Arxiv:2006.06019V2 [Cond-Mat.Str-El] 16 Oct 2020 Cessing Otocs Using Ordinary Time-Ordered Measure- Models Are Coupled by Interactions [24] Or Tunneling [25]

Traversable wormhole and Hawking-Page transition in coupled complex SYK models

Sharmistha Sahoo,1, ∗ Etienne´ Lantagne-Hurtubise,1, 2, † Stephan Plugge,1 and Marcel Franz1 1Department of Physics and Astronomy & Stewart Blusson Quantum Matter Institute, University of British Columbia, Vancouver BC, Canada V6T 1Z4 2Kavli Institute for Theoretical Physics, University of California Santa Barbara, CA 93106, USA (Dated: October 20, 2020) Recent work has shown that coupling two identical Sachdev-Ye-Kitaev (SYK) models can realize a phase of matter that is holographically dual to an eternal traversable wormhole. This phase supports revival oscillations between two quantum chaotic systems that can be interpreted as information traversing the wormhole. Here we generalize these ideas to a pair of coupled SYK models with complex fermions that respect a global U(1) charge symmetry. Such models show richer behavior than conventional SYK models with Majorana fermions and may be easier to realize experimentally. We consider two different couplings, namely tunneling and charge-conserving two-body interactions, and obtain the corresponding phase diagram using a combination of numerical and analytical tech- niques. At low temperature we find a charge-neutral gapped phase that supports revival oscillations, with a ground state close to the thermofield double, which we argue is dual to a traversable wormhole. We also find two different gapless non-Fermi liquid phases with tunable charge density which we interpret as dual to a ‘large’ and ‘small’ charged black hole. The gapped and gapless phases are separated by a first-order phase transition of the Hawking-Page type. Finally, we discuss an SU(2)-symmetric limit of our model that is closely related to proposed realizations of SYK physics with spinful fermions in graphene, and explain its relevance for future experiments on this system.

I. INTRODUCTION enables the transmission of information between two chaotic systems through ‘revival dynamics’ [26, 27] corre- The Sachdev-Ye-Kitaev (SYK) model [1–4] has sponding, in the gravity interpretation, to sending parti- emerged recently as a powerful toy model allowing to cles through a wormhole [28–34]. Proposals for the phys- glean insight into the exotic behavior of non-Fermi liq- ical realization of such coupled SYK models in condensed uids, such as quantum chaos [5–8], holography [9–11] and matter platforms have also been discussed [35]. strange metallic transport [12–16]. The model consists A key concept at the heart of the traversable wormhole of N Majorana fermions coupled via all-to-all, random proposal is the thermofield double (TFD) state. Given Gaussian interactions. It derives its predictive power two identical copies of a quantum mechanical system, a from the fact that, once averaged over quenched disor- TFD state is defined as der, the model can be solved exactly through a large-N 1 X −βE˜ n/2 saddle-point expansion. Remarkably, at low temperature |TFDβ˜i = q e |ni1 ⊗ |n¯i2. (1) the model can be shown to be maximally chaotic: its Zβ˜ n out-of-time-order correlators (OTOCs) exhibit an expo- ˜ nentially growing regime with a Lyapunov exponent that P −βEn Here Zβ˜ = n e is the partition function of a sin- saturates the bound on many-body quantum chaos [5]. gle system at inverse temperature β˜, and |n¯i = |Θni Such growth indicates fast scrambling of quantum infor- where Θ is an anti-unitary symmetry of the Hamilto- mation in the system and underlies the holographic con- nian. The TFD is a purification of a Gibbs state at in- nection between the SYK model and black holes. Moti- verse temperature β˜: it is an entangled state of the two vated by these exciting predictions, a number of propos- copies, such that tracing over either copy recovers the als have emerged for the physical realization of the SYK thermal density matrix for the other. As such, the TFD model and its variants in atomic [17], optical [18] and can be used as a resource state to study thermal proper- solid-state [19–21] platforms, or using quantum simula- ties in quantum simulators [36]. It also obeys an inverted tors [22, 23]. version of time-translation invariance which enables ac- Interesting physics also occurs when two identical SYK

arXiv:2006.06019v2 [cond-mat.str-el] 16 Oct 2020 cessing OTOCs using ordinary time-ordered measure- models are coupled by interactions [24] or tunneling [25]. ments [35] and teleporting states or operators between At low temperature the coupling can drive phase tran- the two subsystems [37, 38]. Recent work has shown sitions to symmetry-broken states [24] or, remarkably, how to construct model Hamiltonians with a ground state to a phase holographically dual to an eternal traversable close to a TFD [25, 39] by coupling identical systems, the wormhole [25] with an AdS throat (AdS refers to the 2 2 Maldacena-Qi (MQ) wormhole model [25, 40, 41] being 1+1-dimensional anti-de Sitter spacetime). This phase an example of such a construction. Variants of the SYK model built from ordinary complex fermions, rather than real Majorana fermions, have ∗ [email protected] also been studied and exhibit similar properties [2, 42]. † [email protected] The main diﬀerence between the complex and Majorana 2

ending at a critical point. Further, we investigate the dynamics of the system and find that two-point correlations between the two subsystems decay as a power- law in the high-temperature cSYK phase, but show pe- riodic revivals in the gapped phase [26, 27]. These results lead us to conjecture that the gapped phase of this model, similarly to its Majorana counterpart [25], is holographically dual to a traversable wormhole. The first- order phase transition between the wormhole phase with a TFD ground state and the charged black hole (cSYK) phase can thus be interpreted as a Hawking-Page type transition [48]. We then consider two-body interactions with strength α that conserve charge in each system separately – leading to a U(1)⊗U(1) symmetry group. At charge neutrality we find two low-temperature phases depending on the interaction strength α. For α < 0 or α > 4, we observe a -1 0 1 gapped phase with spontaneous symmetry-breaking from U(1)⊗U(1) down to U(1), while for 0 ≤ α ≤ 4 we obtain a gapless non-Fermi liquid phase with tunable charge den- FIG. 1. (a) Illustration of the coupling terms in our model. sity and properties similar to the cSYK phase. Combin- (b) Phase diagram of the model at charge neutrality, for zero ing both types of couplings we find the phase diagram il- and low temperatures. The dashed line indicates a first-order phase transition between a gapless non-Fermi liquid phase and lustrated in Fig.1 (b) which contains a dome-shaped gap- a gapped phase supporting revival oscillations. The α = 0 less cSYK phase, separated from the surrounding gapped line is the generalization of the MQ wormhole model [25] to phase by a zero-temperature first-order phase transition. complex fermions, while the α = 1 line with SU(2) symmetric Surprisingly, away from charge neutrality we uncover an- interactions maps to the graphene flake proposal of Ref. [43]. other first-order phase transition to a different gapless non-Fermi liquid describing a smaller black hole with half of its degrees of freedom gapped out. We finally SYK models is that the former has a conserved U(1) discuss in detail the α = 1 limit of our model, featuring charge. Importantly, these complex SYK variants, hence- SU(2)-invariant interactions, which is directly relevant forth abbreviated as cSYK, might be easier to realize to the proposed graphene flake realization of the cSYK in condensed matter systems [43, 44], where Majorana model [43]. Using this new connection we revisit the re- fermions are notoriously difficult to obtain and control. sults of Ref. [43] and argue that the irregularly-shaped Further, a number of experimental probes such as spec- graphene flake model for weak applied magnetic field ad- troscopy [43, 45], electrical conductance [46] and ther- mits a cSYK non-Fermi liquid phase. mopower [47] have recently been proposed to identify The rest of this paper is organized as follows. In Sec.II, signatures of complex SYK models. we review properties of the TFD and show how to define In this work we investigate the physics of coupled com- it for a pair of identical complex SYK models with U(1) plex SYK models using a combination of analytical ar- symmetry. In Sec.III, we couple the two systems via the guments, exact diagonalization for small N and saddle- tunneling term and discuss the resulting gapped phase, point solutions for large N. We consider two types of which we conjecture is dual to a wormhole in the gravity couplings: a simple tunneling term with strength κ, and description. In Sec.IV, we add interaction terms be- random two-body interactions with strength α that con- tween the two systems and explore the resulting phase serve charge in each system separately. The form of the diagram. In Sec.V we provide a connection of our model couplings, illustrated in Fig.1 (a), is partly motivated by to the graphene flake proposal [43] with SU(2)-symmetric their natural connection to the disordered graphene flake interactions, and discuss its relevance for future experi- proposal of Ref. [43]. ments on this system. More technical contributions, including details of the TFD construction and solutions of We first show that coupling identical cSYK models the saddle-point equations, in both imaginary and real with the tunneling term leads to similar physics as the time, are presented in the Appendices. MQ wormhole model [25]: at low temperature the system is gapped, with a charge-neutral ground state close to a TFD. The correct definition of the TFD state in the presence of a U(1) symmetry is however subtle and we explain II. THERMOFIELD DOUBLE STATE it in detail. We obtain the finite-temperature phase dia- CONSTRUCTION gram of the model, showing that the gapped phase with a TFD ground state is separated from a gapless cSYK Consider two identical copies of a quantum system de- phase at high temperature by a first-order transition line scribed by Hamiltonians H1 = H2, with eigenstates |ni1 3

and |ni2 of a common eigenenergy En. A TFD is an en- and satisfy the symmetry constraints tangled state of the two copies defined by Eq. (1). Such ∗ a definition represents a one-parameter family of states Jij;kl = −Jji;kl = −Jij;lk = Jlk;ji (8) which can be generated from imposed by the fermionic commutation relations. The ˜ −iφ −β(H1+H2)/4 model has a global U(1) symmetry, cj → cje that |TFDβ˜i = e |Ii (2) expresses the conservation of the total charge where |Ii = |TFD0i. (We denote the effective inverse N temperature of the TFD state as β˜ to differentiate it X 1 Q = (c†c − ). (9) from the physical temperature T = 1/β of the system). i i 2 The advantage of the representation in Eq. (2) is that |Ii i=1 is a maximally entangled state between the two copies, A detailed discussion of the cSYK model and its physical and is thus usually simple to write down. properties can be found in Refs. [2, 42, and 49]. The TFD state has a few important properties which Additionally, at charge neutrality (µ = 0) the model follow from its definition, Eq. (1). First, it is a purifica- has an anti-unitary particle-hole symmetry if we con- tion of the thermal density matrix, such that strain the tensor of couplings Jij;kl to be fully antisymmetric. This anti-unitary symmetry is generated by 1 h −βH˜ i hVaiβ˜ = Tr V e (3) Zβ˜ N Y † P = (ca + ca)K, (10) for any operator Va (a = 1, 2) that acts only on one subsystem. Here the expectation value is taken in the state a=1

|TFDβ˜i, and the trace is over the Hilbert space of a sin- with K the complex conjugation operator, and transforms gle subsystem. Although the TFD is not an eigenstate of † cj ↔ cj up to a sign that depends on N. One can check the full system’s Hamiltonian H1 +H2, it is an eigenstate that P −1HP = H and of the difference P 2 = (−1)N(N−1)/2. (11) (H1 − H2) |TFDβ˜i = 0 (4) which implies that two-point correlation functions re- Such a symmetry has been discussed in Refs. [42, 49–51] spect an inverted version of time-translation invariance, and is useful to simplify calculations. Ref. [49] imple- ments it by including additional hopping terms to can- 0 00 0 00 hV1(t)W2(t )iβ˜ = hV1(t + t )W2(t − t )iβ˜ , (5) cel out the unwanted interaction terms with repeated indices. Here we simply consider fully anti-symmetric for any operators V1 and W2. This property has an ana- log in the gravity context, where the TFD is used to Jij;kl, and use the anti-unitary symmetry P to define the describe traversable wormholes and time effectively flows states |n¯i that are required to construct the TFD state. in opposite directions on its two sides [25, 29]. Leveraging the U(1) symmetry, the Hamiltonian Our goal in this section is to construct the TFD state Eq. (6) can be block diagonalized in symmetry sectors labeled by the eigenvalues q of the charge operator Q. when each subsystem is described by a cSYK model. As −1 we shall see this construction entails a subtlety: Whereas Since P QP = −Q, at charge neutrality the model has in the Majorana SYK model the state |n¯i can be chosen a two-fold spectral degeneracy guaranteed by the map- as equal to |ni (up to a phase), this is not the case for ping between states in the q and −q sectors. For even N cSYK where |n¯i carries a different charge quantum num- there is a special zero-charge (q = 0) sector which maps ber. onto itself under P ; this enforces a two-fold Kramers degeneracy when P 2 = −1, for N mod 4 = 2 (see TableI).

