Boosting quantum efficiency by reducing complexity

Giovanni Sisorio,1 Alberto Cappellaro,1, 2, ∗ and Luca Dell’Anna1, 2

1
Dipartimento di Fisica ed Astronomia "G. Galilei",
Università degli Studi di Padova, via Marzolo 8, 35131 Padova, Italy
2
National Institute of Nuclear Physics (INFN), Padova Section, Via Marzolo 8, 35131 Padova, Italy
(Dated: May 26, 2025)
In the context of energy storage at the nanoscale, exploring the notion of quantum advantage
implies walking on the thin line at the boundary between quantum mechanics and thermodynamics,
which underpins our conventional understanding of battery devices. With no classical analogue, the
Sachdev-Ye-Kitaev (SYK) model has emerged in the last years as a promising platform to boost
charging and storage efficiency thanks to its highly-entangling dynamics. Here, we explore how
the robustness of this setup by considering the sparse version of the SYK model, showing that, as
long as chaos is not completely broken, reducing its complexity may lead to more efficient quantum
arXiv:2505.17679v1 [quant-ph] 23 May 2025

batteries.

Introduction and motivations. The quest for efficient becomes significantly costly [30], with exact diagonaliza-
energy storage and work extraction has been a corner- tion seemingly limited to few tens of fermions due to the
stone of several technological breakthroughs, dating back rapid scaling of Hilbert space size [31–33].
at least to the first industrial revolution, with its pro- Here, we investigate the sparse version of the SYK
found connection to the development of thermodynam- model, where some interaction terms are removed with a
ics as a backbone for our scientific understanding of Na- certain probability [34–36]. By pruning a fraction of the
ture [1, 2]. Nowadays, as material samples are brought couplings, quantum chaotic features - instrumental for ef-
to increasingly lower temperature and smaller sizes, we ficient energy storage - are retained while mitigating its
are experiencing a sort of nanoscale industrial revolution, computational and experimental overhead. Even more
where the challenge of energy storage takes on a quan- interestingly, our results demonstrate that such sparsi-
tum mechanical dimension, where centuries-old thermo- fication can also enhance the battery performance, pro-
dynamical concepts are in need to be revisited in the face vided the system remains sufficiently chaotic.
of non-classical resources such as quantum coherence and The sparse SYK model. As mentioned in the intro-
entanglement [3–5]. Indeed, the interplay between entan- duction, the SYK model presented itself as a theory for
glement generation and work extraction from many-body strongly coupled fermions with random (all-to-all) inter-
quantum states has spurred a sustained research effort actions [17–19]. Now, up to the last years, most of the
into quantum batteries [6–14]. existing efforts in elucidating the peculiar properties of
It is now clear that quantum battery engineering hinges this model focused on the case with Majorana fermions.
on balancing complexity and controllability. For in- Here, on the other hand, our starting point is the com-
stance, while being very sensitive to noise and fluctua- plex version of the SYK (cSYK) model [20, 21] for N
tions, the onset of a highly entangling dynamics can pos- spinless fermions, i.e.
itively affect the battery performance, especially when it N
comes to figures of merit such the amount of injected en-
Jijkl ĉ†i ĉ†j ĉk ĉl
X
ĤcSYK = (1)
ergy or the work extraction power [14–16]. This tension
i,j,k,l=1
motivates the exploration of minimal platforms where ac-
cess to such observables is amenable to analytical and
with the usual anticommutation relations {ĉi , ĉ†j } = δij
computational analysis. One of the most interesting can- (†) (†)
didates is certainly the Sachdev-Ye-Kitaev (SYK) model, and {ĉi , ĉj } = 0, while the complex couplings satisfy
∗
a paradigmatic example of a strongly interacting quan- Jijkl = Jklij and Jijkl = −Jjikl = −Jijlk , such that
tum system describing all-to-all randomly interacting Eq. (1) is hermitian and anti-symmetrized. These cou-
fermions [17–21]. This peculiar structure supports the plings are distributed according to a zero-mean Gaussian
onset of maximal chaos, such that the system qualifies density function P(Jijkl ), with variance Jijkl2 = J 2 /N 3 ,
as a fast scrambler, leading to a very efficient charging Rthe average over disorder being defined as O(Jijkl ) =
and stable energy storage [22–24], but also to significant D[Jijkl ]P(Jijkl )O(Jijkl ). The sparse version of Eq. (1)
experimental and computational challenges. The former is simply implemented by introducing, for each coupling,
are a direct consequence of the all-to-all character of SYK an additional random variable xijkl being equal to 1 with
couplings, ending up in significant practical bottleneck, probability p and 0 otherwise, such that the coupling is
like the system’s size [25–29]. In addition, all-to-all in- retained with probability p or pruned (i.e. removed) with
teractions leads to volume-law entangled ground states, probability 1−p [34, 35]. Therefore, for a q-body interac-
such that even powerful techniques such tensor networks tion (in the case of Eq. (1), q = 4), the pruning procedure
 2

