ebook img

Bosonic representation of one-dimensional Heisenberg ferrimagnets PDF

0.52 MB·English
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 Bosonic representation of one-dimensional Heisenberg ferrimagnets

Bosonic representation of one-dimensional Heisenberg ferrimagnets Shoji Yamamoto Division of Physics, Hokkaido University, Sapporo 060-0810, Japan (2 September2003) 4 Theenergy structureandthethermodynamicsofferrimagnetic Heisenbergchainsof alternating 0 spins S and s are described in terms of the Schwinger bosons and modified spin waves. In the 0 Schwingerrepresentation,weaveragethelocalconstraintsonthebosonsanddiagonalizetheHamil- 2 tonian at the Hartree-Fock level. In the Holstein-Primakoff representation, we optimize the free n energy in two different ways introducing an additional constraint on the staggered magnetization. a A new modified spin-wave scheme, which employs a Lagrange multiplier keeping the native energy J structurefreefromtemperatureandthusdiffersfromtheoriginalTakahashiScheme,isparticularly 4 stressedasanexcellentlanguagetointerpretone-dimensionalquantumferrimagnetism. Othertypes 2 of one-dimensional ferrimagnets and theantiferromagnetic limit S =s are also mentioned. ] h PACS numbers: 75.10.Jm, 75.50.Gg, 75.40.Cx c e m I. INTRODUCTION group[22,23]andquantumMonteCarlo[24,25]methods in an attempt to illuminate dual features of ferrimag- - t netic excitations. More general mixed-spin chains were a Significant efforts have been devoted to syn- analyzed via the nonlinear σ model [26] with particu- t thesizing low-dimensional ferrimagnets and under- s lar emphasis on the competition between massive and . standing their quantum behavior in recent years. t massless phases. Quasi-one-dimensional mixed-spin sys- a The first example of one-dimensional ferrimagnets, m MnCu(S C O ) (H O) 4.5H O, was synthesized by tems[27,28]werealsoinvestigatedinordertoexplainthe 2 2 2 2 2 3· 2 inelastic-neutron-scattering findings [29,30] for the rare- - Gleizes and Verdaguer [1] and followed by a series of or- d dered bimetallic chain compounds [2] in an attempt to earth nickelates R2BaNiO5. n design molecule-based ferromagnets [3]. Caneschi et al. Inordertocomplementnumericaltoolsandtoachieve o c [4] demonstrated another approach to alternating-spin further understanding of the magnetic double structure [ chains hybridizing manganese complexes and nitronyl of ferrimagnetism, several authors have recently begun nitroxide radicals. The inorganic-organic hybrid strat- to construct bosonic theories of low-dimensional quan- 1 egyrealizedmorecomplicatedalignmentsofmixedspins tum ferrimagnets. The conventional spin-wave descrip- v 6 [5]. There also exists an attempt at stacking novel tri- tion of the ground-state properties [22,31–33], a modi- 7 radicals into a purely organic ferrimagnet [6]. Mono- fiedspin-waveschemeforthelow-temperatureproperties 4 spin chains can be ferrimagnetic with polymerized ex- [34], and the Schwinger-bosonrepresentationof the low- 1 change interactions. An example of such ferrimagnets energy structure [35] and the thermodynamics [36], they 0 is the ferromagnetic-antiferromagnetic bond-alternating all reveal the potential of bosonic languages for various 4 copper tetramer chain compound Cu(C H NCl) (N ) ferrimagnetic systems. However, considering the global 0 5 4 2 3 2 [7]. The trimeric intertwining double-chain compound argument and total understanding over the bosonic the- / t Ca Cu (PO ) [8] is another solution to homometallic ory of ferromagnets and antiferromagnets [37–47], ferri- a 3 3 4 4 m one-dimensional ferrimagnets, where the noncompensa- magnets are still undeveloped in this context especially tion of sublattice magnetizations is of topological ori- in one dimension. In such circumstances, we represent - d gin. Besides one-dimensional ferrimagnets, metal-ion one-dimensionalHeisenberg ferrimagnets in terms of the n magneticclusterssuchas[Mn O (CH COO) (H O) ] SchwingerbosonsandtheHolstein-Primakoffspinwaves. 12 12 3 16 2 4 o [9] and [Fe (N C H ) O (OH) ]8+ [10], for which res- Basedonamean-fieldansatz,thelocalconstraintsonthe 8 3 6 15 6 2 12 c onantmagnetizationtunneling [11–14]wasobserved,are Schwinger bosons are relaxed and imposed only on the : v also worth mentioning as zero-dimensional ferrimagnets. average. The conventional antiferromagnetic spin-wave i The discovery of ordered bimetallic chain compounds formalism [48,49] is modified, on the one hand following X stimulated extensive theoretical interest in (quasi-)one- the Takahashi scheme [37,38] which was originally pro- r a dimensional quantum ferrimagnets. Early efforts [15] posedforferromagnets,whileontheotherhandintroduc- were devoted to numerically diagonalizing alternating- ing a slightly different new