engine and its quantum signature -...
TRANSCRIPT
The Journey to the quantum Carnot engine and its quantum signature
Ronnie Kosloff
Institute of Chemistry, Hebrew University Jerusalem, Israel The Fritz Haber Center for Molecular Dynamics
November 12, 2019
Talk content
1 Quantum Thermodynamics: Consistency.
2 Reciprocating heat engines: learning from example.
3 Generalized Canonical Invariance.
4 Thermodynamics of the GKLS Master Equation.
5 The Non Adiabatic Master Equation (NAME).
6 The inertial theorem.
7 Shortcut to Equilibrium (STE).
8 Quantum control of open systems.
9 The Quantum Carnot engine with finite power.
Eitan Geva
Jeff Gordon
Tova Feldmann
Jose Palao
Quantum Thermodynamics
Eitan Geva
Jeff Gordon
Tova Feldmann
Jose Palao
Quantum Thermodynamics
LevAmikamAmikam Lev
Morag Am Shalem
YYairair Rezek RezekRobert Alicki
Quantum ThermodynamicsQuantum Thermodynamics
Finding quantum analogies to:
0) System bath partition:Approach to equilibrium.
1) The first law:Energy change.
2) The second law:Entropy change.
3) The third law: Approaching the absolute zero.
Any theory should be consistent with Thermodynamics
Einstein 1905
ConsistencyEmergence of Thermodynamics from quantum mechanics
Can a Thermodynamical viwpoint be relavant to a single device at the quantum limit
What is the limit of minaturization of a quantum heat engine
What is quantum in a quantum heat engine Is there quantum supremacy
?
?
??
single molecular refrigerator
Learning from example Thermodynamic ideals
?How small a quantum engine can be
What is quantum in quantum heat devices
Power or efficiency?
Efficiency at maximum power Maximum efficiency
How far can we mintuaize?
The quantum Otto refrigeration cycle
Hot Isochore (isomagnetic) A #B $=$h .The working mediumis in contact with the hot bath of temperature Th .
Expansion adiabat (demagnetization) B #CThe field changes from $h to $c
Cold Isochore (isomagnetic) C #D $=$c .The working medium is in contact with the cold bath of temperature Tc .
Compression adiabat (magnetization) D #AThe field changes from $c to $h .
Uh
Uhc
Uc
Uch
Cycle Propagator: Ucycle=UhUhcUcUch
[UhUhc,UcUch] % 0limit cycle
UcycleY=Y
Realizations 2016 Science 352 , 325 (2016)
Reciprocating heat engines
Learning from example
1
2
3
Carnot cycle Hot to cold adiabatic stroke hcCold isotherm cCold to hot adiabatic stroke ch
4 Hot isotherm h
Carnotcycle:cyc=hchchc
Operating conditionsfixed point of CPTP map
<n>
ωωω ω
Compression ratio
ω
<n>Τh
Tc
A quantum-mechanlcal heat engine operating in finite time. A model consisting of spin-1 /2 systems as the working fluid
Eitan Geva and Ronnie Kosloff
Departmenו of Physical Chemistry and The Fritz Haber Research Center for Molecular Dynamics, The Hebrew Uniuersity, Jerusalem 91904, Jsrae/
(Received 28 August 1991; accepted 21 October 1991)
_____ כo _____ ס (ו)
2 2' 3 3'
l3c·,
s2
1. 4 4' Sl ..
S2 .:•: w·-._ W w,,,//Qו,
Qc'°' c S1 ········------ ·._ -......w .Q.5 -------=---
FIG. S. Thc cyclc 1'-2' -3' -4'-1' is of the Curzon-Ahlborn
FIG. 1. The revcrsible Camotcyc]c (so]id line) in וhc ש,S) plane (w is the ficld and S the polariz.ation ). The cyclc is composed of t'י''O revcr$iblc isothcrms corresponding 10 the tcmperatures J and /J, (J, > Pג) and of two
adiabatscorrcsponding 10 וhe polari:uוioחs S, and Sג (S 1 <S,; S, ,S0> ג).
Pסsitivc net work production is obtained by going anזiclockwise. The directions ofwork and heat flows along uch branch are indicatd.
J. Chem. Phys., Vol. 96, No. 4, 15 February 1992
Heisenberg picture, reads as follows:
x = i[H,XJ + ax + 2' מ(X>, ar
Vl [X,V0) 0ץ L = (X)ע '2. ] + [V,X]V0 ).