a general bound [38] on the Lyapunov exponent λL which

controls the decay of out-of-time four-point correlators as
Â(t)B̂(0)Â(t)B̂(0) th ∼ 1 − α exp(λL ts ), with  and B̂
some system’s operators, ⟨•⟩th the thermal average and
ts the so-called scrambling time. The onset of maximal
chaos implies that information is scrambled at the fastest
rate enabled by quantum mechanics and it is crucial for
an efficient charging protocol [22–24]. It is then natural
to investigate the relation between sparsity and quantum
chaos, especially when one has to leverage the latter to
implement an efficient setup. Technically, the quantum
chaotical character of the SYK model reflects on specific
features of its energy spectrum, which can be analyzed
by means of tools from random matrix theory (RMT)
[36, 39, 40]. Here, we first examine the behavior of the
nearest-neighbor gap ratio, which is defined as
  
si si+1
r = min , , (3)
si+1 si
Figure 1. Top panel. Pictorial representation of the fully- with si = Ei − Ei−1 the spacing between adjacent eigen-
connected SYK model, as opposed to its sparse version. In the values. The gap ration quantifies the repulsion between
fully-connected version, we are considering a set of N all-to-
all randomly interacting (complex) fermions. Once we move
energy levels at scales of the mean level spacing. Con-
to the sparse version, couplings are retained with probability sistently with predictions from RMT, r remains approx-
p (the so-called sparsity parameter, which is unitary in the imately constant for large p, while a sharp drop is ob-
fully-connected case), and otherwise removed. Also notice served as soon as p approaches a critical value of sparsity
that, in principle, the complex SYK model defined in Eq. (1) usually labelled p2 , signaling the dramatic breakdown of
accounts for spinless fermions, with the mapping to two-level spectral rigidity.
systems (i.e. the battery elementary cells) enabled by the
Jordan-Wigner transformation defined in Eq. (2). Bottom
panel. Charging protocol of a quantum battery in a nutshell,
as described in the main text (cfr. Eq. (5) and discussion
thereafter). After preparing our set of N two-level systems in
its ground state Ψ0 , we turn on the charging Hamiltonian
ĤcSYK at a certain value of the sparsity parameter, and we
let the dynamics unfolds up to τc . Accessing the evolved
state Ψ(τc ) is paramount to unveil the quantum battery
performance, through properly defined figures of merit.

reduces the number of interacting terms Nq by a factor

p. In order to ensure comparability with the energy scale

2
of the original model (i.e. p = 1), Jijkl 2 /p.
−→ Jijkl
Now, the complex SYK can be mapped onto a spin-
1/2 model by means of a Jordan-Wigner transformation,
providing us with a more convenient numerical platform Figure 2. Nearest-neighbor eigenvalue gap ratio r (cfr. Eq.
for our later purposed [23, 37]; by recalling σ̂i± = (σ̂ix ± (3)) as a function of the sparsity parameter p for N = 6
iσ̂iy )/2, the transformation is defined by (cyan, squared dots), N = 8 (red, diamond dots), and N = 10
(green, round dots), with J = 1. For large p, r remains ap-
i−1  i−1 
† +
Y
z
Y
z
proximately constant, consistent with the RMT understand-
ĉi = σ̂i σ̂m and ĉi = σm σ̂i . (2) ing. As p decreases, a sharp decline in r indicates the break-
m=1 m=1 down of spectral rigidity and the transition to non-universal
statistics. The critical sparsity values p2 are highlighted in
As noticed in [24], this transformation make the highly
darker shade: p2 = 0.1263 for N = 6, p2 = 0.0279 for N = 8,
nonlocal character of the cSYK more transparent. As we and p2 = 0.0097 for N = 10. Statistical errors originat-
aim to make clear later, this feature is particularly rele- ing from the disorder average are included, but they end up
vant when we compare the sparse cSYK to other battery smaller than the marker dots.
implementations, such as, for instance, the Dicke-based
one [8]. Even in its complex version, it is important to re- In Fig. 2 we report the gap ration r as in Eq. (3) as
call that the SYK model is maximally chaotic, saturating a function of the sparsity parameter p and for three dif-
 3

