the general linear response theory for RE of the subsystem. Then we discuss general properties of RE dynamics for bosonic systems and fermionic systems. Then in the third section, we apply our theory to SYK model with a non-Markovian bath and we changed the temperature and the spectrum of environment to calculate the RE dynamics. Finally, we summarize our results in the conclusion section.

2 General Entropy Response Theory

Here we consider a general situation for a sudden coupling between a system and a bath. Before the coupling, the hamiltonian is H^0=H^S+H^E\hat{H}_{0}=\hat{H}_{\rm S}+\hat{H}_{\rm E}, and after the coupling the hamiltonian becomes

H^=H^S+H^E+V^.\displaystyle\hat{H}=\hat{H}_{\rm S}+\hat{H}_{\rm E}+\hat{V}. (1)

Here V^=gj(O^jξ^j+ξ^jO^j)\hat{V}=g\sum_{j}(\hat{O}_{j}^{\dagger}\hat{\xi}_{j}+\hat{\xi}_{j}^{\dagger}\hat{O}_{j}), where O^j\hat{O}_{j} is an operator acting on the system and ξ^j\hat{\xi}_{j} is an operator in the environment. jj is the index of the operator. We can find out that the system’s density operator evolutes like follows

ρ^S(t)=TrE(eiH^tρ^0toteiH^t)/Tr(ρ^E(t)),\displaystyle\hat{\rho}_{\rm S}(t)={\rm Tr}_{E}\left(e^{-i\hat{H}t}\hat{\rho}_{0}^{\rm tot}e^{i\hat{H}t}\right)/{\rm Tr}(\hat{\rho}_{\rm E}(t)), (2)
ρ^E(t)=TrS(eiH^tρ^0toteiH^t)/Tr(ρ^S(t)),\displaystyle\hat{\rho}_{\rm E}(t)={\rm Tr}_{S}\left(e^{-i\hat{H}t}\hat{\rho}_{0}^{\rm tot}e^{i\hat{H}t}\right)/{\rm Tr}(\hat{\rho}_{\rm S}(t)), (3)

where ρ^0tot\hat{\rho}_{0}^{\rm tot} is the initial density operator, which we assume as a product state of thermal states in the system and in the bath.

ρ^0tot=ρ^0ρ^E.\displaystyle\hat{\rho}_{0}^{\rm tot}=\hat{\rho}_{0}\otimes\hat{\rho}_{E}. (4)

Because after a unitary evolution, the trace of ρ^S(t)\hat{\rho}_{\rm S}(t) and ρ^E(t)\hat{\rho}_{\rm E}(t) is invariant. Therefore we can simply replace TrE(ρ^E(t)){\rm Tr}_{E}(\hat{\rho}_{\rm E}(t)) to ZE=TrE(ρ^E(0))Z_{\rm E}={\rm Tr}_{E}(\hat{\rho}_{\rm E}(0)). Here we employ TrE{\rm Tr}_{E} as a trace over the environment’s Hilbert space. Tr{\rm Tr} is short for TrS{\rm Tr}_{S} as trace over system’s Hilbert space.

The reduced density operator of the system is evoluting in an interacting picture as

ρ^S(t)\displaystyle\hat{\rho}_{\rm S}(t) =\displaystyle= TrE(eiH^teiH^0teiH0^tρ^0ρ^EeiH^0teiH^0teiH^t)/ZE\displaystyle{\rm Tr}_{E}\left(e^{i\hat{H}t}e^{-i\hat{H}_{0}t}e^{i\hat{H_{0}}t}\hat{\rho}_{0}\otimes\hat{\rho}_{\rm E}e^{-i\hat{H}_{0}t}e^{i\hat{H}_{0}t}e^{-i\hat{H}t}\right)/Z_{\rm E} (5)
\displaystyle\equiv TrE(𝒰(0,t)ρ^0I(t)ρ^EI(t)𝒰(t,0))/ZE\displaystyle{\rm Tr}_{E}\left({\cal U}(0,t)\hat{\rho}^{\rm I}_{0}(t)\otimes\hat{\rho}_{\rm E}^{\rm I}(t){\cal U}(t,0)\right)/Z_{\rm E}

Here we introduce the evolution operator 𝒰(0,t)eiH^0teiH^t=𝒯^texp(i0tV^I(t)𝑑t){\cal U}(0,t)\equiv e^{i\hat{H}_{0}t}e^{-i\hat{H}t}=\hat{\cal T}_{t}\exp(-i\int_{0}^{t}\hat{V}^{\rm I}(t^{\prime})dt^{\prime}), where V^I(t)=eiH^0tV^eiH^0t\hat{V}^{\rm I}(t^{\prime})=e^{i\hat{H}_{0}t}\hat{V}e^{-i\hat{H}_{0}t} is in interaction picture, 𝒯^t\hat{\cal T}_{t} is time-ordering operator. By inserting 𝒰^(0,t)\hat{\cal U}(0,t) and 𝒰^(t,0)=𝒰^(0,t)\hat{\cal U}(t,0)=\hat{\cal U}(0,t)^{\dagger} into Eq. (5), keeping to g2g^{2} order, we have

ρ^S(t)\displaystyle\hat{\rho}_{\rm S}(t) =\displaystyle= TrE((1i0tV^I(t)dt0tdt10t1dt2V^I(t1)V^I(t2))ρ^0ρ^E\displaystyle{\rm Tr}_{E}\left((1-i\int_{0}^{t}\hat{V}^{\rm I}(t^{\prime})dt^{\prime}-\int_{0}^{t}dt_{1}\int_{0}^{t_{1}}dt_{2}\hat{V}^{\rm I}(t_{1})\hat{V}^{\rm I}(t_{2}))\hat{\rho}_{0}\otimes\hat{\rho}_{\rm E}\right. (6)
(1+i0tV^I(t)dt0tdt10t1dt2V^I(t2)V^I(t1)))/ZE.\displaystyle\left.(1+i\int_{0}^{t}\hat{V}^{\rm I}(t^{\prime})dt^{\prime}-\int_{0}^{t}dt_{1}\int_{0}^{t_{1}}dt_{2}\hat{V}^{\rm I}(t_{2})\hat{V}^{\rm I}(t_{1}))\right)/Z_{\rm E}.

More explicitly,

ρ^S(t)\displaystyle\hat{\rho}_{\rm S}(t) =\displaystyle= ρ^0+g20t𝑑t10t1𝑑t2ξ^jI(t1)ξ^I,(t2)E(𝒪^I(t2)ρ^0𝒪^jI,(t1)𝒪^jI,(t1)𝒪^I(t2)ρ^0)\displaystyle\hat{\rho}_{0}+g^{2}\int_{0}^{t}dt_{1}\int_{0}^{t_{1}}dt_{2}\langle\hat{\xi}^{\rm I}_{j}(t_{1})\hat{\xi}_{\ell}^{{\rm I},{\dagger}}(t_{2})\rangle_{\rm E}\left(\hat{\cal O}^{\rm I}_{\ell}(t_{2})\hat{\rho}_{0}\hat{\cal O}_{j}^{{\rm I},{\dagger}}(t_{1})-\hat{\cal O}^{{\rm I},{\dagger}}_{j}(t_{1})\hat{\cal O}^{\rm I}_{\ell}(t_{2})\hat{\rho}_{0}\right)
+g20t𝑑t10t1𝑑t2ξ^I,(t2)ξ^jI(t1)E(𝒪^jI,(t1)ρ^0𝒪^I(t2)𝒪^I(t2)𝒪^jI,(t1)ρ^0)\displaystyle\hskip 12.70132pt+g^{2}\int_{0}^{t}dt_{1}\int_{0}^{t_{1}}dt_{2}\langle\hat{\xi}_{\ell}^{{\rm I},{\dagger}}(t_{2})\hat{\xi}_{j}^{\rm I}(t_{1})\rangle_{\rm E}\left(\hat{\cal O}^{{\rm I},{\dagger}}_{j}(t_{1})\hat{\rho}_{0}\hat{\cal O}_{\ell}^{\rm I}(t_{2})-\hat{\cal O}^{\rm I}_{\ell}(t_{2})\hat{\cal O}^{{\rm I},{\dagger}}_{j}(t_{1})\hat{\rho}_{0}\right)
+g20t𝑑t10t1𝑑t2ξ^jI(t2)ξ^I,(t1)E(𝒪^I(t1)ρ^0𝒪^jI,(t2)ρ^0𝒪^jI,(t2)𝒪^I(t1))\displaystyle\hskip 12.70132pt+g^{2}\int_{0}^{t}dt_{1}\int_{0}^{t_{1}}dt_{2}\langle\hat{\xi}^{\rm I}_{j}(t_{2})\hat{\xi}_{\ell}^{{\rm I},{\dagger}}(t_{1})\rangle_{\rm E}\left(\hat{\cal O}^{\rm I}_{\ell}(t_{1})\hat{\rho}_{0}\hat{\cal O}_{j}^{{\rm I},{\dagger}}(t_{2})-\hat{\rho}_{0}\hat{\cal O}^{{\rm I},{\dagger}}_{j}(t_{2})\hat{\cal O}^{\rm I}_{\ell}(t_{1})\right)
+g20t𝑑t10t1𝑑t2ξ^jI,(t2)ξ^I(t1)E(𝒪^I,(t1)ρ^0𝒪^jI(t2)ρ^0𝒪^jI(t2)𝒪^I,(t1)),\displaystyle\hskip 12.70132pt+g^{2}\int_{0}^{t}dt_{1}\int_{0}^{t_{1}}dt_{2}\langle\hat{\xi}_{j}^{{\rm I},{\dagger}}(t_{2})\hat{\xi}_{\ell}^{\rm I}(t_{1})\rangle_{\rm E}\left(\hat{\cal O}^{{\rm I},{\dagger}}_{\ell}(t_{1})\hat{\rho}_{0}\hat{\cal O}_{j}^{\rm I}(t_{2})-\hat{\rho}_{0}\hat{\cal O}^{\rm I}_{j}(t_{2})\hat{\cal O}^{{\rm I},{\dagger}}_{\ell}(t_{1})\right),

