• Tidak ada hasil yang ditemukan

Wang–Landau simulation

N/A
N/A
Protected

Academic year: 2023

Membagikan "Wang–Landau simulation"

Copied!
5
0
0

Teks penuh

(1)

Contents lists available atScienceDirect

Physics Letters A

www.elsevier.com/locate/pla

Magnetization plateaus of the frustrated Ising Shastry–Sutherland system: Wang–Landau simulation

W.S. Lin

a

, T.H. Yang

a

, Y. Wang

a

, M.H. Qin

a,b,

, J.-M. Liu

c

, Zhifeng Ren

b,

aInstituteforAdvancedMaterialsandLaboratoryofQuantumEngineeringandQuantumMaterials,SouthChinaNormalUniversity,Guangzhou510006,China bDepartmentofPhysicsandTcSUH,UniversityofHouston,Houston,TX 77204,USA

cLaboratoryofSolidStateMicrostructures,NanjingUniversity,Nanjing210093,China

a r t i c l e i n f o a b s t ra c t

Articlehistory:

Received9May2014

Receivedinrevisedform30June2014 Accepted30June2014

Availableonline5July2014 CommunicatedbyR.Wu

Keywords:

Magnetizationplateaus Wang–Landaualgorithm Shastry–Sutherlandmagnets

TheWang–LandaualgorithmisusedtostudythemagneticpropertiesoftheIsingmodelontheShastry–

SutherlandlatticeinordertounderstandtheinterestingmagnetizationplateausobservedinTmB4.The simulated resultsdemonstratethat theequilibriumstateofthemodelproducesonly the1/3 and1/2 magnetizationplateaus atlow temperaturesevenwhen therandom-exchange termorthe long-range interactionsare takenintoaccount.Thisconfirmsourearlierconclusion(Huangetal.,2013)[20] that thosefractionalplateausobservedinexperimentsmaybeduetothemagnetizationdynamics.

©2014ElsevierB.V.All rights reserved.

1. Introduction

During the past decades, a number of frustrated spin sys- temswhichexhibit interesting magneticbehaviorshaveattracted widespread interest. In special, amazing magnetization (M) plateaus have been experimentally reported in these systems such asthe triangular spin-chain system Ca3Co2O6 andShastry–

Sutherland (S–S) magnets[1,2].Up to now, it iswidely accepted that the multi-step magnetization behaviors in Ca3Co2O6 are caused by non-equilibrium dynamics, but those in S–S systems arestillcontroversial[3–5].

In1981,theS–Slattice(Fig. 1,only J1and J2areincluded)was theoreticallyintroducedbyShastryandSutherlandasa frustrated quantum antiferromagnetic (AFM) model with an exact ground state [6]. Subsequently, thismodel has attractedextensive atten- tion because a sequence of magnetization plateaus at fractional valuesofthesaturatedmagnetization(MS)havebeenexperimen- tallyobservedinSrCu2(BO3)2[7].Furthermore,afewofrare-earth tetraboridesRB4(R=Tm,Dy,Tb,etc.),werefoundtobeS–Smag- nets exhibitingcomplex magnetic structures underapplied mag- netic fields (B) [8–10]. For example, magnetization plateaus at M/MS=1/2,1/7,1/9, etc., were reported in TmB4, and several

*

Correspondingauthors at: Department of Physics and TcSUH, University of Houston,Houston,TX77204,USA.

E-mailaddresses:[email protected](M.H. Qin),[email protected](Z. Ren).

theoreticalattemptswere reportedto uncoverthephysicalmech- anismsforsuchaphenomenon.

InTmB4,the largemagnetic momentsTm3+ (∼6.0

μ

B) arelo- catedon a latticethat is topologicallyequivalent to theS–S one.