strategy [50]. The Schwinger spin Heisenberg chains. Numerical diagonalization, bosons and the modified spin waves both interpret the combined with the Lanczos algorithm [16,17] and a low-energypropertiesfairlywellidentifyingtheferrimag- scaling technique [18], further contributed to study- netic long-range order with a Bose condensation, while ing modern topics such as phase transitions of the the two languages are qualitatively distinguished in de- Kosterlitz-Thouless type [17,19] and quantized magne- scribing the thermodynamics. We demonstrate that the tization plateaux [20,21]. Alternating-spin chains were new modified spin-wave scheme of all others depict one- further investigated by density-matrix renormalization- dimensional ferrimagnetic features very well. 1 II. FORMALISM where k is defined as nπ/Na (n = 0,1, ,N 1) with ··· − a being the distance between neighboring spins, and λ A practical model for one-dimensional ferrimagnets is andµaretheLagrangemultipliersduetotheconstraints two kinds of spins, S and s (S > s), alternating on a (2.6). Via the Bogoliubov transformation ring with antiferromagnetic exchange coupling between nearest neighbors, as described by the Hamiltonian, ak↑ = αk↑coshθk+βk†↓sinhθk, a = α coshθ β† sinhθ , N k↓ k↓ k− k↑ k (2.8) H=JnX=1(Sn·sn−1+sn·Sn) , (2.1) bbkk↑↓ == ββkk↑↓ccoosshhθθkk−+αα†k†k↓↑ssiinnhhθθkk,, where N is the number of unit cells. The simplest case, with (S,s)=(1,1), has so far been discussedfairly well using 2 the matrix-productformalism [51], a modified spin-wave 2Ωcosak scheme[50],theSchwinger-bosonrepresentation[36],and tanh2θk = , (2.9) λ+µ modern numerical techniques [22–25]. We make further explorationsintohigher-spinsystemsanddevelopthean- the Hamiltonian (2.7) is diagonalized as alytic argument in more detail. =2NJSs+4NJΩ2 MF H 2NJλ(2S+1) 2NJµ(2s+1)+2J ω A. Schwinger-boson mean-field theory − − k k X Letusdescribeeachspinvariableintermsoftwokinds +J ωk−σα†kσαkσ +ωk+σβk†σβkσ , (2.10) of bosons as Xk σX=↑,↓(cid:16) (cid:17) Sn+ =a†n↑an↓, Snz = 21 a†n↑an↑−a†n↓an↓ , where s+ =b† b , sz = 1 b† b b† b , (2.2) n n↑ n↓ n 2(cid:0)n↑ n↑− n↓ n↓ (cid:1) ω± ω± =ω (µ λ); (cid:0) (cid:1) kσ ≡ k k± − ω = (λ+µ)2 4Ω2cos2ak. (2.11) where the constraints k − p a† a =2S, b† b =2s, (2.3) λ, µ, and Ω are determined through a set of equations nσ nσ nσ nσ σ=↑,↓ σ=↑,↓ X X n¯− cosh2θ +n¯+ sinh2θ +sinh2θ =NS, (2.12) are imposed on the bosons. Then the Hamiltonian can kσ k kσ k k k be written as X(cid:0) (cid:1) n¯− sinh2θ +n¯+ cosh2θ +sinh2θ =Ns, (2.13) N kσ k kσ k k H=2NJSs−2J Ω†n+Ωn++Ω†n−Ωn− , (2.4) Xk (cid:0)n¯− +n¯+ +1 coshθ sinhθ =NΩ(cid:1), (2.14) nX=1(cid:16) (cid:17) kσ kσ k k k where Ω = (a b a b )/2 and Ω = X(cid:0) (cid:1) n+ n↑ n↓ n↓ n↑ n− − (an↑bn−1↓ an−1↓bn↑)/2. The Hartree-Fock treatment where the thermal distribution functions n¯− assumes th−e thermal average of the short-range antifer- α† α and n¯+ β† β are required tokσmin≡i- romagnetic order to be uniform and static as h kσ kσiT kσ ≡ h kσ kσiT mize the free energy and given by Ω† = Ω =Ω. (2.5) h n±iT h n±iT n¯± = 1 . (2.15) The constraints (2.3) are correspondingly relaxed as kσ eωk±σ/kBT 1 − N N The magnetic susceptibility is expressed as a† a =2NS, b† b =2Ns. (2.6) nσ nσ nσ nσ nX=1σX=↑,↓ nX=1σX=↑,↓ χ= (gµB)2 n¯τ (n¯τ +1), (2.16) 4k T kσ kσ In the momentum space the mean-field Hamiltonian B k τ=±σ=↑,↓ XX X reads where we have set the g-factors of spins S and s both =2NJSs+4NJΩ2 4NJ(λS+µs) MF equal to g. The internal energy should be given by H − 2JΩ cosak(a b a b +H.c.) − k↑ k↓− k↓ k↑ 1 Xk E = 2(EMF+2NJSs)−2NJSs, (2.17) +2J λa† a +µb† b , (2.7) kσ kσ kσ kσ where Xk σX=↑,↓(cid:16) (cid:17) 2 E =2NJSs+4NJΩ2 2NJ(2λS+2µs+λ+µ) ω±(k)=(S+s)cosh2θ 2√Sscosaksinh2θ MF − 1 k− k +2J ωk+J n¯τkσωkτσ. (2.18) ±(S−s)≡ωk±(S−s), (2.26a) Xk Xk τX=±σX=↑,↓ ω±(k)= S + s Γ 2Λ cosh2θ Arovas and Auerbach [44] pointed out that relaxing the 0 "(cid:16)rs rS (cid:17) − # k original constraints (2.3) into Eq. (2.6) leads to dou- S s blecountingthenumberofindependentbosondegreesof 2Γ + Λ cosaksinh2θk freedom. Therefore,in Eq. (2.17),we havecorrectedthe −" −(cid:16)rs rS (cid:17) # mean-field artifact reducing the overestimated quantum S s , (2.26b) fluctuation. ± s − S r r (cid:16) (cid:17) B. Modified spin-wave theory: Takahashi scheme γ (k)=2√Sscosakcosh2θ (S+s)sinh2θ , (2.27a) 1 k k − S s γ (k)= 2Γ + Λ cosakcosh2θ Nextweconsiderasingle-componentbosonicrepresen- 0 " − rs rS # k tation of each spin variable at the cost of the rotational (cid:16) (cid:17) symmetry. We start from the Holstein-Primakoff trans- S s + Γ 2Λ sinh2θ . (2.27b) formation −" rs rS − # k (cid:16) (cid:17) S+ = 2S a†a a , Sz =S a†a , The conventional spin-wave scheme naively diagonal- n − n n n n − n n (2.19) ize the Hamiltonian (2.20) and ends up with the num- s+n =bq†n 