ebook img

Multi-stage quantum absorption heat pumps PDF

1.3 MB·
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Multi-stage quantum absorption heat pumps

Multi-stagequantumabsorptionheatpumps Luis A. Correa1,2,3,∗ 1SchoolofMathematicalSciences,TheUniversityofNottingham,UniversityPark,NottinghamNG72RD,UK 2IUdEAInstitutoUniversitariodeEstudiosAvanzados, UniversidaddeLaLaguna, 38203Spain 3Dpto. F´ısicaFundamental,Experimental,Electro´nicaySistemas,UniversidaddeLaLaguna,38203Spain (Dated:January17,2014) Itiswellknownthatheatpumps,whilebeingalllimitedbythesamebasicthermodynamiclaws,mayfind realizationonsystemsas‘small’and‘quantum’asathree-levelmaser.Inordertoquantitativelyassesshowthe performanceofthesedevicesscaleswiththeirsize,wedesigngeneralizedN-dimensionalidealheatpumpsby mergingN−2elementarythree-levelstages. Wesetthemtooperateintheabsorptionchillermodebetween givenhotandcoldbaths,andstudytheirmaximumachievablecoolingpowerandthecorrespondingefficiencyas afunctionofN. Whiletheefficiencyatmaximumpowerisroughlysize-independent,thepoweritselfslightly increases with the dimension, quickly saturating to a constant. Thus, interestingly, scaling up autonomous 4 quantumheatpumpsdoesnotrenderasignificantenhancementbeyondtheoptimaldouble-stageconfiguration. 1 0 PACSnumbers:03.65.-w,03.65.Yz,05.70.Ln,03.67.-a 2 n a I. INTRODUCTION becauseoftheirinherentsimplicity,thatallowstogaininsight J intotheemergenceofthelawsofthermodynamicsfromindi- 6 Anabsorptionheatpumpisamulti-purposethermaldevice vidualopenquantumsystems[33]. 1 inwhichthedrivingforceismostlyheat,ratherthanmechan- Inmanypracticalsituations,absorptionchillersmightalso icalwork[1]. ThefirstdesignthereofwaspatentedbyFerdi- be a preferable alternative to power-driven quantum devices. ] h nand Carre´ in 1860 and played a prominent role in the early Nonetheless, as in the classical case, these are usually much p development of refrigerators. Anecdotally, in 1927, Albert lessefficientandpowerfulthanthebestreversedheatengines - Einstein and Leo´ Szila´rd presented an absorption fridge run- [34], which turns the optimization of their performance into t n ningsolelyonheat,thatoperatedatnearlyconstantpressure, amatterofparamountimportance. Somestepshavebeenal- a requiringnomovingparts[2]. Inspiteoftheobviousadvan- ready undertaken in this direction, by identifying the upper u tages of absorption technologies in terms of safety, reduced bounds on a suitable figure of merit for the cooling perfor- q environmentalimpactorbetterexploitationofthefreelyavail- mance, and the design prescriptions for their saturation [35], [ able thermal resources, mechanical compression cycles are andbyproposingreservoirengineeringtechniques[36,37]in 1 preferredformostindustrialandcommercialapplications,as ordertopushthoseboundsfurther[34]. v theseareusuallymuchmoreefficient. Consequently,consid- Notwithstanding the important differences between clas- 1 erable effort has been devoted over the last decades to im- sical and quantum absorption heat pumps, it is most natu- 1 0 provetheperformanceofabsorptionheatpumps[3]inorder ral to ask whether significant enhancement can be expected 4 tomakethemcompetewiththeirpower-drivenanalogues.For fromscalingupquantumthermodynamiccyclesinto“larger” . instance,onecouldthinkofscalingupabsorptionsystemsinto multi-stage networks. In order to quantitatively assess the 1 largermulti-stageconfigurationscapableofreusingtheresid- sizescalingontheperformanceofheatpumps,weconstruc- 0 4 ualoutputheatforfurthercooling,thusboostingtheiroverall tively build generalized N-level ideal absorption devices by 1 performance[4–7]. couplingtogetherN−2elementarythree-levelcycles,some- : Itwasalreadyacknowledgedinthelate1950sthatathree what in the spirit of parallel multi-stage classical absorption v i levelmaserselectivelycoupledtotwoheatbathsandacoher- devices [1]. We then compute the efficiency at maximum X ent driving field, is a valid embodiment for a quantum heat cooling power and the maximum power itself as a function r engine [8–10], which running in reverse ultimately amounts of N for 3 ≤ N ≤ 10. While the efficiency proves to be a to a power-driven fridge. Ever since, quantum heat engines roughly size-independent, the corresponding power only in- andrefrigeratorshavebeenobjectofextensivestudy[11–20], creases mildly with the number of stages, quickly saturating including detailed experimental proposals [21, 22], and the to a constant asymptotic value. Therefore, and again in to- developmentofstrategiestoovercometheirultimatethermo- tal correspondence with the classical scenario, double-stage dynamicslimitations[19,20,23,24]. absorption heat pumps appear as the most reasonable com- On the other hand, a three-level maser operating between promise between simplicity and performance. Interestingly, heat baths only (i.e. without driving) realizes a quantum ab- thiscaseof N = 4happenstocorrespondwiththetwo-qubit sorption heat pump [25], that can function either as a fridge modelrecentlyintroducedin[26],thusmarkingitasatarget or as a ‘heat transformer’. In particular, quantum absorption forexperimentalimplementations. fridges have attracted a lot of attention [20, 26–32], mostly We also carry out an extensive numerical investigation thatsupportsthegeneralvalidityandtightnessofthemodel- independent performance bound established in [34], for all ideal multi-stage absorption refrigerators. Finally, we com- ∗Electronicaddress:[email protected] parethenon-idealeight-levelabsorptionfridgeof[27]withall 2 its ideal counterparts, when operating under the same condi- Notethatthefrequencyoftheworktransition|2(cid:105)↔|3(cid:105)isjust tions. Wefindthatidealfridgeslargelyoutperformnon-ideal ω ≡ ω − ω . We will take Eq. (1) as the defining prop- w h c ones, not only when it comes to their maximum achievable erty of an ideal heat pump. More generally, the asymptotic efficiencybut,especially,intermsoftheirmaximumcooling stationarityofenergy,translatesinto power. This paper is structured as follows: We start by providing Q˙w+Q˙h+Q˙c =0, (2) a brief review of quantum absorption heat pumps in Sec. II. whichholdsforbothidealandnon-idealdevices. Sincethere In Sec. III, we introduce our constructive scheme to build is no work involved in the operation of the pump, Eq. (2) is generalized multi-stage parallel cycles, discuss their micro- just a statement of the first law of thermodynamics. On an- scopicmodelandcorrespondingmasterequation, andobtain other note, and always provided that the dissipation is suffi- the steady-state heat currents flowing across the system. We cientlyweak, thepositivityoftherateofirreversibleentropy shall be then, in Sec. IV, in the position to quantitatively production [38, 39] makes it possible to write down the sec- comparethemaximumcoolingpowerandthecorresponding efficiency of devices with increasing N and fixed input ondlawofthermodynamicsintheform[33] thermal resources. Finally, in Sec. V, we summarize and Q˙ Q˙ Q˙ drawourconclusions. w + h + c ≤0. (3) T T T w h c Eqs.(2)and(3)encodemostoftherelevantinformationabout absorptionheatpumps. Letustakeforinstance,anidealde- II. IDEALQUANTUMABSORPTIONHEATPUMPS vice,assumingthatitworksinthechiller/heatermode,sothat Q˙ ,Q˙ >0andQ˙ <0. Then,combiningEqs.