Gravity from entropy

Abstract

Gravity is derived from an entropic action coupling matter fields with geometry. The fundamental idea is to relate the metric of Lorentzian spacetime to a density matrix. The matter fields curve spacetime, defining a metric induced by the matter fields. The entropic action is the quantum relative entropy between the metric of spacetime and the metric induced by the matter fields. The modified Einstein equations obtained reduce to the Einstein equations in the regime of low coupling. To simplify the theory, exclusively scalar matter fields are considered. This theoretical framework can be extended in different directions, for instance by including fermionic matter fields and gauge fields. A canonical quantization of this field theory could bring new insights into quantum gravity.

The relation between general relativity, statistical mechanics and information theory is a central research topic in theoretical physics. The interest in the subject has its roots in the discovery that black holes have an entropy [1, 2] and emit Hawking radiation [3]. Recently, important results have been obtained relating information theory, entanglement entropy [4] and gravity [5, 6, 7, 8, 9] involving the holographic principle [10, 11] and entanglement properties of quantum field theory [12]. These results define a very active research direction [13] indicating that the quest for an ultimate gravitational theory based on information and statistical mechanics is ongoing. A comprehensive statistical mechanics approach to gravity is expected to give rise to modified Einstein equations that one side can be testable experimentally [14, 15] while on the other side they can bring important conceptual insights into quantum gravity [16, 17, 18, 19, 20, 21].

In this Letter, a continuum modified theory of gravity based on a statistical mechanics action is considered. This theory treats the metric at each point of spacetime as a density matrix. Specifically, this work proposes an action based on the quantum entropy between the metric and the metric induced by the matter fields. A crucial aspect of the presented theory is the adoption of a topological (Dirac-Kähler like [22, 23]) description of bosonic matter fields. Note that the extension of Dirac-Kähler and staggered fermions formalism to bosonic particles is gaining increasing interest in lattice gauge theory [24, 25]. and network theory as well [26]. These bosonic matter fields are described as the direct sum of a zero-form and a one-form defined on the Riemannian manifold describing spacetime. Moreover the metric induced by the topological matter fields is expressed in terms of the Hodge-Dirac operator [27]. This work greatly expands on previous results [26] obtained in the discrete setting by the same author. On one side, here a continuum and fully Lorentz invariant theory is proposed. This progress is based on the development of the suitable mathematical framework to define the Lorentz invariant entropy and cross-entropy between the metric of the spacetime and the metric induced by the matter fields. On the other side, here the relation of this statistical mechanics/information theory action with the Einstein-Hilbert action [28, 29] is established defining a clear connection to gravity. Indeed by considering a continuum limit, in this Letter the intrinsic difficulty related to the definition of the curvature of networks, simplicial and cell complexes is avoided.

To keep the discussion concise the focus is here exclusively on scalar matter fields and their topological generalizations. Further extensions of the proposed framework to Dirac and gauge fields [26, 30, 31] in the discrete or continuum setting are left for future investigations.
Quantum entropy of a tensor of rank 2- Spacetime is described by a torsion free, $1+d$ Riemannian manifold associated with a Lorentzian metric $g_{\mu\nu}$ of signature $\{-1,1,,1\ldots,1\}$ and a metric compatible Levi-Civita connection $\Gamma_{\nu\mu}^{\sigma}$ determining the covariant derivative $\bm{\nabla}_{\mu}$. In order to formulate our statistical mechanics and information theory action for gravity, we need first to define the entropy and cross-entropy associated to rank $2$ tensors. To this end we first define the eigenvalues $\lambda$ and eigenvectors $V^{(\lambda)}_{\nu}$ of the tensor $\hat{\bf T}$ of elements $\hat{T}_{\mu\nu}$ in a Lorentz invariant way. These satisfy the eigenvalue problem

$$\hat{T}_{\mu\nu}[V^{(\lambda)}]^{\nu}=\lambda V^{(\lambda)}_{\mu}.$$ (1)

We say that a symmetric rank 2 tensor is positively defined if all its eigenvalues are positive. We notice that this definition of the eigenvalue of a rank 2 tensor reduces to the definition of the eigenvalue of the matrix $\hat{\bf T}{g}^{-1}$ as Eq.(1) can be rewritten as

$$\hat{T}_{\mu\sigma}g^{\sigma\nu}[V^{(\lambda)}]_{\nu}=\lambda V^{(\lambda)}_{\mu}.$$ (2)