A. Complex SYK model B. Thermofield double state for complex SYK The complex SYK model is the charge conserving vari- ant of the Majorana SYK model. It is described by the As discussed above, the eigenstates |nqi of the cSYK Hamiltonian model, in the presence of the anti-unitary symmetry P , form particle-hole pairs in charge sectors ±q. We thus N X † † X † define the TFD as a zero-charge state with H = Jij;klci cjckcl − µ ci ci (6) N/2 i,j,k,l=1 i 1 X X −βE˜ n/2 † |TFDβ˜i = q e |nqi1 ⊗|n¯−qi2 (12) where the ci, c are N fermionic operators satisfying i Z ˜ q=−N/2 nq † β {ci, cj} = δij. The coefficients Jij;kl are complex Gaus- sian random numbers with where the state |n¯−qi is equivalent to P |nqi up to a phase J 2 which depends only on the symmetry label q. (An arbi- J = 0 , |J |2 = (7) ij;kl ij;kl 8N 3 trary phase θn is not allowed as it cannot be absorbed 4 by a gauge transformation in |ni and |ni.) This phase N mod 4 0 1 and 3 2 can be fixed most easily by choosing a specific infinite- single cSYK 1 2 2 ˜ temperature (β = 0) TFD state |Ii and identifying it two decoupled cSYKs 1 4 (1,2,1) 4 with |TFD0i. The definition in Eq. (2) then automat- tunneling (κ > 0) 1 1 1 ically fixes the same phases for the finite-temperature interaction (κ = 0, 0 < α < α ) 1 2 (1,0,1) 1 TFD states. The state |Ii must be maximally entangled, c as it is the purification of the infinite-temperature den- interaction (κ = 0, α < 0 or α > αc) 1 2 (0,2,0) 1 sity matrix. This motivates the choice of a product of N TABLE I. Ground state degeneracy of two identical cSYK Bell pairs between the two systems [52], models at charge neutrality µ = 0, for various system sizes and model parameters κ, α. Ground states are in the q = 0 N Y 1 symmetry sector unless indicated in brackets, which show the |I i = √ |1i |0i − e−iφ|0i |1i . (13) φ 1 2 1 2 i number of ground states in the q = (−1, 0, 1) sectors. The i=1 2 critical value αc ∼ 4. In AppendixA we explicitly show that the state in Eq. (13) is equivalent to state of this Hamiltonian is the Bell state |Iφi = |TFD0i, 1 X X whereas for κ/J = 0 it is simply the product state |TFD0i ≡ |nqi ⊗ Θ |nqi (14) 2N/2 1 2 |TFD∞i. As we shall see below using numerical exact q nq diagonalization, the model admits a ground state close to ˜ with the anti-unitary symmetry defined as |TFDβ˜i for all κ/J, with a parameter β that is a monotonically decreasing function of κ/J. Note that a gauge − iηπΓ −iq(φ− π ) Θ = e 4 e 2 P. (15) transformation on the fermion operators in either subsystem can absorb the phase φ. Hence in the following we q+ N Here q and Γ = (−1) 2 are the charge and fermion consider, without loss of generality, a purely imaginary parity, respectively, of the state |nqi and η = ±1 is a tunneling term with φ = π/2, corresponding to the sim- sign which depends on the total number N of fermions plest case where the charge-dependent phases in Eq. (15) −1 † through P ci1P = ηci1. The parity-dependent phase disappear. factor has been discussed for the Majorana version of the The symmetries and the ground state degeneracy of TFD state [40]. The charge-dependent phase is required the model, deduced by simple arguments and verified to cancel the minus signs appearing when fermionic cre- using exact diagonalization, are summarized in TableI. ation or annihilation operators are taken across eigen- For κ = 0 we have two decoupled cSYK models, where states of subsystem 1 to act on subsystem 2. Note that the charge is separately conserved in systems 1 and 2 the TFD for a bosonic Hamiltonian would not have these (U(1)⊗U(1) symmetry). There are two anti-unitary sym- phases. metries P1 = P ⊗ 1 and P2 = 1 ⊗ P , where P acts in one subsystem. The ground state degeneracy is the product of that of each SYK model, which is unique for III. COUPLING cSYK MODELS WITH N mod 4 = 0 and doubly degenerate otherwise. (The TUNNELING TERM ground states of a single SYK model are always found in the q = 0 sector (for even N) or the q = ±1/2 sec- In this section we introduce a coupling between the tors (for odd N)). For even N, the ground states of the two cSYK Hamiltonians which results in a ground state coupled system are in the q = 0 sector, while for odd N close to |TFDβ˜i. Consider the infinite-temperature TFD they are distributed in the q = 0, ±1 sectors as shown state |Iφi defined in Eq. (13). This is the ground state of in TableI. The tunneling term κ breaks the charge con- a simple tunneling term servation in each system down to the total U(1) charge conservation, and also breaks the anti-unitary symme- X † † iφ −iφ tries P1 and P2 – thus the ground state for κ > 0 is K = κ e ci1ci2 + e ci2ci1 (16) i unique. Finally, there is a discrete mirror symmetry that exchanges fermion operators between systems 1 and 2. with real κ and φ. Further, it is clear that the zero- For φ = π/2 this symmetry transforms ci1 → ci2 and temperature TFD state |TFD i = |0i ⊗|0i is the exact ∞ 1 2 ci2 → −ci1, and constrains the two-point correlators of ground state of two decoupled identical cSYK models. the system as discussed in Sec.IIIB. We thus consider the following model, X X X H = J c† c† c c − µ c† c + K, κ ij;kl ia ja ka la ia ia A. Exact diagonalization: TFD ground state ij;kl a=1,2 i,a (17) We first perform an exact diagonalization study of the where the coupling constants Jij;kl are identical in sys- model, Eq. (17), to confirm that it admits a ground state tems 1 and 2. For large coupling κ/J 1 the ground close to a TFD. To do this we construct the family of 5

1.00 3 are Lagrange multipliers which can be interpreted as 10 2N = 12 L 0.99 2N = 14 fermion self-energies. The matrix M = n M(iωn) with 2N = 16 Mab(iωn) = (−iωn −µ)δab +iκab −Σab(iωn) results from 0

0.98 J 2N = 18 x performing the integral over complex Grassman ﬁelds. | 2 a 10 D m Varying the eﬀective action Eq. (19) with respect to F 0.97 T (a) (b) Gab and Σab leads to the large-N saddle-point equa- 0 0.96 tions. Using the time-translation symmetry Gab(τ, τ ) = 101 G (τ − τ 0) and the mirror symmetry which enforces 0.95 ab 2 1 0 2 1 0 10 10 10 10 10 10 G11(τ) = G22(τ) and G12(τ) = −G21(τ), the saddle- /J /J point equations can be reduced to the form

FIG. 2. Exact diagonalization results for 2N up to 18 at −iωn − µ − Σ11(iωn) G11(iωn) = , charge neutrality µ = 0. The shaded area corresponds to D(iωn) the standard deviation obtained from 12 independent disorder −iκ + Σ12(iωn) realizations. (a) Overlap between the ground state |Ψ0i of the G12(iωn) = , coupled cSYK models, Eq. (17) and the best-fit TFD state as D(iωn) ˜ 2 2 a function of κ/J. (b) Effective inverse temperature βmax of Σ11(τ) = −J G11(τ)G11(−τ) , the best-fit TFD state. 2 2 Σ12(τ) = J G12(τ)G12(−τ) , (20) ˜ π TFD states with parameter β from the eigenstates of where ωn = (2n + 1) β are fermionic Matsubara frequen- a single SYK model using the definition, Eqs. (12)-(15). cies,

We then compute the overlap of this family of TFD states 2 2 with the numerical ground state of the coupled model, D(iωn) = (−iωn − µ − Σ11) + (iκ − Σ12) , (21) Eq. (17) and select the TFD with the parameter β˜ that and the parameters J, µ and κ are real. When κ = 0 maximizes the overlap. As shown in Fig.2 this best-fit and thus G = 0, these equations reduce to the usual overlap is always close to 1, with a minimum of ∼ 0.96 12 cSYK model, which can be solved in the low-frequency at a value κ/J ∼ 0.1 which roughly corresponds to the or long-time limit by appealing to an emergent conformal end of the finite temperature first-order transition seen in invariance. The zero-temperature result is [2] the large-N calculation (see Sec.IIIB). The parameter ˜ ( β characterizing the best-fit TFD is monotonically de- bτ −1/2 τ J −1 creasing with κ. In the gravity interpretation of the MQ G11(τ) = −2πE −1/2 −1 (22) model the parameter β˜ is proportional to the length of −be |τ| −τ J the wormhole, or equivalently to the period of the revival 2 1/4 oscillations between the two sides [25, 26]. where b = 1/(4πJ ) . The ‘twist’ parameter E, which leads to a spectral asymmetry in the frequency domain, comes about because the chemical potential µ 6= 0 breaks B. Large N saddle point solution the particle-hole symmetry of the problem. The resulting U(1) charge density Q = hQi /N is related to E by [2, 53] We now derive the large-N saddle point equations of 1 1 the model Eq. (17) in imaginary (Euclidean) time τ, and Q = [1 − tanh(2πE)] − tan−1(e2πE ). (23) 4 π solve them numerically to investigate its properties in the thermodynamic limit. The partition function of the The asymmetry parameter E is also related to the ther- system at inverse physical temperature β is given by modynamic quantity Z Y † R β P † − 0 dτ( i,a=1,2 cia(τ)∂τ cia(τ)+Hκ) 1 ∂S0 Z = DciaDciae . E = (24) i 2π ∂Q (18) where S0 is the residual zero-temperature entropy den- Upon disorder averaging (keeping only replica-diagonal sity of the cSYK model. In the holographic picture E de- terms) and integrating out the fermion fields we arrive scribes the electric field near the charged AdS2 black hole at the following effective action (see AppendixB), horizon, and respects Eq. (24) with S0 the Bekenstein- Z Hawking entropy density of the horizon [2, 54]. S[G, Σ] X h 0 0 − = ln DetM + Σab(τ, τ )Gba(τ , τ) Let us first consider the model at charge neutrality, N 0 a,b τ,τ µ = 0. When κ (and thus G12) is non-zero, an exact J 2 i solution of the saddle-point equations (20) in the low- + G2 (τ, τ 0)G2 (τ 0, τ) (19) energy limit cannot be obtained. Instead we solve them 4 ab ba numerically by iterating until a self-consistent solution 0 1 P † 0 where Gab(τ, τ ) = N hT i cia(τ)cib(τ )i is the averaged is found, choosing the initial seeds for the iteration to 0 1 time-ordered correlator at the saddle point and Σab(τ, τ ) be the non-interacting solution G11(τ) = 2 sgn(τ), and 6

0.16 0.15 = 0.01J 0.5 T/J = 0.001 (a) Egap/J (b) (c) F/N c(T) (d) G 0.3 11 ( /J)2/3 iG 0.18 12 = 0.04J 0.2 0.10 TFD

0.0 0.20 = 0.07J

0.1 0.05 T/J = 0.02 0.22 = 0.10J 2cSYK G11 0.5 iG12 = 0.12J 0.24 0.00 0.5J/T 0 0.5J/T 0.01 0.05 0.10 0.00 0.01 0.02 0.03 0.04 0.05 0.00 0.02 0.04 0.06 J /J T/J T/J

FIG. 3. (a) Imaginary time correlators G11(τ) and G12(τ) for κ/J = 0.03 and temperatures T/J = 0.001 and 0.02. While the solution for G11 at low temperature decays exponentially at long times 1 τJ J/T , at higher temperature it instead −1/2 follows a power-law G11 ∼ sgn(τ)|τ| indicative of SYK behavior. (b) Energy gap extracted from the exponential decay −4 of correlators G11(τ) and G12(τ) as a function of κ at low temperature, T/J = 10 . (c) Free energy density as we decrease temperature from T/J = 0.05 to low temperature, and then increase back to T/J = 0.05. The hysteresis indicates a ﬁrst-order phase transition when the two solutions cross. (d) The transition temperature Tc as a function of the tunneling strength κ/J. The shaded region indicates the parameter range where the two phases coincide in (c).