(1)and(3)one c w h Letusconsiderathree-levelsystemwithHamiltonianHˆ = arrivesto 3 (cid:126)ωc|2(cid:105) + (cid:126)ωh|3(cid:105) in its energy eigenbasis {|1(cid:105),|2(cid:105),|3(cid:105)}. We (T −T )T shall allow for a very weak dissipative interaction between 0<ωc <ωc,max ≡ (Tw−Th)Tcωh, (4) the transition |1(cid:105) ↔ |2(cid:105) and a ‘cold’ heat bath at tempera- w c h ture Tc. Likewise, we couple the transitions |2(cid:105) ↔ |3(cid:105) and whichdefinesthecoolingwindow,thatis,theregioninparam- |3(cid:105) ↔ |1(cid:105)veryweaklytoahightemperature‘work’reservoir eterspaceinwhichcooling(andheating)ispermittedbythe anda‘hot’bathrespectively, suchthatTw > Th > Tc. Once secondlaw. Forωc,max <ωc <ωh,itistheheat-transforming itsasymptoticstatebuildsup,thissystemoperatesasthemost cycle the one that dominates, thus rendering a quantum ab- fundamental quantum absorption heat pump [25], relying on sorptionheattransformer. the imbalance between the steady-state rates associated with From the cooling point of view, the ratio of the incom- thecoolingcycle|1(cid:105)→−c |2(cid:105) −→w |3(cid:105)→−h |1(cid:105),anditscomplement, ing cold heat Q˙c and the energetic cost Q˙w of its process- h w c ing,definestheefficiencyorcoefficientofperformance(COP) theheat-transformingcycle|1(cid:105)→− |3(cid:105)−→|2(cid:105)→− |1(cid:105). ε≡Q˙ /Q˙ ,whichbenchmarkstheusefuleffectofanabsorp- c w Whentheasymptoticrateofthecoolingcycleexceedsthat tionchiller. CombiningEqs.(2)and(3),onegenerallyfinds oftheheat-transformingcycle, net‘heat’perunittimeisex- tracted from the cold and work reservoirs and dumped into (T −T )T ε≤ w h c ≡ε , (5) the hot bath. That is, net energy is moved between the hot (T −T )T C h c w andcoldbathsagainst thetemperaturegradient, withtheas- sistance of the input heat coming from the work bath. De- that is, that the maximum efficiency allowed is the cor- pendingonthewhethertheusefuleffectissoughtinthecold responding Carnot COP ε [1]. It is easy to see that the C orthehotbath,wespeakofaquantumabsorptionchillerora Carnot efficiency is indeed saturated by an ideal fridge at quantum absorption heater. If on the contrary, the stationary ω → ω , where each transition equilibrates with its c c,max rate of the heat-transforming cycle exceeds that of the cool- localbathandconsequently, allheatcurrentsvanish[10]. In ing cycle, low-quality heat coming from the hot bath would general, for a non-ideal absorption chiller [i.e. a system not be upgraded to high-temperature heat leaking into the work obeying Eq. (1)], Eq. (5) would be a strict inequality [35], bath, only by means of the coupling to the cold bath. These whileEq.(4)wouldceasetohold. arethetwomodesofoperationofbothclassicalandquantum absorptionheatpumps[1]. At the steady state and given that each reservoir interacts locally with its corresponding transition [35], it is intuitively III. GENERALIZEDN-LEVELIDEALHEATPUMPS clearthatallthreeheatcurrents(work,hotandcold)mustflow at the same rate, in order to keep the average energy asymp- An autonomous thermal device based on two non- totically stationary [10, 28], i.e. one must have: Q˙ = ω q, interacting qubits was put forward in Ref. [26], which ulti- α α whereQ˙ standsforthesteady-stateheatcurrentflowingfrom matelyamountstotwoelementarythree-levelcyclessharing α bathα∈{w,h,c}intothesystem. Therefore the work transition. It can be seen that it realizes a parallel double-stageidealheatpumpandthatthepartialheatcurrents (cid:12) (cid:12) (cid:12)(cid:12)Q˙α/Q˙α(cid:48)(cid:12)(cid:12)=ωα/ωα(cid:48). (1) from each stage just add up to the total [34]. The overlap 3 |5 … 2𝜔 |5 ℎ |4 |4 |3 𝜔ℎ + 𝜔𝑐 |4 |3 𝜔 |3 ℎ |3 |2 𝜔 |2 𝑐 |2 0 |1 |1 FIG.1:Ontherighthandside,weshowthedetailofthefirstfivelevelsofageneralizedN-dimensionalidealquantumabsorptionheatpump. Thecoloredarrowsrepresentdissipativeinteractionbetweenthecorrespondingtransitionandoneofthethreeheatreservoirs: Bluestands for the cold bath, orange, for the hot bath, and red, for the high temperature work reservoir. Note that the energy difference between two consecutivelevelskeepsalternatingbetweenωc andωw = ωh−ωc. Ontheleft ha n d𝑻side,wes howthequtritbreakupofa5-levelfridge: It consistsofthreeelementarythree-levelcyclesoperatinginparallel. Thefirstandthesecondsharetheworktransition,whilethesecondand 𝒄 𝝎 thethird,shareacoldtransition. Thenextelementaryblocktobeaddedfora6-levelheatpump,wouldsharetheworktransition|4(cid:105) ↔ |5(cid:105) 𝒄 𝑻 withtheprecedingblock,while|5(cid:105)↔|6(cid:105)and|4(cid:105)↔|6(cid:105)wouldbeconnectedwiththecoldandhotbathsrespectively. 