PHYSICAL REVIEW RESEARCH 2, 013095 (2020)

Stabilizing open quantum batteries by sequential measurements

Stefano Gherardini ,1,* Francesco Campaioli,2 Filippo Caruso,1 and Felix C. Binder 3
1
Department of Physics and Astronomy & LENS, University of Florence, via G. Sansone 1, I-50019 Sesto Fiorentino, Italy
2
School of Physics and Astronomy, Monash University, Victoria 3800, Australia
3
Institute for Quantum Optics and Quantum Information - IQOQI Vienna, Austrian Academy of Sciences,
Boltzmanngasse 3, A-1090 Vienna, Austria

(Received 10 October 2019; published 29 January 2020)

A quantum battery is a work reservoir that stores energy in quantum degrees of freedom. When immersed in
an environment, an open quantum battery needs to be stabilized against free-energy leakage due to decoherence,
unavoidably entailing entropy production. For this purpose we here propose a stabilization protocol given
by a nonunitary open-loop control action able to compensate for the entropy increase and to maintain the
open quantum battery in its highest ergotropy state. The protocol relies on nonselective, frequent, projective
measurements that are interspersed by optimized time intervals. In accordance with a second-law-like inequality
derived for the entropy production rate of the controlled battery, the proposed procedure results in minimized
control power. The effectiveness of the method is finally tested on a qubit subject to decoherence, achieving an
average fidelity value around 95%.

DOI: 10.1103/PhysRevResearch.2.013095

I. INTRODUCTION the system, i.e., no postselection is performed, as shown

in Refs. [36–38]. The adoption of quantum measurements
Among recent research in quantum thermodynamics [1–5],
has been recently proposed for the realization of a quantum
the design of quantum energy storage devices, called quan-
Maxwell’s demon engine [39] and in Ref. [40] to fuel a
tum batteries [6–20], is of increasing interest. So far the
cooling engine. Our goal is to neutralize the local increase
main focus has been on multipartite speed-up effects in
of entropy and ensure energy-efficient control operations for
charging [7,9–12,15–17], fluctuations in charging precision
implementing fast, on-demand charging/discharging and sta-
[8,13,17,19], and mitigating imprecise unitary control pulses
bilization protocols, using the lowest amount of energy.
[20]. A specific quantum battery design using decoherence-
free subspaces has been recently proposed as to avoid energy
leakage into the environment [21]. However, no attention has II. OPEN QUANTUM BATTERIES
been paid to the energetics of efficiently stabilizing quantum
A quantum battery is a finite-dimensional quantum system
batteries that do not benefit from such protection-affording
B whose energy is quantified by a bounded internal Hamilto-
symmetries, even if contributions in the area of control theory
nian H0 . We here consider the highest-energy state |e and the
[22,23] touch upon this question both in classical [24] and in
lowest-energy state |g, both eigenstates of H0 , representing
quantum settings [25–28].
the maximally charged and discharged battery states, respec-
In this paper, we introduce the concept of an open quantum
tively. Hence, the battery’s capacity [10] is simply given by
battery (OQB). Here, the quantum system B, acting as a
Emax ≡ Tr[H0 (ρe − ρg )], with ρe ≡ |ee| and ρg ≡ |gg|. If
battery, interacts with the surrounding environment E leading
the battery system were perfectly isolated it would always
to decoherence [29]. Due to this interaction, the entropy of
evolve unitarily. In contrast, an open quantum battery, when
the battery increases [30–33] and thus unitary control pulses
left uncontrolled, evolves under the effect of some open dy-
applied to the system are not generally sufficient to compen-
namics, i.e., ρ̇t = −i[H0 , ρt ] + D[ρt ], where, here and below,
sate such entropy production and then stabilize the system;
h̄ is set to 1, ρt denotes the density operator of the system
rather, we would require a source of free energy, such as a
at time t, and D is the superoperator modeling free-energy
low-temperature heat bath. For this purpose, we propose a
leakage due to decoherence. Equivalently, the evolution of the
stabilization scheme based on a sequence of repeated quantum
system can also be described by means of a time-parametrized
measurements [34,35], each of them preserving the trace of
family of completely positive and trace-preserving (CPTP)
maps t : ρ0 → ρt , where ρt = t [ρ0 ], with steady state ρ ≡
limt→∞ t [ρ0 ] and initial density operator ρ0 .
*
[email protected] A charging protocol must be able to powerfully charge
the battery, bringing it into the excited state |e from an
Published by the American Physical Society under the terms of the arbitrary initial condition, e.g., a thermal state at inverse
Creative Commons Attribution 4.0 International license. Further temperature β or a state in the neighborhood of |g, so as
distribution of this work must maintain attribution to the author(s) to maximize its ergotropy [41], i.e., the amount of energy
and the published article’s title, journal citation, and DOI. that can then be unitarily extracted. While a closed quantum

2643-1564/2020/2(1)/013095(9) 013095-1 Published by the American Physical Society

GHERARDINI, CAMPAIOLI, CARUSO, AND BINDER PHYSICAL REVIEW RESEARCH 2, 013095 (2020)