Furthermore, subjected to strong crystalfield effects, TmB4 is of strong easy-axis anisotropy and can be reasonably described by classical Ising model. Theoretically, the magnetization process of the classicalS–S Ising modelwas extensivelystudied, andonlya singlemagnetization plateauatM/MS=1/3 wasconfirmedinan intermediate Brangeforacertaintemperature(T)rangeandcou- plingconstants[11,12].Itwasstronglysuggestedthattheeffectof additionallong-range interactions should beconsidered to finally explaintheexperimentalresultsinTmB4 [13].Asamatteroffact, the Ising-like X X Z (the in-plane andout-of-plane exchange cou- plingsaredifferentfromeachother)modelontheS–Slatticewith theadditionallong-rangeinteractions wasvisitedusingthequan- tum Monte Carlo Method [14,15]. The experimentally reported M/MS=1/2 plateauintheabsenceofthe M/MS=1/3 onewas reproduced.Inourearlierwork,theeffectofthelong-rangeinter- actionswasextensivelystudiedbasedontheclassicalIsingmodel forTmB4[16].ItwasdemonstratedthattheAFM J3andtheferro- magnetic(FM) J4 couplings(depictedinFig. 1)resultedfromthe RKKY interactions maybe responsible forthe stabilizationofthe M/MS=1/2 state[17].Inaddition,amodelbasedonthecoexis- tenceofspinandelectronsubsystemswasinvestigated,suggesting that the interaction between the electron and spin may play an importantroleinthemagnetizationprocessesin RB4[18,19].

http://dx.doi.org/10.1016/j.physleta.2014.06.047 0375-9601/©2014ElsevierB.V.All rights reserved.

(2)

Fig. 1.(Color online.)EffectivemodelontheShastry–Sutherland latticewiththe diagonalcouplingof J1, J2alongtheedgesofthesquares,andtheadditionalin- teractionsJ3and J4.

On the other hand, in frustrated spin systems, magnetization dynamics may play an essential role in the stabilization of the stateswithfractionalMsteps.Actually,thedynamicmagnetization processinTmB4wasstudiedbymeansoftheGlauberdynamicsin ourearlierwork [20].The experimental 1/7,1/9 and 1/11 mag- netizationplateaus inadditiontothe1/2 onecanbereproduced, suggesting that the magnetization dynamics may be responsible for the emergence of those plateaus. Furthermore, the effect of inhomogeneity on thestep-like magnetization feature was inves- tigated usingtherandomfield method[21]. Itwas observedthat the plateauat M/MS=1/3 split into three plateaus ina proper rangeoftheinhomogeneity.Thus,thereis anurgent needinun- derstandingthe groundstate orequilibrium statesbecause inho- mogeneityisalways inevitable inrealistic materials. However, an exactsolutionofthegroundstateproblemforthismodelwiththe random-exchange termisnotavailablesofar.Inaddition,thecon- ventionalMetropolisalgorithm mayalsofail torelax a trailstate into theequilibrium one atlow T in such a frustrated spin sys- tem[22].Therefore,itisstillanunsolvedissuedemonstratingthe equilibriumstateoftheS–SIsingmodelinordertofurtherunder- standthemagneticbehaviorsofTmB4.

Fortunately, the Wang–Landau algorithm that is very power- fultoreachtheequilibriumstate hasbeensuccessfullyappliedto variousfrustrated systems sinceitwas proposed in2001[23,24].

For example, the equilibrium state of the AFM triangular Ising modelforCa3Co2O6hasbeenobtainedbytheWang–Landausimu- lation.ThecooperationbetweentheWang–Landausimulationand theGlauber dynamicsgavea firmconclusionthat themagnetiza- tionplateausinCa3Co2O6 mustbecausedbythenon-equilibrium magnetization[5,25].Naturally,onemayquestionthatifthesame mechanismalsoworksforTmB4.However, noWang–Landau cal- culationforTmB4 hasbeenreported,asfarasweknow.

In order to clarify this critical issue, we investigate the clas- sical Ising model with the random exchange term on the S–S lattice using the Wang–Landau algorithm. The calculated results demonstrate that the equilibrium state of the model only pro- ducesthe1/3 and1/2 magnetizationplateaus,confirmingtheear- lier conclusionthat the other stepsin TmB4 magnetization curve may arise from non-equilibrium states.In addition, the effect of the additional long-range interactions is also examined in this work.