Carnot cycle
No coherence considered
Generalized Gibbs State
Canonical state: ρ =1
Ze−β H
Maximum entropy state subject to the expectation E = 〈H 〉
Generalized Gibbs state:
Aρ = e ∑k λk
ˆk
Maximum entropy state subject to the expectations〈Ak 〉 = trρAk .
and λk are Lagrange multipliers.
How to incorporate coherence
Canonical invarianceConsider the dynamics
d
dtρ = L ρ
Under what conditions on L and ρ the canonical form stays invariant with β = β (t) .
Generalized Canonical invariance:Under what conditions does the gerealized Gibbs statestay invariant to the dynamics.
Canonocal invariance
Generalized Canonocal invariance
Quantum generalized canonical invariance
Consider the closed Lie algebra of operators Ak :
∑k
[Ai ,Aj ] = CijkAk
where C kIJ is the structure factor of the group.
If the dynamics
is generated b y the Hamiltonian
ρ = −i[H ,ρ]:
H =l∑hl Al
Then the equations of motion of the algebra of operators isclosed under the dynamics. In addition the generalized canonical form is invariant to the dynamics.
.
Proof of closed Heisenberg equations
d
dtAi = i [H, Ai ] =∑
l
hl [Al , Ak ] =∑l
hl ∑k
C kli Ak
Proof of invariance of Generalized Canonical form For a closed Lie algebra:
ρ = e∑k λkAk = ∏k
eαkAk
For the dynamics ddt ρ =−i [H, ρ]:
d
dtρ ρ−1=∑
k
fk (~α,~α)Ak =−i∑l
hlAl+∑j
gj(~α)Aji
using the product form
Canonical invariance for open quantum systems
Consider the dynamics
d
dtρ = L ρ
Generalized Canonical invariance:Under what conditions does the gerealized Gibbs state stay invariant to the dynamics.
Lindblads form of the generator L :
ρ =− i
h[H,ρ]+
N2−1
∑i=1
γi
(AiρA
†i −
1
2ρAi
†Ai −1
2A†i Aiρ
).
The canonical invariance for a product form
ρ = eµ1A1eµ2A2...eµk Ak ...eµN AN
where the set A form a closed Lie algebra andµ = µ(t)
d ρ
dtρ−1 = µ1X1 + µ2e
µ1A1A2e−µ1A1 + µ3e
µ1A1eµ1A2A3e−µ1A2e−µ1 A1 + ...
Using the Baker-Housdorff relation.For canonical invariance to prevail during the evolutionL (ρ)ρ−1 also should become a linear combination of
the set of operators AFor a unitary:L (ρ)ρ−1 =−i h[H, ρ]ρ−1 =−i hH− i hρHρ−1
A1
The canonical invariance for a product formFor an open quantum system LD(ρ)ρ−1 also should
forms a linear combination of the set A. Thiscondition becomes:
LD(ρ)ρ−1 = ∑
j
(Fj ρF
†j ρ−1− 1
2(F†
j Fj + ρFj F†j ρ−1)
)Canonical invariance will be obtained If the operatorsFj F
†j and Fj ρF
†j ρ−1 are part of the set A .
Specifically for the harmonic oscillator set P2 , Q2 andD = (QP+ PQ) and the Lindblad operator F = a this
condition is met. On the contrary for F = H , H2 isnot part of the Lie algebra therefore, canonicalinvariance is not met in the case of pure dephasing.
Closed Heisenberg equations of motion for a Lie algebraof operators, leads to generalized canonical invariance for the form:
ρ = e∑k λkAk
Example
The set I, X, P, X2, P2, (XP+ PX) is a closed Lie
The Hamiltonian which is composed of members of the alebra:
H=1
2mP2+
mω(t)2
2X2
will generate a generalized canonical form for the closed algebra.
algebra
An thermodynamically significant time dependent choice of operators
The Hamiltonian H(t) = 12m P2 + 1
2mω(t)2Q2.
The Lagrangian L(t) = 12m P2− 1
2mω(t)2Q2.
The position momentum correlation C(t) = 12ω(t)(QP+ PQ).
The Casimir operator G commutes with all the operators in the algebra:
G=H2− L2− C2
h2ω2
The coherence:C o =
1
hω
√〈L2〉+ 〈C2〉
The generalized Gibbs state in a product form:
ρ =1
Zeγ a2e−β Heγ∗a†2
,
where H = hω
2 (aa† + a†a), C =−i hω
2 (a2− a†2) , L =− hω
2 (a2 + a†2) and
⟨H⟩
=hω(e2β hω −4γγ∗−1)
(eβ hω −1)2−4γγ∗⟨a2⟩=
2γ∗
(eβ hω −1)2−4γγ∗.