𝒉 𝑻 𝝎 𝒉 betweenthem,however,makestheirindividualcontrib𝝎utions 𝝎HˆS-𝒘B = (cid:80)αΣˆ(αN)⊗𝒘Bˆα,w𝝎here𝒘 𝑻 smallerthanwhatwouldbeexpectablefromtwoindepende𝒉nt three-level heat pumps. Once more, the corres𝒉pondence be- Σˆ(N) =(cid:88)(cid:100)N/2(cid:101)−1|2n(cid:105)(cid:104)2n+1|+|2n+1(cid:105)(cid:104)2n| (7a) tweentheclassicalandthequantumcaseisapparent[1,3,4], w𝑻 n= 1 andthequestionnaturallyarisesaboutthesizescalingofthe 𝝎 Σˆ(N) =(cid:88)𝒄N−2|n(cid:105)(cid:104)n+2|+|n+2(cid:105)(cid:104)n| (7b) coolingpowerQ˙ andtheCOPεofmulti-stagequantumab- 𝒄 h n=1 sorptionrefrigeractors. Doesthesizereallymatterinquantum Σˆ(N) =(cid:88)(cid:98)N/2(cid:99)|2n−1(cid:105)(cid:104)2n|+|2n(cid:105)(cid:104)2n−1|. (7c) c n=1 absorptioncooling? The reservoirs may consist, as usual, of an infinite collec- 𝝎 = 𝝎 − 𝝎 tion of uncoupled harmonic modes in three dimensions. We 𝒘 𝒉 𝒄 choosethebathcouplingoperatorsBˆ tobe α (cid:88) (cid:16) (cid:17) A. Microscopicmodelandmasterequation Bˆ = g bˆ +bˆ† . (8) α αµ αµ αµ µ A generalized N-level ideal heat pump may be built by mergingN−2three-levelsystems,alternativelysharingawork Withbˆαµandbˆ†αµ,wedenotetheannihilationandcreationop- oracoldtransition. InFig.1,thecaseN = 5(triple-stage)is eratorsformodeωµ√inthereservoirα,anddefinethecoupling schematically illustrated (see figure caption for details). The constantsasgαµ ≡ γαωµ,whichyieldsflatspectraldensities (cid:80) totalHamiltonianofsuchN-levelsystemreads (Jα(ω)∝ µg2αµ/ωµ).Thedissipationstrengthsγαarechosen soastodefinethelargestdynamicaltime-scaleintheproblem (cid:98)(cid:88)N/2(cid:99) (cid:100)N(cid:88)/2(cid:101)−1 (cid:40) (cid:126) (cid:41) HˆN = [(n−1)ωh+ωc]|2n(cid:105)+ nωh|2n+1(cid:105), (6) γα−1 ≫ |ωα−ωα(cid:48)|−1,k T . (9) n=1 n=1 B α This allows us to perform the Born-Markov and secular ap- where(cid:98)·(cid:99)and(cid:100)·(cid:101)standforthefloorandceilingfunctions. Let proximations in deriving a master equation for the system usmodelthesystem-bathsinteractionswithatermoftheform [40]. Under the further assumption of a preparation with 4 totally uncorrelated system and equilibrium reservoirs, one Ingeneral,forarbitraryN,onewouldbeleftwith readilyobtains Q˙(N) =ω (cid:88)(cid:100)N/2(cid:101)−1(cid:104)2n+1|L(N)(cid:37)ˆ(∞)|2n+1(cid:105) (15a) d (cid:16) (cid:17) w w n=1 w dt(cid:37)ˆN(t)= L(wN)+L(hN)+L(cN) (cid:37)ˆN(t), (10) Q˙(N) =ω (cid:88)N ((cid:100)n/2(cid:101)−1)(cid:104)n|L(N)(cid:37)ˆ(∞)|n(cid:105) (15b) h h n=3 h where(cid:37)ˆN(t)isthestateofthesystemintheinteractionpicture Q˙(N) =ω (cid:88)(cid:98)N/2(cid:99)(cid:104)2n|L(N)(cid:37)ˆ(∞)|2n(cid:105), (15c) andthedissipatorsL(N)aregivenby c c n=1 w α which provides, together with Eqs. (7) and (11), closed for- (cid:32) (cid:33) L(N)(cid:37)ˆ =Γ σˆ− (cid:37)ˆσˆ+ − 1σˆ+ σˆ− (cid:37)ˆ− 1(cid:37)ˆσˆ+ σˆ− + mulasfortheevaluationofthestationaryheatcurrents. From α α,ωα N,α N,α 2 N,α N,α 2 N,α N,α nowon,weshallfocusonthechilleroperationmode. Atfirst (cid:32) (cid:33) glance,onecouldintuitivelyexpectadifferentscalingofe.g. 1 1 Γα,−ωα σˆ+N,α(cid:37)ˆσˆ−N,α− 2σˆ−N,ασˆ+N,α(cid:37)ˆ− 2(cid:37)ˆσˆ−N,ασˆ+N,α . (11) |Q˙(cN)| for even and odd N: The addition of an even level to the heat pump provides it with an extra cold transition, so Here, the positive decay rates Γ are the diagonal ele- that a new term contributes to the cold current in Eq. (15c) α,ωα while, whentheaddedlevelisodd, thenewcontributionap- ments of the spectral correlation tensor [40], which for a pears instead in the work current of Eq. (15a). It must be three-dimensional free bosonic field in thermal equilibrium are given by Γ ∝ ω3[1 + n (ω)] > 0, with n (ω) = noted, however, that each of the already contributing terms (exp(cid:126)ω/kBTα−α1,ω)α−1. The mutuaαlly adjoint jump opαerators will always decrease as N grows, e.g. |(cid:104)2|L(c3)(cid:37)ˆ3(∞)|2(cid:105)| < σˆ+ =(σˆ− )†arejust |(cid:104)2|L(4)(cid:37)ˆ (∞)|2(cid:105)|whengoingfromEq.