ferent values of N , clearly displaying the behavior we’ve

just outlined. From an operational perspective, we de-
fine the critical sparsity value p2 as the value of p where
r drops below 99% of its value in the fully connected (un-
sparsified) case [36]. At this threshold, the characteristic
ramp in the spectral statistics vanishes, marking a qual-
itative change in the eigenvalue dynamics. The spectral

Figure 4. Top panel. Population dynamics pk (t) as defined in

Eq. (7) as a function of time and eigenvalue index k; the plot
has been obtained for N = 10, J = 2 and sparsity parameter
p = 0.5. We have considered here just a single realization
of couplings’ disorder. Also notice that we have shifted the
spectrum upward, such that the k = −N eigenvalue is labeled
by 0 in the vertical axis. The peak displayed in the bottom-left
corner correspond the initial preparation in the ground state
of Ĥ0 (cfr. Eq. (5)), representing the discharged battery.

Figure 3. Spectral form factor (SFF) as defined in Eq. (4) vs.

time for the SYK model at different sparsity parameters p for lowest eigenstate of Ĥ0 ) up to t → 0− . Then, at t = 0 the
N = 8. The typical markers of quantum chaos are diluted as charging Hamiltonian is turned on, realizing the following
p decrease (i.e. we move towards to a more sparse scenario), protocol
until the dip-ramp is completely suppressed around p ≃ 0.02.
ĤB (t) = Ĥ0 + λ(t)Ĥ1 (5)
form factor (SFF) provides a robust measure of spectral where λ(t) = 1 when 0 ≤ t < τc , and 0 elsewhere, with τc
rigidity and it is commonly defined in terms of the ana- being the duration of the charging protocol. Now, within
lytically continued partition function Z(τ ), with τ ∈ C, our implementation, Ĥ0 is just an ensemble of N two-
as follows: level systems (or spin-1/2 particles, in the context of the
⟨Z(β + it)Z(β − it)⟩ JW transformation defined in Eq. (2)), while the sparse
SFF(β, t) ≡ , (4) cSYK is responsible for charging. With the convention
⟨Z(β)2 ⟩
ℏ = 1,
where β is the inverse temperature. As an exemplary N
case, we compute the SFF for N = 8, as reported in (sparse)
X
Ĥ0 = ω0 σiy and Ĥ1 = ĤcSYK (6)
Fig. 3. Evidently, the fully connected case p = 1 (solid i=1
blue line) displays the expected dip–ramp– plateau struc-
ture characteristic of quantum chaotic systems, with the where ω0 is the level splitting and σyi the usual Pauli
plateau arising from universal long-range spectral corre- matrix for the ith . Now, in order to characterize the
lations. As the sparsity increases (i.e. as p decreases), charging dynamics and its efficiency, an initial insight can
we observe a suppression of the ramp and a lifting of the be gained by looking at how energy levels are populated
plateau. This effect becomes more pronounced for lower throught the protocol (i.e. for t ∈ [0, τc ]). This is indeed
values of the sparsity parameter, such as p = 0.02 for provided by the overlap
N = 8 as displayed in Fig. 3 (green dashed-dotted line). X 2
pk (t) = ⟨k, i Ψ(t)⟩ (7)
Technically speaking, the SFF in Fig. 3 is computed in
i
the high-temperature limit, but the quantum chaotical
features encoded in the energy spectrum are expected to where { k, i⟩} are the eigenstates of Ĥ0 , the index i la-
persist across the whole temperature range [36, 41, 42] belling the degeneracy related to ϵk = kω0 , and Ψ(t)
Implementing the battery: The charging protocol. A is the state evolved according to Ĥ1 . Considering Ĥ0
quantum battery can be simply intended as a collection in Eq. (6), the initial state is reasonable prepared as
NN
of N quantum cells [6, 7, 14], each acting as an individual
PN Ψ0 = i=1 ↓(y) i .
unit of energy storage, such that Ĥ0 = i=1 ĥi . Here, In Fig. 4 we report pk (t) as defined in Eq. (7) for
we focus on the charging protocol, thereby assuming the N = 10, J = 2, p = 0.5 and single realization of cou-
battery is initially prepared in its discharged state (the plings’ disorder. As first noticed in [23] regarding the
 4

