ebook img

Sharp Continuity Bounds for Entropy and Conditional Entropy PDF

0.07 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 Sharp Continuity Bounds for Entropy and Conditional Entropy

SCIENCE CHINA Physics, Mechanics & Astronomy 1 Vol. No.: doi: Sharp Continuity Bounds forEntropy andConditional Entropy ZhihuaChen1,Zhihao Ma2†, IsmailNikoufar3, andShao-Ming Fei4,5* 7 1DepartmentofAppliedMathematics,ZhejiangUniversityofTechnology,Hangzhou,310014,China; 1 2SchoolofMathematicalSciences,ShanghaiJiaotongUniversity,Shanghai,200240,China; 0 3DepartmentofMathematics,PayameNoorUniversity,19395-3697Tehran,Iran; 2 4SchoolofMathematicalSciences,CapitalNormalUniversity,Beijing100048,China 5Max-Planck-InstituteforMathematicsintheSciences,04103Leipzig,Germany n DearEditor, tropyto define thelinear discord. Theyfoundthatthe linear a J The von Neumann entropy captures many operational discordhasdeepconnectionwiththeoriginaldiscorddefined 0 quantities in the quantum information theory such as quan- byvonNeumannentropy. Moreover,theygavetheanalytical 1 tum capacity of the communication channel. Von Neumann formulaforarbitrary2⊗d state of the linear discord. How- entropyiscontinuousandisrepresentedbyFannesinequality, ever,a questionstillremainsopen: if twostates are close, is ] h whichwasoriginallygivenin[1]. Quantumcorrelationssuch theirlineardiscordalso close to eachother? Inotherwords, p as entanglement and quantum discord , et al., are important isthelineardiscordcontinuous? Fortheoriginaldiscord,the t- resourcesinquantuminformationprocessing. Inthelastyear answerisaffirmative,see[10]. Forthelineardiscord,thereis n enormous progress on the generation, concentration, detec- noansweryet. Henceitisworthwhiletostudythecontinuity a u tionandquantificationofentanglementhasbeenachieved[2]. ofconditionallinearentropy. q Fannes inequality has many applications in the quantum in- We have two aims in this work: first, we study the conti- [ formation theory, such as the investigation of continuity of nuity estimation of the Re´nyientropy and presenta tight in- 1 entanglementmeasures,includingentanglementofformation, equalityrelatingtheRe´nyientropydifferenceoftwoquantum v relativeentropyofentanglement,squashedentanglementand states to their trace norm distance, which includes the sharp 8 conditionalentanglementofmutualinformation,andthecon- FannesinequalityforvonNeumannentropyasaspecialcase. 9 tinuity of quantum channel capacities [3]. Recently Fannes Second,westudythecontinuityofconditionallinearentropy, 3 inequality was improved to get a sharp one and it was also andproveausefulpropertyforameasureofthequantumcor- 2 0 generalizedtoTsallisentropy[4,5]. relation:lineardiscord. 