2s−b†nbn, szn =−s+b†nbn. ber of bosons diverging with increasing temperature. In q order to suppress this thermal divergence, Takahashi Treating S and s as O(S) = O(s), we can expand the [38]considered optimizing the bosonic distribution func- Hamiltonian with respect to 1/S as tions under zero magnetization and obtained an excel- lent description of the low-temperature thermodynam- = 2NJSs+E1+E0+ 1+ 0+O(S−1), (2.20) ics for low-dimensional Heisenberg ferromagnets. For H − H H ferrimagnets, this idea is still useful [34,50] but never where Ei and i give the O(Si) quantum corrections applies away from the low-temperature region as it is. H to the ground-state energy and the dispersion relations, The zero-magnetization constraint plays a role of keep- respectively. Via the Bogoliubov transformation ing the number of bosons finite under ferromagnetic in- teractions but does not work so under antiferromagnetic a = α coshθ β†sinhθ , interactions. Takahashi [38] and Hirsch et al. [39] pro- k k k− k k (2.21) b = β coshθ α†sinhθ , posed constraining the staggered magnetization, instead k k k− k k of the uniform magnetization, to be zero as the anti- they are written as ferromagnetic version of the modified spin-wave theory. Their scheme was applied to extensive antiferromagnets E = 2NJ 2√SsΓ (S+s)Λ , (2.22a) in both two [38,39,52–55] and one [40,41] dimensions. 1 − − Theconventionalspin-waveprocedureassumesthatspins h S i s ononesublatticepointpredominantlyup,whilethoseon E0 = 2NJ Γ2+Λ2 + ΓΛ , (2.22b) the other predominantly down. The modified spin-wave − " −(cid:16)rs rS (cid:17) # treatmentrestoresthesublatticesymmetry. Weconsider the naivest generalizationof the antiferromagnetic mod- ified spin-wave scheme to ferrimagnets. Hi =J ωi−(k)α†kαk+ωi+(k)βk†βk The constraint of zero staggered magnetization reads Xk h +γ (k) α β +α†β† , (2.23) a†nan+b†nbn =N(S+s). (2.28) i k k k k n (cid:0) (cid:1)i X(cid:0) (cid:1) where In order to enforce this condition, we first introduce a Lagrange multiplier and diagonalize the effective Hamil- 1 tonian Γ= cosaksinh2θ , (2.24) k 2N 1 Xk H=H+2Jν a†nan+b†nbn . (2.29) Λ= 2N (cosh2θk−1), (2.25) e Xn (cid:0) (cid:1) k Then the ground-state energy and the dispersion rela- X tions are obtained as 3 E = 2NJSs+E ; E =E +4NJΛν, (2.30) Theground-stateenergyandthedispersionrelationsare g 1 1 1 − ω± =ω±(k); ω±(k)=ω±(k)+2νcosh2θ , (2.31) simply given by k 1 1 1 k e e keeping oenly the beilinear terms and as Eg =−2NJSs+E1; ωk± =ω1±(k), (2.40) E = 2NJSs+E +E , (2.32) within the up-to-O(S1) treatment and by g 1 0 − ω± =ω±(k)+ω±(k), (2.33) k 1 0 e Eg =−2NJSs+E1+E0; ωk± =ω1±(k)+ω0±(k), considering the O(S0) interactions as well. In terms of e (2.41) the spin-wave distribution functions in the up-to-O(S0) treatment. They are nothing but the 1 n¯± = , (2.34) T =0 findings in the Takahashi scheme. k eωk±/kBT −1 At finite temperatures we replace α†kαk and βk†βk in the internal energy and the magnetic susceptibility are the spin-wave Hamiltonian (2.23) by expressed as [56] ∞ n¯∓ n∓P (n−,n+), (2.42) E =Eg+ n¯τkωkτ, (2.35) k ≡ k k τ=± n−X,n+=0 XX χ= (gµB)2 n¯τ(n¯τ +1). (2.36) where Pk(n−,n+) is the probability of n− ferromagnetic 3kBT k τ=± k k and n+ antiferromagnetic spin waves appearing in the XX k-momentum state and satisfies θ , defining the Bogoliubov transformation (2.21), is de- k termined through P (n−,n+)=1, (2.43) k nX−,n+ γ (k) 2νsinh2θ γ (k)=0, (2.37) 1 k 1 − ≡ for all k’s. Then the free energy is written as provided we treat as a perturbation to . H0 e H1 F =E +J P (n−,n+) nτωτ g k k C. Modified spin-wave theory: A new scheme Xk nX−,n+ τX=± +k T P (n−,n+)lnP (n−,n+). (2.44) B k k Although the Takahashi scheme overcomes the diffi- Xk nX−,n+ culty of sublattice magnetizations diverging thermally, theobtainedthermodynamicsisstillfarfromsatisfactory We minimize the free energy with respect to Pk(n−,n+) (see Fig. 2 later on). Within the conventionalspin-wave enforcing a condition theory,thequantumspinreduction,thatis,thequantum Sz sz +2δ :Sz sz : fluctuation of the ground-state sublattice magnetization h n− niT ≡h n− n iT per unit cell, reads S+s n¯τ =S+s k =0, (2.45) − N ω ha†naniT=0 =hb†nbniT=0 ≡δ Xk τX=± k π S+s dk 1 aswellasthetrivialconstraints(2.43). Inthesecond-side = , (2.38) Z0 (S s)2+4Sssin2(k/2)2π − 2 compactexpression,thenormalorderingistakenwithre- − spect to both operatorsα and β. Equation(2.45)claims and divergesqat S = s. The Takahashi scheme settles that the thermal fluctuation (S + s) k(n−k + n+k)/ωk this quantum divergence as well as the thermal diver- should cancel the full, or classical, N´eel order (S +s)N gence. However, the number of bosons does not diverge rather than the quantum mechanicallyPreduced one (S + in the ferrimagnetic ground state. Without quantum di- s 2δ)N. Without