where E=TrE(ρ^E)/ZE\langle\cdot\rangle_{\rm E}={\rm Tr}_{E}(\hat{\rho}_{\rm E}\cdot)/Z_{\rm E} is the environment ensemble average. Operators with upper index I{\rm I} are in interacting picture. By introducing Green’s function for ξ^\hat{\xi} field and Lindbald-like operators ^j,𝒪d,e\hat{\cal L}^{d,e}_{j\ell,{\cal O}}, we can simplify the above result as

ρ^S(t)=ρ^0+ig20t𝑑t1𝑑t2(Gξ,j<(t1,t2)^j,𝒪d(t1,t2)+Gξ,j>(t1,t2)^j,𝒪e(t1,t2))ρ^0,\displaystyle\hat{\rho}_{\rm S}(t)=\hat{\rho}_{0}\!+\!ig^{2}\!\!\iint_{0}^{t}\!\!\!dt_{1}dt_{2}\left(G^{<}_{\xi,j\ell}(t_{1},t_{2})\hat{\cal L}^{d}_{j\ell,{\cal O}}(t_{1},t_{2})\!+\!G^{>}_{\xi,j\ell}(t_{1},t_{2})\hat{\cal L}^{e}_{j\ell,{\cal O}}(t_{1},t_{2})\right)\!\hat{\rho}_{0}, (7)
Refer to caption
Figure 1: Diagrammatic illustration of Eq. (7)

where

Gξ,j<(t1,t2)iξ^jI(t1)ξ^I,(t2)E,Gξ,j>(t1,t2)iξ^I,(t1)ξ^jI(t2)E,\displaystyle G^{<}_{\xi,j\ell}(t_{1},t_{2})\equiv-i\langle\hat{\xi}_{j}^{\rm I}(t_{1})\hat{\xi}_{\ell}^{{\rm I},{\dagger}}(t_{2})\rangle_{\rm E},\hskip 12.91663ptG^{>}_{\xi,j\ell}(t_{1},t_{2})\equiv-i\langle\hat{\xi}^{{\rm I},{\dagger}}_{\ell}(t_{1})\hat{\xi}_{j}^{\rm I}(t_{2})\rangle_{\rm E}, (8)

^d\hat{\cal L}^{d} and ^e\hat{\cal L}^{e} are dissipation and enhancement operators which are defined as

^j,𝒪d(t1,t2)ρ^0\displaystyle\hat{\cal L}^{d}_{j\ell,{\cal O}}(t_{1},t_{2})\hat{\rho}_{0} \displaystyle\equiv 𝒪^I(t2)ρ^0𝒪^jI,(t1)θ12𝒪^jI,(t1)𝒪^I(t2)ρ^0θ21ρ^0𝒪^jI,(t1)𝒪^I(t2),\displaystyle\hat{\cal O}^{\rm I}_{\ell}(t_{2})\hat{\rho}_{0}\hat{\cal O}^{{\rm I},{\dagger}}_{j}(t_{1})\!-\!\theta_{12}\hat{\cal O}^{{\rm I},{\dagger}}_{j}(t_{1})\hat{\cal O}^{\rm I}_{\ell}(t_{2})\hat{\rho}_{0}\!-\!\theta_{21}\hat{\rho}_{0}\hat{\cal O}^{{\rm I},{\dagger}}_{j}(t_{1})\hat{\cal O}^{\rm I}_{\ell}(t_{2}), (9)
^j,𝒪e(t1,t2)ρ^0\displaystyle\hat{\cal L}^{e}_{j\ell,{\cal O}}(t_{1},t_{2})\hat{\rho}_{0} \displaystyle\equiv 𝒪^I,(t2)ρ^0𝒪^jI(t1)θ12𝒪^jI(t1)𝒪^I,(t2)ρ^0θ21ρ^0𝒪^jI(t1)𝒪^I,(t2).\displaystyle\hat{\cal O}_{\ell}^{{\rm I},{\dagger}}(t_{2})\hat{\rho}_{0}\hat{\cal O}^{\rm I}_{j}(t_{1})\!-\!\theta_{12}\hat{\cal O}^{\rm I}_{j}(t_{1})\hat{\cal O}^{{\rm I},{\dagger}}_{\ell}(t_{2})\hat{\rho}_{0}\!-\!\theta_{21}\hat{\rho}_{0}\hat{\cal O}^{\rm I}_{j}(t_{1})\hat{\cal O}^{{\rm I},{\dagger}}_{\ell}(t_{2}). (10)

Here θ12=θ(t1t2)\theta_{12}=\theta(t_{1}-t_{2}) is heavyside step function, which is 1 when t1>t2t_{1}>t_{2}, 1-1 when t1<t2t_{1}<t_{2}, and it is 12\frac{1}{2} when t1=t2t_{1}=t_{2}. Here we also assumed that ξ^(t)E=0\langle\hat{\xi}(t)\rangle_{\rm E}=0. This condition is usually satisfied. Here we assume the vacuum of environment is not a symmetry broken state of ξ^j\hat{\xi}_{j}.

Now we introduce a diagram to describe above result. Diagrammatically, Eq. (7) can be expressed as Fig. 1. The diagram rules are the following, we use red dot to represent 𝒪^\hat{\cal O}^{\dagger}, red cross to represent 𝒪^\hat{\cal O}. Here we dropped the subindices of operators 𝒪^\hat{\cal O}. Blue dot is ξ^\hat{\xi}^{\dagger} and blue cross is ξ^\hat{\xi}. The arrow lines are evolution operators. The half circle represent an evolution in imaginary time in the system — eβH^Se^{-\beta\hat{H}_{\rm S}} and the blue full circle is eβEH^Ee^{-\beta_{\rm E}\hat{H}_{\rm E}}. A closed circle means trace. Every dashed line connecting blue dot(cross) and red cross(dot) carries a factor gg. The length of lines in x direction represents real time, representing eiH^te^{-i\hat{H}t} when the arrow is pointing at the origin of the circle and representing eiH^te^{i\hat{H}t} in outgoing direction. A closed blue curve represents Gξ>,<G^{>,<}_{\xi}. One can find, by applying these rules, the diagrams given in Fig. 1 recover Eq. (7).