γ=
hω
2 (⟨L⟩
+ i⟨C⟩
)⟨L⟩2
+⟨C⟩2− ( hω
2 −⟨H⟩
)2
L2
+ C − hω
2 − H2
Inverese temperature β :
eβ hω =L⟨ ⟩2
+ C⟨ ⟩2
− H⟨ ⟩2
+ h2ω2
4
L⟨ ⟩2
+ C⟨ ⟩2
−(hω
2 − H⟨ ⟩)2
.
The expectation values of the operatorscompletely determine the GGS
squeezed themal state
The Heisenberg equations of motion for the thermodynamical set of operators are expressed as:
d
dt
H
L
C
I
(t) = ω(t)
µ −µ 0 0
−µ µ −2 0
0 2 µ 0
0 0 0 0
H
L
C
I
(t) .
(31)
The generalized Gibbs state can be reconstructed from the observables
µ is the adiabtic parameter: µ= ω/ω2.
P = µω(〈H〉 − 〈L〉
).Power
generating coherence reduces power
The equations of motion for the adiabatsFree Dynamics
The Heisenberg equation of motion for thermalization
d
dt
HLCI
(t) =
−Γ 0 0 Γ〈H〉eq0 −Γ −2ω 00 2ω −Γ 00 0 0 0
HLCI
(t)
where Γ = k↓−k↑ is the heat conductance and
k↑/k↓ = e−hω/kBT obeys detailed balance whereω = ωh/c and T = Th/c are defined for the hot or coldbath respectively.The heat current can be identified as:
Q =−Γ(〈H〉−〈H〉eq)
Otto cycle
Engine cycle [H(t),H(t ′)] 6= 0
Adiabatic efficiency
Quantum friction is the inability to stay on the energy shell
Shortcuts to adiabaticity use coherence to reach frictionless conditions.
At high temperature the efficiency at maximum power becomes:
A
C D
B
Uch
UhUc
Uhc
Canonical Invariance for Otto Cycle
The Problem:Derive a dynamical description for a driven systemcoupled to a bath beyond the adiabatic limit:
H = HS(t) + HB + HSB
Carnot cycle: The isotherms
An open system quantum control problem:State to state control:
ρi → ρf
where we have control only on the system Hamiltonian HS(t):
H = HS(t) + HB + HSB
The system dynamics is governed by:
d
dtρS = LS ρS
where LS(t) depends on the bath implicitly and HS(t).
The theory of open quantum systems
The quantum Markovian Master Equation .
A completely positive map:
Λρ = ∑j
W †j ρWj ,
where ∑j W†j Wj = I
The Gorini-Kossakowski-Lindblad-Sudarshan (GKLS) quantumMaster equation
d
dtρ =− i
h[H , ρ] +
1
2 ∑j
([Vj ρ, V†j ] + [Vj , ρV
†j ])≡− i
h[H , ρ] +L ρ .
1975
Kraus 1971
System and bath are in tensor product form in all times Lindblad 1996
G. Lindblad
Thermodynamical properties.
Davies construction: The weak coupling limit:
H = HS + HB + Hint
The system-bath interaction as Hint = ∑k Sk ⊗ Rk
One obtains the following structure of MME which isin the GKLS form
d
dtρ =−i [H , ρ] +L ρ, L ρ = ∑
k,l∑ω
L ωlk ρ
where
L ωlk ρ =
1
2h2Rkl(ω)
[Sl(ω)ρ, S†
k (ω)] + [Sl(ω), ρ S†k (ω)]
.
Here, the operators Sk (ω) originate from the Fourier decomposition
Davies 1974
Thermodynamical properties.ω- denotes the set of Bohr frequencies of H.
e i/hHt Ske−i/hHt = ∑
ωe−iωt Sk(ω),
Rkl (ω) is the Fourier transform of the bath correlation function 〈 Rk ( t)Rl 〉bathcomputed in the thermodynamic limit
Rkl (ω) = −+
∞
∞ e iωt 〈Rk (t)Rl 〉bathdt.
The derivation of Davis makes sense for a generic stationary state of the bath
and implies two properties:
1) the Hamiltonian part [H , ·] commutes with the dissipative part L ,
2) the diagonal ( in H -basis) matrix elements of ρ evolveindependently of the off-diagonal ones according to the Pauli Master Equationwith transition rates given by the Fermi Golden Rule.
No mixing of energy and coherence
Thermodynamical properties.
