Recently there has been extensive theoretical interest in the remarkable Haldane prediction that spin-1 antiferromagnetic Heisenberg-like chains should display an energy gap between singlet ground state and excited continua. This behavior has been confirmed numerically, and it has been shown that the energy gap is of a magnitude to be observable experimentally. The phenomenon has captured the interest of experimental groups in neutron scattering and magnetism. A recent experimental study obtains very good experimental agreement with the theoretical T=0 magnetization curve obtained numerically by Parkinson and Bonner.1,2 An explanation will be provided for the unusual phase diagram and other anomalous features observed by the authors of Ref. 2. Numerical susceptibility studies of Heisenberg antiferromagnetic and general XXZ spin-1 chains will be presented, and discussed in terms of a novel quantum-classical cross-over phenomenon.2 Finally, a comparison will be made of T=0 magnetization isotherms for several spin-1 chains, including the Takhtajan–Babujian chain and the chain with pure biquadratic interactions.

The spectral properties, and associated critical singularities, of spin-1 Heisenberg antiferromagnetic chains are of high current interest, both theoretical and experimental. Theoretical studies using the single mode approximation and also a Monte Carlo calculation indicate that the spectrum has a gap at k=π, as predicted by Haldane. They also predict the unusual feature that the excited state continuum is asymmetric about the mid-point of the Brillouin zone, k=π/2. However, neither approach is sufficiently accurate to predict reliably the magnitude of the gap at k=0 which was not discussed by Haldane. Our approach is to use static (spectral) calculations of excited states of chains of N=4 through 14 spins for all values of k, a novel feature not associated with previous finite-size calculations, to discuss the dispersion spectra of the S=1 Heisenberg chain. In addition, we have calculated dynamical properties, in particular S(q,ω), exactly for finite chains up to N=14 spins. This allows us to estimate the gap at k=0 to good accuracy, and we find that to within this accuracy the gap at k=0 is twice the Haldane gap at k=π. This result is important in connection with very recent neutron scattering and magnetization measurements. In addition, we have calculated the correlation length and critical singularities associated with dynamical quantities.

The antiferromagnetic (AFM) s=1 chain with biquadratic exchange coupling has recently been the subject of considerable attention, and some controversy. Theoretical arguments by Affleck that the model has dimerized character associated with a small spectral excitation gap have not received unanimous support from numerical calculations. For finite systems the field-dependent dispersion spectra are interesting and informative concerning the nature of the model. The excitations divide into two classes having very different character, corresponding to SzT even and SzT odd. Reflection symmetry about k=π/2 is apparent if states with a given total spin ST, rather than SzT, are plotted as a function of wavevector k, consistent with the Affleck prediction of dimerized character. The high degeneracy of the eigenstates reflects the existence of new symmetry operators which (a) shift the SzT value, (b) change wave vector k to π/2−k, and (c) shift the total spin quantum ST, of the energy eigenstates. Parkinson has recently observed that finite N states with SzT even and SzT≥2 map into states of the finite N s= 1/2 XXZ model at a special value of uniaxial anisotropy. Barber and Batchelor (unpublished) have shown the existence of an analytic mapping between states of the s=1 biquadratic chain and the 2D 9-state Potts model. They further show that all states of the Parkinson s= 1/2 XXZ model map into states of the s=1 biquadratic model. Hence, they find the ground state and first excited state energies exactly, and verify the Affleck picture. Their approach falls short of establishing complete integrability, however. Numerical studies on finite chains indicate that the T=0 magnetization curve is determined by the lowest-energy states for given SzT even, which are states of the s= 1/2 XXZ chain. Hence the magnetization curve of the s=1 AFM biquadratic chain is the same as the analytically known magnetization curve of the s= 1/2 XXZ chain!