τ (ii) Quantum measurements and Zeno protection. After

driving B into the state ρi , a projective energy measurement
E ρe (in the eigenbasis of H0 ) is performed on the battery: With
ρi ··· probability Pe ≡ Tr[ρi ρe ] the state of B collapses into the
ρα excited state, while with probability Pg ≡ 1 − Pe the collapse
ρi occurs into one of the other energy eigenstates. After the
measurement, if B has collapsed into the maximum energy
ρ0 state ρe , then a Zeno protection protocol is applied. The latter
ρg consists of a sequence of frequent projective measurements
(again in the energy eigenbasis) at discrete periodic times
(i) (ii) (iii) ··· t with the aim of freezing the dynamics of the battery and thus
stabilizing it in the excited state. As proved in Refs. [43–45],
the time interval τ between two consecutive measurements
has to be chosen according to the relation 2 HZeno τ 2  1,
FIG. 1. Stabilization scheme—single-run illustration. After the
where 2 HZeno is the variance of the effective Zeno Hamil-
initialization step (with initial state ρ0 ), the battery stabilization
protocol consists of intermittent free evolutions and fast unitary
tonian HZeno ≡ ρe H0 ρe = Ee ρe (Ee ≡ Tr[H0 ρe ]) with respect
controlled dynamics [dotted points, corresponding to steps (i) and to the freezing state. This physically means that the battery
(iii) of the procedure] and projective measurements [solid brown is repeatedly brought back to the maximum energy state ρe
line, step (ii)] in time intervals of duration τ . In particular, the with a probability almost equal to one as long as ρe —the
green dots denote the maximum energy state ρe , while the blue dots state to be stabilized—and the quantum state after the evo-
represent the state ρα of the battery immediately before a projective lution are statistically indistinguishable, i.e., their difference
measurement in the Zeno regime. ρi and ρi are the nearest states to is nondetectable by any measurement device [46]. Thus, for
ρe on the unitary orbit of ρ0 and ρg, respectively. an experimental realization of the protocol, τ needs to be
significantly smaller than both the timescale of the system
dynamics and the characteristic decoherence time.
battery can be charged by means of cyclic unitary operations, (iii) Reinitialization. If the projective energy measurement
an open quantum battery experiences nonequilibrium free- results in one of the lower-energy eigenstates, the stabilization
energy leakage. Hence, a unitary process no longer suffices procedure is repeated from the beginning, and unitary driving
to restore the battery state or to avoid the energy losses during is applied as in step (i). This means that the whole procedure
its dynamics. Moreover, the target state ρe does not generally is repeated until the Zeno protection protocol starts.
belong to the unitary controllability space [23] of the system, To summarize, ρe is an unstable state of the battery due to
especially if the charging time tc is comparable with 1/γ , interactions with the environment. Hence, in order to stabilize
where γ denotes the relevant decoherence coefficient of the B, we apply the nonunitary process given by a sequence of
superoperator D. So, how can an open quantum battery be projective measurements. Despite the probabilistic nature of
charged and stabilized? √ √ 2
this scheme, a very high fidelity F ≡ (Tr ρe ρt ρe ) in
stabilizing the battery can be achieved, as shown in Fig. 2 for
III. STABILIZATION SCHEME an example process. Further details on numerical simulations
are provided in Appendix B.
In this paper, we propose a nonunitary (NU) control pro-
tocol that counteracts the increase of entropy induced by
the interaction with the environment. The control scheme Performance measures
is realized by a sequence of projective measurements and For each time t we define two figures of merit for the
intermittent driving as illustrated in Fig. 1. The first objective stabilization scheme. First, the ratio
is to charge the battery by bringing it towards the excited Wstab (t )
state ρe . The second one is to maintain the system in the ςstab (t ) ≡ (1)
neighborhood of ρe during the time interval [0, tfin ]. Now, we Emax
introduce each step of the stabilization protocol. is named the relative stabilization cost, where · denotes
(i) Initialization. Given an input state ρ0 , the battery B is the average over a sufficiently large number of protocol re-
driven to that out-of-equilibrium state ρi on its control orbit alizations and Wstab (t ) is the energy expended to stabilize B.
which lies closest to ρe [e.g., in terms of the trace distance Second, we want to identify the excess cost of the stabilization
T (ρ, σ ) ≡ 21 Tr[ (ρ − σ )2 ]]. In general, this operation com- procedure besides the energy cost spent to just compensate
bines the uncontrollable open system dynamics with Hamilto- decoherence. To this end, we introduce the relative excess
nian control H (t ) = H0 + V (t ). While magnitude constraints stabilization cost, i.e.,
on the driving may impose bounds on the maximum achiev- |Wstab (t ) −  L(t )|
able power [11], we here assume that driving fields of arbitrary ξstab (t ) ≡ , (2)
magnitude are permitted. This allows for this step of the pro- Emax
tocol to be performed sufficiently fast so as to be considered where  L(t ) is the average energy leakage that would
unitary, despite the presence of decoherence. However, also spontaneously occur if the battery were left uncontrolled. Note
regimes of slow control could be equally treated by properly that ςstab (t ) is a cumulative energy term: It is zero when no
adapting existing control methods to the present scenario [42]. control is applied, but can also diverge since Wstab (t ) is an

013095-2
STABILIZING OPEN QUANTUM BATTERIES BY … PHYSICAL REVIEW RESEARCH 2, 013095 (2020)

0.7 (11) (11)

| (12) | | (12) |
t e t e 0.6
0.6 (a)
0.4
0.5
0.5 0.2
(d)
0.4 (b) 0
0 1 1.5 t 2
0 2 4 6 8
0.3 1
(11) t
t 0.6
0.8

Fidelity
0.2 (11)
e 0.4
| (12)
t
| 0.6
0.1 0.2
| (12) | (c) 0.4
e (e)
0 0
0 5 10 0 2 4 6 8 0 5 t 10
t t
FIG. 2. Stabilization scheme—numerical results for a qubit with internal Hamiltonian H0 = 3σx + σz (in natural units). (a) Average
behavior over time of the battery density matrix, obtained by repeating the stabilization procedure 1000 times. (b), (c) Behavior over time
of the battery density matrix in single realizations of the scheme: Being probabilistic, the charging process could require the application of
more than one projective measurement. In the subplots, (ρe(11) , ρe(12) ) and (ρt(11) , ρt(12) ) are the top diagonal elements and the coherence terms,
respectively, of the maximum energy state ρe of the qubit battery and of the corresponding time-evolved density matrix ρt . (d) Zoom of (c) in
the time interval [0.75,2.15], showing the occurrence of a failure collapse and the resulting reinitialization procedure. Further details can be
found in Appendix B. (e) Stabilization fidelity F over 1000 realizations of the stabilization procedure. At t = 0, F starts from a value in the
range [0.3,0.4] since also the initialization step has been taken into account.