If additionaly the bath is a heat bath, i.e. an infinite systemin a KMS state the additional relation implies that:
3) Gibbs state ρβ = Z−1 exp−β H is a stationary solution.
4) Under the condition that only scalar operators
commute with all Sˆk (ω),Sˆ
k†(ω).
Any initial state relaxes asymptotically to the Gibbs state: The 0-Law of Thermodynamics.
The bath is able to "measure" the energy level structure of the systemand transfer heat according to the detailed balance conditionsQuantum conditions of isothermal partition
The derivation can be extended to slowly varying time-dependentHamiltonian within the range of validity of the adiabatic theorem
and an open system coupled to several heat baths at the inversetemperatures βk = 1/kBTk .
The MME in Heisenberg form:
d
dtY (t) =−i [H(t), Y (t)] +L ∗(t)Y (t) +
∂
∂ tY
, L ∗(t) = ∑k
L ∗k(t).
Each L k(t) is derived using a temporal Hamiltonian
H(t), L k(t)ρj(t) = 0 with a temporary Gibbs state
ρj(t) = Z−1j (t)exp−βj H(t).
Thermodynamical properties.
Dynamical I-law of thermodynamics
R. Alicki , J.phys. A Math. Gen. 12 L103 (1979)
Power +Heat currentSpohn, Herbert, and Joel L. Lebowitz. Adv. Chem. Phys 38 (1978): 109-142
Adiabatic limit
The task: Isothermal Dynamics
Starting from a thermal initial state ρi = e−β Hi
Transform as fast and accurate to the state: ρf = e−β Hf
while the system is in contact with a bath of temperatureT = 1/kβ
The protocol: HS (t) with HS (0) = Hi and HS (tf ) = Hf
The ProblemWe can control directly HS (t) but only indirectly the relaxation rate.We need the dissipative equation of motion with a time dependent HS (t) with a time dependent protocol.
Carnot cycle: The isotherms
Non-adiabatic open system dynamics
Time-dependent Markovian Master Eq., R. Dann, A. Levy, and R. Kosloff, Phys. Rev. A 98, 052129 (2018).The Inertial Theorem, R. Dann and R. Kosloff, arXiv:1810.12094 (2018).
How can we obtain the Master equation?
Analytic tools:
Non-Adiabatic Master Equation (NAME) Inertial theorem
NAME - The driven Non-Adiabatic Master Equation.
d
dtρS(t) = Ls ρS
We change H S(t) from H
S(0) to H S(tf ) while coupled to the bath.
Then we expect:d
dtρS(t) =−i [HS(t) + HLS(t), ρS ]
+∑k
ck(t)
(Lk(t)ρS L
†k(t)− 1
2L†
k(t)Lk(t),ρS)
.
Lk are the Lindblad jump operators.
Here HLS (t) is the time dependent Lamb shift Hamiltonian, HLS (t) = ∑k Skk ′ (α(t)F †j (t)Fj (t)).
The Non-Adiabatic Master Equation (NAME)
H(t) = HS(t) + HB + HSB . HSB = ∑k
gkAk⊗ Bk .
Following Davies’s derivation,1) Transformation to the interaction picture:
UB†(t,0)U†
SH(t)(t,0)US(t,0)UB(t,0) = ˆHSB(t) ,
Where the system evolution operatori ddt US(t) = HS(t)US(t) , US(0) = I .
.2) Second order perturbation theory lead to theMarkovian quantum master equation
d
dtρS(t) =−
∫∞
0ds trBHSB(t), [HSB(t− s), [ρS(t)⊗ ρB ]] .
Consistancy with thermodynamics
The Lindblad Jump operators:For free evolution of a static Hamiltonian:The propagator in Heisenberg form is:
U (t) = eith [HS,•]
The eigenoperators of U (t) are:
U (t)|m〉〈n|= e iωnmt |m〉〈n|
where H S|n〉 = εn|n〉 and ωnm = h
1¯(εn − εm) is the Bohr frequency.
.
The H SB can be expanded: with Fk = |m〉〈n|
HSB = ∑k
gkeiωk t Fk ⊗ Bk
leading to Fk ≡ Lk the Lindblad jump operators.
Excitations
The Lindblad Jump operators:For a driven evolution: HS(t)
We choose a time dependent operator base: Xj(t)The propagator in Heisenberg form is:
U (t) = T eih
∫ t0 ([ ˆHS(t′),•]+ ∂
∂ t′ )dt′
We find eigenoperators of U (t):
U (t)Fk = e iθk(t)Fk , U†(t)FkU(t) = e iθk(t)Fk
where Fk are time independent..HSB can be expanded in interaction frame with Fk
HSB = ∑k
gkeiθk(t)Fk ⊗ Bk
leading to Fk ≡ Lk the Lindblad jump operators.