Then we are ready to calculate the Rényi entropy of the system up to g2g^{2} order. Now we calculate the second Rényi entropy, which is SRE(2)=log(Tr(ρ^S2(t)))S_{\rm RE}^{(2)}=-\log({\rm Tr}(\hat{\rho}_{\rm S}^{2}(t))). Here Tr{\rm Tr} is trace over system. First we calculate Tr(ρ^S(t)2){\rm Tr}(\hat{\rho}_{\rm S}(t)^{2}),

Tr(ρ^S(t)2)\displaystyle{\rm Tr}(\hat{\rho}_{\rm S}(t)^{2}) =\displaystyle= Tr(ρ^02)+2ig2j,0tdt1dt2(Gξ,j<(t1,t2)Tr(ρ^0(^j,𝒪d(t1,t2)ρ^0))\displaystyle{\rm Tr}(\hat{\rho}_{0}^{2})+2ig^{2}\sum_{j,\ell}\iint_{0}^{t}\!\!\!dt_{1}dt_{2}\left(G^{<}_{\xi,j\ell}(t_{1},t_{2}){\rm Tr}(\hat{\rho}_{0}(\hat{\cal L}^{d}_{j\ell,{\cal O}}(t_{1},t_{2})\hat{\rho}_{0}))\right. (11)
+Gξ,j>(t1,t2)Tr(ρ^0(^j,𝒪e(t1,t2)ρ^0))).\displaystyle\left.\hskip 116.24963pt+G^{>}_{\xi,j\ell}(t_{1},t_{2}){\rm Tr}(\hat{\rho}_{0}(\hat{\cal L}^{e}_{j\ell,{\cal O}}(t_{1},t_{2})\hat{\rho}_{0}))\right).

Above expression can also displayed in diagram language as is shown in Fig. 2.

Refer to caption
Figure 2: Perturbation digrams for Tr(ρ^S2(t)){\rm Tr}(\hat{\rho}_{\rm S}^{2}(t)), up to g2g^{2} order. Diagram representation of Eq. (11)

Now we discuss two situations, the first case is 𝒪^j\hat{\cal O}_{j} and ξ^j\hat{\xi}_{j} are both bosonic annihilation operators. Both in Green’s function and the ^\hat{\cal L} operators, there are sub-indices jj and \ell. Here we assume Gξ,j>,<=Gξ,j>,<δjG^{>,<}_{\xi,j\ell}=G^{>,<}_{\xi,j}\delta_{j\ell}. Then only ^j,𝒪d,e\hat{\cal L}^{d,e}_{j\ell,{\cal O}} contributes. In this diagonal basis, we assume Gξ,j>,<G^{>,<}_{\xi,j} and ^j,𝒪d,e\hat{\cal L}^{d,e}_{j,{\cal O}} are jj independent, then we denote Gξ,j>,<G^{>,<}_{\xi,j} asGξ>,<G^{>,<}_{\xi} and ^j,𝒪d,e\hat{\cal L}^{d,e}_{j,{\cal O}} as ^𝒪d,e\hat{\cal L}^{d,e}_{{\cal O}} for simplicity. According to the relation between the greater (lesser) Green’s function and the spectral function, we have

Gξ>(t1,t2)=i𝑑ω𝒜ξ(ω)1eβEω1eiω(t1t2),\displaystyle G_{\xi}^{>}(t_{1},t_{2})=-i\int d\omega^{\prime}{\cal A}_{\xi}(\omega^{\prime})\frac{1}{e^{\beta_{E}\omega^{\prime}}-1}e^{i\omega^{\prime}(t_{1}-t_{2})}, (12)
Gξ<(t2,t1)=i𝑑ω𝒜ξ(ω)eβEωeβEω1eiω(t1t2),\displaystyle G_{\xi}^{<}(t_{2},t_{1})=-i\int d\omega^{\prime}{\cal A}_{\xi}(\omega^{\prime})\frac{e^{\beta_{E}\omega^{\prime}}}{e^{\beta_{E}\omega^{\prime}}-1}e^{i\omega^{\prime}(t_{1}-t_{2})}, (13)

where 𝒜ξ(ω){\cal A}_{\xi}(\omega) is the spectral function of ξ^\hat{\xi} field, 𝒜ξ(ω)=1πImGξR(ω){\cal A}_{\xi}(\omega)=-\frac{1}{\pi}{\rm Im}G^{R}_{\xi}(\omega) where GξR(t)=iθ(t)[ξ^(t),ξ(0)]G_{\xi}^{R}(t)=-i\theta(t)\langle[\hat{\xi}(t),\xi^{\dagger}(0)]\rangle is the retard Green’s function of ξ^\hat{\xi} field. 1/βE1/\beta_{E} is the environment temperature. Here we can see, a white noise bath can be generated in two cases. At high temperature, βE\beta_{E} is close to zero, therefore when 𝒜ξ(ω)ω{\cal A}_{\xi}(\omega)\sim\omega, Gξ>,<δ(t1t2)G_{\xi}^{>,<}\propto\delta(t_{1}-t_{2}). This is Ohm bath at high temperature. At low temperature, we find Gξ>0G^{>}_{\xi}\sim 0, and a white noise is generated by a constant density of states 𝒜ξconst.{\cal A}_{\xi}\sim{\rm const.}. This is the other case of white noise.

On the other hand, we find Tr(ρ^0^𝒪d(t1,t2))=Tr(ρ^0^𝒪e(t2,t1)){\rm Tr}(\hat{\rho}_{0}\hat{\cal L}^{d}_{{\cal O}}(t_{1},t_{2}))={\rm Tr}(\hat{\rho}_{0}\hat{\cal L}^{e}_{{\cal O}}(t_{2},t_{1})) is a two-point Wightman Green’s function,

Tr(ρ^0𝒪^(t1)ρ^0𝒪^(t2))=Tr(ρ^0𝒪^ρ^0𝒪^(t2t1))=dω2π𝒜𝒪(ω)eβωe2βω1eiω(t2t1),\displaystyle{\rm Tr}\left(\hat{\rho}_{0}\hat{\cal O}^{\dagger}(t_{1})\hat{\rho}_{0}\hat{\cal O}(t_{2})\right)={\rm Tr}\left(\hat{\rho}_{0}\hat{\cal O}^{\dagger}\hat{\rho}_{0}\hat{\cal O}(t_{2}-t_{1})\right)=\int\frac{d\omega}{2\pi}{\cal A}_{\cal O}(\omega)\frac{e^{\beta\omega}}{e^{2\beta\omega}-1}e^{i\omega(t_{2}-t_{1})}, (14)

where 𝒜𝒪(ω){\cal A}_{\cal O}(\omega) is the spectral function of 𝒪{\cal O} field. The last equation in above line can be proved by Lehmann spectrum Theorem. By inserting Eq. (12), Eq. (13) and Eq. (14) into Eq. (11), the deviation of Renyi entropy δSRE(2)(t)SRE(2)(t)+log(Tr(ρ^02))\delta S_{\rm RE}^{(2)}(t)\equiv S_{\rm RE}^{(2)}(t)+\log({\rm Tr}(\hat{\rho}_{0}^{2})) is

δSRE(2)(t)=2g2NjTr(ρ02)0t𝑑t0𝑑ω𝑑ωsin(ωω)t0ωω𝒲ββEB(ω,ω)𝒜ξ(ω)𝒜𝒪(ω)\displaystyle\delta S_{\rm RE}^{(2)}(t)=\frac{2g^{2}N_{j}}{{\rm Tr}(\rho_{0}^{2})}\int_{0}^{t}dt_{0}d\omega d\omega^{\prime}\frac{\sin(\omega-\omega^{\prime})t_{0}}{\omega-\omega^{\prime}}{\cal W}_{\beta\beta_{E}}^{B}(\omega,\omega^{\prime}){\cal A}_{\xi}(\omega^{\prime}){\cal A}_{\cal O}(\omega) (15)

where NjN_{j} is the mode number, and

𝒲ββEB(ω,ω)=(eβω1)(eβωeβEω)(e2βω1)(eβEω1).\displaystyle{\cal W}_{\beta\beta_{E}}^{B}(\omega,\omega^{\prime})=\frac{(e^{\beta\omega}-1)(e^{\beta\omega}-e^{\beta_{E}\omega^{\prime}})}{(e^{2\beta\omega}-1)(e^{\beta_{E}\omega^{\prime}}-1)}. (16)

At high temperature limit and 𝒜ξ{\cal A}_{\xi} being Ohm bath, we have δSRE(2)λ0t\delta S_{\rm RE}^{(2)}\sim\lambda_{0}t where λ0=2g2/Tr(ρ02)𝑑ω(eβω1)𝒜𝒪(ω)/(eβω+1)\lambda_{0}=2g^{2}/{\rm Tr}(\rho_{0}^{2})\int d\omega(e^{\beta\omega}-1){\cal A}_{\cal O}(\omega)/(e^{\beta\omega}+1) is the entropy growth rate. According to this formula, we learn, the entropy response is only dependent on the distribution function and the spectrum of bath and the system.