consideration of the quantum fluc- − vergence, it is not necessary to modify the dispersion tuation 2δ, which is absent from ferromagnets but pe- relations (2.26a) into the temperature-dependent form culiar to ferrimagnets, the present scheme breaks even (2.31). While the thermodynamics should be modified, the conventional spin-wave achievement at low temper- the quantum mechanics may be left as it is. atures. Numerically solving the thermodynamic Bethe- Such an idea leads to the Bogoliubov transformation Ansatz equations,TakahashiandYamada [57]suggested freefromtemperaturereplacingEq. (2.37)byγ (k)=0, that the conventional spin-wave theory correctly gives 1 that is, the low-temperature leading term of the specific heat. Boththe Takahashischeme with Eq. (2.28)andthe new 2√Sscosak scheme with Eq. (2.45) indeed keep unchanged the con- tanh2θ = . (2.39) k ventional spin-wave findings S+s 4 C 3 S sζ(3) considerablyunderestimatetheantiferromagneticexcita- − 2 t1/2 (T 0), (2.46) NkB ∼ 4r Ss √2π → tion energies. The quantum correlation has much effect ontheantiferromagneticexcitationmodeandsuchanef- where t = k T/J within the up-to-O(S1) treatment, fectiswellincludedintotheSchwinger-bosoncalculation B while t=k T/γJ with γ =1+Γ/√Ss (S+s)Λ/Ssin even at the mean-field level. B − the up-to-O(S0) treatment. The conventional spin-wave Nextwecalculatethethermodynamicproperties. Fig- approach gives no quantitative information on the mag- ure 2 shows the temperature dependence of the specific netic susceptibility, whereas the modified theory reveals heat. The Schwinger-boson mean-field theory is still highlysuccessfulatlowtemperatures,whilewithincreas- χJ Ss(S−s)2 t−2 (T 0). (2.47) ing temperature, it rapidly breaks down failing to repro- N(gµ )2 ∼ 3 → duce the Schottky-type peak. The mean-field order pa- B rameter Ω monotonically decreases with increasing tem- In terms of the optimum distribution functions perature and reaches zero at 1 n¯± = , (2.48) k T S+s+1 k e[Jωk±−ν(S+s)/ωk]/kBT −1 BJ = ln(1+1/S)+ln(1+1/s). (3.1) the free energy at the thermal equilibrium is written as Above this temperature, Ω sticks at zero suggesting no F =E +ν(S+s)N k T ln(1+n¯τ), (2.49) antiferromagnetic correlation in the system. The onset g − B k of the paramagnetic phase at a finite temperature is a k τ=± XX mean-field artifact and the particular temperature (3.1) where ν is the Lagrangemultiplier due to the constraint is an increasing function of S and s. The modified spin- (2.45). wave theory based on the Takahashi scheme also fails to describe the Schottky peak. Because of the Lagrange multiplier ν, which turns out a monotonically increasing III. RESULTS function of temperature, the dispersion relations (2.31) lead to endlessly increasing energy and thus nonvanish- First we calculate the ground-state energy E and the ingspecificheatathightemperatures. Only the modified g antiferromagneticexcitationgapω+ andcomparethem spin-wave theory based on the new scheme succeeds in k=0 with numerical findings in Table I. At T = 0, the Taka- interpreting the Schottky peak. Since the antiferromag- hashi scheme and the new scheme lead to the same re- netic excitation gap is significantly improved by the in- sults both giving ν = 0. We are fully convinced that clusionoftheO(S0)correlation,theinteractingmodified the spin-wave treatment better works for larger spins. spin waves reproduce the location of the Schottky peak We further learn that the spin-wave approach is bet- fairly well. Mixed-spin trimeric chain ferrimagnets have ter justified with increasing S/s as well as Ss, which is recently been synthesized [5] and their low-temperature because the quantity S s fills the role of suppressing thermal properties were well elucidated by the modified − the divergence in Eq. (2.38). On the other hand, the spin-wave theory [34]. However, it was unfortunate that Schwinger-boson approach constantly gives highly pre- the additional constraint was imposed on the uniform cise estimatesofthe low-energyproperties. Figure1 fur- magnetizationandthereforethehiger-temperatureprop- ther demonstrates that the Schwinger-boson mean-field erties were much less illuminated. Controlling the stag- theoryis highlysuccessfulindescribingthelow-lyingex- geredmagnetizationinsteadbasedonthenewscheme,we citations. Both the bosonic languages well interpret the canfully investigatesuch polymeric chaincompounds as ferromagnetic excitations, whereas the linear spin waves well. 4.0 QMC SB 3.0 LMSW PIMSW k2.0 w 1.0 (a) (1,1/2) (b) (3/2,1/2) (c) (3/2,1) 0.0 0.0 0.2 0.4 0.6 0.8 1.00.0 0.2 0.4 0.6 0.8 1.00.0 0.2 0.4 0.6 0.8 1.0 2ak/p FIG. 1. The Schwinger-boson (SB), linear-modified-spin-wave (LMSW), perturbational interacting-modified-spin-wave − (PIMSW), and quantum Monte Carlo (QMC) calculations