One striking consequence of this definition is that the eigenvalues of the metric $g_{\mu\nu}$ are all identically equal to one. Assuming that the symmetric tensor $\hat{T}_{\mu\nu}$ is positively defined, we define the logarithm of this tensor as

$$[\ln(\hat{\bf T})]_{\mu\nu}=V_{\mu}^{(\lambda)}V^{(\lambda)}_{\nu}\ln(\lambda).$$ (3)

and the inverse of a tensor as

$$[\hat{\bf T}^{-1}]^{\mu\nu}=[V^{(\lambda)}]^{\mu}[V^{(\lambda)}]^{\nu}\lambda^{-1}.$$ (4)

It follows that the logarithm of the inverse of a positively defined tensor is given by

$$[\ln(\hat{\bf T}^{-1}]^{\mu\nu}=-[V^{(\lambda)}]^{\mu}[V^{(\lambda)}]^{\nu}\ln(\lambda).$$ (5)

Warm-up scenario: Entropic action- Before we develop our theory, let us consider an instructive warm-up scenario that will help justify the theory that we will present in the following. Having defined the logarithm of positively defined rank 2 tensors, we are now in the position to define their Lorentz invariant quantum entropy $H$. This is inspired by the expression of the Von Neumann entropy albeit we do not require the tensor to have trace one at every point in spacetime. Specifically $H$ is defined as

$$H=\mbox{Tr}\hat{\bf T}\ln\hat{\bf T}^{-1}=\hat{T}_{\mu\nu}[\ln\hat{\bf T^{-1}}]^{\nu\mu}=-\sum\lambda\ln\lambda.$$ (6)

Recalling that the metric $g$ has all the eigenvalues equal to one it follows that the entropy of the metric is null, i.e.

$$H=\mbox{Tr}{g}\ln{g}^{-1}=0.$$ (7)

We consider a Lagrangian $\bm{\mathcal{L}}$ proportional to the quantum relative entropy between the metric $g$ and the metric ${\bf G}$ induced by the matter field (that we will specify later) defined as

$$\bm{\mathcal{L}} = -\mbox{Tr}{g}\ln{g}^{-1}+\mbox{Tr}{g}\ln{\bf G}^{-1}.$$ (8)

We thus obtain:

$${\mathcal{L}}=\mbox{Tr}g\ln{{\bf G}}^{-1}.$$ (9)

We observe that since $g$ has all the eigenvalues equal to the identity, and Eq.(2) holds of $\hat{\bf T}={\bf G}$, the Lagrangian $\bm{\mathcal{L}}$ can be also expressed as

$$\bm{\mathcal{L}}=-\mbox{Tr}_{M}\ln\ {\bf G}{g}^{-1},$$ (10)

where with $\mbox{Tr}_{M}\ln\ {\bf G}{g}^{-1}$ we indicate the usual trace of the logarithm of a matrix here expressed as the trace of the Taylor power series of the matrix $\ln\ {\bf G}{g}^{-1}$. We consider an action $\mathcal{S}$ associated to the Lagrangian $\mathcal{L}$, as given by

$$\mathcal{S}=\int d{\bf r}\sqrt{|-g|}\mathcal{L},$$ (11)

where $|-g|$ indicates the absolute value of the determinant of $g$.
Warm-up scenario: scalar matter fields- We now apply this warm-up scenario in presence of scalar matter fields. To this end, we consider the metric induced by the complex valued scalar matter field $\phi$. The metric ${\bf G}$ induced by this matter field is given by [32]

$${\bf G}=g+\alpha{\mathcal{G}}$$ (12)

where $\alpha$ is a real positive parameter, and the matrix ${\mathcal{G}}$ is given by

$${\mathcal{G}}_{\mu\nu}={\bm{\nabla}}_{\mu}\bar{\phi}{\bm{\nabla}}_{\nu}\phi,$$ (13)

where here and in the following $\bar{}$ indicates complex conjugation. Since ${\bf G}$ should be adimensional it is convenient to work in the units $\hbar=c=1$ and to put $\alpha=\alpha^{\prime}\ell_{P}^{d}$ where $\ell_{P}$ is the Planck length and $\alpha^{\prime}$ is adimensional. Let us define the scalar product $|{\bm{\nabla}}\phi|^{2}$ as

$$|{\bm{\nabla}}\phi|^{2}={\bm{\nabla}}_{\mu}\bar{\phi}{g}^{\mu\nu}\hat{\bm{\nabla}}_{\nu}\phi.$$ (14)