The remainder of this paper is organized as follows: in Sec- tion 2the model andthe simulationmethod are described. Sec- tion 3 is attributed to the simulation results and discussion. At last,theconclusionispresentedinSection4.

i,j

=

A

·

RAMi,j

,

(2)

wheretheAFMexchangecouplings J1=1 and J2=1/2,Sirepre- sentstheIsingspinwiththevalue ±1, B isappliedalongthe+z axis,RAMi,j isarandomnumberin[−1,1],and A representsthe magnitudeof therandom exchangeterm. J1 and J2 are fixed to thevaluesthesameasthoseusedpreviously,makingouranalysis simple[21].

Ontheotherhand,whentheadditionallong-rangeinteractions areconsidered,themodelcanbedescribedas

H

=

J1

diagonal

Si

·

Sj

+

J2

edges

Si

·

Sj

+

J3

i,j Si

·

Sj

+

J4

i,j

Si

·

Sj

B

i

Si

,

(3)

wherei,jandi,jdenotethesummationsoverallpairsonthe bondswith J3 and J4 couplingsasshowninFig. 1,respectively.

Our simulation is performed on a 12×12 S–S lattice with period boundary conditions. In the Wang–Landau algorithm, the densityofstate(DOS) g(E,M)inenergyandmagnetizationspace (E istheenergyofagivenspin configurationoftheHamiltonian intheabsenceof B)needs tobeaccuratelyestimatedinorderto investigatethemagneticpropertiesofasystem.Inthisreport,the simulationprocedureischosentobethesameasthatstatedinthe pathbreakingworkofWangandLandau.Foreverypossible(E,M) states,wesetall entriestotheDOS g(E,M)=1 andahistogram RH(E,M)=0 atthevery beginning.Thenwe flipspinsrandomly tobeginrandomwalkintheenergyandmagnetizationspace.The transitionprobabilityfromstate(E1,M1)tostate(E2,M2)is p

(

E1

,

M1

E2

,

M2

) =

min

g

(

E1

,

M1

)

g

(

E2

,

M2

) ,

1

.

(4)

Each time a state (Ei,Mi) is visited or retained, the histogram RH(Ei,Mi) (the number of visits) is accumulated and the exist- ingDOSismodifiedby g(Ei,Mi)=g(Ei,Mi)f0.Inthisreport,the initialmodificationfactorissetto f0=exp(1),andthefactorisre- ducedtoafineroneaccordingtotherecipe fi+1= fi1/2whenthe histogrambecomes“flat”.Afterfinishingtheinitialrunweperform 27cycles,resultinginafinalmodificationfactorof1.00000000745.

Forsimplicity,the“flat”histogramisverifiedwhenthehistogram for all possible (E,M) is not less than 80% of the average his- togram.

Afterg(E,M)hasbeenobtained,wemaycalculatethethermo- dynamicandmagneticquantitiesatanyT andB.Forexample,the magnetization M(T,B)canbecalculatedfrom

M

(

T

,

h

) =

E,MM g

(

E

,

M

)

exp

(

H

/

kBT

)

E,Mg

(

E

,

M

)

exp

(

H

/

kBT

) .

(5)

(3)

Fig. 2.(Color online.) (a) Evaluated DOS ln[g(E,M)]of the model (1) withA=0 and curves in (b) are DOS at various negative energy valuesE.

Fig. 3.(Color online.)(a)MagnetizationM/MSasafunctionofTandB forA=0.

Spinconfigurationin(b)theup–up–downstate,and(c)the1/2 plateaustate.Solid andemptycirclesrepresenttheup-spinsandthedown-spins,respectively.