As the entropy response theory at g2Njg^{2}N_{j} order is indeed an exponential contribution in Tr(ρ^S2(t)){\rm Tr}(\hat{\rho}^{2}_{S}(t))’s perturbation, therefore the response theory for entropy may have larger applicable region. Actually, we can apply a perturbation calculation to next order g4g^{4}, where we can capture how the spectrum function of the system and the environment is changed by interactions. But these back action effects are not considered in present calculation.

Quite similarly, we can work out the Rényi entropy response for fermionic bath and fermionic operator 𝒪^\hat{\cal O}, which is

δSRE(2)(t)=2g2NjTr(ρ02)0t𝑑t0𝑑ω𝑑ωsin(ωω)t0ωω𝒲ββEF(ω,ω)𝒜ξ(ω)𝒜𝒪(ω),\displaystyle\delta S_{\rm RE}^{(2)}(t)=\frac{2g^{2}N_{j}}{{\rm Tr}(\rho_{0}^{2})}\int_{0}^{t}dt_{0}d\omega d\omega^{\prime}\frac{\sin(\omega-\omega^{\prime})t_{0}}{\omega-\omega^{\prime}}{\cal W}_{\beta\beta_{E}}^{F}(\omega,\omega^{\prime}){\cal A}_{\xi}(\omega^{\prime}){\cal A}_{\cal O}(\omega), (17)

where

𝒲ββEF(ω,ω)=(eβω1)(eβωeβEω)(1+e2βω)(1+eβEω).\displaystyle{\cal W}_{\beta\beta_{E}}^{F}(\omega,\omega^{\prime})=\frac{(e^{\beta\omega}-1)(e^{\beta\omega}-e^{\beta_{E}\omega^{\prime}})}{(1+e^{2\beta\omega})(1+e^{\beta_{E}\omega^{\prime}})}. (18)

Notice that in fermionic case, the Markovian condition is requiring a frequency independent spectrum together with frequency independent distribution function. The latter condition is satisfied only for zero and infinite temperature. Therefore there are two ways to break down Markovian approximation, one is by finite temperature; another is by adding strong frequency dependence in the noise spectrum of the environment.

2.1 General Features for Bosonic systems

In this section, we first discuss the short time and long time behavior of entanglement entropy growth for bosonic environment. Then we derive a “ Fourier law ” in heat transport when the system and the environment have small temperature difference.

First, we discuss the short time and long time dynamics of δSRE(2)\delta S_{\rm RE}^{(2)}. When tt is small, in a sense 𝒥t1{\cal J}t\ll 1, 𝒥t1{\cal J}^{\prime}t\ll 1 ( 𝒥{\cal J} and 𝒥{\cal J}^{\prime} are typical energy scale for the system and the bath respectively ), we have

sin(ωω)t0ωωt0\displaystyle\frac{\sin(\omega-\omega^{\prime})t_{0}}{\omega-\omega^{\prime}}\approx t_{0}

For this reason, the short time dynamics is always proportional to t2t^{2}.

δSRE(2)(t)=12κBt2\displaystyle\delta S_{\rm RE}^{(2)}(t)=\frac{1}{2}\kappa_{\rm B}t^{2} (19)

where coefficient is

κB2g2NjTr(ρ02)𝑑ω𝑑ω𝒲ββEB(ω,ω)𝒜𝒪(ω)𝒜ξ(ω).\displaystyle\kappa_{\rm B}\equiv\frac{2g^{2}N_{j}}{{\rm Tr}(\rho_{0}^{2})}\int d\omega\int d\omega^{\prime}{\cal W}_{\beta\beta_{E}}^{B}(\omega,\omega^{\prime}){\cal A}_{\cal O}(\omega){\cal A}_{\xi}(\omega^{\prime}). (20)

Bosonic environment is unique because the chemical potential of Bose distribution is negative ( although in our formula we take chemical potential to be zero ), and the frequency of bosonic excitations are always positive. Therefore when βE>β\beta_{\rm E}>\beta, and the environment has a lower temperature, κB\kappa_{\rm B} is usually negative. That is due to 𝒜𝒪(ω){\cal A}_{\cal O}(\omega), 𝒜ξ(ω){\cal A}_{\xi}(\omega) are positive spectral functions. Further

𝒲ββEB(ω,ω)=eβωeβEω(eβω+1)(eβEω1)eβωβEω1eβω+1\displaystyle{\cal W}_{\beta\beta_{E}}^{B}(\omega,\omega^{\prime})=\frac{e^{\beta\omega}-e^{\beta_{\rm E}\omega^{\prime}}}{(e^{\beta\omega}+1)(e^{\beta_{\rm E}\omega^{\prime}}-1)}\approx\frac{e^{\beta\omega-\beta_{\rm E}\omega^{\prime}}-1}{e^{\beta\omega}+1}

As ω\omega, ω\omega^{\prime} are positive, as long as βω<βEω\beta\omega<\beta_{\rm E}\omega^{\prime}, the factor 𝒲ββEB(ω,ω){\cal W}_{\beta\beta_{E}}^{B}(\omega,\omega^{\prime}) is negative. As long as the low energy spectrum of the bath and the system are large, then κB<0\kappa_{\rm B}<0. Later we will see it is completely different in fermionic bath case.

Now we turn to long time case. Here we assume time is long and 𝒥t0𝒥t01{\cal J}t_{0}\sim{\cal J}^{\prime}t_{0}\sim 1, we have

sin(ωω)t0ωωδ(ωω)\displaystyle\frac{\sin(\omega-\omega^{\prime})t_{0}}{\omega-\omega^{\prime}}\sim\delta(\omega-\omega^{\prime})

Therefore the long time dynamics is approximately

δSRE(2)(t)=2g2NjTr(ρ02)t𝑑ω𝒲ββEB(ω,ω)𝒜ξ(ω)𝒜𝒪(ω)\displaystyle\delta S_{\rm RE}^{(2)}(t)=\frac{2g^{2}N_{j}}{{\rm Tr}(\rho_{0}^{2})}t\int d\omega{\cal W}_{\beta\beta_{E}}^{B}(\omega,\omega){\cal A}_{\xi}(\omega){\cal A}_{\cal O}(\omega) (21)

The sign for the entropy growth rate in long time limit is determined by the sign of 𝒲ββEB(ω,ω){\cal W}_{\beta\beta_{E}}^{B}(\omega,\omega). 𝒲ββEB(ω,ω)>0{\cal W}_{\beta\beta_{E}}^{B}(\omega,\omega)>0 when βE<β\beta_{\rm E}<\beta. In this situation the environment’s temperature is higher than the system. 𝒲ββEB(ω,ω)<0{\cal W}_{\beta\beta_{E}}^{B}(\omega,\omega)<0 when βE>β\beta_{\rm E}>\beta. In this situation, the environment has a lower temperature than the system.