unbounded quantity. Thus, it can be easily adapted for the with Pk ’s probabilities that the battery collapses in one of
definition of the relative stabilization rate the energy eigenstates. Indeed, to each of those probabilities
is associated the information content of the measurement
ςstab (t ) outcomes, which are stored in a classical memory [47]. This
Rstab ≡ lim , (3)
t→∞ t means that, while the measurement procedure locally reduces
which would be the same if defined in terms of ξstab (t ) the battery’s entropy, the reading and storing of the mea-
rather than ςstab (t ), due to the long term limit. This leads surement outcomes entails an additional entropy production
us to just consider the stabilization power Pstab = Ẇstab as a which cannot be neglected. According to Landauer’s principle
performance measure in the following section. [48–50], the irreversible erasure of such information leads to
an energy consumption, proportional to the temperature of the
thermal bath used in the erasure procedure.
IV. MINIMUM CONTROL POWER Once again, it is worth noting that, since B is affected by
decoherence, the evolution of the battery admits at least one
Here, we prove a bound providing the minimum power
fixed point denoted as ρ dec . For the case of a qubit, the steady
required to stabilize the OQB. It originates from a second-
state ρ dec (with no coherence in the energy eigenbasis) can
law-like inequality for the battery’s irreversible entropy pro-
always be described by an effective temperature Tρ dec . The
duction rate (ρt ) (see Appendix A for the proof). A similar
latter is interpreted as the physical temperature of a fictitious
result can be found in Ref. [26], but concerning the energy
quantum system that would lead to the same decoherence
cost to coherently control a mesoscopic quantum system.
effects. In particular, as shown in Appendix A, by defining
In our open-loop control framework, the entropy produc-
E (ρt ) and ED (ρt ), respectively, as the battery’s total energy
tion rate (ρt ) equals the sum of two contributions, D (ρt )
and the energy driven into B by the environment, the control
and NU (ρt ), denoting respectively the entropy production
power Ẇstab obeys the following inequality,
rates due to environmental decoherence and the effect of the
observer/experimenter, responsible for the nonunitary control Pstab ≡ Ẇstab (ρt )  Ė (ρt ) − Tρ dec ṠD (ρt ), (5)
of the battery. In particular, as discussed in Appendix A, the
entropic contribution NU (ρt ) is equal
 to the time derivative with SD denoting the von Neumann entropy of the uncon-
of the Shannon entropy H (P) ≡ − k∈{e,g} Pk log Pk , i.e., trolled battery (note that ṠD ∝ D ). The lower bound (5) can
  be recast into the inequality
Pe
NU (ρt ) = Ḣ (P) = −Ṗe log , (4)
1 − Pe Ḟ (ρt )  0, with F ≡ ED − Tρ dec SD (6)

013095-3
GHERARDINI, CAMPAIOLI, CARUSO, AND BINDER PHYSICAL REVIEW RESEARCH 2, 013095 (2020)

the battery free energy. This inequality represents the second since the stabilization procedure is repeated with probability
law of thermodynamics: The free energy of the uncontrolled Pe until the battery is charged, i.e., ρt reaches ρe , the average
battery reduces due to the increase of the von Neumann total work needed to keep the energy storage in the battery
entropy SD resulting from the open systems dynamics. There- until t = tfin is given by the following relation,
fore, the minimum value of Ẇstab implies the equality Ḟ (ρt ) = ⎛ ⎞
0, with the result that the lowest energy Wstab(min)
required to
N

(min) Wstab (tfin ) ≈ ⎝1 + Pgk ⎠ Eevol +  EZeno , (10)

control B (note that, apart from a constant term, Wstab is k=1
equal to E − Tρ dec SD ) is such that the free energy is con-
stant, i.e., the increase in entropy due to the environment is where N is the average number of times the stabilization
compensated by the control operation. Note that Eq. (5) is procedure is repeated with probability Pg.
valid regardless of the control action applied on B, and for Then, let us derive the average total energy leakages
this reason the symbol · has not been used. However, for  L(tfin ). Since the battery is an open quantum system, its
the probabilistic stabilization procedure we are proposing, the evolution entails energy leakages, which are equal to
results from Eq. (5) just hold true only on average.
Levol = Tr[H0 D[ρt ]]dt, (11)
V. ENERGETIC EFFICIENCY
the integrated energy flow between B and the environment.
Returning to Eq. (2), we now derive the average control In contrast, the projective measurements, assumed as dis-
energy Wstab (t ) and environmental losses  L(t ) for an continuous operations, are not affected by the environment.
open quantum battery controlled by a sequence of projective This assumption is clearly just an abstraction, which may be
measurements. In particular, the energetic balance equation of extended to nonideal measurements with an inherent energy
the OQB is characterized within the total time interval [0, tmax ] consumption [52]. During the Zeno protection procedure,
and evaluated in the limit of fast control, i.e., V (t ) = 0 almost losses are on average equal to

tk +τ
for any t.
The battery’s energy is determined by the time-  LZeno  = Tr[H0 D(ρt )]dt, (12)
independent Hamiltonian H0 . Thus, the cost for the initializa- tk
k
tion of the battery and its dynamical evolution is exactly equal
to where τ is the time interval between two consecutive Zeno
measurements. In conclusion, this implies that the average
Eevol = Tr[H0 (ρi − ρ0 )]. (7) total energy leakage  L(tfin ) at the final time instant tfin is

Indeed, for fast control,
 the integral Tr[V (t )(ρt − ρ0 )]dt ⎛ ⎞
is negligible and ρt dt ≈ ρi . On the other hand, the cost