We observe that the inverse ${\bf G}^{-1}$ of the induced metric ${\bf G}$ has metric given by

$$[{{\bf G}^{-1}}]^{\mu\nu}=g^{\mu\nu}-\alpha\frac{{\mathcal{G}}^{\mu\nu}}{1+\alpha|{\bm{\nabla}}{\phi}|^{2}}.$$ (15)

Adopting this notation we observe that the logarithm of the induced metric $\ln{\bf G}$ and the logarithm of $\ln{\bf G}^{-1}$ are given by

$$\ln{\bf G} = f(|{\bm{\nabla}}\phi|^{2}){\mathcal{G}}$$

$$\ln{\bf G}^{-1} = -f(|{\bm{\nabla}}\phi|^{2}){\mathcal{G}}^{-1}$$ (16)

where

$$f(w)=\frac{\ln(1+\alpha w)}{w}.$$ (17)

where $f(w)\to\alpha$ for $w\ll 1$. With this choice of the metric induced by the matter field we obtain that that the Lagrangian $(\ref{L})$ reads

$${\mathcal{L}} = -\ln(1+\alpha{|{\bm{\nabla}}{\phi}|^{2}}).$$ (18)

We observe that in the limit $\nabla\phi\to 0$ the we have $\bm{\mathcal{L}}\to-|{\bm{\nabla}}{\phi}|^{2}$, i.e. we recover the Lagrangian corresponding to the massless Klein-Gordon equation.

By minimizing the action $\mathcal{S}$ with respect to the field $\phi$ and to the metric $g$ we obtain the Euler-Lagrange equations of motion:

$$\bm{\nabla}_{\mu}h(|\nabla\phi|^{2})g^{\mu\nu}\bm{\nabla}_{\nu}\phi = 0,$$ (19)

where $h(w)$ is given by

$$h(w)=\frac{\alpha}{1+\alpha w}.$$ (20)

In the limit $w\to 0$ we have $h(w)\to\alpha$ and the equation for the scalar field reduces to the simple massless Klein-Gordon equation.

By putting equal to zero the variation of the action $\delta\mathcal{S}=0$ with respect to the metric $g$, we get

$$\delta\mathcal{S}=-h(|\nabla\phi|^{2})\mathcal{G}^{\mu\nu}-\frac{1}{2}\mathcal{L}g^{\mu\nu}=0.$$ (21)

In empty spacetime ${\mathcal{G}}=0$ and $\mathcal{L}=0$, thus this equation is automatically satisfied independently on the value of $g$. Thus in this limit, the metric $g$ is not determined by the action.
Let us now make some remarks about this warm-up derivation. We proposed a statistical mechanics framework that is very inspiring as we get the massless Klein-Gordon equation as the outcome of the minimization of a quantum entropy action for low coupling, i.e. $0<\alpha^{\prime}\ll 1$. However this approach has two important limitations. The first limitation is that the Klein-Gordon equation does not contain the mass term. The second limitation is that in absence of matter fields the metric is not determined. In the following we provide a more comprehensive framework to address these two limitations. This framework can be related to gravity as it gives rise to modified Einstein equations that reduce to the Einstein equations for low coupling.

Topological matter fields and their associated metrics-

In order to derive gravity from our entropic action we need to consider the topological bosonic matter field. The topological bosonic matter field is a type of Dirac-Kähler [22, 23] boson given by the direct sum of a zero-form and a one-form. Topological bosonic fields are receiving increasing attention in discrete theories developed in network [26] and lattice gauge theories [24, 25]. Taking into consideration topological bosonic fields will allow us to introduce in the metric induced by the matter fields, terms depending on the mass of the bosonic field. Thus, in this way, we address the first limitation of the warm-up scenario that we have presented above.