Now we are going to derive a Fourier heat transport formula from our Eq. (21). Now we assume two temperature T=1/βT=1/\beta and TE=1/βET_{\rm E}=1/\beta_{\rm E} are close. Here we have taken kB=1k_{\rm B}=1. Further we assume βE\beta_{\rm E} is small, then we have

𝒲ββEB(ω,ω)=1eβω+1e(ββE)ω11eβEωTETT1(eβω+1)\displaystyle{\cal W}_{\beta\beta_{E}}^{B}(\omega,\omega)=\frac{1}{e^{\beta\omega}+1}\frac{e^{(\beta-\beta_{\rm E})\omega}-1}{1-e^{-\beta_{\rm E}\omega}}\approx\frac{T_{\rm E}-T}{T}\frac{1}{(e^{\beta\omega}+1)} (22)

Insert Eq. (22) into Eq. (21), we have

TδSRE(2)(t)=2g2NjTr(ρ02)(0𝑑ω𝒜𝒪(ω)𝒜ξ(ω)eβω+1)(TET)t\displaystyle T\delta S_{\rm RE}^{(2)}(t)=\frac{2g^{2}N_{j}}{{\rm Tr}(\rho_{0}^{2})}\left(\int_{0}^{\infty}d\omega\frac{{\cal A}_{\cal O}(\omega){\cal A}_{\xi}(\omega)}{e^{\beta\omega}+1}\right)(T_{\rm E}-T)t (23)

If we can define dQTδSRE(2)(t)dQ\equiv T\delta S_{\rm RE}^{(2)}(t) as some kind of heat, therefore above equation shows a Fourier law of heat transport. The heat transport rate is proportional to the temperature difference between the system and the bath. If we define FQ=dQ/dt/(TET)F_{\rm Q}=dQ/dt/(T_{\rm E}-T), then the transport coefficient FQF_{\rm Q} is predicted as

FQ=2g2NjTr(ρ02)0𝑑ω𝒜𝒪(ω)𝒜ξ(ω)eβω+1.\displaystyle F_{Q}=\frac{2g^{2}N_{j}}{{\rm Tr}(\rho_{0}^{2})}\int_{0}^{\infty}d\omega\frac{{\cal A}_{\cal O}(\omega){\cal A}_{\xi}(\omega)}{e^{\beta\omega}+1}. (24)

2.2 General Features for Fermionic systems

In this section, we discuss the short time and long time entropy growth for fermionic bath. Meanwhile we will give the “ heat transport ” coefficient for fermionic bath.

First we discuss the short time behavior of δSRE(2)(t)\delta S_{\rm RE}^{(2)}(t). Similar to the case of bosonic system, we have

δSRE(2)(t)=12κFt2,\displaystyle\delta S_{\rm RE}^{(2)}(t)=\frac{1}{2}\kappa_{\rm F}t^{2}, (25)

where

κF=2g2NjTr(ρ02)𝑑ω𝑑ω𝒲ββEF(ω,ω)𝒜𝒪(ω)𝒜ξ(ω).\displaystyle\kappa_{\rm F}=\frac{2g^{2}N_{j}}{{\rm Tr}(\rho_{0}^{2})}\int d\omega\int d\omega^{\prime}{\cal W}_{\beta\beta_{E}}^{F}(\omega,\omega^{\prime}){\cal A}_{\cal O}(\omega){\cal A}_{\xi}(\omega^{\prime}). (26)

The only difference with Eq. (20) is the distribution function 𝒲ββEB{\cal W}_{\beta\beta_{E}}^{B} is replaced by 𝒲ββEF{\cal W}_{\beta\beta_{E}}^{F}. For fermions, negative frequency is meaningful because of the presence of Fermi level. Here Fermi level is taken to be zero. At low temperature,

𝒲ββEF(ω,ω)(eβω1)(eβωβEω1)e2βω+1\displaystyle{\cal W}_{\beta\beta_{E}}^{F}(\omega,\omega^{\prime})\approx\frac{(e^{\beta\omega}-1)(e^{\beta\omega-\beta_{\rm E}\omega^{\prime}}-1)}{e^{2\beta\omega}+1} (27)

It is clear that when ω<0\omega<0, ω>0\omega^{\prime}>0, 𝒲ββEF(ω,ω)>0{\cal W}_{\beta\beta_{E}}^{F}(\omega,\omega^{\prime})>0. Therefore κF\kappa_{\rm F} is usually positive rather than being negative. This is very different from boson case.

Now we consider the long time behavior of δSRE(2)(t)\delta S_{\rm RE}^{(2)}(t). In long time limit, we have

δSRE(2)(t)=2g2NjTr(ρ02)t𝑑ω𝒲ββEF(ω,ω)𝒜ξ(ω)𝒜𝒪(ω).\displaystyle\delta S_{\rm RE}^{(2)}(t)=\frac{2g^{2}N_{j}}{{\rm Tr}(\rho_{0}^{2})}t\int d\omega{\cal W}_{\beta\beta_{E}}^{F}(\omega,\omega){\cal A}_{\xi}(\omega){\cal A}_{\cal O}(\omega). (28)

One can prove that for βE>β\beta_{\rm E}>\beta, 𝒲ββEF(ω,ω)<0{\cal W}_{\beta\beta_{E}}^{F}(\omega,\omega)<0 for any ω\omega. Therefore if the environment has a larger initial temperature, δSRE(2)(t)\delta S_{\rm RE}^{(2)}(t) experienced a linear growth and if the environment temperature is lower than the system, δSRE(2)(t)\delta S_{\rm RE}^{(2)}(t) is decreasing linearly in long time evolution. However, as we state previously, the short time dynamics of low temperature fermionic bath is an entropy growth, therefore the entropy dynamics could be non-monotonic. Later in the next section, we will show this case in a specific example. Here we stress that the kink in δSRE(2)(t)\delta S_{\rm RE}^{(2)}(t) could be very general in fermionic case, not necessarily as an interpretation of gravity duality.

Finally, we discuss the situation when the system and the bath’s temperature difference is small. Again a Fourier-alike law can be presented,

𝒲ββEF(ω,ω)=eβω1e2βω+1e(ββE)ω11+eβEω(eβω1)ωββE(e2βω+1)(1+eβEω)(TET)\displaystyle{\cal W}_{\beta\beta_{E}}^{F}(\omega,\omega)=\frac{e^{\beta\omega}-1}{e^{2\beta\omega}+1}\frac{e^{(\beta-\beta_{\rm E})\omega}-1}{1+e^{-\beta_{\rm E}\omega}}\approx\frac{(e^{\beta\omega}-1)\omega\beta\beta_{\rm E}}{(e^{2\beta\omega}+1)(1+e^{-\beta_{\rm E}\omega})}(T_{\rm E}-T)

Again we can get a “ Fourier law ” in “ heat transport ” if we define TδSRE(2)(t)T\delta S_{\rm RE}^{(2)}(t) as “ heat flow ” — dQdQ. Again we define FQ=dQ/dt/(TET)F_{\rm Q}=dQ/dt/(T_{\rm E}-T), then

FQ=2g2NjTr(ρ02)0𝑑ω𝒜𝒪(ω)𝒜ξ(ω)(eβω1)ωβE(e2βω+1)(1+eβEω).\displaystyle F_{\rm Q}=\frac{2g^{2}N_{j}}{{\rm Tr}(\rho_{0}^{2})}\int_{0}^{\infty}d\omega\frac{{\cal A}_{\cal O}(\omega){\cal A}_{\xi}(\omega)(e^{\beta\omega}-1)\omega\beta_{\rm E}}{(e^{2\beta\omega}+1)(1+e^{-\beta_{\rm E}\omega})}. (29)

One can observe that the temperature dependence in FQF_{\rm Q} is very different from bosonic bath case.

3 Renyi Entropy Response in SYK model

In this section, we apply the general Rényi entropy response theory to Sachdev-Ye-Kitaev (SYK) model with an non-Markovian environment. We assume the system is SYK4 whose hamiltonian is,

H^S=jklmJjklmχ^jχ^kχ^lχ^m,\displaystyle\hat{H}_{\rm S}=\sum_{jklm}J_{jklm}\hat{\chi}_{j}\hat{\chi}_{k}\hat{\chi}_{l}\hat{\chi}_{m}, (30)

where χ^j\hat{\chi}_{j} is Majorana fermion operator. JjklmJ_{jklm} are random, and we have

Jjklm¯=0,Jjklm2¯=3!N3𝒥2,\displaystyle\overline{J_{jklm}}=0,\hskip 17.22217pt\overline{J_{jklm}^{2}}=\frac{3!}{N^{3}}{\cal J}^{2}, (31)

where ¯\overline{\cdot} is disorder average, NN is the total modes number. Throughout our discussion we are in conformal limit of SYK model, so that Nβ𝒥1N\gg\beta{\cal J}\gg 1.