(13c)toEq.(14c). N,α N,α c 4 ItcanbeseenbynumericalinspectionofEqs.(15)thatin- σˆ−N,w =(cid:88)n(cid:100)N=/12(cid:101)−1|2n(cid:105)(cid:104)2n+1| (12a) cmreeanstainlgfothreanllucmubrreernotsf,sit.aeg.e|sQ˙f(rNo+m1)|ev≤en|Q˙t(oNo)|d,dwihseinndNeeidsdeveterni-. σˆ− =(cid:88)N−2|n(cid:105)(cid:104)n+2| (12b) Still for arbitrary N, one alwaαys finds |Qα˙(αN+2)|−|Q˙(αN)| ≥ 0, N,h n=1 thoughtheenhancementdecreasesasthesystemscalesup. σˆ− =(cid:88)(cid:98)N/2(cid:99)|2n−1(cid:105)(cid:104)2n| (12c) For practical purposes, one would wish to optimize the N,c n=1 incoming cold heat current to cool as quickly as possible. Given a set of heat baths T and dissipation rates γ , one ThedissipatorsinEq.(11)areinthestandardLindbladform α α has to find the cold frequency 0 ≤ ω ≤ ω that renders [39],whichguaranteesthatthereduceddynamicsofthesys- c c,max the maximum cooling power Q˙ for every fixed ω . The temiscompletelypositiveandtrace-preserving.Eachofthem c,max h correspondingCOPε isasensiblefigureofmeritforcooling individually,generatesafullycontractivedynamicsofthesys- ∗ tem towards its ‘local’ equilibrium states ∝ exp{−Hˆ /k T } and, asrigorouslyprovenin[34], itistightlyupperbounded N B α by d ε both in the single and double-stage cases, where [38],whichinturn,impliesthesecondlawofthermodynam- d+1 C d stands for the spatial dimensionality of the cold bath, i.e. icsasexpressedbyEq.(3)[30,33]. in our case d = 3. Once parameter optimization has taken placeatthesingle-stagelevel,weshallbeconcernedwiththe potentialenhancementinbothε andQ˙ asthenumberof B. Stationaryheatcurrents ∗ c,max stages,andthusthesystem’scomplexity,grows. In the light of Eqs. (10) and (11), our definition of steady-state heat current from Sec. II translates into Q˙(N) ≡ α tr{HˆNL(αN)(cid:37)ˆN(∞)}[25], wheretheasymptoticstate(cid:37)ˆN(∞)re- IV. RESULTSANDDISCUSSION sultsfromequatingtozerotheright-handsideofEq.(10). InthesimplestcaseofN =3,oneeasilyfinds A. MaximumpowerandCOPatmaximumpower Q˙(3) =ω (cid:104)3|L(3)(cid:37)ˆ (∞)|3(cid:105) (13a) w w w 3 The dependence of the maximum cooling power on the Q˙(h3) =ωh(cid:104)3|L(h3)(cid:37)ˆ3(∞)|3(cid:105) (13b) numberofstagesisillustratedinFig.2(a). Thetemperatures Q˙(3) =ω (cid:104)2|L(3)(cid:37)ˆ (∞)|2(cid:105). (13c) Tα,dissipationratesγαandthehotfrequencyωhwerechosen c c c 3 atrandomsoastofixthetimescalesforthermalfluctuations, Theadditionofasecondstagetotheheatpumpintroducesthe dissipation and system’s free evolution. Then, the optimiza- further cold and hot transitions |3(cid:105) ↔c |4(cid:105) and |2(cid:105) ↔h |4(cid:105) that tioAnsincoωucldwabseceaxrprieecdteoduftrwomithtihnetheveecno/oodlidngoswciinlldaotiwon.softhe contributetothetotalsteady-stateheatcurrentstoyield stationaryheatcurrents, Q˙ (N)scalesdifferentlyforeven c,max Q˙(4) =ω (cid:104)3|L(4)(cid:37)ˆ (∞)|3(cid:105) (14a) andodd N. Whileintheevencase,thecoolingpowerbarely w w w 4 increases, odd heat pumps significantly benefit from scaling (cid:16) (cid:17) Q˙(4) =ω (cid:104)3|L(4)(cid:37)ˆ (∞)|3(cid:105)+(cid:104)4|L(4)(cid:37)ˆ (∞)|4(cid:105) (14b) up. Onealsoseesthatanyevenconfigurationoutperformsall h h h 4 h 4 (cid:16) (cid:17) theoddones.Incontrast,thecorrespondingCOPatmaximum Q˙(4) =ω (cid:104)2|L(4)(cid:37)ˆ (∞)|2(cid:105)+(cid:104)4|L(4)(cid:37)ˆ (∞)|4(cid:105) . (14c) c c c 4 c 4 power ε proves to be roughly size-independent, as shown ∗ 5 0.10 a b 0.75 0.08 (cid:72) (cid:76) (cid:72) (cid:76) (cid:182) x (cid:42) ma 0.06 (cid:182) (cid:144) (cid:144) c, C (cid:144) (cid:144) Q (cid:160) 0.744795 0.04 0.744792 0.02 2 4 6 8 10 2 4 6 8 10 N N FIG.2: (a)(Squares)Maximumcoolingpowerasafunctionof N forgeneralizedmulti-stageheatpumpsinthechillerconfiguration,with parameters: T = 7.1×103,T = 1.57×103,T = 54.25,γ = 3.5×10−3,γ = 5.1×10−3,γ = 8.8×10−3 andω = 102.6((cid:126) = k = 1). w h c w h c h B (Rhombs)MaximumcoolingpowerforincreasingNwhenthemodesoftheworkreservoirareallsqueezedby7dB,whicheffectivelyraises theworktemperaturetoT (cid:39)1.8×104.