G12 = i with small . We find that only the real and 0.50 imaginary part of G11 and G12, respectively, are non- zero. For κ = 0 we recover the conformal result Eq. (22) 0.4 for long times and low temperatures J −1 τ β. Gapped (wormhole) 0.25 As shown in Fig.3(a), when turning on a small coupling 0.3 J κ/J = 0.03, a gap opens at low temperature, as indicated / 0.00 0.2 by the exponential decay of the correlators G11 and G12. For high temperatures the correlators instead decay as 0.25 a power law. In Fig.3(b) we show that the energy gap 0.1 extracted from the exponential decay of Gab(τ) at low temperature T/J = 10−4 scales as ∼ ( κ )2/3 for κ/J 1. 0.0 0.50 J 0.4 0.2 0.0 0.2 0.4 This is consistent with the scaling of the analogous MQ /J model [25]. The MQ model also exhibits a first-order phase transition at finite temperature for small values of κ. In the FIG. 4. Charge density Q as a function of κ and µ, extracted gravity context this transition was interpreted [25, 55] numerically from the saddle-point solutions through Eq. (26) as a Hawking-Page transition [48] because it separates a at low temperature T/J = 0.002. The gapped wormhole stable AdS2 black hole at high temperature from a low- phase is stable to inclusion of a finite chemical potential µ temperature phase (the wormhole) which appears ther- and remains charge neutral throughout. For larger |µ| the mal for an observer having access to only one subsystem. system transitions to a gapless phase with tunable charge den- This is also manifest in the wormhole phase admitting sity, then finally to a gapped polarized phase with Q = ±1/2. a TFD ground state (see Eq.3). Such a transition can The dashed lines denote first-order phase transitions. be identified from the thermodynamics of our complex fermion model. The free energy F = −T ln Z is obtained by substituting the saddle point solutions in the action, with the converged solution at the previous temperature), and vice versa. For high temperatures the gap- F h X D(iωn) less cSYK solution is favored, whereas for low temper- = −T 2 ln 2 + ln 2 (25) atures the gapped solution prevails. These two phases N (iωn) ωn can be identified from the temperature dependence of 3 X i + (Σ (iω )G (iω ) − Σ (iω )G (iω )) . the free energy: a constant negative slope S = −∂F/∂T 2 11 n 11 n 12 n 12 n at low temperatures indicates an SYK phase with resid- ωn ual entropy S0, while the gapped wormhole phase with Here we regularized the free energy using its value for 2N a unique ground state shows zero slope. We obtain a P non-interacting complex fermions, 2 n ln(iωn) = 2 ln 2, clear hysteresis between the two solutions indicating a to cancel out divergences at large frequencies in the nu- first-order phase transition. The transition temperature merical evaluation of D(iωn). In Fig.3c we present the Tc, identified with the crossing point of the gapped and free energy density numerically obtained by sweeping gapless free energies in Fig.3(c), increases monotonously from high to low temperatures (starting each iteration with κ/J until the phase transition line terminates at a 7 critical point, as shown in Fig.3(d). When introducing a graphically dual to a traversable wormhole. We rely on non-zero chemical potential µ, the system is not necessar- the following observations: (i) the presence of a TFD ily charge neutral. For a single cSYK model this results ground state with large β˜, (ii) a first-order phase transi- in a spectral asymmetry or a ‘twist’ E in the conformal tion separating the (presumed) gapped wormhole phase limit of the imaginary-time Green’s functions [2, 42], see from a gapless cSYK phase at high temperature and (iii) Eq. (22). The corresponding U(1) charge density Q can revival dynamics showing the transmission of excitations be read off from the value of the imaginary-time Green’s between the two chaotic subsystems. functions near τ = 0,

+ 1 − 1 G11(0 ) = − Q ,G11(0 ) = − − Q. (26) IV. COUPLING cSYK MODELS WITH 2 2 INTERACTION TERMS In Fig.4 we show the U(1) charge, obtained numerically from Eq. (26) as a function of parameters κ and µ. In this section we couple the two cSYK models with We find that the gapped wormhole phase is stable to the four-fermion interactions that conserve charge on each inclusion of a finite chemical potential µ, even though µ system. We consider the Hamiltonian introduced in the breaks the microscopic anti-unitary symmetry P used to previous section, modified by an extra term define the TFD state. The charge density of the worm- X † † † † hole phase remains zero throughout. Increasing µ drives H = Hκ + α Jij;kl ci1cj2ck1cl2 + ci2cj1ck2cl1 , the system into a gapless phase with a tunable charge i,j,k,l 1 1 density Q ∈ [− 2 , 2 ] and then finally to a gapped, po- (28) larized state with the maximal charge Q = ±1/2. The transition to the polarized state is of first order, where the where the coupling constants Jijkl are identical to those extensive entropy of the non-Fermi liquid phase jumps to within each cSYK system. Related models have been zero, similar to the transition seen in Ref. [56]. studied before in the context of symmetry-broken ground states [24] and superconducting instabilities of SYK non- Fermi liquid phases [57]. An additional motivation to C. Real-time dynamics study such an interaction term, as explained in more details in Sec.V, is that it naturally arises in pro- In order to probe the dynamical behavior of the model posed physical realizations of the SYK model in graphene we now switch to real-time representation of the saddle- flakes [43]. The additional α term alters the symme- point equations (see AppendixD for details). Following tries of the model, resulting in different degeneracies. For Ref. [26] we focus on the transmission amplitude α 6= 0 but κ = 0, P1 and P2 are no longer symmetries but the combined anti-unitary symmetry P12 = P1P2 re- 2 N(2N−1) > > θ(t) X † mains, where P12 = (−1) . This symmetry guar- Tab(t) = 2 G (t) ,G (t) = hcja(t)c (0)i ab ab N jb antees a two-fold degeneracy for odd N (see TableI). As j before, degeneracies are lifted at κ 6= 0 as the tunnel- (27) ing term breaks the anti-unitary symmetry for odd N, which expresses the probability amplitude of recovering −1 P12 KP12 = −K. a fermion in system a at time t after inserting the cor- The large-N saddle-point equations are obtained in responding fermion in system b at time 0, averaged over a similar way as Eqs. (20), with details delegated to all fermionic modes j in the system. Fig.5a shows the App.B. The equations for the Green’s functions Gab(iωn) transmission amplitudes for small κ in the low tempera- are unchanged while the expressions for the self-energies ture regime T/J = 10−4 κ/J. They exhibit sharply Σab(τ) acquire additional terms peaked revival oscillations in both T11 and T12 that are out-of-phase, consistent with the propagation of fermions h α2 Σ (τ) = −J 2 1 + G2 (τ)G (−τ) (29) back and forth between the two chaotic systems. As with 11 2 11 11 the Majorana case [26] we find that the sharp revivals rely 2 on a tower of equally-spaced states in the spectral func- α 2 i − 2α G11(τ)G12(τ)G12(−τ) − G11(−τ)G12(τ) , tion, see the inset in Fig.5a, which occur at harmonics of 2 the gap ∼ κ2/3. The overall decay of oscillations is due to h α2 Σ (τ) = J 2 1 + G2 (τ)G (−τ) (30) the width of those spectral peaks, which increases when 12 2 12 12 going to higher frequencies and/or temperatures [27]. By α2 i comparison, for temperatures above the first-order tran- − 2αG (τ)G (−τ)G (τ) − G (−τ)G2 (τ) . 11 11 12 2 12 11 sition shown in Fig.3d we observe a smooth, power-law decay ∼ |t|−1/2 of the transmission amplitude character- The low-temperature phase diagram of this model, ob- istic of the SYK non-Fermi liquid phase. tained from the self-consistent numerical solution of the Based on the results of this Section we conjecture that above equations, is analyzed in Fig.6. At charge neu- two identical cSYKs models coupled with a weak tunnel- trality we find two phases whose properties are discussed ing term admit a low-temperature phase which is holo- in the next subsection: a gapless phase with tunable 8

1.2 1.2 1.2

T11(t) (a) 11 (b) T11(t) T11(t) (c) 11 1.0 T12(t) 10 12 1.0 T12(t) 1.0 T12(t) 10 12 TSYK(t) TSYK(t) TSYK(t)

0.8 0 0.8 0.8 0 1 0 1 0.2 0.0 0.2 /J /J 0.6 0.6 0.6

11 0.4 0.4 0.4 10 12

0.2 0.2 0.2 0 2.5 0.0 2.5 0.0 0.0 /J 0.0 0 100 200 300 400 500 0 20 40 60 80 100 120 0 100 200 300 400 500 tJ tJ tJ

FIG. 5. Transmission amplitudes T11,12(t), as in Eq. (27), for (a) κ = 0.005 and α = 0, (b) κ = 0.005 and α = −1, (c) κ = 0.1 −4 and α = 1, with T/J = 10 . All cases show an initial decay of T11(t) following the SYK power-law behavior TSYK(t) (with κ = α = 0). For parameters as in (a,b) that lead to a gapped phase (see the spectral functions in the respective insets) we find oscillations in both transmissions T11,12(t) that are out of phase as expected in the wormhole scenario [26]. For parameters as in (c) the transmission T11(t) instead tracks the SYK curve, reflecting a black hole-like decay of two-point functions in system 1. The spectral function ρ11(ω) exhibits SYK behavior, while ρ12 6= 0 is anti-symmetric about ω = 0. Hence T12(t) is non-zero and shows a similar decay, reflecting persistent correlations between the two systems.

0.81 0.6 0.4 (a) (b) (c) 0.72 Egap c(T) 0.3 0.5 0.63 = 0 .0 0.3 = 1 0.54 0.4 = 1 = 0.6

J 0.2 0.45 = 4 / .4 0.3 0.2 = 0 0.36 .2 0.27 0.2 = 0 0.1 0.18 0.1 0.0 0.1 = 0.09 0.0 0.00 0.0 0.0 1 0 1 2 3 4 5 0.0 0.1 0.2 0.3 0.4 0.5 0.00 0.01 0.02 0.03 0.04 0.05 0.06 2/3J1/3 T/J

FIG. 6. (a) Spectral gap as a function of parameters κ, α in Eq. (28), for temperature T/J = 10−4 and at charge neutrality 2 2 1 2 µ = 0. (b) Gap scaling as a function of κ for different values of α, with Jα = J (1 + 2 α ) taken constant across all cases to allow a direct comparison. (c) First-order phase transition lines for α in the range from 0 to 1, which separate the gapless and gapped phases of the coupled cSYK models. charge density similar to the cSYK non-Fermi liquid, and a range of κ inside a dome-shaped region, where the −1/2 a gapped phase that is adiabatically connected to the correlator G11(τ) ∼ τ decays as a power-law with α = 0 wormhole solution of Sec.III. The gapped phase the same exponent as in the SYK phase. For α = 4 persists down to κ = 0 for α < 0 or α > 4 through a U(1) a gap opens for any κ 6= 0, similarly to the α = 0 symmetry-breaking mechanism, where a finite expecta- case analyzed in Sec.III. This can be easily understood tion value hKi is generated spontaneously. We discuss from the saddle-point equations (29,30). Using the sym- the physics away from charge neutrality in Sec.IVE. metry of the two-point correlators at charge neutrality, G11(−τ) = −G11(τ) and G12(−τ) = G12(τ) we see that at α = 4 the saddle-point equations reduce to A. Phase diagram at charge neutrality 2 2 Σ11(τ) = −9J G11(τ)G11(−τ) (31) 2 2 Σ12(τ) = 9J G (τ)G12(−τ) (32) The low-temperature phase diagram of the model, 12 Eq. (28) consists of two phases near charge neutrality which are just the equations for two decoupled cSYK µ = 0. For α < 0 or α > 4 the system is in a models (α = 0) but with a renormalized Jα = 3J. When gapped phase for all values of κ, as indicated by the ex- turning on a finite κ we thus expect the same low tem- ponential decay of the two-point correlators Gab(τ) at perature ‘wormhole’ physics as for α = 0, but with a 2/3 1/3 late times 1 τJ J/T . In Fig.6a,b we show the renormalized gap Egap ∼ κ Jα . We verify this scal- gap extracted from that exponential decay at low tem- ing from our numerical simulations, as shown in Fig.6b . perature T/J = 10−4 as a function of κ and α. For The mapping between α = 0 and α = 4 is reminiscent of 0 ≤ α ≤ 4, we find that a gapless phase survives for the duality that exists in the analogous Majorana model 9