fully connected cSYK model (i.e., the p = 1 case), such a

charging protocol induces a peculiar non-local dynamics
in energy space. Indeed, as soon as the dynamics unfolds
for t > 0, excited states are populated almost immedi-
ately, singling out a band centered around k ≃ 8 persist-
ing at long times. This behaviour is intimately related
to the peculiar properties displayed by the SYK model,
such as the absence of quasiparticle excitations and the
onset of quantum chaos, reading local thermalization at
the fastest timescale of ∼ ℏ/(kB T ) as T → 0 [43] (T being
the final state temperature), where after-quench evolved
states can be locally described in terms of thermal ones
[32, 44]. This is particularly transparent when we con-
sider alternative platforms for battery implementation,
such as the one based on the Dicke model [8, 45, 46],
Figure 6. Battery efficiency, as defined in Eq. (9), with
where the population dynamics induced by the charging
respect to different values of the sparsity parameter p, for
protocol is markedly local, with evident periodic revivals, three different system’s size. Here, we consider ω = 1, J = 1
a feature tracing back to the model integrability [47, 48]. and the disorder average is performed over Ndis = 103 for
N = 6, Ndis = 5 · 102 for N = 8 and Ndis = 150 for N = 10.
Statistical uncertainties arising from the disorder average are
smaller than marker dots. In the fully connected regime (p =
1), it is already transparent how smaller batteries provides a
more efficient avenue for work extraction. In the intermediate
regime, moderate sparsity slightly improves performance by
reducing interference. As p approaches the critical value p2
(dashed vertical lines), the breakdown of spectral rigidity and
loss of chaotic features lead to a sharp decline in e.

from an all-to-all interacting implementation. A first hint

supporting this understanding is again reported in Fig.
5, where we see EN (τc ) saturating approximately p ∼ p2 ,
Figure 5. Stored energy (cfr. Eq. (8)) as a function of τc , a confirmation that the battery performance are predi-
the duration of the charging protocol, and normalized to the cated on retaining at least some degree of its quantum-
number of battery cells. Here we consider N = 8, J = 1 and
chaotical behavior in the fully connected (p = 1) case. A
four different values of the sparsity parameter. Additionally,
EN (τc )/N is also averaged over Ndis = 103 different disor- more complete insight into this trade-off can be acquired
der realization of the model’s couplings {Jijkl }. The fully- by investigating a different figure of merit, ideally en-
connected cSYK (p = 1) is reported as a solid black, while conding additional information concerning the work ex-
dash-dotted green line follows the case of p2 ≃ 0.0279 (for traction at later times t > τc (i.e. after the charging
N = 8). protocol is turned off). This is the case for the so-called
battery efficiency [11], defined as the ratio between the
Now, in order to assess the performance of a battery, amount of extractable work and the energy stored in the
throughout the charging process, it is worth examining battery during the charging process. Technically, this is
specific figures of merit with particular attention to the computed as
role played by the sparsity parameter. In Fig. 5, we plot
EN/2 (τc )
 
the stored energy, given by
e(τc ) = (9)
(0) EN/2 (τc )
EN (τc ) = Ψ(τc ) Ĥ0 Ψ(τc ) − EN (8)
as a function of τc , i.e. the duration of the charging pro- with EN/2 (τc ) the stored energy as in Eq. (8) and
tocol, and normalized to the number of cells. The second EN/2 (τc ) the maximum extractable work, also called er-
term in the equation above is simply the energy of the gotropy [14, 49–51]. The latter is defined by
(0)
initial state, namely EN = Ψ0 |Ĥ0 |Ψ0 . Remarkably,
   