Non Adiabatic Master Equation (NAME)
1. Weak coupling
2. Born- Markov approximation
3. Fast bath dynamics relative to the external driving
R. Dann, A. Levy, and R. Kosloff, Phys. Rev. A 98, 052129 (2018).
Lamb-shift
1. 2. 3.
How to solve the free dynamics with driving ?
Liouville space representation:
Operator basis:
Elements with inner product
The inertial theorem approximates the evolution of a quantum system, driven by anexternal field. The theorem is valid for fast driving provided the acceleration rate is small.
The Inertial Theorem, R. Dann and R. Kosloff, arXiv:1810.12094 (2018).
Inertial Theorem
Non-adiabatic driving
Time-ordering problem We want to solve this!
The Inertial theorem.
For a closed Lie algebra [Ai , Aj ] = ∑k cijk Ak
the Heisenberg equation of motion, for the set A =~v are
d
dt~v (t) =
(i[H (t) ,•
]+
∂
∂ t
)~v (t) , (1)
In a vector notation (1) becomes
d
dt~v (t) =−iM (t)~v (t) , (2)
where M is a N by N matrix with time-dependentelements and ~v is a vector of size N .If we can factor:
M (t) = Ω(t)B (~χ) . (3)
Here, Ω(t) is a time-dependent real function, andB (~χ) is a function of the constant parameters χ.
For this decomposition, the dynamics becomes
d
dθ~v (θ ) =−iB (~χ)~v (θ ) , (4)
θ ≡ θ (t) =∫ t0 dt
′Ω(t ′) is scaled time.The solution
~v (θ ) =N
∑k
ck ~Fk (~χ)e−iλkθ , (5)
where ~Fk and λk are eigenvectors andeigenvalues of B and ck are constantcoefficients. Each eigenvector ~Fk correspondsto the eigenoperator Fk .
Inertial Theorem
R. Dann and R. Kosloff, arXiv preprint arXiv:1810.12094 (2018).
Inertial solution
Geometric phase
Inertial parameter
can very slowly in time Inertial condition
R. Dann and R. Kosloff, arXiv preprint arXiv:1810.12094 (2018).
Protocol: Illustration
Experimental verification of the Inertial Theorem
Inertial solutionExperimentalNumerical
The Inertial Theorem, R. Dann and R. Kosloff, arXiv:1810.12094 (2018).Experimental Verification of the Inertial Theorem, C.K.Hu, R.Dann, et al. , arXiv:1903.00404
Experimental verification of the Inertial Theorem
Experimental Verification of the Inertial Theorem, C.K.Hu, R.Dann, et al. , arXiv:1903.00404
Example Parametric harmonic oscillator
H(t) =p2
2m+
1
2mω
2 (t) q2 , (8)
The setH(t), L(t) = p2
2m −12 ω2 (t) q2, C(t) = ω(t)
2 (qp + pq), K(t) =√
ω (t)q, J(t) = p
m√
ω(t)
is a Lie algebraThe free dynamics in terms of the vector ~v = H, L, C,K, J, IT
d
dθ~v (θ) =−iB~v (θ) (9)
with,
B = i
χ −χ 0 0 0 0−χ χ −2 0 0 00 2 χ 0 0 00 0 0 χ
2 1 00 0 0 −1 − χ
2 00 0 0 0 0 0
. (10)
Here, χ = µ = ω
ω2 , where µ is the adiabatic parameter, θ =∫ t
0 dt ′ω (t)′.
The Non-Adiabatic Master Equation (NAME)for the Harmonic oscillator(in the interaction picture)
d
dtρS(t) =−i
[HLS(t), ρS
]+
k ↑ (t)
(b†
ρS b−1
2
bb†, ρS
)+k ↓ (t)
(bρS b
†− 1
2
b†b, ρS
),
b≡ b(0) =
√mω(0)
2h
(κ + iµ)
κ
(Q +
µ + iκ
2mω(0)P
)[b, b†] = 1
µ = ω
ω2 , κ =√
4−µ2. α(t) = κ
2 ω(t) k ↑ (t)
k ↓ (t)= e− hα(t)
kBT
γna = k ↓ (α(t)) = πmα(t)J(α(t))(N(α(t)) + 1)
Comparing adiabatic to nonadiabatic rates
Nonadiabatic rate:
γna(α(t)) = πmκ
2ω(t)J(
κ
2ω(t))(N(
κ
2ω(t)) + 1)
κ =√
4−µ2
.Adiabatic rare:
γad(ω(t)) = πmω(t)J(ω(t))(N(ω(t)) + 1)
for J(ω) ∝ ω2 and low temperature:
γna(α(t)) =1
8κ
3γad(ω(t))
Slowing down the relaxation rate
Comparing non-adiabatic to adiabatic equation
Adiabatic: Lidar 2012
Both equations have time dependent Lindblad form. Thisguarantee’s complete positivity and consistency withthermodynamics
Doppler like change in frequency:
κ =√
4−µ2 , α(t) =κ
2ω(t)
slowing down the relaxation rate and changing theinstantaneous target of relaxation.