0.5 1.0 0.6 2N = 10 0.5 0.4 0.8 2N = 14 2N = 18 0.4 0.3 0.6 N large N / J

| 0.3 /

0.2 0.4 K

| 0.2

0.1 0.2 0.1 0.0 0.0 0.0 2 0 2 4 6 2 0 2 4 6

FIG. 8. Symmetry-breaking mechanism for κ = 0, whereby a FIG. 7. Overlap between the ground state |Ψ0i of the coupled cSYK models, Eq. (28), at charge neutrality and the best-fit finite expectation value of the tunneling operator hKi /κN is TFD state. Results are obtained from exact diagonalization generated spontaneously. The large-N result is plotted along- with 2N = 12 and averaged over 20 disorder realizations. side exact diagonalization results for odd N, which represent the eigenvalues of hKi /κN in the ground state manifold. The ED results are averaged over 8 disorder realizations. in Ref. [24]. However, here the duality is only emergent in the large-N saddle-point equations, and is not present in the microscopic Hamiltonian. This spontaneous U(1) symmetry breaking is accompa- We then compute the free energy of the model, us- nied by a gapless Goldstone mode reflecting phase fluc- ing the same approach as in Sec.III, Eq. (25). The free tuations of the corresponding order parameter [58]. energy shows hysteresis across the phase transition be- For finite N the spontaneous symmetry breaking can tween the gapless and gapped phases for 0 ≤ α ≤ 4. The be analyzed using exact diagonalization, providing a sim- phase transition lines for various α are shown in Fig.6c. ple explanation of the phase transition observed at κ = 0. As α increases the first-order transition lines move up For even N the ground state is unique and we always the κ axis and have a non-zero intercept, such that the find that hKi = 0. For odd N however, we have two de- gapless phase extends down to zero temperature. Thus generate ground states which, in the gapped phase, are for 0 ≤ α ≤ 4, the first-order Hawking-Page phase transi- located in the q = 0 sector (see TableI). We can per- tion occurs at zero temperature upon varying the tunnel- form a basis rotation in this twofold degenerate space ing strength. This is also indicated by the discontinuous to obtain two eigenvectors of K with opposite eigenval- jump in the gap magnitude across the transition shown ues, as shown in Fig.8. The system can thus sponta- in Fig.6b. neously choose a ground state which breaks U(1)⊗U(1) Using exact diagonalization we obtain the overlap be- symmetry, as in the saddle-point result. In the process the anti-unitary symmetry is also spontaneously broken tween the ground state of the coupled model and the −1 TFDs defined in Eq. (12), as shown in Fig.7. In the as P12 KP12 = −K for odd N. In contrast, in the gap- gapped phase the overlap with the TFD decreases con- less phase (for 0 ≤ α ≤ 4) the two ground states are tinuously when moving away from the α = 0 ‘wormhole’ in charge sectors q = ±1 and have hKi = 0 since K line, and sharply drops to zero upon entering the gap- conserves charge. There is thus no possible symmetry less phase. Interestingly, at α = 4 the ground state is breaking, in accordance with the large N result. not well approximated by a TFD, a further indication of the duality between α = 0 and 4 being only valid in the large-N limit. C. Revival dynamics

The transmission amplitudes [Eq. (27)] for non-zero α B. Spontaneous symmetry breaking at low temperature T/J = 10−4 are shown in Fig.5b,c. For α = −1 and small κ, deep inside the gapped phase, When κ = 0 and α < 0 or α > 4, the system spon- we again find revival oscillations. Those are notably less taneously develops a non-zero expectation value for the sharp than at the MQ point α = 0, consistent with the tunneling operator K, which can be read off numerically observation that the ground state is not well approxi- hKi from κN = 2iG12(τ = 0), as shown in Fig.8. We mated by a TFD. The reason is that the gap remains therefore conclude that the gap opens at κ = 0 through large as κ → 0. Therefore, there is only a small num- a spontaneous symmetry-breaking mechanism, from the ber of states in the conformal tower (at harmonics of the U(1)⊗U(1) charge conjugation symmetry down to U(1). gap) that can fit within the energy scale J which limits 10

the conformal scaling behavior [26]. To this end compare ln|G11| 0 = 0.1J the spectral functions at α = 0 (Fig.5a inset), show- ln|G12| 1/2 ing a large number of evenly spaced peaks, and α = −1 1 ln(b11| | ) 1/2 (Fig.5b inset), showing only few and far-spaced spec- 2 ln(b12| | ) tral peaks. Thus at α = −1 the revivals are controlled 3 by a few spectral peaks rather than an extensive tower of states. For α = 1 and small κ (deep inside the gap- 4 0.10 5 b4 b4 less dome) we ﬁnd a power-law decay of the transmis- 0.05 11 12 2 sion down to the lowest accessible temperatures, closely 6 1/(6 J ) 0.00 tracking SYK behavior, but now with T12(t) 6= 0 also de- 7 0.1 0.2 caying as a power-law. The spectral function ρ11 (Fig.5c 8 inset) shows gapless power-law behavior as expected in 4 2 0 2 4 6 8 the SYK phase, however ρ12 6= 0 indicates the presence ln| J| of correlations between the two subsystems.

FIG. 9. Imaginary-time correlators G11(τ) and G12(τ) versus time, on log scales, for α = 1, κ = 0.1J and low temperature D. Conformal solution and SU(2) symmetry T/J = 0.002. Fitting to the conformal scaling form for Gab(τ) gives the coefficients bab. The inset shows the difference of fourth powers of the coefficients as function of κ, verifying We now consider whether a low-energy conformal so- the analytical solution discussed in the text. lution of the saddle-point equations (29), (30) can be found to describe the gapless phase. When κ = 0, we have G12(τ) = 0 and the saddle-point equations can be crossing, which occurs at the first-order transition at κc solved at low energies to show that G11(τ) is a conformal 2 2 2 1 2 where the gapless phase becomes favored. cSYK correlator with J replaced by Jα = J (1 + 2 α ). Note that this connects smoothly with the behavior at α = 0, 4 discussed below Eqs. (31)-(32). With κ > 0, G12(τ) does not vanish which renders the analytical solu- E. Moving away from charge neutrality: a tale of tion of the saddle-point equations more difficult. In the two black holes low energy limit, we find that a conformal solution for both G11 and G12 is in general not possible, except at the special point α = 1 discussed in AppendixC. This We now discuss the physics away from charge neutral- limit admits a power-law solution with the same power ity. We focus on the SU(2) symmetric point α = 1 and in- −1/2 vestigate its low-temperature phase diagram in the κ−−µ for both correlators, Gab(τ) ∼ bab|τ| in the long-time limit 1 τJ J/T , and where the two coefficients are plane. To this end we show in Fig. 10 the U(1) charge 4 4 2 obtained numerically from Eq. (26) and the residual en- related by b12 = b11 − 1/ 6πJ . This is demonstrated numerically in Fig.9. Note that the saddle-point equa- tropy density S0, as a function of parameters κ and µ tions alone are not sufficient to fix the coefficients of the at low temperature T/J = 0.002. We recover the known two power laws. Instead one must impose a constraint gapped phases discussed above: for large κ, which ad- linking microscopic and conformal physics, similar to how mits the non-interacting charge-neutral |TFD0i ground U(1) charge enters the conformal solution in the cSYK state, and for large |µ|, which corresponds to the polar- model [2, 53] (see AppendixC). ized Q = ±1/2 states. The boundaries of the two gapped The α = 1 point is special because it has SU(2) sym- phases host first-order phase transitions to gapless non- metric interactions [59]. An important consequence is Fermi liquids, as indicated by the discontinuous jump in that the tunneling term K is now a symmetry of the entropy density at the phase boundaries. Surprisingly, we find not one but two such gapless phases. Near charge model, [Hα=1,K] = 0. Hence hKi is a conserved quantity that can be tuned by the tunneling parameter κ, in neutrality, we obtain a phase smoothly connected to the analogy with the U(1) charge density Q tuned by the conformal solution discussed above. In this phase both chemical potential µ. However, the two symmetries have correlators G11 and G12 show power-law decay which in- different signatures: introducing κ generates a non-zero dicates strong correlations between the two subsystems. This phase can be thought of as a single cSYK phase value of iG12, but not a twist parameter leading to a spectral asymmetry, as occurs with non-zero µ. Another with 2N fermions, dual to a ‘large’ black hole comprising consequence of the SU(2) symmetry is that the non- all degrees of freedom in the combined system. interacting ground state |TFD0i of K is an eigenstate Farther from charge neutrality we find another first- of the full model for any κ. In fact, exact diagonaliza- order phase transition to a different gapless phase with tion shows that for κ > κc ' 0.27J the model admits charge density Q' 0.25 and about half of the residual the |TFD0i state as an exact ground state. Because the entropy at charge neutrality. To understand this, note interaction and tunneling terms commute, the only way that at α = 1 where the interactions are SU(2) invariant, we can rotate to a new basis c = √1 (c ± ic ), such to change the ground state is through an energy level j± 2 j1 j2 11

1.6 0.50 sgn( )G + + 0.4 Gapped (TFD0) 10 1.4 sgn( )G 0.25 sgn( )G 1.2 SYK 0.3 5 Gapless J / (small BH) 0.00 1.0 0.2 0 1 0 1 0.8 Gapless /J 0.1 0.25 (large BH) 0.6