E(t) ≡ E(ρ(t)) = Tr Ĥ0 ρ(t) − Tr Ĥ0 σρ , (10)
by lowering p (i.e. by increasing sparsity) we see a signif-
icant increase in the energy we are able to stored in the with σρ being a completely passive passive state (ther-
battery. This seemingly points to a trade-off between the mal states are good candidates [52]), and ρ(t) the state
onset of quantum chaos (or rather the ability of the sys- evolved according to the charging Hamiltonian, cfr. Eq.
tem to be a fast scrambler) and the complexity emerging (5). Also notice that in Eq. (9) we are considering the
 5

stored energy (and the corresponding extractable work) magnetic field, by considering the electrons in the lowest
in half of the battery. Assuming that Ĥ0 can be recast Landau level [26]. Here, the bottleneck is represented by
as a sum of local terms (this is evident from our setup the magnetic field strength, up to B ∼ 3 · 103 T for a
Eq. (6)), we just need to restrict Ĥ0 in defining E(t) flake with 5 nm radius (2 · 103 carbon atoms mapping to
to the desired subset of elementary cells (i.e. we stop the N ≲ 20 scenario). While larger flakes enables more
the sum at N/2). Technically, this restriction reflects realistic conditions [53], implementing a sparse version
the reasonable assumption that, in a realistic experimen- represents an appealing alternative. This is even more
tal platform, only a subset of the total number of qubits compelling when considering the recent advances with
can be accessed. In Fig. 6 we report our numerical re- novel platforms based on cavity quantum electrodynam-
sults concerning the battery efficiency as defined in Eq. ical simulations [28, 29]. Starting from a single-mode
(9) for different values of the sparsity parameter, with optical cavity filled with quasi-2D 6 Li atoms, through
different comments now in order. First of all, even in fast cycling and engineered time-dependent disorder, it
the fully connected case (p = 1) increasing the battery is possible to simulate the evolution of a random quan-
size worsens its efficiency, a feature persisting even for tum circuit through discrete steps, converging to the de-
different sparse realizations. This is in line with the sem- sired model. Remarkably, this hybrid digital-analog tech-
inal observation made in [24], where it was first shown nique (resembling conventional Trotterization) is able
how energy extraction from large SYK batteries is less to target sparse version of the desired target model
favourable than dealing firsthand with smaller batteries. (the full SYK, in this case) [29]. In order to estimate
Now, as sparsity gradually increases some interactions sparseness, one can compute the statistical distance from
are removed, thus reducing complexity and, at the same the fully connected model via the so-called Kullback-
time, crucially retaining the quantum chaotic character Leibler divergenceR(or relative Shannon entropy), namely
typical of the fully-connected regime. We discussed this DKL (J ||Jsp ) = D[J ]P(J ) log(P(J )/Q(Jsp )) where
previously, in relation to important markers of quantum P(J = Jijkl ) is the distribution of the fully-connected
chaos such as the nearest-neighbour ratio and the SFF model while Q(Jsp ) described the sparse one.
dynamics (cfr. respectively Fig. 2 and 3). Here the
crucial point is that, in this intermediate regime of re- Conclusions. Our work underscores the potential
duced complexity, the battery appears to operate more of sparse quantum systems as a viable path toward the
efficiently likely because of the reduced interference be- goal of implementing robust quantum batteries, offering a
tween the interaction terms. This is evident by looking, blueprint for harnessing quantum resources without suc-
for instance, at the N = 10 case (bottom panel in Fig. cumbing to impractical complexity. More specifically, we
6) where, by approaching p2 from above, efficiency can have first shown that, starting from the fully-connected
be boosted by ∼ 10%. Remarkably, while bigger cSYK SYK model, quantum chaotical features persists up to a
batteries are in principle less convenient for energy ex- critical value (p2 ) of the sparsity parameter, a threshold
traction, at the same time they seem to be more posi- where spectral rigidity breaks down. The stored energy
tively affected by an increased sparsity, when compared saturates at this value and, more remarkably, moving
to smaller implementations (N = 6 and N = 8 in Fig. closer to p2 from above (cfr. Fig. 6) positively affects
6). Finally, as we approach the critical value p2 , spec- the battery performance. In the end, this approach also
tral rigidity breaks down and eigenvalue statistic deviate sheds light on the fundamental interplay between com-
markedly from RMT predictions, signaling the loss of plexity, chaos and energy scale down at the nanoscale.
quantum chaos and, consequently, the system’s ability to
efficiently scramble information. Therefore, the battery Acknowledgements. The authors acknowledge finan-
performance severely deteriorates in the regime p < p2 , cial support from the Project PARD 2024 “Role of disor-
as confirmed by our numerical results in Fig. 6. der in work extraction and energy storage for Sachdev-
Experimental considerations. It is certainly relevant Ye-Kitaev quantum batteries” within the Project ”Fron-
that all the crucial features of the fully-connected SYK tiere Quantistiche” (Dipartimenti di Eccellenza) of the
model persist in its sparse version, especially (at least Italian Ministry for Universities and Research, and from
for our goals) when it comes to the onset of maximal the European Union-Next Generation EU within the “Na-
chaos. Indeed, implementing a sparse model may enable tional Center for HPC, Big Data and Quantum Comput-
experimentalists to overcome significant bottlenecks dis- ing” (Project No. CN00000013, CN1 Spoke 10 - Quan-
played by the full version. Even within the framework tum Computing). The authors thank F. Campaioli for
of ultracold atomic gases, where almost perfect isola- stimulating discussion.
tion from the external environment and control on many
physical parameters are at hand, implementing the fully- Code availability. All the codes employed in this
connected SYK would require 4 · 103 lasers just for the paper are available upon reasonable request. They have
N = 16 fermions [25]. A similar N is achieved by consid- been developed thanks to the open-source Python frame-
ering mesoscopic graphene flakes at strong disorder and work QuTiP [54–56].
 6