3. Simulationresultsanddiscussion

Fig. 2(a) showsthe simulated g(E,M) of the model (1) with A=0. ln[g(E,M)] shows a parabolic shape in the low-energy range (E <0) and reaches its single maximum value at M=0 foragivenenergy.Furthermore,thegroundstate forsuch afrus- tratedsystemis highly degenerate,leading to a ratherlarge DOS atthelowest energy Emin (∼−72 in thiscase),asclearly shown inFig. 2(b).On theother hand,only two FM states(M=1,1) arepossibleatthehighestenergy.Similarbehaviorshavebeenre- portedforsome other systemssuch asthetwo-dimensional AFM triangularmagnets[5].

Subsequently,MasafunctionofT andBiscalculated,andthe simulatedresultsareshownin Fig. 3(a).At low T (T<0.4), two stepsareobserved.AtT=0.02,asanexample,Mrapidlyreaches thefirstplateauatM/MS=1/3 whenB increasesfromzero,and thenswitchesto MS above B3.The sameasearlierreport,the MS/3 plateau results from the up–up–down state (each triangle containstwoup-spinsandonedown-spin,asdepictedinFig. 3(b)) [11]. With the increasing of T, the steps can be progressively washedoutduetotheenhancedthermalactivation.WhenT isin- creasedtoabout0.5,theM0/3 stepdisappearscompletelyandthe systemexhibits the paramagnetic property with the linear MB relation. Thus, our simulation results show that the equilibrium state ofthe Ising modelfor the perfectS–S lattice produce two- step magnetization curve at low T, the same as those obtained fromothermethodssuchastherenormalization-groupapproach.

However,earliersimulationofthemodelincludingtherandom exchange term based on the Metropolis algorithm gave a rather differentmulti-stepmagnetizationcurve.Similarly,onemaydoubt

Fig. 4.(Color online.)Comparisonof(a)M/MSand (b)internalenergiesU asa functionofB calculatedfromthe Wang–Landaualgorithmwith computedusing theMetropolisalgorithmatT=0.02 forA=0.2.

Fig. 5.(Color online.)SimulatedcurvesofM/MSasafunctionofBforvariousAat T=0.02.

that this algorithm may fail to relax the system into the equi- librium state. This argumenthas been verified in oursimulation by comparing the magnetizationsand internal energies obtained by theWang–Landau andMetropolisalgorithms. IntheMetropo- lis simulation,theinitial1×105 stepsare discardedandanother105 steps are retainedforstatistic averaging. Fig. 4(a) shows

(4)

Fig. 6.(Color online.) MagnetizationM/MSas a function ofTandBat (a) J3=0.10 and J4= −0.15, and (c) J3=0.18 and J4= −0.25.

thesimulatedmagnetizationcurvesforA=0.2 atT=0.02,which exhibits a big discrepancy between the two simulations. In de- tails, the MS/3 step which is stable below B3 in the Wang–

LandausimulationsplitsintothreestepsintheMetropolissimula- tion.Importantly,therelevantinternalenergiesobtainedfromthe Metropolis simulation are obviouslyhigher than the correspond- ingWang–Landausimulationresults,asclearlyshowninFig. 4(b).

Furthermore,itisnotedthatthetemperatureconcernedhereisso low that contribute littleto thevalue of thefree energies ofthe system.Asaconsequence,theaboveresultsallowustoarguethat theequilibriumstateofthemodelproducesonlythe1/3 magneti- zationplateauatlowT evenwhentherandom-exchangetermare takenintoaccount.

Moreover,theeffectoftherandom-exchangetermonthestep- likemagnetization feature hasalso beenexamined, andthesim- ulatedresultsarepresentedinFig. 5,wheremagnetization curves uponvarious A atT=0.02 aregiven. Itis obviousthatno addi- tionalstepscanbegeneratedbytherandom-exchangeterm.How- ever,theperfectup–up–downstatewiththe M/MS=1/3 plateau canbe partiallydestroyednearthecritical B (0 and3)whenthe random-exchangetermisconsidered,leadingtothesmoothnessof themagnetizationcurve.Thus,thebordersbetweenthetwosteps becomemoreandmorefaintwiththeincreasingofA,asrevealed inoursimulation.