1. However, in non-asymptotic settings, the natural quanti- ThevonNeumannentropyofaquantumstateρisdefined by 0 ties that arise are Re´nyi entropies [6] and the properties of 7 Re´nyi entropieswere also investigatedin many papers, such S(ρ):=−Tr[ρlog2ρ]. (1) 1 as[7]. Re´nyientropieshavemanyapplications,asinthecase For the classical probability distributions, the von Neumann : v of one-shot problems, typically arising in cryptographicset- entropyreducestotheShannonentropy, Xi tings,themin-andmax-entropiesarewidelyused[6]. In[8], the authors found that Re´nyi-2 entropy was a proper mea- d d ar sure of information for any multimode Gaussian state, and H(p):=XH(pi)=−Xpilog2pi, (2) they defined and analyzed the measures of Gaussian entan- i i glement and quantum correlation by using Re´nyi-2 entropy, where p:=(p)=(p ,p ,...,p )isad-dimensionalprobabil- i 1 2 d andfounditspropertiessuchasmonogamy. Inourwork,we ityvector,p >0, d p =1andH(p):=−p log p. study the continuity property of Re´nyi-α entropy, which in- i Pi i i i 2 i In[1]Fannesprovedhisfamousinequalityforthecontinu- cludes Re´nyi-2 entropy as a special case. Our result is also ityofthevonNeumannentropy, usefulinstudyingthecontinuityoftheentanglementmeasure oftheGaussianstateinquantumharmonicsystems. |S(ρ)−S(σ)|62Tlog (d)−2Tlog (2T), (3) 2 2 Ontheotherhand,theauthorsfoundthatTsalli-2entropy (i.e.,linearentropy)wasnaturaltodefinethemeasureofquan- whereT := ||ρ−σ||1 ishalfofthetracenormdistancebetween 2 tum correlation for the the discrete system [9]. They called thestatesρandσ,||ρ−σ|| =Tr[|ρ−σ|],and|X|:= (X)†(X) 1 p thismeasureaslineardiscordandusedconditionallinearen- denotesthe absolute value of an operator X. ObviouslyT ∈ 1Correspondingauthor(email:†[email protected],*[email protected]) (cid:13)c ScienceChinaPressandSpringer-VerlagBerlinHeidelberg phys.scichina.com link.springer.com SciChina-PhysMechAstron () Vol. No. -2 [0,1].Theinequality(3)isvalidfor06T 61/2e,whereeis Theorem 3. For bipartite quantumstates ρ and σ , if AB AB Euler’snumber.Theinequality(3)isfurtherimprovedtobea ǫ :=||ρ −σ || <1,then AB AB 1 sharponebyAudenaert[4]: |D (ρ )−D (σ )|68ǫ+4h (ǫ,1−ǫ). (9) 2 AB 2 AB 2 |S(ρ)−S(σ)|6Tlog (d−1)−H((T,1−T)). (4) 2 In summary, we haveinvestigatedthe continuityof Re´nyi TheRe´nyientropyisamoregeneralformofthevonNeu- entropyandconditionallinearentropy. Throughthecontinu- mannentropy, ity of conditionallinear entropy,the continuityof linear dis- cordhasalsobeenobtained,whichmeansthatthelineardis- 1 H (ρ):= log[Tr(ρα)], α>0, (5) cordvaries as the quantumstate changescontinuously. This α 1−α fact can guarantee that the errors in state tomograph would notsignificantlyaffectthe resultof the quantumcorrelations whenαgoestoone,Re´nyientropybecomesthevonNeumann inthestate. AsthesharpFannesinequalityisthespecialcase entropy. InthefollowingweshowthatfortheRe´nyientropy, ofourtheoremaboutthecontinuityofRe´nyientropy,ourre- animprovedsharpFannes-typeinequalityexists. sultscanalsobeusedtoverifythecontinuityofentanglement Theorem1.Foralld-dimensionalquantumstatesρandσ, measuresforcontinuousvariablequantumstates. dα−1 |H (ρ)−H (σ)|6 [1−(1−T)α−(d−1)1−αTα], α>1, α α 1−α Acknowledgments ThisworkissupportedbytheNSFCun- (6) dernumber11371247,11275131,11675113and11571313. 