Mixing Energy and coherence: generating squeezing.
Instantaneous attractor
Carnot cycle: The isotherms
Shortcut to Equilibration (STE)
Shortcut to Equilibration of an Open Quantum System, R. Dann, A. Tobalina, and R. Kosloff, PRL 122, 250402 (2019)
Entropy change
Solving NAME for fast isothermal strokes
Change of variable in interaction representation:
d
dtρS(t) = γ ↓
(bρS b
†− 1
2
b†b, ρS
)γ ↑(b†
ρS b−1
2bb†, ρS
)(2)
we can try a solution in a generalized canonical form:
ρS(t) =1
Z (t)eγ(t)b2
eβ (t)b†beγ∗(t)b†2
2Maximum entropy subject to constraints: 〈b†b〉, 〈b† 〉, 〈b2〉
Canonical invariance: Openheim 1964, Alhassid & Levine 1978. Andersen, H. C., Oppenheim, I., Shuler, K. E., & Weiss, G. H., Jour. of Math. Phys. (1964)
Y. Alhassid and R. D. Levine, Phys. Rev. A 18, 89 (1978
Dynamics for any squeezed thermal state.
_β = k↓
(eβ −1
)+k↑
(e−β −1 + 4eβ |γ|2
), (11)
_γ =(k↓+k↑
)γ−2k↓γe
−β ,
We assume that the system is in a thermal state atinitial time, which infers γ(0) = 0. This simplifies to
~ρS (β (t) ,µ (t)) =1
Zeβ b†b(µ) . (12)
The system dynamics are described by
_β = k↓ (t)(eβ −1
)+k↑ (t)
(e−β −1
), (13)
with initial conditions β (0) = hω(0)kBT
and µ (0) = 0.
R. Dann, A. Tobalina, and R. Kosloff, PRL (2019).
For an initial thermal state
Engineering the shortcut to equilibration protocol Guessing a solution in the form of Generalized Canonical form
Control
Shortcuts to Equilibrium (STE)The shortcut protocol H
S (t) → ω(t):
CompressionExpansion
Overshoot
3 fold improvement in time
Compression
Shortcuts to Equilibrium (STE) .The fidelity F and A = − log10(1 −F ):
R. Dann, A. Tobalina, and R. Kosloff, PRL 122, 250402 (2019)
Expansion
Rapid driving costs!
STE- How much does it cost?STE Quench Adiabatic
CompressionExpansion
Efficiency:
Expansion
Compression
STE – Thermodynamic analysis
STE Quench
CompressionExpansion
The cost of shortcuts W , Su
Adiabatic shortcutsStarting on the energy shell:
〈H 〉 6= 0 , 〈L〉 = 0 , 〈C〉 = 0
Nonadiabtic dynamics generates coherence and requires extra work.Shortcuts to Adiabaticity STA retrieve this work cashing on the coherence.The shortcut duration is inversely related to the stored energy.The system entropy remains constant ∆Ssys = 0. Irreversible cost is only in the controller ∆SU ≥ 0.
The process can be classified as catalysis.
The cost of shortcuts W , Su
Shortcuts to Equilibrium (STE)Starting on the energy shell:
〈H 〉 6= 0 , 〈L〉 = 0 , 〈C〉 = 0
Nonadiabtic dynamics generates coherence and requires extra work.The coherence is dissipated generating quantum frictionThe system entropy changes ∆Ssys 6= 0. Irreversibility is inherent ∆SU > 0.
The speedup cost work and entropy production.
Quantum control of open systemsThe task is classified as a state-to-state objectiveρ iS → ρ f
S , governed by open system dynamics:
d
dtρS = L (t)ρS
Controllability: what are the conditions on thedynamics which allow to obtain thestate-to-state objective?
Constructive mechanisms of control, the problemof synthesis.
Optimal control strategies and quantum speedlimits.