Here we are going to consider two kinds of baths. One bath is a SYK2 model with zero chemical potential. In this model, when the coupling strength is large, then the spectral function around zero frequency is a constant. This kind of bath has a Markovian limit. Another bath is two SYK2 model with opposite on site energies, which is more like a gapped system and is far away from Markovian environment. Here we stress that non-Markovian effect comes from both the frequency dependence in spectral function of the environment as well as the the distribution function kernel 𝒲F{\cal W}^{F}. Therefore we separate these two factors with a constant spectral function and a strongly frequency dependent spectral fucntion. In the former, the non-Markovian effect comes from the distribution function, while in the latter, the non-Markovian effect comes from the spectral function.

3.1 Heating and Evaporation with Markovian-like Environment

Here we assume the environment is random free fermions, SYK2, whose hamiltonian is

H^E=α=1Mjk=1NJjkψ^j,αψ^k,α,\displaystyle\hat{H}_{\rm E}=\sum_{\alpha=1}^{M}\sum_{jk=1}^{N}J^{\prime}_{jk}\hat{\psi}_{j,\alpha}^{\dagger}\hat{\psi}_{k,\alpha}, (32)

where ψ^j\hat{\psi}_{j} are complex fermion annihilation operators. JjkJ^{\prime}_{jk} is random, and it satisfies

Jjk¯=0,(Jjk)2¯=1N(𝒥)2\displaystyle\overline{J^{\prime}_{jk}}=0,\hskip 17.22217pt\overline{(J_{jk}^{\prime})^{2}}=\frac{1}{N}({\cal J}^{\prime})^{2} (33)

The interaction between the system and the bath is V^\hat{V},

V^=gj=1Nα=1M(χ^jψ^j,α+ψ^j,αχ^j)\displaystyle\hat{V}=g\sum_{j=1}^{N}\sum_{\alpha=1}^{M}(\hat{\chi}_{j}\hat{\psi}^{\dagger}_{j,\alpha}+\hat{\psi}_{j,\alpha}\hat{\chi}_{j}) (34)

One advantage is that SYK model is known to have a gravitaional duality. SYK model’s low energy effective action can be described by a Schwarzian theory, which is dual to the Jackiw-Teitelboim (JT) gravity theory in AdS2 spacetimeKitaev17 ; Maldacena16 . The present situation of sudden coupling between the SYK4 and SYK2 model can be viewed as an black hole evaporation problem if the environment’s temperature is much lower than the black hole. In gravitational side, it is known that evaporation experience an entanglement entropy growth and drop, which is well-known as Page curve. The entanglement entropy calculation based on RT formula is done recentlyPenington19 ; Almheiri19 ; Maldacena20 towards an understanding of Page curve. Therefore, according to holographic principle, we expect similar non-monotonic behavior in SYK4’s entanglement entropy dynamics.

Now we apply the perturbation theory to calculate the Rényi entropy dynamics after a sudden coupling between SYK4 and SYK2 by V^\hat{V} to see how the second Rényi entropy response to the sudden coupling between these two systems.

From our formula, we find the most important information for Rényi entropy response is the spectrum of the system and the spectrum of the environment. These spectrum can be obtained by calculating retard Green’s function. Compared with the general theory, we find 𝒪^j=χ^j\hat{\cal O}_{j}=\hat{\chi}_{j} and ξ^j=ψ^j,α\hat{\xi}_{j}=\hat{\psi}_{j,\alpha}. Then the spectral function for 𝒪^j\hat{\cal O}_{j} field is

𝒜𝒪(ω)=bβΓ(14iβω/2π)Γ(34iβω/2π),\displaystyle{\cal A}_{\cal O}(\omega)=\frac{b}{\sqrt{\beta}}\frac{\Gamma(\frac{1}{4}-i\beta\omega/2\pi)}{\Gamma(\frac{3}{4}-i\beta\omega/2\pi)}, (35)
Refer to caption
Figure 3: Heating of the “ Black Hole ”. The figure shows the Rényi entropy response of SYK4 with 𝒥=4{\cal J}=4 and β=10\beta=10 coupling to SYK2 model with 𝒥=8{\cal J}^{\prime}=8 and βE=2\beta_{\rm E}=2. A short time t2t^{2} growth and a long time tt linear growth are shown.

where b=(1/4π𝒥2)1/4b=(1/4\pi{\cal J}^{2})^{1/4}. The spectral function of the environment is

𝒜ξ(ω)=(𝒥)2ω2.\displaystyle{\cal A}_{\xi}(\omega)=\sqrt{({\cal J}^{\prime})^{2}-\omega^{2}}. (36)

According to the entropy response theory

δSRE(2)(t)=4g2NMTr(ρ^02)0t/2𝑑t0𝒥𝒥𝑑ω𝒥𝒥𝑑ωsin(ωω)t0ωω𝒲ββEF(ω,ω)𝒜𝒪(ω)𝒜ξ(ω)\displaystyle\delta S_{\rm RE}^{(2)}(t)=\frac{4g^{2}NM}{{\rm Tr}(\hat{\rho}_{0}^{2})}\int_{0}^{t/2}dt_{0}\int_{-{\cal J^{\prime}}}^{\cal J^{\prime}}\!\!\!d\omega^{\prime}\int_{-{\cal J}}^{\cal J}\!\!\!d\omega\frac{\sin(\omega-\omega^{\prime})t_{0}}{\omega-\omega^{\prime}}{\cal W}_{\beta\beta_{E}}^{F}(\omega,\omega^{\prime}){\cal A}_{\cal O}(\omega){\cal A}_{\xi}(\omega^{\prime}) (37)

In extremely short time, sin(ωω)t0(ωω)t0\sin(\omega-\omega^{\prime})t_{0}\sim(\omega-\omega^{\prime})t_{0}, therefore we have

δSRE(2)(t)=12κFt2,\displaystyle\delta S_{\rm RE}^{(2)}(t)=\frac{1}{2}\kappa_{F}t^{2}, (38)

where κF=8g2NM/Tr(ρ^02)𝒥𝒥𝑑ω𝒥𝒥𝑑ω𝒲ββEF(ω,ω)𝒜𝒪(ω)𝒜ξ(ω)\kappa_{F}=8g^{2}NM/{\rm Tr}(\hat{\rho}_{0}^{2})\int_{-{\cal J^{\prime}}}^{\cal J^{\prime}}d\omega^{\prime}\int_{-{\cal J}}^{\cal J}d\omega{\cal W}_{\beta\beta_{E}}^{F}(\omega,\omega^{\prime}){\cal A}_{\cal O}(\omega){\cal A}_{\xi}(\omega^{\prime}). In longer time, sin(ωω)t0/(ωω)=δ(ωω)\sin(\omega-\omega^{\prime})t_{0}/(\omega-\omega^{\prime})=\delta(\omega-\omega^{\prime}), therefore

δSRE(2)(t)=λt,\displaystyle\delta S_{\rm RE}^{(2)}(t)=\lambda t, (39)

where λ\lambda is the entropy rate, whose value is

λ=4g2NMTr(ρ^02)min(𝒥,𝒥)min(𝒥,𝒥)𝑑ω𝒲ββEF(ω,ω)𝒜𝒪(ω)𝒜ξ(ω).\displaystyle\lambda=\frac{4g^{2}NM}{{\rm Tr}(\hat{\rho}_{0}^{2})}\int_{-{\rm min}({\cal J^{\prime}},{\cal J})}^{{\rm min}({\cal J^{\prime}},{\cal J})}\!\!\!d\omega{\cal W}_{\beta\beta_{E}}^{F}(\omega,\omega){\cal A}_{\cal O}(\omega){\cal A}_{\xi}(\omega). (40)

Here min(a,b)\min(a,b) takes the smaller one in aa and bb. Here λ\lambda can be negative. The general picture for the entanglement entropy growth is a quadratic growth at early time and then crossover to a linear growth ( or decrease ) at late time. As the late time behavior can be a decrease of entropy, therefore we can see the signal of evaporation in our perturbative calculation even if it is not early time behavior.