N
Emeas of each projective measurement is given by the dif-  L(tfin ) ≈ ⎝1 + Pgk ⎠ Levol +  LZeno , (13)
ference between the battery energies, respectively, after and k=1
before the measurement,
with Levol counted N times until ρt = ρe .
Emeas The minimal requirement for the battery stabilization is
 the equality between the average total work Wstab (tmax ) and
Tr[H0 (ρe − ρi )], with probability Pe = Tr[ρi ρe ],
= the accumulated losses until the time instant tmax denoted
Tr[H0 (ρi − ρg )], with probability Pg = 1 − Pe .
as break-even time. An upper bound of the break-even time
(8)
can be computed just by inverting the relation Wstab (tmax ) =
However, on average there is no energetic cost associated Emax . However, to make B a high-performance battery, the
with the measurement, i.e.,  Emeas  = 0, independently of collapse probability Pg has to be as close as possible to zero in
ρi . Only the entropic cost for the erasure of the measure- the shortest time interval, and this is in contrast with the need
ment information has to be considered. The latter is equal to for low-energy consumption. Thus, during the initialization
β −1 H (P), where β denotes the inverse temperature associated step of the procedure, a trade-off between precision and
with the thermal reservoir allowing for the resetting of the energy cost is unavoidable. Similar results are also observed
memory after each measurement. The same holds true for the during the Zeno protection protocol. In this regard, let us
Zeno protection protocol, whereby the energy cost of a projec- consider the average stabilization power Pstab  = Ẇstab  ≈
tion on the energy basis is equal on average to β −1 H (P(ρα )), Wstab (τ )/τ . By relating Eqs. (10) and (13), one determines
such that overall one has that that Pstab (τ ) is approximately equal to
 EZeno  = m β −1 H (P(ρα )), (9) Pstab (τ ) ≈ [Tr[H0 (ρi − ρ0 )] + m β −1 H (P(ρα ))]/τ, (14)
where m denotes the average number of projections in the where the first and second terms of Eq. (14) denote, respec-
Zeno regime (Zeno measurements) [51] and ρα is the average tively, the average cost per cycle to initially bring the battery
state of the battery immediately before each Zeno measure- close to ρe , and the Landauer cost per cycle to reset the
ment, depending on the measurement frequency 1/τ . It is (classical) memory register. We can thus conclude that the
worth noting that also during the Zeno protection proce- longer is τ , the smaller is the value of the power required to
dure the average measurement cost Pe (ρα )Tr[H0 (ρe − ρα )] + stabilize on average the battery, but the less accurate will be
Pg (ρα )Tr[H0 (ρg − ρα )] is a vanishing quantity. As a result, the precision to bring it on ρe .

013095-4
STABILIZING OPEN QUANTUM BATTERIES BY … PHYSICAL REVIEW RESEARCH 2, 013095 (2020)

As a final remark, it is worth observing that for a vanish-

ing value of Pg we can perform a first-order expansion of 10 TZeno= 6
1.5
both Wstab (tfin ) and  L(tfin ) as a function of Pg, so that TZeno= 5
Wstab (tfin ) ≈ (1 + Pg ) Eevol +  EZeno  and  L(tmax ) ≈

m Pg ( )
TZeno= 4
(1 + Pg ) Levol +  LZeno . Therefore, if we also reasonably 8 TZeno= 3
1
assume that in the Zeno regime the sum of the energy losses TZeno= 2
is on average almost equal to the energy required to charge TZeno= 1
0.5
the quantum system (i.e.,  EZeno  ≈  LZeno ), the relative

Zeno
6 TZeno= 0.4
excess stabilization cost at t = tfin is given by the following 0
relation, 0 2 4 6