Recently an elegant and quite powerful finite-system approach to determine the exponents ηx and ηz from simple spectral properties has been proposed. For critical systems, the two exponents can be expressed in terms of finite-size spectral gaps as follows: η(N)x=2ΔE01(N)/ΔE(N), η(N)z=2ΔE00(N)/ΔE(N). Here ΔE(N) is the finite-size gap between the ground state (SzT=0,k=0) and the lowest excitation at k=2π/N; ΔE01(N) is the gap to the lowest ‖SzT‖=1 excitations (at k=π), and ΔE00(N) is the gap to the next lowest SzT=0 excited state. The η(N) sequence is then extrapolated to N→∞. For XY models, differences between s=1/2 and s≥1 appear. For s=1/2, the excitations which determine ΔE00(N) and ΔE(N) are degenerate, which implies that ηz=1/2, in agreement with the exact analytic result. For spin-1, however, the next lowest SzT=0 state is located at k=2π/N instead of k=π, and is therefore identical to the state which determines the gap ΔE. The resulting equality ΔE=ΔE00 implies ηz=2, as in the spin-1/2 case. In fact, our result corresponds to power-law decay for all s, and hence we differ from Schulz and Ziman, who claim the out-of-plane correlation function decays exponentially for s>1/2. For the in-plane correlation function, the spectral gap method again agrees with the exact result ηx=0.5 for s=1/2. The consensus of this and other numerical methods for s=1 gives a value ηx≂0.20, considerably different from the case of s=1/2. Hence it is tempting to conjecture that ηx is s dependent, implying that XY models belong to different universality classes for different s. However, a finite-size study of the conformal anomaly produces the result that c=1, independent of s. This situation is further discussed.

Quantum spin chains of higher spin are interesting in the case of both ferromagnets and antiferromagnets. They comprise a wide range of model types—Bethe–Ansatz integrable, nonintegrable, and with limited integrability. It is not well appreciated that the two-magnon (two-body) problems is always solvable, by Bethe–Ansatz and Green’s-function techniques, regardless of the integrability character of the entire model, with interesting implications in the context of solitons and bound solitons. Chubukov and Kveschenko have suggested using the two-magnon problem in a test for integrability and claim to find additional classes of integrable spin chains for s=1 and s= (3)/(2) . Here we investigate this situation.

We compare the excitation spectra in the presence of a magnetic field of a number of integrable (exactly solvable) and nonintegrable quantum spin chains of various spin value s. The archetypal Bethe-ansatz integrable model is the s= 1/2 Heisenberg antiferromagnet (HB AFM). The excitation spectra are characterized by a soft mode which tracks across the Brillouin zone as the field increases to its saturation value. A class of Bethe-ansatz integrable models with SU(2) symmetry and the general spin s display excitation spectra qualitatively similar to the spin- 1/2 model above, for all s. A second class of Bethe-ansatz integrable models has SU(n) symmetry, where n=2s+1. Like the SU(2) integrable chains, these models have gapless excitation spectra, but the basic Brillouin zone changes from k=±2π/(2s+1)a. Studies show that periodicity of the SU(3) member of the class changes (increases) as the field increases to saturation. For both classes of integrable models, there is a single type of excitation pattern which is generically similar for all s. In the case of the other models, on the other hand, numerical studies show that the excitations divide into at least two distinct classes. In the case of the s=1 HB AFM, at high fields (corresponding to SzT=N,N−1, . . .,N/2) the excitations map approximately onto the complete set of excitations for s= 1/2 , whereas at low fields (SzT=N/2,N/2−1,. . .,0) the excitations have notable classical character. In the case of the s=1 model with pure biquadratic exchange, one set of excitations, corresponding to SzT even (SzT=N,N−2,. . .,2,0), again shows an approximate mapping to the complete excitation set for s= 1/2 . The second class of excitations, corresponding to SzT odd, are very different. They are symmetric about k=±π/2a for all SzT, i.e., correspond to a basic Brillouin zone of ±π/2a.