of the dispersion relations of the ferromagnetic (ω ) and anti- k ferromagnetic (ω+) elementary excitations for thespin-(S,s) ferrimagnetic Heisenberg chains at zero temperature. k 5 2.0 (a) (1,1/2) (b) (3/2,1/2) (c) (3/2,1) QMC B SB Nk1.0 LMSW(Takahashi) C/ PLIMMSSWW((YTaamkaahmaoshtoi)) PIMSW(Yamamoto) 0.0 0.0 1.0 2.0 3.00.0 1.0 2.0 3.00.0 1.0 2.0 3.0 k T/J B FIG. 2. The Schwinger-boson (SB), linear-modified-spin-wave (LMSW), perturbational interacting-modified-spin-wave (PIMSW),andquantumMonteCarlo(QMC)calculationsofthespecificheatC asafunctionoftemperatureforthespin-(S,s) ferrimagnetic Heisenberg chains. The modified spin waves are constructed in two different ways, theTakahashi scheme (Taka- hashi) and thenew scheme (Yamamoto). 4.0 (a) (1,1/2) (b) (3/2,1/2) (c) (3/2,1) SB PIMSW(Takahashi) 3.0 0 k=2.0 w 1.0 0.0 0.0 1.0 2.0 3.0 0.0 1.0 2.0 3.00.0 1.0 2.0 3.0 k T/J B FIG.3. Theferromagnetic(ω− )andantiferromagnetic(ω+ )excitationgapsasfunctionsoftemperatureforthespin-(S,s) k=0 k=0 ferrimagnetic Heisenberg chains calculated by the Schwinger bosons (SB) and the perturbationally interacting modified spin waves (PIMSW) based on theTakahashi scheme. IntheSchwingerrepresentationandthemodifiedspin- function of temperature. wave treatment based on the Takahashi scheme, the en- Figure 4 shows the temperature dependence of the ergy spectrum depends on temperature. Since the low- magnetic susceptibility-temperature product, which elu- energy band structure is well reflected in the thermal cidates ferromagnetic and antiferromagnetic features co- behavior and can directly be observedthrough inelastic- existinginferrimagnets[25]. χT divergesatlowtempera- neutron-scatteringmeasurements,we investigatethe fer- turesinaferromagneticfashionbutapproachesthehigh- romagnetic (ω− ) and antiferromagnetic (ω+ ) excita- k=0 k=0 temperature paramagnetic behavior showing an antifer- tion gaps as functions of temperature in Fig. 3. The romagneticincrease. Themodifiedspinwavesmuchbet- Schwinger-boson mean-field theory claims that the an- ter describe the magnetic behavior than the Schwinger tiferromagnetic gap should first decrease and then in- bosons. The spin waves modified along with the Taka- crease with increasing temperature, while the modified hashi scheme better work at high temperatures, while spin-wave theory predicts that the excitation energies those along with the new scheme precisely reproduce of both modes should be monotonically increasing func- the low-temperature behavior. Both calculations con- tionsoftemperature. We findasimilarcontrastbetween verge into the paramagnetic behavior χk T/N(gµ )2 = the twolanguagesappliedto ladderferrimagnets[35,58]. B B [S(S+1)+s(s+1)]/3athightemperatures,whereasthe In the case of Haldane-gap antiferromagnets, both the Schwinger-bosonmean-fieldtheoryagainbreaksdownat Schwinger-boson and modified-spin-wave [41] findings, the particular temperature (3.1). Considering that nu- togetherwiththenonlinear-σ-modelcalculations[59,60], merical tools less work at low temperatures, we realize commonlysuggestthatthe Haldanegapisasimplyacti- the superiority of the new-scheme-based modified spin- vated function of temperature. Extensive measurements wave theory all the more. onspin-1antiferromagneticHeisenbergchaincompounds [61–63] also report that the Haldane massive mode is Finally we calculate another type of ferrimagnet in shifted upward with increasing temperature. We en- order to demonstrate the constant applicability of the courageneutron-scatteringexperimentsonferrimagnetic present new scheme. Figure 5 shows the thermody- chain compounds to solve the present disagreement be- namic properties of the ferromagnetic-ferromagnetic- tween the Schwinger-boson and modified-spin-wave cal- antiferromagnetic-antiferromagnetic bond-tetrameric culations of the antiferromagnetic excitation gap as a spin-1 Heisenberg chain, 2 6 2.0 6.0 3.0 QMC SB LMSW(Takahashi) 2)B 4.0 PLIMMSSWW((YTaamkaahmaoshtoi)) 2.0 mg PIMSW(Yamamoto) N(1.0 T/ kB 2.0 1.0 c (a) (1,1/2) (b) (3/2,1/2) (c) (3/2,1) 0.0 0.0 0.0 0.0 1.0 2.0 3.0 0.0 1.0 2.0 3.0 0.0 1.0 2.0 3.0 k T/J B FIG. 4. The Schwinger-boson (SB), linear-modified-spin-wave (LMSW), perturbational interacting-modified-spin-wave (PIMSW), and quantum Monte Carlo (QMC) calculations of the susceptibility-temperature product χT as a function of temperatureforthespin-(S,s)ferrimagneticHeisenbergchains. Themodifiedspinwavesareconstructedintwodifferentways, theTakahashi scheme (Takahashi) and the new scheme(Yamamoto). 