Controlability of open systems.Under what conditions can an initial quantum state ρ i
S
be transformed into a final state ρ fS , employing open
systems dynamics?
1 For any initial state ρ iS couple the system to a
bath of temperature T and change theHamiltonian to H f
S until an equilibrium thermal
state ρ fS ,T , with common eigenvalue with ρ f
S .
2 Apply a fast unitary transformation US such that
ρ fS = US ρ f
S ,T U†S .
Quantum controllability of open systemsThe protocol can be achieved under the followingconditions:
1 The isolated quantum system is completelyunitary controllable, i.e. any unitarytransformation US is admissible.
2 There is a complete freedom in modifying theHamiltonian HS (t) of the system, embedded inthe bath of temperature T .
The control of a target state ρ fS requires engineering
the asymptotic invariant of L to become unitarilyequivalent to the target state In the adiabatic limitunder the Davis construction, the thermal state withthe Hamiltonian H f
S becomes the invariant of theGKLS equation
Shortcut to equilibration of a qubit.The Hamiltonian:
HS (t) = ω (t) Sz + ε (t) Sx ,
Define the Liouville vector
~v (t) = HS (t) , L(t) , C (t)T .
L(t) = ε (t) Sz −ω (t) Sxand C (t) = Ω(t) Sy ,
where Ω(t) =√
ω2 (t) + ε2 (t)
Shortcuts to Equilibrium for a qubit.For a protocol satisfying a constant adiabaticparameter
µ ≡ ωε−ωε
Ω3,
the dynamical equation in Liouville space
d~v
dt=
(˙Ω
Ω2I + Ω(t)B (µ)
)~v
, d~wdt = M (t)~w , with M (t) = Ω(t)B (µ). B (µ)
and the diagonalizing matrix V are given by
B (µ) = i
0 µ 0−µ 0 1
0 −1 0
, V =
1 −µ −µ
0 iκ −iκµ 1 1
.
Shortcuts to Equilibrium for a qubit
eigenoperators .
vi (t) =Ω(t)
Ω(0)
3
∑j ,k=1
Vik (µ (t))e−i∫ t
0 λk(t ′)
(d θ(t′)dt′
)dt ′×V −1
kj (µ (t)) vj (0) ,
Fk (µ (t) ,0) = ∑j
V −1kj (µ (t)) vj (0)
and θ (t) =∫ t
0 Ω (t ′)dt ′.
Shortcuts to Equilibrium for a qubit.Next, we utilize the inertial theorem to obtain thedynamics of the eigenoperators of the free propagators
Fk = ξ , σ , σ†
ξ (µ,0) =1
κΩ (0)
(HS (0) + µC (0)
)σ (µ,0) =
1
2κ2Ω (0)
(−µHS (0)− iκ L(0) + C (0)
)and σ† (µ,0), with corresponding eigenvaluesλ1 = 0,λ2 (t) = κ (µ (t)) and λ3 (t) =−κ(µ(t)),where
κ (µ (t)) =√
1 + µ2 (t) .
Shortcuts to Equilibrium for a qubitWe consider a system interacting with an OhmicBoson bath at temperature T , through a single spincomponent
HI = gSy ⊗ B .
The Master equation for the two-level-system:
d
dtρS (t) = k↓ (t)
(σ (µ)ρS σ
† (µ)− 1
2σ† (µ) σ , ρS
)+k↑ (t)
(σ
† (µ) ρS σ − 1
2σ (µ) σ
† (µ) , ρS)
,
k↑ (t) =2α (t)
hcN (α (t)) = k↓ (t)e−α(t)/kBT ,
α (t) = κ (t) Ω(t)
and σ (µ)≡ σ (µ (t) ,0) depends on µ (t).
Solution of the driven open systems.
ρS (t) = Z−1eγ(t)σ(µ)eß(t)ξ (µ)eγ∗(t)σ†(µ)
where Z ≡ Z (t) = tr(ρS (t)) is the partition function,with time-dependent parameters γ (t) and ß(t).The dynamics
ß =γeß
2κ2γ∗−k↓
4κ2(eß + 1
)+ |γ|2eß
16κ4
+k↑
(|γ|2 + e−ß4κ2
)(4(eß + 1
)κ2 + eß|γ|2
)64κ6
γ = k↓γ
8κ2−k↑
γ(2(1 + 2e−ß
)κ2 + |γ|2
)16κ4
,
Design of the control protocol.We assume no initial coherence:
ρS (t) = Z−1eß(t)ξ (µ)
Throughout the evolution, and the equation of motionreduces to a single differential equation
ß (t) =1
4κ2 (µ (t))
[k↑ (t)
(1 + e−ß(t)
)−k↓ (t)
(eß(t) + 1
)].