In order to address the second limitation of the warm-up scenario discussed previously, we include in the expression of the metric induced by the matter field, also terms depending directly on the curvature of the manifold. These terms will be expressed in terms of the Ricci scalar $R$ and the Ricci (covariant) tensor ${\bf R}$ of elements $R_{\mu\nu}$. Moreover we also introduce a cosmological constant $\Lambda$.

We define the topological fields $\ket{\Phi}$ as the direct sum between a zero (complex valued) form $\phi$ and a (complex valued) one form $\omega_{\mu}dx^{\mu}$, i.e.

$$\ket{\Phi}=\phi\oplus\omega_{\mu}dx^{\mu},$$ (22)

and its conjugate topological field $\bra{\Phi}$ as

$$\bra{\Phi}=\bar{\phi}\oplus\bar{\omega}_{\mu}dx^{\mu}.$$ (23)

The considered metric matrix $g$ is defined as the direct sum of the metric matrix for zero-forms (the identity) and the metric matrix $g$ for $1$-forms introduced previously,

$$\tilde{g}=1\oplus g_{\mu\nu}dx^{\mu}\otimes dx^{\nu}.$$ (24)

The scalar product among topological fields is defined as

$$\left\langle{\Phi|\Phi}\right\rangle=|\phi|^{2}+\bar{\omega}^{\mu}\omega_{\mu},$$ (25)

where $\omega^{\mu}\omega_{\mu}=\omega_{\nu}g^{\nu\mu}\omega_{\mu}$, while the outer product is given by

$$\ket{\Phi}\bra{\Phi}=\bar{\phi}\phi\oplus\Big{(}\bar{\omega}_{\mu}{\omega}_{\nu}dx^{\mu}\otimes dx^{\nu}\Big{)}.$$ (26)

Indicating here with ${d}\phi$ the differential operator applied to $\phi$ and with ${\delta}\omega$ the codifferential operator applied to $\omega$, we define the Dirac operator $D$ as

$${D}\ket{\Phi}={\delta}\omega\oplus{d}\phi=\nabla^{\mu}\omega_{\mu}\oplus\nabla_{\mu}\phi dx^{\mu}$$ (27)

The metric $\tilde{\bf G}$ induced by the topological matter field will have structure similar to $\tilde{g}$ that we can write in full generality as

$$\tilde{\bf G} = G_{(0)}+[G_{(1)}]_{\mu\nu}dx^{\mu}\otimes dx^{\nu}.$$ (28)

Proceeding as in the previous case we might wish to define the metric induced by the topological field as

$$\tilde{\bf G} = \tilde{g}+\alpha\Big{(}{D}\ket{\Phi}\bra{\Phi}D\Big{)},$$ (29)

where

$${D}\ket{\Phi}\bra{\Phi}D=\nabla^{\mu}\bar{\omega}_{\mu}\nabla^{\nu}\omega_{\nu}\oplus\nabla_{\mu}\bar{\phi}\nabla_{\nu}\phi\ dx^{\mu}\otimes dx^{\nu}.$$ (30)

However, since we interpret the metric as a quantum density matrix, it make sense to add further terms in the metric induced by the topological matter field $\tilde{\bf G}$ depending on the topological field and the geometry of the space. First, we introduce in $\tilde{\bf G}$ a term proportional to $(m^{2}+R)\ket{\Phi}\bra{\Phi}$ where $R$ is the Ricci scalar. Secondly, we introduce a term depending explicitly on the curvature of the manifold. Since ${\bf G}_{(0)}$ is a metric for $0$-forms it is natural to consider a further term involving the Ricci scalar $R$. Similarly, since ${\bf G}_{(1)}$ is a metric for $1$-forms it is natural to consider a further term involving the Ricci tensor ${\bf R}$ of elements $R_{\mu\nu}$. Therefore we define the direct sum of the Ricci scalar and the Ricci tensor $\bm{\mathcal{R}}$, given by

$$\bm{\mathcal{R}}=R\oplus(R_{\mu\nu}dx^{\mu}\otimes dx^{\nu}\Big{)}.$$ (31)