phy (part 2), https://online.kitp.ucsb.edu/online/

entangled15/kitaev2/ (2015), talk at KITP, University
of California, Santa Barbara.
∗
[email protected] [20] S. Sachdev, Bekenstein-hawking entropy and strange
[1] A. Saggion, R. Faraldo, and M. Pierno, Thermodynamics: metals, Phys. Rev. X 5, 041025 (2015).
Fundamental Principles and Applications, UNITEXT for [21] Y. Gu, A. Kitaev, S. Sachdev, and G. Tarnopolsky, Notes
Physics (Springer International Publishing, 2019). on the complex sachdev-ye-kitaev model, Journal of High
[2] H. S. Leff and A. F. Rex, eds., Maxwell’s Demon: En- Energy Physics 2020, 10.1007/jhep02(2020)157 (2020).
tropy, Information, Computing (Princeton University [22] R. Belyansky, P. Bienias, Y. A. Kharkov, A. V. Gorshkov,
Press, 1990). and B. Swingle, Minimal model for fast scrambling, Phys-
[3] N. Halpern, Quantum Steampunk: The Physics of Yester- ical Review Letters 125, 10.1103/physrevlett.125.130601
day’s Tomorrow (Johns Hopkins University Press, 2022). (2020).
[4] G. Kurizki and A. Kofman, Thermodynamics and Control [23] D. Rossini, G. M. Andolina, D. Rosa, M. Carrega, and
of Open Quantum Systems (Cambridge University Press, M. Polini, Quantum advantage in the charging process of
2022). sachdev-ye-kitaev batteries, Phys. Rev. Lett. 125, 236402
[5] P. Strasberg, Quantum Stochastic Thermodynamics: (2020).
Foundations and Selected Applications, Oxford Graduate [24] D. Rosa, D. Rossini, G. M. Andolina, M. Polini, and
Texts (Oxford University Press, 2024). M. Carrega, Ultra-stable charging of fast-scrambling
[6] R. Alicki and M. Fannes, Entanglement boost for ex- syk quantum batteries, Journal of High Energy Physics
tractable work from ensembles of quantum batteries, 2020, 10.1007/jhep11(2020)067 (2020).
Phys. Rev. E 87, 042123 (2013). [25] I. Danshita, M. Hanada, and M. Tezuka, Creating and
[7] F. C. Binder, S. Vinjanampathy, K. Modi, and J. Goold, probing the sachdev–ye–kitaev model with ultracold
Quantacell: powerful charging of quantum batteries, New gases: Towards experimental studies of quantum gravity,
Journal of Physics 17, 075015 (2015). Progress of Theoretical and Experimental Physics 2017,
[8] D. Ferraro, M. Campisi, G. M. Andolina, V. Pellegrini, 10.1093/ptep/ptx108 (2017).
and M. Polini, High-power collective charging of a solid- [26] A. Chen, R. Ilan, F. de Juan, D. I. Pikulin, and M. Franz,
state quantum battery, Phys. Rev. Lett. 120, 117702 Quantum holography in a graphene flake with an irregu-
(2018). lar boundary, Phys. Rev. Lett. 121, 036403 (2018).
[9] G. M. Andolina, D. Farina, A. Mari, V. Pellegrini, [27] O. Can, E. M. Nica, and M. Franz, Charge transport
V. Giovannetti, and M. Polini, Charger-mediated energy in graphene-based mesoscopic realizations of sachdev-ye-
transfer in exactly solvable models for quantum batteries, kitaev models, Phys. Rev. B 99, 045419 (2019).
Phys. Rev. B 98, 205423 (2018). [28] P. Uhrich, S. Bandyopadhyay, N. Sauerwein, J. Sonner,
