Spinons in Spatially Anisotropic Frustrated Antiferromagnets

Transcription

1 Spinons in Spatially Anisotropic Frustrated Antiferromagnets June 8th, 2007 Masanori Kohno Physics department, UCSB NIMS, Japan Collaborators: Leon Balents (UCSB) & Oleg Starykh (Univ. Utah)

2 Introduction Outline Elementary excitations in antiferromagnets : Magnon & Spinon Experimental indications of 2D spin liquids Model & Method Spatially anisotropic frustrated antiferromagnetic Heisenberg model Weak interchain-coupling approach Results Momentum-dependent spectral features Bound State of Spinons ~ Triplon ~ Comparison with Experiments Dynamical Structure Factor observed in Cs2CuCl4

3 Elementary Excitations in Antiferromagnets Excitation Physical Magnon Spinon Spin 1 1/2 G.S. Long-range order Spin-singlet S(k,ω) S(k,ω) Systems ω Unfrustrated in D >1 1D ω

4 Elementary Excitations in Antiferromagnets Excitation Physical Magnon Spinon Spin 1 1/2 G.S. Long-range order Spin-singlet S(k,ω) S(k,ω) Systems ω Unfrustrated in D >1 1D ω Highly frustrated antiferromagnets in D>1?

5 Experimental Indications of 2D Spin liquids Ground state without long-range order: Triangular lattice: κ-(bett-ttf) 2 Cu 2 (CN) 3 Solid 3 He in 2D Kagome lattice : ZnCu3(OH)6Cl2 No signature of long-range order down to T<<J [ex. T~32 mk, J~250 K ] Y.Shimizu et al, PRL 91, (2003).

6 Experimental Indications of 2D Spin liquids Ground state without long-range order: Triangular lattice: κ-(bett-ttf) 2 Cu 2 (CN) 3 Solid 3 He in 2D Kagome lattice : ZnCu3(OH)6Cl2 No signature of long-range order down to T<<J [ex. T~32 mk, J~250 K ] Non-magnon excitation: Y.Shimizu et al, PRL 91, (2003). Triangular lattice with spatial anisotropy : Cs 2 CuCl 4

7 Experimental Indications of 2D Spin liquids Ground state without long-range order: Triangular lattice: κ-(bett-ttf) 2 Cu 2 (CN) 3 Solid 3 He in 2D Kagome lattice : ZnCu3(OH)6Cl2 No signature of long-range order down to T<<J [ex. T~32 mk, J~250 K ] Non-magnon excitation: Y.Shimizu et al, PRL 91, (2003). Triangular lattice with spatial anisotropy : Cs 2 CuCl 4

8 Large Tail of S(k,ω) in Cs2CuCl4 Typical Magnon ω Dynamical structure factor S(k, ω) at kx=π[1] Spinon : Power-law divergence with a large tail. Magnon : δ-functional (for comparison) Intensity Intensity Line shape in Cs2CuCl4 peak with negligible tail. [1] R.Coldea, et al., Phys. Rev. B 68, (2003). ω Line shape in Rb2MnF4 [ square lattice, S=5/2 at k=(π,0.3π) ] T.Huberman, et al., Phys. Rev. B (2005).

9 Large Tail of S(k,ω) in Cs2CuCl4 Line shape in Cs2CuCl4 Theoretical proposals for 2D spinons Intensity J.Alicea, O.I.Motrunich & M.P.Fisher: Phys. Rev. Lett. 95, (2005). S.V.Isakov, T.Senthil & Y.B.Kim: Phys. Rev. B 72, (2005). Y.Zhou & X.-G.Wen: cond-mat/ F.Wang & A.Vishwanath: Phys. Rev. B 74, (2006). ω Dynamical structure factor S(k, ω) at kx=π[1] Spinon : Power-law divergence with a large tail. Magnon : δ-functional C.-H.Chung, K.Voelker & Y. B. Kim: Phys. Rev. B 68, (2003). 2D spin liquid state gives S( k,ω ) peak with negligible tail. 1 [ω η=0.74±0.14 [1] R.Coldea, et al., Phys. Rev. B 68, (2003). 2 ω L (kx ) 2 1 η 2 ], (experimental fit) (c.f. Heisenberg chain : η=1)