0.0 0.50 = 0.3, = 0.3 0.4 0.2 0.0 0.2 0.4 0.4 /J 0.2

0.0 20 0 20 40 60 80 0.4 J 0.75

0.3 FIG. 11. Imaginary-time correlators G±(τ) of fermions c±,

J 0.50 −4

/ cf. Eq. (33), for α = 1, κ = µ = 0.3J and T/J = 10 . In- 0.2 set: corresponding spectral functions ρ±. The behavior of G+ and ρ+ closely follows that of a charge-neutral cSYK model, 0.25 0.1 while ρ− is gapped with finite spectral weight only at negative frequencies. Hence G−(τ) is zero for τ > 0 and shows exponential decay for τ < 0. 0.0 0.00 0.4 0.2 0.0 0.2 0.4 /J law scaling at low frequency in the ρ+ channel while the ρ− channel is gapped, confirming the argument above. In FIG. 10. Charge density Q (top) and residual entropy density imaginary time, the corresponding correlators G±(τ) = † S0 (bottom) as a function of κ and µ, at the SU(2) invariant 1 P N jhT cj±(τ)cj±(0)i = G11(τ) ∓ iG12(τ) show power- point α = 1 and low temperature T/J = 0.002. The charge law and exponential decay, respectively, at long times. density is tunable in the compressible gapless phases. Large In the vicinity of the µ = ±κ lines the residual en- |µ| leads to a gap opening when Q = ±1/2, corresponding tropy and charge density change smoothly, see Fig. 10, to a fully polarized state, while large κ leads to a charge- as one half of the system is in a compressible cSYK state neutral gapped phase with a |TFD0i ground state. The two gapless phases show extensive residual entropy consistent with while the other half remains gapped. We interpret this a ‘large’ and ‘small’ black hole with 2N and N degrees of phase as dual to a ‘small’ black hole, comprising half freedom, respectively. All phases are separated by first-order of the degrees of freedom of the combined system, with phase transitions indicated by dashed lines. The cuts µ = ±κ the other half decoupled and frozen into a fully polarized discussed in the text are shown by black dotted lines. state. Interestingly, we find that this “small black hole” phase persists away from α = 1 even though the SU(2) symmetry allowing the basis change argument is absent. that X X X H = J c† c† c c − µ c† c . (33) ij;kl ia jb ka lb a ja ja V. PHYSICAL REALIZATION IN GRAPHENE a,b=± i,j,k,l a=±,j FLAKES In this basis one can interpret the system as two cSYK models with different chemical potentials µ± = µ∓κ and We now turn to potential physical realizations of the SU(2) invariant interactions between them. Let us first model introduced in this work. As originally described focus on the κ = µ line where the chemical potentials are in Ref. [43], a promising platform for realizing cSYK simply 0 and 2µ. When µ increases, eventually one of the physics is a mesoscopic graphene flake under a perpendic- cSYK models undergoes a first-order phase transition to ular magnetic field B, with the chemical potential µ ly- a gapped, polarized state with Q = 1/2. At low energies ing within the zeroth-Landau level (LL0). An Aharonov- (below the gap), its degrees of freedom thus decouple Casher argument [60] implies that LL0 remains sharp as and we are left with the other cSYK model at charge long as the chiral (sublattice) symmetry of the model is neutrality Q = 0. The combined system thus has exactly unbroken, thus forbidding two-fermion terms that would Q = 1/4 and S0 = ScSYK, as observed in the saddle-point destroy the non-Fermi liquid physics at low energies. Dis- solutions. In Fig. 11 we show the spectral functions ρ± order that preserves this chiral symmetry (such as an for the rotated basis fermions c±. We observe a power- irregular boundary) then imprints disorder on the LL0 12 wavefunctions without lifting their degeneracy, leading to model are invariant under SU(2) rotations in spin space. random and all-to-all interactions between them. Following the analysis in Sec.IVE we can perform a basis Due to the negligible spin-orbit coupling in clean change c = √1 (c ±ic ), which corresponds (up to a j,↑↓ 2 j1 j2 graphene, it is reasonable to assume identical wavefunc- π gauge transformation) to a rotation by 2 along the x-axis tions for the two spin components. This should still in spin space, hold in the presence of non-magnetic disorder (such as an irregular boundary) which preserves the SU(2) sym- X X † † X † † H = Jij;klc c ckaclb + igµBB (c ci2 − c ci1). metry of Coulomb interactions. The graphene flake setup ia jb i1 i2 ijkl a,b i thus naturally leads to two identical copies of the cSYK (38) model, one for each spin component. In Ref. [43] the authors argued that the Zeeman splitting obtained by This has the same form as the model in Eq. (28) with applying a magnetic field to the sample generates a large α = 1, if we identify gµBB with the tunneling amplitude spin gap (augmented by exchange interactions), which κ. As mentioned in the previous section, the SU(2) sym- effectively reduces the problem to a single cSYK model. metric interactions commute with the tunneling term. In In this Section we revisit this analysis by looking more the original basis of Eq. (37), this is easily seen by noting carefully at the role of spin in the above proposal. Using that the Zeeman term is proportional to the total spin a mapping to the model studied in Sec.IV we conclude tot P z projection Sz = i σi . that the graphene flake model with a weak Zeeman split- Using this mapping we thus expect that the graphene ting is in a gapless cSYK phase with fermion scaling di- flake remains in the gapless cSYK non-Fermi liquid phase mension 1/4 and tunable charge density. In contrast to up to a threshold value κ = gµ B ∼ 0.27J. Above the expectations that a strong Zeeman splitting should give c B c critical field strength Bc the system becomes gapped and rise to a cSYK phase [43], we find that it instead leads the ground state is the infinite-temperature TFD state to a gapped phase with an exact |TFD0i ground state. |Ii, which in the graphene basis is a fully spin-polarized The Coulomb interactions between electrons in the state, Q c† |0i. In Ref. [43] the authors estimate that graphene flake read j j↓ for B ∼ 20 T the Zeeman splitting gµBB ∼ 2.4 meV 1 X 0 while the Coulomb interaction strength J ∼ 25 meV, H = ρ V (r − r )ρ 0 (34) int 2 r r thus placing the system within the gapless phase. Given r,r0 various uncertainties in these estimates J could well be where ρr = ρr↑ + ρr↓ is the total charge density at a smaller in which case the system would realize the gapped point r in space and V (r − r0) is the screened Coulomb phase indicated in Fig. 10. Further, the power-law scal- potential. The electronic charge densities can be ex- ing characteristic of the conformal regime is expected for 3 pressed in terms of the eigenfunctions φj(r) of the non- temperatures T Jα ∼ 2 J at the SU(2) symmetric interacting Hamiltonian which, neglecting spin-orbit cou- point α = 1, corresponding to T 430 K which should pling effects, are independent of spin σ =↑, ↓, enable exploration of the low temperature regime. As shown in Sec.IVE the cSYK non-Fermi liquid † X ∗ † ρrσ = crσcrσ = φi (r)φk(r)ciσckσ (35) phase for B = 0 is stable against inclusion of a chem- ik ical potential up to a threshold value µc which corresponds to adding charge density Q = 1/2, filling all the Projecting to the LL0 wavefunctions, Eqs. (34, 35) lead to states in LL0. For B 6= 0 the system first transitions a (normal-ordered) interaction Hamiltonian with all-to- to an intermediate non-Fermi liquid phase corresponding P † † all couplings Hint = ijkl Jij;klciσcjσ0 ckσclσ0 and spin- to a small black hole. This transition could be explored independent coupling constants by tuning the chemical potential in the graphene flake 1 X by external gates. Experimentally, the non-Fermi liq- J = − φ∗(r)φ (r)V (r − r0)φ∗(r0)φ (r0). (36) ij;kl 2 i k j l uids could be distinguished from each other, and from r,r0 the gapped phases at large κ or large µ, by measuring their charge compressibility ∂Q/∂µ [10, 42] or their spec- Assuming spatially random wavefunctions and strong tral function through spin-polarized scanning probe tech- screening these become complex random Gaussian vari- niques. Indeed, the small black hole phase corresponds ables [43] (see Ref. [61] for a discussion of varying screen- to an SYK-type non-Fermi liquid for one spin component ing lengths in a related model). and a gapped, polarized state for the other. This should Adding the Zeeman term, the Hamiltonian describing be contrasted with the large black hole phase which con- the low-energy physics of the graphene flake becomes sists of a non-Fermi liquid in both spin components, and X X † † X the fully gapped states which comprise either filled or H = Jij;klc c 0 ckσclσ0 + gµBB (ni↑ − ni↓), iσ jσ empty spin-polarized Landau levels . ijkl σ,σ0 i (37) An important caveat of our analysis is that in Eq. (38) the coupling constants Jij;kl are only restricted to be anti- where g ∼ 2 and µB are the Landéfactor and the Bohr symmetric under exchanging (iσ, jσ) and (kσ, lσ) – with magneton, respectively. The Coulomb interactions in this the same spin component – by fermionic commutation 13 relations. This is a weaker requirement than the full an- symmetry P which enables the TFD state construction, tisymmetry present for the Majorana SYK model and we restricted the model to only contain interactions that assumed in this work (the same assumption was made are completely antisymmetric in the indices i, j, k, l. In in Ref. [43]). For example, interaction terms with two the large-N limit these should dominate as their num- pairs of matching indices, corresponding to direct (den- ber scales as N 4 (in contrast, the number of terms with sity) interactions or to exchange (spin) interactions [62] one or two pairs of identical indices, of the form Jij;ik 3 2 are excluded from our model. A detailed analysis of such or Jij;ij, respectively scales as N and N ). However, effects is left for future work. for mesoscopic realizations of SYK physics with finite N these terms could be important; understanding their effect will be an important step towards connecting our VI. CONCLUSION AND OUTLOOK results with ongoing experimental efforts. A different approach to define a TFD for cSYK models In this work we generalized the ‘eternal traversable could be to rely on an anti-unitary time-reversal (rather wormhole’ construction of Maldacena and Qi [25] to a than particle-hole) symmetry. Time-reversal symmetry system of coupled complex SYK models with a global is obviously broken for a single cSYK model (as manifest U(1) charge symmetry. We explained how to define the by the coupling constants Jijkl being complex), but can TFD state in the presence of a U(1) symmetry, and be restored globally by considering a pair of time-reversed showed that the model admits a gapped phase with a cSYK models. We anticipate that the TFD construction, charge-neutral ground state close to a TFD. Whether the saddle-point physics and physical realizations will be dif- weak-tunneling and low temperature limit of the model ferent in this case, and leave its detailed study for future admits a gravitational dual similar to the wormhole of work. Another interesting topic concerns the quantum Ref. [25] remains an intriguing open question for the high- chaotic properties of our model: how does scrambling, energy community. The presence of a gapped ground as captured through out-of-time-ordered correlators, be- state close to a large β˜ thermofield double, a first-order have across the Hawking-Page transition which separates Hawking-Page phase transition to a gapless cSYK non- the wormhole and black hole phases? Fermi liquid at high temperature and sharp revivals in Finally, in light of the rich phase diagram of our model, fermion transmissions are however highly suggestive. it is interesting to ask about potential physical realiza- Further, we considered the effect of four-fermion in- tions for generic α 6= 1. In the graphene flake proposal, teractions between the two cSYK models that are disor- α = 1 is enforced by the SU(2) symmetry of Coulomb dered identically to the interactions within each system. interactions. However, in other model systems where the We explored the phase diagram of the system as a func- two subsystems are realized by two surfaces, such as mul- tion of tunneling, interactions and chemical potential in tilayer graphene [35] or a topological insulator flake, one Figs.6, and 10. At low temperature we obtain three non- could imagine changing the distance between the surfaces trivial phases: a gapped, charge-neutral phase which is as a way to tune the ratio of inter-system to intra-system adiabatically connected to the conjectured wormhole and interactions α. In this way one could potentially explore two gapless, compressible non-Fermi liquid phases which the Hawking-Page phase transition between the worm- describe either a ‘large’ or a ‘small’ (i.e. with half of hole and black hole phases discussed in this work. its degrees of freedom gapped out) charged black hole Note added. – Recently, an independent study of the with an AdS2 horizon. All phases are separated by first- spontaneous U(1) symmetry breaking in coupled complex order phase transitions exhibiting extensive residual en- SYK models, focusing on the α term, was posted [58]. tropy jumps. The transition out of the gapped wormhole Our results match where they overlap. phase can be understood as a Hawking-Page transition, as it separates a black hole from a phase appearing lo- cally thermal, a consequence of its TFD ground state. ACKNOWLEDGMENTS The phase diagram also contains the special case α = 1 with SU(2)-symmetric interactions which admits a con- We are grateful to Oguzhan Can, Chengshu Li, Moshe formally invariant solution at low energies, and is directly Rozali and Xiao-Liang Qi for illuminating discussions. relevant to the graphene flake proposal of Ref. [43]. This research was supported in part by NSERC, CIfAR, We conclude by highlighting a few caveats of our anal- the Heising-Simons Foundation, the Simons Foundation, ysis and point out interesting directions for further work. and National Science Foundation Grant No. NSF PHY- First, in order to define the anti-unitary particle-hole 1748958.

[1] Subir Sachdev and Jinwu Ye, “Gapless spin-ﬂuid ground [2] Subir Sachdev, “Bekenstein-hawking entropy and strange state in a random quantum heisenberg magnet,” Phys. metals,” Phys. Rev. X 5, 041025 (2015). Rev. Lett. 70, 3339–3342 (1993). 14