1 |H (ρ)−H (σ)|6 [1−(1−T)α−(d−1)1−αTα], α<1, α α 1−α (7) 1 M.Fannes,Commun. Math. Phys. 31: 291–294(1973);R.Alicki,M. whereT is the trace normdistanceof ρand σ. See proofin Fannes,J.Phys.A:Math.Gen.37(2004)L55. Appendix. 2 MaimaitiW,LiZ,ChesiS,etal,ScienceChinaPhysics,Mechanicsand WeinvestigatedthecontinuityestimationoftheRe´nyien- Astronomy,58(5),50309(2015). 3 D.Yang,M.Horodecki,Z.D.Wang,AnAdditiveandOperationalEn- tropy,bypresentinganinequalitywhichrelatestheRe´nyien- tanglementMeasure:ConditionalEntanglementofMutualInformation, tropydifferenceoftwoquantumstatestotheirtracenormdis- Phys.Rev.Lett.101:140501(2008). tance. In our inequality, equality can be attained for every 4 K.M.R.Audenaert,J.Phys.A:Math.Theor.40(2007)8127. prescribed value of the trace norm distance. It is direct to 5 AlexeyE.Rastegin,LettersinMathematicalPhysics94(3)(2009). verifythat for α → 1, our inequality(6) and(7) give rise to 6 M. Tomamichel, PhD thesis, Department of Physics, ETH Zurich, thesharpFannesinequalityforvonNeumannentropy. Ithas arXiv:1203.2142(2012). potentialapplicationsininvestigatingthecontinuityofentan- 7 Lin S M, Tomamichel M, Quantum Information Processing, 2015, 14(4):1501-1512. glementmeasureandmoregeneralcorrelationsformultimode 8 GerardoAdesso,DavideGirolami, AlessioSerafini,Phys. Rev. Lett. Gaussianstates,sinceRe´nyi-2entropyisaproperinformation 109,190502(2012). measureforthiskindofstate[8]. 9 ZhihaoMa,ZhihuaChen,FelipeFernandesFanchini,Shao-MingFei, BesidesRe´nyientropy,linearentropyisalsousedtomea- ScientificReports5,10262(2015). sure quantum correlations, such as linear discord [9], which 10 ZhengjunXi,Xiao-MingLu,XiaoguangWang,YongmingLi,J.Phys. is defined as the minimal difference of the two conditional A:Math.Theor.44,375301(2011). linear entropy,beforeand after the localprojectivemeasure- 11 M.M.-Lennert,F.Dupuis,O.Szehr,S.Fehr,M.Tomamichel,J.Math. ment,D (ρ ) := min(S (A|B)−S (P|B)),whereS (A|B)is Phys.54,122203(2013). 2 AB 2 2 i 2 Pi 12 I.Nikoufar,AdvancesinMathematics259(2014),376-383. theconditionallinearentropyoftheoriginalstate ρAB, while 13 M.A.NielsenandI.L.Chuang,QuantumComputationandQuantumIn- S2(Pi|B)istheconditionallinearentropyofthepostmeasure- formation(CambridgeUniversityPress,Cambridge,UK,10thAnniver- mentstateafterlocalmeasurementP,andtheminimumruns saryEdition,2010). i overalllocalprojectionmeasurementsP. Thereforeitisalso 14 K.M.R.Audenaert,J.Math.Phys.48(2007),083507. i importantto study the continuityof the linear entropy,espe- ciallyconditionallinearentropy. Itcanhelpusgettheconti- nuityofthelineardiscord. Wecanprovethefollowingconclusion: conditionallinear entropyiscontinuous,seeproofinAppendix. Theorem 2. For bipartite quantumstates ρ and σ , if