where $R=g^{\mu\nu}R_{\mu\nu}.$

We are now able to postulate that the metric induced by the geometry and the matter fields is given by

$$\tilde{\bf G} = \tilde{g}+\alpha\Big{(}{D}\ket{\Phi}\bra{\Phi}D+(m^{2}+R)\ket{\Phi}\bra{\Phi}\Big{)}$$ (32)

$$-\beta\bm{\mathcal{R}}.$$

where $\alpha,\beta$ are positive constants. In particular, since we require $\tilde{\bf G}$ to be adimensional, we need to consider $\alpha=\alpha^{\prime}\ell_{P}^{d}$ and $\beta=\beta^{\prime}\ell_{P}^{2}$ where $\ell_{P}$ is the Planck length and $\alpha^{\prime},\beta^{\prime}$ are adimensional in the units $\hbar=c=1$. According to this choice of the induced metric we obtain that $\tilde{\bf G}$ has the structure given by Eq.(29) with ${G}_{(0)}$ and ${\bf G}_{(1)}$ given by

$$G_{(0)} = 1+\alpha\Big{(}\nabla^{\mu}\bar{\omega}_{\mu}\nabla^{\nu}\omega_{\nu}+(m^{2}+R)\bar{\phi}\phi\Big{)}$$

$$-\beta R,$$

$${[G_{(1)}]}_{\mu\nu} = g_{\mu\nu}+\alpha\Big{(}\nabla_{\mu}\bar{\phi}\nabla_{\nu}\phi+(m^{2}+R)\bar{\omega}_{\mu}\omega_{\nu}\Big{)}$$ (33)

$$-\beta R_{\mu\nu}.$$

Possibly this approach could be extended to matter fields formed by the direct sum between a zero-form one -form and a two-form, thus involving an induced metric depending explicitly not only on the Ricci scalar and the Ricci tensor but also on the Riemann tensor. However, for simplicity, we consider here only topological matter fields formed by the direct sum between a zero-form and a one-form.

Figure 1: The considered Lagrangian is given by the negative of the quantum relative entropy between the topological metric $\tilde{g}$ of spacetime and the topological metric induced by the topological matter fields $\tilde{\bf G}$ minus the cosmological constant $\Lambda$. Since, by definition, the entropy of the metric vanishes identically, the quantum relative entropy in the Lagrangian $\bm{\mathcal{L}}$ reduces to two terms: the quantum cross-entropy and the cosmological constant term.

Entropic topological action- We propose a statistical mechanics action formulated in terms of the quantum relative entropy between the metric $\tilde{g}$ and the metric $\tilde{\bf G}$ induced by the topological matter fields. First of all we observe that also in this case the entropy $\tilde{H}$ of the metric $\tilde{g}$ vanishes. Indeed we define $\tilde{H}$ as

$$\tilde{H}=\mbox{Tr}\tilde{g}\ln{\tilde{g}^{-1}}=1\ln 1+\mbox{Tr}g\ln g^{-1}=H.$$ (34)

and thus, we observe that since Eq.(7) holds, then $\tilde{H}=0$. The Lagrangian that we consider is given by the quantum relative entropy between $\tilde{g}$ and $\tilde{\bf G}$, minus the cosmological constant $\Lambda$

$$\bm{\mathcal{L}} := -{\mbox{Tr}}\tilde{g}\ln{{g}^{-1}}+{\mbox{Tr}}\tilde{g}\ln{\tilde{\bf G}^{-1}}-\Lambda$$ (35)

$$= {\mbox{Tr}}\tilde{g}\ln{\tilde{\bf G}^{-1}}-\Lambda.$$

Since $\tilde{H}=0$ we thus obtain (see Figure 1 for an illustration)

$$\bm{\mathcal{L}} = {\mbox{Tr}}\tilde{g}\ln{\tilde{\bf G}^{-1}}-\Lambda.$$ (36)

By treating separately the contributions of ${G^{(0)}}$ and ${\bf G}^{(1)}$ we get the explicit expression for the Lagrangian $\bm{\mathcal{L}}$ given by

$$\bm{\mathcal{L}} := \ln[{G_{(0)}}]^{-1}+\mbox{Tr}g\ln[{\bf G}_{(1)}]^{-1}-\Lambda.$$ (37)