[3] A. Kitaev, “A simple model of quantum holography,” [21] Marcel Franz and Moshe Rozali, “Mimicking black hole (2015), KITP Strings Seminar and Entanglement Pro- event horizons in atomic and solid-state systems,” Nat. gram. Rev. Mater. 3, 491 (2018). [4] Juan Maldacena and Douglas Stanford, “Remarks on [22] L. Garc´ıa-Alvarez,´ I. L. Egusquiza, L. Lamata, A. del the sachdev-ye-kitaev model,” Phys. Rev. D 94, 106002 Campo, J. Sonner, and E. Solano, “Digital quantum (2016). simulation of minimal AdS/CFT,” Phys. Rev. Lett. 119, [5] Juan Maldacena, Stephen H. Shenker, and Douglas 040501 (2017). Stanford, “A bound on chaos,” J. High Energy Phys. [23] Zhihuang Luo, Yi-Zhuang You, Jun Li, Chao-Ming 2016, 106 (2016). Jian, Dawei Lu, Cenke Xu, Bei Zeng, and Raymond [6] Shao-Kai Jian and Hong Yao, “Solvable sachdev-ye- Laflamme, “Quantum simulation of the non-fermi-liquid kitaev models in higher dimensions: From diffusion to state of sachdev-ye-kitaev model,” npj Quantum Inf. 5, many-body localization,” Phys. Rev. Lett. 119, 206602 53 (2019). (2017). [24] Jaewon Kim, Igor R. Klebanov, Grigory Tarnopolsky, [7] Yingfei Gu, Xiao-Liang Qi, and Douglas Stanford, “Lo- and Wenli Zhao, “Symmetry breaking in coupled syk or cal criticality, diffusion and chaos in generalized sachdev- tensor models,” Phys. Rev. X 9, 021043 (2019). ye-kitaev models,” J. High Energy Phys. 2017 (2017). [25] Juan Maldacena and Xiao-Liang Qi, “Eternal traversable [8] Sumilan Banerjee and Ehud Altman, “Solvable model for wormhole,” (2018), arXiv:1804.00491. a dynamical quantum phase transition from fast to slow [26] Stephan Plugge, Etienne´ Lantagne-Hurtubise, and Mar- scrambling,” Phys. Rev. B 95, 134302 (2017). cel Franz, “Revival dynamics in a traversable wormhole,” [9] David J. Gross and Vladimir Rosenhaus, “The bulk dual Phys. Rev. Lett. 124, 221601 (2020). of SYK: cubic couplings,” J. High Energy Phys. 2017 [27] Xiao-Liang Qi and Pengfei Zhang, “The coupled syk (2017). model at finite temperature,” (2020), arXiv:2003.03916. [10] Richard A. Davison, Wenbo Fu, Antoine Georges, Yingfei [28] P. Gao, D. L. Jafferis, and A. C. Wall, “Traversable Gu, Kristan Jensen, and Subir Sachdev, “Thermoelectric wormholes via a double trace deformation,” J. High En- transport in disordered metals without quasiparticles: ergy Phys. 12, 151 (2017). The sachdev-ye-kitaev models and holography,” Phys. [29] Juan Maldacena, Douglas Stanford, and Zhenbin Yang, Rev. B 95, 155131 (2017). “Diving into traversable wormholes,” Fortschr. Phys. 65, [11] Gábor Sárosi,“Ads2 holography and the syk model,” 1700034 (2017). (2017), arXiv:1711.08482. [30] Juan Maldacena, Alexey Milekhin, and Fedor [12] Xue-Yang Song, Chao-Ming Jian, and Leon Balents, Popov, “Traversable wormholes in four dimensions,” “Strongly correlated metal built from sachdev-ye-kitaev arXiv:1807.04726 (2018). models,” Phys. Rev. Lett. 119, 216601 (2017). [31] Dongsu Bak, Chanju Kim, and Sang-Heon Yi, “Bulk [13] Xiaochuan Wu, Xiao Chen, Chao-Ming Jian, Yi-Zhuang view of teleportation and traversable wormholes,” J. High You, and Cenke Xu, “Candidate theory for the strange Energy Phys. 2018, 140 (2018). metal phase at a finite-energy window,” Phys. Rev. B 98, [32] Ping Gao and Hong Liu, “Regenesis and quantum 165117 (2018). traversable wormholes,” J. High Energy Phys. 2019 [14] Aavishkar A. Patel, John McGreevy, Daniel P. Arovas, (2019). and Subir Sachdev, “Magnetotransport in a model of [33] Zicao Fu, Brianna Grado-White, and Donald Marolf, a disordered strange metal,” Phys. Rev. X 8, 021049 “Traversable asymptotically flat wormholes with short (2018). transit times,” Classical and Quantum Gravity 36, [15] Debanjan Chowdhury, Yochai Werman, Erez Berg, and 245018 (2019). T. Senthil, “Translationally invariant non-fermi-liquid [34] Dongsu Bak, Chanju Kim, and Sang-Heon Yi, “Experi- metals with critical fermi surfaces: Solvable models,” mental probes of traversable wormholes,” J. High Energy Phys. Rev. X 8, 031024 (2018). Phys. 2019, 5 (2019). [16] Peter Cha, Nils Wentzell, Olivier Parcollet, Antoine [35] Etienne´ Lantagne-Hurtubise, Stephan Plugge, Oguzhan Georges, and Eun-Ah Kim, “Linear resistivity and Can, and Marcel Franz, “Diagnosing quantum chaos in sachdev-ye-kitaev (SYK) spin liquid behavior in a quan- many-body systems using entanglement as a resource,” tum critical metal with spin-1/2 fermions,” Proceedings Phys. Rev. Research 2, 013254 (2020). of the National Academy of Sciences 117, 18341–18346 [36] Jingxiang Wu and Timothy H. Hsieh, “Variational ther- (2020). mal quantum simulation via thermofield double states,” [17] Ippei Danshita, Masanori Hanada, and Masaki Tezuka, Phys. Rev. Lett. 123, 220502 (2019). “Creating and probing the sachdev–ye–kitaev model with [37] Adam R. Brown, Hrant Gharibyan, Stefan Leichenauer, ultracold gases: Towards experimental studies of quan- Henry W. Lin, Sepehr Nezami, Grant Salton, Leonard tum gravity,” Prog. Theor. Exp. Phys. 2017, 083I01 Susskind, Brian Swingle, and Michael Walter, “Quan- (2017). tum Gravity in the Lab: Teleportation by Size and [18] Chenan Wei and Tigran A. Sedrakyan, “Optical lattice Traversable Wormholes,” arXiv:1911.06314 (2019). platform for the syk model,” (2020), arXiv:2005.07640. [38] Ping Gao and Daniel Louis Jafferis, “A Traversable [19] D. I. Pikulin and M. Franz, “Black hole on a chip: Pro- Wormhole Teleportation Protocol in the SYK Model,” posal for a physical realization of the sachdev-ye-kitaev arXiv:1911.07416 (2019). model in a solid-state system,” Phys. Rev. X 7, 031006 [39] William Cottrell, Ben Freivogel, Diego M. Hofman, and (2017). Sagar F. Lokhande, “How to build the thermofield double [20] Aaron Chew, Andrew Essin, and Jason Alicea, “Ap- state,” J. High Energy Phys. 2019 (2019). proximating the sachdev-ye-kitaev model with majorana wires,” Phys. Rev. B 96, 121119(R) (2017). 15

[40] Antonio M. Garc´ıa-Garc´ıa,Tomoki Nosaka, Dario Rosa, (2017). and Jacobus J. M. Verbaarschot, “Quantum chaos tran- [51] Jan Behrends and Benjamin Béri,“Symmetry classes, sition in a two-site sachdev-ye-kitaev model dual to an many-body zero modes, and supersymmetry in the eternal traversable wormhole,” Phys. Rev. D 100, 026002 complex sachdev-ye-kitaev model,” Phys. Rev. D 101, (2019). 066017 (2020). [41] Fabien Alet, Masanori Hanada, Antal Jevicki, and [52] Other choices of maximally-entangled states are in prin- Cheng Peng, “Entanglement and confinement in coupled ciple possible. Our choice of this particular Bell state is quantum systems,” (2020), arXiv:2001.03158. motivated by the physical coupling introduced in Sec.III, [42] Yingfei Gu, Alexei Kitaev, Subir Sachdev, and Grigory and is reflected in the phase factors appearing in the def- Tarnopolsky, “Notes on the complex sachdev-ye-kitaev inition of the anti-unitary symmetry, Eq.(15). model,” J. High Energy Phys. 2020 (2020). [53] A. Georges, O. Parcollet, and S. Sachdev, “Quantum [43] Anffany Chen, R. Ilan, F. de Juan, D. I. Pikulin, and fluctuations of a nearly critical heisenberg spin glass,” M. Franz, “Quantum holography in a graphene flake with Phys. Rev. B 63, 134406 (2001). an irregular boundary,” Phys. Rev. Lett. 121, 036403 [54] Subir Sachdev, “Universal low temperature theory of (2018). charged black holes with AdS2 horizons,” Journal of [44] Alexander Altland, Dmitry Bagrets, and Alex Mathematical Physics 60, 052303 (2019). Kamenev, “Sachdev-ye-kitaev non-fermi-liquid correla- [55] Juan Maldacena and Alexey Milekhin, “Syk wormhole tions in nanoscopic quantum transport,” Phys. Rev. Lett. formation in real time,” arXiv:1912.03276. 123, 226801 (2019). [56] Tatsuo Azeyanagi, Frank Ferrari, and Fidel I. Schapos- [45] N. V. Gnezdilov, J. A. Hutasoit, and C. W. J. Beenakker, nik Massolo, “Phase diagram of planar matrix quantum “Low-high voltage duality in tunneling spectroscopy of mechanics, tensor, and sachdev-ye-kitaev models,” Phys. the sachdev-ye-kitaev model,” Phys. Rev. B 98, 081413 Rev. Lett. 120, 061602 (2018). (2018). [57] Debanjan Chowdhury and Erez Berg, “Intrinsic super- [46] Oguzhan Can, Emilian M. Nica, and Marcel Franz, conducting instabilities of a solvable model for an inco- “Charge transport in graphene-based mesoscopic real- herent metal,” Phys. Rev. Research 2, 013301 (2020). izations of sachdev-ye-kitaev models,” Phys. Rev. B 99, [58] Igor R. Klebanov, Alexey Milekhin, Grigory Tarnopolsky, 045419 (2019). and Wenli Zhao, “Spontaneous breaking of u(1) symme- [47] Alexander Kruchkov, Aavishkar A. Patel, Philip Kim, try in coupled complex syk models,” arXiv:2006.07317 and Subir Sachdev, “Thermoelectric power of sachdev- [hep-th]. ye-kitaev islands: Probing bekenstein-hawking entropy in [59] The full symmetry group at α = 1 is quantum matter experiments,” Phys. Rev. B 101, 205148 U(2) = SU(2) ⊗ U(1), as the U(1) charge conserva- (2020). tion is also present. [48] S. W. Hawking and Don N. Page, “Thermodynamics [60] Y. Aharonov and A. Casher, “Ground state of a spin-1/2 of black holes in anti-de sitter space,” Commun. Math. charged particle in a two-dimensional magnetic field,” Phys. 87, 577–588 (1983). Phys. Rev. A 19, 2461–2462 (1979). [49] Wenbo Fu and Subir Sachdev, “Numerical study of [61] Etienne´ Lantagne-Hurtubise, Chengshu Li, and Marcel fermion and boson models with infinite-range random in- Franz, “Family of sachdev-ye-kitaev models motivated by teractions,” Phys. Rev. B 94, 035135 (2016). experimental considerations,” Phys. Rev. B 97, 235124 [50] Yi-Zhuang You, Andreas W. W. Ludwig, and Cenke (2018). Xu, “Sachdev-ye-kitaev model and thermalization on the [62] Alexander Altland and Ben D. Simons, Condensed Mat- boundary of many-body localized fermionic symmetry- ter Field Theory, 2nd ed. (Cambridge University Press, protected topological states,” Phys. Rev. B 95, 115150 2010).

Appendix A: TFD construction

Consider the occupation number basis in the Fock space of 2N complex fermions in the doubled system. The product state

Y 1 −iφ |Ii = √ (|1i1|0i2 − e |0i1|1i2)i (A1) i=1,N 2

P † is a unique ground state of the tunneling Hamiltonian term, K = κ i ci1ci2 + H.c. with κ a complex coefficient iφ |κ|e . Showing that the overlap hI|TFD0i = 1, where |TFD0i is defined in Eq. (14), is equivalent to showing that the expectation value of K in the TFD state has minimum value i.e. −|κ|N. Consider the general definition of TFD in Eq. (1) at β˜ = 0 and evaluate the expectation value

! 1 X X X † hKi = hTFD |K|TFD i = hn¯ 0 | ⊗ hn 0 | κ c c + H.c. |m i ⊗ |m¯ i . (A2) 0 0 2N −q 2 q 1 i1 i2 q 1 −q 2 q,q0 m,n i 16

We insert an identity operator using complete set of basis states with appropriate normalization,

X X 0 0 0 0 I = |mq00 i1 ⊗ |nq000 i2 hnq000 |2 ⊗ hmq00 |1 q00,q000 m0,n0 to separate the operators acting on system 1 and 2,   X X 1 X X X  +iπ(q+ N ) 0 † 0 0 0 hKi = κe 2 hn¯ 0 | n 000 i hn 0 | c |m 00 i hm 00 | m i hn 000 | c |m¯ i 2N −q 2 q 2 q 1 i1 q 1 q 1 q 1 q 2 i2 −q 2 q00,q000 m0,n0 q,q0 m,n i   | {z 0 000} | {z }  δn,n¯ 0 δ−q ,q δm0,mδq00,q  (A3)  00 N †  ∗ +iπ(q + 2 ) 0 0 0 0 + κ e hn¯−q0 |2ci2|nq000 i2 hnq0 |1mq00 i1hmq00 |1ci1|mqi1 hnq000 |2m¯ −qi2 . | {z } | {z } δn,m0 δq0,q00 δn0,m¯ δq000,−q 

+iπ(q+ N ) The phase factor e 2 arises when we commute ci2 across states in system 1 to act on site i of system 2. Summing over Kronecker delta symbols we ﬁnd

1 X X X +iπ(q+ N ) † ∗ +iπ(q0+ N ) † hKi = κe 2 hn 0 | c |m i hn¯ 0 | c |m¯ i + κ e 2 hn¯ 0 | c |m¯ i hn 0 | c |m i . 2N q 1 i1 q 1 −q 2 i2 −q 2 −q 2 i2 −q 2 q 1 i1 q 1 q,q0 m,n i (A4) We now observe that matrix elements in Eq. (A4) can be evaluated separately for system 1 and 2 which, importantly, are identical. The matrix elements can thus diﬀer between the systems at most by a phase. We choose this phase so that it cancels the phase factors present in Eq. (A4), namely

−iπ(q+ N ) −iφ hn¯−q0 |2ci2|m¯ −qi2 = −e 2 e hmq|1ci1|nq0 i1. (A5) This corresponds to the following deﬁnition of the anti-unitary symmetry

iπΓ π −η 4 −iq(φ− 2 ) |n¯−qi2 = Θ|nqi1 = e e P |nqi1 , (A6)

N † (N−1)(N+2) q+ 2 Q 2 where Γ = (−1) is the fermion parity of the SYK eigen-state |nqi and P = i(ci1 + ci1) and η = (−1) −1 † is a sign that depends on total number of fermion in SYK model such that P ci1P = ηci1. To check this note that the expectation in Eq (A5) is only non-zero when q = q0 − 1 and a similar argument holds also for κ∗ term when q = q0 + 1 s.t.