(Triangles)MaximumcoolingpowerversusNforreversedmulti-stageCarnotengines,i.e.heatpumps w withsaturatedworktransitions. Thesolidorange,dottedredanddashedbluecurvesaremerelyaguidetotheeye,highlightingthedifferent sizescalingofdeviceswithevenandoddnumberofparallelcycles. Thedomainreachablebyabsorptionchillersisdepictedinshadedgray. (b)EfficiencyatmaximumpowerasafunctionofN,forthesameparametersasin(a). Theshadedgrayregioncorrespondthethedomainof ε allowedbyabosoniccoldbathinthreedimensionswithflatspectraldensity[34]. ∗ in Fig. 2(b). Consequently, the double-stage absorption heat sible, to benchmark the operation of realistic devices. The pump [26] tuned to operate at maximum power, appears as paradigmaticexampleistheCurzon-Ahlbornbound[41]: Al- thebestpossiblechoiceintermsofsystemoptimization. Our thoughobtainedforaspecificphenomenologicalmodelingof extensivenumericsshowthatthesepropertiesarecompletely the sources of irreversibility, i.e. the endoreversible approx- generalintherangeofapplicationofEqs.(10)and(11). imation, it emerges naturally as a limiting case in different Itisalwayspossibletoenhanceboththecoolingpowerand instances of heat engine [42, 45, 46] and describes well the thecorrespondingCOP,bysuitablyengineeringthespectrum performanceofactualpowerplants. of the environments [22–24]. Perhaps the simplest thing to The bound ε < 3ε , derived from first principles for the ∗ 4 C try is squeezing the modes of the work bath, which effec- three-level maser and the four-level double stage chiller in tively raises its temperature T and boosts the output power [34], has been also seen to hold [35] in the non-ideal eight- w of the heat pump [34]. The Carnot bound ε is also effec- levelrefrigeratormodelof[27],andevensucceedsinlimiting C tivelyincreased,whichallowsforlargerefficienciesatmaxi- from above the performance of classical endoreversible ab- mumpower. Theeffectofreservoirsqueezinginthecooling sorptionfridges[47]. Inordertoestablishwhetheritcontin- power is also illustrated in Fig. 2(a) for a squeezing of 7 dB ues to hold for multi-stage ideal absorption refrigerators be- or r (cid:39) 0.8 (see figure caption for details). As T increases, yondthecasesN =3andN =4,weperformedglobalnumer- w the corresponding contact transitions tend to saturate. In the ical optimization of ε over ω , T , γ and N ∈ [3,10]. An ∗ h α α limitofT →∞,onerealizesgeneralizedmulti-stageCarnot histogram on the resulting COP at maximum power for 105 w enginesrunninginreverse[10]andthus,abandonstherealm randomlydrawnrefrigeratorsispresentedinFig.3(a),clearly of absorption chillers. The limiting cooling powers are also showingthattheboundholdstightlyregardlessofthesizeof depictedinFig.2(a). thesystem. B. BoundontheCOPatmaximumpower C. Idealvs.non-idealabsorptionfridges ThequestfortightboundstotheCOPatmaximumpower Alotofinteresthasbeengeneratedbythenon-idealeight- for heat engines and refrigerators has attracted a lot of inter- level (three-qubit) absorption fridge, first introduced in [27], est over the last decades, both in the classical [41–43] and including experimental proposals for its implementation on quantum [16, 35, 44, 45] scenarios. By making minimal as- super-conductingqubits[48]andquantumdots[49]. Itisle- sumptionsaboutthedominatingsourcesofirreversibility,the gitimatetoaskhowdoesitcompare,intermsofabsolutecool- aimistoderivepracticalbounds,aswidelyapplicableaspos- ingpower,withitsidealeight-levelcounterpart. Interestingly, 6 aa bb 0.001 33 44 (cid:72)(cid:72) (cid:76)(cid:76) (cid:72)(cid:72) (cid:76)(cid:76) (cid:144)(cid:144) 10(cid:45)5 (cid:160) Q c 10(cid:45)7 10(cid:45)9 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 (cid:182) (cid:182) (cid:182) (cid:182) (cid:42) C C FIG.3: (a)HistogramoftheCOPatmaximumpowerfor105 multi-stageabsorptionfridgeswithω ,T ,γ andN chosenatrandom. The h α α blueverticallinemarksthe 43εCultimate(cid:144)boundonε∗[34].