10 2D Dispersion Relation in Cs2CuCl4 Dispersion relation in Cs2CuCl4 Spin-wave theory kx=π kx M.Y.Veillette, A.J.A.James & F.H.L.Essler: Phys. Rev. B 72, (2005). D.Dalidovich, R.Sknepnek, A.J.Berlinsky, J.Zhang & C.Kallin: Phys. Rev. B 73, (2006). R.Coldea, D.A.Tennant & Z.Tylczynski: Phys. Rev. B 68, (2003). Asymmetry with respect to kx=π is due to interchain couplings. [1] R.Coldea, et al., Phys. Rev. B 68, (2003). ω k = ( J k J Q )[( J k Q + J k +Q ) 2 J Q ], J k = J cos + 2J cos k x 2 cos k y 2, J = 0.61(1)meV, J = 0.107(10)meV. c.f. bare couplings J=0.374(5)meV, J =0.128(5)meV.

11 Different Types of Line Shapes in Cs2CuCl4 Line shapes in Cs2CuCl4 ky=4π Intensity ky=2π kx=-π/2 ky/(2π) Depending on the momentum, S(k,ω) can have a sharp peak or broad continuum. R.Coldea, et al., Phys. Rev. B 68, (2003).

12 Unusual Features in Cs2CuCl4 S(k,ω) k y=0 ω Large tail in S(k,ω) 2D Spinon? k y=2π k y=4π k x=π 2D dispersion relation Spin-wave theory? Momentum dependence of line shape??? R.Coldea, et al., Phys. Rev. B 68, (2003). Interpretation of experimental results is controversial. Consistent explanation by a controlled approximation basically without fitting parameter

13 Model Model Spatially anisotropic (J < J) spin-1/2 antiferromagnetic Heisenberg model (J, J > 0). ( ) S x,y H = JS x+1,y + J 1 S x,y +1 + J S + J 2 x+1,y +1 3 S x 1,y +1. x,y [ J : intra-chain coupling, J 1, J 2, J 3 : inter-chain coupling] Highly frustrated antiferromagets : J 1=J 2+J 3.

14 Model Model Spatially anisotropic (J < J) spin-1/2 antiferromagnetic Heisenberg model (J, J > 0). ( ) S x,y H = JS x+1,y + J 1 S x,y +1 + J S + J 2 x+1,y +1 3 S x 1,y +1. x,y [ J : intra-chain coupling, J 1, J 2, J 3 : inter-chain coupling] Highly frustrated antiferromagets : J 1=J 2+J 3. Spatially anisotropic triangular lattice: J 1=J 2, J 3=0 (x,y+1) J 1 J 2 (x,y)

15 Model Model Spatially anisotropic (J < J) spin-1/2 antiferromagnetic Heisenberg model (J, J > 0). H = ( JSx +1,y + J1 Sx,y +1 + J 2 Sx +1,y +1 + J 3 Sx 1,y +1 ) Sx,y. x,y [ J : intra-chain coupling, J 1, J 2, J 3 : inter-chain coupling] Highly frustrated antiferromagets : J 1=J 2+J 3. Spatially anisotropic triangular lattice: Spatially anisotropic J1-J2 model: J 1=J 2, J 3=0 J 2=J 3 =J 1/2 : critical (x,y+1) (x-1,y+1) (x,y+1) (x+1,y+1) J1 J2 (x,y) J1 J3 J2 (x,y)

16 Decoupled Chain Limit (J 0) At J =0, the system is decomposed into decoupled Heisenberg chains. The S=1/2 Heisenberg chain is exactly solved by the Bethe ansatz. Ground state : S=0, E0=J(-ln2+1/4)L (L:length). Triplet states : 2-, 4-,... spinon states

17 Decoupled Chain Limit (J 0) At J =0, the system is decomposed into decoupled Heisenberg chains. The S=1/2 Heisenberg chain is exactly solved by the Bethe ansatz. Ground state : S=0, E0=J(-ln2+1/4)L (L:length). Triplet states : 2-, 4-,... spinon states 2-spinon states can be specified by 2 quasi-momenta (kx1, kx2). π εkx / J π/2 Alternatively, we can specify them by total momentum kx and excitation energy εkx. 0 π/2 π kx = kx1 + kx2 εkx =πj sin[kx/2]cos[(kx1-kx2)/2] 2-spinon spectrum of Heisenberg chain kx

18 Basis At J =0, exact ground state and triplet states are given as follows: 0 y+2 G.S. 0 = y 0 y J,ε kx, y =,ε kx 0 y+2 y y' y 0 y' J 0 y+1 J 0 y +1 J 0 y J,ε kx y J 0 y 1 J 0 y 1 J 0 y : Ground state of y-th chain.,ε kx y : 2-spinon state of y-th chain with S=1, momentum kx and energy εkx+e0.