−iφ i π −iη π (Γ0−Γ) −1 hn¯−q0 |2ci,2|m¯ −qi2 = e e 2 e 4 hnq0 |1P ci1P |mqi1δq,q0−1 (A7)

0 N where phases are complex conjugated. Noticing (Γ − Γ) is 2 or −2 when q + 2 is odd or even, we can replace the −iφ −iπ(q+ N ) −1 † phase simply by −ηe e 2 . Using P ci1P = ηci1, the above expression becomes N † N −iφ −iπ(q+ 2 ) ∗ −iφ −iπ(q+ 2 ) −e e (hnq0 |1ci1|mqi1) = −e e hmq|1ci1|nq0 i1 (A8) Eq. (A6) gives the TFD deﬁnition quoted in the main text, Eq. (14), with the required phase factors. The remaining task is to show that such an expression for TFD gives the proper expectation value for the tunneling operator. Substituting Eq. (A5) in Eq. (A2), the expectation value becomes

−iφ ∗ iφ κe + κ e X X X † hKi = − hn 0 | c |m i hm | c |n 0 i , (A9) 2N q 1 i1 q 1 q 1 i1 q 1 q,q0 m,n i

iφ where both the states and operators now refer to system 1. Summing over |mqi and recalling that κ = |κ|e , we have 2|κ| X X 2|κ| X N hKi = − hn | c† c |n i = − (n ) = −|κ|N. (A10) 2N q 1 i1 i1 q 1 2N q n n,q i q q

N N We used the fact that the number of states in charge sector q with fermion number nq = q + is and that 2 nq N X N N N nq = 2 . (A11) nq 2 nq =0 17

Appendix B: Large N Schwinger-Dyson equations

The partition function of our model, in the Euclidean time formalism, is    Z Z β Y Y † X † Z = DciaDcia exp − dτ  cia(τ)∂τ cia(τ) + H (B1) i a=1,2 0 i,a=1,2 where ! X X † † † † † † X † X † † H = Jijkl ciacjackacla + α ci1cj2ck1cl2 + ci2cj1ck2cl1 − µ ciacia + iκ (ci1ci2 − ci2ci1) (B2) i,j,k,l a i,a i and the imaginary-time dependence of the Grassmann variables, cia(τ) is implied. In order to perform the disorder average, one must ﬁrst rewrite the Hamiltonian H in a way which makes its symmetries explicit. In other words, we only want to sum over independent couplings Jijkl. The SYK term (diagonal in a) becomes

X X † † X X † † Jij;kl ciacjackacla = 4 Jij;kl ciacjackacla (B3) i,j,k,l a i

Similarly, using permutations the α dependent term can be written as

X h † † † † † † † † i 2α Jij;kl ci1cj2ck1cl2 + ci2cj1ck1cl2 + ci1cj2ck2cl1 + ci2cj1ck2cl1 i

Let us now focus on the interacting part of the action (involving the coupling constants Jijkl). We can perform a quenched disorder average to calculate the averaged partition function Z Z Z ∗ † ∗ ∗ Zint = D[J, J ]P (Jijkl)Zint = D[c , c] D[J, J ]P (Jijkl) exp − Jijklφijkl + Jijklφklij (B4) where we deﬁned a short-hand notation combining the Hermitian conjugate terms into a single permutation

∗ Y ∗ † Y Y † D[J, J ] ≡ dJij;kldJij;kl , D[c , c] ≡ DciaDcia. (B5) i

Here φijkl are four-fermion terms

Z ! X † † † † † † † † † † φijkl = dτ 4 ciacjackacla + 2α ci1cj2ck1cl2 + ci2cj1ck1cl2 + ci1cj2ck2cl1 + ci2cj1ck2cl1 (B6) a

2 |Jijkl| − 2 2 2 3 and P (Jijkl) = e σ2 is the complex Gaussian distribution with variance σ ≡ h|Jijkl| i = J /8N . Integrating over Gaussian random variables one gets the averaged expression(upto a multiplicative constant)

Z 1 2 2 ∗ 2 2 ∗ − 2 (Jijkl+σ φklij )(Jijkl+σ φklij ) σ φijklφklij σ φijklφklij dJijkldJijkle σ e ≡ e . (B7)

Expressing the averaged partition function with all possible permutations we ﬁnd   Z Y J 2 Z J 2 X Z ≡ D[c†, c] exp φ φ = D[c†, c] exp φ φ . (B8) int 16N 3 ijkl klij (4N)3 ijkl klij i

The reason is that, for the saddle-point solution of the SYK model, the replica off-diagonal terms can be ignored as they do not contribute to zeroth order in 1/N. Performing the replica trick with only replica-diagonal terms is formally equivalent to the quenched disorder average. Combining with the free part of the action, we finally obtain the averaged partition function for the fermions Z = R D[c†, c]e−S with the effective action

 N  Z X X J 2 X S = dτdτ 0 c† (τ 0)((∂ − mu)δ + iκ )δ(τ − τ 0)c (τ) − φ φ . (B10)  ia τ a,b a,b ib (4N)3 ijkl klij i=1 a,b i,j,k,l

0 1 PN 0 † We now integrate out fermions by introducing the averaged Green’s functions Gba(τ , τ) = N i=1 hT cib(τ )cia(τ)i through the identity

 " N # Z Z X 1 X 1 ∼ DΣ exp N dτdτ 0 Σ (τ, τ 0) G (τ 0, τ) − c (τ 0)c† (τ) (B11)  ab ba N ib ia  a,b i=1

0 where the Lagrange multipliers Σab(τ, τ ) play the role of fermionic self-energies. After integrating out fermions, the eﬀective (G, Σ) action for the averaged Green’s functions and self-energies becomes

1 Z n X J 2 X − S [G, Σ] = ln Det((∂ − µ)δ + iκ − Σ ) + dτdτ 0 Σ (τ, τ 0)G (τ 0, τ) + G (τ, τ 0)2G (τ 0, τ)2 N τ ab ab ab ab ba 4 ab ba a,b a,b 0 0 0 0 0 0 0 0 + 2α G11(τ, τ )G11(τ , τ) + G22(τ, τ )G22(τ , τ) G12(τ , τ)G21(τ, τ ) + G21(τ , τ)G12(τ, τ ) 2 0 0 0 0 0 0 0 0 + α G11(τ , τ)G22(τ , τ)G11(τ, τ )G22(τ, τ ) + G12(τ , τ)G21(τ , τ)G11(τ, τ )G22(τ, τ ) 0 0 0 0 0 0 0 0 o + G11(τ , τ)G22(τ , τ)G21(τ, τ )G12(τ, τ ) + G12(τ , τ)G21(τ , τ)G21(τ, τ )G12(τ, τ ) (B12)

0 0 This expression can be simpliﬁed using time translation invariance Gab(τ, τ ) = Gab(τ − τ ) and the R symmetry transformation that sends c1 → c2 and c2 → −c1, and implies G11(τ) = G22(τ),G12(τ) = −G21(τ). This leads to 1 Z n − S [G, Σ] = ln Det((∂ − µ)δ + iκ − Σ ) + 2β dτ Σ (τ)G (−τ) + Σ (τ)G (−τ) N τ ab ab ab 11 11 12 21 J 2 α2 + 1 + G2 (τ)G2 (−τ) + G2 (τ)G2 (−τ) − 4αG (τ)G (−τ)G (τ)G (−τ) 4 2 11 11 12 12 11 11 12 12 α2 o − G2 (τ)G2 (−τ) + G2 (τ)G2 (−τ) , (B13) 2 11 12 12 11 with a factor of β coming from R dτ and a factor 2 coming from adding identical terms. The saddle point equations can be written using δS[G, Σ]/δΣ = 0 and δS[G, Σ]/δG = 0. It is convenient to take the functional derivative with respect to Σ in Fourier space using the convention

β 1 X Z f(τ) = e−iωnτ f(ω ), f(ω ) = dτeiωnτ f(τ) β n n ωn 0 with ωn = (2n + 1)π/β the Matsubara frequency. The action with only the Σ(ω) dependent terms thus reads

S X = ln Det(M) + 2 [Σ (ω )G (ω ) − Σ (ω )G (ω )] + ··· , (B14) N 11 n 11 n 12 n 12 n n where the dots represent terms not explicitly containing Σ(ωn) and ! M −iω − µ − Σ iκ − Σ M = n 11 12 . (B15) n −iκ + Σ12 −iωn − µ − Σ11

−1 Finally using δ ln Det(M(ωn))/δΣmn(ωn0 ) = 2(M )nmδ(ωn − ωn0 ) one obtains the saddle point equations (20) and (29) quoted in the main text. 19

Appendix C: Conformal solution for generic α

In this Appendix we examine whether a low-energy scale-invariant solution of the saddle-point equations Eqs. (20,29,30) for both correlators G11 and G12 is possible in the gapless phase. We find that this is only possible for the special cases where either κ = 0 or α = 1, but not for generic parameter choices. Our strategy is to assume a power-law ansatz for G11 and G12 and iterate through the saddle-point equations, in the low-energy limit where the term −iωn can be neglected, to check for consistency. We adopt a power-law ansatz with the same exponent ν for both correlators. This is necessary since the saddle-point equations involve a sum of squares of the two self energies Σab – thus a conformal solution with different exponents for the two correlators can never be self-consistent. Similar to the case of decoupled SYK models, we find that the exponent is enforced to be ν = −1/2. Using the imaginary time reflection symmetries valid at charge neutrality µ = 0, we thus write the ansatz

−1/2 −1/2 G11(τ) = b11|τ| sgn(τ) ,G12(τ) = ib12|τ| , (C1) where b11 and b12 are real numbers. Inserting (C1) into the saddle point equations the self-energies become α2 α2 Σ (τ) =J 2 (1 + )b3 − (2α − )b b2 |τ|−3/2 sgn(τ) = s |τ|−3/2 sgn(τ), 11 2 11 2 11 12 11 α2 α2 Σ (τ) = + iJ 2 (1 + )b3 − (2α − )b b2 |τ|−3/2 = is |τ|−3/2, (C2) 12 2 12 2 12 11 12 where we implicitly defined the real constants s11 and s12. Note that at the special point α = 1 we have a simple relation s11b12 = −s12b11. Our goal is now to determine the constants b11 and b12 using the other two saddle-point equations. At β = ∞, Fourier transforming the first two-point correlator Eq. (C1) we get Z ∞ Z ∞ r iωτ −1/2 −1/2 −1/2 π G11(iω > 0) = b11 dτe |τ| sgn(τ) = 2ib11 dτ sin(ωτ)τ = 2ib11 ω (C3) −∞ 0 2 For negative frequencies the result has an overall negative sign because of the sin(ωτ) function inside the integral. Repeating this calculation for G12 and the self energies in Eq. (C2) one obtains √ √ −1/2 −1/2 G11(iω) = i 2πb11 |ω| sgn(ω),G12(iω) = i 2πb12 |ω| (C4) √ √ 1/2 1/2 Σ11(iω) = 2 2πis11|ω| sgn(ω) , Σ12(iω) = −2 2πis12|ω| . (C5) For non-zero κ, the self-energies given in Eq. (C5) will be consistent with the saddle point equations if we perform a uniform shift and define the modified self-energy