AB AB ǫ :=||ρ −σ || <1,thenthefollowinginequalityholds, AB AB 1 |S (ρ |ρ )−S (σ |σ )|64ǫ+2h (ǫ,1−ǫ). (8) 2 AB B 2 AB B 2 Byusingthemethodof[10], itisstraightforwardtoshow thatthelineardiscordD (ρ )isalsocontinuous: 2 AB SciChina-PhysMechAstron () Vol. No. -3 Appendix =S (1− p −T)−S (1−p −T(1−s )) α 1 α 1 1 ProofofTheorem1: −S (T(1−s )φ) α 1 Lemma1.1.Foralld-dimensionalquantumstatesρandσ, ,∆. |H (ρ)−H (σ)|6dα−1|S (ρ)−S (σ)| α>1, (10) α α α α As S (T(1− s )φ) gets its maximumwhen φ is the uniform α 1 distribution=(1,1,··· ,1)/(d−2),∆hastheminimum, |H (ρ)−H (σ)|6|S (ρ)−S (σ)| α<1. (11) α α α α whereS (ρ):=[Tr(ρα)−1]. ∆=Sα(1−p1−T)−Sα(1− p1−T(1−s1)) α TheaboveinequalitiesareobtainedbyusingCauchymean −(d−2)1−αS (T(1−s )). α 1 valuetheorem. (c). From Lemma1.2.Foralld-dimensionalquantumstatesρandσ, ∂∆ |Sα(ρ)−Sα(σ)|61−(1−T)α−(d−1)1−αTα. (12) ∂s =−Tα[(1−p1−T(1−s1))α−1−(d−2)1−α(T(1−s1))1−α]=0, 1 Lemma1.3. Forallprobabilitydistributions p = (p)and weget i q=(qi),thefollowinginequalityholds: T(1−s )= (1−p1)(d−2) ≡ω. (15) 1 (d−1) |Sα(p)−Sα(q)|61−(1−T)α−(d−1)1−αTα, (13) When 0 < T < ω, there is no local minimum of ∆ from (15). For T(1 − s ) = T, i.e. s = 0, ∆ has a minimum d 1 1 wSh(eqr)eT== 12[(iPq=1)|αpi−−qq]i|,=Sα(p)S=(qPi)[,(ppi)α>− p0i,]q= P>i S0α(apni)d, i−m(dum−S2)α((1−pα1)S+αT(T))−. SThα(epre1f)o−re(Sdα−(q2))−1−αSSαα((pT))g.etMs oitrseomvienr-, Piαpi =Pi qPii =1i,Sα(pii):=[(Ppii)αα−pii]. i i dpu1,ewtohethnatp1Sα=(p11−+T(dd−)−1−2)TS, αS(αp(1q))i−s aSαd(epc)regaestisngitsfumncintiiomnuomf Let λi, i = 1,2,...,d, be the eigenvalues of ρ, one has (1− dT−2)α−(1− (dd−−12)T)α−(d−2)1−αTα. Sα(ρ) = P[(λi)α−λi] := PSα(λi),whereSα(λi) := [(λi)α− When ω 6 T 6 1− p1, (15) can be satisfied and ∆ gets i i itsminimum(1− p −T)α−(1−p )α/(d−1)α−1. Therefore λ]. 1 1 i S (q)−S (p)getsitsminimum Firstly, we prove Lemma 1.3. Then Lemma 1.2 is also α α provedforthediagonalquantumstatesρandσ. (1−p )α ProofofLemma1.3:Letq= p+δ+−δ−,whereδ+ =(δ+i) −Sα(p1)+Sα(p1+T)+(1−p1−T)α− (d−1)1α−1. and δ− = (δ−) are two vectors such that δ+ > 0, δ− > 0, i i i iLe=mm1,a21,..3..,idn,thδr+ee·cδa−se=sa0cc,oPrdiiδn+igt=othTe.vaWlueesproofvαe:(1α3<) o1f, tThhaendzeerriov.atHiveenocfe,thwehaebnovpe1f=orm1u−laTw,iSthα(rqe)sp−ecStαto(pp)1giestsleistss 16α<2andα>2. minimum1−(1−T)α−(d−1)1−αTα. (CaseI)α<1: SinceSα(x)−Sα(x−T)isadecreasingfunctionof xand (a).S (p)isconcave,S (p+δ+−δ−)−S (p)isaconcave 1>1− T , α α α d−2 function with respect to δ+. It gets its minimumat a certain T (d−1)T point,say,δ+ =e1. 