Here we stress that for sin(ωω)t0/(ωω)δ(ωω)\sin(\omega-\omega^{\prime})t_{0}/(\omega-\omega^{\prime})\propto\delta(\omega-\omega^{\prime}), we need a really large t0t_{0}. Indeed this approximation is better satisfied when min(𝒥,𝒥)t0{\rm min}({\cal J}^{\prime},{\cal J})t_{0} is really large. Therefore before that time scale, there are always oscillations in δSRE(2)(t)\delta S^{(2)}_{\rm RE}(t), but these oscillations are slow. In long time limit, we find only if β=βE\beta=\beta_{\rm E}, we have δSRE(2)=0\delta S^{(2)}_{\rm RE}=0. However, in some period of time, δSRE(2)0\delta S^{(2)}_{\rm RE}\sim 0 for quite remarkable time scale, which may seems like prethermalization. In this section, we only focus on the situation where β\beta and βE\beta_{\rm E} has large difference.

Refer to caption
Figure 4: Evaporation of "Black Hole", In (a) we show the spectral function of the system ( SYK4 model , β=2\beta=2, 𝒥=2{\cal J}=2 ) and the environment ( SYK2 model with β=20\beta=20, 𝒥=2{\cal J}^{\prime}=2 ); In (b), we show the Rényi entropy response of the sudden coupling between SYK model and the environment. As the environment is at low temperature, it is an evaporation process. We can see a linear growth follows a quadratic growth, then a sharp linear drop follows. The shape of the curve is alike the Page curve in black hole evaporation. g2NMg^{2}NM is tuned for showing the figures, but these values are not important at this level.

In the following, we present the numerical calculation for mainly two kinds of process. First of all, we discuss the situation of heating where the system is at low temperature and the environment is at high temperature. As is shown in Fig. 3, we couple a SYK4 model with 𝒥=4{\cal J}=4, β=10\beta=10 to a SYK2 model with 𝒥=8{\cal J}^{\prime}=8 and βE=2\beta_{\rm E}=2. This situation mimic a Black Hole coupling to a hot environment. Just as our analysis in the last section, the Rényi entropy experienced a quadratic growth in the short time limit and then crossover to a linear growth in longer time scale.

Second, we study a SYK4 model with 𝒥=2{\cal J}=2, β=2\beta=2 coupling to a SYK2 model with 𝒥=2{\cal J}^{\prime}=2 and βE=20\beta_{\rm E}=20. The environment is at low temperature, hence it describes a “ Black Hole ” evaporation process. A Page curve alike behavior is expected and we find that even if the method is perturbative, the non-monotonic behavior can be captured as is shown in Fig. 4.

3.2 Heating and Evaporation with non-Markovian Environment

Here we assume the environment’s hamiltonian is

H^E=α=1Mjk=1N(Jjk,αLμ)ψ^j,αLψ^k,αL+α=1Mjk=1N(Jjk,αR+μ)ψ^j,αRψ^k,αR,\displaystyle\hat{H}_{\rm E}=\sum_{\alpha=1}^{M}\sum_{jk=1}^{N}(J^{\prime}_{jk,\alpha L}-\mu)\hat{\psi}_{j,\alpha L}^{\dagger}\hat{\psi}_{k,\alpha L}+\sum_{\alpha=1}^{M}\sum_{jk=1}^{N}(J^{\prime}_{jk,\alpha R}+\mu)\hat{\psi}_{j,\alpha R}^{\dagger}\hat{\psi}_{k,\alpha R}, (41)

where

Jjk,αL(R)¯=0,(Jjk,αL(R))2¯=1N(𝒥)2.\displaystyle\overline{J^{\prime}_{jk,\alpha L(R)}}=0,\hskip 17.22217pt\overline{(J_{jk,\alpha L(R)}^{\prime})^{2}}=\frac{1}{N}({\cal J}^{\prime})^{2}. (42)
Refer to caption
Figure 5: Here we take 𝒥=4{\cal J}=4, β=10\beta=10, 𝒥=0.8{\cal J}^{\prime}=0.8. In (a) we give the spectral function of the system and the environment. The dashed line is for μ=0.\mu=0., dot-dashed line is for μ=1.2\mu=1.2 and the solid line is for μ=2\mu=2. The gap of the spectrum is increasing from 0 to 1.2. In (b) we show the Rényi entropy response for βE=2\beta_{\rm E}=2, μ=0.\mu=0. as dashed line; μ=1.2\mu=1.2 case as dot-dashed line and μ=2.\mu=2. as solid line.

The interactions between the system and the environment is

V^=gj=1Nα=1Mσ=L,R(χ^jψ^j,ασ+χ^jψ^j,ασ)\displaystyle\hat{V}=g\sum_{j=1}^{N}\sum_{\alpha=1}^{M}\sum_{\sigma=L,R}(\hat{\chi}_{j}\hat{\psi}^{\dagger}_{j,\alpha\sigma}+\hat{\chi}_{j}\hat{\psi}_{j,\alpha\sigma}) (43)

For our problem, this model change is equivalent to a simple change in environment’s spectral function.

𝒜ξ(ω)=1π𝒥2(𝒥,2(ωμ)2+𝒥,2(ω+μ)2)\displaystyle{\cal A}_{\xi}(\omega)=\frac{1}{\pi{\cal J^{\prime}}^{2}}\left(\sqrt{{\cal J}^{\prime,2}-(\omega-\mu)^{2}}+\sqrt{{\cal J}^{\prime,2}-(\omega+\mu)^{2}}\right) (44)

Here we assume μ>𝒥\mu>{\cal J}^{\prime}, such that we have a gapped system. Here we take 𝒥=0.5{\cal J}^{\prime}=0.5, μ=1\mu=1. Again we apply Eq. (15), then we show the numerical results in Fig. 5 . We compared the results with the same parameter 𝒥{\cal J}^{\prime} with previous spectral function. For a case where βE<β\beta_{\rm E}<\beta, the presence of a gap can speed up the entanglement entropy decreasing.

Refer to caption
Figure 6: In (a) we show the RE response for environment with gapless spectrum, β=10\beta=10, 𝒥=4{\cal J}=4, βE=4.5\beta_{\rm E}=4.5, 8 and 12 respectively with solid line, dashed line and dot-dashed line, 𝒥=0.8{\cal J}^{\prime}=0.8; In (b) we show an oscillation behavior of δSRE(2)\delta S_{\rm RE}^{(2)} when βEβ\beta_{\rm E}\approx\beta. Here β\beta is fixed as 1010. Here the solid line has a temperature βE=4.5\beta_{\rm E}=4.5. The dashed line is at the temperature βE=8\beta_{\rm E}=8 and the dot-dashed line is at the environment temperature βE=12\beta_{\rm E}=12.

A very non-trivial and interesting dynamical behavior is observed that the entropy experiences an oscillation before a late time linear growth. This situation happens when the environment temperature is close to the system’s temperature. We find the gap in the spectrum can result non-trivial memory effect in entropy dynamics where the heat flow direction can change at least twice. Deeper reasons for the extra oscillation behavior in RE dynamics is remain to be further understood. By comparing the behavior of the gapless spectrum and the gapped spectrum under similar circumstance, we find the gapped spectrum bring the oscillation of RE to shorter time scale.

4 Conclusion

In summary, we construct a perturbative method for calculating the Rényi entropy for a sudden coupling between a system and an environment. On condition that the coupling between two systems are linear coupling, we obtained general formula for Rényi entropy response for bosonic systems and fermionic systems respectively. One can find the entropy response is related to the spectral function of the system and the environment as well as a distribution kernel function 𝒲B,F{\cal W}^{B,F}. We find the short time behavior of Rényi entropy response follows a t2t^{2} law, and later it follows a tt linear law. Fermi statistics can result a non-monotonic behavior in Rényi entropy dynamics which may have an gravitational explanation as a Page curve in black evaporation. Further, among two reasons for a bath being non-Markovian, we find the distribution kernel plays the vital role in the behavior of entanglement entropy dynamics, while the spetral function of environment controls the detail behavior of the entropy dynamics. Further understanding of these perturbative results in gravity language, as well as the back action of the environment to the sudden coupling of the system are left for further study.

At the closing stage of this paper, we notice a similar paper on this topic as arXiv: 2011.09622 by Dadras and KitaevKitaev20 .

Acknowledgements.
We thank Pengfei Zhang, Yiming Chen and Hui Zhai for discussions. This work is supported by NSFC under Grant No. 11734010, Beijing Natural Science Foundation (Z180013).