(b)Coolingpowerversusnormalizedefficien(cid:144)cyforthethree-qubitnon-idealfridge of[27](solidorange),andforaneight-levelidealchiller(dottedblue)forω =60,T =130,T =60,T =5andγ =10−3.Thethree-body w w h c α interactionstrengthbetweenthequbitsinthenon-idealmodelwassettog=0.1.Again,naturalunitsareassumed. allidealmulti-stagefridges(eventhethree-levelmaser), can thesizeofthesystem. beseentolargelyoutperformtheeightlevelnon-idealchiller, Even if the heat currents were formally given in Eqs. (11) typicallybyseveralordersofmagnitude. and (15), the explicit formulas for any N are quite involved In Fig. 3(b) this is illustrated by plotting together the per- and nothing but insightful. The whole analysis is therefore formancecharacteristicsofboththenon-idealthree-qubitand based on extensive numerics. It must be acknowledged as theideal‘sextuple-stage’(N =8)chiller,forthesamechoice well, that the working substance of an ideal multi-stage ab- of parameters. A closed curve in the Q˙ – ε plane is indica- sorption refrigerator is generally a rather artificial quantum c tive of irreversibility and prevents the refrigerator from ever system. Interestinglyenough,thecaseofN = 4,thatappears realizingtheCarnotefficiency. TheCOPatmaximumpower as the optimal compromise between performance and com- is roughly the same in both cases (close to its ultimate 3ε plexity,alsoturnsouttocorrespondtothetwo-qubitmodelof 4 C bound), but the power Q˙ of the ideal fridge exceeds that [26]. This suggests that it should be considered for practical c,max ofthenon-idealonebyoverthreeordersofmagnitude. applicationsofabsorptioncoolingtoquantumtechnologies. This suggests that any realistic application of absorption The double-stage fridge (more generally, any ideal heat technologies to quantum cooling, should rely upon physical pump) was seen to outperform by several orders of magni- implementations of ideal (multi-stage) heat pumps, out of tudethemaximumcoolingpowerdeliveredbytheonlyavail- which the double-stage design [26] stands out with its opti- able (non-ideal) alternative model: The three-qubit refriger- malbalancebetweenperformanceandsimplicity. ator [27]. Arrangements of superconducting qubits [48] or coupled quantum dots [49], similar to those proposed to re- alizethethree-qubitdevice,couldalsobegoodcandidatesto supportthemorepromisingdouble-stageabsorptionchillers. V. CONCLUSIONS Further enhancement of both the power and the efficiency can always be achieved by applying quantum reservoir Wehavestudiedthesizescalingofthepowerandtheeffi- engineering techniques [36, 37] meant to render exploitable ciencyofparallelmulti-stagequantumabsorptionheatpumps. nonequilibrium environmental fluctuations, thus raising Whenworkinginthechillerconfiguration, heatpumpscom- absorptiontechnologiestothelevelofthebestcompression- prisedofanoddandevennumberstages,scaledifferentlyin basedquantumthermodynamiccycles. terms of their maximum achievable cooling power: Devices with even N always provide a larger power, which is almost size-independent,whileinheatpumpswithoddN,themaxi- mumpowerincreaseswiththenumberofstages,quicklysat- urating to a constant asymptotic value. Contrarily, the effi- Acknowledgements ciency at maximum power does not significantly scale with N and, upon global optimization over all free parameters, it