The resulting statistical mechanics action $\mathcal{S}$ associated to the Lagrangian $\mathcal{L}$ is given by

$$\mathcal{S}=\int d{\bf r}\sqrt{|-{g}|}\bm{\mathcal{L}}.$$ (38)

This action defines a modified theory of gravity. In the linearised limit $\alpha^{\prime}\ll 1$, $\beta^{\prime}\ll 1$ this action reduces to the Einstein-Hilbert action coupled with the scalar topological field. Indeed in this limit $\bm{\mathcal{L}}$ reduces to

$$\bm{\mathcal{L}} = 2{\beta}(R-2\Lambda^{\prime})-\alpha\Big{(}|\nabla\phi|^{2}+|\nabla^{\mu}\omega_{\mu}|^{2}\Big{)}$$ (39)

$$-\alpha(m^{2}+R)\Big{(}|\phi|^{2}+\bar{\omega}^{\mu}\omega_{\mu}\Big{)},$$

where $\Lambda^{\prime}$ is the rescaled cosmological constant given by $\Lambda^{\prime}=\Lambda/(4\beta).$ In the interesting limit in which $\omega_{\alpha}=0$ and the bosonic matter field is only scalar, Eq.(39) reduces to the widely studied Einstein-Hilbert action [28, 29] coupled with the scalar field [33],

$$\bm{\mathcal{L}} = 2{\beta}(R-2\Lambda^{\prime})-\alpha|\nabla\phi|^{2}-\alpha(m^{2}+R)|\phi|^{2}.$$ (40)

Equation of motion for the matter field- Following similar steps used to derive Eq. (10) we can express the Lagrangian $\bm{\mathcal{L}}$ given by Eq.(36) as

$$\bm{\mathcal{L}}=\mbox{Tr}_{M}\ln\tilde{g}{\tilde{\bf G}}^{-1}=-\mbox{Tr}_{M}\ln\ {\tilde{\bf G}}\tilde{g}^{-1}.$$ (41)

Therefore the variation of the Lagrangian is given by

$$\delta\bm{\mathcal{L}} = -\mbox{Tr}_{M}\tilde{g}{\tilde{\bf G}}^{-1}[\delta{\tilde{\bf G}}\tilde{g}^{-1}+{\tilde{\bf G}}\delta\tilde{g}^{-1}]$$ (42)

$$= -\mbox{Tr}_{M}{\tilde{\bf G}}^{-1}\delta{\tilde{\bf G}}-\mbox{Tr}_{M}\tilde{g}\delta\tilde{g}^{-1}$$

$$= -\mbox{Tr}_{M}\delta{\tilde{\bf G}}{\tilde{\bf G}}^{-1}+\mbox{Tr}_{M}\delta\tilde{g}\tilde{g}^{-1}.$$

In this scenario, by putting to zero the variation of the Lagrangian with respect to $\bra{\Phi}$ we get the following equation of motion

$${D}{\tilde{\bf G}}^{-1}{D}\ket{\Phi}+{\tilde{\bf G}}^{-1}(m^{2}+R)\ket{\Phi}=0.$$ (43)

In the interesting case in which $\omega_{\mu}=0$, i.e. the bosonic matter field reduced to the scalar $\phi$ these equations of motions reduce to the Klein-Gordon equations with a non-zero mass term, i.e.

$$\bm{\nabla}_{\mu}g^{\mu\nu}\bm{\nabla}_{\nu}\phi+(m^{2}+R)\phi=0.$$ (44)

Thus this framework effectively overcomes the first limitation of the warm-up scenario as we recover the Klein-Gordon equation in curved spacetime in full.
Modified gravity- In order to obtain the modified Einstein equations we consider the variation of the action $\mathcal{S}$ with respect to $g$. Putting $\delta\mathcal{S}=0$ we note that, since the Lagragian is non-linear, we also need to consider the contributions due to the variation of the Ricci tensor not contributing to the Einstein equations derived from the Einstein-Hilbert action. Taking into account these terms, and the Eq.(42) we thus get