19 Basis At J =0, exact ground state and triplet states are given as follows: 0 y+2 G.S. 0 = y 0 y J,ε kx, y =,ε kx 0 y+2 y y' y 0 y' J 0 y+1 J 0 y +1 J 0 y J,ε kx y J 0 y 1 J 0 y 1 J 0 y : Ground state of y-th chain.,ε kx y : 2-spinon state of y-th chain with S=1, momentum kx and energy εkx+e0. In the small J regime, restricting the Hilbert space to that spanned by this basis, states with momentum kx and ky are expressed as ψ(,k ) y = dε kx D ( kx ε kx )c ( kx,k y ε ) kx,ε kx,k y,,ε kx,k y 1 L e ikyy,ε kx, y. y

20 Interchain Interaction Expressing spin operators in k-space, H1 term is rewritten as follows: H1 = J ( ) Skx,y S kx,y +1, J ( ) J1 + J 2 e ikx + J 3 e -ikx. kx,y The expectation value of the interaction H1 by ψ(kx,ky)> has terms like y+2: y+1: y +2 y +1,ε kx y : y 0 y-1 : y Skx,y +1 (,ky ) ψ (,k y ) H1 ψ = J ( kx,k y ) dεkx S kx,y 1 * = A k, ε S 0 x kx kx,y +1 (kx,ε kx ) y +1,ε x 0 y+2 y 1 y 0 S kx,y,ε kx 1 = A( kx,εkx ) α A k, ε 0 S ( x kx ) kx,y,ε kx y (α=x, y or z). * * d ε D ε D ε c ε c ε A kx kx ( kx ) kx ( kx ) kx,ky ( kx ) kx,ky ( kx ) (kx,εkx ) A(kx,ε kx ), where J ( kx,k y ) 2{J1 cos ky + J 2 cos( + ky ) + J 3 cos( k y )}.

21 Effective Hamiltonian for Small J ψ(,k ) y H ψ(,k ) * y = c kx,k y ( ε i ( kx )H k,k ) x y i, j c ( kx,k y ε j ) j kx, c kx,k y ε kx i, j ( ) 2M c,k y ( j ε kx ). where ( ) = i εkx k H x,k y i, j ( ) 2 { J 1 cosk y + J 2 cos( + k ) y + J 3 cos( k )} y J,k y ( ) α 0 S kx,y A,ε kx ( + E 0 L)δ i, j + 2M J (,k y )A * ( i,ε kx )A( j,ε kx ),,ε kx (α=x,y or z) y ( M is the number of points in the ω space.)

22 Effective Hamiltonian for Small J ψ(,k ) y H ψ(,k ) * y = c kx,k y ( ε i ( kx )H k,k ) x y i, j c ( kx,k y ε j ) j kx, c kx,k y ε kx i, j ( ) 2M c,k y ( j ε kx ). where ( ) = i εkx k H x,k y i, j ( ) 2 { J 1 cosk y + J 2 cos( + k ) y + J 3 cos( k )} y J,k y ( ) α 0 S kx,y A,ε kx ( + E 0 L)δ i, j + 2M J (,k y )A * ( i,ε kx )A( j,ε kx ),,ε kx (α=x,y or z) y ( M is the number of points in the ω space.) Dynamical structure factor S(k,ω) is obtained as where i,e kx,k y α,k y S kx,k y ( ) G.S. S α k S k,ω G.S. 2 i( k) 2 δ( ω e i ( k) ), j i i* c kx,k y j ( ε kx )A * j,ε kx ( ) 1 J (,k y ) l 2M l A(,ε kx ) 2 j ε kx l + ε kx 2.

23 Transition Rate in the Heisenberg Chain The transition rate M(k,ω) between the ground state and 2-spinon states is exactly obtained for the Heisenberg chain by an algebraic analysis based on the (infinite-dimensional symmetry) quantum group[2]: A(,ε kx ) = M(,ε kx ), α M(,ε kx ) 0 S kx 2 1,ε kx = 4π exp 0 cosh 2x dx ( )cos xt x sinh( 2x)cosh x ( ) 1 ( ) e x, where t = 4 π ln ω U k 2 ω x k 2 U ω x L = πj sin 2, ω L + ω U 2 ε kx 2 ω L 2 = πj 2 sin. 2 ε kx, [2] A.H. Bougourzi, M. Couture and M. Kacir, PRB 54, R12669 (1996).