[10] J. Liu, D. Segal, and G. Hanna, Loss-free excitonic quan- J.-P. Brantut, and P. Hauke, A cavity quantum elec-
tum battery, The Journal of Physical Chemistry C 123, trodynamics implementation of the sachdev–ye–kitaev
18303 (2019). model (2023), arXiv:2303.11343 [quant-ph].
[11] D. Rossini, G. M. Andolina, and M. Polini, Many-body [29] R. Baumgartner, P. Pelliconi, S. Bandyopadhyay, F. Orsi,
localized quantum batteries, Phys. Rev. B 100, 115142 N. Sauerwein, P. Hauke, J.-P. Brantut, and J. Sonner,
(2019). Quantum simulation of the sachdev-ye-kitaev model us-
[12] J. Q. Quach, K. E. McGhee, L. Ganzer, D. M. Rouse, ing time-dependent disorder in optical cavities (2024),
B. W. Lovett, E. M. Gauger, J. Keeling, G. Cerullo, D. G. arXiv:2411.17802 [quant-ph].
Lidzey, and T. Virgili, Superabsorption in an organic mi- [30] V. Bettaque and B. Swingle, Nora: A tensor network
crocavity: Toward a quantum battery, Science Advances ansatz for volume-law entangled equilibrium states of
8, eabk3160 (2022). highly connected hamiltonians, Quantum 8, 1362 (2024).
[13] B. Ahmadi, P. Mazurek, P. Horodecki, and S. Barzanjeh, [31] W. Fu and S. Sachdev, Numerical study of fermion and
Nonreciprocal quantum batteries, Phys. Rev. Lett. 132, boson models with infinite-range random interactions,
210402 (2024). Phys. Rev. B 94, 035135 (2016).
[14] F. Campaioli, S. Gherardini, J. Q. Quach, M. Polini, and [32] C. Liu, X. Chen, and L. Balents, Quantum entanglement
G. M. Andolina, Colloquium: Quantum batteries, Rev. of the sachdev-ye-kitaev models, Phys. Rev. B 97, 245126
Mod. Phys. 96, 031001 (2024). (2018).
[15] K. V. Hovhannisyan, M. Perarnau-Llobet, M. Huber, and [33] P. Zhang, Quantum entanglement in the sachdev-ye-
A. Acín, Entanglement generation is not necessary for kitaev model and its generalizations, Frontiers of Physics
optimal work extraction, Phys. Rev. Lett. 111, 240401 17, 43201 (2022).
(2013). [34] S. Xu, L. Susskind, Y. Su, and B. Swingle, A sparse model
[16] F. Caravelli, G. Coulter-De Wit, L. P. García-Pintos, and of quantum holography (2020), arXiv:2008.02303 [cond-
A. Hamma, Random quantum batteries, Phys. Rev. Res. mat.str-el].
2, 023095 (2020). [35] A. M. García-García, Y. Jia, D. Rosa, and J. J.
[17] S. Sachdev and J. Ye, Gapless spin-fluid ground state in Verbaarschot, Sparse sachdev-ye-kitaev model, quan-
a random quantum heisenberg magnet, Phys. Rev. Lett. tum chaos, and gravity duals, Physical Review D 103,
70, 3339 (1993). 10.1103/physrevd.103.106002 (2021).
[18] A. Kitaev, A simple model of quantum hologra- [36] P. Orman, H. Gharibyan, and J. Preskill, Quantum chaos
phy (part 1), https://online.kitp.ucsb.edu/online/ in the sparse syk model, Journal of High Energy Physics
entangled15/kitaev/ (2015), talk at KITP, University 2025 (2025).
of California, Santa Barbara]. [37] L. García-Álvarez, I. L. Egusquiza, L. Lamata, A. del
[19] A. Kitaev, A simple model of quantum hologra- Campo, J. Sonner, and E. Solano, Digital quantum sim-
 7