References

  • (1) G. D. Mahan, Many Particle Physics. New York and London: Plenum Press (1981).
  • (2) L. Pan, X. Chen, Y. Chen and H. Zhai, Non-Hermitian Linear Response Theory, Nat. Phys. 16 (2020) 767-771.
  • (3) R. Bouganne, M. B. Aguilera, A. Ghermaoui, Beugnon, J. and Gerbier, F. Anomalous decay of coherence in a dissipative many-body system, Nat. Phys. 16 (2020) 21-25.
  • (4) J.M. Deutsch, Quantum statistical mechanics in a closed system, Phys. Rev. A 43 (1991) 2046.
  • (5) M. Srednicki, Chaos and quantum thermalization, Phys. Rev. E 50 (1994) 888.
  • (6) P. Calabrese and J. Cardy, Entanglement entropy and quantum field theory, J. Stat. Mech. 06 (2004) P06002 [hep-th/0405152].
  • (7) J.R. Garrison and T. Grover, Does a single eigenstate encode the full Hamiltonian?, Phys. Rev. X 8 (2018) 021026 [arXiv:1503.00729].
  • (8) P. Hosur, X.L. Qi, D.A. Roberts and B. Yoshida, Chaos in quantum channels, JHEP 02 (2016) 004 [arXiv:1511.04021].
  • (9) R. Fan et al., Out-of-time-order correlation for many-body localization, Sci. Bull. 62 (2017) 707 [arXiv:1608.01914].
  • (10) M. Rangamani, and T. Takayanagi, Holographic Entanglement Entropy, Springer (2017).
  • (11) J. Maldacena, The large-N limit of superconformal Field theories and supergravity, Int. J. Theor. Phys. 38 (1999) 1113.
  • (12) S. Ryu and T. Takayanagi, Holographic derivation of entanglement entropy from AdS/CFT, Phys. Rev. Lett. 96 (2006) 181602 [hep-th/0603001].
  • (13) V.E. Hubeny, M. Rangamani and T. Takayanagi, A Covariant holographic entanglement entropy proposal, JHEP 07 (2007) 062 [arXiv:0705.0016].
  • (14) N. Engelhardt and A. C. Wall, Quantum Extremal Surfaces: Holographic Entanglement Entropy beyond the Classical Regime, JHEP 01 (2015) 073 [arXiv:1408.3203].
  • (15) R. Islam et al., Measuring entanglement entropy in a quantum many-body system, Nature 528 (2015) 77.
  • (16) A. M. Kaufman, M. E. Tai, A. Lukin, M. Rispoli, R. Schittko, P. M. Preiss, and M. Greiner, Quantum thermalization through entanglement in an isolated many-body system, Science 353 (2016) 6301.
  • (17) J. Li et al., Measuring out-of-time-order correlators on a nuclear magnetic resonance quantum simulator, Phys. Rev. X 7 (2017) 031011 [arXiv:1609.01246].
  • (18) A. Lukin, M. Rispoli, R. Schittko, M. E. Tai, A. M. Kaufman, S. Choi, V. Khemani, J. Léonard, and M. Greiner, Probing entanglement in a many-body-localized system, Science 364 (2019) 254-260.
  • (19) T. Brydges, A. Elben, P. Jurcevic, B. Vermersch, C. Maier, B. P. Lanyon, P. Zoller, R. Blatt, C. F. Roos, Probing Rényi entanglement entropy via randomized measurements, Science 364 (2019) 260-263.
  • (20) A. Kitaev, Hidden correlations in the Hawking radiation and thermal noise, talk given at the 2015 Breakthrough Prize Fundamental Physics Symposium, November 10, San Francisco U.S.A. (2015).
  • (21) S. Sachdev and J. Ye, Gapless spin fluid ground state in a random, quantum Heisenberg magnet, Phys. Rev. Lett. 70 (1993) 3339 [cond-mat/9212030].
  • (22) J. Maldacena and D. Stanford, Remarks on the Sachdev-Ye-Kitaev model, Phys. Rev. D 94 (2016) 106002 [arXiv:1604.07818].
  • (23) A. Kitaev and S.J. Suh, The soft mode in the Sachdev-Ye-Kitaev model and its gravity dual, JHEP 05 (2018) 183 [arXiv:1711.08467].
  • (24) J. Maldacena, D. Stanford and Z. Yang, Conformal symmetry and its breaking in two-dimensional nearly Anti-de Sitter space, Prog. Theor. Phys. (2016) 12C104 [arXiv:1606.01857].
  • (25) D. J. Gross, V. Rosenhaus, The Bulk Dual of SYK: Cubic Couplings, JHEP 05 (2017) 092 [arXiv:1702.08016]
  • (26) R. Jackiw, Lower dimensional gravity, Nucl. Phy. B 252 (1985) 343-356.
  • (27) C. Teitelboim, Gravitation and hamiltonian structure in two spacetime dimensions, Phys. Lett. 126B (1983) 41-45.
  • (28) Y. Gu, A. Lucas and X.-L. Qi, Spread of entanglement in a Sachdev-Ye-Kitaev chain, JHEP 09 (2017) 120.
  • (29) Y. Huang and Y. Gu, Eigenstate entanglement in the Sachdev-Ye-Kitaev model, Phys. Rev. D 100 (2017) 041901 [arXiv:1709.09160].
  • (30) C. Liu, X. Chen and L. Balents, Quantum entanglement of the Sachdev-Ye-Kitaev models, Phys. Rev. B 97 (2018) 245126 [arXiv:1709.06259].
  • (31) P. Zhang, Evaporation dynamics of the Sachdev-Ye-Kitaev model, Phys. Rev. B 100 (2019) 245104 [arXiv:1909.10637].
  • (32) Y. Chen, X.-L. Qi and P. Zhang, Replica wormhole and information retrieval in the SYK model coupled to Majorana chains, JHEP 06 (2020) 121 [arXiv:2003.13147].
  • (33) P. Zhang, C. Liu and X. Chen, Subsystem Rényi entropy of thermal ensembles for SYK-like models, SciPost Phys. 8 (2020) 094 [arXiv:2003.09766].
  • (34) X.-L. Qi, and P. Zhang, The coupled SYK model at finite temperature, JHEP 05 (2020) 129. [arXiv: 2003.03916].
  • (35) A. Haldar, S. Bera and S. Banerjee, Renyi entanglement entropy of Fermi liquids and non-Fermi liquids: Sachdev-Ye-Kitaev model and dynamical mean field theories, Phys. Rev. Research 2 (2020) 033505 [arXiv:2004.04751].
  • (36) P. Zhang, Entanglement entropy and its quench dynamics for pure states of the Sachdev-Ye-Kitaev model, JHEP 06 (2020) 143. [arXiv: 2004.05339]
  • (37) J. Maldacena, X.-L. Qi, Eternal traversable wormhole, arXiv: 1804.00491 (2018).
  • (38) G. Penington, Entanglement wedge reconstruction and the information paradox, arXiv:1905.08255 (2019).
  • (39) A. Almheiri, N. Engelhardt, D. Marolf and H. Maxfield, The entropy of bulk quantum fields and the entanglement wedge of an evaporating black hole, JHEP 12 (2019) 063 [arXiv:1905.08762].
  • (40) A. Almheiri, R. Mahajan, J. Maldacena and Y. Zhao, The Page curve of Hawking radiation from semiclassical geometry, JHEP 03 (2020) 149 [arXiv:1908.10996].
  • (41) Sagar F. Lokhande, Gerben W.J. Oling and Juan F. Pedraza, Linear response of entanglement entropy from holography, JHEP 10 (2017) 104 [arXiv: 1705.10324].
  • (42) Y. Chen, and P. Zhang, Entanglement Entropy of Two Coupled SYK Models and Eternal Traversable Wormhole, JHEP 07 (2019) 033 [arXiv: 1903.10532].
  • (43) D.N. Page, Information in black hole radiation, Phys. Rev. Lett. 71 (1993) 3743 [hep-th/9306083].
  • (44) D.N. Page, Time Dependence of Hawking Radiation Entropy, JCAP 09 (2013) 028 [arXiv:1301.4995].
  • (45) P. Dadras, A. Kitaev, Perturbative calculations of entanglement entropy, arXiv: 2011.09622 (2020).