The authors have used ' Y NMR to measure the internal magnetic field of a sample of YBa:Cu30& z (T, =90 K) in the superconducting state to correct for the effects of the Meissner shielding currents. They used this information to separate the magnetic-shift tensor into orbital (chemical-shift) and spin (Knight-shift) contributions. They find that the temperature dependence of the chain Cu Knight shift fits the classic Yosida function of weak-coupling, orbital-s-state, spinsinglet BCS theory. For the planes, the Knight shift also requires a spin singlet, but with a strongcoupling Yosida function. The best fit is for an orbital s state, but an orbital d state is also possible.They find a zero-temperature gap b(0) of 1.76k& T, for the chains. For the planes, 6(0) can range from 1.9k8T, to 3.lk~T, .

The authors have measured the change in the anisotropy of the NMR spin-lattice relaxation rates for 63 Cu(2), 17 7. These results cannot be explained by a straightforward extension to T < T c of current theories, such as the theory of Millis, Monien, and Pines of normal-state NMR relaxation.PACS numbers: 74.70.Vy, 74.30.Gn, 76.60.Es, 76.60.Gv NMR measurements have shown that spin fluctuations in high-temperature superconductors have strong antiferromagnetic correlations with a characteristic correlation length £ of several lattice constants. £ varies with temperature in most of the normal state but has been assumed to be independent of temperature throughout the superconducting state. We have measured the anisotropy of the NMR spin-lattice relaxation rate W\ in the normal and superconducting states of YBa2Cu3C>7 (7V =93 K) for the 63 Cu and 17 0 nuclei in the planes [the Cu(2) and 0(2,3) sites] and for the 89 Y nuclei. The 63 Cu anisotropy decreases dramatically upon entering the superconducting state. When combined with our measurements of the 17 0 and 89 Y anisotropy at 100 and 77 K, we see that the change in the copper relaxationrate ratio appears to require a modification of theories such as that by Millis, Monien, and Pines 1 (MMP) in which the temperature dependence of the NMR ratios comes only from the temperature dependence of the (assumed isotropic) antiferromagnetic correlation length. We describe four possible modifications.The data presented below involve six samples. All samples but one exhibit 100% shielding fractions, and T c (H 0 =0) was 93 K as measured by the change in coil inductance of a nuclear quadrupole resonance (NQR) coil and SQUID measurements. 2 Sample 1 was composed of -50 single crystals whose c and alb axes were optically aligned. Samples 2 and 3 were unaligned powder samples. Samples 4, 5, and 6 were aligned powder samples sealed in Stycast 1266 epoxy. Sample 6 was isotopically enriched ( I7 0) and exhibited a smaller shielding fraction (80%); however, all static and dynamic NMR properties agreed with previous measurements. 3,4 Since 63 Cu and ,7 0 have spins of /=y and f, respectively, their spin-lattice relaxation curves are multiexponential. 5 We report the rate W\ a (a=x,y,z) utilized by MMP which is y of the single-exponential relaxation rate which would occur in a strong magnetic field in the absence of electric quadrupole splittings. It is the temperature-dependent anisotropy of the quantities W\ a that we report. W\ a and W\t for Cu(2) and Y are not experimentally distinguishable. However, three different oxygen resonances are distinguishable. Following Takigawa et al., 3 we define xl W\ a and xl W\b to be the rates when the field is in the CuC>2 planes parallel and perpendicular, respectively, to the Cu-O-Cu bond axis.The values of W\ a as measured by NMR for the Cu(2) site in single-crystal sample 1 and aligned-powder sample 4 are shown in Fig. 1, demonstrating that the measured W\ a \ are sam...