TheauthorisgratefultoG.Adesso,J.P.Palao,D.Alonso, saturatestothesamemodel-independentbound,regardlessof K. Hovhannisyan, P. Skrzypczyk, N. Brunner and P. Ramos- 7 Garc´ıa for fruitful discussions and useful comments. This Canary Islands Government through the ACIISI fellowships workwassupportedbytheCOSTActionMP1006andbythe (85%cofundedbyEuropeanSocialFund). [1] J.M.GordonandK.C.Ng,Coolthermodynamics(Cambridge 056130(2001). internationalsciencepublishingCambridge,2000). [26] A.LevyandR.Kosloff,Phys.Rev.Lett.108,070604(2012). [2] A.EinsteinandL.Szilard,Appl.USPatent16(1927). [27] N.Linden,S.Popescu,andP.Skrzypczyk,Phys.Rev.Lett.105, [3] P. Srikhirin, S. Aphornratana, and S. Chungpaibulpatana, Re- 130401(2010). newableandsustainableenergyreviews5,343(2001). [28] P.Skrzypczyk,N.Brunner,N.Linden,andS.Popescu,J.Phys. [4] F.Ziegler,R.Kahn,F.Summerer,andG.Alefeld,International A:Math.Theor.44,492002(2011). journalofrefrigeration16,301(1993). [29] N.Brunner, N.Linden, S.Popescu, andP.Skrzypczyk, Phys. [5] M.Lawson,R.Lithgow,andG.Vliet,ASHRAETrans.;(United Rev.E85,051117(2012). States)88(1982). [30] A. Levy, R. Alicki, and R. Kosloff, Phys. Rev. E 85, 061126 [6] R.DeVault,J.Marsala,etal.,Tech.Rep.,OakRidgeNational (2012). Lab.,TN(USA)(1990). [31] N. Brunner, M. Huber, N. Linden, S. Popescu, R. Silva, and [7] G.Grossman,A.Zaltash,P.Adcock,andR.DeVault,ASHRAE P.Skrzypczyk,e-printarXiv:1305.6009v1. journal37(1995). [32] A.MariandJ.Eisert,Phys.Rev.Lett.108,120602(2012). [8] H.E.D.ScovilandE.O.Schulz-DuBois, Phys.Rev.Lett.2, [33] R.Kosloff,Entropy15,2100(2013),ISSN1099-4300. 262(1959). [34] L. A. Correa, J. P. Palao, D. Alonso, and G. Adesso, e-print [9] J.Geusic,E.Bois,R.DeGrasse,andH.Scovil,JournalofAp- arXiv:1308.4174. pliedPhysics30,1113(1959),ISSN0021-8979. [35] L.A.Correa,J.P.Palao,G.Adesso,andD.Alonso,Phys.Rev. [10] J.E.Geusic,E.O.Schulz-DuBios,andH.E.D.Scovil,Phys. E87,042131(2013). Rev.156,343(1967). [36] N.Lu¨tkenhaus,J.I.Cirac,andP.Zoller,Phys.Rev.A57,548 [11] R.KosloffandA.Levy,e-printarXiv:1310.0683. (1998). [12] H.T.Quan,Y.-x.Liu,C.P.Sun,andF.Nori,Phys.Rev.E76, [37] E.ShahmoonandG.Kurizki, PhysicalReviewA87, 013841 031105(2007). (2013). [13] H.T.Quan,Phys.Rev.E79,041129(2009). [38] H.Spohn,JournalofMathematicalPhysics19,1227(1978). [14] E.GevaandR.Kosloff,J.Chem.Phys.104,7681(1996). [39] G.Lindblad,Comm.Math.Phys.48,119(1976). [15] D.Gelbwaser-Klimovsky,R.Alicki,andG.Kurizki,Phys.Rev. [40] H. Breuer and F. Petruccione, The Theory of Open Quantum E87,012140(2013). Systems(OxfordUniversityPress,USA,2002). [16] A. E. Allahverdyan, K. Hovhannisyan, and G. Mahler, Phys. [41] F.CurzonandB.Ahlborn,Am.J.Phys.43,22(1975). Rev.E81,051129(2010). [42] M.Esposito,R.Kawai,K.Lindenberg,andC.VandenBroeck, [17] Y.ZhouandD.Segal,Phys.Rev.E82,011120(2010). Phys.Rev.Lett.105,150603(2010). [18] S.W.Kim,T.Sagawa,S.DeLiberato,andM.Ueda,Phys.Rev. [43] S.Velasco, J.M.M.Roco, A.Medina, andA.C.Herna´ndez, Lett.106,070401(2011). Phys.Rev.Lett.78,3241(1997). [19] E.BoukobzaandH.Ritsch,Phys.Rev.A87,063845(2013). [44] S.Abe,Phys.Rev.E83,041117(2011). [20] D. Gelbwaser-Klimovsky, R. Alicki, and G. Kurizki, e-print [45] J. Wang, J. He, and Z. Wu, Physical Review E 85, 031145 arXiv:1309.5716. (2012). [21] O.Abah,J.Roßnagel,G.Jacob,S.Deffner,F.Schmidt-Kaler, [46] M.Esposito,K.Lindenberg,andC.VandenBroeck,Phys.Rev. K.Singer,andE.Lutz,Phys.Rev.Lett.109,203006(2012). Lett.102,130602(2009). [22] J.Roßnagel,O.Abah,F.Schmidt-Kaler,K.Singer,andE.Lutz, [47] J.ChenandZ.Yan,Phys.Rev.A39,4140(1989). e-printarXiv:1308.5935. [48] Y.-X.ChenandS.-W.Li,EPL(EurophysicsLetters)97,40003 [23] X.L.Huang,T.Wang,andX.X.Yi,Phys.Rev.E86,051105 (2012). (2012). [49] D. Venturelli, R. Fazio, and V. Giovannetti, Phys. Rev. Lett. [24] O.AbahandE.Lutz,e-printarXiv:1303.6558. 110,256801(2013). [25] J. P. Palao, R. Kosloff, and J. M. Gordon, Phys. Rev. E 64,

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.