24 Effective Schrödinger Equation for Small J Alternatively, ε kx ψ kx ( ε ) kx + ( ) d J k ε kx D( * ε kx )A ( kx ε kx )A ( kx ε kx )ψ ( kx ε ) kx = e k ψ ( kx ε kx ).

25 Effective Schrödinger Equation for Small J Alternatively, ε kx ψ kx ( ε ) kx + ( ) d J k ε kx D( * ε kx )A ( kx ε kx )A ( kx ε kx )ψ ( kx ε ) kx = e k ψ ( kx ε kx ). This equation can be solved analytically, and S(k,ω) is obtained as S( k,ω) = ( ) [ ( )], 2 S 1D,ω [ 1+ J ( k) χ 1D (,ω)] 2 + π J ( k)s 1D,ω where ( ) = d χ 1D,ω ( ) ω S 1D, ω, S 0 ω 1D,ω ω This result is similar to RPA: Re χ 1D,ω ( ) = D kx ( ω) A kx ( ω) 2. ( ) = d ( ) ω S k, ω 1D x. ω ω

26 Features of S(k,ω) J (kx,ky) = 0 ky =π/2 Interchain interactions effectively vanish. 1D spinons persist: ( ) ln ω ω k L ( x) ( ) 2 S k,ω ( ) ω 2 ω L near ω ωl(kx). Spatially anisotropic J1-J2 model with J 2=J 3=J 1/2=0.24J

27 Features of S(k,ω) J (kx,ky) < 0 J (kx,ky) = 0 ky =π Bound state of spinons appear below continuum. It is similar to magnon, but does not require ordered ground states. It will be continuously connected to magnon. ky =π/2 Interchain interactions effectively vanish. 1D spinons persist: S( k,ω ) ln(ω ω L ( )) ω 2 ω L (kx ) 2 near ω ωl(kx). Spatially anisotropic J1-J2 model with J 2=J 3=J 1/2=0.24J

28 Features of S(k,ω) J (kx,ky) < 0 J (kx,ky) = 0 ky =π Bound state of spinons appear below continuum. It is similar to magnon, but does not require ordered ground states. It will be continuously connected to magnon. J (kx,ky) > 0 ky =π/2 ky =0 Interchain interactions effectively vanish. 1D spinons persist: S( k,ω ) ln(ω ω L ( )) ω 2 ω L (kx ) 2 near ω ωl(kx). Spectral weight shifts upwards. The peak is broadened in continuum. Anti-bound state appears above the 2-spinon continuum for small kx. Spatially anisotropic J1-J2 model with J 2=J 3=J 1/2=0.24J

29 Bound State in RPA The bound state is also obtained by RPA as a pole of dynamical susceptibility: Re χ 1 2d ( k,ω) ( ) [ ] = Re[ χ 1 1d k,ω ] + ( ) = 0. J k From this relation, the energy gap opens as ΔE J 2 with logarithmic Energy gap ΔE between bound state and lower edge of continuum on a spatially anisotropic triangular lattice at k=(π/4,π). : present approach : RPA

30 Bound State in the Ising Limit In the continuum, spinons are dispersed: L L L. (a) Density-density correlation function n(δx) in the lowest-energy state of the XXZ model in the Ising limit on a spatially anisotropic triangular lattice. (a) Incoherent state at kx=0.708π and (b) bound state at kx=0.667π.

31 Bound State in the Ising Limit In the continuum, spinons are dispersed: L L L. In the bound state, spinons are close to each other (Δx~0): L L. (a) (b) Density-density correlation function n(δx) in the lowest-energy state of the XXZ model in the Ising limit on a spatially anisotropic triangular lattice. (a) Incoherent state at kx=0.708π and (b) bound state at kx=0.667π.

32 Cs2CuCl4 Spatially anisotropic triangular lattice J=0.374(5)meV, J =0.128(5)meV J /J=0.34(3) (Interlayer J and DM ~ 0.05J) R.Coldea, et al., PRL (2001); PRB 68, (2003).