The conformal anomaly c determines the universality class of a model system in statistical mechanics. The value of c characterizes both 2D classical models and their 1D quantum counterparts. The conformal anomaly may therefore be determined numerically for quantum spin chains using the relation: E0(N)≂E0(∞)−(NΔE/12)c(1/N2), where E0 (N) is the ground-state energy of an N-spin finite system, E0 (∞) is the ground-state energy in the thermodynamic limit, and ΔE is the energy gap between the ground state at k=0 and the first excited state of the dispersion curve at k=2π/N. The numerical approach is highly successful when tested on the integrable s= 1/2 Heisenberg antiferromagnetic XXZ chain and the integrable s=1 SU(2) model. The method gives c=1 to within 2% accuracy for the s=1 and (3)/(2) XY chains, placing them in the universality class of the 2D XY model. The result c=1 (2% accuracy) is obtained for the s= (3)/(2) Heisenberg antiferromagnetic chain, in agreement with the Haldane prediction. The s=1 pure antiferromagnetic biquadratic chain and the s=1 XXZ model with uniaxial anisotropy in the vicinity of the critical point Δ=Δ2 ∼1.15 −1.18 have also been studied.

Since the Heisenberg spin chain can be considered the simplest realistic model of magnetism, surprise and some degree of controversy have resulted from recent work of Haldane. The prediction is that quantum spin chains with half-integer spin should all display T=0 phase behavior equivalent to that of the Bethe Ansatz integrable (solvable) spin-1/2 quantum chain. More remarkably, the class of integer spin chains is predicted to show very different phase behavior. In particular, a gap should be present in the spectrum of a Heisenberg antiferromagnetic chain. This remarkable feature is counterintuitive in terms of accepted wisdom in magnetism (spin-wave theory, spin-Peierls theory) and critical phenomena. Consequently the vertification of the prediction is of great interest. A considerable amount of numerical work has been done, involving finite-chain, finite-size scaling, variational, Monte Carlo and other calculations, which will be reviewed here. The present consensus is that the weight of numerical evidence supports the prediction, although puzzling features still remain. Adding additional interactions to the basic Heisenberg Hamiltonian such as spin (XXZ) anisotropy, single-ion anisotropy, biquadratic exchange, and an applied magnetic field, generates a rich and complicated phase diagram for chains with spin >1/2, particularly for the case of integer spin. The s=1 phase diagram seems to display critical behavior of a type not previously encountered. A theoretical appraisal of the Haldane phenomenon will include a discussion of the possible role of nonintegrability. Mention will also be made of current progress in experimental investigation of the phenomenon, including problems that might be encountered. More recent work of Affleck has greatly generalized the field-theoretic mappings which underlay the original work of Haldane. A number of interesting problems have been mapped into quantum spin chains of various types, including field theoretic phenomena and the localization problem of the quantum Hall effect.

This study concerns the concept of nonintegrability in quantum many-body systems, which is related to the important and unresolved problem of quantum chaos. Our findings strongly indicate that nonintegrability affects the reliability of many approximation techniques which have proved to be successful in the study of integrable models. This report is based on finite-size studies of the low-lying spectral excitations of both integrable and nonintegrable 1D quantum spin models. In integrable cases, the characteristic excitation pattern of the infinite system is apparent even in relatively short chains. This is generally not the case in nonintegrable systems where we observe several classes of excitations with qualitatively different character. In some situations, the nature of the lowest-lying excitations actually changes with increasing system size, which makes finite-size studies very vulnerable to misleading conclusions if care is not taken.

Affleck has made predictions for the critical exponents of a class of integrable spin chains of general spin $s$. The expressions are $s$ dependent. Our interest here is a finite-size scaling analysis of spin-1 systems to determine the validity of the prediction for the alternation critical exponent. We discuss these results in the light of our numerical results for the conformal anomaly $c$, the exponent $\ensuremath{\eta}$, and the mass-gap exponent associated with biquadratic exchange. The numerical results are not generally in good agreement with theory, and hence the possible presence of logarithmic corrections is discussed.

The problem of the critical behavior of the spin-1/2 Heisenberg ferromagnetic chain has challenged many research workers since 1964. We discuss the critical exponents α and γ recently determined by Schlottmann and by Takahashi and Yamada in the context of previously known Tc=0 critical properties of the 1D s=1/2 XXZ model and the classical Heisenberg model, and we comment on the manifest breakdown of scaling in the s=1/2 Heisenberg ferromagnet.