1−(1−T)α 6(1− )α−(1− )α. ThenT =e1,p=(p ,(1−p )r),q=(p +T,(1−p )r−Ts), d−2 d−2 1 1 1 1 wherer and sared−1dimensionalprobabilityvectorssuch Therefore1−(1−T)α −(d −1)1−αTα 6 (1− T )α −(1− that p1+T 61,(1−p1)r−Ts>0,andTs=δ−. Wehave (d−1)T)α−(d−2)1−αTαandthen1−(1−T)α−(dd−2−1)1−αTα d−2 istheminimumofS (q)−S (p). S (q)−S (p)=S (p +T)−S (p ) α α α α α 1 α 1 (CaseII)16α<2: +Sα((1−p1)r−Ts)−Sα((1−p1)r). (a). Sα(p)−Sα(q) = Sα(p)−Sα(p+δ+ −δ−)isconcave withrespecttoδ+. Takeδ+ =e1.Wehave (b). Denote(1−p )r−Ts=(1−p −T)η,then 1 1 S (p)−S (q)=S (p )−S (p +T) α α α 1 α 1 S ((1− p )r−Ts)−S ((1−p )r) (14) α 1 α 1 +S ((1−p )r)−S ((1−p )r−Ts), α 1 α 1 =S ((1−p −T)η)−S ((1−p −T)η+Ts). α 1 α 1 where SinceS (x)−S (x+y)isconcaveandamonotonouslyin- α α creasingfunctionofx,theright-handsideof(14)getsitsmin- Sα((1−p1)r)−Sα((1−p1)r−Ts) imumatcertainpointofη,say,η=e1.Lets=(s1,(1−s1)φ) =Sα((1−p1−T)η+Ts)−Sα((1−p1−T)η). withφad−2dimensionalprobabilityvector,wehave (b).S (x+y)−S (x)isconcavewithrespecttoxforα<2, α α S ((1−p −T)η)−S ((1− p −T)η+Ts) S ((1−p −T)η+Ts)−S ((1−p −T)η)getsitsminimum α 1 α 1 α 1 α 1 SciChina-PhysMechAstron () Vol. No. -4 atη = e1. Let s = (s ,(1− s )φ)withφad−2dimensional Definition WedefinetheTsalliαrelativeentropyas: 1 1 probabilityvector.Weget 1 Tˆ (ρ||σ)= (Tr(ρα)−Tr(ρσα−1)), (16) S ((1−p −T)η+Ts)−S ((1−p −T)η) α α−1 α 1 α 1 =S (1−p −T(1−s ))+S (T(1−s )φ) here α ∈ (0,∞). When α goes to one, the Tsalli α relative α 1 1 α 1 −S (1−p −T),∆. entropy becomes the (quantum) relative entropy, S(ρ||σ) = α 1 −Tr(ρlogσ)−S(ρ). Whenφ= (1,1,···,1)/(d−2),S (T(1−s )φ)getsitsmini- Lemma 2. The relative entropydefined by (16) is jointly α 1 mum,and∆getsitsminimum,thatisS (1−p −T(1−s ))− convexforα∈[0,1)∪(1,2]. α 1 1 S (1− p −T)+(d−2)1−αS (T(1−s )). Proof. Thefunction f(t) = 1 (1−tα−1)isaconvexfunc- α 1 α 1 α−1 (c)From ∂∆ =0,wehavetheformula(15)again. tionforα∈[0,1)∪(1,2]whent>0,then When0<∂Ts1 <ω,∆hasnolocalminimum.ForT(1−s )= T,∆hasaminimum(d−2)(1−α)S (T). S (p)−S (q)get1sits g(Lρ,Rσ)= Lρ1/2f(L−ρ1/2RσL−ρ1/2)Lρ1/2 α α α minimumSα(p1)−Sα(p1+T)+(d−2)1−αSα(T),whichtakes = 1 (L −L2−αRα−1) theminimumvalue(1− (d−1)T)α−(1− T )α+(d−2)1−αTα α−1 ρ ρ σ d−2 d−2 at p1 =1− (dd−−12)T. isjointlyconvex(cf.[12]). Itfollowsthat For ω 6 T 6 1 − p , at T(1− s ) = ω, ∆ gets its min- 1 1 imum ((d1−−1p)1α)−α1 −(1− p1 −T)α. Sα(q)−Sα(p) gets its min- (ρ,σ)7−→hg(Lρ,Rσ)(X),Xi=Tr(X∗g(Lρ,Rσ)(X)) imum Sα(p1) − Sα(p1 + T) − (1 − p1 − T)α + ((d1−−1p)1α)−α1 = is also jointly convex on ρ,σ, where X is an arbitrary oper- (1−T)α+(d−1)1−αTα−1at p =1−T. Therefore ator and h·,·i is the Hilbert-Schmidt inner product. Taking 1 X =ρα−21 wehave (1−T)α+(d−1)1−αTα−1 <(1− (d−1)T)α−(1− T )α+(d−2)1−αTα hg(Lρ,Rσ)(ρα−21),ρα2−1i d−2 d−2 1 and(13)isvalidwhen16α<2. = α−1Tr(ρα−21(Lρ−L2ρ−αRασ−1)(ρα−21)) (CaseIII)α > 2 :FollowingthesamestepasCaseII,we =Tˆα(ρ,σ). get that S ((1− p )r)−S ((1− p )r−Ts) is concave with α 1 α 1 Thiscompletestheproof. respecttos. HenceS ((1−p )r)−S ((1−p )r−Ts)getsits α 1 α 1 minimuminoneoftheextremepointsofs,say,s=e1. Corollary1. ThelinearentropyS2(ρ)andtheconditional Letr =(r ,(1−r )φ)withφad−2dimensionalprobability linearentropyS2(A|B)areconcave. 