1.2 2.0 1.0 0.4 (a) QMC (b) (a) QMC (b) 0/CNkB .9 m2T/N(g)BB LPIMMSSWW((YYaammaammoototo)) 0/LkCB .8 m2J/L(g)B SLPFBIDMMISMSWWS(W(TTa(aTkkaaahkhaaashshahis)ih)i) 0.3 k c c 0.2 0.0 0.0 0.0 0.0 0.0 1.0 kBT / J 2.0 3.0 0.0 1.0 kBT / J 2.0 3.0 0.0 1.0 kBT / J 2.0 3.0 0.0 1.0 kBT / J 2.0 3.0 FIG. 5. The linear-modified-spin-wave (LMSW), pertur- FIG. 6. The bationalinteracting-modified-spin-wave(PIMSW),andquan- linear-modified-spin-wave (LMSW), perturbational interact- tum Monte Carlo (QMC) calculations of the specific heat C ing-modified-spin-wave (PIMSW), full-diagonalization inter- and the susceptibility-temperature product χT as functions acting-modified-spin-wave (FDIMSW), and quantum Monte of temperature for the spin-1 bond-tetrameric ferrimagnetic Carlo (QMC) calculations of thespecificheat C andthesus- 2 Heisenberg chain of JF = JAF. The modified spin waves are ceptibility χ as functions of temperature for the spin-1 anti- constructed on thenew scheme. ferromagneticHeisenbergchain. Themodifiedspinwavesare constructed on theTakahashi scheme. N = J (S S +S S ) AF 4n−3 4n−2 4n−2 4n−1 H · · n=1 X(cid:2) IV. SUMMARY AND DISCUSSION J (S S +S S ) , (3.2) F 4n−1 4n 4n 4n+1 − · · where we have set all the g-factors equal fo(cid:3)r simplicity. We have demonstrated the Schwinger-boson mean- The new modified spin-wave scheme again successfully field representation and the modified spin-wave treat- reproduces the Schottky peak of the Specific heat. The ment of one-dimensional Heisenberg ferrimagnets. The interactingmodifiedspinwavesfurtherinterpretthelow- Schwingerbosonsformanexcellentlanguageatlowtem- temperature Shoulder-like structure. The characteris- peratures but rapidly lose their validity with increasing ticminimumofthesusceptibility-temperatureproductis temperature. The modified spin-wave theory is more re- unfortunately less reproduced but the calculation again liable in totality provided the number of bosons is con- correctly gives the paramagnetic susceptibility at suffi- trolled without modifying the native energy structure. ciently high temperatures. A recent experiment [64] on On the other hand, the Schwinger-boson representation asingle-crystalsampleofCu(C H NCl) (N ) [7],which can be extended to anisotropic systems [67] more rea- 5 4 2 3 2 maybedescribedbytheHamiltonian(3.2),hasreported sonably because it is rotationally invariant in contrast that the specific heat exhibits a double-peakedstructure to the modified spin-wave theory. While the tempera- as a function of temperature. There is indeed a possi- turedependence ofthe anitiferromagneticexcitationgap bility of an additional peak appearing at low tempera- ω+ islefttosolveexperimentally,wearenowconvinced k=0 tures as the ratio J /J moves away from unity [65]. that the bosonic languages remain effective in low di- F AF However, no parameter assignment has yet succeeded in mensions and may be applied to extensive ferrimagnets interpreting all the observations consistently. There are [68]. Besidesground-statepropertiesandthermodynam- further chemical attempts to synthesize novel ferrimag- ics, quantum spin dynamics [69,70] can be investigated nets. Organic ferrimagnets [6,66] are free from magnetic through the modified spin-wave scheme. anisotropyandthussuitableforanalyzingintermsofthe We further mention our findings in the antiferromag- modified spin waves. netic limit with the view of realizing the close relation 7 between the two bosonic languages. We equalize s with fermioniclanguage[74,75],whichisinprinciplecompact, S and set 2N, the number of spins, equal to L for the being superior to any bosonic representation. Hamiltonian (2.1). At S = s, the ground-state sublat- In the case of ferromagnets, the Holstein-Primakoff tice magnetization(2.38)diverges and therefore the new bosons are already diagonal in the momentum space modified spin-wave scheme is no more applicable. We [37,56],suggestingnoquantumfluctuationintheground havetosettlethequantum,aswellasthermal,divergence state, and therefore the present new scheme turns out inevitably employing the Takahashi scheme. Besides the equivalent to the Takahashi scheme. The new-scheme- perturbationaltreatmentof ,wemayconsiderthefull based modified spin-wave theory is the very method for 0 H diagonalization of + , where the ground-state en- low-dimensional ferrimagnets and is ready for extensive 1 0 H H ergy and the dispersion relations are still given by Eqs. explorations. (2.32) and (2.33), respectively, but with θ satisfying k γ (k)+γ (k)=0. (4.1) 1 0 ACKNOWLEDGMENTS Such an idea applied to ferrimagnets ends in gapped e ferromagnetic excitations and misreads the low-energy The authors are gratefulto M. Takahashiand H. Hori physics. The perturbational series-expansion approach for useful discussions. This work was supported by the is highly successful in the case of ferrimagnets [32,33]. Ministry of Education, Culture, Sports, Science, and Focussing ourinterestonHaldane-gapantiferromagnets, Technology of Japan, and the Nissan Science