On the other hand, additional long-rangeinteractions are be- lieved to be responsible forthe stabilization ofthe M/MS=1/2 plateauinearlierreports.Inthiswork,thedependenceofthemag- netic propertieson theadditional long-rangeinteractions J3 and J4 isalso examinedusing theWang–Landau simulation.Fig. 6(a) showsthe simulated magnetization curves atvarious T forAFM J3 =0.1 and FM J4 = −0.15. In addition to the plateaus at M/MS=0,1/3, and 1,the experimentally reported M/MS=1/2 plateau is reproduced at low T. Furthermore, with the increas- ing T, both the 1/3 and 1/2 steps are progressively washed out due to thermal fluctuation. However, the disappearing T of the 1/2 step(∼0.6)ismuch higherthan thatofthe1/3 step (∼0.3), demonstratingthat the1/2 plateaustate is morestablethanthe 1/3 one.Asrevealed inearlierwork,morelocal AFM J3 andFM J4 interactions canbe satisfiedin the1/2 plateau state (spinar- rangementconsistingofalternativeAFMandFMstripes,asshown in Fig. 3(c)) than those in theup–up–down state, leading to the stabilization of the 1/2 plateau accompanied by the destabiliza- tionofthe1/3 plateau[16].Asaresult,withtheincreasingofthe magnitudes ofthe AFM J3 andFM J4,the 1/3 plateau are pro- gressivelyreplacedbythe1/2 one.For J3=0.18 and J4= −0.25, theup–up–downstate is completelyreplacedby the 1/2 plateau one, resulting in the stabilization of the 1/2 plateau in the ab- senceofthe1/3 one,asclearlyshowninFig. 6(b).Thus,theeffect ofthelong-rangeinteractions ontheequilibriummagnetizationis examinedinourWang–Landausimulation.However,theadditional

plateausatM/MS=1/7,1/9,etc.,reportedinexperimentscannot bestabilizedbytheadditionallong-rangeinteractions.

Anyway,it iswellknownthat afrustrated spinsystemcanbe easilylocalizedinoneofthesemetastablestatesatlowT,andthe timeavailableexperimentallymaynotbesufficientforthesystem to relaxto theequilibriumstate.Althoughthe magneticproperty ofTmB4 hasattractedattentionsformanyyears,butthereisstill a bigchallengeindevelopinga reliableexperimental approachto uncover thegroundstatesorequilibriumstatesofthissystem. In this report, we used Wang–Landau simulation to clearly demon- strate the equilibriummagnetization ofthe theoretical modelfor TmB4. No fractional magnetization plateaus other than those at M/MS=1/3 and1/2 canbereproducedevenwheninhomogene- ityoradditionallong-rangeinteractionsaretakenintoaccount.As a consequence,thecooperationbetweentheWang–Landau simu- lationandtheGlauberdynamicsgivesastrongsuggestionthatthe emergenceofthoseplateausinTmB4 maybecausedbythemag- netizationdynamics.

4. Conclusion

In thisreport, we havestudied theequilibriummagnetization oftheS–SsystemTmB4basedontwo-dimensionalIsingmodelus- ing theWang–Landaumethod.Thesimulatedresultsdemonstrate thattheequilibriumstateofthemodelonlyproducesthe1/3 and 1/2 magnetization plateaus at low temperatures even when the random-exchange term or the long-range interactions are taken into account. Thus, the present simulation indicates that those fractional plateaus atsmall magnetization values reportedin ex- perimentsmaybeduetothemagnetizationdynamics.

Acknowledgements

ThisworkwassupportedbytheTexasCenterforSuperconduc- tivity of the University of Houston (TcSUH) andthe U.S. Depart- mentofEnergy underContractNumberDOEDE-FG02-13ER46917 (DE-SC0010831),theNationalNaturalScienceFoundationofChina (11204091,11274094,51332007,and61106061),andtheNational KeyProjectsforBasicResearchofChina(2011CB922101).