Heisenberg spin chains which represent the simplest realistic models for magnetic insulators were thought to be well understood and generically similar for any spin-value s. This is expressed in the spin-wave approach to Heisenberg spin chains [1]. Consequently surprise and some degree of controversy resulted from recent work of Haldane [2,3], who proposed a dramatically different picture. Consider the spin-s XXZ Hamiltonian with anisotropy parameter △: $${\rm{H}} = {\rm{J}}\mathop \sum \limits_{\ell {\rm{ = 1}}}^{\rm{N}} \left\{ {{\rm{S}}_\ell ^{\rm{x}}{\rm{S}}_{\ell {\rm{ + 1}}}^{\rm{x}}{\rm{ + S}}_\ell ^{\rm{y}}{\rm{S}}_{\ell {\rm{ + 1}}}^{\rm{y}}{\rm{ + \Delta S}}_\ell ^{\rm{z}}{\rm{S}}_{\ell {\rm{ + 1}}}^{\rm{z}}} \right\}{\rm{.}}$$ (1) For half-integer s, the region O ≤ △ < 1 is a gapless phase with power-law decay of the two-spin correlation functions, terminating in an essential singularity at △ = 1. For △ > 1, the ground state consists of two degenerate singlet states associated with long-range order and a gap to an excitation continuum. For integer s, on the other hand, the gapless phase associated with planar anisotropy extends only over a range 0 ≤ △ ≤ △1 (△1 < 1) and the phase with gap and ordered ground state extends over △ ≥ △2 (△2 > 1). A new phase (called hereafter the Haldane phase) appears in the region △1 < △ < △2 encompassing the Heisenberg point at △ = 1. The Haldane phase ground state is a non-ordered singlet with exponentially decaying spin correlation functions, and there is a gap to an excitation continuum, which has its maximum value at △ = 1. The spin-dependent gap at △ = 1 is given by △E/J ∼ s2 exp(-πs).

We have performed a variety of numerical studies on the general bilinear-biquadratic spin-1 Hamiltonian H/J=∑Ni=1[Si⋅Si+1 −β(Si⋅Si+1)2], over the range 0≤β≤∞. The model is Bethe Ansatz integrable at the special point β=1, where the spectrum is gapless, but is otherwise believed to be nonintegrable. Affleck has predicted that an excitation gap opens up linearly in the vicinity of β=1. Our studies involving spectral excitations (dispersion spectra), scaled-gap, and finite-size scaling calculations are not consistent with the Affleck prediction. The situation appears complex, with novel crossover effects occurring in both regimes, β<1 and β>1, complicating the analysis.

The integrable S=1 isotropic quantum spin chain of N atoms, introduced by Takhtajan (1982) and Bubujian (1983), is studied as a function of magnetisation and hence of magnetic field. The methods used are (i) direct numerical diagonalisation for N

The authors have investigated the general 1D spin-1 bilinear-biquadratic exchange Hamiltonian by analytic and numerical techniques over the range 01.

An unusual crossover mechanism has been discovered by numerical investigation of the dispersion spectrum of Heisenberg antiferromagnetic chains with various spin values in a magnetic field. This result is reflected in novel behavior of static properties such as the integrated intensity. A study of various excitation gaps using finite chain calculations extended by quantum Monte Carlo studies indicates unusual behaviour in the T=0 magnetization isotherms.

A comprehensive investigation has been made of the spectral excitations and static properties of Heisenberg antiferromagnetic chains of spin 1/2, 1, 3/2, and 2, using Lanczös, Bethe Ansatz, and Monte Carlo techniques. An unusual and unanticipated crossover mechanism for spin chains with 1/2≤S≤∞ has been discovered. The validity of the Haldane conjecture concerning the presence of a spectral excitation gap for integer-spin chains has been investigated by exact finite chains calculations of (a) the primary singlet-triplet excitation gap, (b) higher excitation gaps, and (c) the Fourier transform of the ground state correlation functions. A new Monte Carlo method has extended the spin-1 gap calculations to N=32.