In a recent letter (PRL 79, 2117 (1997)), Bobroff et al. presented novel $^{ 17}$O NMR measurements for YBa_2(Cu_{ 1-x}Ni_x)_3O_{ 6+y} which offer a new probe of the momentum dependence of the static spin susceptibility $\chi'(q)$. They analyzed the $^{ 17}$O line width by assuming a Gaussian form for $\chi'$, and came to the conclusion that the correlation length $\xi$ is independent of temperature $T$. Furthermore, they claim that a Lorentzian form of $\chi'$ yields the same result. We will demonstrate that the Lorentzian and Gaussian form of $\chi'$ actually yield qualitatively different results for the $^{ 17}$O line width and its dependence on $\xi$. We show that their data when combined with $T_{ 2G}$ data from Takigawa require that $\xi$ be T-dependent. Furthermore, assuming a Gaussian form of $\chi'$ leads to the unphysical result that $\xi$ increases with $T$. In contrast, the Lorentzian form of $\chi'$ yields that $\xi$ decreases with increasing $T$ and is thus fully compatible with the experimental results.
Categories: cond-mat.str-el