1 1 Lemma 3. (Projective measurements increase entropy) vector.Then SupposethatP isacompletesetoforthogonalprojectorsand i S ((1−p )r)−S ((1−p )r−Ts) ρ is a density operator. Then the linear entropy of the state α 1 α 1 =Sα((1−p1)r1)−Sα((1−p1)r1−T),∇. ρ′ :=Pi PiρPiafterthemeasurementwillnotdecrease,thatis, S (ρ′)>S (ρ). Since 2 2 Proof.WeknowthatthesquareofHilbert-Schimidtmetric ∂∇ is definedby D2(ρ′,ρ) := Tr(ρ′ −ρ)2, which is nonnegative. =α(1−p )αrα−1−α(1−p )((1−p )r −T)α−1 >0, 2 ∂r 1 1 1 1 1 WehaveTr(ρ′ −ρ)2 =S (ρ′)−S (ρ)>0. 1 2 2 Lemma4. Supposeρ= pρ,where{p}aresomesetof when (1 − p1)r1 = T, ∇ has the minimum Tα. Therefore Pi i i i Sα(p)−Sα(q)takes its minimum pα1 −(p1 +T)α +Tα. Be- probabilitiesand{ρi}aredensityoperators. Thenwehavethe cause pα1 −(p1+T)α+Tα decreaseswiththedecreaseof p1, followingupper bound: S2(ρ) 6 h2(pi)+PpiS2(ρi), where Sα(p)−Sα(q)takesitsminimumTα+(1−T)α−1atp1 =1−T. h (p):=1− p2. i Nowwe haveprovedthe inequality(13), i.e. Lemma1.3, 2 i P i i namelytheinequality(12)ofLemma1.2forthecasethatboth Proof. TheproofusesthemethodsimilartothevonNeu- states ρ and σ are diagonal ones. For general ρ and σ, the mannentropycase(seeTheorem11.10of[13]). Fromdirect inequality can be directly proved by taking into account the calculation and the Cauchy- Schwartz inequality, we get the factthattheRe´nyientropyisunitaryinvariant[11]. Byusing result. Lemma1.1,Theorem1isproved. In the following text, we consider the bipartite quantum ProofofTheorem2: statesonH⊗H,withdbeingthedimensionofHilbertspace First, we define the relative linear entropy D(ρ||σ) := H. Tr(ρ2)−Tr(ρσ). ThelinearentropyisthengivenbyS2(ρ) = Lemma5. ForthebipartitequantumstateρAB,thefollow- −D(ρ||I) = 1 − Tr(ρ2), while the conditional linear entropy inginequalityholds, is given by S2(ρAB|ρB) := S2(A|B) := −D(ρAB||I ⊗ ρB) = d−1 Tr(ρ2)−Tr(ρ2 ). |S2(A|B)|6 . (17) B AB d SciChina-PhysMechAstron () Vol. No. -5 Proof. S (A|B) 6 d−1 comesfromthesubadditivityofthe ǫ)S (ρ ) + ǫS (τ ). From the upper bound S (γ ) 6 2 d 2 B 2 B 2 AB Tsallisentropy(see[14]),S (ρ )6S (ρ )+S (ρ ). Onthe h (ǫ,1−ǫ)+(1−ǫ)S (ρ )+ǫS (τ ),wegetS (γ |γ )= 2 AB 2 A 2 B 2 2 AB 2 AB 2 AB B otherhand,wehaveS (A|B)=S (ρ )−S (ρ )>−S (ρ )> S (γ ) − S (γ ) 6 h (ǫ,1 − ǫ) + (1 − ǫ)S (ρ |ρ ) + 2 2 AB 2 B 2 B 2 AB 2 B 2 2 AB B −d−1,whichcompletestheproof. ǫS (τ |τ ). ThereforeS (ρ |ρ )−S (γ |γ )>−h (ǫ,1− d 2 AB B 2 AB B 2 AB B 2 Lemma 6. For bipartite quantum states ρ and τ , as- ǫ) − ǫ(S (ρ |ρ ) − S (τ |τ )) > −h (ǫ,1 − ǫ) − 2ǫd−1. AB AB 2 AB B 2 AB B 2 d sumethat06ǫ 61,defineγ :=(1−ǫ)ρ +ǫτ ,then Second, as the conditionalentropyis concave, S (γ |γ ) > AB AB AB 2 AB B (1 − ǫ)S (ρ |ρ ) + ǫS (τ |τ ), we obtain S (ρ |ρ ) − d−1 2 AB B 2 AB B 2 AB B |S2(ρAB|ρB)−S2(γAB|γB)|62ǫ d +h2(ǫ,1−ǫ). (18) S2(γAB|γB) 6 ǫ(S2(ρAB|ρB) − S2(τAB|τB)) 6 2ǫd−d1, which completestheproof. Proof. The proof is similar to that of [1]. First, from UsingLemma6,andthemethodof[1],weprovetheresult the concavity of the entropy, we have S (γ ) > (1 − ofTheorem2. 2 B

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.