ulation of minimal AdS/CFT, Phys. Rev. Lett. 119, librum dynamics in the strongly excited inhomogeneous
040501 (2017). dicke model, Phys. Rev. B 86, 195101 (2012).
[38] J. Maldacena, S. H. Shenker, and D. Stanford, A [49] A. E. Allahverdyan, R. Balian, and T. M. Nieuwenhuizen,
bound on chaos, Journal of High Energy Physics 2016, Maximal work extraction from finite quantum systems,
10.1007/jhep08(2016)106 (2016). Europhysics Letters 67, 565 (2004).
[39] T. Guhr, A. Müller–Groeling, and H. A. Weidenmüller, [50] S. Tirone, R. Salvia, and V. Giovannetti, Quan-
Random-matrix theories in quantum physics: common tum energy lines and the optimal output ergotropy
concepts, Physics Reports 299, 189–425 (1998). problem, Physical Review Letters 127, 10.1103/phys-
[40] L. D’Alessio, Y. Kafri, A. Polkovnikov, and M. Rigol, revlett.127.210601 (2021).
From quantum chaos and eigenstate thermalization to [51] A. Touil, B. Çakmak, and S. Deffner, Ergotropy from
statistical mechanics and thermodynamics, Advances in quantum and classical correlations, Journal of Physics
Physics 65, 239–362 (2016). A: Mathematical and Theoretical 55, 025301 (2021).
[41] J. Sonner and M. Vielma, Eigenstate thermalization in [52] A. Lenard, Thermodynamical proof of the Gibbs formula
the sachdev-ye-kitaev model, Journal of High Energy for elementary quantum systems, J. Statist. Phys. 19, 575
Physics 2017, 10.1007/jhep11(2017)149 (2017). (1978).
[42] P. Saad, S. H. Shenker, and D. Stanford, A semiclassical [53] M. Brzezińska, Y. Guan, O. V. Yazyev, S. Sachdev, and
ramp in syk and in gravity (2019), arXiv:1806.06840 [hep- A. Kruchkov, Engineering syk interactions in disordered
th]. graphene flakes under realistic experimental conditions,
[43] S. A. Hartnoll, A. Lucas, and S. Sachdev, Holographic Phys. Rev. Lett. 131, 036503 (2023).
quantum matter (2018), arXiv:1612.07324 [hep-th]. [54] J. Johansson, P. Nation, and F. Nori, Qutip: An open-
[44] A. Eberlein, V. Kasper, S. Sachdev, and J. Steinberg, source python framework for the dynamics of open quan-
Quantum quench of the sachdev-ye-kitaev model, Phys. tum systems, Computer Physics Communications 183,
Rev. B 96, 205123 (2017). 1760 (2012).
[45] R. H. Dicke, Coherence in spontaneous radiation pro- [55] J. Johansson, P. Nation, and F. Nori, Qutip 2: A python
cesses, Phys. Rev. 93, 99 (1954). framework for the dynamics of open quantum systems,
[46] P. Kirton, M. M. Roses, J. Keeling, and E. G. Dalla Torre, Computer Physics Communications 184, 1234 (2013).
Introduction to the dicke model: From equilibrium [56] N. Lambert, E. Giguère, P. Menczel, B. Li, P. Hopf,
to nonequilibrium, and vice versa, Advanced Quantum G. Suárez, M. Gali, J. Lishman, R. Gadhvi, R. Agar-
Technologies 2, 10.1002/qute.201800043 (2018). wal, A. Galicia, N. Shammah, P. Nation, J. R. Johans-
[47] J. Dukelsky, G. G. Dussel, C. Esebbag, and S. Pittel, son, S. Ahmed, S. Cross, A. Pitchford, and F. Nori,
Exactly solvable models for atom-molecule hamiltonians, Qutip 5: The quantum toolbox in python (2024),
Phys. Rev. Lett. 93, 050403 (2004). arXiv:2412.04705 [quant-ph].
[48] C. Sträter, O. Tsyplyatyev, and A. Faribault, Nonequi-