This equation is the basis for the suggested controlscheme.
Ω Ω
Control dynamics of the qubit
The control strategy goes as follows: We perform a change of variables y (t) = eß(t), and introduce an polynomial solution for y (t) which satisfies the boundary conditions of the control. The initial and target Gibbs states correspond to values
ß(0) = −hΩi /kBT0, ß(tf ) = −hΩf /kBTf , where
i = Ω(0) and ¯f = Ω(tf ) are the generalized Rabi
frequencies of the initial and final Hamiltonians. A fifth order polynomial for y is sufficient to satisfy the boundary condition, leading to
ß(t) = ln
(5
∑k=0
bktk
),
Control tasks.The control scheme allows addressing various controltasks:
1 Transformation between two equilibrium stateswith different Hamiltonians (STE).
2 Transformation between an initial nonequilibrium state to an equilibrium state,accompanied by a change in the Hamiltonian(STE).
3 Transformation to a final state which is colderthen the bath Tf < TB .
Control protocol
Thermodynamic analyasisFrom conservation of energy, heat is then determinedas Q = ∆ES −W , with
∆ES = tr(HS (tf ) ρS (tf )
)− tr
(HS (0) ρS (0)
).
Q =∫ t
0J(t ′)dt ′
J (t ′)≡ tr(HS (t ′) d
dt ′ ρS (t ′))
is the heat current..The work efficiency ηW , for an expansion processηW = W /Wadi , and for compression ηW = Wadi/Wwhere Wadi is the adiabatic work.For fast protocols tf ∼ 6
(2π/Ωref
)the efficiency
obtains values of ηW = 0.71 for expansionand ηW = 0.5 for compression,
At last: Shortcut to four stroke Carnot cycle
Carnot cycle:cyc=hchchc
arXiv:1906.06946Quantum Signatures in the Quantum Carnot CycleRoie Dann, Ronnie Kosloff
Performance of Shortcut to Carnot
minimum cycle time
0.2
0.12
o.04 .o.o4
L 1.0 1.2 1.4 1.6 1.8 2.0
Quantum eqivalence
The propagator: U = eL t
Four stroke cycle propagator:
Ucyc = UcUhcUhUch = eLc teLhc teLhteLcht
In the limit of small action: s = ||L t|| h
Ucyc = eLc t/2eLhc teLhteLchteLc t/2
Ucyc ≈ e(Lc+Lhc+Lh+Lch)t +O(s3)
Raam Uzdin, Amikam Levy, and Ronnie KosloffEquivalence of Quantum Heat Machines, and Quantum-Thermodynamic.(Phys. Rev. X 5, 031044 2015
The Voyage:Seeking for quantum open system description of the Carnot cycle
Non Adiabatic Master Equation NAME.
The inertial theorem.
Shortcuts to non unitary maps with entropy change.
Finite time quantum Carnot cycle.
Thank you
The end
Coherence generated “by” the bath.R. Dann, A. Levy, and R. Kosloff, Phys. Rev. A 98, 052129 (2018).
Protocol
Solution of the NAME for the HO
Slowing down the rate
Solution of the NAME for HOComparison to numerical simulation
R. Dann, A. Levy, and R. Kosloff, Phys. Rev. A 98, 052129 (2018).
Protocol
Solution for the propagatorFor can be solved in terms of
where and
Kosloff, R., & Rezek, Y. (2017 Entropy, 19(4), 136.
The NAME in Heisenberg form:.The Heisenberg picture is given by the equation ofmotion:
d
dtO = V † (t,0)L †(t)O .
For such a case the adjoint propagator has the form:
V †(t, t0) = Texp∫ t
t0dsL †(s)
where T is the anti-chronological time ordering.V †(t, t0) is defines by the adjoint generator L †
by the differential equation
∂
∂ tV †(t, t0) = V †(t, t0)L †(t)
.Breuer, H. P., & Petruccione, F. (2002). The theory of open quantum systems. (pg. 124-125)
......................................................................................................................... ...... .... ............ ........................... ............... ....................... .... ............ ................. . ...... .... .... ..... .. .......... .............. ... . ... ... ..... .. .......... .............. .... .... ..... .. .......... ....... . ...................... ........................................... ....................................... ................. . ...... .... .... ..... .. .......... .................. ... ... ..... .. .......... .............. .... .... ..... .. .......... ....... .