33 Cs2CuCl4 Spatially anisotropic triangular lattice J=0.374(5)meV, J =0.128(5)meV J /J=0.34(3) (Interlayer J and DM ~ 0.05J) k x=kx k y=kx+2ky k y J ( kx,k y ) = 4 J cos cos. 2 2 k y Sign of J (k) 2π π 0 π k x 2π R.Coldea, et al., PRL (2001); PRB 68, (2003).

34 Unusual Features Observed in Cs2CuCl4 by Coldea, et al. Intensity ω Large tail in S(k,ω) at k x=π Similar to 1D spinons R.Coldea, et al., PRL (2001); PRB 68, (2003).

35 Unusual Features Observed in Cs2CuCl4 by Coldea, et al. Intensity k y=0 ω Large tail in S(k,ω) at k x=π Similar to 1D spinons k x=π 2D dispersion relation Asymmetry with respect to k x=π R.Coldea, et al., PRL (2001); PRB 68, (2003).

36 Unusual Features Observed in Cs2CuCl4 by Coldea, et al. Intensity k y=0 k x=π ω Intensity Large tail in S(k,ω) at k x=π Similar to 1D spinons k y=2π 2D dispersion relation Asymmetry with respect to k x=π k y=4π Strong momentum dependence of line shape R.Coldea, et al., PRL (2001); PRB 68, (2003).

37 Large Tail at k x=π At k x=π, S(k,ω) reduces to that of the Heisenberg chain in the present approach, k y J ( kx,k y ) = 4 J cos cos. 2 2 because J (k)=0. ー : Present result ๐ : Experimental result of Cs2CuCl4 by Coldea, et al.[1] Inset: log-log plot. There is only one fitting parameter for the overall scale of intensity. [ This parameter is inevitable.] S( k,ω ) ln(ω ω L ( )) 2 ω ω L (kx ) 2 near ω ωl(kx). [1] R.Coldea, et al., PRB 68, (2003).

38 Comparison with Power-Law Fit Intensity The experimental data at kx=π have been fitted in a power-law form as S( k,ω ) where ωl(kx) Intensity ω[mev] 1 [ω 2 ω L (kx ) 2 1 η 2 ], : lower edge of continuum. η=0.74±0.14 (c.f. Heisenberg chain: η=1) : Heisenberg chain (2-spinon) : power-law fit with η=0.74 : experimental data[1] [1] R.Coldea, et al., PRB 68, (2003).

39 Dispersion Relation along k x At k y=0 and k y=2π, dispersion relation is asymmetric with respect to k x=π. At k y=3π, dispersion relation is symmetric, because J (k x,k y=3π)=0. No fitting parameter k y=0 J (k)>0 J (k)<0 k y=2π J (k)>0 J (k) <0 J (k) >0 k y=3π J (k)=0 k y Main peak( ), upper edge( ) and lower edge( ) J ( kx,k y ) = 4 J cos cos. 2 2 estimated experimentally by Coldea, et al.[1].

40 Dispersion Relation along k x At k y=0 and k y=2π, dispersion relation is asymmetric with respect to k x=π. At k y=3π, dispersion relation is symmetric, because J (k x,k y=3π)=0. No fitting parameter k y=0 J (k)>0 J (k)<0 k y=2π J (k)>0 J (k) <0 J (k) >0 k y=3π J (k)=0 k y Main peak( ), upper edge( ) and lower edge( ) J ( kx,k y ) = 4 J cos cos. 2 2 estimated experimentally by Coldea, et al.[1].

41 Dispersion Relation along k y At k x=-π/2, the sign of J (k) changes at k y=3π. No fitting parameter For k y<3π [ J (k)<0 ], a bound state is formed below the continuum. For k y>3π [ J (k)>0 ], the peak is broadened in the continuum. J (k)>0 J (k)<0 S(k,ω) at k x=-π/2 near the lower edge of the continuum by the present method. k y J ( kx,k y ) = 4 J cos cos. 2 2

42 Dispersion Relation along k y At k x=-π/2, the sign of J (k) changes at k y=3π. No fitting parameter For k y<3π [ J (k)<0 ], a bound state is formed below the continuum. For k y>3π [ J (k)>0 ], the peak is broadened in the continuum. k x=-π/2 J (k)>0 J (k)<0 J (k)>0 J (k)<0 S(k,ω) at k x=-π/2 near the lower edge of the continuum by the present method. k y J ( kx,k y ) = 4 J cos cos. 2 2 Main peak( ), upper edge( ) and lower edge( ) estimated experimentally by Coldea, et al.[1].