Founda- welistthebosoniccalculationsoftheground-stateprop- tion. erties in Table II. The bosonic languages interpret the ground-state correlation very well but underesti- mate the Haldane gap considerably. Indeed they can- not detect the topological terms responsible for vanish- ing gap [72], but they are still qualitatively consistent with the nonlinear-σ-model quantum field theory, yield- ing the low-temperature limiting behavior ω+ ∆ [1] A. Gleizes and M. Verdaguer, J. Am. Chem. Soc. 103, k=0− 0 ∝ 7373 (1981); 106, 3727 (1984). e−∆0/T [41,60] and the large-spin asymptotic behavior [2] O.Kahn,Y.Pei, and Y.Journaux, in Inorganic Materi- ∆ e−πS [44,46,72]. The Schwinger-boson mean- 0 als, edited by D. W. Bruce and D. O’Hare (Wiley, New ∝ fieldtheoryandthefull-diagonalizationinteractingmod- York, 1995), p. 95. ified spin-wave treatment give the same estimate of the [3] O. Kahn, Y. Pei, M. Verdaguer, J.-P. Renard, and J. Haldane gap. The Schwinger-boson dispersion relation Sletten, J. Am. Chem. Soc. 110, 782 (1988). (2.11) indeed coincides analytically with that of the full- [4] A.Caneschi,D.Gatteschi,P.Rey,andR.Sessoli, Inorg. diagonalization interacting modified spin waves at zero Chem. 27, 1756 (1988); A.Caneschi, D.Gatteschi, J.-P. temperature. This is interesting but not so surpris- Renard, P. Rey,and R.Sessoli, ibid.28, 1976 (1989). ing, because the Holstein-Primakoff bosons (2.19) are [5] K.Inoue,F.Iwahori,A.S.Markosyan,andH.Iwamura, obtained by replacing both a (b ) and a† (b† ) Coord. Chem. Rev.198, 219 (2000). n↑ n↓ n↑ n↓ [6] Y. Hosokoshi, K. Katoh, Y. Nakazawa, H. Nakano, and by 2S−a†n↓an↓ ( 2s−b†n↑bn↑) in the transformation K. Inoue, J. Am. Chem. Soc. 123, 7921 (2001). (2.2q). q [7] A. Escuer, R. Vicente, M. S. El Fallah, M. A. S. Goher, Figure6showsthethermodynamiccalculationsforthe and F. A.Mautner, Inorg. Chem. 37, 4466 (1998). spin-1antiferromagneticHeisenbergchain. Welearnthat [8] M.Drillon,M.Belaiche,P.Legoll,J.Aride,A.Boukhari, the Schwinger-bosonmean-field theory does not work at and A.Moqine, J. Magn. Magn. Mater. 128, 83 (1993). [9] T. Lis, Acta Crystallogr. Sect. B 36, 2042 (1980). all for spin-gapped antiferromagnets at finite tempera- [10] K.Wieghardt,K.Phol,I.Jibril,andG.Huttner,Angew. tures, which is in contrast with its fairly good represen- Chem., Int.Ed. Engl. 23, 77 (1984). tation of the low-temperature thermodynamics for fer- [11] J. R.Friedman,M. P.Sarachik,J. Tejada, and R.Ziolo, rimagnetic chains. On the other hand, the modified Phys. Rev.Lett. 76, 3830 (1996). spin-wave treatment maintains its validity to a certain [12] L.Thomas,F.Lionti,R.Ballou,D.Gatteschi,R.Sessoli, extent. Indeed the Takahashi scheme still fails to re- and B. Barbara, Nature (London) 383, 145 (1996). producetheantiferromagneticSchottky-typepeakofthe [13] C. Sangregorio, T. Ohm, C. Paulsen, R. Sessoli, and D. specific heat, but it describes the susceptibility very well Gatteschi, Phys. Rev.Lett. 78, 4645 (1997). except for the low-temperature findings attributable to [14] W.WernsdorferandR.Sessoli, Science284,133(1999). the underestimate of the Haldane gap. We may expect [15] M.Drillon,J.C.Gianduzzo,andR.Georges,Phys.Lett. themodifiedspinwavestoefficientlydepictthedynamic, 96A, 413 (1983); M. Drillon, E. Coronado, R. Georges, as well as static, susceptibility for extensive spin-gapped J. C. Gianduzzo, and J. Curely, Phys. Rev.B 40, 10992 antiferromagnets including spin ladders [73]. As for the (1989). thermal properties of one-dimensional antiferromagnets, [16] S. Yamamoto and T. Sakai, J. Phys. Soc. Jpn. 67, 3711 whether spin gapped or not, there is a possibility of a (1998). 8 [17] M. Abolfath and A. Langari, Phys. Rev. B 63, 144414 (1988); A. Auerbach and D. P. Arovas, Phys. Rev. Lett. (2001). 61, 617. [18] T.SakaiandS.Yamamoto,Phys.Rev.B60,4053(1999); [45] J. E. Hirschand S.Tang, Phys.Rev.B 39, 2850 (1989). T. Sakai and K. Okamoto, Phys. Rev. B 65, 214403 [46] S.Sarker,C.Jayaprakash,H.R.Krishnamurthy,andM. (2002). Ma, Phys. Rev.B 40, 5028 (1989). [19] S. Yamamoto and T. Sakai, J. Phys.: Condens. Matter [47] M.-R. Li, Y.-J. Wang, and C.-D. Gong, Z. Phys.B 102, 11, 5175 (1999). 129 (1997). [20] T. Kuramoto, J. Phys. Soc. Jpn. 67, 1762 (1998); 68, [48] P. W. Anderson: Phys. Rev. 86, 694 (1952). 1813 (1999). [49] R. Kubo: Phys.Rev. 87, 568 (1952). [21] S.YamamotoandT.Sakai,Phys.Rev.B62,3795(2000). [50] S. Yamamoto and T. Fukui, Phys. Rev. B 57, R14008 [22] S.K.Pati, S.Ramasesha, and D.Sen, Phys.Rev.B 55, (1998). 8894 (1997); J. Phys.: Condens. Matter 9, 8707 (1997). [51] A.K.Kolezhuk,H.-J.Mikeska,S.Yamamoto,Phys.Rev. [23] K.Maisinger, U.Schollw¨ock,S.Brehmer,H.