References

[1]H.Kageyama,K. Yoshimura,K.Kosuge,M. Azuma,M. Takano,H.Mitamura, T. Goto,J.Phys.Soc.Jpn.66(1997)3996.

[2]H.Kageyama, K.Yoshimura, R.Stern, N.V.Mushnikov,K. Onizuka,M. Kato, K. Kosuge,C.P.Slichter,T.Goto,Y.Ueda,Phys.Rev.Lett.82(1999)3168.

[3]Y.B.Kudasov,Phys.Rev.Lett.96(2006)027212.

[4]X.Y.Yao,S.Dong,J.-M.Liu,Phys.Rev.B73(2006)212415.

[5]M.H.Qin,K.F.Wang,J.-M.Liu,Phys.Rev.B79(2009)172405.

[6]B.S.Shastry,B.Sutherland,PhysicaB&C108(1981)1069.

[7]R.W.Smith,D.A.Keszler,J.SolidStateChem.93(1991)430.

[8]K.Siemensmeyer,E.Wulf,H.J.Mikeska,K.Flachbart,S.Gabani,S.Matas,P. Pri- puten,A.Efdokimova,N.Shitsevalova,Phys.Rev.Lett.101(2008)177201.

(5)

[9]S.Yoshii,T.Yamamoto,M.Hagiwara,S.Michimura,A.Shigekawa,F.Iga,T.Tak- abatake,K.Kindo,Phys.Rev.Lett.101(2008)087202.

[10]R.Watanuki,G.Sato,K.Suzuki,M.Ishihara,T.Yanagisawa,Y.Nemoto,T.Goto, J.Phys.Soc.Jpn.74(2005)2169.

[11]M.C.Chang,M.F.Yang,Phys.Rev.B79(2009)104411.

[12]Y.L.Dublenych,Phys.Rev.Lett.109(2012)167202.

[13]H.Cencarikova,P.Farkasovsky,P.Puchala,ActaPhys.Pol.A126(2014)46.

[14]T.Suzuki,Y.Tomita,N.Kawashima,Phys.Rev.B80(2009)180405(R).

[15]T.Suzuki,Y.Tomita,N.Kawashima,Phys.Rev.B82(2010)214404.

[16]L.Huo,W.C.Huang,Z.B.Yan,X.T.Jia,X.S.Gao,M.H.Qin,J.-M.Liu,J.Appl.Phys.

113(2013)073908.

[17]J.J.Feng,L.Huo,W.C.Huang,Y.Wang,M.H.Qin,J.-M.Liu,Z.F.Ren,Europhys.

Lett.105(2014)17009.

[18]P.Farkasovsky,H.Cencarikova,ActaPhys.Pol.A126(2014)44.

[19]P.Farkasovsky,H.Cencarikova,S.Matas,Phys.Rev.B82(2010)054409.

[20]W.C.Huang,L.Huo,J.J. Feng,Z.B.Yan,X.T.Jia,X.S.Gao,M.H. Qin,J.-M.Liu, Europhys.Lett.102(2013)37005.

[21]M.H.Qin,G.Q.Zhang,K.F.Wang,X.S.Gao,J.-M.Liu,J.Appl.Phys.109(2011) 07E103.

[22]D.P. Landau, K. Binder, A Guide to Monte Carlo Simulations in Statistical Physics,CambridgeUniversityPress,Cambridge,England,2008.

[23]F.Wang,D.P.Landau,Phys.Rev.Lett.86(2001)2050.

[24]F.Wang,D.P.Landau,Phys.Rev.E64(2001)056101.

[25]Y.B.Kudasov,A.S.Korshunov,V.N.Pavlov,D.A.Maslov,Phys.Rev.B78(2008) 132407.

[26]H.Shinaoka,Y.Tomita,Y.Motome,Phys.Rev.Lett.107(2011)047204.

Referensi

Dokumen terkait

Dari penjelasan diatas dapat disimpulkan bahwa pabila sebuah URL atau alamat web tersebut sudah dapat diakses oleh orang yang tidak bertanggung jawab maka akan