43 Dispersion Relation along k y At k x=-π/2, the sign of J (k) changes at k y=3π. No fitting parameter For k y<3π [ J (k)<0 ], a bound state is formed below the continuum. For k y>3π [ J (k)>0 ], the peak is broadened in the continuum. k x=-π/2 J (k)>0 J (k)<0 J (k)>0 J (k)<0 S(k,ω) at k x=-π/2 near the lower edge of the continuum by the present method. k y J ( kx,k y ) = 4 J cos cos. 2 2 Main peak( ), upper edge( ) and lower edge( ) estimated experimentally by Coldea, et al.[1].

44 Dispersion Relation along k y At k x=-π/2, the sign of J (k) changes at k y=3π. No fitting parameter For k y<3π [ J (k)<0 ], a bound state is formed below the continuum. For k y>3π [ J (k)>0 ], the peak is broadened in the continuum. k x=-π/2 J (k)>0 J (k)<0 J (k)>0 J (k)<0 S(k,ω) at k x=-π/2 near the lower edge of the continuum by the present method. k y J ( kx,k y ) = 4 J cos cos. 2 2 Main peak( ), upper edge( ) and lower edge( ) estimated experimentally by Coldea, et al.[1].

45 Bound State and Broad Peak For J (k)<0, S(k,ω) has a sharp peak of the bound state. No fitting parameter For J (k)>0, S(k,ω) has a broad peak in the continuum. k x=-π/2 k y=2π J (k)<0 ー k x=-π/2 k y=4π J (k)>0 : Present result (2-spinon) ๐ : Experimental result of Cs2CuCl4 by Coldea, et al.[1] [1] R.Coldea, et al., PRB 68, (2003).

46 Bound State and Broad Peak For J (k)<0, S(k,ω) has a sharp peak of the bound state. No fitting parameter For J (k)>0, S(k,ω) has a broad peak in the continuum. k x=-π/2 k y=2π J (k)<0 ー k x=-π/2 k y=4π J (k)>0 : Present result (2-spinon) broadened by experimental resolution ๐ : Experimental result of Cs2CuCl4 by Coldea, et al.[1] (ΔE=0.019 mev) [1] R.Coldea, et al., PRB 68, (2003).

47 Bound State and Broad Peak For J (k)<0, S(k,ω) has a sharp peak of the bound state. No fitting parameter For J (k)>0, S(k,ω) has a broad peak in the continuum. k x=-π/2 k y=2π J (k)<0 ー k x=-π/2 k y=4π J (k)>0 : Present result (2- & 4-spinon, Lx=288) broadened by experimental resolution ๐ : Experimental result of Cs2CuCl4 by Coldea, et al.[1] (ΔE=0.019 mev) [1] R.Coldea, et al., PRB 68, (2003).

48 Summary ~ Features in S(k,ω) ~ Three distinctive features of dynamical structure factor S(k,ω) are obtained depending on the momentum, which is classified by the sign of J (k). J (k) = 0 : J (k) < 0 : J (k) > 0 : The interchain interactions effectively vanish. Spinons of the Heisenberg chain persist at these momenta. Bound states of spinons appear below the continuum, where a spinon pair moves coherently like a magnon. This bound state does not require ordered ground states. Spectral weight shifts upwards. The peak is broadened in the continuum. Anti-bound state is observed above 2-spinon continuum. In the present approach, power-law behaviors appear only at J (kx,ky)=0. The exponent η=1 (1D Heisenberg chain).

49 Summary ~ Comparison with Experiments ~ Unusual experimental features in Cs2CuCl4 have been consistently explained by the present approach basically without fitting parameter. Asymmetry of the dispersion relation : Bound state below the continuum for J (k)<0. Broad peak in the continuum for J (k)>0. Momentum dependence of line shape : Bound states are actually observed experimentally for J (k)<0 as a sharp peak. J (k) <0 J (k)<0 J (k) >0 J (k)=0 kx=π Large tail at kx=π : present approximation S(k,ω) at J (k)=0 S1D(kx,ω) (1D Heisenberg chain). Interpretation : descendants of 1D spinons strongly persist at J (k)=0.