1 + Pg 4
ξstab (tfin ) ≈ | Eevol − Levol |, (15)
Emax
with the result that ξstab (tfin ) = 0 if the energy cost to drive 2
the battery up to the state ρi perfectly equals the decoherence
losses during the battery evolution (i.e., if the losses are just
compensated by the control action). 0
0 1 2 3 4 5 6
Entropic cost in the Zeno regime
FIG. 3. Entropic cost of the Zeno protection procedure: σZeno
Now, let us analyze in more detail the entropic cost of the
as a function of the time interval τ between Zeno measurements.
Zeno protection procedure, based on applying a sequence of
The results have been numerically derived for the same quantum
projective energy measurements. In this regard, at the level of
system used in Fig. 2. Each curve has been obtained by choosing
the battery, the shorter is τ , the closer the density operator a fixed duration TZeno ≡ tfin − tmax , among a set of values (see the
of B approaches ρe . However, such a condition does not legend of the figure), and letting τ vary, so that also m≈ TZeno /τ
imply that the global variation of energy during the procedure τ
of Zeno measurements changes every time. The integral 0 Ḣ (P)dt
decreases in the same way. Indeed, the smaller is τ , the greater (black line), numerically solved with the initial condition ρ = ρe ,
is the average number of Zeno measurements, each of them has a monotonically increasing behavior for greater values of τ ,
entailing an entropic cost proportional to the irreversible loss thus identifying m as the dominant factor. Inset: Amount of not
of the information content. From a purely dynamical point stored energy mPg (τ ) (normalized by Emax ) as a function of τ for
of view, this corresponds to the cost of purifying the state TZeno = 0.4, 1, 2, 3, 4, 5, and 6, in natural units. Here, the black
of the open quantum battery in correspondence to ρe . As a line denotes Pg (τ ), and an unavoidable worsening of the battery
result, an increasing of the Landauer cost, corresponding to a stabilization is observed when τ increases.
more frequent memory erasure, is expected. Such behavior is
illustrated by Fig. 3, in which the entropy production
τ τ
the control power is minimized. The introduced method can
σZeno ≡ m NU (ρt )dt = m Ḣ (P)dt (16)
0 0
also be seen as a procedure to make the energy (Hamiltonian)
basis of the system a decoherence-free subspace [53–55]. This
is plotted as a function of τ by fixing the duration TZeno of
implicitly corresponds to having engineered a superoperator
the procedure. In Fig. 3, the Landauer entropic cost, being
C[ρt ], modeling the average effects of applying projective
proportional to 1/τ , diverges as τ decreases to zero, and the
measurements, so that for ρt ≈ ρe ,
behavior of σZeno as a function of τ is a decaying exponential.
Its exponent has the dimensionality of an energy (in natural
units); thus, it represents the global energy variation of B in ρ̇t = −i[H0 , ρt ] + D[ρt ] + C[ρt ] (17)
the Zeno regime. In summary, the value of τ has to follow
a trade-off condition: τ cannot be too small so as to prevent
high-energy and entropic costs, but neither too large in order is approximately equal to zero at any time instant t.
to avoid an inadequate value of the stabilization fidelity. As the main outlook, one could take into account the pos-
sibility that the projective measurements adopted in the sta-
bilization scheme are nonideal, at least according to the
VI. CONCLUSION AND OUTLOOK
definitions recently introduced in Ref. [52], so as to prevent
Thermodynamics and control theory have been combined an unbounded energy cost for their performance. Finally, other
with the aim to stabilize an open quantum system that acts promising directions for future research lie in (i) the explicit
as a battery. In particular, in accordance with the laws of treatment of charging and stabilization fluctuations, as was
thermodynamics, we have analyzed the variation of the en- recently done for Gaussian quantum batteries [13], (ii) the
tropy generated on average by a quantum system, to be used adoption of optimal quantum control theory [22] to improve
as a device for the storage of energy, in interaction with an the stabilization procedure, (iii) the analysis for open quantum
external environment and manipulated by a properly designed batteries with non-semigroup dynamics, e.g., non-Markovian
control action. In this respect, we have proposed a nonunitary behavior leading to memory effects, and (iv) the extension to
control strategy, represented by a sequence of projective mea- multipartite quantum systems [9], exploiting the physics of
surements interspersed by optimized time intervals, so that quantum Zeno dynamics [55].

013095-5
GHERARDINI, CAMPAIOLI, CARUSO, AND BINDER PHYSICAL REVIEW RESEARCH 2, 013095 (2020)

ACKNOWLEDGMENTS cost Wstab . Our goal is to find a lower bound for Ẇstab (ρt ). The
This work was financially supported by the Fondazione CR energy ED (ρt ) due to decoherence is given by
Firenze through the project Q-BIOSCAN and QUANTUM- ED (ρt ) ≡ Tr[ρt H0 ] − Tr[ρ0 H0 ], (A4)
AI, PATHOS EU H2020 FET-OPEN Grant No. 828946, and
UNIFI Grant Q-CODYCES. F.C.B. acknowledges funding and the corresponding infinitesimal energy leakage is equal to
from the European Union’s Horizon 2020 research and inno- ĖD (ρt ) = Tr[D[ρt ]H0 ]. (A5)
vation programme under the Marie Skłodowska-Curie Grant
Agreement No. 801110 and the Austrian Federal Ministry of If the battery is a two-level system, the energy current ĖD (ρt )
Education, Science and Research (BMBWF). can be written as
ĖD (ρt ) = −Tρ dec Tr[D[ρt ] log ρ dec ] = Ė (ρt ) − Ẇstab (ρt ),
APPENDIX A: SECOND-LAW-LIKE INEQUALITY (A6)
FOR STABILIZING AN OQB
where Tρ dec is the effective temperature of the battery in
In this Appendix, we provide more details about the correspondence to the steady state ρ dec . We point out that
derivation of the minimum power required to stabilize an Tρ dec is more than a parameter introduced for mathematical
OQB affected by decoherence. Our proof follows a similar convenience: It corresponds to the physical temperature of a
procedure to the analysis presented in Ref. [26]. fictitious quantum system leading to the same decoherence
Let us consider a quantum system dynamics described by effect as the general dynamics assumed here.
the dynamical semigroup V (t ) ≡ eLt with a (not necessarily In this way, a lower bound for Ẇstab (ρt ) can now be derived.
unique) fixed point ρ dec = V (t )ρ dec and time-independent L, We first reconsider Eq. (A1) again for the OQB in the absence
where ρ dec denotes the steady state induced by the presence of of control,
battery decoherence alone. As shown in Ref. [30], the corre-
sponding entropy production rate (ρt ) related to the battery D (ρt ) = −Tr[D[ρt ] log ρt ] + Tr[D[ρt ] log ρ dec ]  0.
density operator at time t is convex and given by the following (A7)
relation, Since −Tr[D[ρt ] log ρt ] = ṠD (ρt ) is the time derivative of the
(ρt ) ≡ −Tr[L[ρt ](log ρt − log ρ dec )]  0. (A1) von Neumann entropy for the uncontrolled battery, we have
that
Recasting this general picture to the OQB model discussed
in the main text, without applying external control for now, Ẇstab (ρt ) − Ė (ρt )  −Tρ dec ṠD (ρt ), (A8)
one has L[ρt ] = D[ρt ]. Therefore, with our control knob which leads to the analytical expression of the lower bound of
given by a sequence of projective measurements, the entropy Ẇstab ,
production rate (ρt ) of the controlled OQB is greater than
or equal to the entropy contribution D (ρt ) due to the envi- Ẇstab (ρt )  Ė (ρt ) − Tρ dec ṠD (ρt ). (A9)
ronment alone. More formally, (ρt ) = D (ρt ) + NU (ρt ), Notice that Eq. (A9) has to fulfill the second law of thermody-
with NU (ρt ) denoting the entropy production rate of the namics. Indeed, by substituting Ė = ĖD + Ẇstab from the first
battery given by the nonunitary control transformation. In law of thermodynamics, the lower bound (A9) can be recast
other words, the total entropy production is lower bounded in the following inequality,
as
Ḟ (ρt ) ≡ ĖD (ρt ) − Tρ dec ṠD (ρt )  0, (A10)
(ρt )  −Tr[D[ρt ](log ρt − log π )] ≡ D (ρt ). (A2)
where F (ρt ) stands for the free energy of the uncontrolled
Let us observe that by means of the control procedures the battery. Inequality (A10) implies that, without controlling the
battery is stabilized in the sense that its density operator ρt battery, its entropy unavoidably grows due to decoherence,
approaches the maximum energy state ρe , which thus becomes leading thus to a progressive decreasing of the battery free
an equilibrium state induced by the control. Moreover, since energy.
we are assuming that any operation on B preserves the trace Here, it is worth observing that the lower bound (A9) is
of its density operator, the total entropy production (ρt ) is quite conservative, in the sense that the value provided for the
non-negative due to the monotonicity of relative entropies minimum control power Ẇstab could be overestimated. This is
under CPTP maps. This means that only energy exchanges because we have not directly expressed Ẇstab as a function of
are allowed, such that the evolution of the uncontrolled battery the entropy production rate NU (ρt ) given by controlling B by
can be always described by a CPTP quantum map. means of the proposed nonunitary transformation (sequence
As a next step, we quantify the rate of change of the battery of projective measurements). NU (ρt ) is equal to the sum of
total energy E (ρt ) under stabilizing control by using the first the entropy production rates associated with each projective
law of thermodynamics, with Ė (ρt ) given by the relation measurement, and it can be obtained by evaluating the energy
Ė (ρt ) = ĖD (ρt ) + Ẇstab (ρt ), (A3) cost in storing and erasing the measurement outcomes in re-
lation to Landauer’s principle [48–50]. In this regard, by con-
where ĖD (ρt ) is the energy current driven into the battery by sidering the expression for the entropic contribution st of each
the environment, while Ẇstab (ρt ) denotes the power required to single measurement result, i.e., st ≡ − log P, with P (equal to
charge B and stabilize it against decoherence. Here, the cost Pe or Pg) denoting the probability that the battery collapses in
of the sequential measurements is included within the control one of the two energy eigenstates [47], we find that NU (ρt ) =

