Lawrence Berkeley National Laboratory
Dynamical properties of the S=12 random Heisenberg chain
- Author(s): Shu, Yu-Rong
- Dupont, Maxime
- Yao, Dao-Xin
- Capponi, Sylvain
- Sandvik, Anders W
- et al.
Published Web Locationhttps://doi.org/10.1103/physrevb.97.104424
We use numerical techniques to study dynamical properties at finite temperature ($T$) of the Heisenberg spin chain with random exchange couplings, which realizes the random singlet (RS) fixed point in the low-energy limit. Specifically, we study the dynamic spin structure factor $S(q,\omega)$, which can be probed directly by inelastic neutron scattering experiments and, in the limit of small $\omega$, in nuclear magnetic resonance (NMR) experiments through the spin-lattice relaxation rate $1/T_1$. Our work combines three complementary methods: exact diagonalization, matrix-product-state algorithms, and stochastic analytic continuation of quantum Monte Carlo results in imaginary time. Unlike the uniform system, whose low-energy excitations at low $T$ are restricted to $q$ close to $0$ and $\pi$, our study reveals a continuous narrow band of low-energy excitations in $S(q,\omega)$, extending throughout the Brillouin zone. Close to $q=\pi$, the scaling properties of these excitations are well captured by the RS theory, but we also see disagreements with some aspects of the predicted $q$-dependence further away from $q=\pi$. Furthermore we find spin diffusion effects close to $q=0$ that are not contained within the RS theory but give non-negligible contributions to the mean $1/T_1$. To compare with NMR experiments, we consider the distribution of the local $1/T_1$ values, which is broad, approximately described by a stretched exponential. The mean value first decreases with $T$, but starts to increase and diverge below a crossover temperature. Although a similar divergent behavior has been found for the static uniform susceptibility, this divergent behavior of $1/T_1$ has never been seen in experiments. Our results show that the divergence of the mean $1/T_1$ is due to rare events in the disordered chains and is concealed in experiments, where the typical $1/T_1$ value is accessed.