Σ˜ 12(iω) = Σ12(iω) − iκ (C6) Here we consider for simplicity the µ = 0 limit where the original symmetry of the correlators about τ = 0 in Eq. (C1) remains intact. For non-zero µ, one can use a similar redeﬁnition of the self-energy Σ˜ 11(iω) = Σ11(iω) + µ to solve the saddle-point equations in the conformal limit. This choice leads to an asymmetry in G11(τ) about τ = 0 (see Eq. (22)) and, as shown in Ref. [2], to a twisted G11(z) with a µ-dependent phase factor in the complex frequency z plane. Note however that adding a constant frequency shift does not aﬀect the long-time conformal scaling of Σab(τ), as it translates to a delta function at early times. The Schwinger-Dyson equations now read

Σ˜ (iω) Σ˜ (iω) G (iω) = − 11 ,G (iω) = + 12 (C7) 11 ˜ 2 ˜ 2 12 ˜ 2 ˜ 2 Σ11(iω) + Σ12(iω) Σ11(iω) + Σ12(iω) which lead to s11 s12 2 2 = 4πb11 , 2 2 = −4πb12. (C8) s12 + s11 s12 + s11

We thus have two constraints for the two unknown scaling parameters b11 and b12 which should give us a solution for all values of α. However for generic α we ﬁnd that the only real solution has b12 = 0 and !1/4 1 b11 = (C9) 2 α2 4πJ (1 + 2 ) 20

This solution represents decoupled SYK models with no correlations (at the saddle-point level) between the two sides, where the only eﬀect of α was to renormalize the constant b11. This solution can thus only represent the κ = 0 limit of our model. This is indeed the numerically obtained solution for κ = 0 and 0 < α < 4, in the gapless phase. On the other hand, we do not ﬁnd a conformal solution for α < 0 or α > 4 as the system develops a gap through the symmetry breaking mechanism discussed in the main text. For non-zero κ we always obtain a non-vanishing correlator G12 which is inconsistent with the solution above. Thus a conformal solution cannot be found for generic points inside the gapless dome. However, at the SU(2) symmetric point α = 1 the equations above have additional structure. Using the relation s11b12 = −s12b11, it is clear that the two equations (C8) are now equivalent. We thus have an under-constrained system, which we can solve for

1 b4 = b4 − . (C10) 12 11 6πJ 2

As shown in Fig.9, this relation appears to be satisﬁed numerically for α = 1 and small κ. To ﬁx the value of b11, we notice that there is another constraint coming from the relation

hKi 2iG (τ = 0) = , (C11) 12 κN where K is the tunneling operator. For α = 1 the value of hKi is a good quantum number of the system, because [K,Hα=1] = 0. This is similar to the case of ﬁnite chemical potential µ, for which[2] 1 G (τ → 0+) = − Q, (C12) 11 2 where Q is the conserved U(1) charge of the system, and is related to the asymmetry parameter E appearing in the low-energy Green’s function, Eq. (22) in imaginary time [1,2]. It is possible to directly compute the value of Q from the microscopic theory [53], and relate it to the conformal scaling parameter E through Eq. (23). Similarly, here we have at zero temperature

hKi Z ∞ dω = iG12(τ → 0) = i G12(ω) (C13) 2κN −∞ 2π which provides the second constraint allowing to ﬁx b11 and b12. It is not clear if an analytical result can be obtained for this constraint, as the full form of G12(ω) at all energies is needed.

Appendix D: Saddle-point equations in real time and frequency

In this appendix we show how to analytically continue the imaginary-time saddle point equations (20) and (29-30), to real time and frequency. The basic scheme is the same as the one described in Refs. [4,8, 26, and 35], but we here summarize the essential steps and results to keep this work self-contained.

1. Analytic continuation of self-energy

First, let us note a generic U(1)-symmetry conserving self-energy such as the ones appearing in Eqs. (29-30) as

Σ(τ) = Ga(τ)Gb(τ)Gc(−τ) . (D1)

Here a, b, c are arbitrary labels, and prefactors like J 2 are omitted. We ﬁrst Fourier transform into imaginary frequency 1 X Σ(iω ) = G (iω )G (iω )G [i(ω + ω − ω )] . (D2) n β2 a n1 b n2 c n1 n2 n n1,n2

R ρx(ω) Using the spectral representation (Hilbert transform) of the Greens function, Gx(iωk) = dω , we obtain iωk−ω Z ∞ Σ(iωn) = dω1,2,3ρa(ω1)ρb(ω2)ρc(ω3) · Y (iωn; ω1, ω2, ω3) (D3) −∞ 21

where, noting the constraint ωn3 = ωn1 + ωn2 − ωn, we have 1 X 1 1 1 Y (iωn; ω1, ω2, ω3) = 2 . β (iωn1 − ω1) (iωn2 − ω2) (iωn3 − ω3) n1,n2

We now perform the Matsubara summations and use Bose- and Fermi-function identities to evaluate Y (iωn; ω1,2,3).

The n2 sum can be evaluated by deﬁning Ω = iωn − iωn1 + ω3 to obtain

1 X 1 −1 iω 0+ nF (Ω) − nF (ω2) nF (ω3) − nF (ω2) e n2 = = . (D4) β (iωn2 − ω2) (Ω − iωn2 ) Ω − ω2 iωn − iωn1 + ω3 − ω2 n2

In the last step we used the fact that the imaginary part of Ω is a multiple of 2πT , hence nF (Ω) = nF (ω3). To evaluate the n1 sum we deﬁne Ω˜ = iωn + ω3 − ω2 and get ˜ 1 X 1 1 iω 0+ nF (ω1) − nF (Ω) nF (ω1) + nB(ω3 − ω2) e n1 = = . (D5) ˜ ˜ β (iωn1 − ω1) (Ω − iωn ) Ω − ω1 iωn + ω3 − ω2 − ω1 n1 1

Here we used that Ω˜ is a fermionic Matsubara frequency, hence nF (Ω)˜ = −nB(ω3 − ω2). To get Y , we now take the product of the numerator in Eq. (D4) and the expression Eq. (D5). The product of both numerators simpliﬁes to

[nF (ω1) + nB(ω3 − ω2)][nF (ω3) − nF (ω2)] = nF (ω1)nF (ω3) − nF (ω1)nF (ω2) − nF (−ω2)nF (ω3)

= −nF (ω1)nF (ω2)nF (−ω3) − nF (−ω1)nF (−ω2)nF (ω3) .

In the last step we inserted identities 1 = nF (ωj) + nF (−ωj) to obtain a more symmetric expression. Finally we get

nF (ω1)nF (ω2)nF (−ω3) + nF (−ω1)nF (−ω2)nF (ω3) Y (iωn; ω1, ω2, ω3) = − . (D6) iωn − ω1 − ω2 + ω3

Note the frequency-symmetric form ωj → −ωj of the numerator. Inspecting Eq. (D6) and the self-energy in Eq. (D3), the remaining imaginary frequency iωn now appears only in the denominator of Y . We hence can analytically continue iωn → ω + iη to obtain the retarded self-energy from Eq. (D3) with Y (ω + iη; ω1, ω2, ω3) given in Eq. (D6). We then 1 R ∞ i(Ω+¯ iη)t ¯ use the identity Ω+¯ iη = −i 0 dte with Ω = ω − ω1 − ω2 + ω3 to obtain Z ∞ Z ∞ ret i(ω+iη−ω1−ω2+ω3)t Σ (ω) = i dt dω1,2,3e ρa(ω1)ρb(ω2)ρc(ω3)[nF (ω1)nF (ω2)nF (−ω3) + (ωj ↔ −ωj)] . (D7) 0 −∞ This allows us to perform the three frequency integrals, and ﬁnally we obtain Z ∞ ret i(ω+iη)t ++ ++ −+ +− +− −− Σ (ω) = i dte na nb nc + na nb nc . (D8) 0 Here we deﬁned the “time-dependent occupations” Z ∞ ss0 0 −iωt nx (t) = dωρx(sω)nF (s ω)e , (D9) −∞ that can be calculated directly from the spectral function, and hence from the retarded Greens functions. As we will note below, analytically continuing the Greens functions is essentially trivial, and hence the expressions (D8-D9) are convenient for the numerical solution of the saddle-point equations in real time and frequency [4, 26, 35].

2. Application to the coupled complex SYK model

ret The retarded self-energies Σ11,12(ω) are obtained from the analytical continuation of Eqs. (29-30), according to the recipe outlined above. It is useful to simplify these expressions further by taking some of the observed spectral ret 2 R ∞ i(ω+iη)t symmetries into account. The general form reads Σx=11,12(ω) = −iJ 0 dte Kx(t) with 1 K (t) = (1 + α2) [(n++)2n−+ + (n+−)2n−−] − 2α [n++n++n−+ + n+−n+−n−−] (D10) 11 2 11 11 11 11 11 12 12 11 12 12 1 − α2 [(n++)2n−+ + (n+−)2n−−] 2 12 11 12 11 22 and 1 −K (t) = (1 + α2) [(n++)2n−+ + (n+−)2n−−] − 2α [n++n++n−+ + n+−n+−n−−] (D11) 12 2 12 12 12 12 12 11 11 12 11 11 1 − α2 [(n++)2n−+ + (n+−)2n−−] 2 11 12 11 12

Note that, up to an overall minus sign, K11(t) and K12(t) are directly related by replacing 11 ↔ 12 everywhere. To make further progress, note the simple form of the non-interacting retarded Green’s functions

−1 −1 [g11(ω)] = ω − µ + iη , [g12(ω)] = −iκ . (D12)

ret ret Following the convention for spectral functions in Ref. [27], and using G21 (ω) = −G12 (ω), we obtain 1 i ρ (ω) = − ImGret(ω), ρ (ω) = ReGret(ω). (D13) 11 π 11 12 π 12

In plots in the main text, cf. Figs.5 and 11, when referring to ρ12 we implicitly consider Im[ρ12(ω)] with the above deﬁnition. One can simplify the above expressions for K11,12 by using properties of the spectral functions ρ11,12. First ρ11 (ρ12) is purely real (imaginary), and for zero chemical potential µ = 0 they also have a deﬁnite frequency parity: ∗ ρ11(ω) = [ρ11(ω)] , ρ11(ω) = ρ11(−ω) , ∗ ρ12(ω) = −[ρ12(ω)] , ρ12(ω) = −ρ12(−ω) . ±± ±± Using these properties, one can express all n11 and n12 by a single n11 and n12. We note

++ ω −+ Re +− ∗ ω −− ∗ n11 ≡ n11 = +n11 = +[n11 ] = +[n11 ] , (D14) ++ ω −+ Im +− ∗ ω −− ∗ n12 ≡ n12 = −n12 = +[n12 ] = −[n12 ] , (D15) where at ω we used the frequency parity, and Re/Im means we used the real/complex-valuedness of ρ11/12. Then

2 3 2 2 K11(t) = (2 + α )Re[n11] + (4α − α )Re[n11n12] , (D16) 2 3 2 2 K12(t) = (2 + α )Re[n12] + (4α − α )Re[n12n11] . (D17) This simplified version makes apparent the symmetry of self-energies under exchange 11 ↔ 12. It also suggests that there is a second ‘decoupling point” at α = 4, similar to the situation at α = 0. However the SYK interaction strength at this point is enhanced to an effective Jα=4 = 3J. We further discuss this around Eq. (31) and in Fig.6. The above SD equations are solved numerically by repeated self-energy evaluations and re-insertion into the Dyson equation (20) until a fixed point solution is found. As the starting point we use g11(ω) and g12(ω) in Eq. (D12) with 2/3 the initial value κ0 = 0, κ or κ . We then check that the same solution is obtained independent of the starting point and the iteration parameters.