-J.Mikeska, B 55, R3336 (1997). and S. Yamamoto, Phys. Rev. B 58, R5908 (1998); S. [52] J.E.HirschandS.Tang,Phys.Rev.B39,R2887(1989). Yamamoto, T. Fukui, K. Maisinger, and U. Schollw¨ock, [53] J.H.XuandC.S.Ting,Phys.Rev.B42,R6861(1990). J. Phys.: Condens. Matter 10, 11033 (1998); [54] H. A. Ceccatto, C. J. Gazza, and A. E. Trumper, Phys. [24] S.Yamamoto,S.Brehmer,andH.-J.Mikeska,Phys.Rev. Rev. B 45, 7832 (1992). B 57, 13610 (1998); S. Yamamoto, T. Fukui, and T. [55] A. V. Dotsenko and O. P. Sushkov, Phys. Rev. B 50, Sakai, Eur. Phys.J. B 15, 211 (2000). 13821 (1994). [25] S.Yamamoto, Phys.Rev. B59, 1024 (1999). [56] M.Takahashi,Prog.Theor.Phys.Suppl.87,233(1986). [26] T. Fukui and N. Kawakami, Phys. Rev. B 55, 14709 [57] M. Takahashi and M. Yamada, J. Phys. Soc. Jpn. 54, (1997); 56, 8799 (1997). 2808 (1985); M. Yamada and M. Takahashi, ibid. 55, [27] S.Maslov andA.Zheludev,Phys.Rev.B 57,68 (1998). 2024 (1986). [28] Y. Takushima, A. Koga, and N. Kawakami, Phys. Rev. [58] T. Nakanishi, S. Yamamoto, and T. Sakai, J. Phys.Soc. B 61, 15189 (2000). Jpn. 70, 1380 (2001). [29] A.Zheludev,J. M. Tranquada, T. Vogt, and J. Buttrey, [59] I. Affleck,Phys. Rev.Lett. 62, 474 (1989). Europhys. Lett 35, 385 (1996); Phys. Rev. B 54, 6437 [60] Th. Jolicœur and O. Golinelli, Phys. Rev. B 50, 9265 (1996). (1994). [30] A.Zheludev,J. M. Tranquada, T. Vogt, and J. Buttrey, [61] J. P. Renard, M. Verdaguer, L. P. Regnault, W. A. C. Phys.Rev.B 54, 7210 (1996). Erkelens,J.Rossat-Mignod,J.Ribas,W.G.Stirling,and [31] S. Brehmer, H.-J. Mikeska, and S. Yamamoto, J. Phys.: C. Vettier,J. Appl.Phys.63, 3538 (1988). Condens. Matter 9, 3921 (1997). [62] R.M. Morra, W.J. L.Buyers,R.L. Armstrong,and K. [32] N.B.Ivanov,Phys.Rev.B 57,R14024 (1998); 62, 3271 Hirakawa, Phys. Rev.B 38, 543 (1988). (2000). [63] T.Sakaguchi,K.Kakurai,T.Yokoo,andJ.Akimitsu,J. [33] N. B. Ivanov and J. Richter, Phys. Rev. B 63, 144429 Phys. Soc. Jpn. 65, 3025 (1996). (2001). [64] M.Hagiwara,K.Minami,andH.A.Katori,Prog.Theor. [34] A. S. Ovchinnikov, I. G. Bostrem, V. E. Sinitsyn, A. S. Phys. Suppl.145, 150 (2002). Boyarchenkov, N. V. Baranov, and K. Inoue, J. Phys.: [65] T.NakanishiandS.Yamamoto,Phys.Rev.B65,214418 Condens. Matter 14, 8067 (2002). (2002). [35] X. Y. Chen, Q. Jiang, and W. Z. Shen, J. Phys.: Con- [66] A.Izuoka,M.Fukada,R.Kumai,M.Itakura,S.Hikami, dens.Matter 15, 915 (2003). and T. Sugawara, J. Am.Chem. Soc. 116, 2609 (1994). [36] C.Wu,B.Chen,X.Dai,Y.Yu,andZ.-B.Su,Phys.Rev. [67] M.-R. Li, Y.-J. Wang, C.-D. Gong, Z. Phys. B 102, 129 B 60, 1057 (1999). (1997). [37] M. Takahashi, Phys. Rev.Lett. 58, 168 (1987). [68] M. Drillon, E. Coronado, M. Belaiche, and R. L. Carlin, [38] M. Takahashi, Phys. Rev.B 40, 2494 (1989). J. Appl.Phys. 63, 3551 (1988). [39] J.E. Hirsch andS.Tang, Phys.Rev.B 40, 4769 (1989); [69] S. Yamamoto and T. Nakanishi, Phys. Rev. Lett. 89, S.Tang,M.E.Lazzouni,andJ.E.Hirsch,Phys.Rev.B 157603 (2002). 40, 5000 (1989). [70] H. Hori and S. Yamamoto, Phys. Rev. B 68, 054409 [40] S.M. Rezende,Phys. Rev.B 42, 2589 (1990). (2003). [41] S. Yamamoto and H. Hori, J. Phys. Soc. Jpn. 72, 769 [71] S.TodoandK.Kato,Phys.Rev.Lett.87,047203(2001). (2003). [72] F. D.M. Haldane, Phys.Lett. 93A, 464 (1983). [42] V.Y.Irkhin,A.A.Katanin,andM.I.Katsnelson,Phys. [73] H. Hori and S. Yamamoto, to be published in J. Magn. Rev.60, 1082 (1999). Magn. Mater. [43] M. Kollar, I. Spremo, and P. Kopietz, Phys. Rev. B 67, [74] X. Dai and Z. Su,Phys. Rev.B 57, 964 (1998). 104427 (2003). [75] H. Hori and S. Yamamoto, J. Phys. Soc. Jpn. 71, 1607 [44] D. P. Arovas and A. Auerbach, Phys. Rev. B 38, 316 (2002). 9 TABLE I. The Schwinger-boson (SB), linear-modified-spin-wave (LMSW), perturbational interacting-modified-spin-wave (PIMSW),andnumericaldiagonalization (Exact)calculations oftheground-stateenergyEg andthezero-temperatureantifer- romagnetic excitation gap ∆0 for thespin-(S,s) ferrimagnetic Heisenberg chains. (S,s)=(1,1) (S,s)=(3,1) (S,s)=(3,1) Approach 2 2 2 2 Eg/NJ ∆0/J Eg/NJ ∆0/J Eg/NJ ∆0/J SB −1.45525 1.77804 −1.96755 2.84973 −3.86270 1.62152 LMSW −1.43646 1 −1.95804 2 −3.82807 1 PIMSW −1.46084 1.67556 −1.96983 2.80253 −3.86758 1.52139 Exact −1.4541(1) 1.759(1) −1.9672(1) 2.842(1) −3.861(1) 1.615(5) TABLE II. The Schwinger-boson (SB), linear-modified-spin-wave (LMSW), perturbational interacting-modified-spin-wave (PIMSW),full-diagonalization interacting-modified-spin-wave(FDIMSW),andquantumMonteCarlo(QMC)[71]calculations of the ground-stateenergy Eg and thelowest excitation gap ∆0 for thespin-S antiferromagnetic Heisenberg chains. S =1 S =2 S =3 Approach Eg/LJ ∆0/J Eg/LJ ∆0/J Eg/LJ ∆0/J SB −1.396148 0.08507 −4.759769 0.00684 −10.1231 0.00295 LMSW −1.361879 0.07200 −4.726749 0.00626 −10.0901 0.00279 PIMSW −1.394853 0.07853 −4.759760 0.00655 −10.1231 0.00287 FDIMSW −1.394617 0.08507 −4.759759 0.00684 −10.1231 0.00295 QMC −1.401481(4) 0.41048(6) −4.761249(6) 0.08917(4) −10.1239(1) 0.01002(3) 10

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.