013095-6
STABILIZING OPEN QUANTUM BATTERIES BY … PHYSICAL REVIEW RESEARCH 2, 013095 (2020)

Ḣ (P) = −Ṗe log ( 1−P

Pe
), where Ḣ denotes the time derivative (a) 1
e 
of the Shannon entropy H (P) ≡ − k∈{e,g} Pk log Pk , with
Pg = 1 − Pe . It is worth noting that the entropy production
NU is zero if and only if Pg = Pe = 1/2, i.e., the probabilities 0.5
that the battery collapses in the maximum or lowest-energy (11)
t
state are both equal to 1/2. (22)
t
0
APPENDIX B: DETAILS ABOUT THE NUMERICAL 0 2 4 6 8 10
IMPLEMENTATION (b) t
0.5
The results of Fig. 2 in the main text have been obtained
by considering as a quantum battery the following two-level
system, with an internal Hamiltonian H0 = σx + ωσz , where
ω = 1,  = 3 (in units such that h̄ = 1), and σx , σz Pauli 0
matrices. Thus, in the basis of σz , given for convention by the { (12)
t
}
eigenstates |0 ≡ [0, 1]T and |1 ≡ [1, 0]T [the superscript { (12)
}
t
(·)T denotes the transposition symbol], the corresponding -0.5
maximum and minimum energy states are respectively equal 0 2 4 6 8 10
to t
   
0.658 0.474 0.342 −0.474
ρe ≈ and ρg ≈ . FIG. 4. (a) Behavior over time of the battery density matrix
0.474 0.342 −0.474 0.658
elements ρt(11) and ρt(22) = 1 − ρt(11) . (b) Behavior over time of the
(B1) real and imaginary part of the battery density matrix element ρt(12) .
Here, we have reasonably chosen as the input density
operator ρ0 the minimum energy state ρg. Moreover, in order so that Pg  1/2. In this regard, in the numerical simula-
to fulfill the fast control condition, we have assumed to use tions we have verified that also in this case  Emeas (t ) ≡
a time-dependent term V (t ) in the driving Hamiltonian only PgTr[H0 (ρt − ρg )] + Pe [H0 (ρe − ρt )] ≈ 0 for each t within
to slightly bring out of equilibrium the battery from ρg to the entire evolution of the system, and thus also for t ∗ .
≡|00|. Given the internal time-independent Hamiltonian H0 , The results in Figs. 4 and 5 could induce the experimenter
this operation is achieved by taking V (t ) = e−iφσy (rotation to take as t ∗ a sufficiently long time interval so as to minimize
around the y axis), with the trace distance T (ρt , ρe ) and at the same time maximize
  the probability Pe . However, because the stabilization scheme
ρg(11)
φ ≡ arctan . (B2) is a probabilistic procedure, this choice could bring the main
ρg(21) disadvantage of waiting for a long time interval and then
Then, to transfer the battery population from |00| to
the initialization state ρi , only the dynamical evolution
governed by H0 has been exploited. Accordingly, under this (a)
assumption, we need to determine the optimal value of t (i.e., 0.8
t ∗ ) in correspondence to which it is worth performing the first
) e

energy projective measurement of the protocol.