$$\beta R^{\mu\nu}[{G}_{(0)}]^{-1}-\frac{1}{2}g^{\mu\nu}\bm{\mathcal{L}}+B^{\mu\nu}=\tilde{T}^{\mu\nu}.,$$ (45)

where $B^{\mu\nu}$ is given by

$$B^{\mu\nu} = g^{\mu\nu}-[{\bf G}^{-1}_{(1)}]^{\mu\nu}+\beta\Big{(}g^{\mu\nu}\nabla_{\sigma}\nabla^{\sigma}-\nabla^{\mu}\nabla^{\nu}\Big{)}{G}^{-1}_{(0)}$$

$$+\frac{1}{2}\beta\Big{(}g^{\mu\nu}\nabla^{\sigma}\nabla^{\rho}{[{\bf G}^{-1}_{(1)}]}_{\sigma\rho}+\nabla_{\sigma}\nabla^{\sigma}{[{\bf G}^{-1}_{(1)}]}^{\mu\nu}$$

$$\hskip 28.45274pt-\nabla_{\sigma}\nabla_{\rho}g^{\rho\nu}\{{[{\bf G}^{-1}_{(1)}]}^{\sigma\mu}+{[{\bf G}^{-1}_{(1)}]}^{\mu\sigma}]\}$$

$$\hskip 28.45274pt+\nabla_{\sigma}\nabla_{\rho}\left\{g^{\sigma\nu}{[{\bf G}^{-1}_{(1)}]}^{\mu\rho}-g^{\sigma\mu}{[{\bf G}^{-1}_{(1)}]}^{\nu\rho}\right\}\Big{)}.$$

In Eq.(45) we have indicated with $\tilde{T}^{\mu\nu}$ the tensor given by

$${\tilde{T}}^{\mu\nu} = \alpha\Big{(}R^{\mu\nu}+g^{\mu\nu}\nabla_{\sigma}\nabla^{\sigma}-\nabla^{\mu}\nabla^{\nu}\Big{)}\bra{\Phi}{\tilde{\bf G}}^{-1}\ket{\Phi}$$ (46)

$$+\alpha K^{\mu\nu}[{G_{(0)}}]^{-1},$$

with

$$K^{\mu\nu}=\nabla^{\mu}\bar{\omega}^{\nu}\nabla^{\alpha}\omega_{\alpha}+\nabla^{\alpha}\bar{\omega}_{\alpha}\nabla^{\mu}\omega^{\nu}.$$ (47)

As expected the Eq.(45) reduce to the Einstein equations for $\alpha^{\prime}\ll 1$ and $\beta^{\prime}\ll 1$. This implies, among the other things, that by adding to $\tilde{\bf G}$ the term proportional to $\bm{\mathcal{R}}$ we solve the second limitation of the warm-up scenario, and we obtain that the entropic action allows for the full determination of the metric in the vacuum.
The study of the solutions of the modified Einstein equations (45) is beyond the scope of this paper and will be considered in future works.

In conclusion this work proposes a modified theory of gravity emerging from statistical mechanics and information theory action. The approach is based on the interpretation of the metric as a density matrix albeit its trace is not unitary, pointing out to possible connections of this approach to von Newmann algebras and the theory of entanglement in quantum field theories [12]. The statistical mechanics action is given by the quantum relative entropy between the metric and the metric induced by the matter fields. Here a Lorentzian theory consistent with this fundamental physical interpretation of gravity is formulated. In order to do this we have built the necessary mathematical background to define the entropy and the cross-entropy of rank 2 tensors in a Lorentzian well defined way. A modified gravity is emerging from this framework when the matter fields are described by topological Dirac-Kähler bosons formed by the direct sum between a zero and a one form and the induced metric also depends on the curvature of spacetime. The modified Einstein equations reduce to the Einstein equations in the regime of low coupling.

We hope that this approach can help identify the deep connections between gravity and statistical physics. Given the interpretation of the metric as a density matrix our hope is that this approach will also be instrumental to formulate new approaches to quantum gravity.

This work was partially supported by a grant from the Simons Foundation. The author would like to thank the Isaac Newton Institute for Mathematical Sciences, Cambridge, for support and hospitality during the programme Hypergraphs: Theory and Applications, where work on this paper was undertaken. This work was supported by EPSRC grant EP/R014604/1.