T( ,

The dynamical evolution of the system is given by the 0.6

Markovian master equation ρ̇t = −i[H0 , ρt ] + D[ρt ] (h̄ = 1),
where the superoperator D[ρt ] modeling decoherence within
the battery dynamics has been chosen equal to D[ρt ] = 0.4
γ (−{N , ρt } + 2N ρt N ), i.e., as an operator inducing pure 0 2 4 6 8 10
dephasing, with {·, ·} the Poisson bracket, N ≡ |11|, and (b) t
γ = 2/3. Here, it is worth noting that also pure-dephasing
master equations, despite being energy preserving, involve 0.5
dynamical behaviors worthy of being studied, since stabilizing
an OQB implicitly implies the protection (in our case) of
Pe

coherence in the battery energy basis. Moreover, the moti- 0.4

vation under the choice of γ = 2/3 will be clear below. In
Fig. 4 we show the behavior over time of both the populations
and coherence of the two-level system for a whole duration 0.3
of the dynamics taken equal to 10 (always in natural units) 0 2 4 6 8 10
by starting from the state |00|. Instead, in Fig. 5 we plot t
the trace distance of ρt with respect to ρe , as well as the
probability Pe (t ) ≡ Tr[ρt ρe ]. We can observe that the trace FIG. 5. (a) Behavior over time of the trace distance T (ρt , ρe ).
distance T (ρt , ρe ) is always greater than or equal to 1/2, (b) Behavior over time of the probability Pe .

013095-7
GHERARDINI, CAMPAIOLI, CARUSO, AND BINDER PHYSICAL REVIEW RESEARCH 2, 013095 (2020)

observing the battery staying for most of the time not in the proposed stabilization scheme, based on sequential quantum
maximum energy state but in correspondence to ρg, and thus measurements, turns out to be extremely efficient from an
leading to a very low stabilization fidelity F. To make a better energetic point of view only if the probability to collapse
choice of t ∗ , it is worth analyzing the reason why Pe is always onto ρe after each quantum measurement is sufficiently high,
1/2. We find that, in order to achieve the maximum energy ideally close to 1. If not, a greater energy cost (if compared
state ρe , we need to stabilize both populations and coherence with Emax ) is required, so as to bring the system into the
of the battery, but with the chosen internal Hamiltonian H0 the maximum energy state and at the same time compensate the
stabilization of populations and coherence cannot be reached presence of the external environment leading to decoherence.
at the same time. In this regard, there are three possibilities: (a) Finally, as it can be observed in Fig. 2 in the main text, the
Minimize only the difference between the modulus squared of probability that the state of the battery collapses in the mini-
coherence terms of ρt and ρe ; (b) minimize only the difference mum energy state ρg while the procedure of Zeno protection
between the modulus squared of the diagonal terms of ρt and is turned on is very low (smaller than 1%). This is due to our
ρe ; and (c) find a trade-off between (a) and (b) by ensuring that choice to take the time interval τ between Zeno measurements
the value of t ∗ is not too large and at the same time Wstab  is as to be equal to 0.0662 (in natural units), five times smaller than
small as possible with respect to the average energy leakages t ∗ . However, there does exist the possibility that the Zeno
 L. For the specific implementation of Fig. 2 in the main protection procedure would fail; in such a case we simply
text we have chosen solution (c) corresponding to a value of reinitialize the system and the stabilization scheme is repeated
t ∗ equal to 0.33 (in natural units). Notice that, since Pe  1/2 from the beginning. In this regard, it is worth noting that for
by starting from the state |00|, we can at most minimize the larger values of τ the fidelity F decreases and consequently
average total control work and get a very high fidelity F, but Wstab  unavoidably increases, since for a fixed value of TZeno
without achieving the best possible energetic efficiency. We the stabilization scheme needs to be reinitialized a greater
have deliberately chosen this example in order to show that the number of times.

[1] R. Kosloff, Entropy 15, 2100 (2013). [17] S. Julià-Farré, T. Salamon, A. Riera, M. N. Bera, and M.
[2] J. Goold, M. Huber, A. Riera, L. del Rio, and P. Skrzypczyk, Lewenstein, arXiv:1811.04005.
J. Phys. A: Math. Theor. 49, 143001 (2016). [18] F. Barra, Phys. Rev. Lett. 122, 210601 (2019).
[3] S. Vinjanampathy and J. Anders, Contemp. Phys. 57, 545 [19] L. P. Garcia-Pintos, A. Hamma, and A. del Campo,
(2016). arXiv:1909.03558.
[4] Thermodynamics in the Quantum Regime, edited by F. Binder, [20] A. C. Santos, B. Çakmak, S. Campbell, and N. T. Zinner, Phys.
L. A. Correa, C. Gogolin, J. Anders, and G. Adesso (Springer, Rev. E 100, 032107 (2019).
Berlin, 2018). [21] J. Liu, D. Segal, and G. Hanna, J. Phys. Chem. C 123, 18303
[5] S. Deffner and S. Campbell, Quantum Thermodynamics (Mor- (2019).
gan & Claypool Publishers, San Rafael, CA, 2019). [22] C. Brif, R. Chakrabarti, and H. Rabitz, New J. Phys. 12, 075008
[6] F. Campaioli, F. A. Pollock, and S. Vinjanampathy, in Ther- (2010).
modynamics in the Quantum Regime, edited by F. Binder, [23] C. P. Koch, J. Phys.: Condens. Matter 28, 213001 (2016).
L. Correa, C. Gogolin, J. Anders, G. Adesso, Fundamental [24] J. M. Horowitz and J. L. England, Entropy 19, 333
Theories of Physics Vol. 195 (Springer, Cham, 2018), Chap. 8, (2017).
pp. 207–225. [25] F. Ticozzi, K. Nishio, and C. Alta, IEEE Trans. Autom. Control
[7] R. Alicki and M. Fannes, Phys. Rev. E 87, 042123 (2013). 58, 74 (2013).
[8] P. Skrzypczyk, A. A. J. Short, and S. Popescu, Nat. Commun. [26] J. M. Horowitz and K. Jacobs, Phys. Rev. Lett. 115, 130501
5, 4185 (2014). (2015).
[9] F. C. Binder, S. Vinjanampathy, K. Modi, and J. Goold, New J. [27] J. J. W. H. Sørensen, M. Dalgaard, A. H. Kiilerich, K. Mølmer,
Phys. 17, 075015 (2015). and J. F. Sherson, Phys. Rev. A 98, 062317 (2018).
[10] F. C. Binder, Work, heat, and power of quantum processes, [28] R. Dann, A. Tobalina, and R. Kosloff, Phys. Rev. Lett. 122,
D.Phil. thesis, University of Oxford, 2016. 250402 (2019).
[11] F. Campaioli, F. A. Pollock, F. C. Binder, L. Céleri, J. Goold, [29] H.-P. Breuer and F. Petruccione, The Theory of Open Quantum
S. Vinjanampathy, and K. Modi, Phys. Rev. Lett. 118, 150601 Systems (Oxford University Press, Oxford, UK, 2002).
(2017). [30] H. Spohn, J. Math. Phys. 19, 1227 (1978).
[12] D. Ferraro, M. Campisi, G. M. Andolina, V. Pellegrini, and M. [31] P. A. Camati, J. P. S. Peterson, T. B. Batalhão, K. Micadei, A. M.
Polini, Phys. Rev. Lett. 120, 117702 (2018). Souza, R. S. Sarthour, I. S. Oliveira, and R. M. Serra, Phys. Rev.
[13] N. Friis and M. Huber, Quantum 2, 61 (2018). Lett. 117, 240502 (2016).
[14] G. M. Andolina, D. Farina, A. Mari, V. Pellegrini, V. [32] S. Gherardini, M. M. Müller, A. Trombettoni, S. Ruffo, and F.
Giovannetti, and M. Polini, Phys. Rev. B 98, 205423 Caruso, Quantum Sci. Technol. 3, 035013 (2018).
(2018). [33] T. Batalhão, S. Gherardini, J. Santos, G. Landi, and M.
[15] D. Farina, G. M. Andolina, A. Mari, M. Polini, and V. Paternostro, in Characterizing Irreversibility in Open Quantum
Giovannetti, Phys. Rev. B 99, 035421 (2019). Systems, edited by F. Binder, L. Correa, C. Gogolin, J. Anders,
[16] G. M. Andolina, M. Keck, A. Mari, M. Campisi, V. Giovannetti, G. Adesso, Thermodynamics in the Quantum Regime, Vol. 195
and M. Polini, Phys. Rev. Lett. 122, 047702 (2019). (Springer, Cham, 2018), Chap. 16, pp. 395–410.

013095-8
STABILIZING OPEN QUANTUM BATTERIES BY … PHYSICAL REVIEW RESEARCH 2, 013095 (2020)

[34] A. De Pasquale, K. Yuasa, and V. Giovannetti, Phys. Rev. A 96, [45] M. M. Müller, S. Gherardini, A. Smerzi, and F. Caruso, Phys.
012316 (2017). Rev. A 94, 042322 (2016).
[35] S. Gherardini, A. Smirne, M. M. Müller, and F. Caruso, [46] W. K. Wootters, Phys. Rev. D 23, 357 (1981).
Proceedings 12, 11 (2019). [47] P. Strasberg, Phys. Rev. E 100, 022127 (2019).
[36] M. Campisi, P. Talkner, and P. Hänggi, Phys. Rev. Lett. 105, [48] S. Lorenzo, R. McCloskey, F. Ciccarello, M. Paternostro, and
140601 (2010). G. M. Palma, Phys. Rev. Lett. 115, 120403 (2015).
[37] F. W. J. Hekking and J. P. Pekola, Phys. Rev. Lett. 111, 093602 [49] L. Mancino, M. Sbroscia, E. Roccia, I. Gianani, F. Somma, P.
(2013). Mataloni, M. Paternostro, and M. Barbieri, npj Quantum Inf. 4,
[38] S. Gherardini, L. Buffoni, M. M. Müller, F. Caruso, M. Campisi, 20 (2018).
A. Trombettoni, and S. Ruffo, Phys. Rev. E 98, 032108 (2018). [50] K. Abdelkhalek, Y. Nakata, and D. Reeb, arXiv:1609.06981.
[39] C. Elouard, D. Herrera-Martí, B. Huard, and A. Auffèves, Phys. [51] For a fixed value both of τ and tfin also the number of Zeno
Rev. Lett. 118, 260603 (2017). measurements is a random variable, since an unsuccessful
[40] L. Buffoni, A. Solfanelli, P. Verrucchi, A. Cuccoli, and M. projection on ρg can happen, in spite of a very small occurrence
Campisi, Phys. Rev. Lett. 122, 070603 (2019). probability for such phenomenon.
[41] A. E. Allahverdyan, R. Balian, and T. M. Nieuwenhuizen, [52] Y. Guryanova, N. Friis, and M. Huber, Quantum 4, 222 (2020).
Europhys. Lett. 67, 565 (2004). [53] D. A. Lidar and B. K. Whaley, in Irreversible Quantum Dy-
[42] N. Suri, F. C. Binder, B. Muralidharan, and S. Vinjanampathy, namics, edited by F. Benatti and R. Floreanini (Springer, Berlin,
Eur. Phys. J.: Spec. Top. 227, 203 (2018). 2003), pp. 83–120.
[43] A. Smerzi, Phys. Rev. Lett. 109, 150410 (2012). [54] P. G. Kwiat, A. J. Berglund, J. B. Altepeter, and A. G. White,
[44] F. Schäfer, I. Herrera, S. Cherukattil, C. Lovecchio, F. S. Science 290, 498 (2000).
Cataliotti, F. Caruso, and A. Smerzi, Nat. Commun. 5, 3194 [55] M. M. Müller, S. Gherardini, and F. Caruso, Ann. Phys. 529,
(2014). 1600206 (2017).

013095-9